Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

PLL Divider

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Synchronous programmable divider design for PLL

Using 0.18 um cmos technology


Suchitav Khadanga
RFIC design consultant, Bangalore, India

suchitav@ieee.org
Which operate at RF frequency, Due to the dividers
complexity, high operation frequency normally leads
to high power dissipation. Other crucial aspects of the
present-day consumer electronics industry are the
short time available for the introduction of new
products in the market, and the short product lifetime.
On top of that, the lifespan of a given CMOS
technology is also short, due to the aggressive scaling
of minimum feature sizes. Short time-to-market
demands architectures providing easy optimization of
power dissipation, fast design time and simple layout
work. High reusability, in turn, requires an
architecture, which provides easy adaptation of the
input frequency range and of the maximum and
minimum division ratios of existing designs.

Abstract
In the wireless communication market,
trends are moving towards smaller size, fewer parts,
longer lifetime and higher frequency operation. These
trends imply that wireless communications circuits
must incorporate higher integration and that their
design and IC technology must be optimized for low
power and high frequency system. One innovative
method to increase the frequency of programmable
divider is discussed. The new method not only
increases the frequency of operation but also
decreases circuit complexity and power dissipation.
This new design use synchronous counters instead of
asynchronous counters. The digital gates are
optimized for minimum propagation delay and
loading effect using progressive sizing of the
transistors. This is better configuration in every
aspect in terms of frequency, power dissipation and
chip area.

The choice of the divider architecture is


therefore essential for achieving low-power
dissipation, high design flexibility and high
reusability of existing building blocks. A modular
architecture complies with these requirements, as
shall be demonstrated in this paper.

Index terms: CMOS integrated circuits, frequency


synthesizers, phase locked loop, programmable
divider, Prescaler, PLL.

The divider can be implemented using static


or dynamic cmos circuits. Dynamic circuit techniques
evolved in the last few years into several cmos circuit
technique such as domino, NORA, TSPC circuits.
Nevertheless, the field of CMOS integrated circuits
has reached a level of maturity, where it is now
mainstream technology for higher integration density,
low power consumption. Therefore, we had chosen
cmos to design the programmable divider. The
designed dividers have increased frequency
capability with reduction of propagation delay using
synchronous counters instead of asynchronous
counters. The focus of the paper is first on the truly
modular architecture and on the implementation of
the circuits. Then the description of circuits with their
behavior inside the programmable divider is
described. Finally, a collection of measured data and
the conclusions are presented.

1. Introduction
The scaling of CMOS technologies to deep
submicron has made CMOS a technological option
for the low-gigahertz frequency range. However, for
CMOS to become a commercial option for RF
building blocks requires compliance to all trends of
the consumer market: miniaturization, low cost, high
reliability and long battery lifetime. Bulk CMOS
technologies presently available satisfy the low cost
and reliability trends by standard design practice.
Complying with miniaturization and long battery
lifetime, on the other hand, demands CMOS building
blocks with low-power dissipation and good
electromagnetic compatibility (EMC) characteristics.
A critical RF function in this context is the frequency
synthesizer, more particularly the programmable
frequency divider. The divider consists of logic gates

1
Proceedings of The 3rd IEEE International Workshop on System-on-Chip
for Real-Time Applications ISBN 0-7695-1929-6/03 $17.00 2003 IEEE

4. Design Requirement
2. Background
The synthesizer generates frequencies in the
range of 2.4 to 2.5GHz in steps of 1MHz. The
synthesizer architecture is given in figure 1. This is a
single-loop PLL. VCO oscillates in the range of 2.4
to 2.5GHz and gives out a frequency component
proportional to the control voltage input. The
programmable divider N divides the VCO frequency.
The PFD compares the Phase & frequency of the
VCO-divided-by-N and the reference-divided-by-R
signals and applies a correction voltage to the VCO
tuning input. This keeps the VCO phase locked to the
reference.

