Tunable RF Components Circuits: Jeffrey L. Hilbert
Tunable RF Components Circuits: Jeffrey L. Hilbert
Components
and Circuits
APPLICATIONS IN MOBILE HANDSETS
E D I T ED B Y
Jeffrey L. Hilbert
Tunable RF
Components
and Circuits
APPLICATIONS IN MOBILE HANDSETS
Devices, Circuits, and Systems
Series Editor
Krzysztof Iniewski
CMOS Emerging Technologies Inc., Vancouver, British Columbia, Canada
FORTHCOMING
Telecommunication Networks
Eugenio Iannone
Optical, Acoustic, Magnetic, and Mechanical Sensor Technologies
Krzysztof Iniewski
Biological and Medical Sensor Technologies
Krzysztof Iniewski
Tunable RF
Components
and Circuits
APPLICATIONS IN MOBILE HANDSETS
EDITED BY
Jeffrey L. Hilbert
W i S p r y, I n c o r p o r a t e d , I r v i n e, C a l i f o r n i a , U S A
Krzysztof Iniewski
MANAGING EDITOR
C M O S E m e r g i n g Te c h n o l o g i e s R e s e a r c h I n c.
V a n c o u v e r, B r i t i s h C o l u m b i a , C a n a d a
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2016 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a photo-
copy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
For Cyndie, who makes everything possible
and worthwhile, and who proves every day
that dreams can come true
Contents
Preface.......................................................................................................................ix
Acknowledgments................................................................................................... xiii
Series Editor.............................................................................................................. xv
Editor......................................................................................................................xvii
Contributors.............................................................................................................xix
vii
viii Contents
Tunable RF’s time has come. Arguably, it may be long overdue. With the continuing
explosive growth of mobile communications on a global scale, whether measured in
terms of users, handsets, data traffic, networks, and/or frequency bands, ubiquitous,
reliable wireless connectivity has never been more important. But innovation in the
portion of the mobile handset architecture that most directly enables network con-
nectivity, the radio frequency (RF) front end, has taken a backseat to continuing
advancements in other parts of the handset. Whereas most mobile handset users
have some idea about the functions and value of a faster processor, more memory,
or a larger screen since these components are more easily connected to observable
user benefits, the same is not true about the value or function of the antennas, ampli-
fiers, switches, and filters in the RF front end. (Witness the “Antennagate” episode
a few years ago, and the resulting calls I received from various reporters, some of
whom wanted to know what an antenna does and why it is so important.) Of course,
without a robust front-end solution, accessing the benefits provided by advance-
ments in other parts of the handset becomes, at a minimum, problematic. And it is
perhaps interesting to note that, when surveyed, a majority of handset users regu-
larly cite battery life, dropped calls, and data speed as among the biggest concerns
with their mobile service, all of which are heavily driven by the performance of the
RF front end.
So, why have advancements in the RF front-end architecture lagged those in the
rest of the handset? Perhaps the simplest answer is, until recently, complementary
metal oxide semiconductor (CMOS) technologies and design techniques have not
supported the levels of performance required of RF components. As a result, a col-
lection of various higher-performance specialty technologies (acoustics, ceramics,
magnetics, gallium arsenide semiconductors) have been utilized to implement the
various RF functions. These functions have been interconnected into chains of RF
components through 50 ohm interfaces, with each such chain optimized to work for
a specific mode or frequency of operation. As more modes and frequencies are added
to the handset, more chains are added, and interconnected by higher and higher
throw count switches. A second answer is that by throwing more processing power
and memory into the handset, and shifting more hardware functionality into soft-
ware, in other words, by using Moore’s law scaling, we have been able to (barely)
keep pace with demand. However, with no end to data demand in sight, and net-
work performance improvements already defined for future releases of 4G, we have
reached a tipping point where traditional RF front-end architectures have become
a major bottleneck in enabling the industrial design and performance attributes of
handsets that customers are demanding.
ix
x Preface
At first glance, in the context of the handset RF front end, the words “tunable”
and “RF” may not seem to belong together. We define a tunable RF function to be
one whose target frequency of operation and performance parameters are dynami-
cally adjustable under software control. Such adjustments are typically in response
to changes in the micro- and/or macro-environment of the handset. Current front-
end architectures are generally “switchable” at the RF chain level (as discussed ear-
lier) or between instances of components within a chain and as such, can broadly
be considered to be tunable but at a lower level of precision than possible with indi-
vidual RF components or modules. We will discuss both levels of “tunability” within
this book along with the substantial progress toward future RF front-end architec-
tures that are completely tunable without the use of switches at the RF chain level
(“switchless” architectures).
It is difficult to establish the exact start of today’s tunable RF market; however,
our company, WiSpry, along with Paratek Microwave (later purchased by Research
in Motion), was engaged in working with handset original equipment manufacturer
(OEM) customers on tunable impedance matching for antennas beginning about the
better part of 10 years ago. Since then, the market has grown rapidly, and today it would
be difficult to find a smartphone that does not employ at least one tunable RF func-
tion implemented using either electrostatic microelectromechanical systems (MEMS),
barium strontium titanate (BST), silicon-on-insulator (SOI) field effect transistors
(FETs), or, for some tunable functions, high-performance RF CMOS technology.
This book serves as a snapshot of the state of the art in tunable RF circa 2015.
The book is written by leading practitioners in the field, and between us, we comprise
a majority of today’s commercial market share for tunable devices in mobile handset
applications. The goal of the book is to provide a technical introduction to the field
and to document the foundational work that has been done to date. Chapter 1 serves
as an introduction and provides an overview of the tunable RF market along with a
glimpse into the future. Chapters 2 and 3, contributed by Qorvo (Greensboro, North
Carolina), ON Semiconductor (Nashua, New Hampshire) and BlackBerry (Waterloo,
Ontario, Canada), focus on RFSOI technology and BST technology, respectively,
along with an introduction to the applications of these technologies to tunable func-
tions. Chapters 4 and 5, contributed by Skycross (Viera, Florida) and Ethertronics
(San Diego, California), discuss the applications of tuning to mobile handset antenna
structures and systems. Beginning with Chapter 6, the next three chapters (con-
tributed by Cavendish Kinetics [San Jose, California], Peregrine Semiconductor
[San Diego, California], and ON Semiconductor [Nashua, New Hampshire]) dive
into antenna tuning applications utilizing RF-MEMS, SOI, and BST technologies,
respectively. Chapter 9 concludes the detailed discussion of antenna tuning by offer-
ing a perspective on the topic from a handset OEM, Huawei (San Diego, California).
In Chapter 10, Nujira (Cambridge, United Kingdom) provides an in-depth discussion
of power amplifier envelope tracking, a rapidly emerging and important technique
for improving efficiency. In Chapter 11, DelfMEMS (Villenuve D’Ascq, France)
discusses using RF-MEMS switches for a next-generation implementation of a
switchable RF front end. A case study of tunable radio architectures by Interdigital
(Melville, New York) is presented in Chapter 12, while Chapter 13 provides AT&T’s
(Atlanta, Georgia) network operator perspective on the evolution of the handset front
Preface xi
end. The book concludes with a chapter from LitePoint (Sunnyvale, California) on
production testing of wireless devices in the face of the continuing evolution of hand-
set architectures and the increasing drive toward shifting such testing from a con-
ducted to a radiated (over-the-air) performance basis.
Some readers may note a substantial portion of the book is devoted to antennas
and antenna tuning. This is by design, as the antenna tuning application dominates
the commercial market for tunable RF today. The rate of adoption of tunable RF and
the evolution of RF front-end architectures in mobile handsets continue to accelerate
so there is little doubt that future books on the topic will present a more diverse and,
perhaps arguably, a more balanced overview of a then more mature market.
Over my almost 40-year career in the semiconductor industry, I have been lucky
enough to contribute to a number of advances in technology, design, and design
tools. In the 1970s, it was CMOS technology and the start of commercial design
tools. In the 1980s, it was gate array and standard cell ASICs as both a supplier
and a user. In the early 1990s, it was systems-on-a-chip (SOC), followed in the late
1990s and early 2000s by the mainstreaming of MEMS technology. Then, since
about 2005, I have participated in developing the foundation of tunable RF. As
we move toward a world in which always-on connectivity and on-demand access
to information is available to everyone on a global scale, it seems possible that
sensing, computing, and wireless communications technologies will become so
pervasive as to become invisible in the fabric of society, perhaps as a step before
they become biologically integrated within us. I hope my luck holds and I get
another chance to play. If not, I will hope to find some small level of satisfaction
at having played a small role in bringing about (with apologies for paraphrasing
a portion of a Winston Churchill quote) “the end of the beginning” of something
that ultimately should be wonderful for all.
Jeffrey L. Hilbert
Dana Point, California
xiii
Series Editor
Krzysztof (Kris) Iniewski manages R&D at Redlen Technologies Inc., a start-up
company in Vancouver, Canada. Redlen’s revolutionary production process for
advanced semiconductor materials enables a new generation of more accurate, all-
digital, radiation-based imaging solutions. Dr. Iniewski is also a president of CMOS
Emerging Technologies Research Inc. (www.cmosetr.com), an organization of high-
tech events covering communications, microsystems, optoelectronics, and sensors.
In his career, Dr. Iniewski has held numerous faculty and management positions at
the University of Toronto, the University of Alberta, Simon Fraser University, and
PMC-Sierra Inc. He has published over 100 research papers in international journals
and conferences. He holds 18 international patents granted in the United States,
Canada, France, Germany, and Japan. He is a frequent invited speaker and has con-
sulted for multiple organizations internationally. Dr. Iniewski has written and edited
several books for CRC Press/Taylor & Francis, Cambridge University Press, IEEE
Press, Wiley, McGraw-Hill, Artech House, and Springer. His personal goal is to
contribute to healthy living and sustainability through innovative engineering solu-
tions. He can be reached at: kris.iniewski@gmail.com.
xv
Editor
Jeffrey L. Hilbert is the president and founder of WiSpry, Inc., a fabless semicon-
ductor company utilizing CMOS-integrated radio frequency microelectromechani-
cal system (RF-MEMS) technology to develop tunable RF products for the cellular
communications and wireless consumer electronics markets. WiSpry defined and
pioneered the rapidly growing tunable RF market segment. Hilbert has more than
37 years of executive management and technical experience in a number of lead-
ing semiconductor and MEMS companies, including LSI Logic, Compass Design
Automation, AMCC, Motorola, Harris, and Coventor. Early in his career, Hilbert
did pioneering work in CMOS technology and in IC design tools leading to today’s
design automation tools that are supplied by companies such as Cadence. As an
experienced entrepreneur, he has raised over $120 million in financing to fund two
consecutive start-up semiconductor companies over the past 15 years. He also has
board of director and advisory board experience in the commercial, government, and
academic arenas. Hilbert holds a BS in chemical engineering from the University of
Florida, an MS in computer science from the Florida Institute of Technology, and
has done course work toward a PhD in computer engineering from North Carolina
State University.
xvii
Contributors
Rob Brownstein Yuang Lou
Litepoint AT&T Network Architectures,
Sunnyvale, California Radio Access and Devices
Atlanta, Georgia
Frank Caimi
Skycross Paul McIntosh
Viera, Florida ON Semiconductor
Nashua, New Hampshire
Julio Costa
Qorvo Larry Morrell
Greensboro, North Carolina Cavendish Kinetics
San Jose, California
Alpaslan Demir
Interdigital James G. Oakes
Melville, New York BlackBerry
Waterloo, Ontario, Canada
Laurent Desclos
Tero Ranta
Ethertronics
Peregrine Semiconductor
San Diego, California
San Diego, California
Tanbir Haque
Sebastian Rowson
Interdigital
Ethertronics
Melville, New York
San Diego, California
xix
1 Tunable RF Market
Overview
Jeffrey L. Hilbert
CONTENTS
1.1 Introduction....................................................................................................... 1
1.2 The Mobile Handset Market..............................................................................3
1.3 Evolution of Handset RF Front-End Architectures...........................................4
1.4 What Is Tunable RF?.........................................................................................8
1.5 Market Snapshot: RF Front-End and Tunable RF........................................... 13
1.6 Tunable RF Component and Circuit Outlook.................................................. 16
1.7 Conclusions......................................................................................................20
References................................................................................................................. 21
1.1 INTRODUCTION
Mobile (wireless) communications has rapidly become embedded in the fabric of
society enabling and changing the ways in which we communicate in the broadest
sense. Cutting across geographic and political boundaries, age, differences in ethnic-
ity and religion, and independent of sharing and exchanging information one-to-one,
one-to-many, or many-to-many, the mobile or cellular (cell) phone has become a
ubiquitous personal appliance on a global scale.
It is difficult to imagine a world without mobile phones yet it has only been about
42 years since the first mobile phone call was placed by Martin Cooper of Motorola
in New York City in April of 1973. About 10 years later, Motorola began selling the
Dynatac 8000x. With an initial price of $3995, this phone offered mobile, analog
voice communications with up to 30 minutes of talk time in a form factor of only
13 in. × 1.75 in. × 3.5 in. (Figure 1.1).
From these humble beginnings, the mobile communications market has exploded
to change the way in which we live. At the same time, the mobile device has contin-
ued to rapidly grow in capabilities, performance, and talk time while shrinking in
size and weight. Figure 1.2 shows a current generation, 4G mobile phone, the LG G3.
This phone provides up to 21 hours of talk time in a form factor of 5.42 in. × 2.74 in.
× 0.41 in., not to mention almost unbelievable improvements in functionality, perfor-
mance, and ease of use. The contrast with the Dynatac 8000x is apparent.
In this chapter, we begin with a review of the status and projections for the mobile
handset market. As mobile phone form factors continue to evolve in response to
consumer demands, and performance and capabilities continue to grow with the use
1
2 Tunable RF Components and Circuits: Applications in Mobile Handsets
front-end components for Wi-Fi connectivity and GPS, we will focus solely on the
cellular RFFE and include the associated antennas as a part of the discussion (since,
as we will see later, tunable RF solutions both enable and potentially require, new
approaches to antenna design and new levels of antenna functionality). From an
architectural perspective, the RFFE can be viewed as one or more chains of RF
devices whose purpose is to enable and optimize high-frequency transmission and
reception of voice, data, and control signals (connectivity) between the handset and
the network. The functions performed by each RF chain include power amplifica-
tion, selection and rejection of signals (filtering), the actual radiation and reception
of signals by the antenna, and the switching of signals.
While the transceiver, baseband, applications processor and other digital and
relatively lower frequency components in the mobile handset have benefited greatly
from continuing advances in complimentary metal oxide semiconductor (CMOS)
technology and Moore’s law scaling, components in the RFFE have only recently
and selectively begun to enjoy similar benefits. Why has this been the case? Potential
reasons include the following:
This is not to say that there have not been consistent advances in the RFFE archi-
tecture. There have been albeit at a much slower pace than in the digital portions
of the handset. Newer combinations of design techniques and technologies have
largely supplanted some of the widely used specialty technologies in the past such as
magnetics. Other technologies have continued to scale in size while maintaining (or
improving) performance resulting in physically smaller, more power efficient, and
lower cost solutions. And advances in passive device, substrate, and packaging tech-
nologies have fueled continued integration at the module level providing multiband
and multimode capabilities within a single footprint. But a new set of requirements
and constraints in the form of 4G or LTE and LTE-Advanced (LTE-A) technology
6 Tunable RF Components and Circuits: Applications in Mobile Handsets
Low band
High band
WCDMA
AMPS
Single-band 3G
transceiver Transceiver
AMPS
has clearly highlighted that a new approach to scaling and implementing the RFFE
is required.
From the initial deployments of first-generation cellular technology, through
today’s explosive growth of 4G networks, one item has remained consistent—find
ways to do more at the product level with less (space, volume, power, cost, time).
Without general access to the benefits of CMOS technology, scaling the RFFE has
been achieved historically by duplication. Need support for a new frequency band or
mode of operation? Add an additional chain of RF components targeted at the spe-
cific frequency or mode. Repeat as required. Figure 1.3 provides a visual comparison
between the structural complexity of typical “1G” and 3G RFFEs.
While some of the advances in RFFE implementation mentioned earlier allowed
this approach to remain the de facto scaling technique up until the 3.5G time frame,
new form factor, performance, connectivity, power, time-to-market, and loss consid-
erations began to mitigate the benefits and limit usefulness with 4G.
Table 1.1 shows a snapshot of the frequency band plan for 4G/LTE on a global
basis. Once this plan is fully implemented, a truly global handset could need to pro-
vide support for up to 43 different frequency bands, up from an average of approxi-
mately 10 bands in 2010.6 In addition to the large increase in band count, 4G targets
maximizing network capacity by increasing peak data rates to the 100 Mbps to
1 Gbps range depending on the mobility of the user. To support these data trans-
fer rates, LTE-A includes support for multiple input/multiple output (MIMO) archi-
tectures employing multiple antennas and carrier aggregation (CA). CA is a new
capability which allows (depending on the LTE release level) between two and five
channels to be stitched together dynamically in software to provide a larger data
“pipe” for reception, transmission, or both. While enabling tremendous potential to
maximize network capacity, CA brings significant challenges to the RFFE, which
has relied heavily on fixed channel bandwidths, known spacing between transmit
and receive frequencies, and single feed antennas. The number of CA combinations
being planned and deployed is growing rapidly, and differs between geographies
and carriers making transparent global roaming a harder goal to achieve. CA can
be implemented using multiple channels within a band (“intra-band” CA) or using
different channels in different bands (“inter-band” CA), which along with transmit
CA, sets new, tougher requirements for the linearity and harmonic performance of
the RFFE in 4G handsets.
Tunable RF Market Overview 7
TABLE 1.1
LTE Frequency Bands and the Corresponding Regions5
Duplex
LTE Spacing BW Duplex Deployment in
Bands Uplink (MHz) Downlink (MHz) (MHz) (MHz) Mode the World
Band 1 1920–1980 2110–2170 190 60 FDD China, Japan, EU,
Asia, Australia
Band 2 1850–1910 1930–1990 80 60 FDD North and South
America
Band 3 1710–1785 1805–1880 95 75 FDD EU, China, Asia,
Australia, Africa
Band 4 1710–1755 2110–2155 400 45 FDD North and South
America
Band 5 824–849 869–894 45 25 FDD North and South
America,
Australia, Asia,
Africa
Band 6 830–840 875–885 45 10 FDD Japan
Band 7 2500–2570 2620–2690 120 70 FDD EU, South
America, Asia,
Africa, Australia
Band 8 880–915 925–960 45 35 FDD EU, South
America, Asia,
Africa, Australia
Band 9 1749.9–1784.9 1844.9–1879.9 95 35 FDD Japan
Band 10 1710–1770 2110–2170 400 60 FDD North and South
America
Band 11 1427.9–1447.9 1475.9–1495.9 48 35 FDD Japan
Band 12 698–716 728–746 30 18 FDD North America
Band 13 777–787 746–756 31 10 FDD North America
Band 14 788–798 758–768 30 10 FDD North America
Band 17 704–716 734–746 30 12 FDD North America
Band 18 815–830 860–875 45 15 FDD North and South
America,
Australia, Asia,
Africa
Band 19 830–845 875–890 45 15 FDD North and South
America,
Australia, Asia,
Africa
Band 20 832–862 791–821 41 30 FDD EU
Band 21 1447.9–1462.9 1495.9–1510.9 48 15 FDD Japan
Band 22 3410–3500 3510–3600 100 90 FDD
Band 24 1626.5–1660.5 1525–1559 101.5 34 FDD
(Continued )
8 Tunable RF Components and Circuits: Applications in Mobile Handsets
Diversity RFIC
FEM
Diversity
antenna port SPnT
Diversity
SPnT RX LNAs
mipi RFFE
mipi RFFE
SPnT
High
bands
mipi RFFE
FEMid
Main
antenna port
mipi RFFE
mipi RFFE
loop or closed loop. As an example, we can look at a tunable RF filter function. For
such a filter, we can think of the range of coverage as being the range of frequency
bands the filter can be tuned to cover. The number of steps would denote the num-
ber of discrete filter settings we can realize between a minimum and a maximum
value (the range of coverage) while the precision would indicate how precisely we
could set each filter value or state (± a percentage value). If the filter were being
tuned in operation by the baseband processor (a typical scenario) using a predefined
lookup table, we would refer to the filter as being controlled in an open-loop fashion.
Alternatively, if real-time information from the environment is being used to control
the filter (either directly or via the baseband), we would call this closed-loop opera-
tion. As with most options, decisions between open- and closed-loop control are
based on many variables including complexity, ease of integration, and cost.
To demonstrate another compelling advantage of tunable RF, we can expand
on our tunable RF filter example. Let us suppose that a certain handset needs to
support the following LTE (low) frequency bands: (1) Band 17 (704–746 MHz);
10 Tunable RF Components and Circuits: Applications in Mobile Handsets
(2) Band 13 (746–787 MHz); and (3) Band 5 (824–894 MHz). Using traditional fixed-
frequency filters, we would need to implement one filter (in this case, one duplexer)
per frequency band whereas if we had a tunable RF filter supporting the required
levels of performance and which could be tuned between 704 and 894 MHz with a
sufficient number of steps and precision, we could replace the three fixed filters with
one tunable device. Now let us further suppose that the handset supplier desires to
supply the same model of phone in a different geography or to a different carrier
who desires to cover Band 20 (791–862 MHz) instead of Band 5. The same tunable
RF filter can be used to cover the B17/B13/B20 combination in the second phone as
the B17/B13/B5 combination in the first phone. All that is required is a change in
software and/or lookup table contents. Thus, tunable RF provides the potential for
the type of superlinear scaling the RFFE of handsets needs to meet the challenges
of multi-geography (or super-geography) and global LTE handsets by enabling one
physical tunable function to replace multiple instances of its static counterpart.
Correspondingly, as shown in Figure 1.5, one tunable RF chain can replace multiple
fixed-frequency or mode-specific chains of components. In the limit (assuming suf-
ficient tuning capabilities), the number of physical chains of RF components can be
reduced to the maximum number of radio paths that can be active concurrently.
Tunable RF has the capability to assist the designer in meeting an increasingly dif-
ficult set of RFFE design constraints. Benefits to marketing can include r ealization
of desired form factors (industrial designs) including increasingly thinner designs,
all glass/screen fronts, and metal case/frame construction. Equally, if not more so,
are the business benefits available to the handset OEM. These include faster time
to market, superior performance and usability, reduction in the number of SKUs,
reduction in bill-of-materials (BOM) cost and complexity per handset, and the pos-
sibility of field upgrades to accommodate access to a new or re-farmed spectrum.
With all of the potential advantages provided by tunable RF, why have all handset
OEMs not adopted a tunable approach to implementing the RF front end? Reasons
include the following:
4G 4G
RF
Multimode tuning Multimode
transceiver transceiver
Tunable RF
components
Compelling reasons for adoption can come in many forms. Whereas there were clear
benefits to be realized with the adoption of tunable RF solutions for 3.5G and 3.9G
phones, it was not clear that adoption was more than a nice-to-have feature. But with
4G, there is a clear mandate for implementation of tunable solutions. Lower operat-
ing frequencies, higher data rate expectations from consumers enabled by CA and
MIMO architectures, return on investment (ROI) expectations from network opera-
tors on investments in 4G infrastructure, and continued ID design constraints have
resulted in a clear call for the evolution of the RF front end.
However, given the handset OEM concerns over risk, differences in implemen-
tation approach, and cost—How have tunable RF suppliers driven the adoption
of tunable components? By identifying an initial application where there is a high
need and substantial benefit which still allows the OEM to maintain their platform
investment—antenna tuning. Rather than starting in the middle of an existing RFFE
architecture (where a fan-in/fan-out challenge is created by inserting a tunable com-
ponent between static components), the antenna is located on one end of the RFFE.
Figure 1.6 shows a WiSpry impedance tuner located on a production phone board
next to the antenna contacts (and at one end of the handset).
Given the location, the addition of a tuner is possible with minimum perturbations
to the remainder of the existing RFFE design. This lowers the barrier to adoption and
as we will see in the next section, has driven initial production revenue growth of
tunable RF.
Antenna tuning can be categorized as follows based on the physical location/
connection of the tuner and the intended function:
We can also categorize antenna tuners based on the technology used for their
implementation. Today, three basic technology approaches are being used for the
implementation of antenna tuning solutions7:
• Switched capacitors: Typically, SOI FETs are used to switch between fixed
MIM capacitors.
• Barium strontium titanate (BST) capacitors: Are used to provide a smoothly
varying change in capacitance through a combination of a control chip and
one or more BST capacitor die.
• RF microelectromechanical systems (MEMS) capacitors: These use
mechanical structures that are movable (actuated) using electrostatic forces
to provide a set of digitally tunable capacitor values, typically monolithi-
cally integrated in a CMOS process.
(a) (b)
FIGURE 1.7 WiSpry RF-MEMS capacitor pair (a) and (b) array of 48 capacitors with an
integrated charge pump and digital control logic implemented in an 180 nm RF-CMOS.
Quality factors
4
3
Cost Intercept points
2
1
SOI
0 BST
$300
$150
$100
$50
$0
2012 2013 2014 2015 2016 2017 2018
FIGURE 1.9 Total revenue projection for discrete tunable components in the total RF
front-end market.
Tunable RF Market Overview 15
Finally, according to Taylor,14 the estimated market size for power amplifier (PA)
envelope tracking in 2015 is 251 M units increasing to 1 billion units in 2018. And
Yole15 forecasts initial sales of MEMS-based tunable filters and PAs starting in 2017.
Based on the market data presented earlier it is clear that adoption of tunable RF,
while growing rapidly, is still very much at an early stage. Meanwhile, the overall
RF front end TAM continues to grow at a substantial rate from its approximate
$9 billion base creating both a tremendous opportunity and need for rapid inno-
vation and evolution. Figure 1.10 provides a graphical summary of some of the
major innovations in handset architecture over the 14-year period between 1998 and
2012.14 As impressive as these achievements are, more progress is needed at a faster
rate to keep pace with the demands of 4G users and networks. In the next section, we
explore how we can best accelerate the evolution of the RFFE through the adoption
of tunable RF.
Antenna
2012: Antenna tuner Varicap
ICs (switch-
capacitors) Antenna
Filter Duplexer
switch
2013: CMOS multimode
PA and filter
multiband-PA-switch
(Qualcomm) PA
Drain
2014: CMOS envelope
modulator
tracker-PA (claimed by
RF axis) Filter Filter
2012: PA-transceiver (Intel)
Rx LNA
2003: CMOS transceiver
(GSM/GPRS) Transceiver
Tx Synth PMIC
2005: Baseband-transceiver
(GSM/GPRS)
Chipset
DAC ADC
1998: DBB with CPU + DSP
cores Baseband
CPU
DSP GPU
2010: Baseband-apps
processor (Qualcomm)
Apps processor
Memory
has continued and subsequent prototypes covering both low band and high band are
now operational in the lab.
Figure 1.13 provides a schematic representation of a complete LTE-A capable
tunable RFFE concept that WiSpry refers to as CAFE (Compact, Agile Front End).
As shown, this architecture inherently supports LTE CA and MIMO functions,
2800
2600
2400
Frequency (MHz)
2200
HB1
2000
1800
1600
1400
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Capacitance (pF)
High-band
measured filter response
0
–10
–20
S21 (dB)
–30
–40
–50
1200 1400 1600 1800 2000 2200 2400 2600 2800 3000
Frequency (MHz)
FIGURE 1.11 Dual multiband tunable filter module measured results. (Continued )
18 Tunable RF Components and Circuits: Applications in Mobile Handsets
1000
Frequency (MHz)
900
800
LB1
700 LB2
600
500
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Capacitance (pF)
Low-band
measured filter response
0
–10
–20
S21 (dB)
–30
–40
–50
500 600 700 800 900 1000
Frequency (MHz)
FIGURE 1.11 (Continued) Dual multiband tunable filter module measured results.
requiring only one additional tunable RF path for each potentially concurrently
active radio link.
Viewed in the context of the $7.00–$17.00 BOM opportunity for RF front ends
in 4G handsets, architectures such as CAFE look compelling. But at the same time,
they are revolutionary and, as such, they carry a great deal of perceived, and some
actual, risk. Fundamentally, one of the requirements of such an architecture is for
production, radiated test of the handset. Great progress toward this capability is
Tunable RF Market Overview 19
Tunable
Tunable Distributed
power
filters tunable
amplification
antennas
700–1000 RX
Antenna
integrated
700–1000 TX
Transceiver
Baseband
tunable
duplexing
MIMO RX and
impedance
1700–2700 RX matching
1700–2700 TX
being made by companies such as LitePoint (see Chapter 14), but it will take several
years until the necessary certification test cases and specifications are ratified, and
the production test infrastructure is in place. In the interim, tunable RF can continue
to grow market share and revenue bottom up through the implementation of addi-
tional and next-generation antenna tuners, tunable antennas, tunable PAs and LNAs,
tunable filters for applications such as coexistence, and perhaps subsets of CAFE-
like architectures on the receive side, while pursuing the top-down development of a
complete tunable narrowband front-end solution in parallel.
Ultimately, it is likely that no single design approach or technology will prove to
be the silver bullet but rather that future RFFE architectures will evolve to encom-
pass the best of both fixed and tunable functions, along with CMOS, acoustic, and
other existing and new implementation technologies. Whatever form progress takes,
20 Tunable RF Components and Circuits: Applications in Mobile Handsets
time is of the essence. The year 2020 will be upon us sooner than we think, and with
it, an entirely new level of even more daunting performance challenges with 5G.
1.7 CONCLUSIONS
Over the 42-year period since the first mobile phone call was placed, mobile
communications has changed the world in ways that we could not have imagined.
Mobile communications technology has evolved from voice communications over
an analog link, to a digital, data-centric architecture that provides for high-speed
downlink and uplink of data, downlink of high-speed streaming video, and voice
over data. Communications standards have evolved to the present fourth-generation of
technology, whose deployment in the market has ramped faster than any previous such
standard. Today, an astounding 62% of the global population is estimated to be mobile
phone users. Mobile handsets, and the RF semiconductors they employ have become
the growth engine of consumer electronics and in turn, of the overall semiconductor
industry. In 2015, it is estimated that a total of 1.8 billion new handsets will be sold,
an increasing majority of which are smartphones whose processing, memory, and
software capabilities will far surpass those of the leading PCs of just a few years ago.
As the mobile handset market has matured, different portions of the handset
architecture have evolved at different rates. Whereas the lower frequency, digital
portions of the handset have enjoyed the scaling benefits of Moore’s law, the RF
front end, the increasingly crucial link between the handset and the network, has
evolved much more slowly. Increasingly more challenging performance specifica-
tions coupled with the rapid proliferation of frequency bands on a global basis with
LTE, have positioned the RFFE as a key bottleneck to the next stages of innovation
in mobile communications.
To address this bottleneck, a new design paradigm has emerged over the past
decade. The concept of tunable RF, high performance RF components whose
frequency and functions can be dynamically adjusted via software, has entered
the front-end architecture initially in the form of antenna tuners. Such tunable RF
components enable the replacement of multiple components with a single hardware
device and reduce the number of chains of such components required to support
multiband operations to a number equal to that of the maximum number of concur-
rently active radios. From this perspective, we can consider tunable RF front ends as
the final step in the implementation of software defined radios.
The current tunable RF market is in the early stages of growth with only a
limited number of players in production revenue. Total revenues for the segment are
measured in the hundreds of millions of dollars, a small percentage of the overall
$9 billion RFFE market, which is growing rapidly. Of the three implementation
technologies being utilized for tunable RF devices today—MEMS, BST, and SOI—
the majority of the current market belongs to antenna tuners implemented using SOI
FET switches.
Looking ahead, tunable RF can be expected to grow faster than the RFFE
market. Antenna tuners will enjoy deeper penetration with average utilization per
handset pushing well beyond one per unit; antenna vendors will bring new types
Tunable RF Market Overview 21
of active and adaptive antenna systems to market that meet the volume constraints
imposed by desirable industrial handset designs while providing the levels of over-
the-air performance required to keep both consumers and network operators satis-
fied. MEMS-enabled tunable filters will roll out in the market and advances in the
application of tunability to power and low-noise amplifier applications will become
apparent. Finally, and perhaps most dramatically, completely new fully tunable
RFFE architectures will begin to appear on the market, most likely beginning with
coexistence filtering and diversity receive applications. Once the required standards
and production manufacturing capabilities supporting radiated test are deployed,
these architectures will complete their evolution into software defined, narrowband
designs, with broadband tunability. Such architectures will most likely be imple-
mented using hybrid approaches of fixed and tunable devices implemented in CMOS
and specialty technologies as most appropriate.
And finally, it is estimated that in the 2020 time frame, commercialization of 5G
technology will begin, bringing a new wave of challenges and opportunities as wire-
less communications and connectivity continue their seemingly inevitable march
toward becoming so pervasive and tightly integrated in the fabric of society as to
become invisible in daily life.
REFERENCES
1. Young, J., Mobile front end module—The battle between SIP & SOC, The 12th
International System-on-a-Chip (SOC) Conference, Exhibit, and Workshops, Irvine,
CA, October 2014.
2. International Telecommunications Union, Measuring the Information Society Report,
May 5, 2014, pp. 2–3.
3. IDC Press Release, Smartphone momentum still evident with shipments expected to
reach 1.2 billion in 2014 and growing 23.1% over 2013, May 28, 2014, www.idc.com.
4. Canaccord/Genuity, Smartphone demand, Daily Letter, pp. 2–3, November 3, 2014.
5. Derived from https//en.wikipedia.org/wiki/E-UTRA, Frequency bands and channel
bandwidths, accessed April, 2015.
6. Qorvo 2014 Investor Day Presentation, www.qorvo.com, accessed April 2015.
7. Steel, V. and Morris, A., Tunable RF technology overview, Microwave Journal, 55(11),
26–36, November 2012.
8. BMO Capital Markets, Digging into China, handset model update, Communications
Equipment Report, May 8, 2014.
9. Madden, J., Private communication, March 2015.
10. Charter Equity Research, Post December 2014 quarter wireless and semiconductor
industry analysis, February 13, 2015.
11. Barclays, Triquint semiconductor, filters still the best way to play LTE, Equity Research,
November 1, 2013.
12. Charter Equity Research, iPhone 6: A detailed teardown analysis, September 19, 2014.
13. Oppenheimer, LTEvolution, The implication for wireless semiconductor vendors
beyond LTE, Equity Research, March 25, 2013.
14. Taylor, C., Radio front-end integration driving cellular component market dynamics,
Strategic Analytics Presentation, Private communication, October 22, 2014.
15. Yole Development, RF MEMS market 2015 update, Private communication, March
2015.
2 RFSOI Technologies on
HR Silicon Substrates
for Reconfigurable
Wireless Solutions
Julio Costa
CONTENTS
2.1 Introduction to RFSOI on High-Resistivity Silicon Technology..................... 23
2.1.1 Technology Drivers for RFSOI-on-HR Silicon Substrates..................24
2.2 RFSOI-on-HR Silicon Substrates and Devices...............................................28
2.3 Switch Branch Design..................................................................................... 32
2.3.1 RONCOFF................................................................................................ 33
2.3.2 PSP Electrical Compact Model........................................................... 35
2.4 PACs: Programmable Array of Capacitors...................................................... 36
2.4.1 PAC Dynamic Range........................................................................... 38
2.5 RFSOI Applications in the Cellular Handset.................................................. 39
2.5.1 Standalone PACs.................................................................................. 39
2.5.2 RFSOI Impedance Tuners................................................................... 41
2.6 Future Trends in RFSOI-on-HR Silicon Technologies................................... 43
Acknowledgments.....................................................................................................44
References.................................................................................................................44
23
24 Tunable RF Components and Circuits: Applications in Mobile Handsets
technology offering the RF designer the needed versatility for multiple design RF
device configurations. This chapter will present a chronology of RFSOI technol-
ogy development in the wireless industry, as well as the basic device fabrication
and relevant physics, the different approaches to switch cell design, and traditional
trade-offs, including relevant RF device metrics such as linearity, insertion loss, and
intermodulation distortion. This chapter will also cover typical configuration of tun-
able RFSOI switch cells in tunable networks, including novel uses of RFSOI capa-
bilities in cellular 4G systems as well as the critical specifications needed in this
application space.
The chapter will focus entirely on mainstream RFSOI technologies fabricated on
high-resistivity (HR) silicon substrates, since these technologies are readily available
from many high-volume silicon foundries, leaving out other captive offerings for
wireless RF switching such as silicon-on-sapphire (SOS) or microelectromechanical
systems (MEMS) that also offer equivalent tunable switching devices for the RF
mobile space. RFSOI-on-HR silicon technologies are now offered through the very
popular MOSIS (www.mosis.com) multi-project-wafer (MPW) shuttle service, which
gives low-cost access to a high-performance RFSOI technology to virtually anyone
in industry/government/academia. Although the majority of applications for RFSOI
in the mobile space today consist of switching solutions (RF switches, impedance,
and aperture tuners), several RFSOI-on-HR linear power amplifiers (PAs) are also
starting to gain acceptance, suggesting high levels of integration in future RFSOI
product roadmaps.
It is important to note that the maturity of RFSOI-on-HR silicon technologies
in the mobile marketplace did not happen overnight. It was the product of constant
technological advances and excellent collaboration between foundries, SOI substrate
wafer suppliers, and the RF device community, showing very steady progress since
the first attempts were made in the early 1990s. In this chapter, we discuss some of
the many challenges RFSOI-on-HR silicon has gone through in the last two decades
to solidify its dominance in the cellular RF front end.
7.4
4.7
2.8
0.9 1.6
FIGURE 2.1 Mobile data growth. (From RFMD Cellular Group; CISCO’s public website,
http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-
vni/white_paper_c11-520862.html.)
FIGURE 2.2 Illustration of peak-to-average power ratios (PAPRs) for 3G, 3.5G, and 4G
(LTE) waveforms.
26 Tunable RF Components and Circuits: Applications in Mobile Handsets
The increase of PAPR values impose severe requirements in the voltage and
power handling as well as harmonic generation and reliability of the technologies
used in the front end of mobile systems.
Figure 2.3 lists the current frequency band spectrum allocation to the different
bands, as defined by the 3GPP Mobile Broadband Standard Group.*†
This group is a consortium of industry, academia, and government agencies that
is responsible for the definition and standardization of the communication protocols
of the different bands used in mobile applications. The characteristics and definition
of each frequency band is an intricate product of several different factors, owing to
the fact that spectrum licenses are allocated in each separate countries to different
carriers and operators.
The introduction of carrier aggregation (CA) in the newer 4G mobile releases
added a significant layer of complexity and enhanced linearity requirements on all
RF front-end components. In the first releases of carrier aggregation, a second band
Duplex spacing
Width Width
Uplink Gap Downlink
band band
Frequency
FIGURE 2.3 LTE frequency division duplexing (FDD) and time division duplexing (TDD)
as defined in the 3GPP Mobile Standards Group in September, 2012.
* Official releases of the 3GPP Mobile Standards Group are kept at www.3gpp.org/releases.
† GPP Standards Update, Moray Rumney, http://www.home.agilent.com/upload/cmc_upload/All/
25Oct12LTE.pdf?&cc=DE&lc=ger.
RFSOI Technologies on HR Silicon Substrates 27
can be utilized at the same time to improve the downlink capacity of the handset to
receive more data; TX CA will also be available in the near future to improve uplink
capacity although currently a clear priority exists in the launch of RX carrier aggre-
gation RF front ends since that is currently where the peak demand for data is con-
centrated. With CA, the technology requirements for RF devices become even more
accentuated, especially for frequency division duplexing (FDD) band combinations
where TX and RX functions are done at the same time. Figure 2.4 depicts the bands
combinations that are planned for LTE 4G carrier aggregation.1
The inset of Figure 2.4 also illustrates why linearity requirements are much
more demanding in the CA mode in the FDD operation; for this particular example
of Bands 4/17 carrier aggregation—and there are numerous other troubling
combinations—a harmonic caused by the TX B17 signal in the 700 MHz band can
fall exactly on the RX B4 2100 MHz band. If this harmonic power level is suffi-
ciently high, it will “jam or block” the intended RX signal. This unwanted harmonic
product will, of course, look just like RX noise and the transceiver will be unable
to receive the desired RX B4 signal. This particular CA problem is referred to as
“carrier desense” and will invariably cause a condition where the communication
between the mobile terminal and the base station is seriously affected.
Especially in FDD bands—where the TX/RX (transmit/receive) functions take
place at the same time—there are extremely demanding requirements in the lin-
earity of the devices and components that are on the TX/RX signal paths. Such
concurrent RX/TX paths occur, of course, at the antenna of the handset and at all
of the components in the downstream RF path directly connected to the antenna.
The critical devices where such high linearity comes at an absolute premium are at
Region Bands CA
bands Third
harmonic
4 + 17, 2 + 17,
2 + 29, 4 + 29,
2 + 13, 4 + 13,
2G/3G: 1, 2, 4, 5, 8 2 + 30,
LTE: 12 + 30, Ftx-B17 Desense in B4 RX
5 + 30,
US 2, 4, 5, 12, 13, 17, 26, 29, 30, 17 + 30, B17 Tx B17 Rx B4 Rx
41 29 + 30,
Roaming LTE: 1, 3, 7, 20, 28 704 M 716 M 734 M 746 M 2110 M 2155 M
4 + 30, 4 + 5,
5 + 17,
Receiver B17-
26 + 41,
Receiver B17
Receiver B4-
Receiver B4
4 + 12
MIMO
MIMO
FIGURE 2.4 Carrier aggregation (CA) modes as defined in current 3GPP standards. The
right side of the figure illustrates a potential problem created by having harmonics of the TX
signal fall exactly on the band of the CA-RX signal.
28 Tunable RF Components and Circuits: Applications in Mobile Handsets
100
90
4G LTE + CA
IIP3(dBm)
80
70 4G LTE
60 3G
50 2G
2012 2013 2014 2015 2016 2017 2018
FIGURE 2.5 Required third-order input intercept point (IIP3) for RF front-end components
as a function of evolving cellular standards. (IIP3: From Larry Schumacher, Intel Mobile
Corporation, Challenges for radios due to carrier aggregation requirements, International
Wireless Industry Consortium [IWPC] Workshop, November 6, 2012.)
the front-end (FE) switch and any of the impedance/antenna tuner components that
are now commonly attached to smartphones being released in the cellular market.
Transceiver chipset makers are anticipating that 4G LTE/CA modes will require RF
front-end components with extremely high IIP3 (third-order input intercept point)
values—in the 87–90 dBm range—substantially higher than any linearity level
previously accomplished with any RF switching technology (see Figure 2.5). This
will require that every possible source of distortion in the front-end components be
thoroughly investigated and minimized to the absolute minimal degree. In addition
to high linearity requirements, these front-end switching blocks are also required to
have very low insertion loss for the ON state and high isolation for the OFF state.
Through intensive collaboration between academia, silicon wafer foundries, SOI
substrate providers, and the mobile RF component makers, RFSOI-on-HR today
has largely displaced traditional III–V and silicon-on-sapphire FET power switching
solutions due to its improved performance, reduced size and excellent manufactur-
ing yields, and low costs. RFSOI solutions are now present in essentially every major
handset platform, providing state of the art performance in increasingly complex
multi-throw RF switches, tunable RF capacitance arrays, and antenna RF tuners.
and power dissipation over conventional bulk digital and analog technologies
were possible. Honeywell,2 Motorola,3 and Harris4 as well as many in academia in
the United States and several other research laboratories in Japan published sev-
eral leading articles on RFSOI on silicon substrates, identifying the benefits and
trade-offs for RF building blocks inherent in a thin film silicon layer over oxide.
Motorola published extensively in the mid-1990s on the demonstration of several
key RF building blocks such as switches, LNAs, mixers, and power amplifiers built
on RFSOI technologies,5 also demonstrating voltage charge pumps and digital con-
trol with a high degree of integration. Similar development efforts were also being
conducted in Asia and Europe at the same time. Honeywell, however, is generally
credited with commercially offering the very first RF switches and RF components
built on RFSOI on silicon substrates in the late 1990s, with many of the design fea-
tures that are used to this day in our 3G/4G switches; the company also pioneered the
offering of the RFSOI-on-silicon process as a general foundry option and very early
realized the potential benefits of high-resistance substrates and cross talk isolation
for superior RF performance.* On the cellular phone side, Peregrine Semiconductor
is generally credited with achieving the first SOI design win in a mobile handset in
the early 2000s, with its own proprietary silicon on sapphire (SOS) technology.6
In order for RFSOI-on-HR silicon technologies to compete—and eventually
replace—other conventional switching technologies such as III–V PHEMT (pseudo-
morphic high electron mobility transistor) and SOS solutions, three completely
separate technological challenges had to be resolved in order to bring about the nec-
essary improvements in RF performance, yield, and cost that would eventually make
RFSOI-on-HR the leading technology that it is today:
1.
SOI device yield: RFSOI-on-HR silicon technologies typically employ sili-
con device regions with layer thicknesses in the 800–2000 Å range; such
thin layers must be controlled very uniformly across the wafer and must be
free of defects and other issues that negatively affect CMOS device charac-
teristics and reliability. Conventional SOI wafer techniques involving wafer
bonding and thinning of the silicon epitaxy layer proved to be extremely
challenging for layer thicknesses under 1000 Å. An alternative technique
called SIMOX yielded the first generation of RFSOI substrates used in early
demonstrations in the 1990s, where the BOX (buried oxide) layer was built
by the thermal oxidation of a high-density oxygen implant. Whereas the
SIMOX technique appeared on paper to offer a significant improvement
in cost and simplicity over other SOI bonding techniques, it was quickly
determined to be fraught with high levels of defectivity and severe prob-
lems with wafer uniformity and device yield. The “SMART-CUT” process
developed by SOITEC in France is credited with providing the key needed
AM
OL
TH I
CA
CA
. LOV
. VNIVERSITAS .
MIM
ANIENSIS .
MT
AE
SE D
1425 M1
TI
S.
S A PIEN FET
E
RES
(a) (b)
AM
MIM
MT
M1
FET
RES
Poly Trench
contact
Box
High-dose implant
FIGURE 2.7 IBM 7RFSOI NFET transistor cross section, also showing the implementa-
tion of an MIM capacitor and polysilicon resistor as well as the harmonic suppression process
(HSP) trenches and implants.
RFSOI Technologies on HR Silicon Substrates 33
The substrate resistivity must be high enough with respect to the frequency of
operation to allow the substrate to behave strictly as a capacitive network, which is
easily accomplished with today’s available values in the range of 1–3 kΩ· cm. The
stacked group of the RFSOI NFETs is the basic building block for essentially any
application in this field. At its core, lies the basic thick-gate NFET switch device.
Today, nearly all RFSOI on-HR silicon switch and tuner solutions use a 0.25 um
NFET output (also known as “thick gate”) device built on either a 0.18 um or a 0.13 um
SOI CMOS foundry technology, employing many different layout configurations to
optimize the RF performance of the switch branch. The thinner gate oxide option in
the technology is typically omitted to minimize costs and cycle time, although with
the current drive to integrate low-noise amplifiers (LNAs) in front-end modules, more
foundries are offering the thinner gate FET as a device option for such applications.
PFET devices are typically not utilized in RF switch branches due to their inherently
poorer carrier mobility and higher RDSON characteristics, although PFETs gener-
ally play a very important role in the implementation of i ntegrated power controllers
for RF applications.
RFSOI switch branches are typically biased using gate switching between +VDD/
VNEG in the ON/OFF states respectively, where VDD and VNEG are high and low
voltage levels supplied by the power management block of the IC. Typical VDD and
VNEG voltages used in today’s RFSOI applications employ values around ±2.5 V,
respectively. The drain and source voltages are biased at 0 V. Conventionally, body-
contacted FETs are used, where the body would be biased through high-value resis-
tors and typically be switched between 0 V/VNEG in the ON/OFF states respectively.
2.3.1 RONCOFF
A key figure of metric (FOM) utilized in RFSOI technologies is the product of
the ON resistance of a switch device with its respective OFF state capacitance
(RONCOFF), generally expressed in femtoseconds. When associated with a specified
device breakdown value and power handling, the RONCOFF product is a valid pre-
dictor of the device performance in an RF switch application. RFSOI-on-HR sili-
con technologies currently report RONCOFF figures of merit in the 240–150 fs range.
Several new improvements are currently being developed to continuously decrease
the switch FOM, such as the introduction of all copper metallization, finer lithogra-
phy for improved gate pitches, and several other proprietary techniques will reduce
this figure to the 110–150 fs range in 2015–2016 time frame.* It is important to keep
in mind that this FOM is only meaningful when comparing RFSOI technologies of
similar RF power handling and DC breakdown. Constant improvements in process-
ing techniques and materials will undoubtedly push RFSOI figures of merit to the
sub-100 fs range in 2017–2018 time frame.
In the OFF state, the power handling of a given switch branch is determined by
the number of stacked NFETs in series, as summarized in Figure 2.8.
RG RG RG RG RG RG
RGC
1
Rgate >>
VGATE 2πf(Cgs + Cgd)
FIGURE 2.8 Typical RFSOI switch branch topology, showing N× stacked switch FETs,
each DC-biased via a high-value gate resistor according to the frequency of operation and
the technology OFF state capacitance. High-value resistors may also be employed across the
source–drain terminals of the device for biasing and voltage distribution.
The goal of the combination of stacked switch devices and gate and body resis-
tors is to equally divide a large RF voltage into smaller RMS values consistent with
the breakdown characteristics of each individual transistor. This stacking property
of RFSOI NFET devices is, of course, identical to the strategies utilized in the
past by III–V switch designers as summarized in a classic paper by M. Shifrin.14
Shifrin also highlighted the need for high-value DC biasing gate resistors, with the
required resistor value being a function of the RF of operation and the OFF state
capacitance associated with the given switch branch transistor. In RFSOI switches
and tuner applications, typical stack numbers may vary between N = 6 and N = 24,
depending on the magnitude of the RMS voltage the branch must withstand in
the OFF state. The switch designer must pay close attention to the harmonics of
the OFF state device under the highest rated RF power levels and mismatched, as
harmonics levels will start increasing dramatically prior to the onset of voltage
breakdown.
Once properly designed for a given power level/insertion loss/isolation require-
ment, switch branches become the fundamental building block of a myriad of differ-
ent RF applications such as power switches, programmable attenuators, impedance
tuners, tunable networks, and programmable filters, employing a combination of
series and shunt branches and other internal and external passive and active compo-
nents.15,16 Figure 2.9 illustrates the block diagram of a typical single pole–four throw
(SP4T) antenna switch application with the different series and shunt branches and
its associated small-signal equivalent circuit. The shunt branches are key elements
that provide necessary port-to-port isolation and the series branches are the blocks
that connect to the antenna (the single pole) and hence are the key elements that
determine the insertion loss of each branch.
