EQUIVALENT CIRCUIT MODELLING OF BIPOLAR TRANSISTORS

by

Monica Dowson

A thesis submitted to the University of Leicester

for the degree of Doctor of Philosophy

March 1982
 ACKNOWLEDGMENTS

The author gratefully acknowledges the advice and encouragement

given by her supervisor, Dr. O. P. D. Cutteridge; the helpful discussions

with and the provision of data by Mr. P. J. Rankin and Mr. K. W. Moulding

of Philips Research Laboratories; and the understanding of her husband,

who provided encouragement, food and drink at all the appropriate times.
 EQUIVALENT CIRCUIT MODELLING OF BIPOLAR TRANSISTORS

Monica Dowson

ABSTRACT

Existing equivalent circuit models of bipolar transistors are

reviewed together with techniques for the evaluation of suitable values
of the model elements. A method enabling the optimisation of the element
values of any particular model in order to match the measured S parame-
ters of a device that is to be modelled is described. This method uses
a modified Gauss Newton algorithm to minimise an objective function
defined as the sum of the squares of the weighted errors between the re-
quired S parameters and those of the model.

Details are then given of a new modelling algorithm for the

development of accurate equivalent circuit models which was developed
from this original optimisation method. The new modelling algorithm
requires some S parameter measurements of the device to be modelled'
over an appropriate range of frequencies, together with a potentially
suitable model. The initial model elements are optimised and then,
if necessary, suitable topological changes, involving the addition or
deletion of both elements and nodes, are made until a model having the.
required accuracy or complexity is obtained.

A number of examples are given of small-signal equivalent circuit

models of bipolar transistors developed using the algorithm. These
particular transistors were operating at frequencies up to lGHz .A
further example is given of the use of the modelling algorithm in the
development of a bias dependent small-signal model of two similar bipolar
transistors frequencies up to 2GHz Additional S para-
operating at .
meter data for the same two transistors is also used to demonstrate that
the algorithm can be used successfully for the development of non-linear
models.
 CONTENTS

Page

CHAPTER 1 INTRODUCTION 1

1.1 Background 2

1.2 Aims of the Research 3

1.3 Achievements 4

1.4 Computing Facilities Used 5

CHAPTER 2 MODELLING THE BIPOLAR TRANSISTOR 7

2.1 The Bipolar Transistor 10

2.2 Hybrid Pi Model 13

2.2.1. Basic Hybrid Pi Model 13

2.2.2. Modified Hybrid Pi Model 14

2.3 Ebers Moll Model 16

2.3.1. Basic Ebers Moll Model EMI 16

-
2.3.2. Extended Ebers Moll Model - EM2 18

2.4 Other Non-Linear Models 20

2.5 Choice of Models 21

CHAPTER 3: TRANSISTOR PARAMETERMEASUREMENTS

AND ANALYSIS OF MODELS 23

3.1 y Parameters 25

3.1.1. Nodal Analysis 26

3.1.2. Generation of the Network Polynomial Coefficients

from the Nodal Admittance Matrix 28

3.1.3. Calculation of the y Parameters Using Pivotal

Condensation 30

3.1.4. Accuracy of Calculations 32

3.1.5. Further Considerations on the Choice of Algorithm 39

3.2 s Parameters 39

3.2.1. Measurement of the s Parameters 41

3.2.2. Calculation of the s Parameters 42

CHAPTER4: NUMERICAL OPTIMISATION 43

4.1 Available Methods 46

4,1.1. Algorithms Using Function Values Only 46

4.1.2. Algorithms Using Derivatives 47

4.1.3. Algorithms Designed for Hinimising Sum of

Squares Functions 50

4.2 Some Details of the Optimisation Algorithm Chosen 51

CHAPTER5 DEVELOPMENTOF THE MODELLING ALGORITHM 54

5.1 Details of the First Modelling Exercise 54

5.2 Construction of an Algorithm to Optimise Model Element

Values 57

5.2.1. Construction of the Error Function 57

5.2.2. Generation of the Jacobian Matrix 59

5.2.3. Application of Constraints 60

5.2.4. Exit Criteria 61

5.3 Example of Results 61

S. 4 Improvement of Models 63

5.5 Example of Results 65

5.6 Description of the Final Modelling Algorithm 70

5.6.1. Stage 1- Obtaining the First Local Minimum 71

5.6.2. Stage 2- Element Addition 76

5.6.3. Stage 3- Node Addition 81

5.7 Checking the Final Modelling Algorithm 87

CHAPTER 6 : MODELLING A VERTICAL N-P-N TRANSISTOR 92

6.1 Common Collector Configuration 96

6.2 Common Emitter Configuration 100

6.2.1. First Attempt 100

6.2.2. Second Attempt 102

6.3 Common Base Configuration 107

.
6.4 General Model 112
 6.4.1. First Attempt 112

6.4.2. Second Attempt 116

6.4.3. Suggested Alternative Approaches 123

CHAPTER 7: A BIAS DEPENDENTMODEL FOR TWO SIMILAR BIPOLAR TRANSISTORS 125

7.1 Optimisation of the Model 127

7.2 Examination of the Bias Dependent Elements 133

7.3 Resultant Bias Dependent Model 141

7.4 Models for High Emitter Currents 147

CHAPTER 8: CONCLUSIONSAND SUGGESTIONSFOR FURTHERV,

'ORK 151

APPENDIX 1: S PARMIETER DATA PROVIDED BY PHILIPS RESEARCH

LABORATORIES 156

APPENDIX 2: RELEVANT PUBLICATIONS BY THE AUTHOR 176

REFERENCES 194
 I

CHAPTER 1

INTRODUCTION

Since the invention of the transistor in 1948, models have been

sought which simulate the behaviour of the device. The work undertaken

by the author and described here, has been concerned entirely with

modelling bipolar transistors, the construction and applications of which

are discussed in Chapter 2.

Models of a transistor can be either in terms of mathematical rela-

tionships or in the form of equivalent circuits. Equivalent circuit

models are popular for several reasons: circuit designers tend to feel more

at ease with a circuit diagram rather than a set of numbers or mathematical

relationships; transistor and integrated circuit designers prefer a model

that reflects the construction of the device enabling modifications to the

construction to be mapped. onto the model; and there are many circuit

analysis packages now available that use equivalent circuit models of tran-

sistors and other devices to incorporate in larger circuits to be analysed.

The author has been fortunate in receiving data on a number of bipolar

transistors from Philips Research Laboratories for which they were inter-

ested in obtaining equivalent circuit models. In addition to proving that

the electronics industry is interested in the modelling of transistors, this

also provided a definite direction for the research. The requirement was
for small-signal (implying linear) models for transistors designed to

operate at frequencies in ranges between 0.5 MHz and 2 GHz. MacleanI has

also expressed an interest in this type of problem.

Although the techniques described here were originally devised for the

generation of linear models of bipolar transistors, the techniques involved

are equally applicable to other modelling problems. In the final problem

here, which was a bias dependent model, the author obtained models for the

non-linear operation of a bipolar transistor. Mathews and Ajose2 have

 2

used one optimisation technique to model both bipolar and field effect

transistors.

1.1. Background

In recent years, there has been a great deal of research interest

at Leicester in network synthesis and optimisation. Theses on passive

have been presented 3,4

by di Mambro . Wrights, Krzeczkowski6.,
network synthesis
7 8, 9, have presen-
Hegazi and Savage whilst Dowson Henderson" and Dimmer"

ted theses more concerned with the optimisation processes. This expertise

has been of direct benefit to the author.

The techniques used for the analysis of passive networks, particularly

3,4 12
those developed by di Mambro and di Mambro and Cutteridge , were con-

sidered for the author's application as described in Chapter 3; these were,

however, only used for a short time until the analysis routine written by

the author, also described in Chapter 3, was developed. The technique of

coefficient matching which all these earlier network synthesis programs

used was not considered suitable for the problems being studied. by the author.

In addition to the local experience in the analysis of networks, there

was also a considerable knowledge of optimisation techniques available.

In particular the technique developed by Henderson", to which the author9

also made some contributions, were of great importance. This optimisation

technique involved the use of a two-part algorithm and was designed to solve

difficult optimisation problems. It was discovered that for the type of

problems to be solved by the author one part, the Gauss Newton section,

with adaptations to suit the problem type, was an ideal technique. Details

of the optimisation problem and the techniques used are given in Chapter 4.
 3

1.2. Aims of the Research

As already mentioned, some data was provided by Philips Research

Laboratories on some devices for which they were interested in obtaining

equivalent circuit models. The data was provided in three stages.

The first set of data13 was for a lateral p-n-p transistor. This was

an integrated circuit transistor and was described13 as "a notoriously

difficult device to model" and "one of current interest to us [Philips

Research Laboratories]". In addition to the data, a first stage model of

the device was provided. The requirement by Philips Research Laboratories

was that a more accurate model of the device be developed.

This requirement, as discussed in Section 2.5, is quite common.

Frequently, standard models of devices do not give adequate correlation

with the measured characteristics of the devices. Thus, the aim of this

research was to develop techniques whereby models of devices could be

verified, and if necessary, improved models be produced. These improved

models could be produced merely by optimising the element values of the

original models. If this is still not adequate then it becomes necessary

to optimise the topology of the models together with the element values

until the required accuracy or complexity is obtained.

These aims necessitated the development of analysis and optimisation

techniques which were then combined together to form an overall modelling

strategy. The development of an improved model for the lateral p-n-p

transistor is described in Chapter 5 and the modelling technique devised

from this experiment is detailed in Section 5.6.

This new technique was then tested on the second set of data provided
by Philips Research Laboratories 14 This data was for integrated
- another
circuit transistor, this time a vertical n-p-n transistor, with data for
each of the common emitter, common base and common collector configurations.
Finally, a set of data'5'16 was provided for two similar n-p-n tran-
 4

sistors at different bias conditions. Measured data was available at

two voltages and at six different currents for the first type and at two

voltages and at seven different currents for the second type. Slatter17

had published a model for a limited bias range based on physical aspects

of these devices. The author's aim was to improve on this model by ob-

taining more accurate bias dependent models covering a wider bias range

using as few bias dependent elements as possible consistent with the use

of simple 'relationships for those elements. This problem is described

in detail in Chapter 7.

Achievements

As will be seen later the author has produced a method by which equi-

valent circuit models of bipolar transistors can be verified against a set

of s parameter measurements of the devices and if they are inadequate

.
the element values of the* models can be optimised. Then..' if necessary,

the topology of the model can be modified until the required accuracy or

complexity is obtained.

In order to achieve this, the author has produced a computer program

to analyse a network in terms of its s parameters and combined this with

an optimisation routine together with a new modelling strategy.

Using this technique a number of models have been produced using data

supplied by Philips Research Laboratories. The first of these was a

model for a lateral p-n-p transistor which operated in the frequency range
0.5 MHz to 10 MHz. The maximum absolute errors in the parameters
s

of the final model were 0.1 dB in the modulus and 1.2* in the phase with

r. m. s. (root mean square) errors of 0.04 dB and 0.4*. This final model

consisted of 6 nodes with 12 elements and included one current source

only.

The next device modelled was a vertical n-p-n transistor model which
 5

operated in the frequency range 0.1 GHz to 1 GHz. Separate models were

produced for each of the common emitter, common collector and common base

configurations. These gave maximum absolute errors of 1.5-dB and 4"

and r. m. s. errors of 0.3 dB and 3* An attempt was made to produce

a single general model applicable for any of the configurations. This

became a task of tremendous volume and was only partially successful giving

r. m. s. errors of 1.3 dB and 12.9* Whilst the author believes that

.
this type of task is possible, it requires a large, fast computer with

extremely efficient code in order to generate sufficiently accurate models.

Finally, a bias dependent model was produced for two similar bipolar

transistors operating at frequencies of 0.1 GHz to 2 GHz The same

.

model was produced for both transistors and was valid for the linear oper-

ation of the devices at voltages of 0.5 V and 3V and for emitter

currents in the range 0.5 mA to 4 mA for the first and 0.5 mA to

8 mA for the second of the two transistors. Simple-expressions for

the bias dependent elements were developed and good agreement with the

measured characteristics of the transistors was achieved. Accurate models

were also developed for the non-linear operation of the transistors.

Thus, this modelling technique has been extensively tested and although

these tests have been with s parameter data for bipolar transistors, the

author is of the opinion that the techniques could be applied to many

electronic modelling problems.

1.4. Computing Facilities-Used.

The majority of the computing work has been done on the CDC Cyber 73

at the University of Leicester and the CDC 7600 at the University of

Manchester (UMRCC). These machines are compatible and programs are

easily transportable between the two. The main feature of these machines

is the word length which is 60 bits. This means that high precision can
 6

be easily obtained. The UMRCC CDC 7600 is approximately 12 times more

'powerful' than the Leicester Cyber 73, thus for larger jobs where the

Leicester time limit of 33 minutes of c. p. u. time proved an encumbrance,

the UMRCCfacilities were used. In general, the time problem depends

on the amount of s parameter data available and the complexity of the

models involved. The p-n-p transistor model described in Chapter 5 was

produced on the Cyber 73 but the n-p-n transistor models described in

Chapter 6 required the CDC 7600, whilst the majority of the computing for

the bias dependent model in Chapter 7 was done on the Cyber 73.

Graphical output was generally obtained using the Cyber 73 together

with the GHOST library of graphical subroutines. The model circuit diagrams

were produced on the University of Leicester PDP-11/44 using an interactive

circuit layout program written by the author and which used the GINO

library. Flow charts were also produced using the PDP-11/44, using

the interactive program FLOW18.

All the computer programs were written in FORTRANIV.

 7

CHAPTER2

MODELLING THE BIPOLAR TRANSISTOR

A special report on the transistor in Electronics19 describes the

stages leading to the announcement in 1948 by Bardeen, Brattain and

Shocklaythat they had invented the transistor. This first transistor

was a point contact device. Shockleythen went on to develop the

junction transistor. These three inventors were awarded the Nobel Prize

in 1956 for their work.

Since the invention of the transistor, models describing their

behaviour have been sought. Both circuit and device designers need

models that will simulate the performance of the transistor over a wide

range of operation. Device designers generally require models which are

based on the physical structure of the device in order that the effects of

changes in the construction or geometry of the device can be predicted.

On the other hand, circuit designers are more concerned with the accuracy

of the model and are less interested whether the elements in the model have

a physical equivalent in the manufactured device.

Equivalent circuit models of transistors can be used in conjunction

with computer aided design techniques to assist in the design and analysis

of discrete or integrated circuits without the time and expense of manu-

facturing the circuits. GetreU20 lists a number of such uses of models,
from performing analyses of waveforms and frequency responses of circuits
to predicting the performance of an integrated circuit at high frequencies

without the parasitics a breadboard would introduce.

This chapter starts with a brief description of the bipolar transistor

and its applications. Then some of the popular models are described

beginning with the classical hybrid pi model and going on to more compli-

cated models such as the Ebers Moll model and its variants.
 8

2.1. The Bipolar Transistor

A bipolar transistor consists of an emitter, base and collector and

may be either an n-p-n or p-n-p type. Block diagrams of both types are

given in Figure 2.1 together with their circuit symbols. Navon2l

defines the modes of operation of the bipolar transistor as follows.

1. Active mode the emitter junction is forward biased and the

-

collector reverse biased.

2. Saturated mode both the emitter and collector junctions

-

are forward biased.

3. Cutoff mode both the emitter and collector junctions are

reverse biased.

4. Inverse mode the collector junction is forward biased and

the emitter is reverse biased.

In the active mode the transistor may be used as an amplifier. A

schematic of the active region biasing arrangements for an n-p-n transistor

in common base configuration is given in Figure 2.2. Figure 2.3 shows an

amplifier circuit described in Reference (22) where the biasing is achieved

by using a single battery together with the voltage divider network con-

sisting of resistors R3 and R4 In the diagram R, is the source

resistance and R2 is the load resistance.