Phase-Locked Loops (PLL) is standard


components on almost all-modern Integrated
Circuits. Their uses range from frequency synthesis
(i.e. to increase or decrease the period of a given
periodic signal), to data recovery, to minimizing
clock skew on large ICs.
Because of the extreme importance of PLL
in todays architectures, a great deal of research and
literature on the topic emanates from some of the top
minds in the IC industry and the academic field. The
current state of knowledge of general PLL
mathematical interpretation, design and fabrication is
vast; however, research does not exist for the design
of a PLL in the specific implementation we are
proposing. PLL is very challenging because internal
circuits must be both digital and analog. In fact,
although both the inputs and outputs of the PLL are
digital signals, all internal circuitry is essentially
analog. It is the analog areas of the PLL that require
the most design experience and are the most difficult
to test.

Dual modulus prescaler and the counters P


& S form the programmable divider. This divides
the VCO output to 1 MHz. Their operation is as
follows. The prescaler divides the VCO frequency
output by 17 until S counter counts down to zero.
At this point, it switches over and divides by 16 until
P counter counts down to zero. Then the counters
are reloaded and the DMP switches back to divideby-17. External crystal oscillator provides a stable
reference of 10MHz that is internally divided down
by the R counter to 1MHz.

3. PLL Overview
PFD and charge pump in combination with
the external loop filter give out a voltage proportional
to the phase or frequency difference between the
reference and the divided VCO output. The external
loop filter establishes the Phase noise of the
synthesizer output, the locking time and suppression
of reference spurii.
Frequency setting: P, S & R are
programmable counters. Loading P, S & R
counters (fig.1) set the synthesizer frequency. Three
9-bit latches are present, which load the counters
every time they count down to 0. A 9-bit control
word is programmed into the latches from external
glue logic. Two latch-select bits and a strobe are
used. Hence, 12 pins (9 data pins + 2 latch select + 1
strobe) are required for programming.

The function of the PLL is to lock the phase


and frequency of the VCO to the reference oscillator.
The VCO output is coupled into the PLL circuit
where the VCO frequency is divided down in a dual
modulus Prescaler and feed back divider to the phase
comparison frequency or step size of PLL. The block
diagram of PLL is shown in figure 1. The shaded
portion is the programmable divider section.

F REF

DIVIDER
OUTPUT

PFD and
Reference
Divider(R)

Ve

CP &
FILTER

COUNTER (P)

Vc

Fout
VCO

Prescaler
N/N+1

DIVIDER

INPUT

SWALLOW
COUNTER (S)

There are three programmable dividers in


this PLL chip. We will discuss one programmable
divider. The other two dividers are implemented in
the same manner. Such a programmable divider
consists of three parts. The block diagram of 9 bit
programmable divider is shown in figure 2. It has N
bit counter, control logic and SR latch. After the

MODE
CONTROL

Figure 1: Block diagram of PLL

2
Proceedings of The 3rd IEEE International Workshop on System-on-Chip
for Real-Time Applications ISBN 0-7695-1929-6/03 $17.00 2003 IEEE

5. 9 bit Programmable divider

minus one. Therefore, we had added an incrementer


circuit in the divider circuit, so that the output
frequency will be divided by the loaded value.
Here, first we invert the loaded value, then we
increment by one using incrementer. Without the
incrementer the output frequency is divided by
loaded value minus one.
To get a clear picture of the function of the
divider take for example design of 4 bit
programmable divider. It consists of one 4 bit
synchronous counter and control block. The outputs
of the counters are QB4, QB3, QB2, and QB1. The
MSB is QB4, For normal operation the QB4 is
divided form of the clock by 15. If we are loading 3
(0011) then it will be inverted to 1100 and
incremented by one so the final value is 1101. So
when the counter value reaches at 1101, then all the
flip-flops will be cleared. So we will get three clocks,
1111 to 1101 high value and one low value. So in
total there are three clock means the frequency of the
clock is divided by three, which is the loaded value.

loaded (programmed value) the out put of the control


logic will be one and all the flip flops are cleared.