Typical RFSOI on-HR silicon switch products in the market today integrate all of
the necessary power controllers and digital logic blocks needed for a given applica-
tion. Most RFSOI technologies offered by foundries also offer “e-fuse” elements,
which are one-time programmable elements useful for calibration or identification
of a switch or tuner IC.
RFSOI Technologies on HR Silicon Substrates 35
ANT
ANT RSE
RF1
CSE CSE CSE
RF1 RF3 CSH
RF2 RF3 RF4
RSH RSH RSH
RF2 RF4
16
Switch branch bias
voltages and control
Controller
FIGURE 2.9 A typical block diagram of a single pole–four throw (SP4T) RFSOI switch
showing the combination of switch branches in a given application and its small-signal equiv-
alent circuit. The power handling of the switch is a function of the configuration and design
of each of the individual OFF state switch branches. Nearly all RFSOI circuits employ inte-
grated biasing controllers.
RFSOI on-high-HR silicon technologies today provide the designer with a complex
process design kit (PDK) that includes design rule checking (DRC), layout versus
schematic (LVS) checking, and PEX, in addition to extensive digital/analog/ESD
and effuse library elements.
(a)
(b)
1
Qmin = (2.1)
((Cmax /Cmin ) −1)⋅ R ON ⋅ COFF
The Qmin equation assumes that the Cmin is solely determined by the COFF of the
switches and that the FETs have been sized to meet the target Cmin. In moderate Q
applications the target Cmin can be achieved by placing a high-quality MIM or MOM
capacitor in parallel to a PAC structure. This gives only a modest Q improvement
compared to scaling the FET switch size but has the benefits that the Cmin correlates
closer to Cmax and the die area is not dominated by a huge switch.
The design of a PAC consists in trading off chip area for the necessary capaci-
tance tuning ratio and effective Quality factor for a maximum RF voltage waveform;
obviously the better the RONCOFF figure of merit for a given RFSOI technology, the
lesser the impact of the switch branch. This trade-off between capacitor tuning ratio
(Cmax/Cmin) and switch branch resistance and technology figure of merit RONCOFF was
summarized in a classic RFSOI paper by Whatley et al.,18 as depicted in Figure 2.11.
Clearly, moving forward with further improving the RONCOFF figure of merit will
play a very positive role in minimizing the trade-off of Q versus tuning ratio for
future RFSOI on high-resistivity silicon technologies.
MIM capacitors themselves are capable of generating second- and third-order har-
monics and also having significant temperature-dependent effects, although these are
typically below the levels of harmonics created by the series RFSOI switch blocks.
38 Tunable RF Components and Circuits: Applications in Mobile Handsets
RF+
CMIM . COFF
Cmin = (2bits–1)
CMIM + COFF
80
Q
60
40
20
0
1 2 3 4 5 6 7 8 9 10
Tuning ratio
FIGURE 2.11 Trade-off of Quality factor Q versus tuning ratio of an SOI switched MIM
capacitor for different RONCOFF figures of merit. (After Whatley, R. et al., RF front-end tunabil-
ity for LTE handset applications, Compound Semiconductor Integrated Circuit Symposium
(CSICS), 201018.)
2.5.1 Standalone PACs
Individual RFSOI PACs are useful and versatile components that find use in many
tuning applications. These components are offered in a number of different binary
configurations (from 4-bit to 32-bit binary tuning) depending on the necessary mini-
mum capacitor resolution needed for the application. A 4-bit binary weighted PAC
design that provides a 0.5–5 pF tuning range is shown in Figure 2.12.1 It contains
all of the digital/analog circuitry, including the serial interface (GPIO or RFFE),
charge pump, ESD network, and switch drivers to provide extremely small and cost-
effective tuning solutions. This very compact and highly integrated design and other
similar configurations are being shipped in extremely high volume in antenna appli-
cations in the cellular handset market.
The measured performance for a 4-bit PAC on using an SOI process with a 220 fs
FOM (IBM 7RFSOI) is shown in Figure 2.13. The device’s Q falls short of the Qmin
predicted by Equation 1. This deviation from the ideal is caused by a combination of
the loading of the biasing network, routing resistance, and parasitic capacitances that
reduce the net Cmax/Cmin ratio. These effects can, however, be readily simulated by
40 Tunable RF Components and Circuits: Applications in Mobile Handsets
1×C
4×C
2×C
CTRL // supply
VDD+ VDD–
VDD+ VDD–
C0 C1 C2 C3
C0 C1 C2 C3
FIGURE 2.12 Tunable array of capacitor built with RFSOI switch branches and MIM
capacitors. (From Costa, J., IEEE Microwave Mag., 15(7), S61, November–December,
2014.)
10
40
4
35
3.5 8
30
3
R (ohm)
Cs (pF)
Q (-)
25 6
2.5
20
2 4
15
1.5
10 f = 1000 MHz
f = 2000 MHz 2
1 5
0.5 0 0
0 5 10 15 0 5 10 15 0 5 10 15
PAC state PAC state PAC state
Total extracted series inductance = 0.131 nH (preceding data are de-embedded for this);
shunt capacitance range = 0.51–4.65 pF
FIGURE 2.13 A 4-bit PAC capacitance, Q and series resistance versus tuning state.
RFSOI Technologies on HR Silicon Substrates 41
adding parasitic-extraction (PEX) generated networks, which will correct for these
parasitic networks.
A recent comparison of published PAC performance21 is shown in Figure 2.14,
including results reported for PACs, implemented in alternative technologies such as
microelectromechanical systems (MEMS) and silicon on sapphire (SOS).20–23
FIGURE 2.15 Antenna tuner implemented with a combination of RFSOI die and external
high-Q inductors. The die area is 2.4 × 2.6 mm and the PAC covers a capacitor range between
Cmin = 0.5 pF and Cmax = 4.6 pF.
Figure 2.15 depicts a broadband SOI impedance tuner that has been utilized in
very high volume in several leading 4G handsets, which covers a capacitor range of
Cmin = 0.5 pF to Cmax = 4.6 pF, for a die size of 2.4 mm × 2.6 mm.
This particular PAC design serves as the configurable element in an antenna
impedance tuner designed to cover the frequency range between 0.7 and 2.2 GHz.
In this particular application, the overall Q of the entire impedance tuner ended up
being limited just as well by the Q of the external inductors used along with the PAC
(typical inductor Q’s at 1 GHz hover in the 40–55 range).
The die is implemented in a popular RFSOI-on-HR silicon technology
(IBM 7RFSOI). The impedance tuner depicted performs very well in matching
and optimizing the varying impedance characteristics to the RF front-end module
for a given antenna. The net gain or loss of this particular RFSOI antenna tuner at
895 MHz is illustrated in Figure 2.16; similar charts are available for this same tuner
in different frequency band.
The RF antenna tuner gain chart of Figure 2.16 is a very useful tool to visualize
the performance of a given RF tuner in a given frequency band under a different set
of load characteristics. This particular chart shows that the insertion loss of the tuner
is about 0.56 dB; it is also very useful to determine the “breakeven” state, which
corresponds to the GTUNER = 0 dB where the impedance tuning essentially compen-
sates for the insertion loss of the tuner itself. Outside of the “breakeven” circle, there
is essentially a net gain in performance due to the RFSOI tuner, so obviously, the
designer seeks the optimal combination of RFSOI tuner and external components to
realize the smallest possible “breakeven” circle.
The antenna tuner depicted in Figure 2.16 shows an insertion loss of 0.56 dB at 1:1
VSWR. This loss, however, is recovered very quickly with VSWRs in the ≥2 range,
and as VSWR increases, the tuner realizes net positive gain for the overall network
showing a clear improvement in typical antenna implementations.
RFSOI Technologies on HR Silicon Substrates 43
3
150 1 30 5
5 2
0.5 1
4
25 0
3
5 4
0.2
0.5
0.2
2
5
–
4
180
3
4
0.50.25
3
3
1
0
2
.2
1
0 –0
0.25 2
5
0.5
4
210 2 330
1
2 1
3 3 4
4 5
5
240 300 0
Tuner PAC states
optimized to give 270
Red circles represent
maximum GTUNER load VSWR of 2, 3, and 5
GTUNER = –0.56 dB @ 50 Ω
FIGURE 2.16 Low-band (895 MHz) impedance tuner performance. The colors in the circle
represent the effective gain/loss of the combined impedance tuner and antenna. At VSWR’s
levels approaching 2, the tuner starts providing a net performance gain for the overall circuit,
and at higher VSWR levels, the gain improvement becomes very significant.
ACKNOWLEDGMENTS
The author would like to acknowledge the invaluable technical contributions of the
following engineers in the preparation of the materials covered in this chapter:
Mike Carroll, Phil Mason, Dan Kerr, Christian Iversen, Ali Tombak, Beth Glass,
Nadim Khlat, Marcus Granger-Jones, and Eddie Spears.
REFERENCES
1. Costa, J., Passing the plateau of productivity: Development of RFSOI technologies on
HR silicon substrates for reconfigurable wireless solutions, IEEE Microwave Magazine,
15(7), S61–S73, November–December 2014.
2. Liu, S.T., Fechner, P., Balster, S., Dougal, G., Sinha, S., Chen, H., Shaw, G., Yue, J.,
Jenkins, W.C., and Hughes, H.L., A 5 nm nitrided gate oxide for 0.25/spl mu/m SOI
CMOS technologies, IEEE Transactions on Nuclear Science, 46(6), 1824–1829,
December 1999.
3. Huang, W.M., Ngo, D., Babcock, J.I., Shin, H.C., Welch, P., Racanelli, M., Foerstner, J.,
Ford, J., and Cheng, S., TFSOI complementary BiCMOS technology for low power RF
mixed-mode applications, IEEE Custom IC Conference, 1996.
4. Krull, W.A., Buller, J.F., Rouse, G.V., and Cherne, R.D., Electrical and radiation char-
acterization of three SOI material technologies, IEEE Circuits and Devices Magazine,
3(4), 20–26, July 1987, doi: 10.1109/MCD.1987.6323129.
5. Lovelace, D., David, N., Costa, J., and Camilleri, N., Silicon MOSFET technology
for RF ICs, Sixth IEEE International Symposium on Personal, Indoor and Mobile
Radio Communications, 1995 (PIMRC’95). Wireless: Merging onto the Information
Superhighway, Toronto, Ontario, Canada, vol. 3, p. 1238, September 27–29, 1995.
6. Kelly, D., Brindle, C., Kemerling, C., and Stuber, M., The state-of-the-art of silicon-
on-sapphire CMOS RF switches, IEEE Compound Semiconductor Integrated Circuit
Symposium (CSIC ’05), 2005.
7. Wittkower, A., Auberton-Herve, A., and Maleville, C., SMART-CUT(R) technology
for SOI: A new high volume application for ion implantation, Conference on Ion
Implantation Technology, Alpbach, Austria, pp. 269–272, 2000.
8. Rong, B., Burghartz, J.N., Nanver, L.K., Rejaei, B., and Van der Zwan, M., Surface-
passivated high-resistivity silicon substrates for RFICs, IEEE Electron Device Letters,
25(4), 176–178, April 2004.
9. Kerr, D.C., Gering, J.M., McKay, T.G., Carroll, M.S., Roda Neve, C., and Raskin,
J.-P., Identification of RF harmonic distortion on Si substrates and its reduction using a
trap-rich layer, IEEE Topical Meeting on Silicon Monolithic Integrated Circuits in RF
Systems, 2008 (SiRF 2008), pp. 151–154, January 23–25, 2008.
10. Ben Ali, K., Roda Neve, C., Gharsallah, A., and Raskin, J.-P., RF SOI CMOS tech-
nology on commercial trap-rich high resistivity SOI wafer, IEEE International SOI
Conference (SOI), pp. 1–2, October 1–4, 2012.
11. Botula, A., Joseph, A., Slinkman, J., Wolf, R., He, Z.-X., Ioannou, D., Wagner, L.
et al., A thin-film SOI 180nm CMOS RF switch technology, IEEE Topical Meeting
on Silicon Monolithic Integrated Circuits in RF Systems, 2009 (SiRF ’09), pp. 1–4,
January 19–21, 2009.
12. Carroll, M., Kerr, D., Iversen, C., Tombak, A., Pierres, J.-B., Mason, P., and
Costa, J., High-resistivity SOI CMOS cellular antenna switches, Annual IEEE
Compound Semiconductor Integrated Circuit Symposium, 2009 (CISC 2009), pp. 1–4,
October 11–14, 2009.
RFSOI Technologies on HR Silicon Substrates 45
13. Lee, T.-Y. and Lee, S. Modeling of SOI FET for RF switch applications, Radio
Frequency Integrated Circuits Symposium (RFIC), 2010 IEEE, pp. 479–482,
May 23–25, 2010.
14. Shifrin, M.B., Katzin, P.J., and Ayasli, Y., IEEE Transactions on Microwave Theory
and Techniques, 37, 2134–2141, 1989.
15. Blaschke, V., Zwingman, R., Hurwitz, P., Chaudhry, S., and Racanelli, M., A linear-
throw SP6T antenna switch in 180 nm CMOS thick-film SOI, 2011 IEEE International
Conference on Microwaves, Communications, Antennas and Electronics Systems
(COMCAS).
16. Wang, D., Wolf, R., Joseph, A., Botula, A., Rabbeni, P., Boenke, M., Harame, D., and
Dunn, J. High performance SOI RF switches for wireless applications, International
Conference on Solid-State and Integrated Circuit Technology (ICSICT), 2010.
17. Wu, W., Li, X., Gildenblat, G., Workman, G., Veeraraghavan, S., McAndrew, C., van
Langevelde, R. et al., PSP-SOI: A surface potential based compact model of partially
depleted SOI MOSFETs, IEEE Custom Integrated Circuits Conference, 2007 (CICC
’07), pp. 41–48, September 16–19, 2007.
18. Whatley, R., Ranta, T., and Kelly, D., RF front-end tunability for LTE handset applica-
tions, Compound Semiconductor Integrated Circuit Symposium (CSICS), 2010.
19. Babcock, J.A., Balster, S.G., Pinto, A., Dirnecker, C., Steinmann, P., Jumpertz, R., and
El-Kareh, B., Analog characteristics of metal-insulator-metal capacitors using PECVD
nitride dielectrics, IEEE Electron Device Letters, 22(5), 230–232, May 2001.
20. Shim, Y., Wu, Z., and Rais-Zadeh, M., A high-performance, temperature-stable,
continuously tuned MEMS capacitor, 2011 IEEE 24th International Conference on
Micro Electro Mechanical Systems (MEMS), pp. 752–755, January 23–27, 2011.
21. Natarajan, S.P., Cunningham, S.J., Morris, A.S., and Dereus, D.R., CMOS integrated
digital RF MEMS capacitors, 2011 IEEE 11th Topical Meeting on Silicon Monolithic
Integrated Circuits in RF Systems (SiRF), pp. 173–176, January 17–19, 2011.
22. Baxter, B., Ranta, T., Facchini, M., Dongjin, J., and Kelly, D., The state-of-the-art in
silicon-on-sapphire components for antenna tuning, 2013 IEEE MTT-S International
Microwave Symposium Digest (IMS), pp. 1–4, June 2–7, 2013.
23. Im, D. and Lee, K., Highly linear silicon-on-insulator CMOS digitally programmable
capacitor array for tunable antenna matching circuits, IEEE Microwave and Wireless
Components Letters, 23(12), 665–667, December 2013.
24. Larry Schumacher, Challenges for radios due to carrier aggregation requirements,
International Wireless Industry Consortium (IWPC) Workshop, Intel Mobile
Corporation, November 6, 2012.
3 BST Technology for
Mobile Applications
James G. Oakes and David W. Laks
CONTENTS
3.1 Introduction..................................................................................................... 47
3.2 BST Tunable Capacitor Technology and Structure......................................... 48
3.2.1 Barium Strontium Titanate.................................................................. 49
3.2.2 Capacitor Structure.............................................................................. 50
3.3 BST Device Design with Application to RF Performance.............................. 51
3.3.1 DC and Low-Frequency Characteristics............................................. 51
3.3.2 Quality Factor...................................................................................... 53
3.3.3 Linearity.............................................................................................. 58
3.3.4 Power Handling...................................................................................60
3.4 Reliability and Qualification........................................................................... 61
3.4.1 Acceleration Factors............................................................................ 61
3.4.2 Mission Life Profile............................................................................. 62
3.4.3 Early Failure Rate................................................................................ 63
3.4.4 Qualification Tests...............................................................................64
3.5 Tuning Applications.........................................................................................64
3.5.1 Generic Tuner......................................................................................64
3.5.2 Commercial Antenna Tuners...............................................................66
3.6 BST Technology’s Future................................................................................ 68
3.7 Conclusions...................................................................................................... 69
Acknowledgments..................................................................................................... 70
References................................................................................................................. 70
3.1 INTRODUCTION
Mobile devices are used in a growing number of consumer applications driving
up the demand for both voice and data from carriers. As a result of this demand,
the carriers must provide high data rates for their customers, and the mobile phone
OEMs have to provide cost-effective, high-performance, and attractive mobile
device designs for the consumers. The carriers have responded to the increase in
data rates by building a roadmap for the increasingly efficient use of the frequency
bands that they have licensed. The mobile phone OEMs have improved the phone
handset by creating slimmer phones using cosmetically attractive materials (such
as metal backs, metal rings) and operating in the frequencies on the carriers’ road-
map. But the mobile phone OEMs are being challenged by the multiple frequency
47
48 Tunable RF Components and Circuits: Applications in Mobile Handsets
bands now used (from 700 to 2700 MHz expanding to 600 to 3500 MHz in the
next 5 years), the aggressive device designs, and the ever-present consumer price
and quality expectations. In particular, these requirements present a significant
issue for the antenna performance in mobile devices. The use of antenna RF (radio
frequency) tuning is the disruptive technology that is being used to respond to the
antenna challenge.
The history of the development of BST (barium strontium titanate) started with
the discovery of ferroelectricity in barium titanate in the 1940s. In 1988, Paratek
Microwave Inc., a private start-up company, was founded to exploit the unique prop-
erties of BST in tunable devices. The basic BST tuning and materials technology
was funded under a DARPA contract for the FAME (frequency agile materials for
electronics) program and the vision of Paratek was to commercialize the technology
for military and commercial applications. Since 2006, the commercial focus of the
company was to implement tunable and adaptive antenna impedance matching in
mobile devices. In 2011, Paratek Microwave introduced their first generation of BST-
based tuning solutions into the mass market, and since then, RF tuning using vari-
ous technology options has become a mainstream solution in modern smartphones.
The mass production and market support for BST-based tunable solutions has been
led by ON Semiconductor and STMicroelectronics—two licensees of Paratek’s IP
for BST devices. Paratek Microwave was acquired by BlackBerry® in March 2012.
Fundamental capacitor and materials work has been pursued at many universities
around the world including significant efforts in the United States at Pennsylvania
State University, North Carolina State University, and the University of California
at Santa Barbara.
This chapter will detail the device characteristics of the BST-based tuning
technology and some specific attributes and considerations in designing tuning solu-
tions to be used in the tunable and adaptive RF front end of a modern mobile device.
5
Capacitance (pF)
0
–25 –20 –15 –10 –5 0 5 10 15 20 25
VBIAS (V)
The dependence of the capacitance on the bias voltage of a typical BST tunable
capacitor is illustrated in Figure 3.1, where the capacitance in picofarads is on the
vertical axis and applied voltage in volts is on the horizontal axis.
At zero applied voltage, the capacitance is 5.7 pF but at 22 V, in either polarity,
the capacitance drops to 1.59 pF. Four to one changes in capacitance can be easily
observed in such capacitors. So rather than fixed increments of capacitance change,
the BST capacitor offers a continuous set of capacitance values, depending on the
applied voltage, which lie between a maximum at 0 V and a minimum at some maxi-
mum applied voltage, which is usually determined by the breakdown voltage of the
capacitor or the long term reliability of the device.
FIGURE 3.2 BaxSr(1 − x)TiO3 unit cell in paraelectric state above Curie temperature.
The Curie temperature, Tc, of BST can be reduced from about 120°C for BaTiO3
to about −45°C for a 50:50 mixture of BTO and STO, Ba0.5Sr0.5TiO3. The relative
ratio of BTO and STO also affects the tuning and loss characteristics of the capaci-
tors. At high barium compositions, the tuning effects are maximized but the micro-
wave loss of the material is high. At low barium levels, the loss is good but the tuning
ratio is low. For most practical applications, the barium composition is between 35%
and 55%, resulting in low-loss and good tuning while operating above the Curie tem-
perature in the paraelectric state. A proprietary BST material composition known as
ParaScan™, developed by Paratek Microwave, further enhances the performance of
BST by adding small quantities of additional elements (dopants) to further lower the
microwave loss and leakage currents.
3.2.2 Capacitor Structure
In its simplest form, the tunable capacitor is no more than a parallel plate capacitor
using ParaScan or BST as the dielectric between the two plates as shown in Figure 3.3.
A DC potential is applied between the top and bottom electrodes to establish the elec-
tric field in the ParaScan and therefore, set the capacitance value as seen in Figure 3.1.
Top electrode
ParaScan
dielectric
Bottom electrode
Substrate
The thickness of the ParaScan layer sets the required tuning voltage as well as
determines the capacitance per unit area (pF/mm 2) which drives the physical size
of the capacitor.
The choice of the barium composition of the tunable material, the substrate mate-
rial, the capacitor electrodes, and the processing used to make the capacitor largely
determine the tuning ratio and leakage current of the capacitor. Current ParaScan
tunable capacitors achieve 4:1 tuning ratios at applied bias voltages from 2 to 20 V.
The leakage currents at 20 V are less than 1 nA. The conductivity and thickness of
the metal layers determine the Q at the microwave frequencies. Of course in real
life, there are additional design elements to consider such as a bias network, inter-
connect metals, dielectric passivation, and the need to address linearity and power
handling for applications such as mobile handsets. These complicate the simple
capacitor layout of Figure 3.3 and lead to an integrated circuit comprising multiple
tunable capacitors and resistors on each die, which is referred to as a passive tun-
able integrated circuit (PTIC). For clarity, we will use tunable capacitor to refer to
a single parallel plate capacitor structure and PTIC when discussing an integrated
circuit with multiple capacitors. In addition, the term stack will be used to describe
the number of capacitors connected in series in a PTIC.
TABLE 3.1
Typical Characteristics of a 4.7 pF PTIC
Characteristic (25°C) Typical Value Units
Operating bias voltage range 2–20 V volts
Capacitance, 2 V 4.7 pF
Capacitance, 20 V 1.175 pF
Tuning range 4:1
Leakage current, 20 V <100 nA
Frequency range 500–2700 MHz
52 Tunable RF Components and Circuits: Applications in Mobile Handsets
maximum capacitance, in this case 4.7 pF, and 20 V for the minimum, or 1.175 pF.
The tuning ratio of the capacitor is, therefore, C(2)/C(20) or 4:1. Higher tuning ratios
can be achieved at higher bias voltages but careful attention must be paid to the reli-
ability of these devices. The leakage currently associated with the bias voltage at its
maximum of 20 V is less than 100 nA and usually less than 10 nA. The power draw
of the capacitor at the maximum leakage is only 0.1 µW. These devices cover fre-
quencies from 500 to 2700 MHz and are specified over the 700 to 2700 MHz range
typical of a modern mobile handset.
To meet the linearity and voltage handling requirements, the PTIC uses many
capacitors connected in series on a single die hence the designation of this as an
integrated circuit. A typical PTIC is shown in Figure 3.4.
This PTIC integrates 24 series connected capacitors, each of which is tunable
to create the desired two terminal capacitance at the RFIN and RFOUT ports of the
IC. The DC bias voltage is applied to the terminal marked VBIAS and the RFIN and
RFOUT terminals must be at DC ground. The bias voltage is distributed inside the IC
through a network of 24 resistors supplying the DC bias to each capacitor while at
the same time providing isolation of the DC and RF signals. The resistor values must
be large enough to avoid the degradation of the Q due to resistive losses while stay-
ing small enough to permit fast charging and discharging of the capacitors as the DC
tuning voltage is changed. In practice, the bias networks limit the Q of the PTIC up
to 500 MHz, below the desired operating range, yet still support capacitor transition
times of 100 microseconds and less.
While the use of multiple capacitors in series improves the voltage handling and
linearity as will be discussed further, the total capacitance on the PTIC die is much
greater than the net series capacitance in the RF circuit. If there are N individual
capacitors in series to produce a net capacitance of CNET between the RFIN and RFOUT
ports of Figure 3.4, each capacitor must be N × CNET and the total capacitance on
the die is now N2 ×CNET. For example, a typical 4.7 pF PTIC with 24 individual
VBIAS
P
Resistors
C
a
p
a
c
RFIN i RFOUT
t
o
r
s
Resistors
FIGURE 3.4 Photo of a PTIC with 24 series-connected capacitors and bias resistors.
BST Technology for Mobile Applications 53
capacitors in series must use a string of 112.8 pF capacitors for a total capacitance on
the die of 2707 pF. Yet, the very high relative dielectric constant of the BST dielec-
tric, between 400 and 450, keeps the size of the capacitors small so that the total die
area including the resistor bias network and pads is below 1 mm2.
3.3.2 Quality Factor
While a tunable circuit can optimize the impedance match of an antenna or amplifier
to get the best response, the loss of the tuning elements may reduce any a dvantage.
Therefore, the loss of the PTIC, expressed as the Q or Quality Factor, is a key factor
and one where the BST-based tunable capacitor technology performs well. At low
frequencies where the bias network loss dominates the PTIC response, the Q is
determined by
Q = ωR pCp (3.1)
where
ω is 2π times the frequency in Hertz
Rp is the parallel equivalent resistance of the bias network in ohms
Cp is the parallel equivalent capacitance in farads
Note that the Q of a parallel RC circuit rises with frequency. At very low frequencies,
the effective resistance of the leakage current of the capacitors could also be included
but the very low leakage of PTICs, below 100 nA at 20 V, makes this a negligible
factor.
The Q at high frequencies, where the series resistance of the capacitor plates and
interconnects dominates is given by
1
Q= (3.2)
ωR sCs
where
Rs is the series equivalent resistance in ohms
Cs is the series equivalent capacitance in farads
TABLE 3.2
Parameters for a Simple Model of a 4.7 pF Tunable Capacitor
Parameter Value Comments
Capacitance 4.7 pF
Parallel resistance 35 kohms
Series resistance 0.1 ohms
Material Q 120 at 1 GHz Frequency dependence is
f−0.12 for frequency in GHz
Tuning ratio 4:1
1000.00
100.00
Composite PTIC Q
700–2700 MHz
operating range
10.00
0.1 Frequency (GHz) 1
FIGURE 3.5 Total Q of the tunable capacitor at zero bias is described by the losses due to
the bias network, the series resistance, and the BST material.
BST Technology for Mobile Applications 55
The Q due to the bias network resistance, labeled bias network, rises with
frequency starting at about 100 at 100 MHz and reaching 1000 at about 1 GHz. The
Q due to the series resistance falls from 1000 at about 340 MHz to a little above 100
at 3 GHz. And the material Q representing losses in the tunable BST material falls
slowly from 158 at 100 MHz to 105 at 3 GHz. This slow drop in the material Q with
frequency is described by
n
⎛f⎞
Qm = Q0 × ⎜ ⎟ (3.3)
⎝ f0 ⎠
Qc_2
40 Qc_2
30
20
10
0
0 500 1000 1500 2000 2500 3000
Frequency (MHz)
FIGURE 3.6 Measured Q at 2 V of five 4.7 pF PTICs, each with 24 tunable capacitors in
series with bias networks.
56 Tunable RF Components and Circuits: Applications in Mobile Handsets
These curves represent the measured responses of five devices across a wafer. The
data is very similar to the curve we generated from a few simple equations in Figure 3.5.
One difference is the use of linear scales now in frequency and Q in Figure 3.6 to
better show the response in the 500–2700 MHz region where this device would
be used. These devices were biased at 2 V for this data, which is the lowest rec-
ommended operating voltage for them. It may be worthwhile to point out that the
devices being used to illustrate the performance of BST-based tunable capacitors are
not test devices but complete production devices with multiple series capacitors, typ-
ically 24, and complete bias networks using a production process designed to create
robust, high-performance PTICs providing reliable performance in mobile phones.
Looking now at the Q variation with frequency when the PTIC is biased to its
lowest capacitance value, the simple Q relationships of Equations 3.1 and 3.2 still
hold but, as we will see, the tunable material now introduces an acoustic response
which must be included in the capacitor design. Looking first at the impact of the
bias network, the series resistance, and the material losses, the bias network losses
become more important than the series resistance losses. Remember that when the
capacitor is biased at high voltage, the capacitance is reduced by about a factor of 4
in these capacitors. So the Q of the bias network, which varies proportionally to the
capacitance, drops by a factor of 4 as shown in Figure 3.7.
Where the bias network Q at 100 MHz was 103, it is now 26. This pulls down
the Q below 1 GHz but the chosen design values maintain a Q above 60 from
500 MHz to 1 GHz. Consider that the impedances in the bias network could be
raised and the Q improved but the response time of the capacitor would then
1000.00
Series resistance
Bias network
100.00
Composite PTIC Q
700–2700 MHz
operating range
10.00
0.1 Frequency (GHz) 1
FIGURE 3.7 Total Q of the tunable capacitor at maximum bias as described by the losses
due to the bias network, the series resistance, and the BST material.
BST Technology for Mobile Applications 57
Qc_16
40 Qc_16
30
20
10
0
0 500 1000 1500 2000 2500 3000
Frequency (MHz)
FIGURE 3.8 Measured Q at 16 V of five 4.7 pF PTICs, each with 24 tunable capacitors in
series plus bias network.
be slowed. Since the Q due to the series resistances is inversely proportional to the
capacitance, the series resistance losses have less effect at these high bias voltages
and the composite Q of the capacitor remains high through 3 GHz.
When we look at the measured Q over frequency of Figure 3.8 for a capacitor
biased at the maximum tuning voltage, we can see that the simple model for Q is
no longer sufficient to describe the Q. Comparing Figures 3.6 and 3.8, the over-
all Q at 100 MHz has dropped from about 60 to about 23 as expected due to the
parallel restive losses of the bias network, and the Q at the higher frequencies is
good except for the appearance of resonances at about 1250 and 1900 MHz, which
cannot be attributed to the causes we have been discussing. So there must be some
new effect.
To understand what is happening to the Q response under bias, let us start again
with the case of no bias or very low bias. When there is no electric field and the tem-
perature is above the Curie temperature, the BaxSr(1 − x)TiO3 unit cell of Figure 3.2
has a titanium atom in the center of a cube, a barium or strontium atom at each of
the eight corners of the cube, and six oxygen atoms at the center of each face of
the cube. This configuration has no dipole moment. But when an electric field is
applied, the position of the atoms shift causing a large dipole moment or polarization
of the crystal, which we exploit as a change in the permittivity of the BST, in other
words, our tuning. But in addition to the change in dielectric constant, the material
becomes electrostrictive, exhibiting a piezoelectric response to an extent determined
by the strength of the applied electric field.6–8 Therefore, the observed dips in Q are
caused by bulk acoustic resonances excited by the piezoelectric response of the BST.
This has practical implications for tunable capacitors as the observed resonances can
destroy the Q at the desired operating frequencies. An acoustoelectric model of the
58 Tunable RF Components and Circuits: Applications in Mobile Handsets
Bottom electrode
Substrate
FIGURE 3.9 Cross section of two-layer capacitor structure used to reduce acoustic excita-
tion of resonances.
capacitor must be used to determine the optimum layer thicknesses and composition
of the materials in the capacitor stack to tune the acoustic response.
To further address the acoustic response, a two-layer capacitor design9 can be
used to reduce the acoustic excitation of the capacitor structure. As shown in the
cross section of Figure 3.9, two capacitors are fabricated, one on top of the other, the
top capacitor from the top electrode to the middle electrode and a bottom capacitor
from the middle electrode to the bottom electrode. When an RF signal is applied
across the two capacitors from top electrode to bottom electrode, the RF signals
on the two capacitors are approximately equal and in phase. By biasing the middle
electrode with the tuning voltage, typically a positive voltage from 2 to 20 V, and
DC grounding the outer two electrodes, the DC electric field in the two capacitors
is inverted relative to the RF signals. That is, in the upper capacitor, the RF signal
flows from the grounded terminal to the positive terminal while in the lower capaci-
tor the RF flows from the positive terminal to the grounded terminal. As a result, the
acoustic excitation of the top capacitor is out of phase and cancels the acoustic excita-
tion of the bottom capacitor. This two-layer configuration is like flipping a switch to
reduce the acoustic coupling coefficient of the tunable material. One more result of
the two-layer structure is the cancellation of the harmonic generation of even-order
harmonics (2, 4, 6,…) as will be seen in our discussion on linearity.
3.3.3 Linearity
Due to the sophisticated modulation used in modern mobile communications systems,
the linearity of any tuning device is very important to the performance of a mobile
phone. This linearity is usually expressed as requirements on third-order intermodu-
lation and harmonic generation. In the PTIC, the use of N multiple capacitors in
series decreases the RF voltage across any individual capacitor by N to improve
the linearity. In fact, a doubling of the number of capacitors in series decreases the
voltage across each capacitor by a factor of 2 and the IP3 drops by 6 dB. For this
reason, even higher stacks of capacitors than the current 24 are being developed.
The third-order intermodulation point is measured using two closely spaced RF sig-
nals, FL and FH, and recording the level of the third-order intermodulation products,
F1 = 2FL – FH and F2 = 2FH – FL. Measuring F1 and F2 in a small signal range for the
BST Technology for Mobile Applications 59
device under test, the intermodulation rises at three times the rate of the fundamental
tones, FL and FH. The intercept point is then found from the extrapolation of the two
curves, one for the fundamental tones and the other for the intermodulation products.
This intercept is the third order intercept point or IP3 for short.
The IP3 of PTICs has been characterized by mounting the PTIC under test in
shunt across a 50 ohm line and also by mounting it in series in the 50 ohm line.
Extensive filtering of the two test signals must be used to ensure that the signals are
clean. Attenuators are used to provide isolation between the two source signals and
prevent reflections from the filters out of band. The measured shunt IP3 of three
4.7 pF PTICs is shown in Figure 3.10. The IP3 in each case is plotted against the
applied DC tuning voltage. The responses of the three PTICs are nearly identical.
The measurement frequencies, FL and FH, were 824 and 846 MHz at power levels up
to 25 dBm. The shunt configured 4.7 pF PTIC exceeds 69 dBm IP3 from 2 to 20 V
and over 75 dBm starting at 4 V.
Any nonlinear device will generate harmonics when excited by an AC signal. The
PTIC design using multiple tunable capacitors connected in series works to reduce
the harmonic generation by decreasing the RF voltage across any one capacitor. The
harmonics are tested by placing the PTIC in shunt with a 50 ohm line and filtering
out the fundamental signal before measuring the harmonics on a spectrum analyzer.
Attenuators are again used to eliminate the effects of mismatches in the test setup.
The measured harmonics of two 4.7 pF PTICs at 824 MHz in shunt are shown in
Figure 3.11.
The second harmonics are −60 dBm and lower over the specified operating range
of 2–20 V, in part because the two-layer capacitor structure suppresses the generation
85
80
75
IP3 (dBm)
70
65
60 4–23 J2
4–34 J2
55
4–35 J1
50
0 5 10 15 20 25
Bias voltage (V)
2nd 5–16, 12 – 2
–40
–50
–60
–70
–80
0 5 10 15 20 25
Bias voltage (V)
FIGURE 3.11 Second and third harmonics of a 4.7 pF PTICs in shunt at 824 MHz.
of all even harmonics. The third harmonic stays at or below −38 dBm from 2 to 20 V
bottoming below −65 dBm at 20 V.
99
90
80
70
60
50
Failure rate (t) (%)
40
30
20
10
3
2
1
10 100 1000
Break voltage
FIGURE 3.12 Breakdown voltage distributions across two wafers of PTICs, 40 devices
per wafer.
The final test for the power handling is conducted using a high-power RF source,
typically a 10 or 20 W TWT amplifier, and a PTIC mounted in series in a 50 ohm
line terminated with a 50 ohm load. Attenuators or isolators placed between the
TWT and the device under test protect the TWT from the high mismatch should
the PTIC fail. With the high breakdown voltage and series connected design of the
PTIC, power levels of 10 W are routinely achieved without failure.
3.4.1 Acceleration Factors
The reliability of ceramic capacitors has been well characterized for electronic
and medical applications.10,11 Under high DC bias voltages and high temperatures,
62 Tunable RF Components and Circuits: Applications in Mobile Handsets
BST capacitors generally fail due to (1) thermal runaway with a gradual increase
in leakage current which eventually shorts the capacitor, (2) an abrupt increase in
current, an avalanche, shorting the capacitor, or (3) a mixture of both failure modes.
The PTICs made with ParaScan dielectric characteristically fail due to an avalanche
short in one of the multiple series capacitors on the chip leading to a high leakage
current, limited by the series resistance of the on-chip bias network. While 1 of the
24 capacitors in the PTIC has failed, the other 23 continue to operate normally and
the device still operates, though with reduced tuning range and higher current draw.
For many applications, this “soft” failure mode provides continued operation of the
tuned product with reduced performance but without a failure of the final product
containing the PTIC.
The failure of a single capacitor in the PTICs is accelerated by both tempera-
ture and voltage factors and is modeled following the work of T. Prokopowicz and
A. Vaskas12 as
where
Ea is the thermal activation energy, eV
T is the temperature, degrees Kelvin
k is Boltzmann’s constant
V is the bias voltage, V
N is the voltage acceleration factor
A is an arbitrary constant
Based on measured data generated over multiple wafer lots at voltages up to 24 V and
temperatures to 150°C, Paratek devices using ParaScan dielectrics typically exhibit
the following acceleration factors.
These two factors allow the PTIC performance over time to be understood and
verified for the intended application.
• Out of each 24 hours, the phone is both operating (PTICs biased) and non-
operating (PTICs unbiased).
BST Technology for Mobile Applications 63
• When the phone is operating, the temperature of the PTICs in the phone is
higher than the circuit board temperature due to heating by other components
surrounding the PTIC components such as the RF power amp. Multiple values
for the temperature rise are based on different operation modes of the phone,
such as hotter for high-power GSM and cooler in less power hungry modes.
Note that the mission profile represents aggressive use of the phone; 4 hours of use
per day for a 30 day month equates to 7200 minutes per month use, more than most
users consume on their phone plan.
1. Measure the MTTF of 50 or more devices, plot the results on a Weibull plot
and extrapolate the resulting curve to the 2-year mission life point. One
problem with this approach is that the first failure out of 50–100 parts tested
lies at only the 1%–2% failure level, but the extrapolation must extend by
two orders of magnitude to the 0.02% point and is subject to any uncer-
tainty in the slope of the failure rate at such low levels.
2. Measure 4500 devices to the 2-year mission life point and show zero fail-
ures to project less than 200 ppm at 60% confidence level. This approach
more directly measures the failures at the 2-year mission life point, but
requires many more devices to be tested and usually covers a small sample
of wafers or wafer lots.
3. Institute a regular sampling of 10–100 devices from every production wafer
to build a significant database of the failure points to allow calculation of
the failure rate at the 2-year mission life point with greater accuracy.
TABLE 3.3
Early Failure Rate Data, 25,760 PTICs
644 Wafers/25,760 2-Year 5-Year 9-Year 30-Year
Devices Life Life Life Life
# Failures @ 2 16 33 129
Failure rate (ppm) 78 621 1281 5008
Failure rate 121 689 1356 5146
(ppm @ 60% CL)
64 Tunable RF Components and Circuits: Applications in Mobile Handsets
0.3
0.1
0.03
0.01
0.1 1 10 100 1000
Lifetime (years)
FIGURE 3.13 Weibull graph of life test of 25,760 PTIC devices to determine failure rate at
2-year handset operation.
As can be seen, there were 2 failures out of 25,760 devices at the 2-year mission
life point or a 78 ppm failure rate. Applying a 60% Confidence Level to the data, the
failure rate is only 121 ppm, nicely below the 200 ppm target. The data is graphed in
the Weibull plot in Figure 3.13.
From this data, we see that not only are the BST-based PTICs better than the
200 ppm failure rate target at the 2-year mission life point, but the extrapolation
of the data to MTTF suggests over 1000 years mean time to failure in the phone
application.
3.4.4 Qualification Tests
While reliable operation with low failure rates is required for PTIC products to be
used in mobile handsets, there are other important qualification tests as well. These
are summarized in Table 3.4. PTIC devices have passed all these qualification tests.
3.5.1 Generic Tuner
A generic tuning circuit is shown in Figure 3.14. This is a Pi circuit using three
PTICs with two in shunt and one in series. A T circuit with two PTICs in series
BST Technology for Mobile Applications 65
TABLE 3.4
Qualification Test Results for PTIC Devices
Qualification Test Qty Per
Test Method Conditions Lot/#Lots
Temp cycle (TC) JEDEC JESD −40°C to +125°C 45/3 PASS
22-A104 1000 cycles (Cond G)
High-temp JEDEC JESD +150°C for 1000 hours 45/1 PASS
storage (HTS) 22-A103C
Low-temp storage JEDEC JESD −40°C for 168 hours 45/1 PASS
(LTS) 22-A119
Drop JEDEC JESD 1500 G Peak 6/3 PASS
22-A119 0.5 ms Pulse
3 Die Level
3 PWB Mounted
Vibration MIL-Std-883 Cond A 20G peak, 6/3 PASS
MTD 2007 20–2000 Hz
Solderability JEDEC JESD Method 1 3/3 PASS
22-B102E
ESD JEDEC JESD Human Body Model 3/3 Class 1A
22A114
L2
C2
C1 C3
Mobile phone Antenna
L1 L3
RF circuits feed
V1 V2 V3
and one PTIC in shunt might equally be used instead. The purpose of the tuner is
to transform the antenna impedance to 50 ohms of the RF circuits in the mobile
phone. The PTICs are C1, C2, and C3 which along with their control voltages V1, V2,
and V3, provide variable capacitance values to alter the impedance transformation
between the mobile phone’s RF circuits on the left and the antenna feed impedance
on the right. The degree to which the impedance transformation can be changed is
related to the tuning range of the PTIC and the number of PTICs in the tuner. While
this example shows three PTICs, many tuners require only two, or in some cases, one
PTIC to achieve the needed response. Likewise, the three inductors shown may not
66 Tunable RF Components and Circuits: Applications in Mobile Handsets
all be required either. However, the two inductors, L1 and L3, on the two RF ports
of the tuner provide the DC path to ground needed to keep the RFIN and RFOUT ports
of the PTICs at 0 V. They also provide protection from ESD damage, particularly at
the antenna feed where the typical phone spec of 6000 V or more greatly exceeds the
capability of the PTIC itself.
In an open-loop tuner, the required values of the bias voltages are determined dur-
ing the design of the tuner for that phone model and that particular antenna and are
stored in memory on the phone. Knowing the operating mode (full or half duplex)
and the frequency of the signal, the proper voltages can be looked up in memory and
applied. This open-loop operation provides a significant advantage for the antenna
match, but it does not correct for the dynamics in the use of a phone such as the shifts
in antenna impedance when the phone is in a pocket, on a table, held in the hand, or
placed at the ear.
Closed-loop operation of the tuner can use a directional coupler to monitor the
incident and reflected power from the antenna directly during use. An algorithm to
perform corrections is incorporated in the software and the phone shifts the tuner
voltages to continually optimize the tuner response. This method provides the best
performance at the expense of a very small additional processor load.
The operation of PTICs in practice requires voltages up to 20 V that are not
readily available in a mobile phone. For this reason, the manufacturers of the PTICs
provide several silicon ICs designed to interface the PTICs to both an existing volt-
age source in the phone and to the digital signals needed to allow the phone to set the
PTIC voltages, whether in open-loop or closed-loop operation. These are referred to
as HVDACs, short for high-voltage digital to analog converters.13,14 Running from
the 1.8 V supply in the phone, the HVDAC generates a 22 V or higher internal volt-
age. Available with either an SPI interface or a MIPI RFFE interface and from two
to six outputs to drive up to six PTICs, the HVDAC uses an individual DAC for each
output to convert the digital input to the desired output voltage for each PTIC.
SEINE_JPN_S
2011. 01. 1
RAFT tuning module
FIGURE 3.15 Tuning module including PTIC elements used in Samsung Galaxy S II
mobile phone released in April 2011.
FIGURE 3.16 Discrete tuning solution used in HTC One models with PTICs in plastic QFN.
WLCSP PTICs
FIGURE 3.17 Antenna tuning section of BlackBerry Z10 with discrete wafer level chip
scale packaged PTICs.
(WLCSP) for the PTIC, Figure 3.17, which created the smallest and lowest cost tuner
implementation. In March 2013, the Z10 was introduced to the market, followed by
the Q10 and Q5 phones.
In September 2013, BlackBerry continued to drive the Paratek BST technology
with one of the first ever, closed-loop implementations of adaptive antenna tuning
technology in their Z30 model. This was also the first implementation where both the
main and diversity antennas in a single phone were tuned—a trend that is continu-
ing to grow. In their press release of September 18, 2013, BlackBerry specifically
referred to the antenna tuning as a key feature:
68 Tunable RF Components and Circuits: Applications in Mobile Handsets
It was a big year in 2014 for the introduction of PTIC-based antenna tuning. A series
of China-based OEMs started using the Paratek BST tuning technology—ZTE in
their Nubia 5S and Huawei in their Asend P7 smartphones. In September 2014,
BlackBerry announced its second phone using closed-loop tuning technology—the
Passport. Also in September 2014, Motorola introduced the first of three phones
with Paratek BST-based tuning technology—the Moto-X phone which was followed
by the Droid Turbo and the Nexus 6. The introduction of these phones brought atten-
tion specifically to the improved antenna performance with what Motorola referred
to as “dynamic tuning.”
TABLE 3.5
Improvements to PTIC Performance on a Near-Term Roadmap
Current Future
Parameter Performance Performance
Tuning ratio 4:1 5:1
Quality factor (Q) 60 90 500–1000 MHz
50 70 1700–2700 MHz
Linearity, IP3 >68 dBm >75 dBm Over entire tuning
range
improvement may also come from extension of the maximum operating voltage but
staying within the voltage range of the existing HVDAC integrated circuits designed
to provide the tuning voltage for PTICs. The increased tuning range will extend the
area of impedances which can be tuned by matching circuits employing the PTICs
and may, in some cases, allow a decrease in the number of PTICs needed for a spe-
cific application, say from three PTICs to two or from two PTICs to one.
The higher Q of these next-generation PTICs will lower the loss of tuning net-
works, particularly where a high standing wave ratio must be matched and the imped-
ance transformations are great. These improvements rely on a lower loss ParaScan
formulation for BST to help across the entire frequency range coupled with lower
bias network losses and good acoustic design.
Finally, higher linearity means an increase in the stacking of capacitors in series
beyond the typical 24 series connected capacitors as used now. As discussed ear-
lier, for a PTIC of N series connected capacitors, the total capacitance on the die
is N2 times the desired net series capacitance. For a 4.7 pF PTIC with 24 capaci-
tors stacked in series, the total capacitance on the die is therefore 2707 pF. If the
stack were increased to 48, the IP3 will be 6 dB lower, but the die will now have
four times the total capacitance or 10,829 pF! Designs with increased stacking have
been shown to improve the IP3 and are now shipping in high volume, production
handsets.
3.7 CONCLUSIONS
Tunable capacitors based on BST material have proven themselves in the production
of antenna tuning solutions used in millions of mobile phones worldwide. BST tun-
able capacitors provide a change in capacitance or tuning ratio of 4:1 combined with
low losses at microwave frequencies, high linearity, and up to 10 W power handling.
The devices are produced as integrated circuits on 6 in. wafers using many of the
process techniques used in the silicon IC industry and are available from multiple
sources. As the need for better broadband antennas covering wider frequency ranges
is driven by the design of attractive phones in small form factors, the use of antenna
tuning will continue to increase and the PTICs of today and tomorrow will be there
to meet the challenges.
70 Tunable RF Components and Circuits: Applications in Mobile Handsets
ACKNOWLEDGMENTS
The development of ParaScan, the improved BST dielectric, the design of a robust
tunable capacitor, and the qualification and production of PTICs are the result of the
efforts of many people at Paratek Microwave, now part of BlackBerry Inc., along
with ON Semiconductor in Burlington, Ontario, Canada, and STMicroelectronics
in Tours, France. To all those involved, many thanks for the hard work bringing this
technology to market. The contributions of Dr. James V. DiLorenzo, Mr. Ralph Pini
and Dr. Louise Sengupta were critical to the successful development of PTIC tech-
nology and its application in commercial mobile phones.
The authors gratefully acknowledge the support of the BlackBerry Technology
Services division of BlackBerry Inc. and ON Semiconductor Corporation in the
publication of this information on BST technology and products. Figures 3.1, 3.4,
3.6, 3.8, and 3.10 through 3.13 as well as the data in Tables 3.3 and 3.4 are used
with the permission of BlackBerry Inc. The data of Table 3.1 on a typical 4.7 pF
capacitor are a combination of data from the data sheets of ON Semiconductor and
STMicroelectronics.
REFERENCES
1. Tagantsev, A.K., Sherman, V.O., Astafiev, K.F., Venkatesh, J., and Setter, N.,
Ferroelectric materials for microwave tunable applications, Journal of Electroceramics,
11, 5–66, 2003.
2. Newnham, R.E. and Cross, L.E., Ferroelectricity: The foundation of a field from form
to function, MRS Bulletin, 30, 845–848, November 2005.
3. Remmel, T., Gregory, R., and Baumert, B., Characterization of barium strontium
titanate films using XRD, JCPDS—International Centre for Diffraction Data, 38–45,
1999.
4. ON Semiconductor Datasheet, TCP-3147H Rev 0, 4.7 pF passive tunable integrated
circuits (PTIC), http://onsemi.com, accessed September, 2014.
5. STMicroelectronics Datasheet, STPTIC, ParaScan™ tunable integrated capacitor,
DOCID023772 Rev 3, http://www.st.com, accessed January 2014.