Navon2l states that in a linear amplifier circuit such as this, the

9
signal transmission can be quite fast, of the order of 10- seconds,

and this rapid response can be used to achieve signal amplification at

frequencies up to several gigahertz. In addition to the common base

configuration, common emitter and common collector amplifiers can also be

used. The common emitter amplifier is probably the most used in

practice because, as stated in Reference (22), it is

 9

EC 71 Ec
IpIn pI npn
L 0
emitter IB collector emitter 'o"
coltector

B
base base

EC EC

B
p-n-p type n-p-n type

Figure 2.1. The Structure and Circuit Symbols for Bi polar Transistors

r T_
L

Ve Ib vc

Figure 2.2. Typical Bias Arran ent for an N-P-N Transistor Biased in
the Active Rer,,ion
 10

V it

Vout
Vin
0- %%ý wo
The 0- u Ttme

Figure 2.3. A Simple Common Base Amplifier

Ic

F
out

Vid
F] ýý, f Time
Vout4
1ime

Figure 2.4. An N-P-N Transistor Used in Switching Mode

 11

the only configuration to provide current, voltage and power gain.

In the saturated and cutoff modes the bipolar transistor can be used

as a switch. When used in switching mode, the transistor is normally

connected in common emitter mode as in the simple switch shown in Figure

2.4. The two parts of the switch are the emitter and collector terminals

and a small signal, V

' applied to the base terminal can cause the
in
normally non-conducting transistor to convert to the conducting state thus

producing a voltage V across the load RL* Navon2l states that in

out
this configuration the current gain, defined as the ratio of the collector

current ificrease to the base current which causes this increase, can be

higher than 50.

If the transistor was symmetrically designed then the operation of

the transistor in the inverse mode would be identical with the active mode
but with the collector and emitter interchanged. However, in practice

the emitter is more heavily doped and the collector cross-sectional area

is greater than that of the emitter. Thus the p-n-p transistor should
be more accurately described as a p+-n-p type.
A more accurate diagram of the construction of a planar n-p-n trans-
istor than that given in Figure 2.1 is now shown in Figure 2.5 . This diagram

will aid the identification of some of the physical elements in the models
described in the following sections of this chapter and is also relevant
to the modelling of the integrated circuit transistors described in later

chapters.
A full description of the theory and manufacturing processes involved

in the construction of these devices is given by Hamilton and Howard23.

Briefly, an integrated circuit n-p-n transistor such as that shown in
Figure 2-5, is built up layer by layer on a p-type substrate. The n+ buried

layer serves to reduce the series collector resistance of the transistor.

The n-type collector is grown by means of epitaxial growth and is part of
the same single crystal structure as the substrate. The p+ channels
 12

n+ emittercontact
p base contacts
n+ coltectorcontact
P+substrate contact

p+isolation channel
AP (base) I/
Ln/epitaxiat
p layer
collector)
ýýL4 n+
buried[aye
r
substrate

Figure 2.5 Construction of an Integrated Circuit N-P-N Transistor

 13

isolate the device from the rest of the integrated circuit. The p-type

base and n-type emitter are formed by a diffusion process and this is

followed by metalisation of the contacts.

2.2. Hybrid ?i Model

The hybrid pi model is the most popular linear model of a bipolar

transistor. Its appeal lies in its simplicity, its accuracy over a

wide frequency range and the ease of parameter determination.

Basic Hybrid Pi Model

The basic seven element, common emitter hybrid pi model is described

24 25.126
by Chua and Lin and Brown and is shown here in Figure 2.6. By com-

paring the model with Figure 2.5 which showed the construction of a tran-

sistor, it is possible to see the relevance of most of the elements.

Wc

b C

rce

e
Figure 2.6 Common Emitter Ilybrid Pi Transistor Model
 14

Of particular interest is r bbt" which Chua and Lin24 term the base

spreading resistance, stating that this is practically constant for a

given transistor and is typically between 5Q and 100 Q. Gray and

Searle27 describe how this element is inserted to allow for the transverse

voltage drops in the base region caused by the drift of base region

majority carriers flowing into the active region between the emitter and

collector. The inclusion of capacitors C and C to account

ble b1c
for the space charge layer capacitances is also described by Gray and

Searle27.

Brown26 describes how measurements of the y parameters (short-

circuit admittance parameters) can be used together with other selected

measurements to determine the values of the elements in the model for any

particular case.

It is generally agreed that the basic hybrid pi model is a good

linear model for frequencies up to a few megahertz.

2.2.2. Modified Hybrid Pi Model

As the frequency increases, the agreement between the basic hybrid

pi model and the performance of the transistor deteriorates.

Brayden28 describes a modified hybrid pi model in which the

collector-base depletion capacitance is split, taking part of it to b

instead of bI phase shift term is also included in the mutual

.A
conductance gm and, as at higher frequencies rbIc and r are swamped
ce
by the capacitance around them, these are eliminated from the model. it

is also generally agreed that interlead and case capacitances are also

significant. The new element C inserted by Brayden2B will thus

be
also include some interlead capacitance. Gray and Searle27 that
state

while the interlead capacitances in an integrated circuit component are

smaller than for the same component in discrete form, the junction
 15

bc

Cb Cce

Figure 2.7 Modified Hybrid_ Pi Model

 16

capacitances must still be accounted for.

The above thus leads to the modified hybrid pi model shown in

Figure 2.7 which Brown 25,26

states should be accurate'at frequencies

higher than 300 MHz, especially if some emitter lead inductance is also

included.

As with the basic hybrid pi model, Brown26 also describes the

measurements which must be made to determine the element values for this

model.

2.3. Ebers Moll Model

The Ebers Moll model is probably the most popular non-linear model

for bipolar transistors. This model was first published by Ebers and

M01127 in 1954 and there have been numerous variations on this original

model published since then. This model is included here because, even

though the work in this thesis was concerned mainly with small-signal

linear models, the final problem involved a transistor which entered the

non-linear region. Also, the non-linear modelling problem and the Ebers

Moll model are too important to be ignored. Here the notation used by

GetreU20 is used to differentiate between the stages in the development of

the models.

2.3.1. Basic Ebers Moll Model - EMI

A diagram of the basic Ebers Moll model, is given in Figure 2.8. The

basic model is a d. c. model that describes the behaviour of the transistor

in all modes of its operat -ion. The model is described by its reference

currents IF and IR' the currents through the diodes. GetreU20

describes this model as follows.

 17 1

Figure 2.8 Basic Ebers Moll Model - EM1

The diodes represent the base-emitter and base-collector junctions

of the transistor. I is the current that would flow across the

F*
base-emitter junction for a given base-emitter
Vbe ' if the voltage
collector region were replaced by an ohmic contact without disturbing the
base. I is the saturation current of this junction. Thus, for a
es
given value of V
be

q Vbe
(2.1)
kT
es

where q= electron charge

Boltzmann's constant

A6soluke. +etr,?
er&ture

Similarly, if the emitter were replaced by an ohmic contact without

disturbing the base and Ics is the saturation current of the collector-

base junction
 18

(2.2)
kT
IR=I e-
cs -11
Coupling between the junctions of the transistor is provided by the

base region and is modelled by the two current dependent current sources.

Thus

Ic=FIF-I (2.3)

where aF is the large signal forward current gain of a common base

transistor and is related to the large signal forward current gain (a

F)

of a common emitter transistor by the expression

aF (2.4)
F1aF

Similarly the base terminal current can be written

Ib ý' (1 -F )I (i -aR )I (2. S)

F+ R

where aR is the large signal reverse current gain of a commonbase

transistor and similarly to equation (2.4)

ol
R
R 1- aR (2.6)

The derivation of these equations is given by Ebers and M01129.

GetreU20 shows how the linearised version of this model for operation
in the forward active region reduces to the d. c. portion of the small-

signal hybrid pi model.

2.3.2. Extended Ebers Moll Model - EM2

The EM1 model is a basic d. c. model of a bipolar transistor. The

next stage in the development of the model is to take into account the

charge storage effects and to improve the d. c. representation.

 19

In the EM2 model described fully by GetreU20 and shown here in

Figure 2.9, the basic EMILmodel has been transformed. Expressions for the

currents I,,, 12 and 13 are derived by Getreu. Around this 011 model a

number of elements have been added. The three resistors r bb" rc-
cl,

and r are added to improve the d. c. characterisation. These resis-

eel
tors represent the ohmic resistance of the active region of the transis-

tor to its base, collector and emitter respectively. The charge

storage effects are modelled by (i) two junction capacitors CJc and
,
C which model the incremental fixed charges stored in the transistor's
Je '
space-charge layers for incremental changes in the associated junction

voltages, (ii) two diffusion capacitors and CDe which model

,C Dc -,
the charge associated with the mobile carriers in the transistor, and

(iii) the substrate capacitor which is important in integrated

,c sub -'
circuit transistors.

-I-
sub
r-- -
rcc Tc
S
III
J12
CDC Cj
r bb' ý
T C
Jljý I EM1 model
b 3T, i (transformed)
1 113
CDe-T
T, CJe
el i

ree/

Figure 2.9 Ebers Moll Model - EM2

 20

As with the E?I1 model, GetreU20demonstrates that this model reduces

to the linear hybrid pi model in the forward active region.

2.4. Other Non-Linear Models

In addition to the Ebers Moll models described in Section 2.3,

GetreU20 goes on to describe a further refinement of the Ebers Moll model

which he calls the EM3 model. This model includes a further two diodes

and a junction capacitor and models the base-width modulation and varia-

tion of current gain S with current and voltage, and some temperature

effects.

Another model described by GetreU20 is the Gummel Poon model which he

states is almost entirely concerned with improvements to the d. c. charact-

eristics of the EM3 model but is still basically equivalent to the 013

model.

Other popular models*include the Linvill30 lumped mpdel which, while

adhering closely to the physics of the device, introduces the new network

parameters of storance, diffusance and combinance, and the Beaufoy SparkeS31

model which is favoured by Brown26. He describes this model as lying

between the Ebers Moll and Linvill models with only two new network

elements namely storance and charge controlled current generators.

Hamilton et al. 32 in their comparison of models conclude that the

Ebers Moll, Linvill and Beaufoy Sparkes models are all similar in their

degree of approximation and give similar results for transient problems.

There are many variations on the models described here, Ruch33 gives

a list of references concerned with high-injection effects and Agajanian34

presents a total of 486 references in his bibliography on semiconductor
device modelling although not all of these are concerned with bipolar

transistors.
 21

2.5. Choice of Models

For each of the models described in this chapter, at least one set of

measurements enabling the evaluation of the element values is published.

For example, Orlik35 describes a set of electrical measurements that can be

made to determine the parameter values of a Gummel Poon model while

Roulston36 describes how the parameters of a similar model may be deter-

mined directly from fabrication data. Another method of determining the

element values for any particular model is to optimise the values of the

elements to give the best fit to a set of measured data. Probably the

first description of this method was by Bassett37.

It is obvious from the discussion in Section 2.4 that there is not a

universally accepted single model for the bipolar transistor. A number

of physical effects have been incorporated into models with varying

degrees of complexity but while Getreu2O states that the simplest existing

model adequate for the analysis to be undertaken should be used, Ruch33

suggests that time and effort spent on modelling even a complex model has
definite advantages, in particular that it saves the wasted time and

expense of evaluating many models or of using unsuitable models.

The author feels that model verification is an important aspect of

modelling, an opinion confirmed by Ruch33, and one which is often overlooked.

Similarly, there is virtually no guidance as to what action should be taken

if the model chosen is inadequate.

Thus, the author's aim was to present here a method by which models

could be verified and if necessary, better models could be evolved to the

complexity and accuracy required. It was also hoped that by so doing, the

knowledge of the further effects that must be incorporated in order to

produce acceptable models could be extended.

 22

The work by Bassett37 on optimising the element values of models

indicated a method by which models could be verified. The optimisation

techniques developed by the author to achieve the aims stated are

described in the later chapters of this thesis.

 23

CHAPTER 3

TRANSISTOR PARWETER MEASUREMENTS

AND ANALYSIS OF MODELS

In order to obtain element values of models of transistors and to check

their validity, a set of measurements characterising the transistor to be

modelled must be available. It was mentioned in the previous chapter that

there are prescribed measurements that may be made to enable the calculation

of element values for specific models.

Similarly, the validation and optimisation of a model require that func-

tions that can be measured for the transistor can also be calculated for the

model in order to make the necessary comparisons. Suitable measurements are

the short-circuit admittance, or y, parameters and the scattering, or s

parameters. These parameters are described later in this chapter and are

related to each other by a set of equations also given later. The y para-

meters of a network can be obtained by nodal analysis as shown in Section

Earlier work on network synthesis at Leicester has used coefficient

matching techniques to generate passive networks. In these cases the network

to be synthesized was represented by network polynomials such as those given

by Luca138, investigated by Ilegazi7 and Savage8 and shown in equations (3.1).

36p4 + 2o58p3_+ 6552p2 + 4638p + 36

36p 3+ 216p2+ 396p + 216

36P4 + 36p3_+ 72p2 + 36p + 36

-Y 12 (3.1)
36p 3+ 216p + 396P + 216

364 + + 1572p2 + 1183P + 36

22 _53323
3+ 2+
36p 216p 396p + 216

where y and y are the short circuit admittance parameters

11 ,y 12 22
and p is the complex frequency variable.
 24

Cutteridge and di Dlambro12 developed a method to calculate the

coefficients of p for a passive two-port network using the nodal admittance

matrix thus enabling network synthesis by coefficient matching to be per-

formed. In addition to his work with passive networks, di Mambr039 later

developed an analysis routine to calculate the network polynomial coeffic-

ients for networks containing active elements.

As it would be possible to convert measured s or y parameters in the

frequency domain into complex rational functions, for example in a similar

manner to that suggested by Enden and Groenendaal40 it would therefore be

,
possible to perform coefficient matching on a transistor model.

Coefficient matching techniques for network synthesis were first sugges-

ted by Calahan4l and this approach has been favoured for many years.

SavageB states that the advantage's of this method include good convergence

properties since the coefficients and derivatives are multi-linear functions

of the network elements with the consequent easy and accurate formulation of

the derivatives. Also, using coefficient matching the minimum size of the

network required to effectively synthesize the functions is well defined and,

as stated by Savage, it is clearly apparent when a realization has been

achieved.

However, there is the disadvantage that the formulation of the rational

functions in p, known as the approximation stage, is bound to introduce

errors before any matching is even begun. It would also be necessary to

decide somehow on the orders of the polynomials in p. One way would be

to choose orders giving less than a specified error at the approximation

stage.

This seems to be over-complicated and e:ýror-prone, thus the author

rejected coefficient matching as a suitable technique for this application,

choosing direct matching of the measur,ed parameters ýn the frequency domain.

This is not without precedent: Bassett37, one of the first to use optimisa-

tion methods for transistor modelling, chose to match the y parameters

 25

directly and most others have followed this example.

3.1. Y Parameters

The y parameters, or short-circuit admittance parameters, are a

means of describing a two-port network as a 'black box' network, i. e. one

whose interior is inaccessible. They are defined in terms of the port

small-signal voltages and currents, which are shown in Figure 3.1, and are

the complex functions y 112 Y122 Y2 , and Y22 where

Ivv Y11 Y12

1 1+ 2
(3.2)
12 Y21V1 + Y22 v2

at any particular bias point and frequency.

By examination of equations (3.2) it can be seen that the parameters

may be defined and measured by short-circuiting the input or output port

and taking the relevant admittance, thus

y1 vi when V0
2

yIl
12 v2 when V, 0
(3.3)
! Z-
Y21 when V
vi 20

!
Y22 = VZ when V
2 .1=0

hence their description as the short-circuit admittance parameters.

121-
v TWO-POTv =T

BLACKBOX 2
I

2
V, and V2are the port smaLL
signal voltages
11and12 are the port small signal currents

Figure 3.1 Two-Port 'Black Box'

 26

3.1.1. Nodal Analysis

Nodal analysis is a method by which the y parameters of a two-port

network can be calculated.

Consider an m-terminal passive network where, conventionally, nodes 1

and 2 are the input and output nodes respectively, and denote, say, node

m as a reference node. Spence 42 describes how applying Kirchhoff's

Current Law to each node in turn we can obtain the equations which in matrix

form give

I= (Y (3.4)
ind3V-

where y is the indefinite admittance matrix

ind
I is the vector of currents entering each node from external sources

and V is the vector of voltages at each node with reference to node m.

For a two-port network, since the internal nodes are inaccessible, there

cannot be any current injected into these internal nodes from external

sources, thus only I, and 12 are non-zero.