OUTPUT

Input

CONTROL
BLOCK

COUNTER

CLR

LATCH

Figure 2: 9 bit programmable divider


The function of the 9 bit divider is based on
digital logic. The counter can be synchronous or
asynchronous down counters. The control logic is
used to load the desired value. The counters are down
counters. The divider output is taken across MSB of
the counters. For normal operation the input clock
frequency is divided by 512 (9 bit). But in order to
divide the frequency for the programmed value, the
counter is to be cleared after the loaded value. The
function of the control block is to generate signal for
counters to be cleared. .
The schematic of the control block is given
in figure 3. It consists of two input XOR gates. As we
know if the two inputs of the XOR gate are same then
output is zero otherwise it is one. We use this
principle for our divider design. For 9 bit divider we
need nine two input XOR gates. One input of the
XOR gate is used for loading and the other input is
for comparing with QB output of individual flip flops
of the counter. When all the loaded value is equal to
all QB then all the output of the XOR gates are zero
and the output of nor gate is one. The output of the
NOR gate is used to clear the flip flops.

Figure 3: Control circuit


programmable divider

for

But the propagation delay from the output of


control circuit to the control input of individual flipflop will vary, so the clear of all the flip-flops will
not occur simultaneously. The last flip flop (MSB of
counter) will take appreciably longer time to reset
than the first flip flop. Wide variations in reset
propagation time will occur if the counters are
unevenly loaded. A method of eliminating this
difficulty with resetting is to use a latch to memorise
the output of nand gate at the Nth pulse. This process
is termed as pipelining of the flip flops. The complete
schematic of the 9 bit divider is shown in figure 4.

bit

Figure.4: Schematic of 9bit programmable divider

6. Programming overview

With this control circuit of XOR gates


having comparison of QBAR of individual flip flops
will give divided output frequency by loaded value

The programmable dividers are serially


programmed using a standard serial input parallel

3
Proceedings of The 3rd IEEE International Workshop on System-on-Chip
for Real-Time Applications ISBN 0-7695-1929-6/03 $17.00 2003 IEEE

output (SIPO) shift register. The programming data is


input using the clock; data and latch enable input
pins. The clock input latches one bit on the data input
into PLL shift register on the rising edge of each
clock pulse. When the latch enable input is high, the
stored data is transferred into the latches. Then the
programmable divider is loaded with the
programmed value and is in the same value until
unless the latch enable is high.

Power supply
Input
frequency
(S,P)
Output
freq.
(S,P)
HIGH voltage

Min

Typ

Max

1.65

1.8

1.9

Volt

400

MHz

25

MHz

0.7Vdd

V
0.2Vdd

LOW voltage
S counter

127

511

127

First, the functionality of the programmable


divider is tested using AHDL models. The uses of
AHDL models have mainly two advantages. One is
the simulation time is less for AHDL models against
using transistor models directly. The second is it is
easy to change the specifications of the models to
meet the design requirement.
Then all the AHDL models are replaced
with transistor models one by one. In each and every
change, the functionality is tested thoroughly. The
circuit diagram of the programmable divider is shown
in figure 6. Each block is already explained
previously. So also, these are fundamental building
blocks for a digital designer. The circuits are self
explanatory.

Figure 5 Schematic of programming


section of programmable divider
The schematic is shown in figure 5. Here we
used 11 bit shift register to load the dividers. The first
nine bit will decide the loaded value for the divider
and the last two bits will select the dividers ( P/S or
R) The last two bits are given to a decoder inputs,
which will send enable to the individual dividers. If it
is (0,1) the divider P is selected, If it is (1,0) then
divider S will be selected and if it is (0,0) then
divider R will be selected.

7. Circuit Design
The programmable dividers are designed
using TSMC 0.18um technology. As all the circuits
are digital only, so we chose CMOS process instead
of BICMOS or SiGe. So also the preliminary analysis
of the 0.18um technology was verified to meet the
design requirements. The design requirements for the
programmable divider is given in table-1

Figure.6: Schematic of the divider without


the programming section.