6. Gevorgian, S. and Vorobiev, A., DC field and temperature dependent acoustic resonances
in parallel plate capacitors based on SrTiO3 and Ba0.25Sr0.75TiO3 films: Experiment and
modeling, Journal of Applied Physics, 99(124112), 1–11, 2006.
7. Vendik, O.G. and Rogachev, A.N., Electrostriction mechanism of microwave losses in
a ferroelectric film and experimental confirmation, Technical Physics Letters, 25(9),
702–704, 1999.
8. Tappe, S., Bottger, U., and Waser, R., Electrostrictive resonances in (Ba0.7Sr0.3)TiO3 thin
films at microwave frequencies, Applied Physics Letters, 85(4), 624–626, 2004.
9. Oakes, J., Martin, J., Kozyrev, A., and Prudan, A., Capacitors adapted for acoustic
resonance cancellation, U.S. Patent 7,936,553 (May 3, 2011).
10. Rawal, B.S. and Chan, N.H., Conduction and failure mechanisms in barium titanate
based ceramics under highly accelerated conditions, Technical Information Publication,
AVX Corporation, Myrtle Beach, SC, S-CFMB00M301-R, 1984.
11. Ashburn, T. and Skamser, D., Highly accelerated testing of capacitors for medical
applications, SMTA Medical Electronics Symposium, Anaheim, CA, 2008.
BST Technology for Mobile Applications 71
12. Prokopowicz, T. and Vaskas, A., Research and development, intrinsic reliability,
subminiature ceramic capacitors, Final report ECOM-90705-F, NTIS AD-864068,
1969.
13. ON Semiconductor Datasheet, TCC-106 Rev 4, Six Output PTIC Control IC, http://
onsemi.com, accessed June 2014.
14. STMicroelectronics Datasheet, STHVDAC-304M, High Voltage BST capacitance
controller, Doc ID 023054 Rev 2, http://www.st.com, accessed November 2012.
4 Tuned Antennas for
Embedded Applications
Frank Caimi
CONTENTS
4.1 Introduction and Overview.............................................................................. 74
4.2 Radiation Theory: Basics, Loss Mechanisms, Bandwidth, and Radiation
Mechanisms..................................................................................................... 74
4.2.1 Design Principles Affecting Small Antenna Design........................... 78
4.2.1.1 Conductor Area..................................................................... 78
4.2.1.2 Radiation Resistance............................................................. 78
4.3 Antenna Requirements: Current and Future—Band Coverage,
Radiation Efficiency, Coupling, Correlation, and Architectural Variants......... 79
4.3.1 Factors Driving Antenna Design.........................................................80
4.3.1.1 Reduction in Size..................................................................80
4.3.1.2 Increased Radiating Power...................................................80
4.3.1.3 Decreased Battery Consumption.......................................... 81
4.3.1.4 Higher Gain.......................................................................... 81
4.3.2 The Evaluation of Antenna Performance: Design Issues.................... 81
4.3.2.1 Traditional Requirements..................................................... 82
4.4 Basic Antenna Types: IFA, PIFA, Monopole, Ring, and Slot...............................87
4.4.1 Monopole............................................................................................. 87
4.4.2 IFA/PIFA............................................................................................. 87
4.4.3 Ring..................................................................................................... 87
4.4.4 Slot....................................................................................................... 88
4.4.5 Loop Antennas.................................................................................... 88
4.4.6 Adaptive Antennas.............................................................................. 89
4.5 Tuning Methodology and System Concepts: Aperture-Based,
Match-Based, Open Loop, Closed Loop......................................................... 89
4.5.1 Open-Loop Tuning.............................................................................. 91
4.5.2 Closed-Loop Tuning............................................................................94
4.6 Tuning Component Requirements: Loss Mechanisms, Parasitics, and
Nonlinearity..................................................................................................... 95
4.6.1 Loss Mechanisms................................................................................ 95
4.6.2 Parasitics.............................................................................................. 95
73
74 Tunable RF Components and Circuits: Applications in Mobile Handsets
TABLE 4.1
LTE Frequency Bands and the Corresponding Regions
Duplex
Downlink Spacing BW Duplex Deployment in
LTE Bands Uplink (MHz) (MHz) (MHz) (MHz) Mode the World
Band 1 1920–1980 2110–2170 190 60 FDD China, Japan,
EU, Asia,
Australia
Band 2 1850–1910 1930–1990 80 60 FDD North and South
America
Band 3 1710–1785 1805–1880 95 75 FDD EU, China,
Asia, Australia,
Africa
Band 4 1710–1755 2110–2155 400 45 FDD North and South
America
Band 5 824–849 869–894 45 25 FDD North and South
America,
Australia, Asia,
Africa
Band 6 830–840 875–885 45 10 FDD Japan
Band 7 2500–2570 2620–2690 120 70 FDD EU, South
America,
Asia, Africa,
Australia
Band 8 880–915 925–960 45 35 FDD EU, South
America,
Asia, Africa,
Australia
Band 9 1749.9–1784.9 1844.9–1879.9 95 35 FDD Japan
Band 10 1710–1770 2110–2170 400 60 FDD North and South
America
Band 11 1427.9–1447.9 1475.9–1495.9 48 35 FDD Japan
Band 12 698–716 728–746 30 18 FDD North America
Band 13 777–787 746–756 31 10 FDD North America
Band 14 788–798 758–768 30 10 FDD North America
Band 17 704–716 734–746 30 12 FDD North America
Band 18 815–830 860–875 45 15 FDD North and South
America,
Australia, Asia,
Africa
Band 19 830–845 875–890 45 15 FDD Africa
Band 20 832–862 791–821 41 30 FDD EU
Band 21 1447.9–1462.9 1495.9–1510.9 48 15 FDD Japan
Band 22 3410–3500 3510–3600 100 90 FDD
(Continued)
76 Tunable RF Components and Circuits: Applications in Mobile Handsets
In 1948, Chu extended Wheeler’s analysis and expressed the fields for an
omnidirectional antenna in terms of spherical wave functions and found limits for
the antenna Quality factor Q, the maximum gain Gm and the ratio G/Q. Later in 1959,
Harrington related the effects of antenna size, Quality factor, and gain for the near
and far field diffraction zones for linearly and circularly polarized waves, and also
treated the case where the antenna efficiency is less than 100%.
Additional work carried out by others4–6 from 1969 through 2001 was directed
to obtain exact expressions for the antenna Q over an expanded size range. These
efforts led to the equation:
1 1
Q= + (4.2)
(ka) ka
3
where
k is 2π/λ
a is the radius of a sphere containing the currents associated with the fundamental
or lowest mode, generally confirming the results obtained by Chu, Harrington,
Collin and Rothschild, and others
(reactive elements such as inductors and capacitors store energy). Antenna Q can be
loosely considered as the ratio of the reactive term of the impedance to the resistive
or dissipative portion.
⎛ reactance ⎞ f
Q∼⎜ ⎟∼ (4.3)
⎝ resistance ⎠ Δf
Chu-limit
1000
(Fractional bandwidth)–1
100
GPS
ISM
PCS'3G
cellular
10
U-NII
FM
UHF
1
Chu-limit Goubau
Narrowband MLA Planar inverted
"F" antenna
Dipolo
Wideband MLA
Inverted "F" antenna Patch
FIGURE 4.1 Chu–Harrington curve and antenna example Q relative to the normalized volume.
78 Tunable RF Components and Circuits: Applications in Mobile Handsets
wave radiation into space and the loss due to heating in the conductors making up
the antenna, the so-called “I-squared R” loss. To achieve efficient radiation of energy
into space with minimum heating of the antenna elements, the resistive loss in the
antenna conductors must be minimal with respect to the radiation resistance (loss).
It turns out that both losses can be controlled somewhat through engineering design.
As a result, poorly designed antennas will exhibit higher radiation than others for the
same input power. A measure of the design quality is the radiation efficiency. Small
antennas can have efficiencies that range from small fractions of 1% for submarine
antennas to near 100% for others.
2
⎛ b2 ⎞
Rradloop = 300,000 ⎜ 2 ⎟ (4.5)
⎝λ ⎠
do more functions while consuming less power, and all device elements need to be
more compact in size. But one element that has received less attention throughout
the wireless revolution has been the very element that makes a device wireless—the
antenna. Until recently, antennas have been incorporated into wireless devices with
little modification to the traditional monopole and loop designs that have been used
for decades. Today, however, as companies compete for ever-smaller devices with
increased functionality, the search for new, more efficient antenna configurations has
once again become a factor in wireless designs.
keep a phone that continually drops calls, no matter how stylish it looks. So design-
ers are tasked with finding a way to overcome the detuning effects of the casing
and of the user’s body in order to maintain acceptable signal strength. Adaptive
tuning of the antenna is one way that is being developed to mitigate these effects.
⎛P ⎞
RL = 10Log ⎜ incident ⎟ (4.6)
⎝ Prefelected ⎠
⎡ (VSWR −1) ⎤
RL = −20Log ⎢ ⎥ (4.7)
⎣ (VSWR +1) ⎦
For handset antennas in the past, RLs of −5 dB (VSWR ~ 3.5:1) has been accept-
able, but with the advent of cost effective tunable components for use in the match-
ing network much better RL can be feasibly obtained dynamically at the operating
frequency.
For example, transmitted power at VSWR 5:1 is about 55%, while at 2:1 is 89%.
constructed with multiple distinctive elements: one for each band at a maximum.
Antenna designers are then faced with the complex task of correctly forming the
size and spacing of the elements and matching their resonant impedance (voltage-
to-current ratio). One of the biggest problems encountered in this process is that
each element acts as a parasitic load on each of the others and, therefore, the tun-
ing of one element has a noticeable effect on the others. Once the elements are
properly constructed, the antenna sends and receives RF signals on different fre-
quencies each corresponding to the resonance of each of the elements. Assuming
that the antenna does not need to operate on all bands simultaneously, a scheme
using adaptive tuning of the front-end electronics to the antenna impedance may be
employed to mitigate the antenna design problem of impedance matching over all
bands simultaneously.
4.3.2.1.3 Bandwidth
Bandwidth is the amount of electromagnetic spectrum needed or allocated for a par-
ticular communications channel or group of channels. It is usually defined in units of
frequency and is computed as the difference between an upper and lower band edge
limit. For instance, Band 2 (Table 4.1) is defined as 1850–1990 MHz, and, therefore,
has an allocated bandwidth of 140 MHz. Individual channel bandwidths, however,
are much narrower requiring only 10’s of kilohertz bandwidth.
Most antennas used in handheld devices are multiband—they only operate on
the band of frequencies for which the device was intended. Broadband antennas
tend to perform less effectively than narrowband antennas and typically cost more to
produce. However, the advantages offered by broadband antennas (covering multiple
frequency bands, ability to transmit more data faster, etc.) can make them attrac-
tive for wireless device makers. The key is finding a happy medium between per-
formance and bandwidth coverage, and in some cases depending on the front-end
electronics architecture, multiple feedpoints are possible thereby splitting operation
between the low bands and the higher bands. Such architectures using more than one
antenna can be effective in reducing losses in both the front end and the antenna,
but at the expense of extra feed lines and potential assembly cost. Increased antenna
space or volume can also be a factor.
The increase in the demands now expected of antennas greatly increases the
importance of two characteristics—efficiency and size.
4π
TIS = 2π π (4.8)
∫ ∫
0 0
[(1 / (EISθ (θ1,φ1 )))+ (1 / (EISφ (θ1,φ1 )))]sinθ1 dθ1 dφ1
UE must support at least downlink capability with two receiving antennas of similar
gain and efficiency. Future plans as indicated by 3GPP and some wireless service
providers suggest that as many as four antennas will be required in wireless handsets
and tablet devices. This imposes new requirements for antenna efficiency, coupling,
and signal correlation between antennas.
2π π
ρp =
∫ ∫
0 0
A12 dΩ
2π π 2π π
∫ ∫
0 0
A11 dΩ ∫ ∫
0 0
A22 dΩ
The subscript indices (m,n) refer to either antenna port 1 or 2, depending on their
integer value, Ω(θ,ϕ) is the spatial angle in steradian as commonly depicted, and
E1,v, E1,h , E 2,v, and E 2,h are the complex envelopes of either the vertical or horizon-
tal field components resulting from excitation of either antenna port 1 or 2, respec-
tively. The cross polarization ratio is defined as the ratio of the mean received
86 Tunable RF Components and Circuits: Applications in Mobile Handsets
power in the v ertical polarization to the mean received power in the horizontal
polarization.
Since the signals may arrive from different directions, the antennas should exhibit
characteristics that allow for highest independence of the received signals. A prob-
ability function P may also be used to weight the computation to match the angular
arrival distribution expected, for the expected scattering functions associated with
the propagation environment. The method involves the use of a simple environmen-
tal model that has been experimentally verified.10 In this model, the signal power is
described as a function of angle of arrival by probability functions Pv and Ph, which
can be included in Equation 4.8 as follows:
Difficulty achieving low values of ECC (ρ² < 0.5) depends on factors such as antenna
separation, polarization, and currents induced on the ground structure. In the latter
case, the structure can be analyzed to determine its modal behavior, and how each
antenna placement excites or contributes to existing in a particular mode.11
4.3.2.1.6.4 Antenna Isolation and Correlation Antennas should not only exhibit
low port-to-port signal correlation, but should also exhibit isolation from one another
so that signals being broadcast or received from one feed port don’t appear at the oppo-
site port or feed. The isolation is typically measured at the antenna terminals by using
the commonly derived S-parameters. Isolation figures are typically given in d ecibels
and are designated as S12 or S21 depending on whether the measurement is from antenna
1 to 2 or vice versa. If the isolation is poor, the source antenna will deliver substantial
power into the adjacent antenna’s termination impedance (typically 50 ohms-resistive),
with a corresponding reduction of the overall radiation efficiency. S21 values of −10 dB
are regularly deemed acceptable, resulting in power loss of less than 10%.
Since the antenna near-field coupling is related to the far-field pattern, it is possi-
ble to approximately express the correlation coefficient in terms of the S-parameters
measured at each antenna terminal. This is given by12
*
S11S12 + S*21S22
ρs = (4.11)
(1− (|S11 |2 + |S21 |2 )1/2 (1− (|S22 |2 + |S12 |2 )1/2
correlation coefficient depends primarily on the phase difference between the two
antenna patterns—indicating that good correlation performance requires careful
consideration of the antenna placement and design in the case where multiple single
antennas are used in close proximity.
Briefly then, the antenna system, its associated pattern, and the environment are
all clearly important in establishing independence between the signals received at
each antenna and, therefore, the overall system data transfer rate.
4.4.1 Monopole
Monopole antennas rely on a single radiator and a corresponding counterpoise as
shown in Figure 4.2a. In the case of a typical wireless handset or other wireless
device, the ground structure and printed circuit board (PCB) ground plane estab-
lish the counterpoise, while the radiator may be a printed conductor on a variety of
substrates, including the case. The radiator may be “loaded” with lumped or con-
tinuous elements that exhibit an electrical reactance that cancels an opposite equiva-
lent reactance associated with the combined structure at the frequency of interest.
The monopole exhibits a theoretical resistance that is one half that of a dipole (see
Equation 4.4) at the same design frequency and can be perpendicular or parallel to
the PCB ground plane—the latter is called a transmission line monopole.
4.4.2 IFA/PIFA
The inverted F or IFA antenna is shown in Figure 4.2b and is a variant of the trans-
mission line monopole and includes a termination to the counterpoise structure. The
position of this termination relative to the feedpoint allows matching of the antenna
impedance at resonance to the feedline.
When the antenna width is increased (Figure 4.2c) to form a planar structure, the
antenna is termed a planar inverted F or PIFA.
Either antenna type shows high concentration of current in the feed to ground
point and, therefore, high near field RF intensity.
4.4.3 Ring
The ring antenna is a recently used antenna type that surrounds the ground plane
associated with the wireless device. It has become popular as it interferes minimally
88 Tunable RF Components and Circuits: Applications in Mobile Handsets
+ +–
–
+
–
–+
+
–
(e) (f)
FIGURE 4.2 Examples of typical antenna types: (a) Monopole, (b) dipole, (c) slot, (d) loop,
(e) PIFA, and (e) ring.
4.4.4 Slot
The slot antennas can be considered an analog of dipole antennas, having mag-
netic and electric fields interchanged. Typically, the slot is formed between two
conductive planes and is a finite length that determines the fundamental resonant
frequency. An electric field is established from one side of the slot to the opposite
side and shows decreasing magnitude from the feed to the terminus. Slot antennas
are used in predominantly metal structures and may be fed with a probe near the
centerline.
4.4.5 Loop Antennas
Loop antennas that take a variety of shapes are fed at a break in the conductive
loop structure. Another property of the loop is the ability to achieve multimode
operation in one antenna structure. Square loops can be used in two radiation
modes that can be excited simultaneously or that can be individually selected
electrically. The fundamental mode is that of a monopole producing a linear
polarized radiation field resembling that of two closely spaced monopole anten-
nas. In the monopole mode, the primary antenna current is perpendicular to the
Tuned Antennas for Embedded Applications 89
antenna ground plane. A second mode of operation is the “loop mode” where the
primary antenna current is parallel to the antenna ground plane and the maximum
radiation is perpendicular to the ground plane.
4.4.6 Adaptive Antennas
Also commonly known as “smart” antennas, these are antennas that focus their RF
energy from the mobile handset to the base station or that direct a radiation pattern
null toward an interferer.
Adaptive antenna arrays have two great advantages: (1) they allow more spectrum
to be reused because omnidirectional signals are not overlapping and (2) they can
send a strong directed signal using less energy because energy is not wastefully radi-
ated in directions it is not needed. In addition, the mobile device will also require
less energy and conserve battery power in some systems.
1
Z in = + sL + R
sC
As can be seen in Figure 4.3b the actual behavior of an antenna is only approximated
in this example as from Equation 4.4, the resistive component should increase with
frequency as shown. Nevertheless, the circuit is illustrative of the need for adaptive
tuning if operation is desired at even 10% above or below resonance (shown here as a
normalized frequency of 5). Also note that an inserted series capacitance or switched
circuit inductance would change the resonance frequency but that the real part of
90 Tunable RF Components and Circuits: Applications in Mobile Handsets
1000
Actual
Rin
Rin
0
–1000
L
C –2000 Xin
Input
R
–3000
0 5 10
(a) (b)
L
Feedpoint C
R
Aperture-based tuning
(c)
FIGURE 4.3 (a) Equivalent circuit example of a dipole at fundamental resonance. (b) Series
circuit model versus actual real (Rin) and imaginary (Xin) components of antenna imped-
ance. (c) Possible tuning methodology, for example, of Figure 4.5a.
the input impedance would still be mismatched as also can be seen from Equation
4.4 and Figure 4.3b. The need for match-based tuning compensating for the change
in the resistive component of Zin might also be called for in some applications. So
from this simple example, it can be seen that a series variable capacitance within the
antenna could result in raising the resonant frequency and that a series selectable
inductance would have the opposite effect (Figure 4.3c). Note also that either com-
ponent could be considered as part of a matching network external to the antenna
(outside the black outline).
Tuned Antennas for Embedded Applications 91
The previous example is illustrative, but real world antennas are complex multi-
resonant devices. As an example, the circuit model shown in Figure 4.4a is exem-
plary of a broadband multi-resonant antenna developed by Foster.15 The broadband
model is reasonably accurate over the three resonant frequencies shown.
Such models can be used to develop a tuning methodology based on basic
circuit design principles. The model parameters can be matched to a particular
antenna design and/or a new model may need to be developed that includes a
lumped model tuning elements if available. Once a candidate tuning architecture
is established, a physical model of the antenna and tuning components can be
simulated using one of the many 3D EM simulators available commercially to
refine the design.
4.5.1 Open-Loop Tuning
There are various motivations throughout the industry for using tuned circuits in
the front-end architecture of a wireless device. Expectations among wireless service
providers are that device RF performance would improve as tuning theoretically
has the ability to mitigate impairments due to absorptive hand loading and other
environmental usage conditions. Component suppliers for power amplifiers, anten-
nas, and other components see a potential for reducing cost or design constraints
that result in loss of power efficiency. Original equipment manufacturers or OEMs
see a potential for reduced space usage as tuned antennas can have higher Q. This
is particularly true in the low-frequency region where the phone ground structure
is the primary radiator and the antenna is a transducer that couples RF energy to
the structure. Each of these separate motivations can drive the tuning architecture
toward different realizations.
Figure 4.5 illustrates the case where simple open-loop control of the antenna reso-
nance is desirable to cover four low-frequency bands but with smaller volumetric
sizes than with a broadband antenna alone. Open-loop control requires only infor-
mation from the baseband processor concerning the frequency of operation. To illus-
trate that a volumetric size reduction is possible, two different antenna configurations
were analyzed using a 3D EM modeling program, one being a broadband design and
another a narrower band design capable of tuning over the same frequency range but
using four tuning states. Figure 4.5a illustrates a 50 × 6 × 14 mm seven-band antenna
configuration and its associated radiation efficiency over just the lower three-band
spectral region from 700–960 MHz. A similar, but smaller (50 × 6 × 7 mm) antenna
configuration is shown in Figure 4.5b illustrating, that tuning using only four states
is able to produce nearly the same efficiency, and total frequency coverage as the
larger broadband antenna.
In this example, it is clear that a physical volume reduction of ½ can be achieved
by tuning the antenna to one of several states, each supporting a certain set of fre-
quency bands. The antenna during operation would, therefore, only be required to
change state when the operating band is changed. The time required for this change
must be compatible with other functions within the radio system. A typical require-
ment might be 10–20 microseconds or less.
92 Tunable RF Components and Circuits: Applications in Mobile Handsets
Feedpoint C1 C2 C3
C
L1 L2 L3
R1 R2 R3
(a)
100 90 80
110 70
120 1.00 60
130 50
0.50 2.00
140 40
150 Model 30
160 20
0.20 5.00
170 10
–170 –10
FIGURE 4.4 (a) Foster’s Second Canonical Model with small added losses. (b) Smith chart
plot of actual dipole impedance versus model of Figure 4.3a.
Tuned Antennas for Embedded Applications 93
50
05191001–7low
100
90
80
70
100
Efficiency (%)
60
50
40
30
20
14
10
0
Feed 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
6 Frequency (GHz)
(a)
50 05181007–7low
100
0001
90 0010
1000
80 0100
100 70
Efficiency (%)
60
50
40
30
20
7 10
0
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Feed Frequency (GHz)
6
(b)
FIGURE 4.5 Comparison of (a) multiband antenna and (b) tuned antenna with respect to
size and radiation efficiency over the region 700–960 MHz. (Dimensions in mm.)
This is just one example, and the antenna must still be able to operate over a
much broader bandwidth characteristic of the many air interfaces supported by the
mobile device—typically 700–2600 MHz in a 4G-world phone. It is, therefore,
necessary to tune the antenna so that its instantaneous bandwidth at any time is
sufficient to cover the expected operating frequencies associated with the usage
model. This is particularly true for LTE-Advanced where interband carrier aggre-
gation (CA) requirements are such that the antenna and radio system must operate
simultaneously on multiple bands at the same time. For example, a phone operating
at B17 (state 0001 in Figure 4.8) may also need to operate in B4 (see Table 4.1) for
CA 17–4 according to 3GPP requirement. Although this requirement could poten-
tially negate the use of tuning to separate bands, it is important to note that many
CA band combinations utilize low-band–high-band operation allowing the antenna
94 Tunable RF Components and Circuits: Applications in Mobile Handsets
designer to separate the low- and high-band elements of the antenna. Since achiev-
ing a broad bandwidth low Q antenna is particularly difficult for the low bands
(600–960 MHz), tuning might be applied to only the low-band element while
achieving broader f ractional bandwidth at the higher band frequencies. These are
details associated with a particular implementation, and are generally beyond the
scope and intent of this chapter.
4.5.2 Closed-Loop Tuning
Closed-loop feedback control of the antenna frequency or impedance matching net-
work is used when conditions that may cause detuning or adverse operation are
unknown to the onboard processors. Such control schemes can be either local or
remote; that is, under direct control of the local device or commanded from received
information obtained remotely and sent back to the local device. Motivation for
closed-loop control is primarily to improve system performance and can take many
forms.16 Local control may use onboard sensors to determine device orientation
and/or hand location from which a tuning command signal can be generated from
a lookup table or expert system. It is also possible to monitor the return loss from
the transmit antenna and make adjustments to minimize that signal by tuning the
matching network. Alternatively, it is possible to “sniff” the emitted RF signal
and make adjustments to the antenna resonant frequency or matching network to
maximize the near field signal. Other schemes, some covered by patents, derive
frequency information or bandwidth information directly and make adjustments
accordingly. An e xample of a local closed-loop system used for optimally tuning a
2 × 2 MIMO antenna is illustrated in Figure 4.6. Coarse frequency control is afforded
RF
(Port 2) Aperture-
Tunable matching tuned
network antenna
Control
Sensor
Aperture-
Tunable matching tuned
network antenna
RF
(Port 1)
FIGURE 4.6 Block diagram local closed-loop control of a demonstrator MIMO antenna.
Tuned Antennas for Embedded Applications 95
4.6.1 Loss Mechanisms
There are a variety of loss mechanisms that contribute to both antenna and tuned
component losses.
Resistive near field coupling is due to radiated fields from the antenna circuit
inducing current into nearby components that exhibit resistive dissipation.
Components exhibiting such loss mechanisms may be either passive or
active devices and may, therefore, be either linearly or nonlinearly related
to the field amplitude.
Conducted resistive loss is attributable to I 2 R losses in the circuit carrying the
antenna radio frequency currents. These losses are related to the antenna
conductor geometry, composition, and frequency as well as any series or
parallel resistive losses resulting from tuning components that are in the
front end or tuning circuits. Therefore, the Q of the tuning components can
become a sizable factor in determining the overall benefit or gain of tuned
versus passive antennas.
Nonlinearity loss stems from components that exhibit a higher order depen-
dence on either the voltage or current at their terminals. This then converts
the desired excitation in one spectral region to undesirable frequency con-
tent. As such, it represents some power loss in the desired spectrum but
more importantly can mix down band into adjacent channels. Typical mea-
sures used are second and third harmonic relative to the carrier frequency
as well as IP2 or IP3 measures.17
4.6.2 Parasitics
Parasitics are couplings or parameters that depart from idealized models. Parasitic
capacitance for instance can be particularly troublesome in tuning circuits and may
result from stray capacitance from the semiconductor die to power or ground con-
nections, or from interelectrode capacitance. This often results in a reduction of the
range of tuning such as the Cmax to Cmin ratio. Additional parasitic coupling can result
96 Tunable RF Components and Circuits: Applications in Mobile Handsets
from RF coupling from the antenna conductors to the tuning component die itself.
Depending on the induced field amplitude, the tuning device may exhibit nonlinear
capacitance and may not allow normal performance such as the selection capaci-
tance values at the active terminals. For this reason, tuning circuits must be tested at
full applied power, and may require individual shielding. Also, package selection is
critical for some applications.
From the antenna perspective, parasitic capacitance from the tuning terminals
to the control circuits within the tuning chip can be particularly problematic at
the highest antenna frequencies where even small capacitances on the order of
1 pF or less exhibit low enough reactance to increase loss to intolerable levels. For
instance at 3.5 GHz, 0.5 pF exhibits a reactance of 90 ohms providing a substantial
coupling to other circuit elements when used in a 50 ohm circuit. Designers can
avoid these effects by careful design of the antenna and tuning architecture.
Antenna 2
SMA
connectors
ON Antenna 2
Antenna 1
APC000005RA
SKYCROSS
Antenna 1
TABLE 4.2
Performance Goals for a Dual Antenna System
Parameter Requirement Units Comments
# Antennas 2 Primary, Secondary
Frequency coverage B17, 13, 5, 8, 3, 2, Band
39, 1, 34, 40, 7, 38
Gain, Primary, >−3 primary dBi (average)
High bands >−6 secondary
Gain, Primary, Per curve dBi
Secondary
Isolation <−10 dB
Power (max) > +25 dBm
VSWR <3.5:1 All bands
Correlation <0.7
coefficient <0.5 high bands
Size (ground plane) 50 × 115 millimeters
Antenna volume <1600 mm3 Including ground
keep-out
Control voltage 2.7 Volts Nominal
Control interface Parallel (2-bit) or
serial
98 Tunable RF Components and Circuits: Applications in Mobile Handsets
–2
–3
Efficiency (dB)
–4
–5
–6
–7
–8
–9
–10
Band 17 Band 13 Band 5 Band 8 Band 3 Band 2 Band 39 Band 1 Band 34 Band 40 Band 7 Band 38
704–746 746–787 824–894 880–960 1718– 1850– 1880– 1920– 2010– 2300– 2500– 2570–
1880 1990 1920 2170 2025 2400 2690 2620
(a) Band (MHz)
4.0
VSWR
3.0
2.0
1.0
Band 17 Band 13 Band 5 Band 8 Band 3 Band 2 Band 39 Band 1 Band 34 Band 40 Band 7 Band 38
704–746 746–787 824–894 880–960 1718– 1850– 1880– 1920– 2010– 2300– 2500– 2570–
1880 1990 1920 2170 2025 2400 2690 2620
(b) Band (MHz)
FIGURE 4.8 (a) Measured antenna performance for primary and secondary antennas
shown in Figure 4.9. (b) Mean values of VSWR for each band. (Continued)
Tuned Antennas for Embedded Applications 99
06231108–1–12band2
10
10
9 00
11
8
6
VSWR
0
1.6 1.8 2 2.2 2.4 2.6 2.8
(c) Frequency (GHz)
FIGURE 4.8 (Continued) (c) Measured high-band VSWR versus tuning state (10, 00, 11)
showing minimal impact of low-band tuning on high band.
As can be seen in Figure 4.8c, the low-band tuning has a minimal effect on the
high-band VSWR so that various CA band combinations would be possible (the tun-
ing state is shown as a binary number 10, 00, or 11).
Battery and
regulator
USB dongle
IMAT dual form factor
port
antenna
SMA
test ports
DTCs (2)
Notebook ground
Microcontroller
plane
and SPI bus
FIGURE 4.9 USB dongle dual port aperture-tuned antenna using IMATTM technology.18
Tuned Antennas for Embedded Applications 101
80
70
Efficiency (%)
60
50 Port1
40
Port2
30
20 Spec
10
700 710 720 730 740 750 760 770 780 790 800
Frequency (MHz)
80
Efficiency (%)
70
60
50 Port1
40
30 Primary ant. spec.
20 Secondary ant. spec.
10
820 830 840 850 860 870 880 890 900 910 920 930 940 950 960
Frequency (MHz)
80
70
Efficiency (%)
60
50
40 Port1
30
20
10
1710 1740 1770 1800 1830 1860 1890 1920 1950 1980 2010 2040 2070 2100 2130 2160
(a)
60
50 Port1
40 Port2
30 Primary ant. spec.
20 Secondary ant. spec.
10
2300 2310 2320 2330 2340 2350 2360 2370 2380 2390 2400
Frequency (MHz)
Antenna radiation efficiency
80
Efficiency (%)
70
60 Port1
50
40 Port2
30 Primary ant. spec.
20 Secondary ant. spec.
10
2500 2520 2540 2560 2580 2600 2620 2640 2660 2680 2700
(b) Frequency (MHz)
FIGURE 4.10 (a) Radiation efficiency plots for tuned-match antenna, low and mid bands
and (b) radiation efficiency plots for tuned-match antenna, high bands.
4.8 CONCLUSIONS
Antenna design continues to be challenged by the increasing number of bands,
the requirement for multiple antennas to support MIMO operation, as well as
the thinner form factors so popular with latest model smartphones and other
wireless devices. With further complications arising from intermodulation and
102 Tunable RF Components and Circuits: Applications in Mobile Handsets
REFERENCES
1. Wheeler, H., Small antennas, IEEE A&P, AP-23(4), 462–469, 1975.
2. Chu, L.J., Physical limitations of omni-directional antennas, Journal of Applied
Physics, 19, 1163–1175, 1948.
3. Harrington, R., Effect of antenna size on gain, bandwidth, and efficiency, Journal of
Research of the National Bureau of Standards, Section D: Radio Propagation, 64D(1),
1–12, 1960.
4. Reading, L.J., Designing dual-band internal antennas, Electronic Design News, 99–104,
November 8, 2001.
5. McClean, J.S., A re-examination of the fundamental limits on the radiation Q of elec-
trically small antennas, IEEE A&P, 44, 672–675, May 1966.
6. Grimes, D. and Grimes, C.A., Minimum Q of electrically small antennas: A critical
review, Microwave and Optical Technology Letters, 28(3), 172–177, February 5, 2001.
7. http://www.ctia.org/content/index.cfm/AID/10021.
8. http://www.3gpp.org/IMG/pdf/2009_10_3gpp_IMT.pdf.
9. Parson, S.D., The Mobile Radio Propagation Channel, 2nd edn., Wiley, London, UK,
2000.
10. Breit, G. and Ozaki, E., Phone level radiated test methodologies for multi-mode multi-
band systems, IWPC China Mobile Workshop, IWPC Transactions Handset Antenna
Technologies for MultiMode-Multi-Band, Durham, NC, January 2007.
11. Fabres, M.C., Systematic design of antennas using the theory of characteristic modes,
PhD thesis, Universidad Polytecnia de Valencia, Valencia, Spain, February 2007. http://
anteny.jeziorski.info/wp-content/uploads/2011/10/2007_Fabres.pdfs, accessed March,
2015.
12. Blanch, S., Romeu, J., and Corbella, I., Exact representation of antenna system diversity
performance from input parameter description, Electronics Letters, 39(9), 705–707,
May 1, 2003.
13. Kalliola, K., Sulonen, K., Laitinen, H., Kivekas, O., Krogerus, J., and Vainikainen, P.,
Angular power distribution and mean effective gain of mobile antenna in different prop-
agation environments, IEEE Transactions on Vehicular Technology, 51(5), 823–838,
2002.
14. Long, B. et al., A simple broadband dipole equivalent circuit model, Proceedings of the
IEEE International Symposium on Antennas and Propagation, vol. 2, Salt Lake City,
UT, July 16–21, pp. 1046–1049, 2000.
15. Ramo, S., Whinnery, J.R., and Van Duzer, T., Fields and Waves in Communication
Electronics, Wiley, New York, 1965, Section 11.13.
16. Oh, S.-H. et al., Automatic antenna-tuning unit for software-defined and cognitive
radio, Wireless Communications and Mobile Computing, Wiley Interscience, 2007.
Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/
wcm.484.
17. Kundert, K., Accurate and rapid measurement of IP2 and IP3, Designers Guide
Consulting, May 22, 2002. http://www.designers-guide.org/analysis/intercept-point.pdf,
accessed March 2015.
18. Caimi, F.M. and Montgomery, M., Dual feed, single element antenna for WiMAX
MIMO application, IJAP, 2008, Article ID 219838, April 2008.
5 Tunable and Adaptive
Antenna Systems
Laurent Desclos, Sebastian Rowson,
and Jeff Shamblin
CONTENTS
5.1 Passive Antenna Limitations......................................................................... 103
5.1.1 Common Passive Antenna Approaches in Wireless Devices............ 103
5.1.2 Limitations in Bandwidth.................................................................. 104
5.1.3 Limitations Due to the Changing Environment................................ 105
5.1.4 A Need to Optimize Radiation Patterns to Improve the
Communication Link......................................................................... 105
5.2 Tunable Matching Circuits............................................................................ 106
5.2.1 Adaptive Matching............................................................................ 107
5.3 Band Switching Techniques.......................................................................... 111
5.3.1 Approaches to Alter the Frequency Response of a Radiator............. 111
5.3.1.1 Band Switched Antenna Configuration 1........................... 111
5.3.1.2 Band Switched Antenna Configuration 2........................... 113
5.3.2 Combination of Adaptive Matching and Band Switching to
Optimize an Antenna........................................................................ 115
5.4 Dynamic Radiation Pattern Optimization..................................................... 116
5.4.1 Description of the Modal Antenna Technique.................................. 117
5.5 Conclusions.................................................................................................... 121
Reference................................................................................................................ 122
103
104 Tunable RF Components and Circuits: Applications in Mobile Handsets
advantages and disadvantages that need to be considered as trade-offs are made to bal-
ance frequency bandwidth with performance. For example, monopole elements tend
to couple strongly to the surrounding structure in the host wireless device, which will
reduce isolation between adjacent antennas; on the other hand, the monopole element
will typically possess wider frequency bandwidth compared to loops and PIFAs. PIFAs
are 3D structures where multiple resonances can be generated to allow for multiband
operation, but like monopoles tend to couple strongly to the host device. Loops do not
couple well to the surrounding structure of a host device, making this type element
better suited for embedded applications where isolation is required. Unfortunately, the
loop is a narrow bandwidth element, which is not suited for multiband applications.
More recently, new antenna structures or topologies have been developed to
address the need to maintain isolation between an embedded antenna and its sur-
roundings while maintaining bandwidth and efficiency. An example of a well-isolated
antenna is the isolated magnetic dipole (IMD). This type of antenna is designed
where consideration is made as to the near-field characteristics of the antenna to mini-
mize coupling between the antenna and surroundings, resulting in improved isolation
between antennas in a device, which is needed as more transceivers are integrated into
wireless devices and as MIMO systems gain favor in these communication systems.
5.1.2 Limitations in Bandwidth
Modern wireless devices, such as smartphones, tablets, laptops, and wearable
devices tend to have small form factors and are thin to increase usability (e.g., smart-
phone that fits in a shirt pocket). Designing a lot of functionality into a smartphone
(cellular, WLAN, and GPS receivers and transceivers) highlights the need to reduce
or limit the volume that an antenna can occupy. Not having adequate volume reduces
the performance of the antenna, with the bandwidth at the lower frequencies being
the first parameter where degradation is observed.
The main problem that arises when limiting the size or volume of an antenna is
achieving sufficient operational bandwidth. Wheeler1 defined the following general for-
mula that links the bandwidth of an antenna to its mode volume at a certain frequency:
This equation shows that the bandwidth ∆f over the central frequency f is linked by a
dimensionless number K to the ratio of the antenna mode volume to the wavelength.
The K factor is a figure of importance when, all things being equal, we want to com-
pare one antenna to another. The K factor is related to the antenna technology and
how it is designed. In Wheeler’s original paper, the antennas used to demonstrate
this were electric or magnetic dipoles and loops in freespace. This allowed the over-
all antenna mode volume to be defined according to its natural boundaries. For the
dipole, it was the sphere enclosing the dipole. In the case of electrically small anten-
nas, the problem is defining the antenna mode volume since in most cases it sig-
nificantly exceeds the physical volume of the antenna itself. Indeed, some antennas
Tunable and Adaptive Antenna Systems 105
Wheeler’s formula
Antenna mode volume required for fixed bandwidth
80
∆f Antenna mode volume
=K× LTE 700 band 460 MHz bandwidth
f (Radio wavelength)3 70
42 MHz 140 MHz bandwidth
50
GSM850/EGSM
40
GSM850 only
30
20 DCS/PCS/WCDMA
10
0
600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Frequency (MHz)
FIGURE 5.1 Antenna mode volume related to frequency of operation and bandwidth.
tend to couple strongly to the ground plane making the whole device the antenna.
This might lead to an apparent advantage when bandwidth alone is considered but
has many drawbacks for system performance and multiple antenna integration. For
this reason, the antenna alone should be designed to provide sufficient bandwidth.
Figure 5.1 shows a graph relating antenna mode to various center frequencies. The
curves shown are curves of fixed bandwidth, varying from 35 to 460 MHz. As can
be seen from this set of curves, as the center frequency of operation decreases and
as the required bandwidth increases, the antenna mode volume needed increases.
An important point to take away from this graph is that it requires a much smaller
volume to service 460 MHz of bandwidth at 2 GHz compared to 700 MHz.
to walls, furniture, and other obstructions cause fading in the propagation channel.
Over the years, antenna diversity schemes have been implemented in cell phones
and WLAN access points to compensate for deep fades, with these antenna diversity
schemes incurring volume, complexity, and cost constraints during product develop-
ment. MIMO systems, on the other hand, rely on a rich multipath environment to
assist in signal discrimination during the processing phase of the coincident transmit
and receive signals. Though multipath is desired for good MIMO system perfor-
mance, the fixed radiation pattern of the antennas that constitute the MIMO system
limit the performance of the wireless device, specifically in terms of MIMO system
coverage across all aspect angles. For both antenna diversity and MIMO applica-
tions, the inability to dynamically alter the radiation pattern of the embedded anten-
nas limits the performance of the antenna system, primarily due to the fact that the
gain maxima of the antenna is typically not optimized for mobile systems due to the
motion on the mobile side.
current flowing through the capacitor. To maintain low-loss tuning circuit character-
istics, a series configuration will require a Q in the range of 30–80, with this range
in Q dependent on the frequency of operation and antenna type. A shunt capacitor
configuration will allow for relaxed Q and power handling requirements compared
to a series configuration, which can translate to a reduction in tuning losses and
reduced component cost. Qs in the range of 10–30 can support low-loss tuning cir-
cuit requirements when the antenna design is taken into consideration and the tuning
circuit and antenna are designed as a system. A proper selection of antenna type can
allow for the use of a shunt tunable capacitor, which will provide the impedance tun-
ing range required to impedance match the antenna for the host device design and
use cases. To demonstrate the effect of Q on shunt capacitor performance in a tuning
circuit, a set of circuit simulations were performed where Q was varied and insertion
loss and return loss were monitored. A shunt capacitor was positioned between two
50 ohm ports with 50 ohm transmission lines connecting the two 50 ohm ports to
the terminals of the capacitor. The Q of the capacitor was varied from 10 to 100 and
the return loss and insertion loss of the circuit were monitored. Figure 5.2 shows the
circuit that was simulated along with the insertion loss and return loss as the Q of the
capacitor is varied. As can be seen at 820 MHz, the insertion loss of the circuit varies
by less than 0.15 dB when the Q is varied from 10 to 100. The return loss varies by
less than 0.2 dB across the 500 MHz to 10 GHz range.
5.2.1 Adaptive Matching
Open- and closed-loop impedance matching have both pros and cons in terms of
capabilities and system integration. Closed-loop impedance matching provides the
capability of impedance matching an antenna without input or control signaling from
the rest of the system (transceiver, baseband), which minimizes the mismatch loss as
the environment changes. Closed-loop matching does need to synchronize with the
transmit and receive functions and requires additional RF circuitry for implementa-
tion, which incurs cost and losses. Open-loop impedance matching is easier to imple-
ment, but typically requires input from the communication system such as frequency
band or channel to optimize for. A third technique discussed here, adaptive match-
ing, which can be considered a hybrid technique where additional capability can be
gained compared to an open-loop matching scheme by using additional information
from a baseband to provide additional tuning capability.
Current commercial wireless devices have one or multiple sensors or devices that
can be used to discern the use case or environment that the antenna system in a wire-
less device is exposed to. These sensors can be proximity sensors, display status,
speaker state, microphone status, and/or orientation sensor. By taking an antenna
element, tunable matching circuit formed with a tunable capacitor and passive com-
ponents, and a lookup table relating sensor inputs to capacitor settings, an adaptive
antenna solution can be implemented in a wireless device that matches the antenna
for a specific frequency band and adjusts the impedance match of the antenna to
compensate for an estimated use case or environment. Figure 5.3 shows a sche-
matic of an adaptive matched antenna, with baseband of the host device providing
information such as frequency band or channel and other stimuli for use in making
108 Tunable RF Components and Circuits: Applications in Mobile Handsets
decisions regarding antenna tuning. Figure 5.4a shows a plot of return loss for an
adaptive antenna embedded in a cell phone; in this plot, there are several traces
related to specific tunable capacitor tuning states. This return loss plot shows three
resonances: low, mid, and high band, which provides the capability to cover a major-
ity of 3G and 4G cellular bands with a single port antenna, which is typical for main
antennas in current cell phones. The low-band resonance can be seen to shift in
MTEE_ADS
Tee1
Subst = "MSub1"
W1 = 1.46 mm
W2 = 1.46 mm
W3 = 1.46 mm
Term
MLIN MLIN
Term 1 Term
TL1 TL2
Num = 1 CAPQ Term 2
Subst = "MSub1" Subst = "MSub1"
Z = 50 Ohm Num = 2
W = 1.46 mm C4 W = 1.46 mm
L = 10 mm C = 1.0 pF L = 10 mm Z = 50 Ohm
Q=Q
F = 820 MHz
Mode = proportional to frequency
MSub
S-parameters
MSUB
S_Param MSub1
SP1 H = 31 mil
Start = 0.10 GHz Var VAR Er = 4.4
Eqn
Stop = 10.0 GHz VAR1 Mur = 1
Step = 0.01 GHz R = 10 {t} {d} Cond = 5.8E + 7
Q = 10 {t} {d} Hu = 3.9e + 034 mil
C = 1 {t} {d} T = 35 um
TanD = 0.02
Rough = 0 mil
Bbase =
Dpeaks =
m12
Frequency = 820.0 MHz
dB(S(2,1)) = –0.128 –0.12 m12
Q = 100.000000
–0.14
m11
Frequency = 820.0 MHz –0.16
dB(S(2,1))
dB(S(2,1)) = –0.226
–0.18
Q = 10.000000
–0.20
–0.22 m11
m14
0
–0.24
819.85 819.90 819.95 820.00 820.05 820.10 820.15 820.20 820.25
–1
Frequency (MHz)
–2
dB(S(2,1))
–3
–4
–5
–6
0 1 2 3 4 5 6 7 8 9 10
Frequency (GHz)
m7
Frequency = 1.120 GHz
dB(S(1,1))= –15.963
m7 Q =10.000000
Frequency = 1.120 GHz m6
dB(S(1,1))= –15.963 Frequency = 1.120 GHz
dB(S(1,1))= –15.914
Q =10.000000 Q =100.000000
m6 –15.90
Frequency = 1.120 GHZ –15.91 m6
dB(S(1,1))= –15.914
–15.92
Q =100.000000
dB(S(1,1))
–15.93
–15.94
0 –15.95
–5 –15.96 m7
–10 –15.97
m6
1.119970
1.119975
1.119980
1.119985
1.119990
1.119995
1.120000
1.120005
1.120010
1.120015
1.120020
1.120025
–15
dB(S(1,1))
–20
Frequency (GHz)
–25
–30
–35
–40
0 1 2 3 4 5 6 7 8 9 10
Frequency (GHz)
RFFE
RF Tunable
Baseband XL
matching
circuit
Lookup XL Tunable
Frequency band table capacitor
Control signal
status
other stimulus
frequency as the tunable capacitor is varied, with the mid- and high-band resonances
either being optimized or detuned as the tunable capacitor is varied. The return loss
traces shown here are for a “freespace” use case defined as a use case where there
are no objects in the vicinity of the cell phone (neither hand nor head loading occurs
for this use case). Figure 5.4b highlights the benefit of an adaptive matched antenna.
Instead of the wideband frequency sweep shown in Figure 5.4a,b shows the return
loss of the low-band resonance for the hand-loading use case. The multiple return
loss traces relate to specific capacitor tuning states. The desired frequency band of
interest is denoted with the markers at 824 and 894 MHz, with the resonances to the
110 Tunable RF Components and Circuits: Applications in Mobile Handsets
–10.00
–25.00
Tuning applied to Bands
–30.00
5, 6, and 8
–35.00
Ch 1: Start 700.000 MHz Stop 3.00000 GHz
(a)
894 MHz
824 MHz
15.00
10.00
5.000
0.000
R-hand
grip
–5.000
1pF
–10.00 2pF
–15.00
3pF
4pF
–20.00
5pF
–25.00
(b)
FIGURE 5.4 (a) Return loss showing three resonances from the single-feed antenna installed
in a cell phone; multiple traces relate to different capacitor tuning values. (b) Return loss cor-
rected for hand loading by adjusting the value of the tunable capacitor in the matching circuit.
left of these markers representing the return loss of the antenna with hand loading
of the cell phone prior to selecting the optimal capacitor tuning state. As the tunable
capacitor is varied, the return loss can be improved at the frequency band of interest.
Proximity sensor status is used to determine which capacitor tuning state to com-
mand, with the lookup table relating such functions as frequency band, proximity
Tunable and Adaptive Antenna Systems 111
sensor status, speaker, or microphone status to tunable capacitor tuning states. The
lookup table is developed during the adaptive antenna design phase.
the common port of the RF switch. One port of the RF switch is configured in
an “open” state, where the port is not connected to any structure or component.
The additional ports of the RF switch are connected to the ground either directly
or using lumped components, such as capacitors and inductors. The frequency
response of the antenna can be shifted by switching from one port of the RF switch
to another port. More specifically, when the open switch port is activated, the
parasitic element has negligible effect on the frequency response of the antenna.
This is due to the negligible level of coupling between the antenna and the para-
sitic and the electrical length of the parasitic element. When a second switch port,
connected to a reactive lumped component and the component in turn connected
to the ground is activated, the frequency response of the antenna will shift. The
shift in frequency response is due to two effects: (1) the electrical length of the
parasitic element is altered when the reactive lumped component is connected to
it, and (2) the level of coupling between the parasitic element and the antenna ele-
ment is increased when the parasitic element is attached to the ground (through a
lumped reactance, in this case) (Figure 5.5). Figure 5.6 shows the antenna struc-
ture with parasitic element for two reactive loading states that provide different
frequency responses, with the first loading state of the parasitic element being an
open circuit and the second loading state being a fixed value capacitor. Figure 5.7
Switch
RFFE module
RF
Baseband
Lookup
Frequency band status table
Other stimulus Control signal
y x y x
z z
FIGURE 5.6 IMD antenna configuration 1 with parasitic element; two reactive loading
states are shown.
Tunable and Adaptive Antenna Systems 113
S1, 1_Capacitor
S-parameter (Magnitude in dB) S1, 1_OPEN
5
–5
–10
–15
80 MHz shift in low-
230 MHz shift of high-
band resonance
band resonance
–20
–25
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7
Frequency (GHz)
FIGURE 5.7 Return loss of the two loading states of the parasitic element for antenna
configuration 1 shown in Figure 5.6.
shows the return loss for both loading states of the IMD antenna; notice that
both low- and high-band resonances have shifted when the reactive loading on
the parasitic element changes from an open circuit to a capacitive reactance. The
lower frequency response shifts lower in frequency while the upper frequency
response shifts higher in frequency.
Feedpoint Feedpoint
y z x y z x
FIGURE 5.8 IMD antenna configuration 2 with the parasitic element; two reactive loading
states are shown.