Spence also shows how Y

ind ' the indefinite admittance matrix, may be

formed by inspection of the network. Thus, each diagonal element Y

11
(i = I, m) is formed by the sum of the admittances incident at node i and

each off-diagonal element Yij (i 2 l, m; iýl, m; iý j) is formed by the

negative of the sum of all the admittances connected directly between nodes

i and j, where in general terms the admittance Y between any two nodes

is given by

Y= PC +G+1 (3.5)
pL

where p is the complex frequency variable

C is the capacitance in Farads between the two nodes

G is the conductance in Siemens between the two nodes

and L is the inductance in Henrys between the two nodes.

 27

It can be seen that for a passive network the indefinite admittance

matrix will always be symmetric.

This analysis can now be extended to include a voltage controlled

having Again to Spence, 42

current source mutual conductance gm referring

the terms in the indefinite admittance matrix due to the controlled source

can be formed by inspection. If the controlling voltage is between nodes

k and Z (voltage high at k) and the controlled current flows from

node i into node j then the additional terms in Y are

ind

gM to element Y of Y
ik ind

gM to element Y of Y
it ind
(3.6)
gM to element Y of Y
jk ind

+gM to element Y of Y
jq. ind

If a time constant T is also required, as in the model referred to in

PT
Section 2.2.2, then the terms become ±gme Thus we are able to form

the indefinite admittance matrix of an active network.

The n xn nodal admittance matrix, where n=m-1, is obtained by

deletion of the mth row and column of Y where node m is the

ind

reference node and is usually the earth node. At any particular value of

p, the nodal admittance matrix may be reduced using pivotal condensation

as described by Aitken 43 to obtain the y parameters as defined in equa-

tions (3.2). This procedure is described in detail in Section 3.1.3.

An alternative approach developed by di Hambro 4 and Cutteridge and

di Mambro12 making use of the network polynomials is described in Section

3.1.2.
 28

3.1.2. Generation of the Network Polynomial Coefficients from the Nodal

Admittance Matrix

As described in the previous section, for a two-port network containing

a total of n+I nodes, in which nodes 1 and 2 are the input and output

respectively, plus the earth node which is designated the reference node,

the nodal equations may be written

n
yijvj = Ii i=1, n (3.7)

where [Y] is the n xn nodal admittance matrix and I, =0, i=3, n .

Cutteridge and di flambro12 describe how the y parameters can be

written
A22
Yll 2'' 61122

Y12 A21
61122
(3.8)
A12
y
21 61122
All
Y22
1ý1122

where A is defined as the determinant of (YI

and Al, and A are the unsigned minors of

1133

They then describe how, for a purely resistive network, the y parameters

can-be obtained using Gaussian elimination and state that since 61,22 is

given by the product of the diagonal elements produced in rows 3 to n

inclusive, A 'ý12P and 622 can be obtained with only slightly

113 'ý2 1

more work than the calculation of A alone.

1122

Extending the analyses to networks containing reactive elements, it can

be seen that all the determinants involved are polynomials in the complex

frequency variable, p, and if inductances are present these include a

division by some power of p. Cutteridge and di Mambro state the upper

 29

limits of the powers of p in the numerator and denominator for RC, RL

and RLC networks. In their method any inverse powers of p are removed

by multiplying through by p as appropriate. Then, if m is the highest

order of the polynomials produced, m +1 values are assigned to p, say

p, ,i=O, m giving

pp2pm a A(p
000 0 0
2m
P, P, P, a A(pl)

(3.9)

pp2pm a A(p
mmm m m

where A(p) is the value of a cofactor evaluated at a particular value of p

and ail, i = Om , are the coefficients of the corresponding polynomial.

The (m +1) x (m +1) matrix on the left hand side of equation (3.9) is

the Vandermonde matrix and this Cutteridge and di Mambro invert using an

algorithm given by Traub 44 thus enabling calculation of the polynomial

coefficients.
12,, 45
Cutteridge and di Mambro also describe how the partial derivatives

of the polynomial coefficients with respect to the network elements may be

efficiently computed.

di Mambro4 further developed this technique to allow for the presence

of active elements without a time constant. A later, development, version

of a computer program written by di Mambr039 which allowed voltage controlled

current sources with time constants was received by the author. Time

introduce te PT into
constants, T, exponential terms the nodal admittance

matrix. di Mambro overcame this problem by approximating these terms using

 30

the first two terms of the series

xx l+T X2 X3
-5-1-+ 31

PT
thus replacing e by 1+ PT .

The author used this version of di Hambro's program J-n the early stages

of research, calculating the y parameters at any particular frequency from

the network polynomials. Unfortunately, this being a development version

there were some errors present, one of these occurring when inductors

were included. However, even on correcting this fault, the method was

still not satisfactory. In particular, the approximation 1+ pT for

the terms involving time constants was considered too inaccurate. Also,

it was felt that a more direct method using pivotal condensation, as des-

cribed in the next section, would be more suitable.

The results obtained using di Mambro's method and the method described

in the next section are compared in Section 3.1.4.

3.1.3. Calculation of the Y Parameters using Pivotal_Condensation

If all the values of the elements present in a network are inserted

into the nodal admittance matrix together with the value of the complex

frequency p at which the y parameters are required, then the Gaussian

elimination procedure described in the previous section can be performed

in a systematic procedure known as pivotal condensation which is described

by Aitken43.

In pivotal condensation, since the currents 1- to In+j of the n+InOAe

3
two-port 'black box' are zero, the n xn nodal admittance matrix [y] is

reduced to an (n -1) x (n -1) matrix using the formula

 31

Figure 3.2 Flow Chart for Pivotal Condensation Procedure

 32

1,
nj in ,
Y
ij y j 1, n-1
nn

where [YI] is the new matrix of size (n -1) x (n -1).

This procedure can then be repeated on (YIj and so on until a2 x2

matrix is obtained thus giving the y parameters.

The flow chart in Figure 3.2 gives the computational procedure for the

calculation of the y parameters from the nodal admittance matrix using

this technique.

In the FORTRANsubroutine written by the author, since the y parameters

for any particular network would be required at a number of frequencies, it

was decided to set up two-dimensional arrays G, L and C to contain the

contributions of conductance, inductance and capacitance respectively, to

the nodal admittance matrix. Then at any given frequency an element in

the nodal admittance matrix [Y] was given by

yG , I", ij PT (3.12)
ij +1+9me
ýLij -

PT
where 9e was the term due to a current source, which was included as
m
appropriate.

As complex arithmetic is available in FORTRANon modern computersthere was

no difficulty in the complex arithmetic due to p which is given by

j2Trf (3.13)

where f= frequency in Hertz.

3.1.4. Accuracy of Calculations

When calculating the y parameters using pivotal condensation it is

possible, just as with Gaussian elimination, to choose the pivotal element

in order to preserve the maximum accuracy. However, as pointed out by

Spence 42, the nodal admittance matrix should always have non-Azero elements
 33

on the diagonal, unless a node is not connected in which case it should be

eliminated anyway, therefore the matrix is generally well-conditioned.

Fairbrother and Basset46 also consider the effect of the frequency on

the calculation. They observe that, in particular, at very low or high

frequencies or at resonant frequencies, some elements in the matrix can

become very large compared with other elements in the matrix and that this

can lead to inaccuracies of calculation. However, they felt that even

this was an infrequent occurrence and took no special precautions.

In the subroutine written by di Mambr039, because of the in-built method

of setting up the frequency components used in the calculation of the network

polynomial coefficients, it was found that the calculations were grossly

inaccurate in the Mz frequency range required by the author. To overcome

this it was necessary to scale the values by multiplying the frequency values
9
by 10- and the reactive by 109.
component and time constant values

Bearinj in mind the above points the author decided first of all, that

in order to preserve the maximum accuracy in the pivotal condensation, that

section should be performed in double precision arithmetic. Remembering

that the computers in use had a 60 bit word length this, it was felt, should

provide adequate protection. Additionally, it was decided to apply the

factors of 109 and 10-9 in the same way as was necessary with di Mambrols

method, partly for reasons of consistency but also to avoid the presence of

both very large and very small values in the arithmetic.

Checks were then made using the circuit shown in Figure 3.3 at frequen-

cies of 0.5 MHz, 10 ?fflz and 100 MHz and with time constantS T on the current

generator of 0, -0.01 ns and +0.01 ns .t This circuit, which appeared

to be prone to loss of accuracy, was analysed manually to obtain as accurate

The notation used here is that a negative value Of T indicates a time

delay.
 34

cl
Ar C-'
1
2

0
Node 1- Input

Node 2- Output

Node 0- Ground

L1L (1-4) = 4.15E+00 nH

L2L (2-6) = 2.38E+01 nH

L3L (3-0) = 5.71E+00 nH

C1C (4-5) = 3.16E-02 nF

C2C (5-3) = 8.89E-02 nF

G1G (4-5) = 2.8SE-02 S

G2G (5-3) = 5.96E-03 S

g, g (6-3) = 6.84E-01 S

T0 -0.01 ns , +0.01 ns
S.
V across nodes 3

p= j2rrf

f=0.5 MHz 10 MHz 100 flHz

J. .

Figure 3.3 Circuit for Testing Analysis Methods

 35

values of the y parameters as possible. The definite nodal admittance

of this circuit, using w= 27f is given in equation (3.14).

matrix ,

00 0
-110
JwLj jwL,

01000 jwL. jwL

2

jWT
0- (G2 +jwc +g w-r)0
0+G 0 3WL +jwc, +gle 2 1ej
3
(3.14)

100 jwc 0
-..JWL jwL -+G
--! +jwC -G
1111

00 jw (C +C? )+GlG2 0
-G 2-jWC2 -G, -jwC, 1

jWT JWT i-I

0 0
-1jwL -gle gle jwL
2 2

Using the notation of Section 3.1.2 and cancelling terms where appropriate

we get

All 0
CiR CI
A12
w l -LlL2L
+GI+jGR+
3

A21 0
(GIC2 + G-,Cj) CJC? GG
ý122 - j (3.15)
w,' LlL2L3 w LlL2L3

L3Gl(G, ) +R) + C, +Cq +GIG2L, W2 (Ll+ L3)CIC2 CII

1&1122 - ý' L2L3
w -L, w LlL2

(G, +G2) (GlC2 (C2G, +CIG2)

4 LIL2 - +CI(G2+R)) + L3 GI
+jw
w L2

jWT
where R= Real part of g1e

I= Imaginary JWT
part of g1

w= 21rf

and f= frequency.
 36

Thus the y, parameters of this circuit are given by

Agg
Yll
1ý1122
A21
Y12 0
61122
(3.16)
A12
y 61122
21

Ali
A1122

The values of y 11 and y were evaluated from these expressions

21
using double precision accuracy on the Cyber 73 at the required values

of f and T. The results to 14 significant figures are given in

Table 3.1. Comparisons were then made between the values obtained using

the variations on the two methods of analysis described. The number of

significant figures that corresponded with those in Table 3.1 are given

in Table 3.2. Starting with a value of T=0 only, the methods

compared were:

di Mambrols method

in but with factors 10 9 9

P9 - as P of and 10- applied to

frequencies, reactive component values and time constants

as appropriate

pivotal condensation using double precision arithmetic

M9 - as in NI but with factors applied as in P9

N19S- as in 1.19 but using single precision arithmetic.

 37

t-1 e te)
00 r- 00 ý4 r*I P-4 «e C 0
M
r- (M
CD \O r-1 m
1--4 (N t'n Ln
CD le
rlý
"
Glý e -1 1-4 00 00
r- "0 le o 00 (2) CD fli
m %0 Lr) r-i tn CD 00
t4) cli r--1 0 m ý4 %0
lqt cq r- r, 1 ci t- e C) CD
t- CO 00
le 1-4 Ln CD r_
M Ln M Ln
-4 Ln --1 cý
"l 00 CD -zr Co CD M oo
1-1 CD \o P-4 CD 'Z --. i CD LA
L', -4
4. -1 r-ý Lt-i -4 Ln
r-i r-- Ln
ýi 0 ý Z (ý "ý c;
';
\o
-e %0 00 m
C% CD m CD c71 %0
P-4
%0 rq ei LA -4 r41 (N tn
4 00 (111 cý M Ln 00 -t cn
tn Itt t- ýo tn r, 4 r-- CD C%
" %0 Cý r- -4 tn CD LI- CD
nt "q
er) 00
t41 00 t41 r, 4 00 00 t41 00 0
P-4 00 r4 ýI r, 1 t-' f.-4 qt it,-
\C tn rq e (N e %0 Lr) r_
Ln m CD ul Cý m Ltli (M CD -, -4
\C m r- 1,0 C) \JD \O m r_ 4-4

c; lý
u
., 4
4.4
. r4

0
CD
m CD CD Ln r- r-
r- rq r- tn M P-4
let 00 00 Cä tn "4 Ul %0 CD
rq - Izt cý Lt') tl rq C% 1-
CD P-4
le (M M-4 Co \o gt N
Vý CD Ln e 00 t-
e Co CD e cý e P-4
rq 00 (71 rq -4 -4 r, 1 Ln r-
00 00 tn 00 NI CD Co Izt bi
cq Izt le r-4 00 c) (N cý 00
t4ý e en tn e CD tn " Ln
le Ln P-4 ýr -e ul 1-t le r-
Ln Ilt ul e-i

CD 0
-i t-ý rn r, In In 44
r-, rn r. ' r-,
0
+ + + + + + + + +
Ln Lri 00
rq IKT, M et 00
00 tn CD vj P-4
Ln P--1
\JD tn Lt)
r- 10 nt (ýq P-4 (N rq cz
le Lti le
Ln 0) Ln te gt
(14 %M cý 00 rý
r- vj "-1 %0 (-4 tn e
-I Lr) bl CD t4ý CD ei t-- 00
CD -4 00 CD \Z Ln (Z e 0
rn \Z (N tn 00 \JD tn tn c
00 \o le 00 Ln 00 00 r-- CD
r-4 Co F-4 CN 00 rý (N 00 Lf)
cý oh cl- 0

.0

Ln (Z CD LA C CD Ln CD (Z
; ; ; ; 7
zi c c c c c
(Z C
 38

0.5 ISIZ 10 MHz 100 Mllz

Yll I I [ I Y21
Y21 Yll Y21 Yll
T ns Method
R I R I R I R I R I R I

P9 13 14 13 14 14 14 14 14 14 14 14 14

M 14 11 14 14 14 12 14 14 14 14 13 14
0
M9 14 11 14 14 14 12 14 14 14 14 14 14

m9s 13 9 13 14 13 11 13 14 14 13 13 14

P9 13 10 9 10 8 7 7 7 6 4 5 5
0.01
M9 14 10 14 14 14 11 14 14 14 13 14 14

P9 13 10 9 10 8 7 7 7 5 4 5 5
0.01
M9 14 11 14 14 14 11 14 14 ý14 13 14 14
1

Table 3.2 Number of Correct Significant Figures in Computed Values

The results using method P are not shown in Table 3.2 because the

errors were so large that none of the values corresponded even to one signi-

ficant figure and many values were an order of magnitude in error.

As can be seen from Table 3.2, with T=0 method P9 was the most

accurate closely followed by methods M9 and M. Whilst the 109 factors

had a major effect on method P, there was only a very slight improvement

when they were applied to method M. However, the differences between the

single and double length arithmetic versions of methods M9 and M9S were

more marked although M9S might still be considered accurate enough for

most purposes.
Taking the two most accurate methods, P9 and M9 for further compari-
,
sons with time constants T of -0.01 ns and +0.01 ns , it can be seen

from Table 3.2 that at 0.5 MHz method P9 has lost some accuracy and that,

as might be expected, this loss of accuracy rises as the frequency increases

so that by a frequency of 100 NIHz the results are becoming unacceptable.

Meanwhile, method M9 has suffered negligible loss of accuracy and is

unaffected by the magnitude of the frequency. Thus, although method P9

 39

is slightly more accurate when no time constants are applied to any current

generators, method M9 is significantly superior when time constants are

applied.

In all cases the computed values for y 12 and y 22 had absolute errors

10- 20.
of less than

Further Considerations on the Choice of Algorithm

In addition to the accuracy of calculations, when choosing a computer

algorithm, consideration should also be given to the speed of calculation

and the storage space required by the algorithm.

For the given analyses, method M9 took 90% of the computation time

taken by method P9 while M9S took 50% of the time taken by M9

, .

The storage space required by N19 was 70t 4"fiadt r-almead. by P9

,

while M9S required BOX of that-raquitred -by M9

.
Thus, method M9 in addition to being more suitable from accuracy
,
considerations, was also better than P9 in terms of speed and storage space.