Table 1 Design requirement for programmable


divider
Parameter

Specification

Unit

8. Characterization

4
Proceedings of The 3rd IEEE International Workshop on System-on-Chip
for Real-Time Applications ISBN 0-7695-1929-6/03 $17.00 2003 IEEE

The programmable divider is completely


characterised with different process, temperature and
power supply variations from normal conditions. The
programmable divider is tested with interfacing with
a general bit pattern generator. The general bit
pattern is implemented using AHDL models. The
property of the generator allows the user to generate
any loading value depending upon the user. The
schematic of the complete programmable divider
with testing part is shown in figure 7.

Figure 8: Simulated output of the divider for


input clock period 2.5nS and P=7 and S=3
Figure 7: Schematic of programmable
divider for characterization.
The simulation is carried out under worst
case conditions. The divider is tested using slow slow
(SS) corner analysis, 50 degree centigrade
temperature and power supply of 0.2V low to 1.6 V
high. The simulation is carried out both in schematic
and parasitic extraction level of the programmable
divider with application of all the worst-case
conditions at the same time. The simulated graph is
shown in figure 8, It shows the divided output for
P=7 and S=3. This output will go to mode control
circuit of prescaler. For low level the prescaler will
select N+1 and for high logic it will select N. The
design of prescaler is not part of this research. So the
design of prescaler is excluded here. We had
concentrated of the divider section of the PLL
without the prescaler.

Figure.7: layout of the Programmable divider

Figure 9: Layout of the divider


9. Conclusion
The architecture, principle and circuit design
of the programmable divider was described here.
From the simulation results it is clear that the
designed programmable divider meets all the
specifications except the frequency requirement. The
difference in delay between schematic and parasitic
simulation can be minimize. As we had concentrated
on less chip area and less power dissipation, so the
propagation delay is little bit more. However for
decreasing the delay, we have to pay for the higher
power dissipation. The future development will be in
terms of further increase of application frequency,
power dissipation and area.

The programmable divider is characterised


with different values of P and S, The layout is shown
in figure 9. The simulation on parasitic extraction is
working well above 2.85nS, whereas the schematic
level simulation is working at 2.5nS. The variation in
the frequency is due to RC delay of the divider in the
interconnections and parasitic.

5
Proceedings of The 3rd IEEE International Workshop on System-on-Chip
for Real-Time Applications ISBN 0-7695-1929-6/03 $17.00 2003 IEEE

[4] F. Piazza and Q. Huang, A low power CMOS


dual modulus prescaler for frequency synthesizers,
IEICE Trans. Electron., vol. E80-C, pp. 314319,
Feb. 1997.
[5] P. Larsson, High-speed architecture for a
programmable frequency divider and a dual-modulus
prescaler, IEEE J. Solid-State Circuits, vol. 31, pp.
744748, May 1996.
[6] J. Navarro Soares, Jr. and W. A. M. Van Noije,
A 1.6 GHz dual-modulus prescaler using the
extended true-single-phase-clock CMOS circuit
technique (E-TSPC), IEEE J. Solid-State Circuits,
vol. 34, pp. 97102, Jan. 1999.

Reference
[1]. J. Craninckx and M. S. J. Steyaert, A 1.75GHz/3-V Dual- Modulus Divide-by-128/129
Prescaler in 0.7-_m CMOS,IEEE Journal of Solid
State Circuits, vol. 31, no. 7, pp. 890- 897, July 1996.
[2] J. Yuan and C. Svensson, High-Speed CMOS
Circuit Technique ,IEEE Journal of Solid-State
Circuits, vol. 24, no. 1,pp. 62-70, Feb. 1989.
[3] J. Craninckx and M. Steyaert, A 1.75 GHz/3 V
dual-modulus divide-by-128/129 prescalar in 0.7 _m
CMOS, IEEE J. Solid-State Circuits, vol. 31, pp.
890897, July 1996.

6
Proceedings of The 3rd IEEE International Workshop on System-on-Chip
for Real-Time Applications ISBN 0-7695-1929-6/03 $17.00 2003 IEEE

You might also like