S1, 1_OPEN
S-parameter (Magnitude in dB) S1, 1_SHORT
5
–5
–10
–15
–20
–30
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7
Frequency (GHz)
FIGURE 5.9 Return loss of the two loading states of the parasitic element for antenna
configuration 2 shown in Figure 5.8.
parasitic element being an open circuit and the second loading state being a fixed
value capacitor. Figure 5.9 shows the return loss for both loading states of the IMD
antenna; notice that in this configuration, the both low resonance has shifted when
the reactive loading on the parasitic element changes from an open circuit to a
capacitive reactance while the high-band frequency response remains practically
unchanged.
Tunable and Adaptive Antenna Systems 115
Feedpoint Tuning
z
Ground plane x
–5
–10
–15
–20
–25
–30
S1, 1_05pF
–35 S1, 1_2pF
S1, 1_OPEN
–40 S1, 1_SHORT
–45
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3
Frequency (GHz)
FIGURE 5.10 Antenna design, return loss, and total efficiency of a band switched antenna
showing frequency response roll-off at low-band resonances. (Continued)
116 Tunable RF Components and Circuits: Applications in Mobile Handsets
1D Results\Efficiencies\dB
0
–1
–2
–3
–4
dB
–5
–6
Tot. Effic. [1]_05pF
–7
Tot. Effic. [1]_2pF
–8 Tot. Effic. [1]_OPEN
Tot. Effic. [1]_SHORT
–9
0.6 0.62 0.64 0.66 0.68 0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 1
Frequency (GHz)
S-parameter (Magnitude in dB)
5
–5
–10
–15
dB
–20
–25
–30 S1, 1_05pF
S1, 1_2pF
–35 S1, 1_OPEN
S1, 1_SHORT
–40
0.6 0.62 0.64 0.66 0.68 0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 1
Frequency (GHz)
FIGURE 5.10 (Continued) Antenna design, return loss, and total efficiency of a band
switched antenna showing frequency response roll-off at low-band resonances.
band switching antenna designed for integration into the edge of a host device where
four low-band tuning states are set up to provide tuning across the 700–960 MHz
frequency range. The roll-off in return loss and total efficiency shows that several
dB of attenuation can be achieved out-of-band for this type of antenna compared to
the broader bandwidth passive antennas, with this additional a ttenuation providing
better isolation.
radiation pattern of an antenna, the antenna can now be optimized for specific mul-
tipath environments. This technique of dynamically changing the radiation pattern
of an antenna can be thought of as a method to match or optimize the second port
of an antenna, the “radiating port,” to the propagation channel of a communica-
tion system. This dynamic adjustment of the radiation pattern of the antenna can
be implemented by adjusting the radiation pattern of a single embedded antenna,
termed a “modal antenna” or can be implemented by using traditional antenna array
techniques where two or more antenna elements are combined to generate a directive
pattern that can be scanned by adjusting the phasing of the feed network.
Frequency
Shorting parasitic to ground generates two modes
z
z
x y
x y
Z Z
f0 resonance with switch open
f1, f2 double resonance when
switch shorts the parasitic X X
element to ground
–5
2
–10 q1 (1.96, –16.402)
q2 (1.876, –10.314)
q3 (2.026, –15.329) f1 1
3 Mode 0
–15 Mode 1
f0 f2
–20
1.6 1.65 1.7 1.75 1.8 1.85 1.9 1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4
Frequency (GHz)
or magnetic (TM) modes and is used simply as a term to specify the radiation state
of the antenna. The antenna structure generates two fundamental radiation states
Mode 0 and Mode 1 associated to when the parasitic is in an open-circuit state
and short-circuit state, respectively. Mode 0 exhibits a single frequency response
centered at frequency f0 generated from the IMD antenna element. Mode 1 exhibits
a dual frequency response at frequencies f1 and f 2, wherein f1 exhibits a radiation
pattern similar to the IMD antenna element (f0), and f2 exhibits a radiation pattern
wherein the nulls in the radiation pattern for f1 are rotated by almost 90° as seen in
Figure 5.14.
Tunable and Adaptive Antenna Systems 119
A/m
12
A/m
12
FIGURE 5.13 Current distribution plots at frequencies f1 and f2 of a null steering antenna.
The surface current distribution plots at frequencies f1 and f2 in Figure 5.13 show
that for f1, the current distribution vector on the parasitic arm is in the same direction
as the inner arm of the IMD antenna element, whereas it is in the opposite direction
compared to the inner arm of the IMD antenna in the case of frequency f2. This
variation in the near-field characteristics of the antenna structure result in a variation
in the far-field characteristics at frequencies f1 and f2. The far-field variation can be
measured in terms of change in the co- and cross-polarization gain between frequen-
cies f1 and f2. The 3D radiation patterns for f1 and f2 in Figure 5.14 also show that the
FIGURE 5.14 3D radiation pattern at frequencies f1 and f2 for a null steering antenna.
120 Tunable RF Components and Circuits: Applications in Mobile Handsets
–5
2
–10
Q1 (1.96, –16.402)
Q2 (1.876, –10.314)
Q3 (2.026, –15.329) 3 Mode 0
1
–15 Mode 1
–20
1.6 1.65 1.7 1.75 1.8 1.85 1.9 1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4
Frequency (GHz)
null in the radiation pattern for f2 (along the +Z axis) is rotated by almost 90° in the
radiation pattern for f1 (along the –X axis) (Figure 5.15).
The parasitic element placed offset from the main IMD antenna element can be
terminated in either open-circuit or short-circuit state to generate to fundamental
radiation states at frequencies f0 (Mode 0) and frequency f2 (Mode 1). These two
frequency points define the frequency bandwidth over which it is possible to generate
two or more radiation patterns, which will be de-correlated in far field. In terms of
design guideline, these two states will be the fundamental states and it is possible to
generate more states by terminating the offset parasitic elements in lumped capaci-
tors, inductors or LC circuits.
This modal antenna technique provides the ability to generate multiple radiation
patterns from a single port antenna, and a direct application of this type antenna is
for a receive diversity in 3G handset applications. A single modal antenna can be
implemented in a 3G application where the receive diversity antenna and second
receive port in the transceiver can be eliminated. Using a modal antenna as the
main antenna in the handset will provide switched diversity from the single modal
antenna. This implementation will result in reduced power consumption due to
the elimination of the second receive port in the transceiver, and will also result
in reduced cost and reduced internal volume required from the antenna system.
The modal antenna can address these issues by proving pattern diversity from
a single radiating structure without adding a second receive chain. The modal
antenna architecture is being designed keeping these aspects in mind with an aim
of providing an antenna system which provides the maximum diversity gain and
improves the link budget by adding several dB of improvement to the fading mar-
gin. In essence, the modal technique utilizes novel antenna architecture capable of
generating two or more unique modes with low ECC and comparable efficiencies.
The technique will utilize a switched combining technique to switch between the
Tunable and Adaptive Antenna Systems 121
RFFE RFIC
RF
Baseband
Algorithm
CQI
Control signal
RSSI
modes. The goal here would be to have possible gains that would be comparable
to maximum ratio combining (MRC), which increases the hardware required and
the associated complexity.
The antenna hardware for the modal antenna is satisfied with an IMD antenna, a
parasitic element, and multiport RF switch. To complete the modal antenna system,
an algorithm is developed and implemented to provide the decision-making pro-
cess for optimal radiation mode selection in a dynamic environment. The algorithm
requires a single metric from the host device baseband processor, with this metric
being a low-latency SINR, RSSI, CQI, or equivalent metric, which characterizes the
quality of the communication link. The algorithm can be located in any processor
in the host device such as baseband or the application processor. The role of the
algorithm is to compare the link quality metric to the sampled radiation modes and
choose the optimal mode for the communication link. A periodic sampling of the
additional modes (the modes not used for communication at a specific instance) is
made to determine, which mode to switch to and when to switch to keep the commu-
nication link optimized. Figure 5.16 shows a schematic of a modal antenna system
implemented in a communication device.
The modal antenna architecture would be designed k eeping these aspects in mind
with an aim of providing an antenna system, which provides the maximum diversity
gain and improves the link budget by adding s everal dB of improvement to the fad-
ing margin.
5.5 CONCLUSIONS
The adaptive antenna techniques described in this chapter, adaptive matching, band
switching, and beam steering, provide the capability of dynamically altering antenna
characteristics, such as impedance, resonant frequency, and radiation patterns.
These adaptive antenna techniques are realized by taking advantage of onboard
sensors and signaling found in today’s commercial wireless devices. These sensor
inputs can be proximity sensor status, frequency band information, display light-
ing status, microphone, speaker state, and so on. One or multiple adaptive antenna
techniques can be combined to optimize multiple parameters simultaneously.
122 Tunable RF Components and Circuits: Applications in Mobile Handsets
REFERENCE
1. Wheeler, H.A., Fundamental limitations of small antennas, Proceedings of the IEEE,
35(12), 1479–1484, December 1947.
6 Effective Antenna
Aperture Tuning
with RF-MEMS
Larry Morrell and Paul Tornatta
CONTENTS
6.1 Antennas and MEMS: Advantages of MEMS for
Aperture-Tuned Antennas............................................................................. 124
6.1.1 Key Performance Parameters of MEMS........................................... 124
6.1.2 Trade Space of MEMS Implementations.......................................... 125
6.1.3 Antenna Selection for Antenna Aperture Tuning.............................. 125
6.2 Trade-Off of Instantaneous Bandwidth and System Requirements.............. 125
6.2.1 System Operation: How Wide Does the Bandwidth Need to Be
at Any One Time������������������������������������������������������������������������������ 125
6.2.2 Antenna Design: Frequency Bands and Simultaneous Usage
Requirements������������������������������������������������������������������������������������ 126
6.2.3 Multimode, Multiband Performance................................................. 127
6.3 Antenna Volume and Efficiency Considerations........................................... 127
6.3.1 The Small Antenna Limit.................................................................. 127
6.3.2 Baseline Requirements for Antennas in Smartphones...................... 127
6.4 Maximizing Efficiency: The Effect of Loading and Unloading
the Antenna............................................................................................... 128
6.4.1 Antenna System Parasitic Effects to Load and Unload
the Antenna....................................................................................... 128
6.4.2 Frequency Shifting and Tuner Precision........................................... 129
6.4.3 Effect of Amplifier Load-Pull on Tuning Component Resolution........ 129
6.5 Tunable Component Placement on the Antenna Structure............................ 130
6.5.1 The Impact of Capacitance Range on Radiation Efficiency.............. 130
6.5.2 The Impact of Quality Factor (Q) on Radiation Efficiency............... 133
6.5.3 The Impact of Cmin on Radiation Efficiency...................................... 133
6.5.4 Linearity Consideration in Tuner Placement..................................... 134
6.5.5 Voltage and Impedance Considerations............................................ 134
6.6 Smartphone Tunable Antenna Design........................................................... 135
6.6.1 Antenna Design Considerations in Modern Smartphones................ 135
6.6.2 Metal Frame Antenna Design............................................................ 135
6.7 Conclusions.................................................................................................... 142
References............................................................................................................... 143
123
124 Tunable RF Components and Circuits: Applications in Mobile Handsets
1 1 ωL (6.1)
Q= , where Qc = and QL =
(1/QL )+ (1/QC ) ωCRC RL
Keeping the resistance low and minimizing parasitic inductance and capacitance
allows the MEMS device to approach the performance of an ideal component.
For antenna aperture tuning, all of the key performance parameters must be
optimized simultaneously. RF MEMS allows for the optimization of all parameters
without having to trade-off between them.
Antenna
performance Untuned
antenna performance
for roaming, hand-off, and other system management functions. Carrier aggregation
(CA) requires the phone to operate in two bands at the same time. CA requirements
depend on the wireless service provider and regional requirements (Figure 6.2).
Δf (a/λ)3 (6.2)
∝
f η
This well-publicized relationship was developed in the 1940s has been shown empir-
ically to reflect antenna performance for electrically small antennas.1,2 This equation
states that antenna size expressed in wavelengths has a direct impact on the antenna
efficiency and bandwidth. If the antenna size is limited by industrial design con-
strains, then the instantaneous bandwidth and efficiency will also be limited.
design constraints are making it difficult for a fixed-tuned antenna to cover the fre-
quencies of interest with adequate performance.
L2 C2
L1 C1 G1 R2
Zin N1 : 1 1 : N2
Antenna Chassis
also determine if the tuning element must present a capacitive or inductive reactance
in order to tune the resonance of the antenna. If the fundamental resonant frequency
of the antenna is at the low end of the frequency band, the reactance of the tuner must
be inductive, to “pull” the frequency up. If the fundamental resonant frequency of
the antenna is on the high end of the frequency band, the reactance of the tuner must
be capacitive to “pull” the frequency down.
Peak Pout
31 dBm max
0.5 dB/step
LTE PAE
45% max
5%/step
1.0
0.9
1.2
0.8
1.4
0.7
1.6
0.6
1.8
Cntl=21.000
Cntl=21.000
Cntl=21.000
Cntl=21.000
Cntl=21.000 Cntl=20.000
Cntl=21.000
Cntl=21.000
Cntl=21.000
Cntl=21.000
Cntl=20.000
Cntl=20.000
Cntl=21.000 Cntl=21.000
2.0
Cntl=21.000
Cntl=21.000
0.5
Cntl=21.000
Cntl=22.000
Cntl=23.000
0.4
3.0
Cntl=24.000
0.3
4.0
S(1,1)[find_index(Cntl, indep(m7)),::]
Cntl=25.000
0.2 5.0
Cntl=26.000
0.1 10
Cntl=27.000
20
S(1,1)
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.2
1.4
1.6
1.8
2.0
3.0
4.0
5.0
10
20
Cntl=28.000
–20
–0.1 Cntl=29.000
–10
Cntl=30.000
–0.
2 Cntl=31.000 .0
–5
.0
–4
.3
–0
.0
–3
.4
–0
.0
.5
–0
–2
.6
.6
–1
6
–0
–1.
7
–1.4
–0.
–0.8
–1.2
–0.9
–1.0
FIGURE 6.6 Locus of antenna impedance as the capacitive load shifts. (Note: Increasing
control state corresponds to an increasing capacitance.)
Changing the resonance of the antenna causes the impedance locus on the Smith
chart to rotate in a predictable way. With fine resolution in the tuning element, it is
possible to make small favorable changes to the impedance not affecting the radia-
tion efficiency but having a big impact power amplifier loading. The overall impact
to power added efficiency of the RF chain can be several decibels (dB). The diagram
in Figure 6.6 shows the trajectory of the antenna impedance as the resonant fre-
quency is changed.
Moving the antenna impedance around the Smith chart allows it to pass through
different regions of the power amplifier load-pull diagram changing output power
and efficiency of the amplifier. Figure 6.7 shows the impact of impedance on TRP
and power amplifier current, Idd.
25
20
TRP versus C-state
15
TRP (dBm)
10
5
0
–5
–10
0 4 8 12 16 20 24 28 32
State index (0..31)
300
250
Idd versus C-state
200
Idd (mA)
150
100
50
0
0 4 8 12 16 20 24 28 32
State index (0..31) ABC
FIGURE 6.7 Total radiated power and power amplifier current over the capacitance range.
Voltage distribution
1 2
Current
distribution
DVC placement region
Feedpoint
FIGURE 6.8 Voltage and current distribution along a typical PIFA structure.
132 Tunable RF Components and Circuits: Applications in Mobile Handsets
FIGURE 6.9 Tuning range as a function of the capacitance range. (Note: [1] is placed nearer
the feedpoint; [2] is placed farther from feedpoint and as a result has a higher ΔFrequency/
ΔCapacitance [tuning ratio].)
Efficiency (%)
0.85 50
0.8 40
0.75 30
2.5–5.0 pF
0.7 20 1.5–3.5 pF
1.0–2.5 pF
0.65 10 0.5–1.5 pF
0.6 0
0 1 2 3 4 5 6 0.6 0.7 0.8 0.9 1
(a) Capacitance (pF) (b) Frequency (GHz)
FIGURE 6.10 Tuning ratio: Impact of the capacitive range on efficiency and frequency
tuning range. (a) Small capacitance ranges can produce an equivalent change in frequency
compared to larger ranges if Cmin is smaller. (b) Smaller capacitance loading on the antenna
produces higher radiated efficiency while still covering the desired frequency range.
From the illustration in Figure 6.8, moving farther out on the PIFA increases the
capacitive loading effect of the tuner and increases the voltage at the terminals of the
tuner. At increased voltage, the loading of the capacitor will have a larger effect on
the resonant frequency; therefore, it is possible to get the same frequency tuning range
for different capacitance tuning ranges depending on where the variable capacitor is
placed along the length of the antenna. Figure 6.9 shows two different capacitor ranges
at different location on the antenna producing the same frequency tuning range.
Although it is possible to get the same frequency tuning range using different
capacitor values, the maximum radiation efficiency is achieved when using the
smallest capacitive loading values. Figure 6.10a shows the same frequency tuning
range achieved with different capacitor ranges and the impact on radiation effi-
ciency (Figure 6.10b).
Effective Antenna Aperture Tuning with RF-MEMS 133
3.5
Slope is a function of antenna design
3
Implementation loss (dB)
2.5
MEMS
2 implementation
loss
1.5
1
Solid-state switch
0.5 implementation
loss
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Equivalent series resistance (ohms)
80
V RMS at +33 dBm power
15 mm
12.5 mm 7.5 mm 0.4 pF
10 mm
60 5 mm 1.0 pF
40
20
0
500 600 700 800 900 1000 1100 1200
Frequency (MHz)
FIGURE 6.12 Antenna voltage at 33 dBm with different tuner locations. (Note: At 15 mm
from the feedpoint, nearly 70 VRMS is generated on the tuner.)
Effective Antenna Aperture Tuning with RF-MEMS 135
and the solid traces are at 1.0 pF. The rightmost trace of each group is at 5 mm from
the feedpoint and the leftmost trace is 15 mm. The increment between traces is
2.5 mm.
Metal frame
Main ground
Ground feedpoint
Antenna aperture tuner
Main antenna
TABLE 6.1
Comparison of Digital Variable Capacitors (DVC) Used for Design
Parameters at the Board Side of Bump Pads
Capacitor Values Used—of 32 Possible States (pF)
Device R (Ohm) L (nH) Cmin C10 C20 Cmax
32CK301 0.3 0.8 0.45 0.64 0.84 1.05
32CK417 0.4 0.8 0.55 0.92 1.29 1.70
32CK402 0.4 0.8 0.70 1.14 1.57 2.05
The impact of ESR was studied using a combination of discrete capacitors and
resistors. The capacitor is placed in parallel across the aperture of the antenna, the
same way the MEMS device is placed on the antenna. A discrete resistor is then
placed in series with the capacitor to simulate a device with a higher ESR. The objec-
tive of the study is to show the impact of ESR on radiation efficiency.
Figure 6.16 shows a mockup of the frame antenna with a discrete capacitor and
series resistor in place of the MEMS aperture-tuning device.
The resistance is changed from 0 to 2 ohms. In each case, the return loss (S11)
magnitude and efficiency are measured. The experiment was performed using a
2.7 pF capacitor at a distance of 5 mm from the feedpoint.
Figure 6.17 shows the S11 measured results and Figure 6.18 shows the measured
efficiency.
–5
–10
S11 Log Mag (dB)
–15
Cmin
–20 C10
–25 C20
Cmax
–30
–35
–40
700 750 800 850 900 950 1000
(a) Frequency (MHz)
–8.00 Cmin
–10.00 C10
–12.00 C20
–14.00 Cmax
–16.00
–18.00
–20.00
700 750 800 850 900 950 1000
(b) Frequency (MHz)
FIGURE 6.14 Comparison of return loss (S11) (a) and efficiency (b) for different DVC
devices and locations. (Continued)
138 Tunable RF Components and Circuits: Applications in Mobile Handsets
–5
–10
S11 Log Mag (dB)
–15
Cmin
–20 C10
C20
–25
Cmax
–30
–35
–40
700 750 800 850 900 950 1000
(c) Frequency (MHz)
Cmin
–8.00
C10
–10.00
C20
–12.00
Cmax
–14.00
–16.00
–18.00
700 750 800 850 900 950 1000
(d) Frequency (MHz)
FIGURE 6.14 (Continued) Comparison of return loss (S11) (c) and efficiency (d) for differ-
ent DVC devices and locations. (Continued)
Effective Antenna Aperture Tuning with RF-MEMS 139
–5
S11 Log Mag (dB)
–10 Cmin
C10
–15 C20
Cmax
–20
–25
700 750 800 850 900 950 1000
(e) Frequency (MHz)
–2.00
–4.00
Efficiency (dB)
–6.00 Cmin
C10
–8.00 C20
Cmax
–10.00
–12.00
–14.00
700 750 800 850 900 950 1000
(f) Frequency (MHz)
FIGURE 6.14 (Continued) Comparison of return loss (S11) (e) and efficiency (f) for differ-
ent DVC devices and locations.
140 Tunable RF Components and Circuits: Applications in Mobile Handsets
0.00
Loss due to DVC <0.25 dB
–2.00
Impact of using
higher capacitance
–4.00
–6.00
Efficiency (dB)
–10.00
Unloaded frame
–12.00 301 Cmin 15 mm
417 Cmin 10 mm
402 Cmin 5 mm
–14.00
–16.00
700 750 800 850 900 950 1000
Frequency (MHz)
Antenna
aperture tuner
Metal frame
Ground
Main antenna Feedpoint
Capacitor
Resistor
Main ground
FIGURE 6.16 Development platform for measuring return loss and efficiency.
Effective Antenna Aperture Tuning with RF-MEMS 141
2.7 pF cap, 5 mm
0
–1
–2
–3
S11 Log Mag (dB)
0 Ohm
–4
0.5 Ohm
–5 1.0 Ohm
–6 2.0 Ohm
–7
–8
–9
700 750 800 850 900 950 1000
Frequency (MHz)
2.7 pF, 5 mm
0.00
–2.00
–4.00
Efficiency (dB)
–6.00 0 Ohm
–8.00 0.5 Ohm
1 Ohm
–10.00
2.0 Ohm
–12.00
–14.00
–16.00
700 750 800 850 900 950 1000
Frequency (MHz)
2.7 pF, 5 mm
–2.00
–3.00
–4.00
Efficiency (dB)
0 Ohm
–5.00 0.5 Ohm
1 Ohm
–6.00 2 Ohm
–7.00
–8.00
720 730 740 750 760 770 780 790 800
Frequency (MHz)
FIGURE 6.19 Magnified scale showing antenna efficiency impact of ESR over frequency.
The chart in Figure 6.19 shows the measured efficiency with a magnified scale in
order to more clearly see the impact of high capacitance and high ESR on the radia-
tion efficiency.
6.7 CONCLUSIONS
Antenna design challenges in smartphones are driven by industrial design constraints
and a proliferation of frequency bands of operation. Designing aperture-tuned anten-
nas to meet the challenge offers significant advantages in total radiated power and
receiver sensitivity over fixed frequency and variable impedance-tuned antenna
designs. When designing an aperture-tuned antenna, it is important to understand
the unique design requirements and select the appropriate tuning element.
The performance characteristics of RF MEMS are particularly well suited to
aperture-tuned antennas. RF MEMS devices have very low equivalent series resis-
tance (ESR). The typical RF MEMS capacitor has an ESR from 0.2 to 0.5 ohm. This
parameter is particularly important as it is directly related implementation loss. As
the ESR increases, the implementation loss increases. ESR > 0.5 ohm can lead to
implementation losses from 1 to 3 dB. This much loss is unacceptable in the front
end of the RF chain. In addition to ESR, minimum capacitance or Cmin is also a criti-
cal parameter. The fundamental antenna structure acts as a distributed network of
R’s, L’s, and C’s. A shunt variable capacitance placed in the antenna aperture will
load the radiating element and lower the resonant frequency. To compensate, the
antenna must be made shorter. To preserve antenna length and maximum radiation
efficiency, the antenna must be loaded as little as possible. A very low Cmin, <0.5 pF,
allows the fundamental antenna to be as big as possible preserving the radiation
efficiency.
Effective Antenna Aperture Tuning with RF-MEMS 143
REFERENCES
1. Wheeler, H.A., Fundamental limitations of small antennas, Proceedings of the IRE, 35,
1497–1484, December 1947.
2. Chu, L.J., Physical limitations of omnidirectional antennas, Journal of Applied Physics,
19, 1163–1175, December 1948.
3. Vainikainen, P. et al., Resonator-based analysis of the combination of mobile handset
antenna and chassis, IEEE Transactions on Antenna and Propagation, 50(10), 1433–
1444, October 2002.
7 RF-CMOS Impedance
Tuners
Performance Metrics and
Design Trade-Offs
Tero Ranta
CONTENTS
7.1 Introduction—Toward a Fully Integrated Reconfigurable Cellular RF
Front End....................................................................................................... 145
7.2 Fundamentals and Design Trade-Offs of Semiconductor Switched
Capacitors...................................................................................................... 146
7.2.1 Basic Concepts and Fundamental Trade-Offs................................... 146
7.2.2 Real-Life Performance Is Dominated by Parasitic Effects............... 149
7.2.3 Component Quality Factor................................................................ 151
7.3 Impedance Tuning Networks......................................................................... 152
7.3.1 Tuning Network Circuit Topology and Trade-Offs........................... 152
7.3.2 Cellular System Level Considerations............................................... 154
7.3.3 Dealing with Exponential Complexity.............................................. 154
7.3.4 Impedance Tuner Compact Model.................................................... 155
7.3.5 Impedance Tuner Visualization......................................................... 158
7.3.6 Smith Chart Constellations................................................................ 158
7.3.7 Smith Chart Surfaces......................................................................... 158
7.3.8 Transducer Gain Contours................................................................. 159
7.3.9 Density Plots for Visualizing Large Amounts of Data...................... 161
7.4 Conclusions.................................................................................................... 162
References............................................................................................................... 162
145
146 Tunable RF Components and Circuits: Applications in Mobile Handsets
throws, and antennas, but also requires fundamental modifications to the radio fre-
quency (RF) front-end architecture to support flexibility and reconfigurability.
Monolithic integration using complementary metal oxide semiconductor
(CMOS) technology has been a growing trend in the baseband, application proces-
sor, and transceiver areas of the cellular device. Over the past several years the
CMOS revolution has also swept the cellular RF front end, enabled by significant
technological advancements in silicon-on-sapphire (SOS) and silicon-on-insulator
(SOI) technologies. The growing use of CMOS SOI technology in the RF front-
end components, such as antenna switches and power amplifiers also makes it a
natural candidate for implementation of impedance tuning devices. This enables
monolithic solid-state integration for a truly fully integrated reconfigurable CMOS
SOI solution and provides a small die area, good cost structure, high yield, and high
reliability required for very high-volume mass production.
Key cellular applications for tuning devices include antenna impedance tuning5–7,10;
antenna aperture tuning8; power amplifier output impedance tuning; tunable phase
shifters9; duplex filter tuning; and diplexer, coupler, and low-pass filter tuning. Typical
specification requirements for tuning devices used in these applications can be sum-
marized,2–11 as supporting capacitance ranges within 0.5–30 pF, tuning ratios from 3:1
to 30:1, Quality factors in the range of 20–80 at 1–2 GHz, ability to withstand instan-
taneous peak RF voltages of 15–90 VPK, and RF power levels above +35 dBm, being
able to reliably handle 10–100 billion switching cycles and being controlled using a
high-speed serial digital interface such as MIPI RF front-end control interface (RFFE).
100
Vdd 80
Q
60
40
Serial
20
bus
0
RF– 1 2 3 4 5 6 7 8 9 10
(a) (b) Tuning ratio
FIGURE 7.1 Basic switched capacitor implementation (a) and tuning ratio versus Quality
factor trade-off versus RONCOFF (b).
5 100
C0
90 C1
4 80 C2
70 C4
Capacitance (pF)
C8
3 60 C15
50 C31
Q
2 40
30
1 20
10
0 0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 0.5 1 1.5 2 2.5 3
State Frequency (GHz)
C0
C1
C2
C4
C8
C15
C31
FIGURE 7.2 Fundamental C, Q, and S11 behavior for a 0.9–4.6 pF digitally tunable capacitor.
between C0 and C31 is a combination of the two behaviors depending on how many
bits are ON and OFF.
The main differentiator between RFSOI and bulk CMOS switched capacitor
implementations is that SOI enables the use of FET stacking to arbitrarily increase
the RF peak voltage handling (VPK) of the tunable component to levels required by
cellular devices. This is made possible by the use of a fully insulating SOI (or SOS)
substrate, which enables nearby FETs to float with respect to each other, enabling
increase of voltage handling from 2 to 4 VPK for a single FET to up to 100 VPK for
the whole stack. This allows the RF power handling capability to be fully tailored
to the specific application requirements by simple circuit design techniques, rather
than having to modify materials properties of dielectrics or mechanical properties
of moving parts.
The overall die area of the switch FET stack is proportional to the square of stack
height in order to maintain the same total DC resistance across the stack regardless
of the stack height (i.e., going from A stack of one to two requires doubling the size
of each device to maintain constant total resistance, which quadruples the total die
area). This causes the die to grow very rapidly with increase in VPK specification. In
practice this is limited by (1) parasitic capacitance resulting from large die area and
(2) limited available die area, which bounds the achievable Cmin, VPK, and FET stack
DC resistance (i.e., Q).
RF-CMOS Impedance Tuners 149
FIGURE 7.3 Example of a product level equivalent circuit model including all parasitic
effects.
150 Tunable RF Components and Circuits: Applications in Mobile Handsets
Another major effect of the distributed parasitic shunt capacitance is that it causes
the RF voltage to divide unevenly across the FET stack. The overall voltage handling
of the switch FET stack is limited by the weakest device (highest relative RF voltage
in the stack). Due to the nature of a series-shunt capacitance ladder, the FETs on top of
the stack see the highest relative voltage across them and start clipping the RF signal
first resulting in high distortion. The voltage division must be equalized in order to get
the full peak voltage handling performance, especially for stack heights above approxi-
mately 8 as uneven voltage division increases almost exponentially with increasing
stack height. Often tunable devices need to support both shunt and series operation.
This makes it even more challenging to achieve high-peak voltage handling, as the
FET stack may experience highest per-device voltages on top or bottom of the stack,
depending on whether the terminals see high- or low-impedance levels when being
used in the tuning network. It is very important to codesign the tuning network and
tunable capacitors as a whole to optimize performance of the whole network.
Another side-effect of increasing stack height is that it makes the circuit more sus-
ceptible to unintended Q degradation (de-Q’ing) due to the bias resistors connected
to the gates of the stacked FETs. The Q degradation can be reduced by increasing
value of the gate resistor at the expense of switching time which is dominated by the
RC time constant between gate bias resistor and FET gate capacitance. To further
reduce the de-Q’ing effect, the MIM capacitors are usually placed on top of the
FET stack for shunt connected devices making them “invisible” to the gate resistors.
MIM capacitors may be placed at both ends of the FET stack if full DC blocking
is required in series configuration, but this typically requires adding a separate DC
path to the FETs between the series capacitors, which will cause additional de-Q’ing.
The main benefit of using SOS substrate is that it is perfectly insulating with a
very high Quality factor. Typical HR SOI (high-resistivity SOI) substrates exhibit
higher parasitic capacitance with lower Quality factors for the parasitic substrate
network (Q = 15–30), which will impact especially high VPK and high-frequency
implementations on SOI.
The CMOS FET devices and MIM capacitors used in switched capacitors are
sensitive to electrostatic discharge (ESD) conditions with typical ratings in sev-
eral kV of human body model (HBM). This is much better than a typical GaAs
switch device with some hundreds of volts of HBM tolerance. However, the typi-
cal cellular device specification for any device connected directly to the antenna
port is 8 kV HBM and there are additional surface-mount device (SMD) assembly
process related machine model (MM) and charged device model (CDM) require-
ments for standalone components. Guaranteeing this level of ESD robustness not
only requires providing high-linearity low-capacitance ESD discharge paths on-
chip between all external terminals of the tunable device, but also making sure the
instantaneous voltage across any on-die MIM capacitor does not exceed the break-
down voltage (~50–100 V) of the dielectric. The switch FET stack is self-protecting
due to the drain-source breakdown of the FETs during ESD event, but ESD damage
on MIM capacitors usually results in short circuits in the dielectric near weak spots
in the dielectric layer. Fully integrated tunable matching networks have typically
less stringent ESD requirements due to less exposed RF ports and integrated shunt
inductors on RF ports.
RF-CMOS Impedance Tuners 151
1 L ωL 1 C R
Qser = = = Q par = R = = ωRC (7.1)
R C R ωRC L ωL
L2 RL2 C2 RC2
+ +
Port 1 L1 C1 L3 C3 Port 2
Z0 Z0
RL1 RC1 RL3 RC3
– –
While placing tunable elements near the high impedance areas of the tuning
network or antenna allow for wider impedance coverage, the downside is that the
RF peak voltage (Vpk) across the tuning element increases significantly. Switched
capacitor networks usually have a very strict limit on the maximum RF voltage. On
the other hand, placing the tunable element near low impedance area reduces the
peak voltage but, at the same time, increases the current element resulting in higher
power dissipation. Careful consideration is required to place the tunable components
in a location that maximizes their performance.
Additionally, tunable capacitors used in series configuration often experience
high RF voltage levels, due to voltage multiplication and resonant conditions. Using
elements in shunt configuration only is significantly easier but reduces frequency and
impedance coverage of the tuning network.
From a network circuit topology point of view, the lower the source or load port
impedance, the higher the current through the circuit elements. The power dissipa-
tion is proportional to the square of the current and, thus, increases rapidly for low
port impedances. This is especially critical for impedance transforming matching
networks often used to match amplifier devices to 50 Ω system impedance.
Regardless of the intrinsic Quality factor of the tunable component, the net-
work loss is determined by the mounted Q factor when the component is used in
the actual tuning circuit. Due to parasitic inductance, capacitance, and resistance,
the network Q is always much lower than the tunable element component Q at
GHz frequencies, as parasitic impedances have values comparable to the actual
impedances. Even more important is the actual total dissipative loss of the tunable
matching network when it is being used to match impedances. The overall loss is
composed of several different factors. Dissipated power loss of an individual com-
ponent can be calculated as half of the real part of the conjugate of complex volt-
age drop multiplied by the complex current (Equation 7.3). The power dissipation
for two ports is the difference between power entering port 1 and power leaving
port 2 (Equation 7.4).
Component real power dissipation:
1
Pcomp = R((V2 −V1 )* I ) (7.3)
2
1 1
Pnetwork = R(V1* I1 ) − R(V2* I 2 ) (7.4)
2 2
A Pareto chart can then be created showing the contribution of each element to the
total loss of the network. This usually highlights the importance of reducing sub-
strate loss effects, impact of finite inductor Q, and importance of minimizing EM
(electromagnetic) effects, such as current crowding, skin effect, spreading resistance,
proximity effect, and radiation loss.
154 Tunable RF Components and Circuits: Applications in Mobile Handsets
The best measurement tool for impedance tuning network S-parameters is the
VNA, even though some custom automation software usually needs to be written to
enable collection of hundreds of thousands or millions of states.
For large signal characterization of parameters such as harmonic and intermod-
ulation distortion (IMD), as well as peak voltage handling, the impedances and
capacitive and inductive parasitics of the tuning network and individual compo-
nents must be well known. Distortion produced by the tuning network cannot be
easily predicted from tunable capacitor harmonic and IMD specifications, because
nonlinear currents and voltages generated by tunable capacitors are significantly
impacted by the rest of the tuning network. The effective peak RF voltage seen by
the switch FET devices must not exceed their design and reliability limits under
typical operation of the tuning network. Depending on configuration (shunt, series)
and port impedance levels and any circuit resonances, the voltage across the FET
stack may be several times higher than effective RF voltage at the ports calculated
from the available RF power levels.
the form of eight real and imaginary components. The total number of complex
S-parameter points becomes 1.25 billion (300*8*220)! Even at a typical VNA sweep
times of 100 ms (including data transfer and storage overhead), it would still take
about 30 h to sweep across the whole tuning space. The number of measurements
could be significantly reduced by, for example, measuring every second or fourth
tuning state, however, narrow resonances may be missed completely.
Regular S-parameter files are stored in ASCII format, which for two-port files
amounts to about 128 bytes of data for every frequency point (16 bytes per float-
ing point number). The total file size would be over 40 GB for 1 million tuning
states and would take several hours to load into a circuit simulator, which is
prohibitively inefficient. A good alternative is to store the file in a binary format.
Not only does this reduce the file loading time to a few seconds, there is also no
waste because data are stored natively in the most efficient way. A typical dou-
ble-precision floating-point format (IEEE754) uses 8 bytes (64 bits) per number,
cutting the file size in half to 20 GB. This is still a very large file, considering
typical corporate e-mail accounts only allow sending and receiving a maximum
of 10 MB files.
Double-precision floating numbers are clearly still too inefficient. The size
of data can be further reduced to 5 GB by using fixed-point integer representa-
tion (i.e., casting floating point S-parameter real/imaginary data) (−1⋯+1) into
−32768⋯+32767 discrete integers such that it can be directly stored as a 16-bit (2
byte) signed integer.
Fixed-point representation only works for passive networks because the maxi-
mum S21 is bounded to 1 (0 dB). For active networks, such as amplifiers, a floating-
point representation is required as the value of S21 has no upper bound. The smallest
IEE754 standardized floating-point number is the half-float (2 bytes, 16 bits), which
is the same number of bits as the fixed-point integer representation, but with less
precision as 5 bits are used for the exponent.
1
Zin =
1 1 1
+ +
RL1 + jωL1 R − j 1 ⎛ 1 ⎞ 1
R + jωL2 ) + ⎜ RC 2 − j
ωC1 ( L 2 ⎟+
C1
⎝ ωC2 ⎠ 1 1 1
+ +
RL 3 + jωL3 R − j 1 Z0
C3
ωC3
(7.5)
The input reflection coefficient (S11) can be calculated from input impedance using.
Solving S11 directly results in an extremely complicated function with too many
terms to reproduce here. However, for the purposes of this analysis a numerical
solution will suffice, which is derived by assuming that all DTCs follow a linear
capacitance relationship Cn = An*xn + Bn and all scalar terms (inductor values Ln, dis-
sipative resistances R Ln and RCn, frequency (w in radians), and DTC slope An and
offset Bn) are set to a value of 1. Substituting these values into Equation 7.5, and
using Equation 7.6 to calculate the numerical input reflection coefficient S11 of net-
work shown in Figure 7.4 is
Closer analysis of Equation 7.7 reveals that input reflection coefficient S11 can be
modeled by a multivariate rational function having denominator and numerator com-
plex polynomial terms that contain all unique combinations (e.g., x1, x2, x3, x1x2, x1x3,
x2x3, x1x2x3) of products of the input variables and a complex constant, but no square
or cubic terms. These can be easily enumerated by treating the combinations as all
binary values of dimension n (0–2n). For example, 3D tuner has eight unique combi-
nations of input variables [x1, x2, x3] = [000, 001, 010, 011,…,111], where 1 means that
input variable exists and 0 means it does not, and 000 refers to the constant.
Similar analysis can be applied to all of the other S-parameters to derive a generic
3D fitting function as shown in Equation 7.8.
Generic complex fitting function for all S-parameters of network shown in
Figure 7.4 is
TABLE 7.1
Examples of Complex Coefficients Fitted to Measured 3D Data at 1.0 GHz
Coefficients S11 S12 S21 S22
v1 0.1006 − 0.7918i 0.0028 − 0.0054i 0.0028 − 0.0054i −0.7150 + 0.2131i
v2 −1.5556 − 1.6865i −0.8548 + 1.9055i −0.8548 + 1.9055i −1.8592 − 1.2029i
v3 −0.5650 + 0.2025i 0.0004 − 0.0009i 0.0004 − 0.0009i 0.0985 − 0.6019i
v4 0.0730 + 0.1356i 0.0004 − 0.0014i 0.0004 − 0.0014i 0.1439 + 0.0327i
v5 0.0467 + 0.0433i 0.0001 − 0.0001i 0.0001 − 0.0001i 0.0518 + 0.0328i
v6 0.1303 + 0.0593i −0.0003 − 0.0000i −0.0003 − 0.0000i 0.0921 + 0.1058i
v7 −0.0040 − 0.0033i 0.0000 − 0.0000i 0.0000 − 0.0000i −0.0044 − 0.0025i
v8 6.0340 + 4.1172i −1.1136 + 2.2268i −1.1136 + 2.2268i 4.5742 + 5.5984i
v9 −0.1486 + 0.7979i 0.1399 + 0.9053i 0.1399 + 0.9053i −0.0151 + 0.7800i
v10 −0.9739 + 2.8708i 0.0132 + 3.4458i 0.0132 + 3.4458i −0.4980 + 2.9187i
v11 −0.2579 + 0.5914i −0.0564 + 0.7325i −0.0564 + 0.7325i −0.1531 + 0.6052i
v12 −0.0667 − 0.1415i −0.1236 − 0.1270i −0.1236 − 0.1270i −0.0853 − 0.1253i
v13 −0.0449 − 0.0467i −0.0655 − 0.0334i −0.0655 − 0.0334i −0.0498 − 0.0379i
v14 −0.0666 − 0.1307i −0.1197 − 0.1164i −0.1197 − 0.1164i −0.0835 − 0.1156i
v15 0.0040 + 0.0035i 0.0055 + 0.0024i 0.0055 + 0.0024i 0.0043 + 0.0028i
v16 8.0640 − 1.5184i 8.0375 − 4.6416i 8.0375 − 4.6416i 7.3763 − 2.7998i
S11 @ 1.0 GHz S12 = S21 @ 1.0 GHz S22 @ 1.0 GHz
Measured Measured Measured
Modeled Modeled Modeled
FIGURE 7.5 Measured versus modeled S-parameter data for a 3D 4-bit impedance tuner
at 0.1 GHz.
(a) (b)
regular constellation. Care should be taken to properly select which DTC is mapped
to which axis such that the resulting graph conveys the information properly. For
higher dimensional tuners, some dimensions will need to be kept fixed while others
are mapped to a surface.
Contours of transducer gain drawn on a Smith chart are immensely useful in tunable
matching network analysis and optimization. Typical Smith chart constellations of
S11 and S22 only show points that can be conjugate matched, but they do not include
any information on the dissipative loss of that particular tuning state or information
about areas of Smith chart that are outside of the constellation. The transducer gain
160 Tunable RF Components and Circuits: Applications in Mobile Handsets
contour, on the other hand, shows the area of the Smith chart that can be matched
to the source with a given loss performance. For best performance, the load imped-
ance of, for example, antenna or filter input impedance should be pre-matched or
otherwise placed inside the low-loss contour. The difference between conjugate of
S22 constellation plot and load transducer gain contours is illustrated in Figure 7.7.
The coverage areas of the S22 constellation do not always overlap with reasonable
(e.g., −1 dB) loss contour.
Transducer gain contours can be created by a mathematical load-pull technique
using S-parameter data, which is valid for linear networks such as impedance tuners.
That is, represent every complex load impedance reflection coefficient on the Smith
chart as a grid of complex values (e.g., values −1 to 1 at 0.01 steps for both real and
imaginary parts for a total of 40,401 complex numbers). Then for every load reflec-
tion coefficient and every tuning state (e.g., 1,048,576 states for a 4D 5-bit tuner) cal-
culate transducer gain using Equation 7.9, and then pick the tuning state with highest
transducer gain. Now this data set can be used to draw contours of transducer gain
overlaid on top of the Smith chart. This technique works but given the brute force
nature it gets very inefficient for large number of tuning states or fine load reflec-
tion coefficient grid. In this example, 40 billion transducer gain function evaluations
(complex equation so computationally very expensive) would be required for every
frequency point.
A novel alternative technique is presented later that is several orders of mag-
nitude less computationally complex. It is based on calculating a transducer gain
circle (for a given contour level, e.g., −1.0 dB) for every tuning state and then graph-
ically tracing the outline of all the G T circles to create the contour (Figure 7.8).
The required calculations are shown in Equation 7.10, where g is the normalized
gain (relative to peak S21) given contour level P (in dB), c is the center of the circle
(in complex coordinates of reflection coefficient and r is the circle radius as an
absolute number).
Equations for calculating transducer gain circles are
2
10 P/10 *
gS22 1− g (1− S22 )
g= 2 2
c= 2
r= 2 (7.10)
S21 (1/(1− S22 )) 1− (1− g) S22 1− (1− g) S22
RF-CMOS Impedance Tuners 161
–1
–1
–1
–1
(a) (b)
FIGURE 7.8 Traditional brute-force transducer gain contour method (a) compared with a
gain circle boundary method (b).
The use of transducer gain circles removes the need to evaluate the transducer gain
function at every load reflection coefficient value, as the gain circle already contains
that information. In the earlier example, this reduces number of function evaluations
by a factor of over 40,000. An efficient sweep-line algorithm or alpha boundary can
then be used to trace the boundary of the gain circles to create the contour in loga-
rithmic time.
The transducer gain circle boundary method is a significant improvement over the
brute-force method but still inefficient as the number of function evaluations is pro-
portional to number of tuning states. This becomes very impractical for tuners that
have a large number of bits per DTC (e.g., 6–8 bit), because number of unique tuning
states grows exponentially with the number of bits per DTC. To completely eliminate
the dependency on number of bits per DTC, the transducer gain circle boundary
method can be combined with tuner compact model equations and coefficients. This
allows that transducer gain contours to be analytically calculated directly from the
complex rational function, without ever having to evaluate the complex function at
discrete DTC values to create discrete S-parameter data. This would provide prac-
tically instantaneous transducer gain contour calculation from the tuner compact
model coefficients, for any complexity.
–0.5 –0.5
Transducer gain (dB)
–1 –1
–1.5 –1.5
–2 –2
–2.5 –2.5
–3 –3
1200 1400 1600 1800 2000 2200 1200 1400 1600 1800 2000 2200
(a) Frequency (MHz) (b) Frequency (MHz)
FIGURE 7.9 Line plot (a) versus density plot (b) for a large number of tuning states.
7.4 CONCLUSIONS
Over recent years, the complexity of cellular devices has grown significantly with
the exponential growth in mobile data demand, which has had a significant impact
to the complexity of cellular RF front end. What once was a simple single-antenna
single-band transmit and receive duplexing solution with no active RF switching
has turned into a complex RF front end supporting 10+ FDD/TDD bands, 16+ RF
switch throws, multiple simultaneous carriers, reconfigurable multimode and multi-
band power amplifiers, and impedance and antenna tuning. Optimization of the per-
formance, size, and cost has practically displaced GaAs in the switching portions of
the RF front end by enabling monolithic integration of switching, tuning, and ampli-
fication functions using RF-CMOS SOI. While silicon varactor, BST, and MEMS-
based alternatives exist for tuning applications, switched CMOS capacitors exhibit
a unique trade-off between RF performance, simplicity, and monolithic integration.
To the first order the RF performance is dominated by the RONCOFF figure-of-merit
of the CMOS switch process, but the peak voltage, impedance tuning range, and
dissipative loss behavior is significantly impacted by second order parasitic effects.
A careful optimization of all parameters is required to achieve best performance.
While RF tuning improves the radiated performance of the mobile device, it relies
on being able to select the one optimal tuning state out of thousands or millions of
possible candidates. Traditional RF circuit simulation, measurement, and optimiza-
tion techniques have to be replaced by size and speed optimized data structures,
analysis, and measurement techniques.
REFERENCES
1. Ranta, T., Ella, J., and Pohjonen, H., Antenna switch linearity requirements for GSM/
WCDMA mobile phone front-ends, 8th European Conference on Wireless Technology
Proceedings, Paris, France, pp. 23–26, October 3–4, 2005.
2. Ranta, T. and Novak, R., New tunable technology for mobile-TV antennas, MWJ, 2008.
3. Ranta, T. and Novak, R., Antenna tuning approach aids cellular handsets, MW& RF,
2008.
RF-CMOS Impedance Tuners 163
4. Ranta, T. and Novak, R., Improve mobile handset antenna performance with new tun-
ing techniques, EE Times, 2008.
5. Ranta, T., Pilgrim, D., and Whatley, R., RF front end adapts for increased mobile data
demand, EE Times, 2010.
6. Whatley, R., Ranta, T., and Kelly, D., RF front-end tunability for LTE handset applica-
tions, Compound Semiconductor Integrated Circuit Symposium (CSICS), 2010 IEEE,
pp. 1–4, October 3–6, 2010.
7. Whatley, R., Ranta, T., and Kelly, D., CMOS based tunable matching networks for cel-
lular handset applications, Microwave Symposium Digest (MTT), 2011 IEEE MTT-S
International, pp. 1–4, June 5–10, 2011.
8. Baxter, B., Ranta, T., Facchini, M., Jung, D., and Kelly, D., The state-of-the-art in
silicon-on-sapphire components for antenna tuning, Microwave Symposium Digest
(IMS), 2013 IEEE MTT-S International, pp. 1–4, June 2–7, 2013.
9. Cheng, C., Facchini, M., Ranta, T., and Whatley, R., High performance 1.8–2.4
GHz phase shifter using silicon-on-sapphire digitally tunable capacitors, Microwave
Symposium Digest (IMS), 2013 IEEE MTT-S International, pp. 1–3, June 2–7, 2013.
10. Ranta, T., Whatley, R., Cheng, C., and Facchini, M., Next-generation CMOS-on-
insulator multi-element network for broadband antenna tuning, Microwave Conference
(EuMC), 2013 European, pp. 1567–1570, October 6–10, 2013.
11. Sekar, V., Cheng, C., Whatley, R., Zeng, C., Genc, A., Ranta, T., and Rotella, F.,
Comparison of substrate effects in sapphire, trap-rich and high resistivity silicon sub-
strates for RF-SOI applications, 2015 IEEE 15th Topical Meeting on Silicon Monolithic
Integrated Circuits in RF Systems (SiRF), pp. 37–39, January 26–28, 2015.
8 Handset Antenna
Tuning Using BST
Paul McIntosh
CONTENTS
8.1 Overview........................................................................................................ 165
8.2 Antenna Matching and the Transition to Tunable Matching Networks........ 165
8.3 Antenna Measurements for Tunable Matching Network Design.................. 168
8.4 Tunable Matching Network Topologies and Circuit Optimization............... 170
8.5 Implementation of PTIC Antenna Tuners in a Mobile Phone....................... 174
8.6 Example of a PTIC Tuner Design.................................................................. 175
8.6.1 Circuit Analysis Results.................................................................... 178
8.7 PTIC Antenna Tuner Deployment................................................................. 185
8.8 Conclusions.................................................................................................... 189
References............................................................................................................... 189
8.1 OVERVIEW
This chapter will review barium strontium titanate (BST)-based handset antenna
tuning, which has become increasingly important for the today’s multiband, m
ultimode,
sleek, and ubiquitous smartphone. As the smartphone screen size has increased, the
battery size has increased to support larger and brighter screens along with more and
more functionality; hence, the volume available for the antenna(s) has been reduced.