If the accuracy of method M9S is considered sufficient then the obvious

savings in calculation time and storage space make it the best choice.

Taking a cautious approach the author elected to use method M9

.

3.2. S Parameters

In Section 3.1 it was described how the y parameters of a two-port

network could be measured by short-circuiting the input and output ports in

turn and measuring the appropriate admittances. However, at frequencies

above 100 NIHz it becomes increasingly difficult to measure the y parameters

because ok the difficulty in obtaining good short circuits and because short

circuits often cause oscillation.

The s, or scattering, parameters are similar to the y parameters in

 40

that they describe the inputs and outputs of a two-port black box, but have

the advantage that they are easily measured up to the GHz frequency range.

Weinert 47 describes how the s parameters are defined in terms of the

incident and reflected parameters (a,, bl) and (a., b. ) which are shown

in Figure 3.4 and are defined in equations (3.17).

1
[a -
vrz, + rz-
0
I,
0

2 rz 0
0 (3.17) '
12
+YZ0 12
/Z--
0

12- r-
b2 'ý '(Z 12
rz- 0
,
0

The s parameters are then given by

b, = s1la, + Sl2a2
(3.18)
s a1S a2'
21 22

12
2
vi
TWO-PORT CL2
V2 z0
bl b2
BLACKBOX

Fig. 3.4 Incident and Reflected Parameters Used to Define the s Parameters
 41

3.2.1. Measurement of the s Parameters

Just as the y parameters can be defined and measured by short-

circuiting the input or output port and taking the appropriate admittance

measurement, so the s parameters may be defined and measured by setting

the incident parameters a, or a2 to zero and taking the appropriate

ratio. Therefore we can express the S parameters as:

b,
s when a2ý0

b
s -L when a=0
12 a21
(3.19)

s221 when a0
a12

s222 when a, 0
a2

The s parameters can therefore be expressed as ratios of the reflected

and incident voltages at the input and output ports when one port is matched.

In the introduction of reference (48) it is explained how the matching

of the input or output port necessary to set a1 or a2 respectively to

zero may be achieved by using SOQ transmission lines to connect the device

under test and then terminating the transmission line in its characteristic
impedance, Z0 in Figure 3.4. A measurement jig is described by Hewlett

Packard 49, incorporating the 500 transmission lines and SOQ loads in

such a way that the device under test is easily reversed thereby enabling the

measurement of the reverse parameters s 22 and s 12 in the same set-up as

the forward parameters s and s The jig is used in conjunction

11 21

with a Hewlett Packard 8405A vector voltmeter which is used to measure the

required voltages and phase angles.

Thus the s parameters may be easily measured over a range of frequen-

cies for any particular device.

 42

3.2.2. Calculation of the s Parameters

The s parameters of a model may be obtained by first calculating the

y parameters of the model using one of the methods described earlier in

this chapter and then applying the conversion formulae given in reference

(49) and reproduced as equations (3.20) below.

Yll)(1 *d Y22) + Y12Y21

sil
I

12 dy
(3.20)
2 y2l
21 dy

(1 + Yll)(1
-d Y22) + Y12Y21
22 =
y

where cý = (1 + Yll)(1 + Y22) - Y12Y21

For conversion from the s parameters to the y parameters equations

(3.21) may be used.

sll)(l-+ S22) + S12S21

Yll
ds

2 sig
Y12- ds
(3.21)
2 IZ1
Y21
dt,
(I + S22) + Si2s2i
Y22 ý-
ds

where ds = (i +s 11)(1 + S22) - S12S21

The y parameters given in equations (3.20) and (3.21) are all norma-

lised to Z, therefore if y! is one of the actual parameters,

0 ij
conversion is performed using

ij yii
z01,2;
j=1,2 (3.22)
 43

CHAPTER 4

NUMERICAL OPTIMISATION

Numerical optimisation is a widely studied subject area with a vast

range of applications, not just in electronics modelling but encompassing

modelling, design and manufacturing problems in many disciplines. At its

most basic, optimisation can be described as a means by which the best value

of some variable may be obtained. The definition of 'best' depends on the

application but usually some error function is defined* which indicates the

difference between the ideal value of a parameter which is dependent on one

or more variables and the value of that parameter currently available. The

optimisation problem thus becomes one of minimising the error function. In

most problems there are a number of criteria, therefore the sum of squares

function

F(x) = (4.1)

where fi (x), i=1, m, are the individual error functions dependent on a

vector of variables x, is often used as a single overall error indicator.

The sum of squares function is also useful in the solution of nonlinear sim-

ultaneous equations. If the equations are

fi (x) = i'm (4.2)

then it can be seen that a zero minimum of the sum of squares function would

give a solution of the equations.

There are numerous other formulations that are suitable for an over-

all error function. Another popular one is the minimax function which

takes the value

F(x) = maxIf (X)l i= i'm (4.3)

i

where f1 (x), i=l, m are individual error functions. The general term

for the single function to be minimised is the objective function.

 44

The optimum value of all sum of squares problems as defined in

equation (4.1) is when F(x) =0, although a zero minimum may not necess-

arily exist. If any set of values of x is found such that F(x) =0

then a global minimum has been found since F(x) cannot be negative for

real values of the individual error functions fi (X) However, whether or

.

not a zero minimum of F(x) exists, the global minimum is defined to be at

the vector of values x where

F (x ) F (x) (4.4)
ý<

for all other values of x.

In any particular region R there might be a non-zero minimum at x1 where

for all values of x within R but where

F(x*) < F(xl) (4.5)

.

The vector xI in this case is termed a local minimum.

Global optimisation is a very difficult and time consuming operation,

not least because of the uncertainty of knowing whether the lowest minimum

found so far is the lowest minimum overall. Dixon, Gomulka and Szeg650

in their survey of global optimisation methods observe that the situation

with regard to solving global optimisation problems is very poor. The most

popular methods are random methods, including multistart methods where local

minimisation is performed many times from different randomly-selected start-

ing points, and those based on Branin's trajectory method, reviewed by Dixon

50. At Leicester University, Price5l, *52 has worked on global opti-

et al.

misation techniques with some success.

The distinction should also be made between constrained and uncon-

strained optimisation. Unconstrained optimisation is when the solution

values may occur at any values between and +- In many real optimi-
-to .

sat ion problems, however, there are restrictions on the acceptable solution
 45

values and so constraints must be imposed. In practice many constrained

optimisation problems can be treated as unconstrained by the application of

transformations to the variables such as those described by BoX53. One of

the simplest transformations, which is used in the author's work as des-

cribed later, is to work in the domain of the logarithms of the variables,

thus constraining the variables to take positive values only. Also,

Lootsma54 has reviewed a number of methods whereby more complex constraints

can be imposed by suitable definition of the objective function to which

unconstrained optimisation methods may then be applied.

It can be seen, therefore, that most design problems are in fact

constrained local minimisation problems.

In the case of the modelling of a bipolar transistor in terms of the

standard elements of resistance, capacitance and inductance together with

voltage controlled current sources, the global minimum could be likened to

the app.roximation to cos x by the series

ý12 ý, X4 X6 (x < (4.6)

Cos x-+ 2! 41 6!

A reasonable approximation to Cos x requires at least the first n

terms of the series, where n is given by

xm<M! (4.7)

where m= 2(n - 1) .

Thereafter each further term improves the accuracy of the approximation but

it could be argued that only an infinite number of terms will give the true

value.

Similarly, in a model of a bipolar transistor, some elements, for

example a voltage controlled current source, are necessary to obtain any

reasonable model whilst the inclusion of other elements is desirable in order

to obtain higher accuracy, but it would require an infinite number of elements

to get a true match. Therefore, in this context a good model should contain
 46

only those elements that h&ve a, majewafficttogether with those that make a

significant contribution to the accuracy of the model. Thus, if a model

is built up element by element, as each new significant element is added to

the model a new local minimum is found.

The modelling strategy developed by the author is dealt with in

Chapter S. The remainder of this chapter is concerned firstly with optimi-

sation methods that have been used in similar applications to those of the

author and secondly with the optimisation method actually used by the author.

4.1. Available Methods

Numerical methods of optimisation can be divided into three main

categories.

Ci) Methods using function values only.

(ii) Methods using derivatives.

(iii) Methods designed for minimising sum of squares functions.

Brief details of some of the algorithms in each category follow, together

with an indication of applications whe-re they have been found useful.

4.1.1. Algorithms Using Function Values Only

In this first category are random and pattern searches. Random

methods were discussed briefly at the beginning of this chapter. Pattern

searches are more systematic. A simple pattern search was used by KnudsenS5

to determine model parameters for MOStransistor models. The attractions

of this method are its simplicity and its suitability for small computers.

The disadvantage, as noted by Knudsen, is its tendency to get trapped in

narrow, curved valleys in the function. The simplex method of Nelder and

Mead56, also studied by Price57 assisted by the author58, forms a regular

polyhedron (or simplex) in n-dimensional space. On examination of the

 47

function values at the vertices of the simplex, it is then expanded, contrac-

ted or reflected about a vertex, the process being repeated until a suitable

minimum is located. A simplex algorithm was used by Mathews and Ajose2 to

obtain optimum values of elements in particular models of microwave FETs and

bipolar transistors by matching s parameter data of the devices. Durbin,

Montaron and Heydemann59 used a combination of simple pattern and random

searches together with a simplex algorithm in the d. c. optimisation of elec-

trical circuits, although they noted that their method was limited due to
51#52
the large computing time required. The global search algorithm of Price

combines a random search with a simplex technique by initially evaluating

the objective function at N random points thereafter selecting each new

point to be evaluated using a randomly selected simplex of points from the N

best points currently available. This algorithm was used by the author6O

to obtain suitable starting values for a trial model of a p-n-p transistor as

described in the next chapter.

4.1.2. Algorithms Using Derivatives

The classic optimisation methods are those of steepest descent and

Newton Raphson which use first and second derivatives respectively.

In the method of steepest descent, the gradient vector g of the

objective function is used as a search direction, hence from any position

given by the vector' xk with associated gradient vector gk, an improved

k+1 k the
position x should be obtainedusing +o inclicate iteraficmý by

k+l k9k
x = xk (4.8)

where Xk is chosen by linear search to minimise the objective function.

It is generally agreed that this method has bad convergence properties,

however, a variation on the method was used by Agnew" who, for the optimisa-

tion of filter circuits, used only active residuals at any particular stage

to generate the smallest step length necessary to produce a specified

 48

"reasonable" reduction in a minimax error function. Mention should also

be made here of the gradient descent method of Cutteridge62 (with later

improvements by Henderson" and Dowson9), which included the Hessian matrix

of second partial derivatives to generate a "curve of steepest descent",

which was used at the initial stage in a two-part algorithm for the solution

of, amongst other applications, a particularly difficult problem encountered

in producing a d. c. Ebers Moll model of a bipolar transistor.

Algorithms using conjugate gradients are more popular and effective

methods using first derivatives. The most well-known of these is by

Fletcher and ReeveS63. In this method, using the superscripts k to

denote the iteration number and T to denote the transpose of a vector, the

sequence of points in the iterative scheme is given by

k+ 1k k+l
x_ Xky
(4.9)
k+l kkk
y9y

where Xk is the linear search parameter minimising the objective function

gk is the gradient vector

ly" is the Vector of Conýuq&tdAirer-tions

and ak is given by

k
a0 for k=O (n+l), 2(n+l)
,

(ýk) T (ýk) (4.10)

(ak-1) otherwise
ýk

On a quadratic function this method locates the minimum in at most itera-

n

tions, where n is the number of variables in x. The method was used

by both Krzeczkowski6 and Savage8 for the initial stages in the synthesis

of lumped linear three-terminal networks.

The Newton Raphson method is based on the Taylor series expansion

F (X + 6) =FW+6Tg+... (4.11)

where g is the gradient vector of F.

 49

Ignoring higher order terms, differentiating and equating the first deriva-

tives to zero, i. e. the equation for a minimum of F, gives

0g+ H6 (4.12)

H is the Hessian matrix of second partial derivatives of F it

where

is usual, as with the steepest descent algorithm, to include a linear search

along the direction given by 6, thus the Newton Raphson iteration becomes

k+l k) -1
=xk-Xk (H g (4.13)

where Xk is chosen to minimise the objective function.

The Newton Raphson algorithm thus requires the evaluation of the Hessian

matrix at each iteration. Variable metric algorithms are a class of algor-

ithms which attempt to approximate equation C4.13) using the first deriva-

tives only. Among these algorithms are those of Davidon 64 and the later,

very successful algorithm by Fletcher and Powell65. Variable metric

algorithms may be classified as quasi-Newton if they satisfy the criteria

0
that S the initial approximation to the inverse Hessian matrix, is posi-

tive definite and that

k+l
s 6khk (4.14)

k k+l k
where x-x

k+l k
hk

thus giving the Newton Raphson iteration for a quadratic function.

A variable metric algorithm was used by Lanca and Nichols" for the design

of networks to time-domain specifications and the method of Fletcher and

Powell was used by Orlik-35 to obtain some of the element values in Gummel

Poon models of integrated circuit transistors.

 so

4.1.3. Algorithms Designed for Minimising Sum of Squares Functions

When minimising sum of squares functions

i-i

where fi Cx) ,i=l, m, are the component functions dependent on the vector of

n variables x, then, in addition to the values of the objective func-

tion and its derivatives, there are also available the values of the compon-

ent functions and their derivatives.

If we say that we are attempting to obtain

f1() =0I=1, m (4.16)

then taking the Taylor series expansion of each component function gives

n afi(x)
(X + 6) =f
fi-i-j11 (x) + 6. + i'm (4.17)
axj 3

which, ignoring higher order terms, gives

f (X + 6) =f (x) + j6

where J is the Jacobian matrix of first derivatives.

Equating this to zero as required from equation (4.16) we get

16
.

If m=n and J is non-singular then

6=- J-1 f (x) (4.20)

.

If m>n then we can use

6=-QT J) -1 JTf (x)

.

DimmerIl describes how equation (4.21) is derived by finding the least

squares solution of equation (4.19).

 51

Thus, introducing a linear search parameter we get the Gauss

Newton iteration

k+1 k kT kkTk
xx_ Xk (P )i (i f (x (4.22)

k) k)=0
Dimmer" shows that provided a point at which (J f(x has not
k)ik
been reached and (J is non-singular, then there will be a value of

X>0 which will reduce the objective function. Dimmer also emphasises

the differences between the Gauss Newton and Newton Raphson methods and

notes a superiority of performance by the Gauss Newton method over the Newton

Raphson method on sum of squares problems.

The Gauss Newton method has been used extensively at Leicester

University on optimisation problems arising from the synthesis of electrical

networks.
Most other methods designed for minimising sum of squares functions

are based on the Gauss Newton method and are concerned either with avoiding

of the matrix to be inverted as with the Levenberý57 Marquardt6 8

singularity

method or with the use of approximations for the Jacobian matrix thus

requiring the calculation of the component function values only as in the

6 9.
method by Broyden

4.2. Some Details of the Optimisation Algorithm Chosen

It can be seen from section 4.1 that just as no single model of the

bipolar transistor has found universal favour, so no single optimisation

algorithm has become the standard even for applications of a similar nature.

It was mentioned in section 4.1.1 that the author used the global

search of Price to obtain the starting values for one of the trial problems;

however it was the author's intention to, initially, use the two-part algor-
62A 10 for
ithm of Cutteridge and Henderson the major part of the model opti-

misation. At this early stage the reason for this choice was largely
 52

because of familiarity with the algorithm together with first-hand observa-

tion of its efficacy.

The two parts of this algorithm consisted of a gradient descent

section, mentioned in section 4.1.2, and a modified Gauss Newton section.

The author soon discovered that the gradient descent section was not in fact

needed in her particular application since the Gauss Newton section proved

quite adequate for the optimisation involved. Furthermore, benefits of

this second section from the modelling point of view also became apparent.

Therefore, this Gauss Newton algorithm was the major optimisation method used

by the author and included several modifications due to Henderson".

One of the modifications was to limit the change in any value x

in one iteration to some predetermined limit. This was achieved in the

k
linear search used to choose the value of X The linear search included

a preliminary search using trial values of Xk calculated from the Fibonacci

kkkkk
series Xo , 2XO , 3X Sx where Xo was a given starting value.
0, 0 ...
The aim of this preliminary search was to attempt to bracket a minimum.

Should this not be achieved, the preliminary search continued until the pre-

set maximum change in value was reached in each variable.

If a bracket was obtained in the preliminary search then the minimum

was located more accurately using a quadratic interpolation algorithm which

was safeguarded against the occurrence of non-unimodal functions and against

slowness of convergence onto the minimum through taking new points for eval-

uation too close to the current three points used in the quadratic interpola-

tion.

Henderson also categorised the possible terminations of the algorithm.

The first of these, which was not really relevant in the author's application,

concerned conditions when it was not advisable to attempt using the algorithm.

The second concerned successful termination which was based on the absolute

values of the correction terms 6 calculated by Gauss Newton. The third

category described criteria for failure due to the JTJ matrix being or
 53

almost being singular. The fourth and final category predicted ultimate

failure of the algorithm. There were three criteria in this category:

(i) if the number of iterations exceeded a certain number, (ii) if 6k>

max
100 60, where 6k max 161 at iteration k, and (iii) if 6k> 6k-1
max max - max max
k k-1
on 10 successive iterations with the increase (6 -6 larger at each
max max
successive iteration, i. e. the rate of increase accelerating.
The author, like Dimmer1l, is of the opinion that termination pre-

diction, especially of ultimate failure, is an important aspect of any opti-

misation algorithm. When the optimisation problem is one of developing

models it is most important that unsuccessful models be detected and dis-

carded as early as possible. We thus have a further justification for the

author's use of this algorithm.

 54

CHAPTER S

DEVELOPMENTOF THE MODELLING ALGORITHM

Philips Research Laboratories (PRL) who provided data for the author

specified certain conditions for their models. One of these was that the

s parameters be fitted directly. The reasons given for this request were

that (i) these were the quantities actually measured, and (ii) operation on

the equivalent sets of y parameters obtained applying the standard trans-

formation formulae would alter the weightings of the fit. PRL also provided

models which they thought would provide a suitable starting point for a model

optimisation exercise.

Thus, this chapter first describes how the component parts of the

three previous chapters were combined together to form a single method des-

igned, initially, to calculate the optimum element values for any particular

model to fit the s parameter measurements of any particular device subject

to the conditions set by PRL. Details are then given of the possible ways

in which models may be improved and of the methods chosen by the author.

At each stage examples are given of the results obtained for the

first model developed for PRL, that of a lateral p-n-p transistor. The

results of this exercise were published jointly by Cutteridge and Dowson60.

It should be noted that although the major criteria for the final modelling

algorithm were developed during this first exercise, the process of refining

and confirming the techniques continued throughout the period of research and

some suggestions for further improvements are included in Chapter 8.

5.1. Details of the First Modelling Exercise

The first set of data provided by PRL was for a lateral p-n-p
transistor connected in common emitter configuration and consisted of the

set of s parameter measurements given in Table 5.1. The data in its

original form is included in Appendix 1.

 55

S S12 S S
Zo -, 5006 11 21 22

Frequency Modulus Phase Modulus Phase Modulus Phase Modulus Phase

Miz dR degrees dB degrees dB degrees dB degrees

0.5 -0.10 -1.40 --- -12,50 178.10 0.00 -0.20

2.0 -5.40 -64.70 88.70 -12.50 169.90 0.00 -0.30
-0.20
5.0 -11.20 -57.20 84.00 -12.80 155.00 0.00 -0.50
-0.40
10.0 -19.40 -52.00 76.30 -14.00 132.90 0.00 -0.90
-1.20

A measurement of S12 at 0.5 NHz was not available

Table 5.1 Measured s Parameters of a Lateral P-N-P. Transistor

It can be seen that the measurements for each of the s parameters are in

terms of modulus in dB and phase in degrees and are given at four frequen-

cies in the range 0.5 DIHz to 10 MHz except for s for which a measurement
12
at 0.5 MHz was not available. The accuracies of these measurements were

given as "about t O. S dBII for the modulus and "phase angles near 0" are

certain to ± 0.50, and near 180* to about ±2*11.

In addition to these measurements, PRL provided a simple model which

they suggested would be a good starting point for the derivation of a model.

This model, which is shown in Figure 5.1, consisted of four nodes (including

the node 0) and five elements of which two elements were voltage con-

trolled current sources. It was suggested that these generators should

take specific values with zero time constants. These values are given in

the diagram. It was felt that this model was too basic and so a further

three elements were added by the author. The global search technique of
51,52
Price was then used to minimise the single overall error function des-

cribed in Section 5.2.1. in order to obtain suitable starting values for the

unknown element values. This model, which is shown in Figure 5.2, was then

used as the init-ial trial model for the lateral p-n-p transistor.
 S6

1 2 No. oF nodes =i
c No. oF elements
(2-3) = 3.53E-03 S
(T :- 0)
(V across nodes P
(0-3) = 2.07E-05 S
(T = 0)
(V across nodes 31)

Node I 9ase
Node 2 Collector
Node 0 Substrate and Emitter

Figure 5.1 Simple Model of Lateral P-N-P Transistor

2 No. oF nodes =i
1 No. oF elements
C.
b
g (2-3) = 3ME-03 S
(T = 0)
(V across nodes 3 1)
(0-3) = 2.0? E-05 S
(T = 0)
(U across nodes 31)
G (1-3) G. 12P--10 S
C (1-3) B. GSE-02 nF
G (2-3) S. 22E-OS S
C (2-3) S. OGP--04nF
C (0-3) G. 77E-03 S
C (0-3) 3.71E-06 nF

Node I- Base
Node 2- rollector
Node 0- Substrote and Ernitter

Figure 5.2 Initial Model of Lateral P-N-P Transistor

 57

5.2. Construction of an Algorithm to Optimise Model Element Values

Given a set of measured s parameters as shown in Table 5.1 and a

trial model as shown in Figure 5.2, the first requirement by the optimisa-

tion algorithm was for an error function to be constructed to give an indica-

tion of how good the model was. Since the author had chosen the Gauss

Newton algorithm described in Section 4.2, the first partial derivatives of

the objective function were also required. The element values of the model

were obviously the variables, but it was necessary to decide whether any con-

straints should be applied to these. Finally, it has already been stated

that, in the author's view, an acceptable first stage model would be

obtained when a local minimum had been found by the optimisation algorithm.

Therefore, the exit criteria of the Gauss Newton algorithm are also discussed.

5.2.1. Construction of the Error Function

Techniques for the calculation of the s parameters of a model were

discussed in Chapter 3. The method first used by the author, and used for

the examples given in this chapter, was to calculate the y parameters using

the method of dillambro described in Section 3.1.2. and to then use the con-

version formulae given in Section 3.2.2. to obtain*the s parameters.

The calculation of the yý parameters was later done by the method of pivo-

tal condensation as described in Section 3.1.3, and all the examples in the

following chapters used that method.

PRL had requested that the s parameters be fitted directly and it

was decided that individual error functions should be calculated at each of

the given frequency values for each of the s parameters. There was a

choice of whether the individual error functions should be in terms of the

real and imaginary parts of the s parameters or whether the modulus and

phase should be matched. One advantage seen in favour of matching the

former was that by weighting equally the errors in the real and imaginary
 58

parts of the s parameters the problems of choosing the appropriate weights

for errors in the modulus in. decibels and in the phase in degrees could be

avoided. However, it was felt that it would be preferable to consider the

errors in terms of decibels and degrees and that by allowing the weightings

on these to be stated individually for the different frequencies and for

each of the s parameters it would give an engineer the opportunity to

emphasise the most important region of operation of the model in question.

The weightings chosen by the author, based on discussions with PRL, were such

that an error of 1 dB in the modulus was equivalent to a 10" error in the

phase. It was felt that absolute errors were more appropriate than rela-

tive errors, particularly in the case of the phase errors where the paradox

exists that a small. absolute error in a phase angle close to 0* gives a

large relative error while a similar absolute error in a phase angle close

to 3600 gives a totally different relative error. A slightly different

problem occurs for the modulus when it is stated in decibels. Since the

decibel is basically a logarithmic unit, a large negative value means that

there is negligible gain and many engineers would feel that an error of ldB

in a modulus of -5dB was less significant than an error of ldB in SdB

.
In view of the stated errors in the measurements by PRL, it was decided that

the weighted absolute errors would be suitable for both types of individual

error functions.

The overall error function to be minimised was then made up of the

sum of the squares of the weighted individual errors in the modulus and phase

of each of the s parameters at each of the frequencies for which measure-

ments were given.

In the early stages, for ease of programming, values of dB

-77.40
and 89.600 were given for the modulus and phase of s 12 at 0.5 MHz
,
these values being chosen by eye from plots of S12 against frequency. Thus

the initial problem was one of minimising a sum of squares function made up

of 32 individual functions in 8 unknowns (the elements in Figure 5.2. ).

 59

5.2.2. Generation of the Jacobian Matrix

When analytic first derivatives are not available as was the case
here, then numerical estimates of the Jacobian matrix can be used. The

author first experimented using central differences so that to evaluate

afi (ý) / axia displacement vector dx is defined such that dxk = d. x k=j
y
and 6xk0O, k ýj, where d is some small value, then

af. (x) fý (X + ax) -fi (X - ax)

1--1----.
3x. 2 6x. (5.1)

By progressively reducing the value of d and recalculating using equation

(S. 1) the accuracy of the value of af (X)/ax can be checked. To do this

, 1 3
during an optimisation run produces prohibitive computing times. However, -

some results were obtained which proved that the value of Sx recommended
I
by Henderson" gave sufficient accuracy. This value was given by

7
2 6x 10- (1 + Ix 1) (5.2)
3= i ,

which was based on the reasoning of Gill and Murray70, who suggest a fixed

small value governed by the word length of the computer in use, assuming
Ix to be of the order unity, but which also allowed for the possibility
iI of

large values of jxjj

.
One further point raised by this choice is that the minimum value of
2.6x 7. in many models
will be 10- Since of bipolar transistors there will
i
be elements which have values smaller than this as seen in the in
example

Figure 5.2, this provides a further reason for the use of the factors

applied to the element and frequency values described in Chapter 3.

The use of central differences requires two function evaluations for

each element of the Jacobian matrix. The number of function evaluations

can be reduced if forward differences are used since the current function

values should already be available, Although the use of forward differences

reduces the accuracy of the estimates of the Jacobian matrix, Brown and
 60

DenniS71 found them to be adequate in methods for sum of squares function

minimisation.

5.2.3. Application of Constraints

Although it is not strictly necessary for model elements to only

take physically feasible values, most engineers find models more acceptable

if they can visualise the component elements. An interesting exception was

one of the initial models provided by PRL, who were most insistent that

model elements should have some physical significance, which included a nega-

tive-valued capacitor. It therefore seemed reasonable to the author that

model elements of capacitance, resistance and inductance should be constrained

to take positive values on.ly.

With regard to voltage controlled current, sources represented by

g Vejwr where V is the controlling voltage and T is the time constant,

then A, negative value of g. would merely indicate that the current was

flowing in the opposite direction to that with a positive value. However,

it could generally be assumed that the model is sufficiently correct that

the value of g could be restricted to positive values only. The time

constant T would normally be expected to take a small negative value indi-

cating a time delay. However it was found that there were many instances

where a small positive value was the optimum value for a particular stage in

the development of a model. The author decided to allow the time constant

to take both negative and positive values since it could be that a positive

value at any particular stage merely compensated for elements which were not

yet present in the model. Certainly there were cases where a positive time

constant at an early stage became negative*as the model development pro-

gressed. So that the same constraints could be applied to all the variables,

it was decided that the value to be optimised should be (1 + T)nS. Since

the value Of T should never be more negative than ns then (1 + T)

-I

should always be positive.

 61

Thus, since all the variables were constrained to take positive

values only, a logarithmic transformation could be applied. Although

other transformations, such as a square transformation, could be applied

in this situation, the logarithmic transformation has been used with some

success at Leicester University and Krzeczkowski6 found that the natural

scaling exhibited by the use of this transformation aided the convergence of

his optimisation algorithms.

5.2.4. Exit Criteria

In Section 4.2 some details were given of the various termination

modes of the Gauss Newton algorithm categorised by Henderson. Obviously

the one aimed for was the successful termination characterised by all the

individual corrections 6 to the element values becoming negligible. if

this occurred then one could say that the elements were all necessary as

part of the present model and that the optimum element values-had been ob-

tained for that particular model. The successful termination criterion

Id 10-8 i=1, n , where 6 is the vector
was set at when iI< of correc-

tions in the logarithmic domain.

If one of the failure termination modes occurred then it could be

that the model was not suitable and that it should be modified. It would

then become necessary to study the model, the element values and the reas-

ons for failure of the optimisation algorithm in order to determine what

action should be taken.

5.3. Example of Results

Applying the algorithm described in Section 5.2 to the model shown

in Figure 5.2, the algorithm failed to converge. The reason for failure

was that there were too many successive increases in the rate of increase of

the maximum modulus correction. This failure mode, which we will refer to
 62

as failure mode 1, was in fact the most common mode encountered in all the

examples. At this stage, failure mode 1 had to occur on 10 successive iter-

ations before the optimisation algorithm terminated and in this example the

maximum modulus correction was being applied to element G the conductor

1-3'
between nodes 1 and 3. , At termination this element had reduced in value from

10 11
6.12 x 10- S to 9.91 x 10- S while the correction term in the logarith-

mic domain had changed from -8-01 X 104 to -4.95 x 105 . Meanwhile, all

the other element values were virtually unchanged. Taking this progression

to its limit, because of the logarithmic transformation this would make G,

_3
zero, i. e. open-circuit. The author therefore removed that element from the

model and restarted the optimisation algorithm. Again the algorithm failed

in failure mode 1. This time the element causing the failure was C
3-0
which was increasing in value. It was decided that this element should

also be removed open-circuit. On restarting the algorithm, one of the

current generators, both of which had been allowed to vary, caused a further

failure mode 1. The value of 9

0-3 was tending to zero. As there were two

generators in the model it was felt that removing one of them would not make

the model invalid. However, there was still the question of whether the

generator should be connected with the current flowing in the opposite direc-

tion. On attempting to optimise the element values with the generator

reversed, its value still tended to zero, therefore it too was removed from

the model. on applying the optimisation algorithm to this model, convergence

was obtained.
The new model is shown in Figure 5.3. It can be seen that the

optimised values of the remaining elements are quite similar to their original

values, the largest change of about 20% being in C At this stage the
2-3*

overall error function had been reduced from 384.2 to 114.9. The individual

errors are shown later in Table 5.2 and it can be seen that the maximum absol-

errors have been reduced from 0.55 dB to 0.43 dB in the modulus and from
ute

11.02* to 6.52* in the phase.

 63

I 1. ISE+02
b No. oF nodes =4
No. oF elements

g (2-3) = 3.94E-03 S
(T = 0)
(V across nodes 1)
C (1-3) = 1. OSE-01 nF
G (2-3) = i. 67E-OS S
C (2-3) = 3.90E-Oi nF
G (0-3) = 6.63E-03 S

Node I- Base
Node 2- Collector
Node 0- Su6strake and Emitter

Figure S. 3 First Model at a Local Minimum

The question of why a model containing fewer elements should be

better than one with more elements could be asked. The author's explana-

tion of this would be that firstly, a model containing more elements would

only be better than one with fewer elements if the extra elements were of

the right types and values and in the correct positions, and secondly,

the original element values in this example had been optimised and two of

the elements had optimum values of zero, while a third had an optimum value

of infinity.

5.4. Improvement of Models

Having found a model which gave a minimum value of the objective

functionp it then became necessary to consider how the model could be '

improved. It would not be expected that the removal of any elements from

a model giving a local minimum of the objective function would make an

improvement, therefore the addition of elements must be the only course of

 64

action. There are many strategies that could be employed for the addition

of elements, some of these will be considered now,

If we assume that we want the model to be as simple as possible then

we should avoid adding extra nodes, Therefore we should first of all con-

centrate on adding elements within the existing structure. Other restric-

tions might be that although we could attempt to replace any current genera-

tor that has been removed, we should not insert any additional generators,

and that we should not insert inductors other than at the terminals since

the most likely physical reason for the presence of inductance would be

lead inductance.