This chapter will review feedpoint antenna tuning that uses BST-based tunable
capacitors in the impedance matching network at the feedpoint of the antenna,
and this chapter will also present a design approach and case study. Antenna feed-
point tuning is becoming increasingly common and is conceptually simple, as it
replaces the fixed match that most handsets employ with a circuit that uses tunable
components.
165
166 Tunable RF Components and Circuits: Applications in Mobile Handsets
Antenna
C2 L1
Matching network
matching networks until recently. However, with the proliferation of bands that are
now used, covering frequencies from as low as 698 MHz to as high as 2.7 GHz rou-
tinely and even as high as 3.5 GHz,1 the performance of a matching network comprised
of fixed components can significantly limit the overall performance of the handset. An
example of a matching network using fixed components is shown in Figure 8.1.
A matching network designer using fixed components must make compromises
during the synthesis by trading off performance at some frequencies in order to
achieve acceptable performance over the whole band. Practically, this is achieved
using a range of techniques, including hand adjustments and optimization using a
vector network analyzer (VNA), impedance matching design using the Smith chart,
and computer-aided design (CAD) using software packages such as Optenni Lab,2
Advanced Design System (ADS),3 and Microwave Office.4 Clearly, once selected the
performance is fixed and cannot be adjusted at all to accommodate any changes in
the antenna’s scattering parameters (S-parameters) due to external loading. Figure 8.2
illustrates the frequency response of a fixed matching network where the performance
has been centered at 800 MHz. The performance cannot be adjusted to improve the
voltage standing wave ratio (VSWR) at either 700 or 900 MHz, nor can the behavior
be modified to compensate for the loading effects of the hand or head that occur
during real-world use.
Tunable antenna matching networks, commonly referred to as antenna tuners,
are not new structures, having been commonly used by amateur radio enthusiasts
for many years. In the amateur radio application, tunable matching networks have
been realized using air variable capacitors and switched inductors. In the case of
cellular handsets though, the lack of suitably high-performance tunable radio fre-
quency (RF) components has limited the use of antenna tuners. There has been a
wide range of work published on antenna tuners, including circuits that use BST-
based capacitors,5 switched capacitor structures,6 and varactor diodes,7 but until
recently, antenna tuners were mostly confined to the published literature. The pub-
lished results are illustrative of the potential capabilities but do not often address
performance in real handsets.
However, three important factors have occurred together leading to antenna tun-
ers now becoming a standard feature in handsets. These are the emergence of high-
performance tunable elements, such as BST-based tunable capacitors, the increasing
frequency range used for handsets, particularly the extension down to 698 MHz band
for fourth-generation Long Term Evolution services (4 G LTE), and the consumer
demand for slim handsets where little volume is available for the internal antenna.
These factors combined have led to handset design teams embracing antenna tuners
in an ever increasing number of commercially available handsets.
The matching network shown previously in Figure 8.1 can be modified to become
a tunable network by simply replacing the fixed capacitors with tunable capacitors,
as shown in Figure 8.3. Conceptually, this is a simple change with BST-based tunable
capacitors, or Passive Tunable Integrated Circuits (PTICs™), being used along with
the addition of a bias controller. This bias controller is used to supply the required
bias voltage to the PTIC and also interfaces with the handset baseband IC, which
controls the operation of the phone including the entire RF chain.
As a result of the replacement of the fixed capacitors with PTICs, the network
response, shown in Figure 8.4, is now variable and can be adjusted to compensate
Antenna
C2 L1
Input from
L2 C1 LANT
transceiver Bias
controller
Tunable matching
network
VSWR
Tuning Tuning
applied applied
for the effect of hand or head loading as well as being tuned to the frequency of
operation. It is also important to note that the frequency response of the BST tunable
capacitor–based network can be tuned in an analog fashion, due to the continuously
variable nature of the PTICs. This, along with the fact that PTICs are analogous to
fixed caps meaning that they can be used as single devices in shunt or series, allows
for tremendous flexibility in the matching network.
SMA connector
Antenna spring
contacts
Micro-coax cable
additional phase shift and loss that is unaccounted for in the design, which will have
a detrimental effect on the performance of the network. It is recommended that a
micro-coaxial semirigid type cable be used, which can be easily routed within the
phone, and the outer jacket of the cable can be soldered to suitable grounding points
along its length, as shown in Figure 8.5.
The cable should be routed away from the antenna to reduce its impact on the
antenna behavior and the S-parameter measurements. The measurements can be
conducted using an automated VNA with port extension capabilities such as those
offered by Rhode and Schwarz8 or Keysight.9 Using the port extension function with
the cable either grounded or open at the antenna feedpoint, that is, the spring contact
in this particular example, the effect of the cable will be accounted for. The phone
is now ready for antenna S-parameter characterization in the freespace use case and
other conditions as required. Figure 8.6 shows the type of phantom that can be used
in the characterization of the antenna in other use cases, to replicate head loading
and hand-loading conditions.
It is also worth noting that this same handset, with the cable installed, can be
used for antenna efficiency measurements. These measurements allow the radiation
performance, known as radiation efficiency of the antenna, to be established and
when combined with the two-port analysis of the matching network, gives the total
performance of the antenna subsystem. Similar measurements can be made across
the entire frequency range of the handset and when cascaded with known amplifier
output power and/or receiver sensitivity in each band provides a good estimate of the
performance of the handset.
The results of the antenna measurement can be plotted on the Smith chart and,
along with the efficiency measurements, provide a good overall picture of the antenna
behavior. The S-parameters for an example antenna measured under three different
use cases—freespace, phantom hand (PH) grip, and for contrast a human hand grip
are shown in Figure 8.7. The S-parameters for each of these use cases can be used
in the design of a tunable matching circuit if desired. Figure 8.7 shows that while a
PH does provide some loading and causes a shift in the S-parameters, the effect of
the human hand is considerably greater. This illustrates two very important points.
First, characterization under real grip conditions, when the hand use case is being
Human hand
grip
Freespace
Phantom hand
FIGURE 8.7 Example antenna S-parameters under multiple use cases (low band on the left,
high band on the right).
considered, is important to establish true loading effects. Second, the impact of the
user’s hand can be quite large. In the case of a fixed-component matching network,
the antenna engineer will have to select a use case to design for. This may lead to
them compromising the other use case(s), or if they attempt to design for multiple
use cases, it is likely that they will end up compromising on performance over all
operating conditions. The benefit of a tunable matching network, where the network
response can be tuned to accommodate the operation of the handset under a variety
of use cases, is conceptually clear.
ANT1
jX1
jX3 jX2
ANT1
jX1 jX2
jX3
ANT1 ANT1
C2 C1 L1 C2 L2
L2
C3 L3 C3 C1 L1
L3
suitable when the antenna is not heavily mismatched (<4 dB mismatch). In this case,
the introduction of a third stage would cause additional loss for very little additional
improvement in matching coverage. In the case of a strongly mismatched antenna this
situation may be reversed; the additional loss of a third stage could be tolerated in
order to achieve better matching performance. Useful examples of two PTIC match-
ing circuits are shown, one previously in Figure 8.3 and one that follows in Figure 8.11.
172 Tunable RF Components and Circuits: Applications in Mobile Handsets
ANT1 ANT1
C2 L2 L4 C2 L2
L4
L1 C1 L3 L1 C3 L3
Since there are several versions of matching circuit that can be used, a computer-
aided simulation is needed in order to optimize and compare the individual circuits.
The general process is to analyze each matching circuit, or tuner, and then determine
the performance of the system containing that matching circuit compared with the
performance of direct connection between the antenna and the system impedance,
typically 50 Ω. This will allow the benefit of the tuner to be established. Comparing
the relative benefit will allow the best circuit to be selected. The improvement
that the tuner provides can be summarized as shown in Equation 8.1 from Gu and
Morris.12 This is a calculation of the relative transducer gain (RTG) and is a measure
of the improvement that the tuner yields when compared to the performance without
a tuner:
When the power wave definition of S-parameters is used with arbitrary complex
input and output termination impedances, then the transducer gain can be rep-
resented as shown in Equation 8.2.13 A useful description of the application of
power waves to the design of conjugate matching impedances can also be found
in Rahola:14
2
GT = S21 (8.2)
In the analysis of a given matching circuit, the return loss or input match, are often
examined and optimized either by hand on a VNA, or using a circuit simulator.
However, in the case of a real circuit with loss from the matching elements, this
approach will not lead to the optimum performance as it ignores the dissipative loss
of the matching network components. Instead, the transducer power gain for each
circuit should be evaluated to ensure that the circuit has the optimum performance.15
The loss of the matching network is composed of the dissipative loss of the compo-
nents, such as inductor and capacitor Q, and the mismatch loss due to non-perfect
impedance matching. The optimization of each circuit should then focus on target-
ing maximum |S21|2 in each band of operation.16 By using this parameter as the main
optimization variable, it will ensure that circuits in which the return loss is better,
but the total loss is higher are avoided. Matching networks with good return loss
results do not necessarily yield optimum performance since the goal of the matching
Handset Antenna Tuning Using BST 173
network is to achieve maximum power transfer to and from the antenna. By comput-
ing the transducer power gain, the best schematic values can be selected.
The optimization process will need to set all of the matching components as
variables and use |S21|2 as the optimization criteria. This optimization process will
need to include all of the frequencies that will be used in the handset if those fre-
quencies are for duplex or simplex operation. In the case of time division duplex
systems (TDD) such as GSM or TDD-LTE, the matching network settings can be
optimized for each transmit and receive frequency. For example, in the GSM 900
band, the transmit frequency of 880 MHz can be optimized independently from the
receive frequency of 925 MHz since transmit and receive operations occur in dif-
ferent time slots. In the case of a PTIC-based matching network, the PTICs can be
tuned to particular values for the transmit operation and then changed for the receive
operation. This reconfiguration will ensure that the optimal performance is obtained
at both frequencies. In the case of frequency division duplex (FDD) systems such as
FDD-LTE or UMTS, the transmit and receive frequencies are paired. The matching
network must operate simultaneously at both frequencies. For example, in Band 1
UMTS (WCDMA) the transmit frequency is 1920 MHz and the receive frequency is
2110 MHz. The optimization of the matching network must be carried out to accom-
modate the transmit/receive frequency spacing of nearly 200 MHz. The PTIC bias
voltages that are used for the transmit frequency must be the same as those used for
the receive frequency. The PTIC bias voltages can be adjusted for optimum perfor-
mance when a different channel is used in the handset, since this means that a new
transmit/receive frequency pair will be used.
The number of PTIC bias voltages used within a band can be as simple or as
complex as desired and the decision on how many to use is part of the optimization
process. In the case of FDD operation, there can be a single setting for each capaci-
tor for each band. However, this may lead to suboptimal performance so a better
approach is to subdivide a given band into two or three sets of subchannels. Using
this approach and three subbands, the matching network will have a different set of
PTIC bias voltages for the lower, middle, and upper third of the band. An example
is shown in Table 8.1.
The use of a circuit simulator is required in order to optimize and compare the
performance of each topology that is examined. There are a wide range of choices
that can be used in which the designer can build their own simulation setup, such as
AWR’s Microwave Office simulator, Ansoft’s Designer suite, Keysight’s (formally
Agilent) Advance Design System, and free packages, such as the open source Quite
TABLE 8.1
Example of Subband PTIC Voltages
Band Mode Channel Range (Tx/Rx) PTIC Settings (Two PTIC Tuner)
I UMTS FDD 9,612–9,703/10,562–10,653 V1, V2
I UMTS FDD 9,704–9,796/10,654–10,745 V3, V4
I UMTS FDD 9,797–9,888/10,747–10,838 V5, V6
174 Tunable RF Components and Circuits: Applications in Mobile Handsets
Universal Circuit Simulator (QUCS) and simulators specifically designed for match-
ing network design such as Optenni Lab. In all cases, it is important to use accurate
antenna S-parameters and realistic models of the fixed elements, such as inductors
and capacitors as well as the PTICs. A data file of PTIC S-parameters can be used
in order to reduce the burden on the designer to extract their own models. These are
available from ON Semiconductor Corp. For the inductors, the manufacturer, Murata
or Coilcraft, for example, can provide accurate device models. Optenni Lab offers a
very simple interface to the simulator where the user can import PTIC S-parameters
and specify the number of elements required in the matching network. After import-
ing the antenna S-parameters and specifying the frequency bands that the network
has to cover, the simulator will automatically optimize the topology with the selected
number of elements.
The simplest optimization process will use a single set of antenna S-parameters,
such as the freespace use case, but additional use cases can also be included to ensure
that the matching network performs adequately under a range of different conditions.
The optimization process can also be constrained so that component values, such as
inductors, remain within a usable range for handset matching circuit application. By
accurately modeling the component losses, the overall performance of the antenna
system (antenna and matching network) can be determined. The total efficiency can
be obtained by cascading the antenna radiation efficiency with the matching network
transducer gain when the network is terminated by the antenna’s S-parameters.
Depending on how many PTICs are being used within a phone, a cost-effective
implementation can be achieved. Since all the control and digital interface circuitry
is centralized, removing any duplication of digital functions, the solution can be
kept to the most efficient implementation. A good example of this is when PTICs
are deployed in matching circuits used for both main and diversity antennas. The
design of the diversity antenna can be just as difficult as the main antenna and, in
some cases, more difficult due to even more challenging space constraints. If we
consider an example where a single PTIC matching network is used for the diversity
antenna and a two PTIC matching network is used for the main antenna, both net-
works can easily be controlled using a three-output controller. There are minimal
special requirements for the DC bias lines, making routing more straightforward.
Since the digital and control functions are in a single IC, they can be kept away
from the antennas to ensure there is no risk of any spurious noise affecting antenna
performance.
In a more complex handset where there are two antennas both used for transmit
and receive, the solution can be further scaled with minimal additional circuitry
needed. In this case, three PTICs could be used in each matching circuit so a six
output control IC is used. The same general approach is used for all designs with the
central control interface communicating with the baseband processor and provid-
ing the DC bias control to each of the PTICs in the matching networks. As it can
be seen, the use of small, DC biased tunable capacitors, combined with the PTIC
controller allows for the support of many different types of architecture. These
PTIC matching networks can be easily deployed at the port of each antenna that
requires tuning.
S(1,1)
S(1,1)
(a) Frequency (b) Frequency
FIGURE 8.12 Antenna S-parameter—S11: (a) 0.6–0.96 GHz and (b) 1.7–2.7 GHz.
0 0
–2
–2 –4
–6
S11 (dB)
S11 (dB)
–4 –8
–10
–6 –12
–14
–8 –16
–18
–10 –20
6e08 7e08 8e08 9e08 1e09 1.7e09 1.9e09 2.1e09 2.3e09 2.5e09 2.7e09
Frequency (Hz) Frequency (Hz)
0 0
Mismatch loss (dB)
–4 –4
–6 –6
–8 –8
6e08 7e08 8e08 9e08 1e09 1.7e09 1.9e09 2.1e09 2.3e09 2.5e09 2.7e09
Frequency (Hz) Frequency (Hz)
0 0
Mismatch loss (dB)
–4 –4
–6 –6
–8 –8
6e08 7e08 8e08 9e08 1e09 1.7e09 1.9e09 2.1e09 2.3e09 2.5e09 2.7e09
Frequency (Hz) Frequency (Hz)
input match of the high band is not compromised too much, but still provide a good
match at low frequencies where the antenna mismatch is high.
In this design example, the antenna behavior under a use case condition will also
be considered. The antenna’s response will alter under loading conditions caused
by the user’s head or hand and a tunable matching network can be adjusted to com-
pensate for these effects. A fixed matching circuit does not have any flexibility so
the designer will either design for a particular use case, typically freespace, or a
compromise between freespace and the various loading conditions. In this simple
example, the beside head use case will be considered. Figure 8.15 shows the mis-
match performance of the antenna under the beside head hand right (BHHR) load-
ing condition.
As can be seen in Figure 8.15, the loading effect of the beside head hand use case
keeps the mismatch loss at 700 MHz to approximately 4 dB and at 960 MHz to more
than 4 dB. Both of these are still significant mismatches and the high VSWR makes
the design of the matching network important to keep the insertion loss as low as
possible. Any loss in the matching network will reduce the output power of the phone
and result in a direct reduction in the total radiated power (TRP) and total isotropic
sensitivity (TIS). The high band has a very similar performance to freespace and is
relatively unchanged by the impact of the PH. As expected, the loading effect is more
prominent in the low bands.
178 Tunable RF Components and Circuits: Applications in Mobile Handsets
The first step in designing the matching network is to set the frequency bands
that will be used in the handset. There are a large number of bands that a handset
needs to address and for the purpose of this design the following bands will be con-
sidered: bands 7, 12, 13, 17, and 40 for LTE operation and bands 1, 2, 3, 4, 5, and 8
for WCDMA operation. These bands will also cover the four GSM bands, GSM850,
900, 1800, and 1900. Operation in either LTE or WCDMA mode will require duplex
settings for the PTICs, but the GSM bands can be optimized for either simplex oper-
ation, with separate settings for transmit and receive operation, or duplex operation.
For simplicity in this example, the GSM bands are also optimized for duplex opera-
tion. The user will also need to decide if the bands will be subdivided at all. This
provides finer control within a band and can yield improved performance. In this
example, the bands will be divided into three with PTIC settings for the low, middle,
and high channels. However, in practice, it may be that some of the tuner states are
sufficient to cover an entire band.
The terminating impedances also need to be defined for the optimization. In
the general case, the antenna impedance file(s) (touchstone format.s1p) are used as
the load impedance(s) and the source impedance can either be 50 Ω or the result
of load-pull measurements of the transceiver. Using load-pull measurements, if
available, can be beneficial since the output of the transceiver when measured at
input of the matching circuit may not be 50 Ω. In the case of source impedances
that are varying over frequency, an additional .s1p file can be specified and used
instead of 50 Ω. For example, the Optenni Lab software allows the user to import
an impedance file for use as the source terminations. In this example, 50 Ω will be
used for simplicity.
Finally, the goals of the simulation need to be established. As discussed previ-
ously, it is important to optimize the circuit values for maximum transducer gain,
or minimum insertion loss, rather than return loss. Optimizing for return loss may
result in a good, but lossy, match being selected that will not give the maximum
power transfer. There may also be the need to weight the performance of a particu-
lar band, such as applying greater weight to the home network bands rather than
the roaming bands, which are used only rarely. This can be achieved by having a
different |S21|2 goal for a particular band, such as specifying |S21|2 < −1 dB for the
home bands and |S21|2 < −2 dB for roaming bands. Once the design criteria have been
established, the simulation/optimization of different topologies can be undertaken in
a simulator of the designer’s choice.
ANT
9.5 nH
2.7 pF 39 nH
In this circuit design, the PTICs are simulated using S-parameter blocks, which
represent the full performance of the devices over frequency and bias. The inductors
are modeled with a Q of 60 at 1 GHz, which represents the performance of com-
monly available RF wirewound-type devices. The maximum capacitance value of
the PTICs is specified at 2 V bias. The device can be continuously tuned from 2 to
24 V with a typical tuning range of 4.5:1. An example of the PTIC capacitance versus
voltage characteristic is shown in Figure 8.17.
In order to understand the benefit of the tunable matching circuit, the perfor-
mance can be examined in a couple of different ways using the RTG method as pre-
viously reviewed. In the first instance, the RTG can be calculated by comparing the
results with the matching network in operation against no matching network. This
will allow the improvement in mismatch loss (MML) to be clearly seen. Even though
7
Capacitance (pF)
6
5
Capacitance (pF)
0
0 2 4 6 8 10 12 14 16 18 20 22 24
Voltage (V)
TABLE 8.2
Tuner Design Case Study Data in the 700 MHz LTE Bands for the FS Condition
Tuner Relative
Frequency MML with MML without Insertion Transducer
(MHz) Band/Mode Channel Tuner (dB) Tuner (dB) Loss (dB) Gain (dB)
698 12/LTE L-Tx −0.4 −5.8 −1.6 4.2
716 12/LTE H-Tx 0.0 −5.4 −1.0 4.4
728 12/LTE L-Rx −0.4 −5.1 −1.1 4.0
746 12/LTE H-Rx −0.8 −4.4 −1.6 2.8
777 13/LTE L-Tx −1.2 −3.3 −1.6 1.6
787 13/LTE H-Tx −1.1 −2.9 −1.5 1.4
746 13/LTE L-Rx −2.0 −4.4 −2.4 2.1
756 13/LTE H-Rx −1.8 −4.1 −2.1 2.0
the tunable matching circuit is lossy, the improvement in the mismatch outweighs the
dissipative loss, thus improving the power transfer of the network and the efficiency
of the entire antenna subsystem (antenna and matching circuitry). The initial data for
this comparison used are the freespace performance. Table 8.2 shows the data for the
LTE bands 12 and 13 that cover from 698 MHz, the lowest frequency of band 12, to
787 MHz, which is the highest frequency of band 13. In each case, the low and high
channel results are shown for illustration. RTG as discussed in Equation 8.1 earlier, is
the calculation of the improvement that the tuner provides over no matching network.
As the results show in Table 8.2, the loss due to mismatch has been consider-
ably reduced with improvements of over 4 dB in some channels, and the overall
benefit is close to 3 dB on average in the 700 MHz LTE bands. The return loss
performance of the tuner under various settings is of some interest since it allows
a designer to quickly see how well the antenna tuner is matching the input imped-
ance. Figure 8.18 shows the return loss performance of the tuner under a number
of different states across the entire frequency band of operation. Each return loss
curve represents a different voltage setting on the PTICs. Since an analog voltage
is used to bias each PTIC, optimum voltages are not constrained by capacitor steps
and can be adjusted as desired to ensure the optimal performance from the net-
work. Note that in Figure 8.18, only a small number of return loss curves are shown
for illustration. This tuning capability leads to the easy adjustment of the PTIC
values based on the handset use case. The use case can be determined either by
triggering a proximity detector if the handset is held to the user’s head or by using
a more sophisticated method where mismatch detection is employed.18 The tuner
can then be adjusted to ensure that the mismatch is reduced to as low as possible.
In the case of head–hand loading, the mismatch loss as shown in Figure 8.15 is
still over 4 dB at 700 MHz, which is slightly better than the freespace value. Since
the mismatch loss is still high and also because there has been a frequency shift in
the antenna performance due to the loading, the tunable matching network needs
to be readjusted. This is carried out by changing the bias voltages and the resulting
performance when compared with the no tuner condition is shown in Table 8.3.
Handset Antenna Tuning Using BST 181
–2
–4
–6
–8
S11 (dB)
–10
–12
–14
–16
–18
–20
7e08 7.5e08 8e08 8.5e08 9e08 9.5e08 1e09
Frequency (Hz)
TABLE 8.3
Tuner Design Case Study Data in the 700 MHz LTE Bands for the
BHHR Condition
Tuner Relative
Frequency MML with MML without Insertion Transducer
(MHz) Band/Mode Channel Tuner (dB) Tuner (dB) Loss (dB) Gain (dB)
698 12/LTE L-Tx −0.1 −4.2 −0.8 3.4
716 12/LTE H-Tx −0.9 −3.6 −1.5 2.1
728 12/LTE L-Rx −1.0 −3.2 −1.6 1.7
746 12/LTE H-Rx −1.0 −2.6 −1.3 1.3
777 13/LTE L-Tx −0.1 −1.6 −0.4 1.2
787 13/LTE H-Tx −0.1 −1.3 −0.4 0.8
746 13/LTE L-Rx −0.7 −2.6 −1.0 1.7
756 13/LTE H-Rx −0.4 −2.3 −0.7 1.6
The results in Table 8.3 show that the benefit of using the tuner is now close to
3.5 dB in some cases, and almost 2 dB on average in these 700 MHz LTE bands.
This is a considerable increase in link margin and will improve handset performance
in the field.
In the previous discussion, the RTG was computed using the performance with
and without an antenna tuner. This is a common approach, but a more stringent
and realistic comparison is to look at the performance of a fixed match under the
182 Tunable RF Components and Circuits: Applications in Mobile Handsets
same operating conditions. This will allow a full understanding of the benefit
of the variable tuner over a fixed match under the two use cases that have been
examined in this section. The design of a fixed matching network is a subjective
exercise since a designer is faced with a trade-off of whether to design for a single
use case, such as freespace; a more realistic condition, such as hand or head-and-
hand loading; or something in between. In many cases, the simple freespace use
case is used as it requires no special test equipment and can be done relatively
quickly by hand using a VNA. Of course, this is a non-optimum approach and
will lead to lower performance than could be obtained at some frequencies and
use cases. In this example, the commercially available Optenni Lab simulator
will be used along with the freespace S-parameters. A four-element network will
be considered and the result of the automatic optimization process is shown in
Figure 8.19.
The automatic optimization carried out using the commercially available simula-
tor provides the user with the options to create all permutations of the four-element
match, or simply output the best result based on the efficiency as calculated in each
band. The results shown here are for the final selected circuit and the performance
has been analyzed using the same Q values for the inductors as used in the tunable
network and typical high Q capacitors values of around 200. The results for the
700 MHz LTE bands have been tabulated in the same way that they were in the tun-
able network case and are shown in Table 8.4 for freespace operation.
Comparing the results for freespace operation between the fixed matching circuit
and the tunable circuit in the 700 MHz frequency range, a few conclusions can be
drawn. Looking at each of the transmit and receive channels, the tunable match-
ing circuit outperforms the fixed match in every transmit and receive channel. This
is an expected result since the tuner can be adjusted on a band-by-band and even
channel-by-channel basis. The fixed-component matching network is by definition
unchanging. It is also worth noting that in Band 12, which also includes the impor-
tant Band 17, which is a subset of Band 12 used in North America and other coun-
tries, the improvement that using the antenna tuner yields is up to 1.3 dB. This can
be simply achieved by using PTICs and associated control components.
ANT
Tx/Rx 3.9 pF
2 nH
11 nH
1 pF
TABLE 8.4
Fixed Match Design Case Study Data in the 700 MHz LTE Bands for the
Freespace Condition
MML Match Relative
Frequency MML with without Insertion Transducer
(MHz) Band/Mode Channel Match (dB) Match (dB) Loss (dB) Gain (dB)
698 12/LTE L-Tx −1.5 −5.8 −2.4 3.4
716 12/LTE H-Tx −1.5 −5.4 −2.3 3.1
728 12/LTE L-Rx −1.7 −5.1 −2.3 2.8
746 12/LTE H-Rx −1.7 −4.4 −2.2 2.2
777 13/LTE L-Tx −1.9 −3.3 −2.2 1.0
787 13/LTE H-Tx −2.3 −2.9 −2.1 0.9
746 13/LTE L-Rx −1.7 −4.4 −2.2 2.2
756 13/LTE H-Rx −1.9 −4.1 −2.3 1.8
TABLE 8.5
Fixed Match Design Case Study Data in the 700 MHz LTE Bands for the
BHHR Use Case Condition
MML Match Relative
Frequency MML with without Insertion Transducer
(MHz) Band/Mode Channel Match (dB) Match (dB) Loss (dB) Gain (dB)
698 12/LTE L-Tx −0.9 −4.2 −1.5 2.7
716 12/LTE H-Tx −1.1 −3.6 −1.5 2.1
728 12/LTE L-Rx −1.4 −3.2 −1.7 1.5
746 12/LTE H-Rx −1.6 −2.6 −1.8 0.8
777 13/LTE L-Tx −1.9 −1.6 −2.0 −0.5
787 13/LTE H-Tx −2.2 −1.3 −2.0 −0.8
746 13/LTE L-Rx −1.6 −2.6 −1.8 0.8
756 13/LTE H-Rx −1.8 −2.3 −1.9 0.3
Under the BHHR use case, the same fixed matching network has been used, but
the loaded S-parameters have been used instead of the freespace S-parameters for
the efficiency and RTG calculations. Table 8.5 summarizes the results for the fixed
match when operating under the BHHR use case condition and is shown later.
Looking at these results it can be seen that at some frequencies, the use of a fixed
matching network actually increases the mismatch loss over a direct connection, and
when compared with the tunable network, the antenna tuner provides up to 1.7 dB
of improvement at some frequencies. The results show that tunable match provides
better performance at every frequency considered here in the 700 MHz band, which
is one of the most challenging bands to design for.
184 Tunable RF Components and Circuits: Applications in Mobile Handsets
–4
–8
Return loss (dB)
–24
0.68 0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96
Frequency (GHz)
–4
–8
Return loss (dB)
–12
FIGURE 8.20 Return loss behavior for an antenna tuner and fixed matching network.
Handset Antenna Tuning Using BST 185
The results of the fixed match and the antenna tuner should also be compared
fully across all of the frequencies of operation to get a full picture of the benefits
and operation. This can be done most effectively by looking at the insertion loss of
the fixed match and comparing it to the tunable match. It is also somewhat helpful
to look at the return loss as well since this is a measure that many designers are
more comfortable with, although this parameter would not be used to optimize
the matching network for the reasons previously discussed. Figure 8.20 shows
the return loss over frequency in both the low and mid/high bands for freespace
operation. The antenna tuner results are shown in discrete bands since this is
more applicable given that the bias voltages are adjusted for each band and set of
subchannels. The data show that the tunable circuit outperforms the fixed match-
ing network in virtually every band and channel and in some cases by quite a wide
margin.
Figure 8.21 shows the companion charts for the insertion loss performance that
show a similar trend with the antenna tuner outperforming the fixed circuit. However,
in both the return loss and insertion loss charts, it can be seen that there are a very
small number of channels where the fixed matching circuit offers slightly better
performance than the tunable circuit. This is expected since it is natural to assume
that the fixed matching circuit can be tuned to be optimal at some frequencies, how-
ever, at the vast majority of channels/bands, the antenna tuner offers superior perfor-
mance. The results are similar for the BHHR use case and the insertion loss results
for this case are shown in Figure 8.22. The tunable matching circuit outperforms the
fixed circuit everywhere except for a few channels.
The performance of the antenna tuner will have a direct impact on the perfor-
mance of the handset, and the improvement in the region of 2 dB shown here in the
antenna subsystem will lead to a direct improvement in TRP, for example. This, in
turn, leads to an improved link margin, which will lead to better coverage and higher
data throughput. As shown in this example, there are very few channels where the
fixed match can actually provide better performance than a tunable network, and
even in those cases, the delta is low. The benefit of using PTICs in a tunable matching
network over the traditional fixed-component approach is clear.
–0.5
–1
–1.5
Insertion loss (dB)
–2
–2.5
Fixed match insertion loss
–3 Band 12 tuner
Band 17 tuner
Band 13 tuner
–3.5
Band 5 tuner
Band 8 tuner
–4
0.68 0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96
Frequency (GHz)
–0.5
–1
–1.5
Insertion loss (dB)
–2
–2.5
Fixed match insertion loss
Band 7 tuner
–3 Band 40 tuner
Band 1 tuner
Band 2 tuner
–3.5 Band 3 tuner
Band 4 tuner
–4
1.71 1.81 1.91 2.01 2.11 2.21 2.31 2.41 2.51 2.61 2.71
Frequency (GHz)
FIGURE 8.21 Insertion loss behavior for an antenna tuner and fixed matching network:
Freespace.
Handset Antenna Tuning Using BST 187
–1 Band 8 tuner
–1.5
–2
–2.5
0.68 0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96
Frequency (GHz)
–0.5
–1
–1.5
Insertion loss (dB)
–2
–2.5
Fixed match insertion loss
Band 7 tuner
–3 Band 40 tuner
Band 1 tuner
Band 2 tuner
–3.5 Band 3 tuner
Band 4 tuner
–4
1.71 1.81 1.91 2.01 2.11 2.21 2.31 2.41 2.51 2.61 2.71
Frequency (GHz)
FIGURE 8.22 Insertion loss behavior for an antenna tuner and fixed matching network:
BHHR.
188 Tunable RF Components and Circuits: Applications in Mobile Handsets
PTIC PTIC
Sense IC
control control
Baseband Baseband
FIGURE 8.24 Block diagram comparing open- and closed-loop antenna tuner systems.
for each band/channel and use case are predetermined and stored in lookup tables,
which are then used by the handset during operation on the relevant channel, and
so on. Figure 8.24 shows simplified block diagrams for both open- and closed-loop
antenna matching systems. Open-loop tuners are only able to adjust to the use cases
that were preset when the design work was undertaken. In contrast, PTIC closed-
loop antenna tuning systems use a feedback mechanism to continuously adjust
the matching circuit to ensure the lowest possible insertion loss of the matching
network. This results in the maximum power transfer and the optimal performance
of the handset. Closed-loop antenna tuner systems have the big advantage that they
can more easily adapt to the changing environment of the phone and antenna. The
system uses a measurement IC that allows the impedance to be measured, which,
when combined with the known S-parameters of the tuner, allows the optimum
capacitance values to be selected irrespective of the use case that the phone is
in. The design of the actual tunable matching network for a closed-loop match-
ing system still follows the same methodology as outlined earlier. The difference
between open and closed-loop operation lies in the additional components used to
control the tuner. PTICs are ideally suited to operate in either open- or closed-loop
implementations and are currently being used in both configurations in a number
of commercially shipping handsets.
Handset Antenna Tuning Using BST 189
8.8 CONCLUSIONS
This chapter has reviewed how the wide frequency range of operation of the
antenna(s) in a modern handset usually results in mismatch-limited performance.
Typically, this mismatch occurs at the band edges and in the 700–960 low-frequency
bands, but can also occur in the mid- and high-frequency bands. Additionally,
the loading caused by use cases, such as head-and-hand or hand, can detune the
antenna, making the design of a fixed match almost impossible to optimize over all
frequencies and use cases, resulting in suboptimal performance. The use of PTICs
to replace fixed capacitors results in a tunable network with considerable flexibility.
The design of a tunable network can accommodate different use cases and should be
optimized for insertion loss rather than return loss to ensure optimal performance.
The tuner networks can utilize one, two, or three tunable elements, and the exact
selection will depend on how strongly mismatched a given antenna is, the required
bands and bandwidth needed, and the insertion loss that can be tolerated. The tun-
able network can be adjusted for each use case and used in either an open-loop or
closed-loop antenna tuner systems. When compared with a fixed match, the tunable
network will provide significant improvements in virtually all bands/channels and
use cases, and this improvement will lead to improved performance in the field. The
improvements that can be achieved using an antenna tuner with PTICs require very
little additional circuitry beyond the tunable capacitors themselves and a central
control IC. The use of a central control IC allows for a very flexible approach to
antenna tuning and multiple antennas can be tuned using a single controller. Many
handsets are already shipping with both open- and closed-loop PTIC antenna tun-
ers, and the technology is now moving from the high-end smartphone segment into
the mass-market segment showing that the adoption is now common and a standard
handset feature.
REFERENCES
1. 3rd Generation Partnership Project Specifications, www.3gpp.org.
2. Optenni Lab Matching Network software, Optenni Ltd., Espoo, Finland, www.optenni.
com.
3. Advanced Design System by Keysight Technologies, www.keysight.com.
4. Microwave Office by National Instruments AWR, www.ni.com/rf/awr/.
5. Chen, L., Forse, R., Chase, D., and York, R., Analog tunable matching network using
integrated thin film BST capacitors, IEEE MTT-S International Microwave Symposium
Digest, June 2004, Fort Worth, TX, pp. 261–624.
6. Dongjiang, Q., Yu, Z., Tsaipi, H., Kimball, D. et al., Antenna impedance mismatch mea-
surements and correction for adaptive CDMA transceivers, IEEE MTT-S International
Microwave Symposium Digest, June 2005, Atlanta, GA, pp. 783–786.
7. Ali, S. and Payandehjoo, K., Tunable antenna techniques for compact handset applica-
tions, IET Microwaves, Antennas and Propagation, 8(6), 401–408, April 2014.
8. Rhode and Schwarz Vector Network Analyzers, www.rhode-schwarz.com.
9. Keysight Technologies’ Vector Network Analyzers, www.keysight.com.
10. Yip, P., High Frequency Circuit Design and Measurements, 1st edn., Chapman & Hall,
London, UK, 1990.
190 Tunable RF Components and Circuits: Applications in Mobile Handsets
11. Schmidt, M., Lourandakis, E., Leidl, A., Seitz, S., and Weigel, R., A comparison of
tunable ferroelectric Pi and T-matching networks, Proceedings of the 37th European
Microwave Conference, 2007, Munich, Germany, pp. 99–101.
12. Gu, Q. and Morris, A., A new method for matching network adaptive control, IEEE
Transactions on Microwave Theory and Techniques, 61(1), 587–595, January 2013.
13. Collin, R.E., Foundations for Microwave Engineering, 2nd edn., McGraw-Hill,
New York, 1992.
14. Rahola, J., Power waves and conjugate matching, IEEE Transactions on Circuits and
Systems—II: Express Briefs, 55(1), 92–96, January 2008.
15. Rahola, J., Estimating the performance of matching circuits for antennas, Proceedings
of the Fifth European Conference on Antennas and Propagation, Barcelona, Spain,
2010, pp. 1–3.
16. Rahola, J., Optimization of matching circuits for antennas, Proceedings of Fifth
European Conference on Antennas and Propagation, 2011, Rome, Italy, pp. 776–778.
17. Tunable Capacitor Control IC Datasheet—TCC-106, www.onsemi.com.
18. Boyle, K.R., de Jongh, E., Sato, S., Bakker, T., and van Bezooijen, A., A self contained
adaptive antenna tuner for mobile phones, Sixth European Conference on Antennas
and Propagation, March 2012, Prague, Czech Republic, pp. 1804–1808.
9 Aperture-Tunable
Antennas
Handset OEM Perspective
Ping Shi
CONTENTS
9.1 Introduction................................................................................................... 191
9.2 Tunable Antenna Design Challenges............................................................. 192
9.3 Survey of Tunable Antenna Technology........................................................ 195
9.3.1 Tunable Impedance Matching........................................................... 195
9.3.2 Aperture-Tunable Antenna................................................................ 197
9.4 Design and Verification Methods.................................................................. 197
9.4.1 Selection of Antenna and Tuning Components................................. 198
9.4.2 Tunable Antenna Design.................................................................... 199
9.4.3 Simulation of Harmonics and Intermodulation from Tunable
Antennas�������������������������������������������������������������������������������������������200
9.4.4 Measurement of Harmonics and Intermodulation from Tunable
Antennas............................................................................................202
9.5 Conclusions....................................................................................................204
References...............................................................................................................204
9.1 INTRODUCTION
Development of mobile communications has drastically changed the handset antenna
design landscape. The handset antenna design target has changed from gain and
radiation pattern to system efficiency and better user experience for multiband and
multimode applications. The traditional approach in passive multiband antenna
design is to trade antenna efficiency with bandwidth. Antenna tuning was proposed
to address the multiband requirements. There are two ways to tune the antenna for
multiple band operation: impedance tuning and aperture tuning. Antenna tuning
has become practical in handsets with recent advancements in radio frequency (RF)
tuning components: varactors, MEMS tunable capacitor, pin-diodes, RF switches,
and so on.
The adoption of LTE-Advanced1 presents a new set of requirements to tunable
antenna technology. In this chapter, tunable antenna design challenges will be pre-
sented first, followed by a survey of tunable antenna technology. Tunable antenna
design and verification methods will be discussed next. A summary of antenna tun-
ing will conclude this chapter.
191
192 Tunable RF Components and Circuits: Applications in Mobile Handsets
TABLE 9.1
LTE Interband CA Classes
Carrier
Aggregation Class Definition Examples
Al Low-band–high-band combination without harmonic B5_B2
relation between bands
A2 Low-band–high-band combination with harmonic relation B12_B4, B17_B4,
between bands B8_B3
A3 Low-band–low-band or high-band–high-band combination B12_B5, B1_B3
without intermodulation problem (low-order IM)
A4 Low-band–low-band or high-band–high-band combination B8_B20
with intermodulation problem (low-order IM)
A5 Combination except for Al–A4
Aperture-Tunable Antennas 193
Harmonic Harmonic
TX/Rx filter Low filter
antenna Low B17 Rx
band
bands
B17 TX
Switch
Duplexer
RF test
connector High
Diplexer band B4 Rx
High
Switch B4 TX
bands
Duplexer
Harmonic Harmonic
TX/Rx filter Low filter
antenna Low band B8 Rx
bands
B8 TX
Switch
Duplexer
RF test
connector High
Diplexer band B1 Rx
High
B1 TX
bands Switch
B3 Rx
B3 TX
Quadplexer
B4 DL
2110 2155
Harmonic Harmonic
TX/PRx filter Low filter
antenna Low band B17 Rx
bands
RF test
connector Switch B17 TX
Diplexer
Duplexer
Tuner High
bands High
band B4 Rx
B4 TX
Switch
Conducted path
Radiated path Duplexer
Harmonic
DRx filter Low
antenna Low B17 Rx
bands band
Switch
Tuner RF test
connector High
band B4 Rx
Diplexer
High
bands Switch
For handset antenna design, traditional passive antenna designs work well due
to their passive and high linearity nature. For a tunable antenna design, tuning
components are placed close to or on the antenna. The generated TX harmonic
or intermodulation product will couple into the receiver chain through two differ-
ent paths: conducted path to the PRx (Primary R) chain and radiated path to DRx
(Diversity R), as shown in Figure 9.4. For both paths, filtering cannot be used to
suppress the harmonic or intermodulation interference falling into RX band. The
generated harmonic or intermodulation product has to be small enough to cause no
desense to either the primary or the diversity receiver. This requirement presents a
serious challenge in the design of a tunable antenna solution.
L
L1
C L
C2 L2
C
C1
9.3.2 Aperture-Tunable Antenna
Tunable impedance matching does not change antenna aperture; the system effi-
ciency of the antenna with tunable impedance matching will not exceed the overall
radiation efficiency. For a handset antenna design, the antenna is much shorter than
the wavelength at low-band frequencies. The radiation efficiency drops quickly as
it approaches the small antenna limit. This is where tunable matching techniques
start to be less useful. On the other side, with aperture tuning, one can change the
antenna radiation aperture and reshape the BW efficiency curve. With aperture tun-
ing, antenna could be designed to have a narrow bandwidth with high efficiency
while keeping the same volume, or to keep the same efficiency with reduced antenna
volume.
With the wide variety of tuning components now available, aperture tuning meth-
ods are being aggressively explored for popular types of antenna designs. Planar
inverted F antenna (PIFA) antennas were popular when PCB reuse between circuit
and antenna was desired and the thickness of phone was not a concern. Aperture
tuning of a PIFA antenna includes tuning the radiation arm length using pin-diodes8
or loading PIFA antennas at different locations using tunable capacitors.9 Monopole
or IFA antennas are popular antennas for their high efficiency and broad band-
width. Aperture tuning of monopole or IFA antennas includes extending/shorting
the antenna with pin-diodes or tunable capacitors, loading the antenna with capaci-
tance,10 changing the shorting arms’ length or shorting location.
By decoupling the high-band radiating element from the low-band radiat-
ing element and aperture tuning the low-band element, the aperture-tunable
antenna technique can be applied to support LTE CA with a high-band–low-band
combination (A1/A2 classes), and high–high combination (A3 class). Pi-shape
IFA antennas and segmented IFA antenna are among the good candidates for this
type of application. To support LTE CA with low-band–low-band combination,
the antenna needs to support dual low-band resonance. Metamaterial inspired
antennas are good candidates.11 This type of antenna provides two low-band reso-
nances: one from the normal radiation mode (L–C) and another from the left-hand
modes (C–L), as illustrated in Figure 9.7.
Aperture tuning is applied to tune the left-hand resonance mode at the coupling
(equivalent C1) or the grounding (L2). Similar designs have been used by several
handset manufactures.
C
L C C L
1 2
C3
It is desirable to have a passive wideband antenna to support all LTE bands and
CA combinations. However, there are cases where antenna design is limited by ID,
for example, metal ID such as metal ring, metal back IDs. In these cases, a tun-
able antenna is typically used to provide the same functionality with less volume.
To meet requirements from LTE CA, a tunable antenna has to support much wider
instantaneous bandwidth, thus imposing new limitation on the antenna design and
performance.
digital control interface like RF-MIPI or SPI/I2C. For MEMS tunable capacitors,
internal charge pumps are implemented to drive each element capacitor to high or
low-capacitance values. These two types of tunable capacitors have some parasitic to
ground and require ground/power/digital interface to control the capacitance values.
These requirements could limit their application in certain aperture-tunable antenna
designs where ground/power/digital interface is not available or parasitic to ground
should be extremely low. Typically, digital tunable capacitor provides a minimum of
16 states with a tuning range about 4:1 or higher. A higher tuning range with lower
parasitic capacitance is preferred as it enables “universal matching network” and
provides flexibility in aperture tuning.
Radio frequency switches make inductance tunable. There are two topologies
for RF switch: serial switch or serial shunt switch. Serial shunt RF switches can
provide extra isolation between ports at the cost of shunting another port to ground.
The serial configuration switch is the most popular RF switch topology for antenna
tuning. RF switches are widely used in aperture tuning, where switches can be used
to extend or shorten radiator length, to create a shorting arm to ground, or to change
the grounding inductance.
The capacitance tuning range (and tuning ratio) and Q (Quality factor) are
among the most important electrical parameters for tunable capacitors. RF switch
structure, on-resistance and off-capacitance determine how the switch is used
for antenna tuning. To have a high linearity antenna, higher IIP2, IIP3, voltage
handling capability, and power handling capability are essential for all tunable
components.
Special attention should be paid to power handling and voltage handling capa-
bility during component selection. Power handling capability is usually specified
in a 50 ohm environment. However, the switches or capacitors used in impedance
matching or aperture tuning are not in a 50 ohm environment. Voltage handling
capability is more useful in a non-50 ohm environment. Different component sup-
pliers specify voltage handling capability differently: some of them use rms voltage,
while others use peak voltage. This voltage is also dependent on the technology, for
example, complementary metal oxide semiconductor (CMOS) and pseudomorphic
high electron mobility transistor (PHMET) will have different breakdown mecha-
nism; while MEMS capacitors have the pull-in voltage and pull-out voltage con-
siderations. These differences should be considered during the tuning component
selection process in order to decide between different technologies and different
vendors.
be optimized for the raw antenna impedance. The antenna and matching network
design is an iterative process.
For aperture-tunable antennas, the tuning component must be included in the ini-
tial design and simulation. It will be desirable to have the tuning component or circuit
in the initial mock-up antenna. Often the tuning components or circuits are not avail-
able when needed for one reason or another. It is a common practice to use a passive
component to simulate the tuning component in the mock-up design with RLC mod-
els provided by the vendors. The same technique is applied for antenna simulation,
where a simple RLC model of device is good enough for the initial design. In the
initial design, the return loss and antenna efficiency are among the major design tar-
gets. Next, antenna performance over different use cases such as beside head, beside
head and hand, various hand positions should be considered to validate the antenna
performance. Specific absorption rate/hearing aid compatibility (SAR/HAC) should
also be evaluated during this phase. Any failure to achieve required performance
specifications for the earlier items will trigger another iteration of the design.
For tunable antenna design, antenna linearity must be evaluated before antenna
design is locked down. Traditionally, the linearity concern is focused on spurious
emissions. Residual spurious emissions are specified by individual government
agencies such as FCC(U.S.)/Industry Canada or EC, as well as by standard organiza-
tions like the European Telecommunications Standards Institute (ETSI) and Cellular
Telecommunication Industry Association (CTIA). Compared to the radiated spuri-
ous emissions (RSE) requirements, the harmonic desense to receiving bands in LTE
CA is more challenging. In a 50-ohm system, this is equivalent to have the imped-
ance tuner with IIP3 over 81 dBm. When the system impedance is not 50 ohm, the
voltage across the component will be higher and correspondingly the requirement to
IIP3 will be higher. It is necessary to simulate or measure actual generated harmonic
or intermodulation to validate antenna design.
Antenna simulation could be performed in a frequency domain and time domain.
Typically, simulation methods could be FDTD, FEM, or a hybrid. All of the earlier
mentioned methods terminate simulation based on the convergence of simulation.
For example, if the difference between two simulations is less than a threshold, the
program will determine that the simulation reaches the convergence and will terminate
the simulation. The actual termination criteria varies, with typically −30 to −40 dBc.
A typical IIP3 of tuning components is much better and the generated harmonic is
far below −40 dBc. Any nonlinear contribution generated from the tuning compo-
nent cannot be simulated by antenna simulator. A joint simulation between antenna
simulator and circuit simulator is required to obtain the harmonic or intermodulation
product level.
Vin
1 2 1 Ref
Ref
P_1Tone X1P
PORT1 S2P XNP2
Num = 1 SNP1
Z = 50 ohm
P = dbmtow (RFpower)
Freq = RFfreq
Vin 1 2
Vout
3 Ref
R
R1
P_1Tone S3P R = 50 ohm
PORT1 SNP1
Num = 1 Ref 1
Z = 50 ohm
P = dbmtow(RFpower) X1P
Freq = RFfreq XNP2
FIGURE 9.9 Model of a main/diversity antenna system with one tuning component.
Spectrum 3 dB
analyzer 1 pad
@f2 @f1
Spectrum DUT
3 dB
analyzer 2 pad Main
Diversity antenna
antenna
Anechoic chamber
Isolator Directional
LPF coupler
Signal
PA 3 dB
generator pad
Forward Reverse
@f0
LPF power power
Power 3 dB
meter pad
Spectrum 3 dB
analyzer 1 pad
@f0
DUT
Spectrum 3 dB
analyzer 2 pad Main
Notch Diversity antenna
antenna
Anechoic chamber
9.5 CONCLUSIONS
This chapter addresses two important aspects of a tunable antenna design simultane-
ous bandwidth and linearity. With the rapidly growing commercial success of the
tuning component industry, it is expected that tuning components with extraordinary
high IIP3/IIP2 and Q will emerge in the near future. High tuning ratio, low para-
sitic tunable capacitor technology will enable true universal matching network and
high(er) order matching networks. Closed-loop tuning solutions based on these types
of capacitors will capture and increase the amount of market share. Aperture-tunable
methods will be explored for more antenna types. It is also expected that aperture-
tunable antenna designs will demonstrate their value in the vast majority of future
handset products.