It might be that the engineer would like there to be present only

those elements for which some physical explanation for their presence could

be given. Certainly, this is a popular strategy and one which Nabawi and

NicolS72 refer to as "intuitive modelling". However, it is possible that

the engineer's preconceived concepts of a suitable topology for a model

could preclude the development of another acceptable and more accurate

model. This factor contributed to the failure by HegaZi7 to discover the

six node realisation to the set of equations by Lucal given in Chapter 3,

which was later found by SavageB.

For the first modelling exercise, the author used an intuitive

approach and experimented by attempting to add one element at a time to the

model at an arbitrary value based on the values of elements of the same type

already present. Each attempt involved restarting the optimisation

algorithm and checking the result. If the algorithm converged and the

overall error function was reduced, then the new element was left in the

model and another new element was tried. If the algorithm failed to con-

verge it was generally because failure mode 1 occurred. Usually, this was
caused by the corrections to the new element value. If this was the case

or if one of the other failure modes occurred it was assumed that the new

element would not make a suitable addition. There was one case, which can
 65

be seen in the example in the next section, where failure mode 1 was caused

by one of the existing elements in the model, In this case the new elem-

in and the causing the failure was removed and the

ent was left element

optimisation algorithm then terminated successfully,

The entire sequence of successful additions to the first model are

described in the next section.

S. S. Example of Results

After failing to replace singly each of the elements removed from the

initial model, it was decided to attempt to add elements of conductance and

between nodes 1 and 0 and nodes 2 and 0. These were all

capacitance

apart from the conductor between nodes 2 and 0 The order

successful

in which they were added is shown in Figure 5.4. It can be seen that the

values of the existing elements changed by only a relatively small amount

between each addition. The new elements of capacitance and conductance

-4
were inserted at values of lpF and 10 S respectively prior to optimisa-

tion. These values too, were reasonably close to the optimum values cal-

and the Gauss Newton algorithm took between 11 and 13 iterations

culated

to converge onto the local minima. The value. of the overall error func-

tion F is given at each stage and it can be seen that it was reduced from

114.9 to 56. SO by the addition of these three new elements. It should be

pointed out that the function being minimised at this stage included the

individual errors calculated against the extrapolated values for S at

12
0.5 MHz. The function values quoted do not include these individual errors

as the weightingswere later adjusted to exclude them from the objective

function.

Next, it was decided to attempt to insert inductors at each of the

external nodes. The two successful attempts are shown in Figure 5.5. The

first of these was originally inserted at a value of 4nH and was optimised
 66

2F=i. iSE+02
C. No. oF nodes
No. nF elements

(2-3) = 3.91E-03 5
(T = 2)
(V across nodes 13)
C (1-3) i. OSE-01 np
C (2-3) 4.67r--QS S
C (2-3) 2.90e-gi nF
G (0-3) G.G'jE--03 F

(S
(I Q

F
b No. nF nodes ai
No. oF elements =6

o (2-3) = 6.35S-03 5
(T = 01
(V a,:ro.; s nodes 13)
(1-3) =I. 99p-ol np
G (2-3) 4..jgE-GSS
C (2-3) nP
r1 (0-3) 4,91E-03 S
C (1-0) 2. ýSF-92 nF
f5

F-S. SGE+gl

No. nP nodes m4
No. oF etements z7

9 (2-3) =i . 36F-02 S
(U across nodes 1)
C (1-3) 21.47E Ol nP
G (2-3) = 4.29F--OS S
c (2-3) - 4.02E-04 nF
G (2-3) - 3.23E-03 S
C (1-0) a 3.20E-02 mp
G (1-9) = 2.1! E-94 S

F=S. GSE+0I

No. nF nodes =I
No. nF elements a8

(2-3) - &. SSE-03 S

(T = e)
(U across noues
C (1-3) 1.2SE-01 nr-
G (2-3) 4.36p-os 3
r. Q-3) 4.10C-94 nP
ri (0-2) 3.48F-03 S
3.03S-02 np
VE-91 S

, Q--O) = i. 9se-33 )F-

os
a

Made i- Base
liode 2- Collectm-
Node a- Substrate *n4 Emitter-

Figure 5.4 Addition of Elements Between Nodes 1 and 0 and Nodes 2 and 0
 b7

2 F=i. 95E+gl
C
No. oF modes =S
No. oF elements

9 (2-3) I. IGE-02 S
(T=
(V across nodes 13)
C (1-3) = 1.53F-01 np
G (2-3) = 1.3sr--OS S
C (2-3) = 4-19r--04 nF
G (4-3) 3.14E-.-03 S
C (1-4) 3.42E-02 nF
G (1-4) 2.27E-04 S
C (2-4) i-SOE-03 nF
L (4-0) 1.27E+01 nH
0

s
I 2 F :-2.2SE+01
b c No. oF nodes =6
No. aF elements

9 (2-3) = 8,79E-03 S
(T = 0)
(V across nodes sl)
C (S-3) 1.31E-01 nF
C (2-3) I. ISP-OS S
C (2-3) 4.04E-Oi nP
G (4-3) 3.33C-03 S
C (5-4) 2. G?E-02 nF
G (S-4) 1.2se-04 S
C (2-1) 2-IOE-03 nF
L (4-0) G. 4SE+Oj nH
L (5-1) 3. S2E+02 nH
0 So

Node I- Base tJote: The values given here

Collector ,
Node 2- were later Found to be in error
Node 0- Substrate and EmIttex due to a FaýAt in the analysis
routine in use at the Lime

Figure 5.5 Addition of Inductors

 68

to a value of 12.7 nH, The second was inserted at a value of 50 nR and

was optimised to 352 nH. In eachcase, the existing elements changed by

factors of up to about two, but when the second inductor was added the

'first inductor changed by a factor of over five, However, both of

these cases converged within 11 iterations of the Gauss Newton algorithm.

The overall error function had now been reduced to 22.45. At this point,

the error in dillambro's analysis routine, referred to earlier, was detected.

After correction of the error the second model in Figure 5.5 was rý-optimised

and in addition to changes in the element values, the value of the overall

error function for this model was reduced to 12.27.

Then several different additions were attempted including G

1-3
C (where, for example, G signifies an element of conductance
3-4 ,G 2-4 1-3

between nodes I and 3) but each of these failed. On attempting to add

C the failure mode 1 was due to the corrections to element C

5-4 '
1-4
C5-4, was therefore removed and on convergence the overall error function

had been reduced to 9.70. This model is shown in Figure S. 6 and it might

have been considered that sufficient improvements had been made. However,

one last attempt was made to replace the current generator that had been

removed from the initial model. This last attempt was successful although

the current was flowing in the opposite direction to that in the initial model.

The values of many of the elements also changed quite considerably as can be

seen from Figure 5.6. However, convergence of the Gauss Newton algorithm was

achieved without difficulty and the overall error function was reduced to 6.13.

The individual errors of this model are shown in Table 5.2 and most

of these are considerably smaller than the stated measurement errors.

 69

1 S
b No. oF nodes =6
No. oF elemenLs = 10

9 (2-3) - 7.9SE-03 S
(T = 0)
(V across nodes sP
C (S-3) = I. IBE--Gl np
0 (2-3) = i. 17F-05 S
C (2-3) = 4.20E-gi nF
G (4-3) = 3.48G-03 S
G (5-4) = 1.29E-04 S
C (2-4) = 2. SGE--03 nP
L (1-0) = 8. SSE+01 nH
L (S-1) = S. 99e+02 nH
C (1-4) = 2. SOE--92 nF

5
1 G. 13E+00

No. oF nodes =6
No. oF elements

9 (2-3) -- I n; --04 S
(T = 0)
(U across nodes '3)
C (S-3) = i. 9GE-03 mp
G (2-3) = 4.43E-OS S
C (2-3) = 8.19E-gi nF
G (4-3) = 2.? iE-03 S
G (5-1) I. ISE-0i S
C (2-4) 2. Gie-03 np
L (4-0) I. SSE+02 mH
L (S-1) 1.9"E+03 nH
C (1-i) 3.23e-02 rip
0 9 (3-4) 2,72ý-03 S
(T =V
(V across noces 13)
Node I- Base
Node 2- Collector
Node 0- Substrate and Emitter

Figure 5.6 Further Improvements

 70
sss 22
12 21

Frequency Modulus Phase Modulus Phase Modulus Phase Modulus Phase

Model MHz dR degrees dB degrees dB degrees dB degrees-

O's 10 -0,37 0ý00 0.00 0,27 0,17 0.03 -0.19

_Q,
2,0 -1.33 O, SS -0,17 0,34 -1,89 0.03 -0.26
-0,14
IFig, 5,0 0,20 -3.41 0,46 -S, 12 0.03 -0.41
S, 2 -0,02 -1.60
10ýO i O. OS -0,36 -9.66 0.46 -11.02 0.03 -0.71
-3,31

O. S 0.09 0,00 0.00 0.52 0.03 -0.19

_ -0.21 -0.30
2.0 ll 0.08 2,19 -0.19. -0.49 0.03 -0.27
'Fig. _O, -0.71
5.0 0.10 -0.05 1.97 0.07 -2.00 0.03 -0.42
5 3 -0.36
. 10.0 0.38 - 0.04 -1.48 0.43 -6.52 0.03 -0.74
-2.21

0.5 0.00 -0.21 0.00 0.00 -0.05 0.67 0.00 -0.14

!Fig. 2.0 0.10 1.03 0.01 0.13 0.00 -0.07
6.6 -0.04 -0.66
5.0 0.04 0.09 0.04 0.07 0.06 0.00 0.07
l(Final -0.13
10.0 0.00 O. OS 0.02 -0.19 . 0.03 -0.0 5 0. 00 0 24
Model: .

A measurement of S12 at 0.5 NSIz was not available

Table 5.2 Errors in the S Parameters of Models of a Lateral P-N-P Transistor

5.6. Description of the Final Modelling Algorithm

Having developed a model in a trial and error manner it was felt that

a more automatic algorithm could be developed. The Gauss Newton algorithm

it was felt, a very reliable indication of whether any particular model

gave,

was suitable or not, but it was extremely time consuming to do one batch com-

puter run for each possible stage in the development of a model. Unfortun-

ately, the program was too large and slow to permit on-line execution.

The author's modelling algorithm was required to continuously attempt

model improvements, rejecting those that were unsuitable and retaining the

successful ones. In addition it was felt that the attempted improvements

should cover as many possible placements as was practicable. As the computer

would have to be run in batch mode, user interaction would not be

program

possible. However, it was felt that each intermediate stage should be stored

so that it was possible for the user to choose a model from any particular

stage, and perhaps modify it, before manually restarting the modelling

process. This way, the model development would not be inhibited by the user's

prejudices, although the user could still exercise some control over the
 71

process. New criteria were introduced to indicate more quickly whether a

model would be suitable or not and techniques for the addition of elements

and nodes to a model were also developed. Some details of the finaI mod-

elling algorithm were published by the author73 in 1981.

The final modelling algorithm can be divided into three separate

stages. The first stage consists of taking the initial model and attemp-

ting to find a local minimum of the objective function, if necessary remov-

ing elements from the model until a minimum is obtained. The second stage

involves the addition of new elements within the existing nu! nherpf, nodes. since

we assume that we want as few nodes as possible. The third stage is for

the addition of a new node. Stages two and three can be repeated altern-

ately until the required accuracy or complexity is obtained. These three

stages will now be described in more detail.

5.6.1. Stage 1- Obtaining the First Local Minimum

In attempting to find the first local minimum, the Gauss Newton

algorithm is applied to the initial model. The criterion for successful

termination is identical to that described in Chapter 4.

The criteria for failure are more stringent than those described

in Chapter 4, in particular, the two most common reasons for failure are

detected sooner. Failure mode 1 is now deemed to be fatal if the rate of

increase in the maximum modulus correction has increased on 5 successive

iterations and the mode we could now call failure mode 2, is deemed fatal if

the number of iterations by the Gauss Newton algorithm exceeds 50 without

obtaining convergence.

Whichever failure mode occurs the next step is to examine the vector

of corrections 6 (which is in the logarithmic domain) to determine which

elements it might be best to remove from the model. Throughout all the

earlier experimentation it was found that removal of the elements having the
 72

maximum modulus corrections aided convergence. Thus the basic aim is to

remove from the model the element giving the maximum modulus correction and

to then restart the Gauss Newton algorithm, generally continuing from the

element values current at the time of the failure.

If the element being removed is a resistor and its corresponding

correction is positive (or similarly a conductor having a negative correc-

tion) then it is assumed that the element is tending to an open-circuit and

so the element is removed open-circuit. Conversely, the removal of a res-

istor having.. a- negative correction (or conductor having a positive corr-

ection) is achieved by a short-circuit. This latter case necessitates the

removal of a node and involves the removal of any other elements which are

connected in parallel with the offending element.

The situation with regard to the removal of elements of inductance

and capacitance is not so straightforward. It was decided that, as any

inductor would. normally occur only at one of the external nodes, elements

of this type should always be removed as a short-circuit. Again this

necessitates the deletion of a node although it is less likely that there

will be other elements in parallel with an inductor than with a resistor.

It was also decided that, in order to reduce the possibility of premature

collapse of a model, elements of capacitance should be removed open-circuit

provided that at least one other element is in parallel with it. If there

is no other element in parallel with a capacitor. that is being removed

then it is replaced by a resistor. The initial value of this resistor

was set quite arbitrarily at 1K 0. The setting of the initial values of

new elements is discussed in more detail in Section 5.6.2.

There are a number of cases when short-circuiting an element is

deemed to have failed. These are

if any of the external nodes are connected together

Cii) if any current generator is short-circuited

(iii) if the controlling voltage of any current generator is

short-circuited.
 73

Furthermore, it was decided that it would not be permitted for a

current generator to be removed automatically since the presence of current

generators is normally critical to a model, particularly as most models

contain only one current generator. It was felt that if there was a sit-

uation where removal of a current generator was the only remaining option,

then it should be done manually.

The method described above for the removal of elements is summar-

ised in the flow chart in Figure 5.7.

The overall scheme for Stage 1 is shown in Figure 5.8. Although

the basic scheme is to remove only the element causing the maximum modulus

correction and to then restart the Gauss Newton algorithm, it was found that

when there were a number of elements causing large modulus corrections, the

removal of several elements at one time could save on the number of failure/

removal cycles. The criterion devised for multiple element removal was
16ij j+2
that all those elements whose modulus correction > 10 , where j is

the average of the positive powers of 10 of all the corrections 6 (in the

logarithmic domain), should be removed. If none of the corrections

satisfy this criterion then only the element having the maximum modulus

correction is removed.

If, following the removal scheme already described, none of the

prescribed elements are successfully removed, then an attempt is made to

remove the element having the second highest modulus correction. If this

is successful then the remaining elements are given their values at the

time of entry to Stage 1 and the process restarted, otherwise Stage 1 is

abandoned. If this should happen then the best course of action is for

the user to first check the initial model for errors and to then try a
different model, perhaps based on the initial model but containing some

additional elements.
 74

Figure 5.7 Scheme for Removal of Elements

 75
STAGE I

lGauss Newton I

Yes
ocaL minimum SuccessPuL exit
i
No

J= Average power oF positive

power corrections
CLIM = 10.0*m((J+2)
CMAX = Max moduLus correction

> CLIM
Yes
N-o

Remove eLement Remove aLL eLements with

giving CMAX moduLus correction > CLIM

Y At Least one
cessFuL removaL

No

Remove onLy the eLement

having the second highest
moduLus correction

--
essPu Lýýo FaiLure exit
Yes

I Reset remaining eLements

to vaLues at entry
to STAGEII

Figure 5.8 Scheme for Stage 1

 76

5.6.2. Stage 2- Element Addition

Having obtained a model giving a local minimum of the objective

function, F the second stage may now be entered. The aim of this stage
min'
is to add extra elements into the model without increasing the number of

nodes. The author decided that only elements of capacitance and resistance

should be added at this stage. Elements of inductance are expected to occur

only at the external modes and are unlikely to be successfully inserted in

parallel with another element even at an external node. Therefore, it was

decided that inductors should only be added with a new node at Stage 3.

The author also felt that users were likely to have strong ideas about the

number of voltage controlled current sources that should be present in a

model, and their positions. This factor, coupled with the problems of

choosing suitable positions and controlling voltages of current sources to

be inserted, led to the decision that they should not be added automatically.