REFERENCES
1. 3GPP, http://www.3gpp.org.
2. Sun, Y. and Fidler, J.K., Practical considerations of impedance matching network, HF
Radio Systems and Techniques, 1994, York, UK, pp. 229–233.
3. Sun, Y. and Fidler, J.K., Design method for impedance matching networks, IEE
Proceedings on Circuits, Devices and Systems, 1996, pp. 186–194.
4. Boyle, K., Bakker, T., de Jongh, M., and van Bezooijen, A., Real-time adaptation of
mobile antenna impedance matching, Antennas and Propagation Conference (LAPC),
2010, Loughborough, UK, pp. 22–25.
5. van Bezooijen, A., de Jongh, M.A., van Straten, F., Mahmoudi, R., and van Roermund, A.,
Adaptive impedance-matching techniques for controlling L networks, IEEE Transactions
on Circuits and Systems, 57(2), February 2010, pp. 495–505.
6. Gu, Q. and Morris, III, A.S., A new method for matching network adaptive control,
IEEE Transactions on MTT, 61(1), January 2013, pp. 587–595.
7. Yarman, B.S., Design of Ultra Wideband Antenna Matching Networks: Via Simplified
Real Frequency Technique, Springer, 2008.
8. Komulainen, M., Berg, M., Jantunen, H., Salonen, E.T., and Free, C., A frequency
tuning method for a planar inverted-F antenna, IEEE Transactions on Antenna and
Propagation, 56(4), April 2008, pp. 944–950.
9. Panayi, P.K., Al-Nuaimi, M., and Ivrissimtzis, L.P., Tuning techniques for the planar
inverted-F antenna, IEE National Conference on Antennas and Propagation, 1999,
York, UK, pp. 259–262.
10. Tornatta, P.A. and Gaddi, R., Aperture tuned antennas for 3G-4G applications using
MEMS digital variable capacitor, IEEE MTT-S International Microwave Symposium
Digest (IMS), 2013, Seattle, WA, pp. 1–4.
11. Lopez, N., Lee, C.J., Gummalla, A., and Achour, M., Compact metamaterial antenna
array for long term evolution (LTE) handset application, IEEE International Workshop
on Antenna Technology, iWAT 2009, 2009, Santa Monica, CA, pp. 1–4.
10 Power Amplifier
Envelope Tracking
Jeremy Hendy and Gerard Wimpenny
CONTENTS
10.1 What Is Envelope Tracking?..........................................................................206
10.2 Generating the Envelope Reference Signal...................................................208
10.2.1 Envelope Path Signal Processing.....................................................208
10.2.2 Timing Alignment...........................................................................209
10.2.3 Upsampling...................................................................................... 210
10.2.4 Magnitude Calculation.................................................................... 210
10.2.5 Pre-LUT Gain Adjustment.............................................................. 210
10.2.6 ET Shaping Table............................................................................. 211
10.2.7 Post-LUT Digital Gain/Offset Correction....................................... 211
10.2.8 Pre-DAC Upsampling...................................................................... 214
10.2.9 ET DAC Requirements.................................................................... 214
10.2.10 Post-DAC Voltage Buffering and Filtering...................................... 214
10.3 Modeling the ET PA...................................................................................... 215
10.3.1 Characterizing the ET PA................................................................ 215
10.3.2 The 3D Surface Model of the ET PA............................................... 217
10.3.3 Noise and Distortion Mechanisms in ET........................................ 218
10.4 ET Shaping Table........................................................................................... 220
10.4.1 Influence of the Shaping Table........................................................ 220
10.4.2 Isogain Linearization....................................................................... 221
10.4.3 Optimizing the PA Phase Response for ET..................................... 223
10.4.4 Approaches to Shaping Table Design..............................................224
10.5 ET Supply Modulator.................................................................................... 225
10.5.1 SMPS-Only ETIC Architectures..................................................... 225
10.5.2 Hybrid ETIC Architectures............................................................. 227
10.5.3 ETIC Figures of Merit..................................................................... 228
10.5.4 ETIC Bandwidth.............................................................................. 228
10.5.5 ETIC Efficiency............................................................................... 228
10.5.6 Low Noise........................................................................................ 229
10.5.7 Slew Rate......................................................................................... 229
10.5.8 Output Impedance........................................................................... 230
10.5.9 “AC Boost” versus “DC Boost”....................................................... 230
10.5.10 ET/APT Mode Transitions............................................................... 231
10.6 How ET Affects the PA Design..................................................................... 231
10.6.1 PA Supply Decoupling..................................................................... 232
205
206 Tunable RF Components and Circuits: Applications in Mobile Handsets
Fixed
Dissipated as
supply
heat
voltage
DC/DC
Conventional PA converter
Baseband/ Power
RF up converter amplifier
Transmitted
ET
power Dissipated as
supply Supply voltage heat
Envelope signal
Baseband/ Power
RF up converter amplifier
70
Envelope tracking PA
ET PA instantaneous
average efficiency trajectory
60 efficiency
Instantaneous PA efficiency (%)
50
4.5 V
DC supply PA 4.0 V
40 average
efficiency 3.5 V
3.0 V
30 2.5 V
2.0 V
20 1.5 V
ET supply
DC supply PA DC supply
10 instantaneous Average Peak
efficiency trajectory output output
power power
0 PAPR
15 20 25 30 35
Instantaneous PA output power (dBm)
significantly increasing the average PA efficiency when used with a high PAPR
waveform.
To implement, ET requires the following hardware components:
Magnitude A
calculator
MIPI eTrak
envelope
Upsample interface
Modem chipset C
and filter
(Baseband or RFIC) B
I
I-DAC
Duplexer
Fine
Delay DPD
balancing (option)
gain
control
∑
Q-DAC To
Power amplifier
To
Q Rxr
antenna
Depending on the architecture of the modem chipset, the envelope path logic may be
implemented in either the baseband IC or RFIC. Architectures with a digital BB/RF
interface will usually implement the ET function in the RFIC, and architectures with
an analog BB/RF interface will usually implement the ET function in the BBIC, that
is, the ET logic and DAC are in the same chip as the TX I/Q DACs.
The envelope processing function is relatively small, occupying 50k–100k gates
of digital logic and consuming around 10 mW in a typical digital process.
IQ or ET path data, whereas fine-grain timing will require a fractional delay fil-
ter (FDF)—usually implemented on the ET path. The OpenET Alliance reference
implementation uses a four-tap farrow filter (FIR) for 20 MHz LTE applications.
Although the eTrak envelope reference signal at point “A” must be generated
slightly in advance of the RF signal at point “B,” it is not always the case that the
delay is added to the IQ path. If the RF path delay from IQ to PA input exceeds the
ET path delay, the coarse-grain delay FIFO must be implemented into the ET path,
rather than the IQ path.
The designer must fully consider all potential delays due to digital signal pro-
cessing pipelining and analog filters in both ET and RF paths, which may also vary
depending on channel bandwidth settings if there are switchable filters in the RF or
ET paths.
The analog ETIC propagation delay will be specific to the ETIC design and may
perhaps also vary with channel bandwidth setting if the ETIC includes switchable
filters. As a guideline, around 10 ns of delay is typical for a handset ETIC.
The ET/RF timing alignment may require some form of temperature compensa-
tion, as the group delay of the ET and RF path filters is likely to vary by more than
0.5 ns over temperature extremes. Production-line calibration of timing alignment
is also likely to be required, since the system timing alignment will be a function of
both the RFIC and ETIC, as well as potential variations in the PA itself.
10.2.3 Upsampling
Prior to calculating the magnitude of the IQ signal, the signal must be upsampled
and filtered to provide sufficient bandwidth for the magnitude calculation, which is
an often overlooked “pinch point” in terms of the sample rate. LTE IQ signals are
normally generated at 1.536× the channel bandwidth (e.g., 30.72 MSPS for a 20 MHz
channel). The magnitude calculation of the unshaped envelope must be performed at
a higher sample rate to avoid aliasing, requiring interpolation of at least 4× to avoid
unwanted distortion—6× (184.32 MSPS) or 8× (245.76 MSPS) are better choices.
10.2.4 Magnitude Calculation
Calculating the magnitude of the IQ signal (i.e., the length of the vector specified by the
IQ coordinates) is normally performed by a CORDIC function, which approximates
the Cartesian-to-polar calculation without needing any hardware multipliers. CORDIC
is an iterative algorithm, which in practice can be implemented as a hardware pipeline
of n stages. Although the rule-of-thumb is that one CORDIC iteration is required for
each equivalent bit of precision, simulations with LTE signals have shown that only
eight iterations are required before the CORDIC becomes an insignificant noise source.
in the analog RF path, for example, where a variable gain amplifier is used to attenu-
ate the RF drive level to the PA input as the average power control level is changed.
This is illustrated in Figures 10.4 and 10.5, which show the envelope waveform as
the average power level is backed off from 0 dBm at the PA input to −9 dBm at the
PA input, for example by adding a 9 dB analog attenuation step into the transceiver
RF path. It can be seen that simply scaling the ET output waveform after the shap-
ing table would not give the correct supply voltage—the magnitude signal must be
scaled before the shaping table, to ensure that the ET supply voltage remains cor-
rectly aligned with the RF drive level at the PA input.
This gain adjustment mechanism can also be used at production calibration
to compensate for analog gain errors in the RF path; if closed-loop DPD is not
employed, the RF and ET path gains must be matched to within ±0.25 dB for
20 MHz LTE.
2.0 V 2.0 V
1.0 V 1.0 V
0.0 V 0.0 V
–30 dBm –20 dBm –10 dBm 0 dBm 10 dBm
Time->
2.0 V
ET_P
ET_N 1.5 V
1.0 V
<-Time
0.5 V
0.0 V
–30 dBm –20 dBm –10 dBm 0 dBm 10 dBm Differential envelope interface
Instantaneous PA RF input power
2.0 V 2.0 V
1.0 V 1.0 V
0.0 V 0.0 V
Power Amplifier Envelope Tracking
1.0 V
<-Time
0.5 V
0.0 V
–30 dBm –20 dBm –10 dBm 0 dBm 10 dBm Differential envelope interface
Instantaneous PA RF input power
The nominal voltage gain and offset (output voltage with zero differential input)
of the ETIC is specified by each manufacturer and is not standardized.
The ET path gain and offset vary from part to part due to analog variation in
the ET DAC, reconstruction filter, and ETIC; since the overall contribution will be
a function of both the modem chipset and ETIC, this requires trim or calibration
at production time. Depending on the behavior of the ETIC and modem chipset
over temperature, some temperature compensation of these parameters may also be
required in “mission mode.”
The final interface to the ETIC is a differential voltage-mode signal; the MIPI
eTrak specification provides for three different classes with different maximum
swing ranges and common-mode voltage ranges.
10.3.1 Characterizing the ET PA
As we have seen, the “standalone” performance of ET PAs cannot be measured
unless the shaping table is first defined. This requires measurement of the PA’s fun-
damental characteristics (output power, efficiency, gain, and phase) over the full
range of instantaneous supply voltages and input powers. There are several methods
for ET PA characterization, summarized in Table 10.1.
In principle, the characterization could be carried out using a CW network ana-
lyzer and a variable DC supply, but results are typically poor due to thermal effects,
ranging errors, and drifts in phase measurements. It is also far too slow to allow
load-pull techniques to be used.
216
TABLE 10.1
ET PA Measurement Methods
Supply
Test PA Current Supply Bandwidth ET Efficiency ET Linearity Parameters
Methodology Measurement Impedance Requirements Prediction Prediction Measured
Swept CW Bench PSU Low (decoupling Low (bench Poor, due to PA Poor, due to PA Gain (AM:AM),
testing capacitor) PSU) die heating die heating efficiency
Pulsed RF/DC Instrumentation grade Low (decoupling Low (bench Good, if short Fair Gain (AM:AM),
testing current probe, ~5 μs capacitor) PSU) pulses (~10 μs, efficiency
resolution 10% duty
cycle)
Dynamic Challenging—high Requires low High (~60 MHz V. Good V. Good Gain (AM:AM),
supply BW with high impedance BW) phase
modulation common mode dynamic supply (AM:PM),
voltage current sense (no decoupling) efficiency
Tunable RF Components and Circuits: Applications in Mobile Handsets
Power Amplifier Envelope Tracking 217
FIGURE 10.7 Alternative views of the ET PA gain surface; the ET shaping table is high-
lighted in black.
The color map (Z axis) represents gain, with the contours of equal (iso) gain high-
lighted, and the selected shaping table again highlighted in black.
Figure 10.8 shows gain, phase, and efficiency surfaces side-by-side, with the tra-
ditional view at the top, and the shaping table view of the surface below.
The 3D “quasi-static” surface model of the PA can also be used in a simulation
environment with the selected shaping table—by applying a sequence of IQ samples
for a defined waveform at a selected mean power level at the PA input, the model
can be used to accurately predict ACLR, EVM and mean efficiency at a system level
across a wide range of waveforms and power levels, provided memory effects are
sufficiently well controlled.
(at the expense of gain). This technique can be used in the shaping table to control
the desired AM:AM characteristic of the PA, and hence the linearity.
When the ET PA is operating in compression/saturation at high instantaneous
power, increasing the RF input amplitude does not provide any more output power—
that is the definition of saturation. The only way to get more output power from a PA
in saturation is to increase its supply voltage. In other words, the output amplitude is
directly controlled by the supply voltage.
The profile of the ET shaping table controls how the supply voltage varies with PA
input signal amplitude. The AM:AM distortion of the PA (i.e., how its output signal
amplitude changes as a function of the input signal amplitude) is, therefore, directly
controlled by the shaping table. A “kink” in the shaping table will give a correspond-
ing “kink” in the AM:AM response and degrade linearity.
Phase is not controllable by the supply voltage in the same way as the gain of the
PA, but the phase response (AM:PM distortion) of the PA in ET mode can certainly
be characterized, and to some extent controlled, via the shaping table—but this also
depends on the design of the PA bias and matching networks.
The ability to directly control and trade-off gain and efficiency, and indirectly
control phase, allows the shaping table to control a wide range of PA performance
characteristics, as listed in Table 10.2.
10.4.2 Isogain Linearization
One obvious ET shaping table is the maximum efficiency shaping, where the sup-
ply voltage is selected to maximize the instantaneous efficiency of the PA at all
signal levels. However, with this approach, the PA gain is usually nonlinear, result-
ing in significant AM:AM distortion as shown in the left side of Figure 10.9. This
response would require correction with digital predistortion (DPD) to achieve the
system linearity requirements. However, the main disadvantage of this approach
is that high-bandwidth predistortion would be needed to effectively correct the
sharp “kinks” in the AM:AM characteristic. In addition to the baseband processing
TABLE 10.2
System Parameters Controlled by the Shaping Table
Primary PA Performance Parameters— Secondary System Performance Parameters—
Directly Controlled by the Shaping Table Influenced by the Shaping Table
PA efficiency ETIC efficiency
AM:AM linearity AM:PM linearity
Average PA gain Envelope signal bandwidth
Waveform PAPR at PA output
PA noise transfer/supply sensitivity
PA peak and minimum supply voltages
Isogain power control range
PA supply port impedance/linearity
70 70
222
50 50
4.5 V 4.5 V
4.0 V 4.0 V
40 40 3.5 V
3.5 V
3.0 V
30 3.0 V 30 2.5 V
Efficiency (%)
Efficiency (%)
2.5 V 2.0 V
20 20 1.5 V
2.0 V ISO26 dB
1.5 V
10 10
opt eff
0 0
15 20 25 30 35 15 20 25 30 35
Pout (dBm) Pout (dBm)
30 30
28 28
26 26
4.5 V 4.5 V
4.0 V 4.0 V
Gain (dB)
Gain (dB)
24 3.5 V 24 3.5 V Envelope tracking 26 dB
3.0 V Envelope tracking 3.0 V Isogain shaping
2.5 V Optimum efficiency shaping 2.5 V
22 2.0 V 22 2.0 V
1.5 V
opt eff 1.5 V
ISO26 dB
20 20
15 20 25 30 35 15 20 25 30 35
Pout (dBm) Pout (dBm)
Tunable RF Components and Circuits: Applications in Mobile Handsets
FIGURE 10.9 PA efficiency: Optimum efficiency (left) versus isogain shaping (right).
Power Amplifier Envelope Tracking 223
implications of the required high bandwidth, it also restricts the ability to filter out
RF path noise.
An alternative mapping of particular interest is “Isogain” shaping in which the
instantaneous supply voltage is chosen to achieve a particular constant PA gain, as
illustrated on the right-hand column of Figure 10.9.
With isogain shaping, the ET PA system achieves low AM:AM distortion, while
still operating in compression over much of the modulation cycle. For typical GaAs
LTE handset PAs, the efficiency penalty for using isogain shaping rather than max
efficiency shaping is often as low as 1% or 2% points, but the ACLR can be 7–10 dB
better.
Depending on the PA design, it is possible to use isogain shaping to linearize
the PA down to ACLR values of −45 dBc or even better without requiring any form
of DPD.
It is important to note that there are an infinite number of potential isogain shap-
ing tables to choose from—selecting the best isogain value is a trade-off between
efficiency, linearity, and the average power control range over which the isogain
remains valid.
To maintain the same isogain over a range of power control levels requires an
isogain value which maximizes the flat (controlled) area of the gain characteristic.
For example, in Figure 10.9, it can be seen that with the highlighted 26 dB isogain
trajectory, a slight “kick-up” in the gain occurs at low instantaneous power levels as
the PA drops out of compression. This introduces some AM:AM distortion in the
troughs of the signal, which may not be significant at high average power levels,
but will degrade ACLR as the average power level is reduced. By contrast, choos-
ing a slightly higher isogain, for example, 26.2 dB, would allow isogain operation
to be maintained at a lower average output power, at the cost of a slight decrease in
efficiency.
APT
ET
(a) (b)
FIGURE 10.10 PA phase surfaces for APT-optimized PA (a) and ET-optimized PA (b).
at that point—this gain being a good starting point for deriving an isogain shaping
table. Although the mean efficiency depends on the statistical distribution of the
signal waveform around the mean power level, the spot efficiency at that power level
is usually a reasonable proxy. Note that the PA will normally operate deeper into
compression at the waveform peaks—perhaps up to 4 dB—but by even more if the
shaping table is also used to provide deliberate “soft clipping” of the waveform.
The influence of the shaping table on the ET-path bandwidth can also be sig-
nificant. Any “sharp corners” in the shaping table, or non-monotonic sections, can
cause a significant increase in the ET-path bandwidth, and lead to mistracking if
the capabilities of the ETIC are exceeded. The minimum shaping table voltage also
serves to “de-trough” the envelope waveform, significantly reducing the envelope
path bandwidth requirement, since the unshaped envelope has very sharp troughs
where the RF envelope falls to zero, as illustrated in Figure 10.11.
4.0 V 4.0 V
3.0 V 3.0 V
2.0 V 2.0 V
1.0 V 1.0 V
0.0 V 0.0 V
Time -> Time ->
Power amplifier
can only deliver theoretical efficiencies of around 40%–50% for the AC amplifier
itself. However, since only a small portion of the energy supplied to the PA is deliv-
ered by the AC amplifier—perhaps only 5%–10%—the effect on the overall energy
efficiency of the ETIC is not dramatic.
• Maximum ET bandwidth
• ET voltage swing range
• Efficiency with high-bandwidth signals
• Efficiency with low-bandwidth signals
• Efficiency over back off
• Output impedance
• Current and voltage slew rates
• Buck, AC-boost, or DC-boost capability
• Supply voltage range
• Ease of programming
10.5.4 ETIC Bandwidth
One of the main implementation challenges for an envelope tracking IC is that of
bandwidth—the power supply requires a small signal bandwidth of 1.5–3.0× the RF
channel bandwidth, that is, 30–60 MHz for a 20 MHz OFDM channel.
The spectral plot of the envelope signal power shows that there is a large DC
component, followed by a significant amount of power within the occupied channel
bandwidth, and then a relatively long “tail” out to about 3× the channel bandwidth.
How much of this tail must be accurately tracked by the ETIC depends on the system
RF linearity requirements.
Some ETICs incorporate switchable bandwidth filters to support different chan-
nel bandwidths—these can be helpful at lower channel bandwidths to attenuate ET
DAC images, particularly if the modem chipset reduces the ET DAC clock frequency
at lower LTE channel bandwidths.
10.5.5 ETIC Efficiency
Perhaps harder than achieving the absolute bandwidth through the ETIC is to main-
tain high energy conversion efficiency for the ETIC itself. While 85% ETIC effi-
ciency is a good figure of merit for low-bandwidth signals such as 3G, maintaining
this efficiency over a wide range of signal types and bandwidths is a significant
architectural and IC design challenge.
Some ETICs are optimized for efficiency at a low bandwidth, for example, 1–4
RB, with a significant falloff in efficiency at higher RB counts. Others may provide a
Power Amplifier Envelope Tracking 229
lower efficiency for low bandwidths but maintain efficiency better as the bandwidth
increases.
Maintaining high efficiency over a wide power control range is also important for
applications, such as LTE and 3G, that employ dynamic power control. ET usually
delivers a power saving benefit over the top 8–10 dB of the power control range and
the efficiency of the ETIC as power is backed off usually determines the “break-
even” point compared with APT.
ETICs for cellular applications typically also include a lower bandwidth APT
mode for operation at reduced power levels, and this must also operate at high effi-
ciency over a very wide power output range—from 10 mW to more than 1 W—mak-
ing circuit design for low quiescent currents an important architectural and design
consideration.
10.5.6 Low Noise
For FDD-LTE systems, analog “noise” created by the ETIC is an important consider-
ation, as it is one of the contributors to overall RX sensitivity degradation. However,
not all “noise” is equal—there are multiple sources of noise and distortion within the
ETIC, all of which contribute to the overall system noise budget.
Some contributors can be considered as true noise, such as thermal noise in
the ETIC amplifier and switching noise from the SMPS blocks. It is not possible
(or practical in the case of switching noise) to predict and precorrect the RF signal to
mitigate these sources of noise.
Other effects, such as crossover distortion, bandwidth limitations, high output
impedance, or slew-rate limitations, lead to mistracking or a “noise-like” distortion
of the ET signal. Some of these effects give rise to deterministic memory effects,
which could, in principle, be corrected using memory effect predistortion, though
this is not typically used in current handset implementations.
10.5.7 Slew Rate
In addition to a high small-signal bandwidth, ET modulators also require a high slew
rate in order to accurately track the peaks and troughs of the waveform.
To accurately track a 20-MHz LTE uplink envelope with a 3-V swing at typical
handset output power levels needs over a 150 V/µs voltage slew rate, requiring the
output current to slew at about 50 A/μs.
Shortfalls in the slew rate will cause mistracking, leading to increased noise
and distortion on the RF output—this may be particularly harmful in FDD systems
where transmit noise in the receive band can degrade the receiver sensitivity.
Although the required voltage slew rate can (to some extent) be controlled at
the system level by the choice of the shaping table, the dynamic behavior of the
PA current consumption is determined primarily by the instantaneous RF output
power characteristics, and cannot easily be reduced. This tends to lead to high-
bandwidth current “spikes” and “clicks” caused by rapid changes in the RF output
amplitude.
230 Tunable RF Components and Circuits: Applications in Mobile Handsets
10.5.8 Output Impedance
If the ETIC has high output impedance, this results in voltage errors due to the
fast current changes in the load, leading to distortion and noise on the RF output.
Typically, the output impedance of a feedback amplifier is primarily inductive
and hence rises rapidly with frequency. Consequently, the ability of the correction
amplifier to “clean up” residual switcher noise falls with frequency. Furthermore,
the inductive output impedance together with the inductance associated with the
ETIC—PA physical interconnect results in memory effect distortion, which may
become significant with very high-bandwidth signals.
The ETIC effectively replaces the supply decoupling capacitor on the PA at video
frequencies, so low ETIC output impedance is also vital to maintain PA stability.
ETICs will typically be connected to multiple PAs on the board, which add para-
sitic capacitance and stray inductance even when the unused PAs are turned off.
Resonances associated with these parasitic elements are difficult to predict and con-
trol and can result in higher than expected out-of-band noise.
RF PAs present a time varying load impedance to the ETIC, making them far
more complex to drive than a simple resistive load. The nature of this “dynamic”
impedance depends on the PA technology and the choice of shaping function, and
typically rises in the waveform troughs as the PA comes out of saturation. The rapid
change of impedance can give rise to transient mistracking or “trough clicks” in
waveform troughs. It also requires more power to be delivered by the linear amplifier
of a hybrid ETIC than when driving an equivalent “linear” resistance—resulting in
the ETIC efficiency reducing by about 2 percentage points.
To optimize a PA for ET operation, there are several further design techniques that
can be employed:
• Increasing the PA load line to match the “AC boost” capability of the ETIC
• Supporting ET co-modulation of the PA driver stage (“VCC1”)
• Optimizing the phase response for ET rather than APT
• Ensuring the bandwidth of the supply bias networks is sufficient for ET
modulation
• Optimizing efficiency around mean rather than max power
• Ignoring any requirement for fixed supply AM:AM linearity at high power
• Ensuring the PA gain falls off as the supply voltage reduces, for good isogain
• Reducing the static PA bias current/Icq (fixed supply gain peaking is not an
issue for ET operation)
232 Tunable RF Components and Circuits: Applications in Mobile Handsets
provision of a dedicated APT supply rail for the driver stage or co-modulation of
both the driver and final stages.
If co-modulation is used, additional gain must be provided in the PA to compen-
sate for the compression in the driver stage. Care must also be taken to when design-
ing the PA to optimize the AM/PM response across both stages and to avoid memory
effects introduced by the propagation delay through the driver stage.
the ET shaping table can be optimized for best system efficiency, while relying on
the DPD to linearize the PA.
A wide range of precorrection techniques fall under the umbrella classification
of “DPD,” with widely varying complexity. It is helpful to categorize DPD systems
as either open- or closed-loop and as correcting memoryless or memory (time-
dependent) terms.
The simplest form of DPD is the open-loop memoryless system, which performs
a static pre-correction of AM:AM and AM:PM. In an ET system, the AM:AM com-
ponent can already be controlled via the ET shaping table, so it is also possible to use
a phase-only correction in an ET system. A difficulty of the open-loop approach is
having certainty that it is “stable enough” with respect to numerous system variables
(temperature, part-to-part variation, VSWR, etc.)—and typically this is only really
known once a large quantity of data has been gathered, which is seldom the case in
the early stages of a development.
The memoryless DPD can also be made adaptive or closed loop, where a math-
ematical model of the PA’s distortion is maintained and used to precorrect the trans-
mitted signal. The PA model is updated when needed, using an observation receiver
to capture a short section of the PA output, via an RF coupler. The system must
also capture the same section of the original digital TX signal fed to the PA input,
time align the reference buffer with the measurement buffer, and perform a series
of relatively complex mathematical operations to update the PA model—which can
take several seconds of DSP processing power. The associated overall power “cost” of
adaptation depends critically on the frequency of update—which could be anything
from once (factory calibration)—to occasional (on power up, or to the track tempera-
ture)—to more frequent (e.g., triggered by a change in load VSWR).
Both open- and closed-loop DPDs can also be implemented with “memory”
terms, that is, the precorrection applied to the TX signal depends not only on the
instantaneous amplitude of the signal, but also the amplitude of one or more prior
samples. This significantly increases the complexity of both the precorrection and
the adaptation tasks and is not commonplace in handset implementations at the time
of writing.
dBm
20 20
B5 LTE Uplink: 10 MHz (50 RB),
10 10
ACLR1 = –51 dBc 27 dBm, Skyworks PA ACLR1 = –53 dBc
0 0
ACLR2 = –56 dBc ACLR2 = –62 dBc
–10 –10
–20 –20
–30 Add PM-only –30
–40 memoryless DPD –40
–50 –50
–60 –60
–70 Isogain linearization –70
–80 –80
Center: 781.9200 MHz Span: 100.0000 MHz Add PM and only Center: 781.9200 MHz Span: 100.0000 MHz
Res BW: 100 kHz Vid BW: 1 MHz Sweep: 1.00 ms
memoryless DPD Res BW: 100 kHz Vid BW: 1 MHz Sweep: 1.00 ms
dBm
20 20
10 10
ACLR1 = –58 dBc ACLR1 = –66 dBc
0 0
ACLR2 = –66 dBc ACLR2 = –68 dBc
–10 –10
–20 –20
Add memory
–30 correction –30
–40 –40
–50 –50
–60 –60
–70 –70
–80 –80
Center: 781.9200 MHz Span: 100.0000 MHz Center: 781.9200 MHz Span: 100.0000 MHz
Res BW: 100 kHz Vid BW: 1 MHz Sweep: 1.00 ms Res BW: 100 kHz Vid BW: 1 MHz Sweep: 1.00 ms
and finally a full memory DPD solution. The ACLR progressively increases from
−51 dBc with the isogain correction to −66 dBc with full memory DPD.
It is important to note that the PA’s AM:AM and AM:PM response depends on
the ET shaping table selected and can be extracted from the PA measurement sur-
faces as previously described. For each ET shaping table, the AM:AM and AM:PM
responses could therefore be predicted and stored alongside the shaping table to con-
figure an open-loop DPD.
DPD systems for fixed supply PAs require additional dynamic range in the IQ
path DACs for “boosting” the peaks of the transmitted signal to compensate for the
roll-off in PA gain. However, the gain of an efficiency optimized ET PA (i.e., one
operating with “max efficiency” shaping) usually expands rather than reduces at
peak power and hence requires slightly less dynamic range in the TX path. If the ET
system is used to correct AM:AM (through the use of isogain shaping), and the DPD
is used to correct only AM:PM, this may reduce the bandwidth requirements of the
DPD, if only benign phase predistortion is necessary.
60 10
ET efficiency
Fixed supply efficiency
50 ET Pout
0
Fixed supply Pout
ET ACLR
Fixed supply ACLR
Efficiency (%), Pout (dBm)
40 –10
ACLR (dB)
30 –20
20 –30
10 –40
0 –50
–135 –90 –45 0 45 90 135 180
Load reflection phase angle (°)
FIGURE 10.15 ET versus fixed supply performance with a 3:1 VSWR mismatch.
Power Amplifier Envelope Tracking 239
operation). The isogain does not remain perfectly flat, resulting in moderate AM:AM
distortion. Figure 10.16 shows gain surfaces captured from an ET PA under 1:1 and
3:1 VSWR, with an isogain shaping table created from the 50 ohm surface (black
line, left plot), and then applied to the mismatched surface (black line, right plot).
By observing the fixed supply curves in Figure 10.16, it can be seen that the onset
of compression in fixed supply operation occurs 2–3 dB earlier under mismatch,
causing the waveform to be severely clipped.
These results illustrate the improvement in linearity when an ET PA operates
under mismatch at the output of the PA. However, in an LTE handset, mismatch
often occurs at the output of the duplex filter which follows the PA. This does gener-
ate a more complex interaction with the ET PA, since the reflected power now arrives
at the PA after being delayed through the filter. The delayed reflection can interact
with the ET supply modulation, and adjustments to the ET/RF timing alignment may
be found to improve performance under mismatch.
• ET path gain
• ET path offset
• ET/RF path gain errors
• ET/RF path timing alignment
240 Tunable RF Components and Circuits: Applications in Mobile Handsets
50 ohm load
ohm50.sfc: Gain (dB)
34
32
30
28
Gain (dB)
26
0.5 V
1.0 V
24 1.5 V
2.0 V
2.5 V
3.0 V
22 3.5 V
4.0 V
4.5 V
5.0 V
20
–15 –10 –5 0 5 10 15 20 25 30
Output RF power (dBm)
32
30
28
Gain (dB)
26
24 0.5 V
1.0 V
1.5 V
2.0 V
22 2.5 V
3.0 V
3.5 V
4.0 V
20 4.5 V
5.0 V
10.10 CONCLUSIONS
Although ET was initially introduced to reduce PA energy consumption in LTE
handsets, the technology also delivers several other system level performance
improvements, such as increased peak and average output power, improved linearity
(even into mismatch), and better stability over temperature.
Almost all LTE chipsets now support ET, which requires the integration of a
relatively simple digital signal processing block and a high-speed envelope DAC into
the modem chipset. The ET path can be integrated into either the RFIC or the BBIC
depending on the chipset partitioning. ET does need relatively high sample rates for
the signal processing, with a good rule of thumb being 6× the RF channel bandwidth.
ETICs are now available from multiple vendors and usually include a combina-
tion of switch-mode power supply technology and high-bandwidth linear amplifiers.
The interfaces between the modem chipset and ETIC have been standardized by
the MIPI Alliance as eTrak and all ETICs deployed in terminals also use the MIPI
RFFE low-speed serial control interface for control and configuration.
ET PAs require some basic modifications to support high-bandwidth supply mod-
ulation, such as removal of video decoupling components, but for optimum perfor-
mance should be designed with ET in mind. For “ET-only” PAs, which do not need
to support high-power APT operation, many traditional design goals (such as lin-
earity) can be sacrificed to improve ET performance, requiring a different mindset
from the PA designer. ET also unlocks the potential for CMOS PAs to be used with
high-PAPR signals, such as LTE and LTE-Advanced.
As ET transforms the PA from a two-port analog component to a three-port soft-
ware defined device, comprehensive system modeling is required to understand how
performance is influenced by different system settings and component parameters.
An accurate three-port model of the PA is needed, requiring new tools for lab mea-
surement and PA characterization, which are now starting to appear from multiple
vendors.
By intelligent design of the ET shaping table, the PA can in some cases be linear-
ized without requiring any DPD, reducing development complexity and bandwidth
requirements in the modem chipset. If ET is combined with DPD, the DPD com-
plexity is significantly reduced, as the PA’s distortion characteristics in ET mode are
much more tightly controlled over temperature and process, allowing use of open-
loop DPD with static gain and phase tables developed at design time.
242 Tunable RF Components and Circuits: Applications in Mobile Handsets
ET does introduce some new complexity into the modem and RF system and has
some impact on production calibration. But just like any other technology, once the
industry overcomes the initial learning curve, ET will quickly become ubiquitous
technology for 4G, 5G, and beyond.
Although ET has been around since the 1930s, it offers significant promise for the
twenty-first century, extending the boundary of “software defined radio” from the
modem into the RF front end.
11 MEMS Switching in the
Handset RF Front End
Igor Lalicevic
CONTENTS
11.1 Overview........................................................................................................ 243
11.2 Data Speed Improvement and Wireless Technology Development............... 243
11.3 RF Front End: Introduction........................................................................... 247
11.4 RF MEMS Switches in the RF Front End..................................................... 250
11.5 RF MEMS Switch Benefits........................................................................... 250
11.6 DelfMEMS RF MEMS Switch Solution....................................................... 254
11.7 Conclusions.................................................................................................... 256
11.1 OVERVIEW
The focus of this chapter is on radio frequency microelectromechanical system (RF
MEMS) switching technology and the benefits that RF MEMS switches provide for
Long Term Evolution (LTE) wireless technology.
RF MEMS benefits are observed through different aspects of mobile phone per-
formance improvement. The history of wireless technology development will be dis-
cussed from the perspective of improving the speed of data.
All of the major components in a typical mobile phone RF front end (RFFE) will
be considered, and we will focus on their impact on performance with respect to the
insertion loss, isolation, and linearity of the system.
Special attention is given to the next generation of RF switching requirements for
the higher end of the LTE performance scale, and a particular emphasis is placed
on the need for high-performance, high-quality, and highly reliable RF MEMS
switches. This chapter includes a description of DelfMEMS RF MEMS switch tech-
nology as a solution for the next generation of switching and how it addresses the
need for high performance while minimizing insertion loss, increasing isolation, and
providing superior linearity performance in the front end.
243
244 Tunable RF Components and Circuits: Applications in Mobile Handsets
generate more than half of the global IP traffic. This is an excellent way to describe
the history of wireless technology development, which has witnessed tremendous
evolution and change in cellular standards over the past 35 years.
First-generation wireless telephone technology, also called 1G, was launched in
Japan by NTT (Nippon Telegraph and Telephone) in 1979. 1G technology used ana-
log transmission techniques for transmitting voice signals. This voice-only standard
used a frequency-division multiple access (FDMA) technique where voice calls were
modulated to a higher frequency of approximately 150 MHz and greater as it is
transmitted between radio towers.
1G issues such as low capacity, unreliable hand-off, poor voice links, and com-
plete lack of security, resulted in development of early digital systems, which became
known as 2G: second-generation wireless telephone technology. These systems
became available in the 1990s and used digital multiple access techniques, such as
time division multiple access (TDMA) and spread-spectrum code division multiple
access (CDMA), which enabled the first low bit rate data services.
2G systems offered higher spectral efficiency, the first data services, advanced
roaming, and for the first time, a single unified standard was provided: global system
for mobile communications or GSM. First employed in Europe in 1991, today GSM
is still utilized across the world.
As the requirements for sending data over the air interface increased, general
packet radio service (GPRS) and wireless application protocol (WAP) technologies
were added to existing GSM systems. This was considered an advancement from 2G
and became known as 2.5G technology, where packet-switching wireless application
protocols enabled wireless access to the Internet.
Even though 2G supported data over the voice path, data speeds were typically a
low 9.6 Kb/s or 14.4 Kb/s, which effectively made 2G a voice-centric system. Data
speed improvement that 2.5G technology introduced with more advanced coding
methods provided theoretical data rates of up to 384 kbps. However, limitations of
packet transfer technology, which behaves in a similar way to a circuit switch called
over the air, combined with low system efficiency and non-standardized networks
across the world led to the birth of the 3G standard.
The initial planning for third-generation or 3G standards started in the 1980s,
and the objective was to enable multimedia applications such as videoconferencing
for mobile phones. However, over the years, 3G’s focus moved toward personal wire-
less Internet access and the need to have connected access worldwide while achiev-
ing increased system and network capacity.
In the year 2001, the first commercial 3G network based on W-CDMA technol-
ogy was launched in Japan by NTT DoCoMo.
The need to create a globally applicable mobile phone system specification
resulted in the 3rd Generation Partnership Project (3GPP), which united seven tele-
communications standard development organizations from Asia, Europe, and North
America known as the “Organizational Partners.” 3GPP has become the focal point
for 3G and mobile systems beyond and it provided the requirements for 3G data
speed specifications of up to 2 Mbps for stationary users.
To additionally enhance Universal Mobile Telecommunications System (UMTS)-
based 3G networks, and to enable higher data speeds, features like High Speed Packet
MEMS Switching in the Handset RF Front End 245
Channel
bandwidth
LTE-Advanced 802.16 m/WiMAX2
Up to 100 MHz 4G (R10,11,12+, FDD, and TDD) (WirelessMAN-Advanced)
Marketing
1.2288 MHz 2.5G iMODE HSCSD GPRS evolution IS-95B (US CDMA)
200 KHz 2G PDC (Japan) GSM (Europe) IS-136/AMPS (US TDMA) IS-95A (US CDMA)
Channel
bandwidth
LTE-Advanced (R10,11,12+, FDD, and TDD), 802.16m/WiMAX2 (WirelessMAN-Advanced)
Up to 100 MHz 4G Max data speed: DL 1 Gbps (target 3 Gbps), UL 500 Mbps (target 1 Gbps)
1.4, 3, 5, HSPA+, E-HSPA, LTE (R8,9, FDD, and TDD), 802.16e (Mobile WiMAX)
10, 20 MHz
3.9G Max data speed: DL 100 Mbps, UL 50 Mbps
WCDMA (FDD and TDD), TD-SCDMA (China), E-GPRS (EDGE), cdma2000 (1 × RTT)
5 MHz 3G Max data speed: DL 2 Mbps, UL 2 Mbps
PDC (Japan), GSM (Europe), IS-136/AMPS (US TDMA), IS-95A (US CDMA)
200 KHz 2G Max data speed: DL 14.4 kbps, UL 14.4 kbps
(DL) and uplink (UL), currently LTE Cat9 will provide up to 450 Mbps downlink
(DL) and 50 Mbps uplink (UL) data rates. But the market demand continues to
grow. Even though the peak data rate targets for LTE-Advanced of 1 Gbps for the
downlink and 500 Mbps for the uplink connection have not been achieved, revised
targets to double or even triple these values are emerging.
Any solutions leading to a dramatic data rate increase will require more frequency
spectrum. 3.5 GHz LTE bands 42 and 43 are a great illustration of this, together with
the target to introduce 100 MHz of bandwidth for downlink carrier aggregation.
Carrier aggregation is one key enabler for LTE-Advanced to meet the International
Mobile Telecommunications–Advanced (IMT-Advanced) requirements of achiev-
ing peak data rates by a bandwidth increase methodology. Carrier aggregation or
channel aggregation enables multiple LTE carriers to be used together to provide a
bandwidth increase.
The LTE cellular standard is divided into categories, each of which caters to dif-
ferent requirements, as shown in Table 11.1.
TABLE 11.1
LTE Cellular Standard Categories
Categories LTE Cat1 LTE Cat2 LTE Cat3 LTE Cat4 LTE Cat6 LTE Cat9
Data Rates DL 10 50 100 150 300 450
(Mbps) UL 5 25 50 50 50 50
Channel DL 20 20 20 20 (CA) 40 (CA) 60 (CA)
Bandwidth UL 20 20 20 20 20 20
(MHz)
DL MIMO Not sup. 2 × 2 2 × 2 2 × 2 2 × 2 2 × 2
Configuration
MEMS Switching in the Handset RF Front End 247
Since the first LTE carrier aggregation solution provided LTE Cat4 that used
10 + 10 MHz aggregation, LTE Cat6 20 + 20 MHz carrier aggregation has delivered
an additional 20 MHz of channel bandwidth that resulted in doubling data rates from
150 to 300 Mbps. The state of the art LTE-Advanced solution today, LTE Cat9, is
using 60 MHz (20 + 20 + 20 MHz), three bands carrier aggregation in a 2 × 2 MIMO
mobile handset configuration, with the capability to offer a 450 Mbps peak down-
link data rate. While downlink data rates from LTE Cat1 to LTE Cat9 has increased
45 times, improving from 10 to 450 Mbps, uplink data rate speeds have increased
only 10 times, improving from 5 Mbps for LTE Cat1 to 50 Mbps for LTE Cat9. It is
noticeable that uplink data speed has stagnated since LTE Cat3.
Diversity RFIC
FEM
Diversity
antenna port SPnT
Diversity
SPnT RX LNAs
Mipi RFFE
High
bands
mipi RFFE
Main
mipi RFFE RX LNAs
FEMid
Main
antenna port
VHF 4G PAM
Antenna SPnT
tuner PA
SPnT
mipi RFFE
modulation allows for the usage of efficient, saturated power amplifiers (PA) and
thereby saves on current consumption; a critical issue for battery powered equipment.
In this low-data rate modulation technique, low-RFFE IL of less than 1 dB had
minimal impact on the RF current, of which the vast majority is spent on highly
efficient PAs.
The situation became progressively worse with 3G standards. Battery life was vis-
ibly shorter compared to 2G phones, but for an average user everyday recharging was
still not necessary. The 3G standard provided higher data speed by way of more effi-
cient modulation schemes, which was an acceptable trade-off for decreased usability.
Binary phase shift keying (BPSK) is a major modulation technique within
Universal Mobile Telecommunications Systems (UMTS) technology, and it requires
PAs working in the linear operating mode. PAs are not operating in the efficient
saturated region due to the high peak to average power ratio of HSDPA/HSUPA
MEMS Switching in the Handset RF Front End 249
10 250
8 200
6 150
4 100
2 50
0 0
–50 –40 –30 –20 –10 0 10 20
Transmit power level (dBm)
FIGURE 11.4 Transmitter current drain and 3G handset power distribution probability ver-
sus handset output power.
modulations that are applied in 3G mobile systems. PA efficiency, which defines bat-
tery supply consumption and PA adjacent channel power ratio linearity performance,
works directly against each other. Therefore, linear PAs are, by default, less efficient
from a current drain point of view. An interesting aspect of the 3G phone’s battery
life is presented in Figure 11.4.
The grey curve is an example of the PA current drain versus the mobile phone
transmitted power. We see the exponential current drain increase versus the trans-
mitted output power level. It follows then, that at a very high or maximum output
power level, with maximum current drain consumption, our battery life will be sig-
nificantly reduced.
From the black curve, which represents a 3G user’s transmit power distribution,
we see the probability of 3G phones operating at the highest power levels with high
current consumption is below 2%. In conclusion, we can state that 3G phones have
limited usage at the highest power levels where current drain is high, and the normal
operating point for 3G phones is at lower power levels, which allow longer battery life.
A simple calculation using Figure 11.4 shows that an output power level of around
0 dBm will have current drain that is around 1/6 of the maximum transmitted output
power current drain level. This will make a 3G phone battery life around 80% longer
at the 0 dBm transmitted power level, compared to the battery life at the +23 dBm
transmitted power level.
Any positive experience with long battery life for the 2G and 3G standards disap-
peared with fourth-generation phones. 4G mobile phones, LTE, and LTE-Advanced,
have to be recharged daily and in a typical user case even more than once per day.
The orthogonal frequency division multiplex (OFDM) modulation scheme requires
PAs that use the same or a very similar design as 3G PAs, and the same linear PAs
today are being used for both 3G and 4G applications.
250 Tunable RF Components and Circuits: Applications in Mobile Handsets
From the battery life point of view, the crucial difference is that LTE networks
typically require short bursts of high RF output power levels for data transmission.
Going back to Figure 11.4, this means 4G phones are almost always used at maxi-
mum (+23 dBm) output power level that implies maximum current drain for 4G
applications and a shorter battery life.
LTE-A with carrier aggregation and global roaming requirements are making
things even worse. The necessity for LTE-Advanced to introduce additional LTE
bands resulted in tremendous RFFE power losses. The typical RFFE IL for 2G was
around 0.5 dB, and with the 3G standard this loss increased to around 1.5 dB. The
IL for the 4G RFFE due to additional filtering needs and complex switching require-
ments can easily reach over 4 dB, which translates into over 60% of the power being
lost. To offset and cover increased power losses and to meet the maximum antenna
radiated power requirements, 4G PA’s have to be capable of providing additional
power that makes 4G PAs operate even less efficiently.
The inherent high linearity of a MEMS switch is probably the most obvious
benefit of RF MEMS switching because it can enable uplink carrier aggrega-
tion. Equally as important, especially at the higher frequency operating bands,
RF MEMS ultra-low IL is seen as an ideal solution to maintain high transmitter
power efficiency and a low-receiver noise figure. MEMS inherent design charac-
teristics provide the high isolation performance needed between the transmitter
and receiver.
The 3GPP standard is used by the industry to determine the degree of linearity
required to avoid interference with other devices on the network. This is done by
specifying the third-order input intercept point (IIP3).
Each new generation of cellular network required progressively higher linearity.
In Table 11.2, we can see that the 2G requirement for switch linearity was an IIP3 of
55 dBm, the 3G switch requirement was 65 dBm, the LTE switch IIP3 requirement
is 72 dBm, and the LTE-Advanced antenna switch with up-link carrier aggregation
capabilities will have to meet an IIP3 of 90 dB. Currently, the dominant solid-state
switch technologies such as SOI or SOS are not capable of achieving an IIP3 target
of 90 dBm.
It is essential to start thinking about the existing switching technology lim-
itations, which will have to be addressed with alternative solutions or a com-
pletely new technology that can meet existing challenges and provide the needed
performance.
RF MEMS technology provides answers for these LTE-Advanced challenges
with the benefits of providing better call quality, longer battery life, and enabling
the 90 dBm requirement for uplink carrier aggregation to improve smartphone
performance.
In order to compare the performance of the different switch technologies, an
industry standard figure of merit (FOM) has been established, which is defined as
RONCOFF, and is expressed in femtoseconds (fs).
TABLE 11.2
Network Linearity Requirements
Network Linearity (IIP3) (dBm)
2G 55
3G 65
3.9G 72
4G (UL CA) ≈90
252 Tunable RF Components and Circuits: Applications in Mobile Handsets
600
FOM
500 485
448 0.18 μm
0.25 μm Thick film
400
Triple gate
300
0.18 μm 250 270 264 0.35 μm
Thin film 224 Single gate
200 207 STM
192
Infineon 160 Next gen STM
CMOS 113 Peregrine ultraCMOS
100
8 DelfMEMS RON
0
0.6 0.8 1 1.2 1.4 1.6 1.8 2
Solid-state switch designs are heavily dependent on the field effect transistor’s
(FET) channel resistance when it is ON (RON), and the switch’s isolation cannot
be better than the FET’s total capacitance when the switch is in OFF state (COFF).
Therefore, it follows that for high throw-count solid-state switches design, OFF state
switch branches will become a dominant source of non-linearity, IL and isolation
degradation. Equally important are degradations that solid-state technology will
experience at higher frequency operation.
By contrast, DelfMEMS RF MEMS switches retain ultra-low IL (0.35 dB @ 3.5
GHz), very high port-to-port isolation (35 dB at 3.5 GHz) and ultra-high linearity
(85 dBm up to 98 dBm) even for high multi-throw switch configurations and at
higher frequency operation.
These parameters make RF MEMS switches a perfect candidate for the 4G
multiband environment where RF components have to demonstrate minimum
power loss degradation at frequencies of up to and above 3.5 GHz. Figure 11.6 com-
pares DelfMEMS versus solid-state switch technologies in the SP12T RF switch
configuration.
The power loss reduction that can be realized replacing existing the solid-state
switch with the DelfMEMS switch leads directly to a significant battery power sav-
ings and increased receiver sensitivity. These improvements become even greater
at the new 3.5 GHz LTE bands, emphasizing RF MEMS broadband design and its
ability to maintain superior performance with increasing frequency.
The block diagram presented in Figure 11.7 assesses the potential power savings
when comparing the DelfMEMS RF MEMS switch to existing solid-state antenna
MEMS Switching in the Handset RF Front End 253
–0.2
–0.4
Insertion loss (dB)
DelfMEMS
–0.6
Diversity RFIC
FEM
Diversity
SPnT
antenna port
Diversity
SPnT
RX LNAs
mipi RFFE
B 7, 40, 41… SOI RF MEMS Current
mipi RFFE
IL (dB) –1 –0.35 savings
SPnT
CDL 21% 8% 13%
High
bands
mipi RFFE
Main
mipi RFFE RX LNAs
FEMid
B 1, 2, 3, 4… SOI RF MEMS Current
Main
antenna port IL (dB) –1.2 –0.3 savings
Antenna
CDL 24% 7% 17%
turner
mipi RFFE SPnT
mipi RFFE
B 5, 8… SOI RF MEMS Current RFIC
IL (dB) –0.9 –0.25 savings TX FE
SPnT
CDL 19% 6% 13%
Envelope tracking
modulator
VPA
mipi RFFE mipi RFFE
mipi RFFE
FE losses while providing exceptional band-to-band and TX/RX isolation and the
needed high linearity to enable uplink carrier aggregation.