The user always has the ability to modify a model at any particular stage

and then restart the modelling process.

The problems associated with intuitive modelling have already been

discussed. The author felt that in Stage 2, all possible placements of

elements of capacitance and resistance should be attempted. Cutteridge74

had developed expressions for the optimum values for the placement of vir-

tual elements which were applicable to network synthesis using coefficient

matching techniques and which had been utilised by Krzeczkowski6.

Savage8 however, eventually rejected the use of this technique in favour of

one based on the Gauss Newton corrections of virtual elements. The author

independently developed a similar scheme for the addition of elements to a

,
model. This scheme is shoum in Figure 5.9.

In this scheme, taking each type of new element in turn, attempts

are systeriatically made to insert the new element between each pair of

nodes where an element of that type does not already exist. If the addition
Figure 5.9 Scheme for Stage 2
 78

is successful, the criteria for which will be given later, then the element

remains in place, otherwise the model retains its former topology. These

placements are attempted repeatedly until neither a capacitor nor a res-

istor has been added in one complete cycle. The reason for these repea--

ted attempts. is that it was found that even if a particular element has

failed to be added at one stage, it might be successful at a later stage

when the model topology has changed slightly. Indeed, in Section 5.7 -

there is an example where an element was deleted from a model at one stage

and then re-inserted later.

An alternative strategy might have been to evaluate each of the

possible placements and to only add the best of these, followed by further

cycles of evaluation and addition of the best until no more elements could

be added. However, the author felt that this latter option would be

more laborious and that the same final topology would be obtained in most

cases.
Obviously, with so many possible placements being attempted, the

efficiency of the algorithm to detect the success or failure of each

addition is paramount. It was observed that, when a new element was

inserted into an existing model, the moduli of the Gauss Newton corrections

to be applied to the existing elements were usually much smaller than the

modulus of the correction to be applied to the new element. This fact

was capatilized upon by the scheme devised for element addition summarised
in Figure 5.10.

Thus, when attempting to add a new element, first the Gauss Newton

corrections are calculated. If the new element does not give rise to

the maximum modulus correction, then the Gauss Newton algorithm proceeds

normally. However, if the maximum modulus correction does apply to the

new element, then this correction, up to a maximum absolute value of ten

in the logarithmic domain, is applied directly to the new element. The
 79
ELEMENT ADDITION

lCaLcuLate Gauss Newton corrections I

L42
Does new eeLement have
0-
max
m correction
xmmod col
No
No
Yy
ess
CCC Correction
Orr
Ensure
sm mod C4 10
AppLyjC C to new eLement

RecaLcuLate
Gauss Newton corrections

Does new eLement have

a--- max mod correction
No again
Yes
Cthse
Is'0 the new correction
1
9- same sign & Larger
No than bePore

YY
ee
ss

s it aR
ILaR
tending 3 to zero
No
Yes
Yes

Remove
appropriate eLement

Continue
Gauss Newton
No

iLed du
ýFaiLed duee to
tc0 2 eLements
: onverged .,:,. ->-< aLready removed
-wL new eLement
eLem
ment
No N
No and F> FMIN es
Yes Yes
Y s

Is iaR
it aL
tending to zerro
z
Yes
No
< O. SGxFMIN FaiLure
exit
I-I
ýSuccessPuL
exit

Figure 5.10 Scheme for Attempted Element Additions Stage

at 2
 80

Gauss Newton corrections are then re-calculated. If the new element

is no longer the one generating the maximum modulus correction then the
0
Gauss Newton algorithm is continued normally. Similarly, if the mod-

ulus correction to the new element is smaller than before, or if the

correction is of the opposite sign, then again the Gauss Newton algor-

ithm is allowed to proceed. If, however, the new correction is of the

same sign and its absolute value is greater than before, then it is

assumed that the new element is tending to a value of either zero or

infinity. If the new element is a resistor and it is tending to zero,

then the removal of the node that this suggests is permitted and the Gauss

Newton algorithm is then applied to the reduced model. Otherwise, it is

assumed that the new element is unsuitable for addition into the model.

By applying the Gauss Newton correction to the new element in this way,

the author is attempting both to detect failure of element addition more

quickly, and to aid more rapid convergence of the Gauss Newton algorithm

by ensuring that the starting value of the new element is reasonable.

It has been found that the Gauss Newton algorithm is very reliable

even when the starting values of new elements are quite inaccurate. Thus,

with this additional aid to rapid convergence, the values at which new

elements of capacitance and resistance are added, were chosen quite arbi-

trarily. It was felt that the new elements should be of such values that

they would not have too great an effect on the existing model. Therefore,

the initial value chosen for new elements of capacitance was lpF and of

resistance was 10 Alternative choices could have been based on

.
the actual values of each type of element already present in the model at

the time of the addition, but it was felt that this would not offer any

improvement.

Those cases where the Gauss Newton algorithm is allowed to proceed

now continue in a similar manner to Stage 1 except that in certain cases

 81

the element addition is deemed to have failed before convergence of the

Gauss Newton algorithm onto a local minimum has been obtained. These

cases are;

(i) if failure of the Gauss Newton algorithm is due to the

new element and it is not a resistor tending to short

circuit a pair- of noJe3

Cii) if further element removals are indicated when two or more

elements have already been removed and the current value

of the objective function F is still greater than the

value for the currently accepted model, F

min*

If the Gauss Newton algorithm converges at a new local minimum, the new

model is only accepted if the value of the objective function is 4% lower

than Fmin' the value for the currently accepted model. This is to ensure

that the model contains only those elements that make a significant contri-

bution to the model accuracy and to avoid adding an excessive number of

elements to the model.

Thus, Stage 2 ends when no further successful additions are possible.

If the model is still not accurate enough then the next step is to attempt

to add elements in such a way that a new node is generated.

S. 6.3. Stage 3- Node Addition

Savage8 in his work on network synthesis discussed a number of

methods by which nodes could be introduced into an existing network.

Among these methods are to:

(i) split an existing element to form a new node to which

attempted element additions should be made

 82

Cii)_ duplicate an existing T-network

perform a delta-wye transformation

(iv) -substitute a T-network in place of a negative virtual

element.

Savage found that the most effective technique was the last of these, where

if a potential element addition indicated that the element should take a

negative value then a T-network, based on that element type and its corres-

ponding proposed negative value, should be inserted in its place. However,

the author felt that each of these methods which, apart from (iii), caused

the generation of at least two extra elements, ignored the most likely

method for the improvement of a model.

If we assume that the current model is basically correct then,

rather than disturb the internal structure of the model, new nodes should be

developed at the external nodes. This way, the addition of a new node

can be achieved by the addition of only one new element as in the examples

in Figure S. S. If internal nodes are to be added then a method which

again involves the addition of only one new element, is to insert a new

element, of a different type, in series with one of the existing elements.

Thus, the author's method of node addition, which is shown in Figure

5.11, is based on the addition of only one element. Since we assume that

the number of nodes in a model should be kept to a minimum, this involves

the evaluation of each possible placement in order to select the best new

node and element. The reSt-Lictions on the placements of new elements can

be seen in Figure 5.11 and include the following.

1. Expansion at an external node is only permitted if there is not an

inductor alieady at that external node. Then the addition of

elements of inductance, resistance or capacitance is permitted.

 83

STAGE 3

FOR I=192 and

Inductor at externaL
node I Yes

No

Attempt addition oP each oP

L, R and C
between node I and new externaL node

NEXT I

FOR I=19 no. oF eLements

JELement I oP type I

R 6 LI IC g

Attempt to add C
at the
Attempt to add L non-externaL node
iP at externaL node
Attempt
to add R
Attempt to add R
lAttempt to add CI at the
non-externaL node

NEXT I

y successFuL additions FaiLure exit

No
Yes

SeLect best addition

ýSuccessFuL exit

Figure 5.11 Scheme for Stage 3

 84 1

2. An element of resistance or conductance may have an inductor

added in series with it only if by so doing the :Lrductor will be

connected to an external node which does not already have an induc-

tor connected.

3. Elements of resistance or conductance may have a capacitor added

in series.

4. Elements of inductance may have an element of resistance or capaci-

tance added in series at the non-external node.

S. An element of capacitance may have a resistor added in series.

6. A voltage controlled current source may not have any element added

in series.

Further clarification of the way in which some of these elements are

added is given below.

1. The addition of a particular element is not attempted if there is

already an element of that type in series in that branch of the

model.

2. The new elements are only added on one side of any particular ele-

ment, usually on the side having the highest node number.

3. When an element is connected at a node at which a controlling voltage

of a current generator is taken, then the controlling voltage is

taken across the new element also.

Thus, although most topologies are possible with repeated entries to

Stages 2 and 3, some could take an inordinate time to develop.

The method used to'determine whether a node addition is successful

is shown in Figure 5.12. As with the element addition at Stage 2, it is

 85

Attempt to add
eLement B in series
with eLement A

there aLready an eLement oF

type B in series with
eLement A> Yes
t
No
F-Inuert
eLement and
ca LC Late Gauss Newton
corrections

ýs the moduLus correction

to either FaiLure
eLement A or B
> 100 Yes exit
'it

No

Continue Gauss Newton

aLgorithm

FNýo
onverged
Yes

Have any eLements

Lready been removed Z>
r e0s
t No --Lý
43 -< < FMI ý,
No
Was PaiLure due to
Yes
eLements A or 8 Yes
No

Remove appropriate IF
eLement

CFaiLure 7e) (Su

ccessPuL
exit exit

Figure 5.12 Scheme for Attempted Node Addition at Stage 3

 86

based on the Gauss Newton algorithm. The initial values for new elements

of resistance and capacitance were set at 100 0 and 0.1 pF respectively,

one order of magnitude smaller than at Stage 2. The initial value for

elements of inductance was set at 10 nH .

It was found more difficult to reduce the computational effort at

this stage and the only early prediction of failure is if the absolute

correction (in the logarithmic domain) to either of the elements connected

to the new node, calculated by Gauss Newton, exceeds 1000. Otherwise, the

Gauss Newton algorithm is entered normally. If the algorithm fails to

converge then, provided the failure to converge was not due to either of the

elements connected to the new node, the appropriate element is removed and

the Gauss Newton algorithm restarted. If it fails to converge again then

the addition is deemed to have failed. If the Gauss Newton algorithm

finds a new local minimum that is lower than F the best minimum obtained
min ,
before entry to Stage 3, then the addition is said to be successful.

Each successful node addition is saved for future reference by the

user and the best of these is then used as the new model. Although the

user may decide at which stage the model should be accepted, the author's

recommendation is that it should be after completion of Stage 2, in order

that the maximum benefit be gained from any particular number of nodes.

It might be suggested that to save time, rather than re-entering Stage 2,

further element additions should be attempted only in the vicinity of the

new node. However the author has found that the addition of a new node can
"
affect the whole model and that, therefore, it is better if all permitted

element additions are attempted using Stage 2 again.

 87

5.7 Checking the Final Modelling Algorithm

As a final check on the modelling algorithm described in the previous

section, after the models described in Chapters 6 and 7 had been developed,

the first modelling exercise, to model a lateral p-n-p transistor, was re-

peated but this time using the final version of the modelling algorithm

instead of the trial and error method used to produce the models in Figures

5.3 to S. 6. In addition, the s parameters of the models were calculated

by the analysis routine using pivotal condensation written by the author, in

place of di Mambro's routine.

From the initial model shown in Figure 5.2, the algorithm proceeded

to remove elements G and C3_,, and then stopped with an indication that
1-3
gO-3 should also be removed. On removing this element manually and re-

starting the algorithm, it then went on to obtain the model shown in Figure

5.3. Continuing from there into Stage 2, the algorithm added the elements

Cl-01 C2-0 and R1_0. This model was thus the same as the last model in

Figure 5.4 even though the element addition was in a different order to that

given in Figure 5.4. Owing to the differences in the analysis routines,

the element values and the overall error function value were slightly differ-

ent and the new value of the overall error function, F, was 50.97 compared

with 56.50. The algorithm then entered Stage 3 and the new element and
,
node added was L 4-0* Thus the algorithm had produced the same model as the

first model shown in Figure S. S. The values of the elements were, here,

quite different from those given in Figure S. S due to the errors in the

analysis routine and the value of F, which had not been recalculated in the

earlier trials, was found to be 13.50.

From this point, the model developed differently from the

v%oaalln
in which r-&3&. the next step was to add an inductor at Wode'l
,
thus immediately generating an additional node. Although at the time
it was thought that this made a significant improvement, it can now be seen
 88

P=3.84E+02

No. oF nodes =4
No. oF etements =8

9 (2-3) = -; IE-03 3
0),
,T=
(V across nades 13)
(0-3) = 2.07E--95 S
(T = 9)
(V across nodes ',)
G (1-3) = S. 12E-ig S
C (1-3) - 9. GSE-02nF
G '2-3) = S. 22E-OSS
S C (2-2) 2 S. OGE-Oinp
(I
e G (9-3) = 6.77E-03 S
C (0-3) = 3.71E-06 np
Delete G( 1-3) C(2-9) 9(0-3)
Add C(1-0) C(2-9) R(1-0)
Add new node with L(4-0)

F=1.3SE+91
No. oF nodes =S
No. oF elements =9

C( 1-3) =I . 2GE--9l nF
6 (2-3) = i. 64E-03 S
C (2-3) x 3.62E-01 np
9 (2-3) = 9.23E-03 S
(T = 0)
(V across nodes 30
G (4-3) = 3.27E-03 S
C (1-4) w 2.97E-02 nF
C (2-4) !. 39E--9i nF
R (1-4) 7.76E+93 ohm
L (0-4) i. 28E+02 nH

Add C(I-2) i Delete C(2-3)

Add C(2-0)
Add new node with C(Q-3)

F=7.3GE-+Q9

No. oF nodes -6
Mo. oF elements = 11
C (1-3) I. ISE-91 nF
6 (2-S) 8.28E-96 S
9 (2-3) 9.14E-03 S
(T = 9)
(V across nodes 1)
G (4-3) = 3.42E-03 S
C (1-4) = 2.8GE-02 nF
C (2-1) = I. G?E-94 nF
R (1-1) = 9.29E+03 ohm
L (0-4) = i. iSE+02 nH
C (1-2) = 3.94E--13i np
C (2-0) = 2.21F-93 nF
C (S-3) = 3.20G-03 nF
e

Add C(2-3) i Delete C(I-2)

Add R(4-5)

Node i- ease
Node 2- Collector
Made 9-S. i:,atrQte, "a Emmet

Figure 5.13 Modelling the Lateral P-N-P Transistor

 89

that the change in the value of F was actually from 13.50 to 12.27.

In the final modelling algorithm, after one successful node addition, Stage 2

is re-entered. In this section the algorithm added two new elements C

1-2
and C deleting C while adding C to obtain a value of F= 12.26.
2-0 , 2-3 1-2 ,
On re-entry to Stage 3, the addition of an inductor at the input node only

gave a very small reduction in the value of F and the new node selected by

the algorithm was formed by the addition of an element of capacitance in

series with G2-3 , and gave a value of F=7.36. This sequence is shown in

Figure 5.13 together with the additions made in the final entry to Stage 2.

These additions were to add C while deleting C followed by the addi-

2-3 1-2 ,
tion of R4-5* It is interesting that elements C and C appear to be
1-2 2-3
mutually exclusive. However, the addition of C at this late stage after
2-3
its earlier deletion serves to confirm the point made in Section 5.6.2 that

attempted element additions should be repeated at each stage of a model's

development.

V, I S.??E+00
No. oF nodes =6
No. oF elements = 12
C (1-3) = 1.19e-01 nF
G (2-S) = 4.63E-05 S
9 (2-3) = B-47E-03 S
(T = 0)
(U across nodes 13)
G (4-3) 2.98E--03 S
C (1-4) 2.99E-02 nF
C (2-4) 1.18E-04 nF
R (1-4) B. iOE+03 ohm
L (0-1) 4. IOE+02 nH
C (2-0) 2. iGF:-03 nF
C (S-3) 2. S7E-01 nF
'e C (2-3) 3. S3C--Oi nF
R (4-S) 2.44E+03 ohm

Node I- Base
Node 2- Coliector
Node 0- Substrate and Emitter

Figure 5.14 Final Model of the Lateral P-N-P Transistor

 90

This'final model, shown in Figure 5.14, had a value of F=5.77,

giving r. m. s. errors of 0.04 dB and ý0.4" and consisted of 6 nodes and 12

elements which included only one voltage controlled current generator.