Diversity RFIC
FEM
Diversity
B 1, 2, 3, 4… SOI RF MEMS RxSI
antenna port SPnT
IL (dB) –1.4 –0.3 29%
Diversity
RX LNAs
SPnT B 5, 8… SOI RF MEMS RxSI
IL (dB) –0.9 –0.25 16%
mipi RFFE
mipi RFFE
Complete control of the MEMS membrane allows for an increased gap between
contact area and transmission line in the OFF state, which is directly linked to the
switch isolation and allows for switch resetting in the unlikely event of stiction.
Moving from the ON state to the OFF state is made through an electrostatic
active actuation, which de-correlates between restoring forces, contact forces, and
the membrane mechanical properties. In addition, the effects of mechanical bounce
are substantially reduced. Importantly, this does not depend only on the elastic resto-
ration forces. This advanced electrostatic actuation results in a very short switching
time, typically less than 3 μs.
The ability to have a reduced gap between the membrane and the transmission line
is a major advantage of the DelfMEMS switch structure. It ensures that an increased
ON-state contact force is achieved with reduced actuation voltages and consequently
delivers ultra-low insertion losses. Due to the reduced gap, the maximum deflec-
tion of the membrane will be reduced as well, and as a result, membrane mechani-
cal stress and the creep effect decreases. The reduced gap combined with a unique
metallurgy for the contacting surfaces results in a highly reliable RF MEMS switch.
Thanks to the DelfMEMS original design approach, RF MEMS switches can be
used effectively for the first time as an RFFE switching solution.
11.7 CONCLUSIONS
LTE technology with its global roaming requirements, carrier aggregation, MIMO
design approach and migration to higher frequency bands has made RFFE archi-
tectures for high-end smartphones exceptionally complex and a key bottleneck in
achieving the RF performance levels required by the market.
Mobile handset battery life, call quality, and data throughput are the most obvious
victims of this excessive complexity.
Extensive work has already been performed to improve IL, isolation, and linear-
ity of existing RFFE systems, but the results remain largely disappointing and, when
compared to 3G, no significant improvements have been demonstrated to date.
A number of companies are endeavoring to design tunable frequency compo-
nent solutions that will reduce existing RFFE complexity. However, many of these
approaches are fraught with difficulty and risk. DelfMEMS simple, yet innovative,
approach to resolving the RFFE complexity problem is to replace specific switching
components with a highly improved switching solution as has been discussed in this
chapter, a solution that provides significant overall system performance benefits.
12 Case Study of Tunable
Radio Architectures
Alpaslan Demir and Tanbir Haque
CONTENTS
12.1 Introduction................................................................................................... 257
12.2 Deployment Scenario..................................................................................... 258
12.3 Context Aware RFFE Resource Manager within the Proposed System
Architecture��������������������������������������������������������������������������������������������������260
12.3.1 System Metrics.................................................................................. 261
12.3.2 Sensor Fusion..................................................................................... 262
12.3.3 User Preferences................................................................................ 262
12.3.4 Network and Regulatory Policies...................................................... 262
12.3.5 RFFE Controller................................................................................ 262
12.3.6 RFFE Hardware................................................................................. 263
12.4 Protocol Stack and Overall System Integration............................................. 263
12.4.1 Macro Procedures and RFFE HW Drivers........................................ 264
12.4.2 CARF RM Optimizer........................................................................ 265
12.4.3 Location and Interference Awareness................................................ 265
12.4.4 User Preference.................................................................................. 265
12.4.5 QoS Awareness.................................................................................. 265
12.4.6 Policy Enforcer.................................................................................. 265
12.5 RF Perspective on Simultaneous Operation of LTE and Wi-Fi RATs..........266
12.6 CARF-Based Messaging Flow Diagram for In-Device Coexistence............ 272
12.7 Conclusions.................................................................................................... 274
References............................................................................................................... 274
12.1 INTRODUCTION
Recent user demand for ubiquitous, instantaneous over-the-air access to multimedia-
rich content has forced related industry segments such as service providers, device
manufacturers, technology developers, as well as regulatory bodies to collaborate
more than ever before. In the United States, for example, the Federal Communications
Commission (FCC) is trying to balance incumbent user demands while creating
opportunities for others to trigger economic growth especially in TV white spaces
(TVWS),1 also known as digital dividend, the spectrum that is released as a result of
television broadcasting companies switching from analog to digital-only platforms.
The FCC’s initiatives in TVWS were followed by countries such as Canada, Australia,
Great Britain, and others. As part of stimulating economic growth, the FCC is working
257
258 Tunable RF Components and Circuits: Applications in Mobile Handsets
together with industry stakeholders and the Department of Defense (DoD) to enable
utilization of the 3.5 GHz Citizens Broadband Radio Service (CBRS) band for mul-
titier communications.2 The FCC is also looking to facilitate mobile radio services in
bands above 24 GHz to enable the fifth-generation (5G) mobile services within the
context of broader efforts to develop technical standards.3
Increasing data traffic demands, especially due to social networking and related
multimedia-rich content delivery, resulted in heavy data traffic burden for LTE tech-
nology over an all IP network. The issues impose a constantly expanding load on
packet data distribution. As a result, data off-loading is defined as a key solution
to ease core network congestion problem in the 3rd Generation Partnership Project
(3GPP) standardization efforts.4
To catch up with user demands and application developers’ limitless imagination, the
device manufacturers are developing very powerful, yet complex and compact handset
designs that enable multiple applications to run simultaneously. Also, the multiband,
multimode enabled devices became a design requirement due to the fractional reuse of
RF spectrum and relevant standardization efforts. However, due to the compact design
constraints, enabling simultaneous operations of multiple radio access technologies
(RATs) can cause in-device self-interference. The deployment of two separate RATs
adjacent in space and spectrum may not be always possible. The operation of one RAT
with high transmit power levels can desensitize another RAT’s receiver in space-lim-
ited compact devices when the spectral separation between the two is limited.5
The tunable radio architectures may address a much broader solution spectrum for
the problems mentioned earlier. However, in this chapter, a deployment scenario that
is related to radio frequency front end (RFFE) in-device coexistence issues for simul-
taneous operations of LTE and Wi-Fi RATs is emphasized. Then, we define a system-
level context aware RF front-end (CARF) approach to create a framework to enable
intelligence in the RFFE domain. Also, an exemplary perspective on the protocol stack
and overall system integration is captured. We conclude our study with some RF level
treatment of the LTE and Wi-Fi coexistence issues followed by an exemplary control
flow implementation that can enable adjacent and simultaneous operation of multiple
RATs.
quality at network edge or anywhere the users are expected to appear. In other words,
the minimum SNR at network edge is guaranteed to be more than +20 dB. Although
the Wi-Fi systems are still operational below +20 dB SNR levels, the connectivity
and data rates may become problematic at around +10 dB SNR. A Wi-Fi deployment
with a single AP may operate at 24 dBm equivalent isotropic radiated power (EIRP)
with less than +6 dBi omnidirectional antenna gain at 2.4 GHz. The effective noise
power can be calculated by using Equation 12.1 at room temperature for bandwidth of
20 MHz as −101 dBm. If the receiver noise figure is assumed to be 7 dB, the effective
noise power can be calculated as −94 dBm. Since we define the network edge SNR
at +20 dB minimum, the target received power at the receiver antenna port must be
at least −74 dBm (receiver noise power + SNR). Therefore, the path loss of 98 dB (Tx
radiated power + receiver noise power − SNR at network edge) can be tolerated. By
using Equation 12.2, the Indoor Hotspot for non–line of sight at 2.4 GHz as described
in Annex B of,7 the path loss of 104 dB corresponds to 66 m in distance. The coverage
may be described as a circle with 66 m in radius where the AP is located at the center.
The coverage we described is based on transmit power, path loss, and receiver char-
acteristics; however, the coverage can also be defined based on service availability:
Core network
Internet
Policies
Media Regulatory
server policies
Home
AP ENodeB
LTE
Macrocell
UE
Network and
regulatory Policy
policies databases
System Protocol
metrics stack
RFFE controller
RFFE CARF
RM User
preferences
RAT State
Sensor
Metrics
fusion
Database
Other inputs
FIGURE 12.2 CARF RM system architecture with RAT state metrics database.
The RFFE controller can be flexible and can leverage additional performance
settings to meet performance objectives with RFFE resources as they become
available. For example, emerging device technologies can be incorporated to the
RFFE controller that may provide new capabilities to further optimize the RFFE
performance.
The RAT State Metrics Database, as depicted in Figure 12.2, includes certain
system parameters such as channel assignment, Tx power limits, and RAT activ-
ity status. In the case of Multi-RAT operations CARF RM utilizes the RAT State
Metrics to enable or disable self-interference cancellation methods. The details about
how the CARF RM make use of these metrics will be given in Section 12.3.1.
In order to ground the CARF RM concept we define what each block entails in
the following sections as shown in Figure 12.2.
12.3.1 System Metrics
CARF RM requires protocol stack integration in order to extract system-level
knowledge to adjust RFFE parameters. System metrics such as QoS indication on
each packet or the group of packets and active channel SNIR levels give CARF
RM necessary inputs to run certain algorithms and create commands to optimize
RFFE HW resources. For example, the low QoS required marking on a packet can
trigger the reduction in linearity requirements on the transceiver chain. Conversely,
the high QoS marking on a packet can push the transceiver chain to be more
linear. Adjustments in battery power management depends on the fact that the
high linearity in RFFE requires more battery power consumption. The example
262 Tunable RF Components and Circuits: Applications in Mobile Handsets
12.3.2 Sensor Fusion
The CARF RM heavily relies on sensory information to scan spectrum in order to
create interference mapping based on location. It is assumed that the sensor fusion
block is capable of location estimation via GPS or other means. The CARF RM
obtains the location info from the sensor fusion block. It will set scanning require-
ments such as threshold, channel bandwidth, search span, and scanning method and
then send commands to the sensor fusion block.
12.3.3 User Preferences
The CARF RM is capable of optimizing user experience based on user prefer-
ences. For example, the user can override RFFE performance criteria based on
time duration and application. A user may choose to set RFFE in high-perfor-
mance mode for an important phone call based on a contact’s higher associated
group set by the user by associating names to different RFFE performance levels
while the penalty drains the battery power faster to receive a higher quality of
experience. A user can also set predefined performance metrics to be adhered by
RFFE resources for specific applications based on packet inspection that identifies
activity for the application.
12.3.5 RFFE Controller
This entity maps the received commands from the CARF RM to the Macro proce-
dures to control RFFE HW resources such as setting the tunable filter effective band-
width and amount of filtering required or enabling/disabling analog and/or digital
self-interference cancellation blocks.
Case Study of Tunable Radio Architectures 263
12.3.6 RFFE Hardware
The RFFE HW is comprised of spectrum and interference detector, circuit level
sensors/detectors, and a transceiver chain that contains PA, LNA, tunable antennas
and filters, baseband processing capabilities, and all other configurable elements.
The RFFE HW is connected to the RFFE controller by means of Macro procedures
that contains SW drivers for each RFFE resources. Each configurable element may
be associated to a unique driver for operation and control.
Regulatory
policies Operator
policies
Application layer
Application
location, proximity,
light, etc., sensors layer
Tunable RF front end
aware optimizer
Logical
Protocol stack
ET
switchplexer
User preference
Transceiver
processing
Antennas
Baseband
PA based optimizer
filters
QoS aware
Physical optimizer
Policy enforcer
Spectrum sensor
and Logical
interference detector
Circuit level sensors/detectors Physical
(i.e., temperature)
The messaging described here can be overlaid onto any existing or future stan-
dard or custom, semi-custom, or proprietary interfaces such as SPI, for example.
These messages specify the control interfaces between the CARF RM and Macro
procedures (i.e., RFFE drivers) via the RFFE controller.
By structuring messages on the interfaces based on CARF concepts described so
far in this chapter, the benefits such as control latency, battery power efficiency, and
radio faults circumvention can be realized. In addition, the learning mode support
may enhance performance and battery power consumption.
As part of context awareness, social setting can also be supported. The user, for
example, may associate higher or lower RFFE performance to a particular contact in
a specific social media platform such as Skype, WhatsApp, Viber, and LINE contacts.
The CARF control requires an efficient and capable real-time interface that exists
between the CARF RM and the RFFE controller as shown in Figure 12.3. The CARF
RM can also affect the interfaces between the RFFE controller and the RFFE resource
such as tunable antennas and tunable filters. The interface must support control of a
diverse collection of RFFE resource elements. A myriad of control parameter formats
require the interface to be highly flexible. The problems confronting the architecture
of a suitable interface include, but are not limited to, issues that are described here.
The interface should be low latency for real-time control of RFFE resource elements.
For example, some controls are only valid for a short time due to fast changing context
that makes low latency essential. Other instances requiring low latency are also possible.
The interface should be low power to conserve battery life and the minimiza-
tion of message traffic could reduce power. The programming and development of
applications should be simplified. Initiating functional modes with abstracted or
combined controls can help to simplify application development.
The real-time debugging should be enabled by the interface. This is desirable to
fix problems and tune performance. The operation during fault conditions and fault
circumvention should be supported where possible.
The emergency communications should be supported. For emergency context,
the RFFE performance should be highest possible to increase the likelihood of suc-
cessful communication. The interface should support social settings chosen by the
user or service provider. For example, the user may require utmost performance for
a mission critical phone call or launching a particular application.
The operation during low battery power conditions should be supported. The
communication bandwidth can be reduced when battery power is at a critical low
level. Other optimizations besides limiting bandwidth are also possible.
The learning should be enabled to facilitate optimization of performance and
battery life. Other benefits from learning are also possible.
The interface should support complete control of all RFFE operating parameters.
It should not limit functionality, operation, or performance.
it may map the command received at the RFFE controller to set high RFFE perfor-
mance into Macro procedures that may do several adjustments such as increasing
biasing current and voltage, changing PLL loop filter bandwidth, and so on.
12.4.2 CARF RM Optimizer
The interfaces within CARF RM enable all the decision-making-related information
among location and interference aware optimizer, user preference–based optimizer,
QoS-based optimizer, and policy enforcer.
12.4.4 User Preference
This entity receives user preferences and provides information to other controlling
blocks under CARF RM. It sends commands to the RFFE controller. For example,
the user may prefer for a period time that the RFFE performance is to be very high
or very low. Upon the reception of user preference, the user request is translated as a
command and sent to the RFFE controller by the user preference–based optimizer.
The user may prefer that a particular application is associated to a RFFE perfor-
mance level.
12.4.5 QoS Awareness
The main functionality of this entity is to observe protocol stack–based system
metrics and optimize RFFE performance. For example, in the case of IEEE 802.11
protocol stack, when communication is done packet by packet, the QoS aware
optimizer observes the SNIR level required for each packet and sends commands
to the RFFE controller in order to prepare the RFFE for each packet transmission.
LTE Wi-Fi
baseband baseband
LTE ISM
band 40 band
FIGURE 12.5 LTE band 40 and ISM band interference scenarios: (a) poor, (b) moderate,
and (c) high spectral separation.
TABLE 12.1
Average Out-of-Band Rejection of LTE and ISM Bandpass Filters
Frequency (MHz)
2300–2380 2380–2400 2402–2414 2415–2418 2419–2480
ISM BPF >39 dB 4 dB — — —
(ξWi-Fi), dB
LTE B40 BPF — — 10 dB >35 dB 50 dB
(ξLTE), dB
hand, the rejection of the LTE bandpass filter in the adjacent ISM band determines
the level of LTE transmitter noise and nonlinear distortion that may fall in the
ISM band.
The effect of OOB blockers and spurious emission interference (nonlinear dis-
tortion and noise) is studied in.9 The total impact of blocking and spurious emis-
sion is considered. The results are summarized in Table 12.2. The table entries are
the level of Wi-Fi (victim) receiver desensitization in dB. Desensitization is defined
as 10 log10(α) where α is the total coexistence interference to the receiver thermal
noise floor ratio. For a given entry in the table, the column represent the aggressor
(LTE) transmit frequency and the row represents the victim receive frequency. It is
assumed that the LTE transmitter output is set to its maximum level (+23 dBm) and
the Wi-Fi receiver is set to its maximum gain setting (i.e., operates at maximum sen-
sitivity). The victim receiver desensitization is categorized into three levels—level 1
(>50 dB), level 2 (>10 dB and <50 dB) and level 3 (<10 dB). In the most extreme case,
268
TABLE 12.2
Coexistence Interference (Blocking and Spurious Emission) Impact from LTE in Band 40 on Wi-Fi in ISM Band
LTE (MHz)
Wi-Fi (MHz) 2310 2315 2325 2335 2345 2355 2365 2375 2385 2390
2412 >10 dB >10 dB >10 dB >10 dB >10 dB >10 dB >10 dB >50 dB >50 dB >50 dB
<50 dB <50 dB <50 dB <50 dB <50 dB <50 dB <50 dB
2422 <10 dB <10 dB <10 dB >10 dB >10 dB >50 dB >50 dB >50 dB
<50 dB <50 dB
2432 >50 dB >50 dB >50 dB
2442 >50 dB >50 dB >50 dB
2452, 2462, 2472 >10 dB >50 dB >50 dB
<50 dB
Tunable RF Components and Circuits: Applications in Mobile Handsets
Case Study of Tunable Radio Architectures 269
where the aggressor and victim are in close spectral proximity, the victim receiver
is desensitized by more than 50 dB. Note that the impact of the high OOB blocker
on the Wi-Fi receiver LNA (low-noise amplifier) is ignored in.9 In typical applica-
tions the LNA gain might be reduced to prevent it from saturating. This, in turn,
will desensitize the receiver. Alternatively, the OOB blocker could be removed or
canceled at the LNA input.
The OOB blocker and the in-band self-interference together reduce the effective sen-
sitivity level of the Wi-Fi receiver. As an example, consider the three scenarios depicted
in Figure 12.5. When the LTE transmitter and Wi-Fi receiver are operated simultane-
ously at the higher end of LTE band 40 and the lower end of the ISM band respectively,
as shown in Figure 12.5a, the filters provide little or no isolation. In such extreme sce-
narios, the Wi-Fi receiver must be able to handle high levels of all three self-interference
types. On the other hand, the Wi-Fi receiver may only need to deal with leaked LTE
transmitter noise for deployment scenarios like that shown in Figure 12.5c.
The required isolation budget allocations for the three scenarios depicted in
Figure 12.5 are summarized in Table 12.3. It is assumed that the LTE transmitter is
set to the maximum output power (+23 dBm) for all three scenarios. The goal is to
ensure that the Wi-Fi receiver desensitization caused by self-interference is limited
to 1 dB. Assume that the Wi-Fi receiver employs direct downconversion. It is known
that the OOB blocker will create unwanted distortion that ranges from DC to the
spectral width of the blocker at the Wi-Fi receiver’s mixer output. This distortion
is caused by even order distortive processes (e.g., quantified by IIP2) in the direct
downconversion Wi-Fi receiver’s front end (LNA, mixers) and falls on top of the
wanted signal at baseband. Additionally, note that the leaked nonlinear distortion
and noise from the LTE transmitter will also interfere with the wanted Wi-Fi signal.
Therefore, assuming that the two processes contribute equally, the unwanted inter-
ference caused by the OOB blocker and the leaked spurious emission interference
from the LTE transmitter must be 16 dB below the Wi-Fi receiver sensitivity level
(i.e., −90 dBm for a 20 MHz channel) if the desensitization is to be limited to 1 dB.
Assume that the Wi-Fi receiver front-end IIP2 is +14 dBm.10 Therefore, OOB blocker
at the input of the Wi-Fi receiver must be reduced to −46 dBm or lower.
In order to determine the required amount of OOB blocker cancellation, we con-
sider the attenuation provided by the Wi-Fi bandpass filter at the LTE transmitter
TABLE 12.3
LTE Band 40 to ISM Band Isolation Budget Allocations
Required
Required OOB Spurious Emission
Blocker Interference
Scenario ξAntenna (dB) ξLTE (dB) ξWi-Fi (dB) Cancellation (dB) Cancellation (dB)
Figure 12.5a 15 10 4 >50 >64
Figure 12.5b 10 39 >15 >44
Figure 12.5c 50 39 >15 >11
270 Tunable RF Components and Circuits: Applications in Mobile Handsets
signal’s center frequency. For the first scenario depicted in Figure 12.5a, the Wi-Fi
bandpass filter provides only 4 dB attenuation (see Table 12.1). Assuming that the
isolation between the LTE and Wi-Fi antennas is 15 dB, the leaked OOB blocker
level appearing at the input of the Wi-Fi receiver is +4 dBm. In order to limit the
second-order (IP2 related) distortion to −106 dBm (16 dB below the sensitivity level
of −90 dBm), the OOB blocker must be canceled to a level of −46 dBm if the Wi-Fi
receiver IIP2 is assumed to be +14 dBm. Therefore, the required OOB blocker cancella-
tion for the scenario depicted in Figure 12.5a is 50 dB or better. For the second and third
scenarios depicted in Figure 12.5b,c, the Wi-Fi bandpass filter provides better than
39 dB attenuation. Therefore, the required OOB blocker cancellation is 15 dB or better.
The leaked spurious emission interference at the input of the Wi-Fi receiver for
scenario 1 (Figure 12.5a) is determined by considering the LTE transmitter adjacent
channel leakage ratio (ACLR1). The standard’s specified minimum ACLR1 require-
ment is 33 dB.11 The ANADIGICS ALT674012 power amplifier (PA) achieves a typical
performance of 40 dB ACLR1 at maximum output power. Note that the LTE bandpass
filter provides 10 dB of attenuation in the Wi-Fi channel of interest for this scenario.
Therefore, the spurious emission interference from the LTE transmitter appearing at
the input of the Wi-Fi receiver is −42 dBm. The required spurious emission interfer-
ence cancellation is 64 dB. A similar approach may be used to determine the OOB
blocker and spurious emission cancellation required for the scenarios depicted in Figure
12.5b,c. Note that the spurious emission created by the LTE transmitter for scenario 2
shown in Figure 12.5b can be determined by considering the alternate channel leakage
ratio (ACLR2).11 The ANADIGICS ALT6740 PA12 achieves a typical performance of
60 dB ACLR2 at maximum output power. The LTE bandpass filter provides 10 dB of
attenuation in the Wi-Fi channel of interest for this scenario. The spurious emission
interference from the LTE transmitter appearing at the input of the Wi-Fi receiver
is −62 dBm. The required spurious emission interference cancellation is 44 dB. For
scenario three, the spurious emission created by the LTE transmitter consists of noise
only. The standard’s specified maximum LTE band 40 noise level in the ISM band is
−30 dBm.11 The LTE bandpass filter provides 50 dB of attenuation in the Wi-Fi channel
of interest for this scenario. The leaked noise in the ISM band is −95 dBn. The required
spurious emission interference cancellation is, therefore, 11 dB.
A conceptual block diagram of an RF system that might be deployed to handle
coexistence-related self-interference is shown in Figure 12.6. The system is in a state
where the LTE transmitter and Wi-Fi receiver are active at the same time. The vic-
tim (Wi-Fi) receiver system is shown to employ analog self-interference cancella-
tion (A-SIC) circuitry and adaptive digital self-interference cancellation (AD-SIC)
logic. These blocks are highlighted in light gray. Additionally, the context aware RF
resource manager (CARF-RM) is shown in dark gray. The CARF-RM is responsible
for configuring and dynamically adapting the canceler blocks in a context driven
manner. The primary purpose of the A-SIC is to remove the OOB blocker at the
input of the main (RX1) and supplementary (RX2) receivers. This prevents the low-
noise amplifier, mixer, and analog-to-digital converters employed by RX1 and RX2
from saturating. A copy of both the OOB blocker and spurious emission is made at
the output of the aggressor (LTE) PA and fed into the victim (Wi-Fi) receiver inter-
ference cancellation blocks as a reference. The A-SIC attempts to recreate a signal
Case Study of Tunable Radio Architectures 271
LTE Wi-Fi
antenna antenna
LTE Wi-Fi
I
TRX BPF TRX BPF
PA PA TX Q
Modem
TRX
+ l
∑ RX 1 Q
canceler (AD-SIC)
self-interference
Adaptive digital
–
Analog
self-interference
Splitter
canceler (A-SIC)
CARF-RM
–
∑ RX 2
+
A-SIC
Wi-Fi TRAX BPF
parameters AD-SIC parameters
A-SIC dynamic RFFE AD-SIC
A-SIC
adaptation algorithm
LUT
algorithms controller selector
that matches the leaked blocker signal for cancellation. The A-SIC design details are
found in references [13,14]. The A-SIC consists of a parallel bank of series connected
programmable delay and gain elements. The A-SIC parameters might consists of a
set of weights and delays, for example. A lookup table (LUT) in the RFFE controller
may contain the initial values of the A-SIC weights and delays for different aggressor
and victim carrier frequency combinations over temperature and aggressor trans-
mitter output power levels. Furthermore, the CARF-RM may employ algorithms to
dynamically adapt or perturb the A-SIC parameter set to optimize its performance
in response to changes in the UE’s operating environment.
The AD-SIC removes the residual in-band self-interference (nonlinear distortion
and noise) left by the A-SIC or when the A-SIC is inactive. Various adaptive filter
techniques (e.g., NLMS) are available for the AD-SIC implementation.15 The opera-
tional states of the A-SIC and the AD-SIC are summarized in Table 12.4. The can-
celers are transitioned from one state to another by the CARF-RM.
TABLE 12.4
Interference Canceler States for Various Self-Interference Levels
A-SIC AD-SIC
Settings Settings
Self-Interference Dynamic Performance
Level State Adaptation State Level
Level 1 ON Yes ON High
Level 2 No
Level 3 Low
272 Tunable RF Components and Circuits: Applications in Mobile Handsets
FIGURE 12.7 Messaging flow diagram for traffic off-load with Wi-Fi RAT.
Case Study of Tunable Radio Architectures 273
implemented within the modem protocol stack to extract messages that are relevant
to the RRC layer targeted for RFFE resources.
In order to enable interference cancellation, the CARF RM utilizes spectrum and
interference detector block as depicted in Figure 12.3 to measure the received signal
strength indication (RSSI) level where the Wi-Fi RAT will be deployed. Although
the Figure 12.7 illustrates the Spectrum and Interference Detector block utilization
for the RSSI measurements, the RSSI measurements may, also, be measured directly
by using the Wi-Fi protocol stack in test mode if the provisions are implemented.
The Wi-Fi modem can be instructed to run under initial state and the RSSI measure-
ments can be periodically generated.
The measured RSSI updates that feed to the cancellation algorithms as shown
in Figure 12.7 are continuously generated until either the A-SIC and/or AD-SIC
drives the system to an equilibrium that results in less than a target threshold or
the reaching of a timeout. The target threshold is chosen such that the Wi-Fi clear
channel assessment (CCA) logic may be successful to access the medium. The CCA
is one of two carrier sense mechanisms in Wi-Fi systems. It is defined in the IEEE
802.11 standards as part of the physical medium dependent (PMD) and physical
layer convergence protocol (PLCP) layer. The CCA utilizes two related functions—
carrier sense and energy detect. The CCA carrier sense (CCA-CS) is a mechanism
that enables the receiver to detect and decode a Wi-Fi preamble.16 If the CCA-CS
threshold is set to −87 dBm (maximum is −82 dBm for 20 MHz bandwidth), which
is +7 dB above −94 dBm that is the noise floor for a receiver utilizing 20 MHz
bandwidth with 7 dB noise figure by using Equation 12.1. The Wi-Fi modem by
using the PLCP header field extracts the time duration for which the medium will be
occupied. The CCA flag is held busy until the end of packet transmission when the
Wi-Fi preamble is detected. The CCA energy detect (CCA-ED) enables the receiver
to detect non-Wi-Fi energy in the operating channel and back off data transmission.
The CCA-ED threshold is typically defined to be 20 dB higher than the minimum
PHY Rx sensitivity. The higher CCA-ED threshold setting enables the Wi-Fi sys-
tems to be more resilient against non-Wi-Fi signals. However, if the in-band signal
energy crosses CCA-ED threshold, the CCA is held busy until the medium energy
is below the threshold.
In the case that the Wi-Fi modem activation is reached based on a timeout, it
should be assumed that the cancellation algorithms settled to the best possible out-
come within the prescribed time and that the measured RSSI value is bigger than the
threshold. The failure to settle to below the target RSSI value may be due to other
interfering sources rather than the LTE RAT interference measured in the Wi-Fi
channel.
Under normal Wi-Fi operations due to the device orientation and proximity as a
result of reflective objects being closer to the antennas, the amount of interference
experienced by the modem may cause increased medium access delay issues. While
the Wi-Fi modem is downloading or uploading data, the CCA failure condition may
trigger to run the cancellation algorithms. A way of approaching the issue is to run
the interference cancellation algorithms periodically even after seeing convergence
to the lowest RSSI value in the desired channel.
274 Tunable RF Components and Circuits: Applications in Mobile Handsets
12.7 CONCLUSIONS
The component-level enhancements and enabling technologies play an important
role allowing the device manufacturers to achieve goals to support ever-increasing
requirements motivated by the user demands for ubiquitous, instantaneous
over-the-air access to multimedia-rich content.
In order to address increasing data traffic demand and help reduce relevant
cellular core network congestion issues, the concept of traffic off-loading has been
introduced. The traffic off-loading is achieved by enabling simultaneous operations
of multiple RATs. However, the deployment of two separate RATs adjacent in space
and frequency may not always be possible. The operation of one RAT with high
transmit power levels can desensitize another RAT’s receiver in space-limited com-
pact devices when the spectral separation between the two are limited.
In this chapter, we have focused on a system approach, namely, the CARF, to
create a framework to enable intelligence in the RF front-end domain. The CARF
approach acquires the system-level requirements to optimize the modem’s RFFE
HW–related resource parameters. Also, we have defined the CARF RM entity to
facilitate the RFFE interaction with protocol stack, application layer, user prefer-
ences, and other system-level inputs. We have established, as an example, how the
cellular data traffic off-loading capability is enabled even when LTE and Wi-Fi
RATs are simultaneously operating in adjacent channels. The CARF RM combined
the appropriate interference cancellation methods with the feedback provided by the
spectrum and interference detector entity in order to reduce the LTE RAT interfer-
ence over the Wi-Fi channel.
REFERENCES
1. FCC, Third Memorandum Opinion and Order FCC 12-36A1, ET Docket No. 04-186
and ET Docket No. 02-380, 2012.
2. FCC, Further Notice of Proposed Rulemaking FCC 14-49, GN Docket No. 12-354,
2014.
3. FCC, Notice of Inquiry FCC 14-154, 2014.
4. Sankaran, C., Data offloading techniques in 3GPP Rel-10 networks: A tutorial, IEEE
Communications Magazine, 50(6), 46–53, June 2012.
5. 3GPP TR 36.816 V11.2.0, Study on signalling and procedure for interference
avoidance, 2011.
6. Hu, Z., Susitaival, R., Chen, Z., Fu, I.-K., Dayal, P., and Baghel, S., Interference avoid-
ance for in-device coexistence in 3GPP LTE-advanced: Challenges and solutions, IEEE
Communications Magazine, 50(11), 60–67, November 2012.
7. 3GPP TR 36.814 V9.0.0, Evolved Universal Terrestrial Radio Access (E-UTRA);
Further advancements for E-UTRA physical layer aspects, March 2010.
8. 3GPP TSG-RAN WG4 meeting #57 R4-104334, Analysis on LTE and ISM in-
device coexistence interference, Motorola, 2010. http://www.3gpp.org/DynaReport/
TDocExMtg--R4-57--28052.htm.
9. Bratislava, S., In-device coexistence interference between LTE and ISM bands, 3GPP
TSG-RAN WG4 Ad-Hoc Meeting #10-03 R4-102416, Qualcomm Incorporated, 2010.
http://www.3gpp.org/DynaReport/TDocExMtg--R4-56--28051.htm.
10. Emira, A.A. et al., A dual-mode 802.11b/Bluetooth receiver in 0.25 um BiCMOS, ISSCC,
San Francisco, CA, Vol. 1, pp. 270–527, 2004. Doi: 10.1109/ISSCC.2004.1332698.
Case Study of Tunable Radio Architectures 275
11. ETSI TS 136 101 V10.3.0, LTE; Evolved Universal Terrestrial Radio Access (E-UTRA);
User Equipment (UE) radio transmission and reception, 3GPP TS 36.101 version 10.3.0
Release 10.
12. Data Sheet—ANADIGICS ALT6740 power amplifier for LTE devices operating in
LTE band 40.
13. Bharadia, D., Mcmilin, E., and Katti, S., Full duplex radios, SIGCOMM, ACM
SIGCOMM Computer Communication Review, New York, Vol. 43(4), pp. 375–386,
2013.
14. Zhou, J. et al., Low-noise active cancellation of transmitter leakage and transmitter
noise in broadband wireless receivers for FDD/co-existence, IEEE Journal of Solid
State Circuits, 49(12), 3046–3062, December 2014.
15. Haykin, S., Adaptive Filter Theory, Prentice Hall, Upper Saddle River, NJ, 1996.
16. Prehia, E. and Stacey, R., Next Generation Wireless LANs, Cambridge University
Press, Cambridge, UK.
13 Network Operator
Perspectives
Yuang Lou
CONTENTS
13.1 LTE Deployments and Challenges............................................................. 278
13.2 Mobile Device Support to Advanced LTE Radio Platform.......................280
13.3 Business Opportunity with LTE Advancements........................................ 283
13.4 Challenges to Advanced LTE Mobile Broadband Devices........................284
13.5 Tunable Devices with MIMO Antenna Array........................................... 285
13.6 Reconfigurable Device Duplexer/Filter..................................................... 287
13.7 Evolved and Innovated Device Power Amplifier....................................... 288
13.8 Power Performance in Mobile Devices...................................................... 290
13.9 Improved Access Capability via Enhanced Network Architecture........... 292
13.10 Innovative Spectrum Policy Leads to Further Mobile FEM Evolution........ 293
13.11 Conclusions to the Mobile Device Evolution under Operator
Perspectives................................................................................................ 294
References............................................................................................................... 296
Long Term Evolution (LTE) mobile services started its deployment at the turn
of 2010 and 2011. By the end of 2013, there were 268 LTE networks launched in
100 countries with a total of 157.7 million LTE users1 worldwide. Among these,
112 million LTE users have been added in 2013 and 79 million of them were acti-
vated from the North America markets. This growth illustrates a strong growing
momentum for the LTE radio platform with an all-IP core network infrastructure to
improve and expand the mobile business opportunities globally.
Presently, mobile network operators are diligently exploring opportunities for
LTE deployments with higher business potentials and new services. The mobile
communication community wants to understand the foreseeable challenges in LTE
advancement and required regulatory innovation and technical support to enhance
LTE service deployment. Based on business expectations and market developments,
LTE RAN equipment and device vendors want to align their R&D activities and
technical investments not only to resolve the performance challenges and improve
the user quality of experience (QoE) but also to support both architecture evolutions
in future LTE RANs and mobile devices.
277
278 Tunable RF Components and Circuits: Applications in Mobile Handsets
Section Maps
Network Operator Perspectives
Sun Clock
4G - marketing term (not all 4G networks are created equal). Major 4G standards:
LTE (Long Term Evolution)—fastest of all 4G networks
HSPA+ (Evolved High-Speed Packet Access)—faster than 3G, however, slower than LTE
WiMAX (Worldwide Interoperability for Microwave Access)—approximately the same speed as HSPA+
(In many countries, LTE, WiMAX, HSPA+, 3G UMTS, GSM networks coexist)
FIGURE 13.1 Worldwide coverage of mobile communications services. (From http://www.worldtimezone.com/4g.html, updated February 12, 2014.)
279
280 Tunable RF Components and Circuits: Applications in Mobile Handsets
of the mobile networks and heavily impact the QoE performance of consumer data
access. In order to meet the challenge of 1000× growth in the user data traffic, small
cell–HetNet complemented with unlicensed Wi-Fi mobile access is an important
option in the network architecture.
Data off-loading is the only one part of developments in future mobile data access.
With data off-loading, a mobile user session may not be continued during the access
change. The device has to stop the existing session and start a new one. As long as
the switching latency from the data off-loading is short enough, users may not even
be able to sense a session interruption. On the other hand, the off-loading scheme
will challenge the QoE of mobile streaming services. The annual report shows that
50% of mobile traffic is from mobile video access.3 Mobile user behaviors have been
continuously shifting to IP video access. Instead of reading the news in text format
from a mobile device, news streaming through mobile audio/video delivery attracts
users’ attentions. For improved QoE, both audio and video streaming services
require seamless session continuity instead of switched data off-loading between the
licensed and the unlicensed network accesses. QoE enhancements, enriched mobile
services, and seamless mobile video streaming delivery drive the need for innova-
tions in spectrum regulation and technical advancements in user radio access.
In the past few years, service providers have made progress with new LTE service
offerings on top of the expanded coverage of LTE network access. For example,
the LTE SON deployment and mobile device access capabilities to Wi-Fi network-
ing9 were both initiated in the past few years; efforts to deploy small cell networks
and Wi-Fi off-loading started in 2013; at the beginning of 2014, mobile application
of TV everywhere10 enables live TV contents to be delivered to and displayed on
user’s smartphones; services of carrier aggregation (CA)11,12 and VoLTE13 have been
launched in markets one after another in the first half of 2014. Looking into the
future, it is not too hard for us to envision that there will be more mobile business
opportunities and deployed service features over LTE and LTE-Advanced platforms
to support medical applications carried over LTE, LTE-enabled broadband vehicle-
to-vehicle communications, dynamically reconfigurable CA bands that logically bind
more RF bands into one wider bandwidth operation, and so on. The LTE radio plat-
form becomes a flexible but core foundation to future mobile broadband data com-
munications. The IP multimedia subsystem (IMS)14 architecture bridges all of the IP
multimedia services. More software-defined capabilities will be deployed in future
LTE networks. Technically, software-defined network (SDN) characterizes the net-
work evolution for virtual network functions and dynamic network management.
One of the challenges to the advanced radio platform is how wide an RF band-
width that an LTE mobile architecture can support and where/how we can have a
wide and even wider RF bandwidth to support the operation. Frequency-selective
fading challenges an efficient and reliable network operation due to the existence of
an inherent irreducible error floor from the setup of wideband radio channels.
Compared to 2G15 and 3G16 mobile systems, higher level QAM modulation in
4G LTE17 across multiple resource blocks (RBs) raises the system peak-to-average-
power ratio (PAPR) challenging the battery power demand and supply in mobile
devices. OFDMA operation in the LTE downlink is not suitable to that of the LTE
device uplink transmission when an optimized and efficient device operation is
considered and configured.
Facing the aforementioned challenges, OFDM18-based LTE becomes an ideal
wideband radio platform. Its higher rate delivery can be achieved with simultaneous
transmission over many orthogonal RF subcarriers.19 Its OFDM architecture breaks
the wideband frequency-selective fading into many narrow but non-frequency-
selective and flat spectrum characterized subchannels. In this way, higher system
performance can be achieved through aggregated narrowband operations. A wider
RF bandwidth can be configured in a flexible and scalable fashion from 1.4 MHz up
to 20 MHz block. In the market places, a continuous 20 MHz RF block pair is rarely
owned by any FDD mobile network operators due to competitive spectrum auc-
tions. It forms another challenge to offer 20 MHz LTE bandwidth through the net-
work operation. The LTE-Advanced platform can support both logically combined
adjacent and even physically separated intraband and/or interband RF blocks into a
wider-channel bandwidth not only up to 20 MHz but also up to 100 MHz via CA.
CA resolution becomes a key advancement to expand wideband LTE operation while
more software-defined features are expected to run on top of it. For example, CA
capability can be defined by a specific band pair combination in 3GPP Band 17 and
band 4. Advanced CA is able to reconfigure more combined band pairs dynamically
according to the consumer demands and also depending on the available spectrum
resources in the field operation.
To properly cover the differences and complications between LTE base station
and device operations, LTE adopts the SC-FDMA20-based mobile platform in its
uplink operation, which runs a simplified device processing to limit the PAPR.21
It reduces the device battery power consumption as compared to that of downlink
OFDMA. Beyond the support of wideband operation, the LTE radio design is spe-
cifically focused on the enhanced implementation of antenna array processing in
terms of raising the RF spectral efficiency. MIMO22 transmitter and receiver over an
antenna array is specified in both downlink and uplink of LTE operations. Varied
MIMO device operations, adaptive modulation and coding schemes (MCS), and CA
make the 4G LTE platform fit in wideband communications much more efficient. In
essence, 4G LTE is characterized by an all-IP-based system with a flat network archi-
tecture, a dynamic RF link adaptation guided by the instant MCS selection, and an
IMS-based core to bridge the legacy mobile operations across multimedia services
with MIMO over both OFDMA downlink and SC-FDMA uplink radio platforms.
Looking at the current LTE deployments, we realize that, among many already
achieved performances, other advanced features need to be developed. For example,
282 Tunable RF Components and Circuits: Applications in Mobile Handsets
15 16.3
Spectral Downlink
(4 × 4 MIMO ) (4 × 4 MIMO )
efficiency
(b/s/Hz) 6.75 4.32
Uplink (2 × 4 MIMO ) (64QAM)
Data rate Downlink 1000 300
(Mbps)
Uplink 100 75
defined in the 3GPP, the goal of LTE RAN radios is to support 8 × 8 MIMO and the
LTE device is to support 4 × 4 MIMO operations, not to mention the performance dif-
ferences between LTE and IMT-Advanced shown in Figure 13.2 with LTE-Advanced
capable of meeting and surpassing the IMT-Advanced specifications. In current field
operations, deployed LTE RAN radios can only support 2 × 2 MIMO and the LTE
mobile device radios can only support 1 × 2 MIMO. It demonstrates a big perfor-
mance gap between the 3GPP goal and the reality from field operations while simul-
taneous radio transmissions in the same mobile device are still challenged by both the
battery power consumption and inter-radio interference. The goal to achieve higher
LTE spectral efficiency via a higher-order MIMO is still on the development agenda.
The latest smartphone design enables multi-active radios to support LTE and
Wi-Fi radio transmissions simultaneously. Further development shows that the
2 × 2 MIMO device chipset for IEEE 802.11ac has been commercially released23
in early 2014. The 2 × 2 MIMO and 4 × 4 MIMO LTE antenna arrays were already
prototyped in mobile devices.24 Their performances have even been evaluated and
compared through market validations. All of these achievements make the device
operation in higher-order MIMO characterized by the simultaneous LTE radio
transmission/receiving more attractive in commercial deployments. OFDM-based
multiuser-MIMO25 (MU-MIMO) opens another venue to raise the spectral efficiency
through the space-division multiple access (SDMA) in the network access. It also
requires mobile device support. Even though the initial MU-MIMO is only applied
to the access of IEEE 802.11ac, its capability should be extended to improve LTE
spectral efficiency as well.
Beyond raising the spectral efficiency to LTE cell coverage, cell edge perfor-
mance attracts the attention of network operators. This interest is strengthened by
small cell–HetNet deployments. Small cell coverage and macrocell coverage are
overlapped (see Figure 13.8), which raises the total network data capacity via a pre-
ferred data off-loading to the small cell access. Enhanced small cell operation at
the cell edge off-loads the traffic volume from the macrocell and thus opens more
macrocell resources to handle higher mobility accesses. Many more LTE-Advanced
features26,27 are going to be developed, such as SON, FeICIC, and coordinated multi-
point (CoMP), which further require support from advanced mobile devices.
Network Operator Perspectives 283
16
Source: Ericsson (June 2013)
14
Monthly exabytes (1018)
12
10
8
6
4
2
0
2010 2011 2012 2013 2014 2015 2016 2017 2018
FIGURE 13.3 Sixty percent of annual growth of mobile video consumption. (From http://
www.cedmagazine.com/news/2013/06/mobile-video-consumption-growing-at-60-percent/
year.)
284 Tunable RF Components and Circuits: Applications in Mobile Handsets
no viewers in the service area at all. In the past MediaFLO operations, the broadcast
spectrum, network infrastructure, and user mobile devices were all standalone and not
shared across any other mobile services. Merging mobile video delivery with LTE ser-
vices and sharing the LTE spectrum/devices can overcome economic barriers in offer-
ing mobile TV broadcasting. Dynamic reconfiguration via capable software control
should make eMBMS broadcasting more efficient and flexible together with LTE radio
resource management to meet the consumer demands. Shared spectrum, networking,
and mobile devices between eMBMS and LTE29,30 infrastructure with switchable uni-
cast, multicast, and broadcast service model will attract the future service development
while the supplemental downlink (SDL) spectrum has the potential to support both
eMBMS and CA operations. Mobile video delivery and LTE operation should not be
limited by the OFDM architecture but be open to a dynamic reconfigurable operation
on an OFDMA basis following market demands. This new model of mobile video
streaming and TV broadcasting as part of the LTE operation demonstrates the oppor-
tunities and also challenges to the network configuration and mobile device support.
The LTE network will continuously support the human connections via voice and
data communications but could also be extended to web-based medical service cov-
erage. It is evolving to support wireless machine-to-machine (M2M)31 communica-
tions for network monitoring and remote control under the Internet of Things (IoT).
It will also be enhanced to support mobile vehicle-to-vehicle (V2V) communications
for connected cars via LTE32 in order to strengthen the in-vehicle security, hands-
free call, turn-by-turn navigation, and remote diagnostic services. Within the IoT
access, a large percent of communication traffic will be between fixed equipment,
between mobile devices and fixed equipment, and also between mobile-to-mobile
devices. To support all of these low power but highly dense and instantly connected
devices under the LTE management is another challenge and opportunity to both the
innovative network configuration and mobile device architecture.
The success of Internet and cellular communication has been merged and raised
onto the track of cloud computing (CC) for data applications, video streaming deliv-
ery, and IoT services over strongly supported mobile platforms. Three major CC
developments, such as the cloud-based data storage, cloud-based online processing,
and cloud-based network services, will be accelerated under the user-initiated SDN
operation where the network equipment can be flexibly managed and reconfigured
by the virtualized functions and enriched service features with more intelligence but
less redundant hardware equipment. Building up a “User-Defined Network Cloud33
(UDNC)” becomes the goal of the mobile network operator who puts user satisfac-
tion at the top priority in network operations. Its MMMB device access and self-
reconfiguration performance directly impacts the user QoE and becomes one of the
keys to qualify the UDNC operation.
of each other from the antenna array of the air interface down to the RF transceiver
units. Separated RAN operations handle their own traffic flow, IP signaling34 to the
backhaul service, mobile radio access, and user data delivery.
Radio architecture in mobile devices is totally different from that of the RAN.
Service options coexist in the same mobile device to support the unique end point
performance of user QoE delivery. Capable mobile devices support MMMB opera-
tion. Supporting a higher number of MMMB access is one of the merits in the device
capability such that a customer-desired device should be able to access multiple ser-
vices from 4G LTE data, simultaneous 2G/3G voice call, GPS tracking, and the
delivery of mobile video streaming, to just name a few. Widely expected service
capability forms specific challenges to the LTE device antenna array, front-end mod-
ule (FEM), RF transceiver, mobile operations for interference control, and effective
battery life management.
In the United States, the RF spectrum for 2G mobile operation is defined from
850 MHz to 1.9 GHz. The 3G RF platform covers the same frequency range of 2G
but extends its high frequency end up to 2.1 GHz. The spectrum coverage of LTE is
much wider—from low 700 MHz to 2.6 GHz, which will be further stretched down
to 600 MHz and also extended up to 3.5 GHz. Wider RF coverage becomes one of
the characteristics of 4G operations. This wide spectrum coverage challenges the
configuration of device antenna to 4G LTE, which requires a multielement antenna
array to support the MIMO operation. Beyond a MIMO antenna array, more than
one antenna is also required to cover the RF range for different band access. On aver-
age, there are four to five antennas in an LTE smartphone to cover bands, modes,
and many other service options, such as GPS, Wi-Fi, FM radio, NFC, and 2G/HSPA/
LTE accesses. Due to limited device volume, the same cellular mobile antenna is
usually shared among 2G/3G and LTE bands. For example, 3GPP band 4 in 2.1 GHz
and band 17 in 700 MHz can share the same piece of antenna. Inserting an addi-
tional antenna normally gains in its Tx/Rx efficiency and RF isolation but is chal-
lenged by the device volume. Envelope correlation among array elements is sensitive
to the distance separation of antenna elements. Sharing the same antenna across
bands improves the device spacing efficiency but challenges the achievable RF per-
formance. Based on all the feasibilities and compromises, tunable antenna becomes
a key contributor to the device antenna design in volume, element spacing, and high-
RF performance. Its contribution is more important to lower frequency operation
where physics require a larger antenna size proportional to the wavelength in terms
of its better RF performance.
antenna design. The gain gap between low- and high-band antenna performance is
usually greater than 6 dB on the same piece of broadband antenna. The receiving
gains of MIMO array antenna elements should also be narrowed down. If the Rx
gain difference among the elements is >3 dB, it impacts the performance of MIMO
receiving process.
Improved performance over a wideband antenna is achievable through software
reconfiguration either via open- or closed-loop tuning. Open-loop tuned antennas
have already been deployed in several mobile devices. Varying input feeds could
be used to tune the antenna resonant frequency. Adjusting digital capacitance can
match the antenna impedance and thus raise the antenna efficiency. Up to this point,
open-loop tuned antenna demonstrates improved total radiated power/total isotropic
sensitivity (TRP/TIS) performance on a single antenna element.
Closed-loop tuned antennas are more complicated than that of the open-loop
ones. Closed-loop tuned antennas have to detect and track antenna detuning and
then match with the best impedance to compensate for the detuning. The process of
detection and best matching is adaptive and continuous. In order to reach large-scale
commercial deployment, solutions of closed-loop tuned antenna need further techni-
cal improvement and market validation.
The basic assumption to the current open- and closed-loop tuned antennas is that
there is only one RF band or mode actively transmitting and receiving among the
many available but inactive. To improve LTE MIMO performance, network opera-
tors are interested to see enhanced tuning in the antenna array instead of TRP/TIS
improvements over a single antenna element. To be more specific, a valid MIMO
antenna tuning will not only focus on the individual TRP/TIS across all antenna
elements, but also includes narrowed Rx gain difference and reduced envelop cor-
relation across all array elements together with the best TRP/TIS at the same time
and even across the CA bands. Less differences in receiving antenna gains across
all array elements improves the MIMO processing. A joint MIMO antenna tuning,
including reduced envelop correlation across array elements, increased RF isolation
and minimized RF interference among the elements together with the best achiev-
able TRP/TIS, boosts the overall performance on all MIMO antenna array elements.
Achieving joint performance of the antenna array is the goal of tunable MIMO
device antennas.
Current implementation of 1 × 2 MIMO antenna arrays in mobile LTE devices
needs qualified tuning support to cover both primary and secondary receiving
antenna elements. A higher-order MIMO chipset such as the 2 × 2 MIMO23 has been
released in the early 2014 to IEEE 802.11ac. Device array antenna architectures to
support 4 × 4 MIMO have gone through an initial market evaluation.24 Both of the
aforementioned higher MIMO cases are implemented in the GHz range. Without a
matured device antenna tuning resolution, it is hard to deploy 2 × 2 MIMO into the
lower frequency range, that is, to the device operation below 900 MHz. Table 13.1
demonstrates a business forecast in 2013 on tunable device antennas as an open but
growing opportunity.
Tunable antennas are active RF components attached to the device FEM. Beyond
the potential benefits from antenna tuning, the tuner itself could possibly generate
RF harmonics and intermodulation products that would impact spurious emission,
Network Operator Perspectives 287
TABLE 13.1
Business Forecast on Tunable Antennas
2011 2012 2013 2014 2015 2016
4G (LTE, LTE 17 81 203 324 476 635
adv) units (M)
Primary antenna — 0.60 0.90 0.80 0.73 0.56
tuner ($)
Diversity antenna — 0.00 0.50 0.50 0.45 0.42
tuner ($)
LTE-Adv — 0.00 0.00 0.00 0.60 0.70
incremental ($)
Diversity tuner — — 0.25 0.5 0.65 0.75
attach rate
Antenna tuning — 48.6 208.1 340.2 772.4 1061.1
business ($M)
increase insertion loss (IL), and narrow the antenna’s bandwidth. These side effects
can curb and challenge the benefits gained from the antenna tuning for flexible RF
band options and combinations in the mobile devices. Providing a full-scale evalua-
tion on the tunable MIMO antenna array is important to network operators for their
device performance support and also important for spectrum regulation. This is a
desired technical outcome but may still need more engineering efforts to accomplish.
GSM/Edge 850
GSM/Edge 950
GSM/Edge 1800
SP9T
GSM/Edge 1900
ASM
GSM/Edge 850/900 PA
GSM/Edge 1800/1900 PA
Power detect
Power detect PA
3GPP Band 1 PA
3GPP Band 2
3GPP Band 5
ET
RF switches
DPD
PAs
LNAs
BB RF XCVR RF FE
(a)
ET
RF switches
DPD PAs
LNAs
BB RF XCVR RF FE
(b)
RF switches ET
DPD PAs
LNAs
BB and RF XCVR RF FE
(c)
FIGURE 13.5 Evolved SoC options for CMOS-integrated device FEM, agile device
RFIC, and SDR: (a) current level of silicon integration for a typical MMMB handset
modem, (b) possible future level of silicon integration of MMMB CMOS SoC for handset
modem, and (c) “ideal” level of MMMB CMOS SoC for handset modem.
SDR is not a pure software solution. It requires mutual support from both hard-
ware and software. The hardware forms an available FEM platform to support the
software reconfiguration for a jointly achievable Q and OOBE in duplexer and PAM,
and is also equipped with temperature-compensated characteristics to meet the
bandwidth requirements particularly in the UHF operations in 600–700 MHz or
even below. More specifically, SDR can only be implemented in the digital domain
and relies on ADC/DAC signal conversion across domains. Current ADC power
consumption limits its performance in its direct sampling of analog RF signals. At
this point, a full-scale SDR expectation serves as a goal of evolution directing the
progressive developments of tunable and reconfigurable agile radios in the mobile
devices.
Antenna
Local wireless connectivity
RAM (Wi-Fi/Bluetooth/FM)
Processor, RF
Processor, cellular modem, modem, front RF front end needs to be
radio, memory, power transceiver end packed into ~15% of total
management, sensors, etc. area
(30%–40% of PCB area)
FIGURE 13.6 Typical layout of current smartphone design. (From Carson, P. and Brown, S.,
Microwave J., 56(6), 24, 2013.)
Network Operator Perspectives 291
Up to 30 Cell_FACH 7–13
disconnect
messages to messages to
connect
RRC
RRC
establish a establish a
connection connection
in 3G PCH/URA in LTE
FIGURE 13.7 Compared state models in 3G and LTE radio architecture mobile devices
with multi-core processing. (From Lou, Y., Evolved Mobile Broadband Services and
Devices, Presentation to IWPC Workshop on Optimizing Advance Smartphone and Tablets
Architecture, March 19, 2014, http://www.iwpc.org/ResearchLibrary.aspx?ArchiveID=210&
Display=doc#.)
more than quadrupled from 150 to 730 W*h/L38 through innovated technologies, the
power gap between the mobile device demands and improved battery supply remains
open and has even been widened. Parallel efforts to the device battery improve-
ments in W*h/L and in raised battery size are also shown in the developments from
improved device power management (PM), effective radio protocol control, and
drastically reduced device PA heat dissipation.
In order to shorten the response latency and improve the QoE in device access,
mobile devices have to maintain a proper power level attached to each radio state
during the device idle periods, which directly impacts device power consumption.
The power level of radio states and the switching timer are controlled by the network
scheduler on the RAN side. To understand the difference in radio states between
3G and LTE systems, Figure 13.7 shows a simplified LTE radio state design that is
much flatter with less layers to achieve efficiencies in radio resource control (RRC).
Connected state discontinuous reception (cDRX) is one of the LTE RRC features.
In the cDRX mode, an LTE mobile radio periodically suspends its RF transmission
and receiving but continuously holds the allocated RF resources and network con-
nection. Beyond gaining a shortened latency, an LTE mobile device radio is fast to
respond with less signaling flow and lowered battery power consumption. A study
shows39 that the cDRX is critical to support VoLTE applications for an effective
device battery power management.
Studies on mobile device power consumption show40 that about 50% of device
power is spent on the PAM operation. Most of the PAM power is wasted through
heat dissipation, which results in a limited PAM power-added efficiency (PAE). A
joint process of averaged power tracking (APT) and envelope tracking (ET) has been
developed41 to reduce the heat dissipation and raise the PAE through the PAM opera-
tion. An analog reference interface to ET (eTrak) has been specified in the MIPI
standard,42 which not only contributes to effective device power management but
also ensures extremely low-noise performance to reduce the RF interference to the
cellular radio receiver.
292 Tunable RF Components and Circuits: Applications in Mobile Handsets
Prior to 2011, the design of mobile devices was dominated by single-core CPU
architecture. Raising both the core processor rate in GHz and bias voltage to the
CPU core processor could increase the device processing power but challenge the
battery life since power consumption in the core processor is proportionate to its
GHz rate and the square of its bias voltage.
Raising the processing power without reducing the device battery life is one of the
expected goals in mobile data processing in order to enable multitasking capability
and also to enhance the audio- and video-related user applications such as the sup-
port of mobile gaming. Dual-core CPU architecture was first introduced into mobile
devices in 2011. Power consumption and coordination of multi-core processing were
the major concerns. Later studies show43,44 that, beyond gaining the processing power
via the multi-core CPU architecture, dual-core CPUs improve the device power bud-
get by 40% relative to that of the single-core device architecture. This achievement
is gained by not raising the CPU processor rate in GHz and not increasing the bias
voltage to the core, but distributing the processing workload evenly through a coor-
dinated parallel core processing.
Progressive developments are followed by more cores added to the device archi-
tecture. Quad-core CPU devices were initiated in 2012, and 8-core CPU with 64-bit
processing has appeared within high-end mobile devices in later 2014 and at the
beginning of 2015 such as the incoming Samsung Galaxy A7.45 The multi-core CPU
approach not only enables parallel processing with a relatively low rate proces-
sor support but also limits the bias voltage to the CPU processor. Along with the
advancement of multi-core CPU device architecture, additional cores are allocated
for the multi-core power management and processing coordination.
The development effort of multi-core processing has been supported by the
mobile’s OS. Current mobile OS is open to support enhanced data processing, effi-
cient core power management, and multitasking-enabled devices up to the 64-core
CPU architecture.46 It forms another branch to the evolution of future mobile
devices.
Macrocell
FIGURE 13.8 HetNets view: Small cell deployed within macrocell coverage.
RF interference and coordinate the small cell and macrocell accesses is a key to the
HetNet success (Figure 13.8).
Within the macrocell coverage, many small cells could be configured and
deployed sharing the same RF spectrum. It requires capable mobile devices to iden-
tify and reliably receive its own session data, but reject the RF interferences from the
neighboring user accesses and maintain a high quality of receiver performance. This
capability requires coordinated efforts between the mobile devices and the HetNet
access network in both time domain and frequency domain operations.
For example, intercell interference coordination (ICIC)47 is defined in the fre-
quency domain and mainly focuses on access interference control over the mobile
data channels. Enhanced ICIC (eICIC)48 and further enhanced ICIC (FeICIC)49 have
been defined in both time and frequency domains where the cross-carrier schedul-
ing and almost blank subframes (ABS) are focused on interference control over the
mobile signaling flows. The SON implementation of CoMP50 operation improves
the network capacity, flexibility, and reliability.
LTE license-assisted access (LAA)51 opens another venue to raise user access
capabilities from the evolution of network architecture. All of the aforementioned
efforts encourage mobile devices to off-load the traffic from the macrocell to the
small cell and unlicensed spectrum access once they are available. This can improve
data capacity, though the use of unlicensed radio bands comes with its own chal-
lenges and is not a substitute for additional licensed RF spectrum.
Along with the growth of mobile data and video traffic, additional mobile traffic
volumes will come from the operation of IoT over wireless. The IoT industrial M2M
communications54 will have a large portion of rising data traffic including the fixed-
to-fixed and fixed-to-mobile device traffic, which might be configured under the
licensed/unlicensed data coverage. Beyond the M2M, the efforts and investments to
build the infrastructure for V2V communications have seen fast progress. As mobile
services shift from data access to video display and mobile TV broadcasting, the
balance strategy of mobile access has evolved from licensed LTE to unlicensed data
off-loading, and then to the required session continuity in support of the streaming
content delivery.
All of the aforementioned evolutions push mobile devices to support the MMMB
access capabilities under a new RF spectrum environment. The device access capa-
bilities will not only include more bands and modes through 2G, 3G, and 4G LTE,
but also covered by the FDD/TDD coexistence across the licensed and unlicensed
spectrums. With SDN virtualization in the network configuration, the trends of SDR
lead the mobile device evolution. Under a given device volume and limited battery
power supply, a configurable and reconfigurable device FEM becomes the most criti-
cal area for technical innovations and developments.
Tuned device RF FEM has already emerged in device antenna development. It
enables the same piece of device antenna to cover a wide RF range operation with
one band active at a time. It will be evolved to track and compensate for a detuned
antenna in a dynamic fashion and also tune the RF performance from a single
antenna to a MIMO antenna array.
Fully reconfigurable and integrated device FEMs are the next phase to this evolu-
tion. RF CMOS developments have been continuously improving the performances
in terms of integrated device FEM as its main target. The RF CMOS architecture
covers the RFIC and has been extended to cover MEMS developments in the RF
switches and PAM developments.
Simultaneous radio operation will be a key milestone to raise the device effi-
ciency in support of higher-order MIMO and multiradio access. It is challenged by
the coordination of device RF interference and effective battery power management.
Phased progress characterizes device development from the near-term to the long-
term evolution.
In the current device FEM architecture, the RF transceiver function is covered by
analog operations where the RF signals are filtered, amplified, and/or modulated/
demodulated by the analog circuits in the device FEM. SDR implementation is
purely in the digital domain. The demarcation between analog and digital operations
is at the ADC/DAC conversion. Raising the ADC RF sampling rate and maintaining
the ADC performance at a high-bit resolution and large dynamic range still funda-
mentally challenges SDR implementation without a trade-off to device battery life.
This becomes one of the expectations of the network operators. Working with the
spectrum regulatory administration and the vendor community, we are at the begin-
ning of this long-term journey. The challenges are highly technical, but the innova-
tions are user QoE directed. All of the technical innovations and advancements are
driven by market demand.
296 Tunable RF Components and Circuits: Applications in Mobile Handsets
REFERENCES
1. www.keysight.com/upload/cmc_upload/All/23Jan14WebCastSlides.pdf; www.gsacom.
com, updated February 5, 2014.
2. http://www.worldtimezone.com/4g.html, updated February 12, 2014.
3. http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-
index-vni/white_paper_c11-520862.pdf.
4. http://www.youtube.com/watch?v=tIv0jhMa64E.
5. http://www.4gamericas.org/documents/2013_4G%20Americas%20Meeting%20
the%201000x%20Challenge%2010%204%2013_FINAL.pdf.
6. http://stakeholders.ofcom.org.uk/consultations/ddr/statement/.
7. http://www.qualcomm.com/research/projects/lte-advanced/asa.
8. http://www.gsma.com/spectrum/wp-content/uploads/2013/04/GSMA-Policy-Position-
on-LSA-ASA.pdf.
9. http://www.fiercewireless.com/tech/story/atts-rinne-using-son-helps-improve-through-
put-and-reduce-dropped-calls/2012-10-30.
10. http://www.multichannel.com/distribution/att-u-verse-expands-live-tv-lineup-tablets-
smartphones/148633.
11. http://www.phonearena.com/news/AT-T-uses-carrier-aggregation-to-quietly-bring-
LTE-Advanced-to-the-Windy-City_id53600.
12. ht t p: // w w w.q u a lc o m m .c o m /s olut io n s / w i r ele s s - n e t wo r k s / t e c h n olog ie s /
carrier-aggregation.
13. http://www.computerworld.com/s/article/9248366/AT_T_to_activate_HD_Voice_
over_4G_LTE_in_four_states_on_May_23?source=rss_latest_content.
14. Camarillo, G. and García-Martín, M.-A., The 3G IP Multimedia Subsystem:
Merging the Internet and the Cellular Worlds, John Wiley & Sons, 2008, ISBN-13:
978-0-470-69513-5.
15. Narang, N. and Kasera, S., Architecture, Protocols, and Procedures and Services, In:
2G Mobile Networks: GSM and HSCSD, Tata McGraw-Hill Publishing Company
Limited, 2007, ISBN 0-07-062106-3.
16. Korhonen, J., Introduction to 3G Mobile Communications, 2nd edn, Artech House,
Inc., 2003, ISBN 1-58053-507-0.
17. Kyriazakos, S., Soldatos, I., and Karetsos, G., 4G Mobile & Wireless Communications
Technologies, CTiF (Aalborg University), Athens Information Technology, Technology
Research Center of Thessaly, River Publishers, 2008, ISBN: 9788792329028.
18. http://en.wikipedia.org/wiki/Orthogonal_frequency-division_multiplexing.
19. Wang, Z. and Giannakis, G.B., Wireless multicarrier communications: Where Fourier
meets Shannon, IEEE Signal Processing Magazine, 17(3), 29–48, 2000.
20. http://en.wikipedia.org/wiki/Single-carrier_FDMA.
21. http://people.rit.edu/grteee/media/IST03.pdf.
22. http://en.wikipedia.org/wiki/MIMO.
23. http://appleinsider.com/articles/14/02/24/broadcoms-new-5g-wi-fi-chip-with-
80211ac-2 × 2-mimo-could-be-tapped-for-future-iphone.
24. SkyCross presentation: Advances in MIMO for LTE-A to IWPC Workshop, San Jose,
CA, March 20, 2014.
25. http://www.anandtech.com/show/7921/qualcomm-announces-mumimo-80211ac-
family-increasing-the-efficiency-of-80211ac-networks.
26. http://en.wikipedia.org/wiki/LTE_Advanced.
27. http://www.4gamericas.org/index.cfm?fuseaction=page§ionid=352.
28. http://reviews.cnet.com/8301-13970_7-57619375-78/at-t-expands-international-
lte-roaming-to-13-more-countries.
Network Operator Perspectives 297
55. http://www.home.agilent.com/upload/cmc_upload/All/23Jan14WebcastSlides.pdf.
56. http://www.cedmagazine.com/news/2013/06/mobile-video-consumption-growing-at-
60-percent/year.
57. Carson, P. and Brown, S., Less is more: The new mobile RF front-end, Microwave
Journal, 56(6), 24–34, 2013.
58. Lou, Y., Evolved Mobile Broadband Services & Devices, Presentation to IWPC
Workshop on Optimizing Advance Smartphone and Tablets Architecture, March 19,
2014, http://www.iwpc.org/ResearchLibrary.aspx?ArchiveID=210&Display=doc#.
14 Testing Wireless Devices
in Manufacturing
Rob Brownstein and Minh-Chau Huynh
CONTENTS
14.1 Introduction................................................................................................... 299
14.2 Wi-Fi and 4G (3GPP LTE)............................................................................300
14.2.1 IEEE 802.11n..................................................................................... 301
14.2.2 4G LTE..............................................................................................304
14.3 Manufacturing Test Methods and Apparatus................................................307
14.4 Wireless Testing of Wireless Devices............................................................309
14.5 Implications................................................................................................... 311
14.6 OTA Method for Antenna Tuning................................................................. 312
14.7 Conclusions.................................................................................................... 316
14.1 INTRODUCTION
Only a decade and a half into the twenty-first century, wireless devices abound.
From wireless car keys that allow you to open a car door and start the engine without
leaving your pocket to smartphones that let you make calls, surf the Internet, take
and transfer photos and videos, and pay for items electronically, people have essen-
tially tether-free access to each other and a multitude of services.
Today’s smartphone eclipses in complexity computing and communications
devices that were densely packed with components, took up cubic yards of space, and
cost tens of thousands of dollars. And, this, by a device one holds in a hand that con-
tains computing, networking, and wireless communications subsystems and costs
only a few hundred dollars to manufacture. It is no wonder, then, that millions of
these devices are produced and sold each month. The amazing part is that despite the
complexities, such devices can be mass produced in huge volumes, with high yields.
Furthermore, manufacturing testing has become so efficient and fast that essentially
every device can be tested to help ensure only a tiny fraction of faulty devices end up
in the hands of customers.
Because such wireless devices are combinations of computers, network clients,
and wireless transceivers, testing them during manufacturing entails making sure
that they work as designed. Testing costs are a combination of capital equipment,
facilities, power, and labor costs. There is a direct correlation between testing cost
and testing time. With very tight testing cost budgets due to price competition and
thinning margins, there is huge pressure on companies that design and sell test
299
300 Tunable RF Components and Circuits: Applications in Mobile Handsets
solutions to wireless device manufacturers to decrease testing cost per device even
as the number of functions to be tested increases significantly.
As with any computer, sophisticated wireless devices such as smartphones and
cellular-enabled tablets have central processing units (CPUs), operating systems,
storage, and I/O. For self-contained applications, such as music playback, taking
photos or videos, or playing games, there are manufacturing tests for ensuring that a
device boots up properly and that all related subsystems, I/O, and transducers (e.g.,
touch screen, LCD display, speaker, microphone) are all working as designed.
When these devices must upload or download data to other devices, they do so
wirelessly. Typically, a smartphone or tablet will have Bluetooth for hands-free
calling in an automobile. They will also have Wi-Fi for using the Internet directly,
rather than through the metered cellular service for Internet access. This, of course,
requires being close enough to a Wi-Fi hotspot to accommodate such use. Cellular
system access typically requires one or more radio-access technologies, such as 3G
and 4G (Third Generation Partnership Project Long Term Evolution [3GPP LTE]).
These will have a wireless subsystem associated with them, too, such as a transmit-
ter and receiver. Thus, testing of the wireless functions of a device will involve tests
different from those for the computing subsystems.
This chapter focuses on the testing of wireless functions during manufacturing.
A device, for example, that features 3G and 4G (3GPP LTE), 802.11n Wi-Fi, and
Bluetooth technologies will be tested to ensure that each one is still operating prop-
erly after manufacture. The assumption is that the design has been verified, earlier,
and that a validated design has moved to production. In other words, manufacturing
test is not used to verify a design, but, rather, to identify any manufacturing defects
that may have occurred and which now render the device’s wireless facilities either
faulty or out of specification.
Several technologies are common to cellular and so-called connectivity wireless
technologies, such as Wi-Fi. These common denominators include orthogonal fre-
quency division multiplexing (OFDM), quadrature amplitude modulation (QAM),
and multiple-input multiple-output (MIMO). In addition, these wireless technologies
require antennas. To fully test a device that replicates real-world operation, the fully
assembled device should be linked to a tester in the same way it would to an access
point or base station, using over-the-air (OTA) signals. Today, this real-world test
typically occurs last. Before that, the partially assembled device has been tested
using non-linked processes for virtually all the specifications and functions; and it
is the partially assembled device that has passed these tests prior to the final fully
assembled device test. There are many reasons for why this process is in place, and
they will be described. Recently, innovative technologies have been developed that
make testing fully assembled devices using OTA signal transfer more practical than
before. These, too, will be described and explored.
14.2.1 IEEE 802.11n
Prior to amendments underlying IEEE 802.11n, wired Ethernet speeds were shoot-
ing past those of wireless Ethernet. In order of their adoption, IEEE 802.11b, a, and
g established a beachhead for wireless Ethernet, but 802.11n really set the stage for
significant performance improvements.
Several changes underpin 802.11n’s performance enhancements. By adopting
OFDM, higher data throughput compared with the previous a, b, and g versions of the
standard is achieved by having a networking system all using the 802.11n s tandard.
A new PHY layer convergence protocol (PLCP) supporting the 802.11n standard with
no legacy provisions (called Greenfield) enables highest data throughputs. A PLCP
called “mixed mode” uses a preamble consistent with 802.11 a and g but incorporates
some performance enhancements, such as a MIMO training sequence. This allows a
performance improvement on a networking system that includes c lients with 802.11
a or g technologies. The third PLCP, legacy mode, simply duplicates 802.11 a or g
modes. Which PLCP modes are brought to bear is determined by the makeup of the
networking system and its clients, and the mode being used determines the char-
acteristics of the PHY layer signal formats being exchanged. Note, also, that the
radio frequency (RF) bands prescribed cover the 5 GHz band of 802.11 a and the
2.4 GHz band of 802.11 g.
Another performance booster in 802.11n is the use of MIMO. Compared to
802.11 a and g, 802.11n can improve range using spatial diversity and can improve
data throughput using spatial multiplexing. Both improvements are a result of using
MIMO technology. With 802.11n, up to four spatial streams can be used, which
would require a device having four transmitters, four receivers, and four antennas
(a 4 × 4 × 4) engaging with an access point having the same capabilities.
By supporting 20 or 40 MHz channel widths, one can trade-off the number of
channels to increase data throughput. Also, IEEE 802.11n supports beamforming
when using two or more transmission subsystems and antennas. Beamforming opti-
mizes the path between the wireless device and the access point, reducing the effect
of multipath destructive interference.
The IEEE 802.11n standard document, known as Part 11: Wireless LAN Medium
Access Control (MAC) and Physical Layer (PHY) Specifications; Amendment 5:
Enhancements for Higher Throughputs, is more than 500 pages long and covers just
the Wireless MAC and Physical Layer (PHY) Specifications. The key portions of the
standard that drive the manufacturing testing of 802.11n devices are: 20.3.21 PMD
transmit specification and all its subsections, and 20.3.22 HT PMD receiver specifi-
cation and all its subsections.
For example, in 20.3.21.1, Transmit Spectrum Mask, which underpins a trans-
mitter (TX) test, the limits are established for devices using 20 MHz and 40 MHz
channels. In the first case, the standard states:
302 Tunable RF Components and Circuits: Applications in Mobile Handsets
When transmitting in a 20 MHz channel, the transmitted spectrum shall have a 0 dBr
(dB relative to the maximum spectral density of the signal) bandwidth not exceed-
ing 18 MHz, −20 dBr at 11 MHz frequency offset, −28 dBr at 20 MHz frequency
offset, and the maximum of −45 dBr and −53 dBm/MHz at 30 MHz frequency offset
and above. The transmitted spectral density of the transmitted signal shall fall within
the spectral mask, as shown in Figure 14.1. The measurements shall be made using a
100 kHz resolution bandwidth and a 30 kHz video bandwidth.
And Figure 14.1 depicts the spectral mask for 20 MHz channel testing.
The document goes on to state that for 40 MHz channels,
When transmitting in a 40 MHz channel, the transmitted spectrum shall have a 0 dBr
bandwidth not exceeding 38 MHz, −20 dBr at 21 MHz frequency offset, −28 dBr at
40 MHz offset, and the maximum of −45 dBr and −56 dBm/MHz at 60 MHz fre-
quency offset and above. The transmitted spectral density of the transmitted signal
shall fall within the spectral mask, as shown in Figure 14.2.
Figure 14.2 depicts the spectral mask for 40 MHz channel testing.
PSD
0 dBr
–20 dBr
–28 dBr
–45 dBr
Frequency (MHz)
–30 –20 –11–9 9 11 20 30
PSD
0 dBr
–20 dBr
–28 dBr
–45 dBr
Frequency (MHz)
–80 –40 –21 –19 19 21 40 60
0
The 802.11n standard also establishes spectral flatness limiting the delta in aver-
age power for the subcarriers for devices that use 20 MHz and/or 40 MHz channels.
In the 20 MHz channel, there are two specifications. In subcarriers that correspond
to constellation indices −16 to −1 and +1 to +16, the average power cannot deviate
more than ±2 dB. For subcarriers corresponding to constellation indices −28 to −17
and +17 to +28, average power cannot deviate more than +2/−4 dB from the aver-
age power measurements of the first set of subcarriers. Maximum transmit power
is determined by the regulations in each country and the standard provides some
tables for the United States, Europe, and Japan in the annex section (Annex I) of the
standard.
Center frequency tolerances are established for devices operating in the 2.4 GHz
spectrum and for those operating in the 5 GHz spectrum. Interestingly, only ±20 ppm
is allowed for the higher frequency spectrum versus ±25 ppm for the 2.4 GHz one.
The numbers are identical for signal clock frequency tolerances.
The 802.11n standard specifies the limit of error in constellations. Ideally, the
subcarriers would have perfectly orthogonal positions relative to each other. In
reality, of course, they do not. The limit to how much they may deviate is related
to constellation size and coding rate as, for example, BPSK with coding rate ½ is
limited to −5 dB of constellation error, whereas 64 QAM with coding rate of 5/6 is
limited to −29 dB of constellation error.
Transmit modulation quality is determined by error-vector magnitude (EVM)
metrics. Here is how the standard describes this test procedure:
The standard goes on to describe a step-by-step procedure or its equivalent, for mea-
suring EVM.
For receiver (RX) specifications, one refers to 20.3.22. The first specification is for
receiver minimum input sensitivity and it is determined by a packet-error rate (PER)
limit of 10%. Minimum sensitivity for 20 MHz channels using 64 QAM modulation
at a rate of 5/6 is −64 dBm, and for a 40 MHz channel, −61 dBm.
Note, also, that the standard specifies the limits for adjacent and nonadjacent
channel rejection that are allowed. Rejection decreases with the higher throughput
QAM modulation compared with binary phase-shift keying (BPSK) and quadrature
phase-shift keying (QPSK). Rejection is determined by setting a channel’s level at
3 dB above the minimum sensitivity value and then increasing the power of the other
channel’s signal until a 10% PER occurs in the first channel.
In addition to a minimum sensitivity level, the specification also sets a maximum
level above which saturation could occur causing the PER to exceed 10%.
304 Tunable RF Components and Circuits: Applications in Mobile Handsets
14.2.2 4G LTE
Wi-Fi from its inception was, like the Internet, meant to be part of a global digital
network. As such, the IEEE 802.11 standards have been adopted virtually every-
where in the world. In fact, using voice-over-IP, one can even use Wi-Fi to facili-
tate global telephone communications using applications such as Skype. Of course,
doing so requires one to be able to link to a Wi-Fi hotspot.
By design, cellular services offer a much broader service area than Wi-Fi.
Elaborate systems of cellular base stations connected to wired telephone infrastruc-
ture allow users to be connected wherever cellular service is accessible. The first cel-
lular standards, such as AMPS, were analog systems. Later, the first digital t elephony
standards and services emerged. These were termed second-generation systems,
or 2G. More sophisticated digital services, including both telephone and data, were
the hot buttons of third-generation standards, also known as 3G. And, today, we have
3GPP LTE standard, which much like 802.11n, has made cellular data services more
robust and higher performance.
The standard that underlies 4G LTE manufacturing testing is 3GPP TS 36.521-1.
Like IEEE 802.11, it has undergone a succession of versions, each amended with
new and enhanced elements. LTE, for example, was introduced initially in Release 8.
It introduced a radically different and new radio interface and network. Downlink
speed maximum was 300 Mbps and maximum speed for uplink was 75 Mbps. Signal
bandwidths of 1.4, 3, 5, 10, 15, or 20 MHz were supported allowing for different
implementation scenarios.
One of the fundamental decisions was to use different access schemes for down-
load and upload. Downlink uses orthogonal frequency domain multiple access
(OFDMA), whereas uplink uses single-carrier frequency domain multiple access
(SC-FDMA). The benefit of SC-FDMA uplink access is better energy efficiency
because of lower average output power of a device’s transmitter power amplifier.
As with 802.11n, LTE takes advantage of MIMO technology for both spatial diver-
sity (e.g., range extension) and spatial multiplexing (e.g., higher throughput). In LTE,
the centralized control of earlier networks (e.g., GSM BSC or UMTS RNC) was
eliminated in favor of a flat distributed architecture (see Figure 14.3).
Further standard releases have followed Release 8 up to and including Release 12,
which was frozen in 2014. As a result, the initial downlink and uplink maximum
speeds have increased to 3 and 1.5 Gbps, respectively (Release 10). In addition, a
feature called “carrier aggregation” has been added that allows up to five carriers to
be aggregated achieving an effective bandwidth of 100 MHz (Release 10).
As with IEEE 802.11n, the PHY layer specifications of 4G LTE drive the PHY
layer manufacturing test of devices that employ it. Thus, for example, physical
layer testing focuses on the lowest layer of the air interface. It seeks to determine
conformance with the key parameters essential to the successful transmission of
a signal over the air. Transmit power, the quality of the TX waveform, the accu-
racy of the TX frequency, are all key to a mobile station’s performance. On the
receive (RX) side, the ability of the mobile to successfully decode the received
signal at the lowest and highest signal levels defines its successful operation in
the network.
Testing Wireless Devices in Manufacturing 305
Packet switched
EPC
S1 interface
eNB X2
eNB
X2
X2 interface
eNB
FIGURE 14.3 4G LTE replaced the centralized architecture of previous standards with a
flat, distributed architecture.
As such, the pertinent specifications are TX power, EVM, frequency error, adja-
cent channel leakage ratio (ACLR), spectrum emission mask, occupied bandwidth,
carrier leakage, TX time mask, and in-band emissions from non-allocated resource
blocks. In many ways, the tests at PHY layer of 4G LTE signals are very similar to
test for the analogous characteristics of an 802.11n signal. To be sure, because of
the complexity of 4G LTE relative to 802.11, there are many tables in the specifica-
tions covering power metrics where there is no carrier aggregation or UL MIMO
compared to cases where there are. The tables that are most relevant to PHY layer
testing are maximum output power, which corresponds to Section 6.2.2 of the speci-
fications, and minimum output power, which corresponds to Section 6.3.2. As noted,
later, a device that passes those two TX power tests can be considered compliant
with the other tables and limitations because these values are controlled by software
and subsystems in the device that are unlikely to be faulty if the basic maximum and
minimum power tests indicate correct operation.
With regard to EVM, the 4G LTE specifications are quite explicit. For example,
the specification states:
The RMS average of the basic EVM measurements for 10 sub-frames excluding any
transient period for the average EVM case, and 60 sub-frames excluding any transient
period for the reference signal EVM case, for the different modulations schemes shall
not exceed the values specified in Table 6.5.2.1.1-1 for the parameters defined in Table
6.5.2.1.1-2. For EVM evaluation purposes, [all PRACH preamble formats 0-4 and] all
PUCCH formats 1, 1a, 1b, 2, 2a and 2b are considered to have the same EVM require-
ment as QPSK modulated.
306 Tunable RF Components and Circuits: Applications in Mobile Handsets
For QPSK and BPSK, the average EVM level cannot exceed 17.5%; and for 16 QAM,
it is a more stringent 12.5%.
Frequency error over a period of one time slot (0.5 ms) is limited to ±0.1 ppm
compared to the carrier frequency received from the network node.
There are related specifications that are essentially looking for the same potential
problem—adjacent channel interference. The adjacent channel leakage ratio looks at
the amount of power found in an adjacent channel due to a signal in the other chan-
nel. The 4G LTE standard specifies two different cases, one where the channels are
adjacent Evolved Universal Terrestrial Radio Access (E-UTRA) channels and one
where they are UTRA channels. Most LTE conformance tests for out-of-band emis-
sions are similar to those used for W-CDMA. These are consistent with pre-LTE
3GPP releases, where UTRA channels are employed, where a root-raised cosine
filter is used for making transmitter measurements. LTE, on the other hand, is not
constrained to root-raised cosine filters in its E-UTRA channels, which result in
better ACLR characteristics. Hence, we have two different specifications.
Essentially, the ratio is of the filtered mean power of the signal centered in
the assigned channel to that measured centered in the adjacent channel. Another
metric—occupied bandwidth—looks at the bandwidth of a signal in an assigned
channel and compares the bandwidth in which 99% of the total integrated power
resides to that of the channel. It is checking to make sure that the signal occupies less
than the channel bandwidth. And the spectrum emission mask specifies the roll-off
of power on either side of a signal. It is analogous to the spectral mask test used in
testing Wi-Fi. If a device’s TX signal passes the mask test, it is unlikely to fail the
ACLR test.
Carrier leakage refers to carrier signal present in the signal output. In a prop-
erly matched transmitter, the IQ modulation suppresses the carrier energy to a point
where it is attenuated far below the normal signal’s energy. The carrier leakage speci-
fication for 4G LTE specifies different leakage levels at different signal power levels.
The TX time mask test looks at the signal in time, verifying that the transmit-
ter power amplifier is turning on and off at the correct time without producing any
extraneous signals. Since LTE signals are shared both in frequency and time, being
accurate in the time domain is just as important as being accurate in the frequency
domain.
The in-band emissions test for non-allocated resource blocks (RBs) is meant to
ensure that RB subcarriers in a non-allocated RB are not emitting more than a very
limited energy, in band. It is a ratio of the power of non-allocated RBs to allocated
RBs over a time slot and averaged across 12 subcarriers.
These 4G LTE TX specifications (e.g., uplink) are the basis for the test flow in
LTE TX manufacturing testing. Similarly, the following 4G LTE RX specifications
drive the test flow in LTE RX manufacturing testing.
The standard states that unless otherwise stated, the receiver characteristics are
specified at the antenna connector(s) of the user equipment (UE). “For UE(s) with
an integral antenna only, a reference antenna(s) with a gain of 0 dBi is assumed for
each antenna port(s). UE with an integral antenna(s) may be taken into account by
converting these power levels into field strength requirements, assuming a 0 dBi gain
antenna. For UEs with more than one receiver antenna connector, identical interfering
Testing Wireless Devices in Manufacturing 307
signals shall be applied to each receiver antenna port if more than one of these is
used (diversity).” Thus, the signals specified are either applied directly to an antenna
connector, or, where an antenna is integral to the device, it is assumed to be a 0 dBi
antenna, that is, having no directional gain. In addition, the specifications assume
a MIMO baseline of 2. That is, the device is assumed to have two RX signal ports.
With Wi-Fi, RX tests, by and large, are performed against a PER specification.
For example, minimum and maximum receiver (RX) power sensitivity is deter-
mined when PER exceeds 10%. With 4G LTE, RX test specifications are based on a
reference sensitivity power level. Maximum and minimum power levels in 4G LTE
are not determined by error rate, irrespective of data rate, but rather by the power
levels at which a device can achieve 95% or better of the maximum throughput. The
standard contains different tables for cases where there is no carrier aggregation and
cases where there is carrier aggregation.
To be sure, the specifications contain a lot of limits, tables, and the like with
regard to RX performance. These include reference sensitivity power level, maxi-
mum input level, UE maximum input level with carrier aggregation, adjacent chan-
nel selectivity, blocking characteristics, and spurious emissions. However, assuming
the device is properly designed and validated, these measurements are not likely to
fail manufacturing testing so long as maximum and minimum power level tests have
been passed. That allows for limiting test steps without risking false negatives.
measuring spectral mask, maximum and minimum power, and EVM with the tes-
ter’s vector analysis subsystem.
During device design verification, virtually all of a standard’s specifications are
tested to make sure the device has been designed properly. The benefit of this is
that the manufacturing test steps and flow can be a subset of verification tests that
determine whether any manufacturing defects have rendered a device faulty or out of
specification without having to reaffirm that all the specification limitations are still
met. Thus, testing can be confined to the PHY layer subsystems without undue fear
that some other failure will go undetected.
The “art” in developing manufacturing test solutions is not necessarily in the
design of the vector signal generation or analysis subsystems, or the control systems.
The art is in developing the test control interface that enables the tester to control the
device through the device’s chipset. Even where two different devices have the same
chipset, the test programs for each device may be significantly different because of
other differences in features and technologies.
In looking at a wireless test regimen, one must include the time spent connect-
ing and disconnecting devices to the test system, the time it takes to boot up the
device, and the time it takes to conduct the actual TX and RX tests. Since 2005, with
Wi-Fi and Bluetooth, test engineers have reduced the time it takes to conduct the
actual TX and RX tests to a point of diminishing return. Further reduction in test
time and costs has been gained by innovations in device-under-test (DUT) handling
as well as reduction of DUT–tester communications so that a larger proportion of the
time spent sending signals back and forth is test, rather than control, related.
As a single device’s overall test time has been reduced to diminishing return
dimensions, further time savings have been gained by testing devices concurrently.
If one can test, say, four devices in a time that is only 20% longer than testing a single
device, the overall test time saving is huge—nearly 75%. If one can test a single
device’s multiple technologies in a quasi-concurrent fashion, one can again achieve
significant time savings. Where the objective is to test millions of devices per month
that test time saving can be the difference between hitting a market window or not.
With intense price competition on many wireless devices, the faster test times can
translate into lower test costs and preserve margins.
One thing, however, should be apparent. If most of the manufacturing tests are
done on a partially assembled DUT, and signals are conveyed using conductive
means rather than radiated means, once the device is fully assembled, including
its antenna(s), there are still possibilities for manufacturing defects. Thus, it is not
uncommon for the assembled devices to be final tested in a way that mimics the way
they will actually be used. That is, the devices are linked to an access point using
that technology’s linking processes, and the signals are conveyed between tester and
DUT using radiated conveyance rather than conducted means.
Testing of the fully assembled device using OTA signals certainly replicates real-
world operation. And if the objective is simply to determine that after final assembly,
when the antenna(s) is (are) connected, the device still works, then current OTA
testing environments are probably adequate. The problem, however, with current
OTA testing environments is that they do not allow the same level of precise testing
that conductive testing affords. As already noted, one can easily determine cable
Testing Wireless Devices in Manufacturing 309
loss. Furthermore, one can create sufficient isolation to ensure that only the signal of
interest is being tested. In an OTA environment, as anyone who has ever dropped a
call while walking can attest, path losses are very position and orientation dependent
and not very predictable.
If there was a way to make the OTA environment more precise and predictable,
where one could control path loss to a point where DUT position was no longer
critical and a reliable loss metric could be applied to both path directions (e.g., from
DUT to tester and from tester to DUT), it would then be possible to do many of the
tests now confined to conductive conveyance by using OTA conveyance. And with
some of the newer device designs featuring “tuned” antennas for optimizing signal
strength and reducing battery power drain, it would be possible to ensure that these
tunable antennas were working properly after final assembly. Recent breakthroughs
in OTA test infrastructure appear poised to do exactly this.
Using this patented technology, one can find an optimal channel that suits the purpose
of a test for each frequency, and do so using a single DUT location and orientation.
Earlier OTA test environments, for factory testing in small metallic enclosures,
used movable antennas and precise positioning screws that could be adjusted to make
path loss more manageable, but the time required to adjust such setups, and the need
to readjust each time a new device was placed inside, makes them less practical for
large-scale manufacturing testing where accuracy and overall speed prevail.
The invention described in the patent has been reduced to practice and is being
productized. Findings, so far, bear out that the path loss after magnitude and phase
delay compensations are consistent with using a coaxial cable. The OTA testing
system prototype showed that a DUT with two antennas (e.g., a MIMO 2 × 2 device)
could also be tested under optimal MIMO performance conditions where all receiv-
ers see the same level of signal strength. It was not intended for testing MIMO
devices in a fading environment. In addition, the prototype allowed testing of mul-
tiple MIMO streams in parallel, thus speeding up testing time. Using conventional
methods, it is often difficult or impossible to find a position and orientation of a
DUT with two antennas that satisfies the needed channel characteristics, as in this
example, where the cross-coupled signals are minimized.
Additional testing showed that one could stack multiple devices in the chamber
and establish predictable and repeatable path loss conditions between the tester and
DUTs as if connecting a vector signal generated source signal, through a splitter, and
routed conductively to those multiple devices.
107a
h11 h12 h1n
H= h21 h22 24b
h2n
20a
hnn 22
h1υ h1υ
(a) 24a
FIGURE 14.4 Ordinarily, radiated signals from multiple antennas to multiple antennas, as
shown, will produce transmitted components of each transmitting antenna at each receiving
antenna. By using an array of antenna elements, each having separately controlled phase
shifts, one can reduce the components of h = ij where I does not equal j creating an environ-
ment where path loss is both predictable and symmetrical. (Continued)
Testing Wireless Devices in Manufacturing 311
302 302
300
130 102a
100a 103a 200a
131aa 102aa
111a 130a 102am
110a 202a 211a 210a
VSG 1 102b RX 1
110b 111b 131am 103b
211b 210b
102ba
VSG 2 RX 2
102bm
131ba 202b
131bm
301
130b Device
Tester under
test dut
201a
135a 135
139a
135b
130a 131aa
132 Ø1
133
111a
136a 131ab
From
VSG G Σ Ø2 To individual antenna
136b array elements
134
138b
137 136m 135m
131 am
Øn
138 Control
136
139m
201aa
From
(b) DUT
FIGURE 14.4 (Continued) Ordinarily, radiated signals from multiple antennas to multiple
antennas, as shown, will produce transmitted components of each transmitting antenna at
each receiving antenna. By using an array of antenna elements, each having separately con-
trolled phase shifts, one can reduce the components of h = ij where I does not equal j creating
an environment where path loss is both predictable and symmetrical.
14.5 IMPLICATIONS
An OTA testing system that can quickly and easily compensate for multipath signals
at a device’s antenna(s), to create a reliable and repeatable path loss, can be used as
an alternative for contemporary testing using multiple ports and conductive means
312 Tunable RF Components and Circuits: Applications in Mobile Handsets
of conveyance. What that means is that some of the tests currently being done
conductively can be shifted to OTA wireless test environments. Rather than having
conductive tests on partially assembled devices, and wireless testing of fully assembled
devices, one can envision a case where a test flow in the OTA test system can do it all.
One of the advantages of shifting more conductive tests over to OTA tests is being
able to test the fully assembled device. In addition, there would be no cabling and unca-
bling handling times or the prospects of damaging a device’s antenna connector(s).
One unique test that an OTA test system can provide is antenna tuning. Doing
it requires that a device’s antenna is connected, thus precluding a conductive test
alternative. One of the unsung areas of device innovations is advances in antenna
technology. These advances can take advantage of OTA testing capabilities.
~ ~
(a) (b)
in results. Since the days of high-frequency (HF) radios and antennas, it was known
that one could “tune” an antenna physically or electrically to obtain maximum radi-
ated power, low-standing wave ratio, and the like. These principles apply equally to
tweaking the antenna system of a smartphone for optimal results.
One can tune antennas in two fundamental ways: aperture tuning and impedance
tuning (see Figure 14.5).
In aperture tuning, one may change antenna geometry, length, or ground-
ing points that, in turn, affects both the antenna radiated efficiency (ARE) and its
impedance. ARE is of great significance in antenna design because it is a metric that
corresponds to the maximum total performance that an antenna can achieve, assum-
ing that impedance mismatch loss is null. With impedance tuning, one conjugate
matches the antenna to the transmitter power amplifier or receiver input impedance
using some changes in capacitance and typically a fixed inductance to minimize the
loss due to impedance mismatch.
Traditionally, one would do the impedance tuning by completing the following
steps:
In Figure 14.6, one sees the application of a matching network for impedance tuning
of an antenna.
In general, one will find it difficult, or impossible, to obtain a simple matching
network that will provide a conjugate match for power amplifier and antenna for
all the bands to be utilized, or even for all the frequencies within a band spectrum.
ANT
ZTX/RX ZANT
TX/RX ~ MN
ANT
ZTX/RX ZANT
TX/RX ~ MN
Additional tuner not
placed on the antenna
Matching network (MN) includes variable capacitors and switches feedline to tune
antenna radiated
performance
Tuner
DUT
Power meter
Search algorithm
to optimize radiated power
Figure 14.9, in contrast, shows a variation on this test approach where the tuner
can be controlled by means that does not require any external cable connected to
the DUT with the tuner. One approach for the computer to control the tuner is to use
Wi-Fi or Bluetooth wireless connections. Another is to use signals from the tester
that could switch different values of capacitance in the tuner while the OTA system
provides fast power measurements. Measured performance, in this case, would be
more consistent with the behavior of the DUT after the test since there would be no
extraneous cables and connections.
Figure 14.10 shows an example of antenna impedance tuning, using a test setup
as in Figure 14.9, where the measured metric is antenna efficiency, and not antenna
impedance. A better impedance match will translate to better antenna efficiency.
Results show overall antenna efficiency for 16 different capacitor values of a capacitor
Tuner
DUT
Tester
FIGURE 14.9 By eliminating any cabling between the DUT and tester during testing, mea-
sured results are more consistent with after-test findings than for methods involving cabling
between DUT and tester.
316 Tunable RF Components and Circuits: Applications in Mobile Handsets
0
Cap value 1
–5 Cap value 2
Cap value 3
Antenna efficiency (dB)
Cap value 4
–10 Cap value 5
Cap value 6
–15 Cap value 7
Cap value 8
Cap value 9
–20 Cap value 10
Cap value 11
Cap value 12
–25 Cap value 13
Cap value 14
–30 Cap value 15
700 720 740 760 780 800 820 840 860 880 900 920 940 960 Cap value 16
Frequency (MHz)
FIGURE 14.10 Antenna efficiency as a function of frequency for each capacitor value of a
variable capacitor tuner.
in the tuner placed in a certain matching network configuration between the antenna
and a transmitter for 27 frequency points from 700 to 960 MHz. With this plot, one
can directly find the capacitor value that gives the best overall radiated performance
for a frequency band. For example, one would choose the following capacitor values
for the following bands: Cap value 13 for LTE band 17 (700–740 MHz), Cap value 9
for LTE B13 (740–790 MHz), Cap value 1 for LTE Band 5 and 8 (820–960 MHz).
As depicted in the plot, there are a lot of measurements, precisely 432 measure-
ment values. Therefore, in order to utilize this method of optimization efficiently,
measurement should be fast.
This method of antenna tuning over the air can be crucial in optimization time
when antenna design includes additional aperture tuning where ARE is a varying
metric. Impedance measurements cannot capture ARE. The only way to capture
radiated efficiency variation from aperture tuning is to measure antenna effi-
ciency. Therefore, OTA optimization on overall antenna efficiency that includes
both radiated efficiency and impedance mismatch loss for antenna tuning makes
a lot more sense.
The example shown here is for developing a device’s antenna for covering the
low-band spectrum while maintaining adequate antenna efficiencies. In a case where
a device has a built-in matching network and where the OTA test system can control
that network’s switching while measuring antenna performance, it would be pos-
sible to fine-tune each device’s antenna performance during the fully assembled
final test phase.
14.7 CONCLUSIONS
Testing of wireless devices after manufacturing is essential to detecting manufactur-
ing defects that render the device inoperable or out of specification. This is important
Testing Wireless Devices in Manufacturing 317
because one does not want to ship a faulty device or one whose operation may cause
interference to other devices and services.
Manufacturing testing is driven by the PHY layer specifications of the standards
that underlie each technology, such as 802.11n or 3GPP LTE. Manufacturing test
should not be used to verify a design. That is the purpose of device design verifica-
tion testing. Rather, manufacturing test must assume a design is valid and look for
indications that a device, whose design is sound, has been affected by manufacturing
defects. This allows for using a subset of design verification tests for faster and lower
cost testing.
Contemporary manufacturing test of wireless devices, ironically, makes copious
use of signals conveyed conductively rather than radiated. As such, because the test
systems must connect to a device’s antenna port(s), the device cannot be fully assem-
bled and intact. In fact, one cannot test the entire device because the antenna will not
be connected during conductive testing.
Consequently, fully assembled devices are tested, again, later to ensure that the
device works in an OTA signal environment. Today, the proportion of tests and test
time done conductively versus OTA is both richer and longer. However, innova-
tions in OTA testing systems and methods are paving the way for more tests to be
done using OTA methods. Whereas the actual testing of TX and RX characteristics
may not change compared to conductive testing, the handling time could certainly
be shortened by the absence of connecting and disconnecting conductive cables to
devices’ antenna ports.
Engineering – Electrical
• Introduces the concept of tunable RF, the forces driving its rapid adoption,
and its potential to transform RF front-end architectures
• Provides in-depth coverage of all deployed and emerging tunable RF
applications in mobile handsets, including estimates of market size, drivers,
and growth projections
• Covers key technologies (such as SOI, BST, and RF-MEMS), design techniques,
and trade-offs involved in the implementation of tunable RF solutions
• Presents an extensive overview of antenna tuning and tunable antennas,
supplemented by examples and case studies
• Offers a broad industry perspective from leading suppliers and customers
in wireless networks, handsets, chipsets, architectures, modules, antennas,
components, and test equipment
K25475
ISBN: 978-1-4987-1889-9
90000
9 781498 718899