The values of the s parameters. and the errors are given in Table 5.3.

Modulus (dB) Phase (degrees)

Frequency j Eodel
MHz Measure dI Model Error Measured Error

0.5 -0.10 -0.11 0.01 -1.40 -1.14 -0.26

2.0 -0.20 -0.16 -0.04 -5.40 -4.52 -0.88
S 0.03
11 5.0 -0.40 -0.43 -11.20 -10.89 -0.31
10.0 -1.20 -1.20 0.00 -19.40 -19.75 0.35

0.5 -76.89 0.00 99.54 0.00

2.0 -64.70 -64.77 0.07 88.70 89.10 -0.40
S
12 5.0 -57.20 -57.12 -0.08 84.00 82.76 1.24
10.0 -52.00 -52.01 0.01 76.30 76.85 -0.55

0.5 -12.50 -12.44 -0.06 178.10 178.28 -0.18

2.0 -12.50 -12.50 -0.00 169.90 169.81 0.09
S
21 5.0 -12.80 -12.88 0.08 155.00 154.72 0.28
10.0 -14.00 -13.97 -0.03 132.90 133.11 -0.21

0.5 0.00 -0.01 0.01 -0.20 -0.02 -0.18

2.0 0.00 -0.01 0.01 -0.30 -0.19 -0.11
S
22 5.0 0.00 0.01 -0.49 -0.01
-0.01 -0.50
10.0 0.00 -0.01 0.01 -0.90 -0.97 0.07

A measurement Of S 12 -at 0-5 MHz was not available

Table 5.3 s Parameters of Lateral P-N-P Transistor and Final Model

Although the value of F is smaller for this model than for the

final intuitive model in Figure 5.6, these values being 5.77 and 6.13 res-

pectively, the maximum absolute errors for this model are 0.08 dB and 1.24*

compared with 0.13 dB and 1.03*. Thus the errors are again comparable with

the errors in the measurements.

In the author's opinion both of these models are equally valid and she

would suggest that the original model proposed by PRL contained two voltage

controlled current*generators in an attempt to model the substrate interaction

of a lateral p-n-p transistor in the manner suggested by Callahan75.

However, Callahan's model was for the lateral p-n-p transistor operating in
 91 1

the saturated region and therefore, probably not applicable to the set of

data provided by PRL. This could explain why the second current generator

in the final intuitive model is reversed. By adding a supplementary

current generator to a model needing only one, a model of much greater com-

plexity, and-potential effectiveness, is produced. The author would suggest

that although the model in Figure 5.14 has one element more than that in

Figure 5.6, it is actually the more simple of the two models.

The time taken to produce this final model was approximately 50 min-

utes of c. p. u. time on the Leicester University CDC Cyber 73. However, the

point should be made that the author's intention was to produce an effective

method for the development of models and that the computing time was not con-

sidered to be of prime importance. In Chapter 8, some suggestions are made

for means of reducing the computing time, but with the present trend of fas-

ter computers constantly being developed, in the author's opinion the main

priority should be the development of an algorithm that produces results ,

regardless of the time taken. There is little point in developing a fast

algorithm if it neglects to produce a good madel.

 92

CHAPTER6

MODELLINGA VERTICAL N-P, N TRANSISTOR

The second set of data provided by PRL was for a vertical n-p-n

transistor. Like the lateral p-n-p transistor this is a component of

an integrated circuit and the term 'vertical' refers to the way in which

the transistor is produced in the integrated circuit. The data

provided is reproduced in Appendix 1, together with an example of the data's

original form.

It can be seen that the data consists of s parameter measure-

ments at 12 frequencies between 0.1 GHz and 1.0 GHz for each of the common

emitter, common base and common collector configurations. The original

modulus measurements were simply the magnitude and were converted to

decibels on input to the computer program. Similarly, the measurements

of the phase angles in degrees were discontinuous in an erratic manner,

for example, the original phase. readings for s1l in the common base con-

figuration (included in Appendix 1) between the frequencies of 0.6 GHz and

0.9 GHz are 355.3*, -11.70, -17.6* and 336.1* on consecutive readings.

Therefore, although the data was read in to the computer program exactly

as it was given, the phase angles were immediately converted to values

between -1800 and t1800. As the calculated phase angles of the models

were produced in the same range, this made the calculation of the errors

easier although the algorithm still had to check for cases where, for

example, the measured phase angle was -1780 and the calculated phase angle

was +176*. In this case the correct absolute error is 6* and not 3540.,

thus the maximum absolute error possible in a phase angle is 1800. In

the graphs that are shown later, the phase angles are made continuous over

the frequency range, therefore the absolute values of the phase angles

exceed 1300 in some cases.

PRL also provided a general model which they had developed. This
 93

E IF E IESa
=. i-_ =. I= C== U- U- LL U- U- ILL LL U- LL LL U- LL
CD oooooccccccccccc(n--%. -%CCC
cu %nw
M V- In CV)It* 0.1 In
Cki m V- CD CD CD CD m CU Ln LO CD MMU)
CD cD CD CD CD CD CD CD CD CD CD CD CD CD CD C CS) CD CD CD tA V
It
cr) ....... 111i! III! I aj !I1 4)
1
m 11.1LU Uj
(1) - Cl)
LU LLILU
M C11 CS)
LLJ
CD LLI
lq-
LU
M
Lli
LID CD
LU LULULU Uj LU ULJCU a
V- CD CD Cl) CD CD CD 0
LL;LLJU.;
CU CD M
- V) V- Cr) CD Cý, - Cý, CD CS) M Cl) LO CD CýJ LO CD IC CD CD CD
................ 4J
a$ S- CU (7) V- CD V- lqr - -, cfý 10; -4 .4 V) - V) MM V1 - V) V) Cd
"o Qj 10 Ln ýo
C) It It it It It H It If it 11 If It It If It It It 0 11 It It c
ý C
C: 0) s- i
0-% -% 0-% 0-% e% 01% ý% 0-ý ý% 0% -% -% 0-% ý% .%-. % e% IQ 0% -% -%

LL Q- U) LO CD CO CD 'q- CU CD CD M LO r,, M Cv) CD CD LD ;10 CD w cy)

00111tI! IIIIIItIIIIItI X
,q- V) Cw)CD 0. CC)IV V) In Cw)M ',r Cl) Cw7 CU CU
1-0
C;
uuu

Ic
Ir

cle) N

0 :2
S. )
-
qj uM gi
-i

ci E0
cm Ui Li cm

CM m CD

r,

-4

-0 14ý

s-I
 94

model is shown in Figure 6.1. In this diagram, node 1 is the base,

node 2 is the emitter, node 3 is the collector and node 0 is ground.

Thus, to obtain any particular configuration, the appropriate node has

to be connected to node 0. This model was primarily a low frequency

model and its s parameters give good agreement with the measurements

at frequencies in the range 0.1 GHz to 0.3 GHz. However, at higher

frequencies the errors become very large as can be seen in the graphs

shown later. In each of these graphs, the black graph refers to this

original model and is used for comparison purposes.

One interesting point about this model is that, although PRL

stated a strong preference for the presence of physically realistic ele-

ments only in models produced for them, this model contains a negative-

valued capacitor C3-6.

This modelling problem was approached by attempting to model each

configuration separately and to produc. e a single general model. The aim

in each case was to improve the frequency range of the model without too

great an increase in the numbers of nodes and elements in the model.

Each of the parameters at every frequency was weighted equally, with a ldB

error in the modulus having the same weighting as a 10* error in the phase.

The first individual models attempted were for the common emitter

and common collector configurations. In each case the initial model

was taken as the PRL general model with the appropriate terminal connected

to ground. Of these two models, the common collector model produced a

good result very quickly and with relatively few changes in the topology

of the model. In the common emitter case, one node and several elements

were deleted in Stage l.. The author therefore decided to try using

the final common collector model, connected appropriately, as another

initial model for the common emitter case. This model developed more

rapidly to give another satisfactory result, which was included in the

author's paper73 describing the modelling algorithm. Following this,

 95

the common base case was also attempted using the final common collector

model as the basis for the initial model.

In attempting to develop a single general model for all three

configurations, the major problem was one of size. In spite of this,

two models were produced. The first of these used the PRL model as the

initial model and this attempt was terminated when the model had reached

such a level of complexity that firstly, it was considered that a more

complex model would be. inappropriate, and secondly, problems of computing

time were being caused by the amount of calculation required. The

second general model that was produced used as the initial model, a com-

bination of those elements of the first individual models that were common

to two or more of those models. The model developed from this contained

one node. more than the first model, but had fewer elements, and its over-

all error function value and maximum individual errors were smaller.

Full details of the development of these models together with

graphs showing the accuracy of the results are given in the following sec-

tions of this chapter. In the graphs, in order to aid comparison of the

errors, the vertical scales for each graph of modulus and of phase cover

the same absolute ranges and the ranges of the modulus and phase are in

the same ratio as the weighting factors applied to these.

Each of the models developed still bears some resemblance to the

PRL model. The major feature of each of the new models is that there

are more interconnections between the internal nodes. This would

probably be expected since the PRL model was based on the primary theore-

tical aspects of the device being modelled. The new models improve on

this approximation by including elements that model less significant

characteristics of the device.

 96

Comnon Collector CoiAfiguration

derived
The initial model for the common collector configuration

from the PRL general model in Figure 6.1 is the first model shown in

Figure 6.2. Since the author's algorithm constrained the element

to take positive values only, the negative-valued capacitor, C4-7'

values

was given a value of 0.1 pF instead of -0.8 pF. This change in value

actually reduced the overall error function, F, from 5.84E + 04 to

5.54E + 04. The r. m. s. errors given by the value of F=5.54E + 04 were

2.4 dB and 240. The maximumabsolute errors were 3.6 dB and 85*, these

occurring at the highest frequency.

After Stage 1, the model had been reduced from 7 nodes and 15

elements to 7 nodes and 12 elements. The new value of F was 1.12E + 03,

already a significant reduction. This is the second model shown in

Figure 6.2.

In Stage 2, there were several additions and deletions of elements

and the details of these are given in Figure 6.2. The model obtained in

Stage 2 was sufficiently accurate that Stage 3 was not entered. The

final model thus obtained, the last model shown in Figure 6.2, consisted of

7 nodes and 15 elements and the value of F had been reduced in Stage 2 by

a further factor of ten to 1.61E + 02. The r. m. s. errors of this final

model were 0.13dB and 1.3" and the maximum absolute errors were 0.32 dB

and 1.74*. This model is extremely accurate and it could be that many

people would find the simpler model obtained at the end of Stage 1 quite

acceptable.

The graphs in Figures 6.3a and 6.3b show the s parameters of the

final model in blue. As has already been mentioned, the scales of the

Y axes of all the graphs in this chapter have been standardised to aid

comparison of the errors. Thus the modulus scales cover a total range of
30 dH and the phase scales cover a total range of 3000. The black points
 97

r- a S.SiE+94
No. (nF nodes w7
No. oF elements = IS
ý R (3-6) a 1,12E+92 ahm
R (4-6) 2 Z. SDE+93 ohm
R (4-9) = 9.43E+04 ohm
R (4-5) - 4.99E+99 ohm
L (1-3) = 4.79E+Iag nH
L (S-2) = 4. igE-+OgM
C (1-9) S. OOE-OSnF
C U-9) I. GOE-04 np
C (3-4) i. 74E-93 nP
C (4-6) 1,99F-02 nF
9C- C (6-9) 9.20E-OS nF
C (4-0) IME-95 nF
C (2-9) S. OGE-93 Mp
C (1-2) S. OOE-OSnF
9 (0-4) 3.90E-02 S
(T u-9.29E-93 ns
Delete R(4-G) i R(4-9) (V across nodes
Replace C(1-0) bg R(1-9)
Delete C(2-9)

r- a 1.12E+03

No. aF nodes -7
No. oF elements = 12

R (3-6) - 9.92E491 ohm

R (4-S) w 2. L2E+gl ohm
L (1-3) x G. S2StG9 nH
L (S-2) a 3.99E+09 nH
R U-9) x 3.29E+93 ohm
C8 9) x 9.49F-91 nF
C (3: 1) 2 I, GSE-93 nF
C (i-G) -t 2. OOE-92 nF
C ý6-9) i. 02E-ei nF
C ý4-9) i. GGSA3 rr-
0 C (1-2) 6.2iE-gi nF
C. 9 (9-1) S. S6E-92 S
Ua3. SSE--02ns )
(V across nodes 'I)
Add C( 1-4)
Replace C(I-2) by R(I-2)
Add C( 1-6)
Add C(2-3) Delete R(I-2)
Add C(2-9)
Add R(I-G) Delete C(I-S) i C(2-3)
Add R(3-0) Delete R(1-0)
Add C(S-6)
Add C(S-0) Delete C(4-9)

F=I. GIE+02
No. oF nodes =7
No. oF elements = IS
R (3-G) - 2. QIE+02ohm
e1 R ýi-S) 2 2.22E+91ohm
L ý1-3) a 1.986+91nH
L (S-V = 7.70E+99nH
C 0-9) = 1.41E-03np
C (3-4) = 2.12E-gi nF
C (i-G) = 1.9iE-92 nF
C (6-9) 1.8CE-04np
9 (0-1) 5.71E-92 S
(T = ?. DOE-92ns
0C. (V across nodes
C (1-4) = 2. S6E-03nF
C (2-0) - 8.25E-04 nF
R (1-6) = 1.41E+92ohm
R (3-9) = 1.48r:+24 ohm
C (S-G) = 9.9SE-04nF
C (5-9) a 1.93E-94nF
tiode I- F-miLter
Mode 2- Base
Node 0- C0114ct*r

Figure 6.2 Modelling the Common Collector Configuration

 98

MEASUREMENTS
PRL MODEL

,EWML

MODOB 511 PHASE DEG 811

5 50

0 FREQ GHZ 0 REG (GHZ

)0
--50

-100

-150

-20 -200

moo DB 521 PHASE DEG 521

10 100

5 50

0 FREQ GHZ FREQ GHZ

)0 )0
-5 -50 -

-10 -100

-15 -150 -

Figure 6.3a s and S in Common Collector Configuration

11 21
 () ()

MEASUREMENTS
PRL MODEL

MODOB 512 PHASEDFG 512

iso

00
-REG GHZ
0

REQ GHZ
3

-50

-1310
-20

MODDB 522 PHASE:OF-G 522

150
5
100
0 FREQ GHZ
)0
150

0 EQ GHZ

-50 -
-15
-100 ý
-20

Figure 6.3b s and S. in Common Collector Configuration

12 21
 100

in Figures 6.3a and 6.3b are the calculated s parameters for tile PRL

general model connected in common coi.'lector configuration (With C4-7

negative). The red points are the measurements and thus show the values

aimed for. The graphs show quite clearly the improvement obtained over

the entire frequency range and especially at higher frequencies. In many

cases the red and blue points are indistinguishable.

The computing time taken to generate this model was approximately

600 c. p. u. seconds on the UMRCCCDC 7600.

6.2. Common Emitter Configuration

Two attempts were made to model the vertical n-p-n transistor in

the commonemitter configuration. The first of these used an initial

model derived from the PRL general model and the second used an initial

model based on the final model produced by the author for the common

collector configuration.

First Attempt

The initial model derived from the PRL general model is the first

model shown in Figure 6.4. As in the common collector case, the negative-

valued capacitor was given a value of 0.1 pF instead of pF. This

-0.8

caused a reduction in the value of F from 8.60E + 04 to 7.72E + 04.

The r. m. s. errors given by this model were 3.0 dB and 30.1-* while the maximum

absolute errors were 10.3 dB and 134.40. As will be seen later, the

largest errors occurred at the higher frequencies, those on S being

12
the most obvious.

After Stage 1, the model had been reduced to 6 nodes and 10 ele-

ments giving a value of F=1.45E + 04. The model giving this first

local minimum is the second model in Figure 6.4. The maximum errors in
 101

40.
,)F nodes
4c. iF elements i7
12 R (3-1) i. i 7E-+02 atim
R (ý-S) 2. skýK YAm
(2-5) 9.43E, 04 ohm
(S-ý) 1.39F-92 ohm
(6-9) 6.2CE-90 &IM
ýi-3) 4.; 2F-09 nH
7-9) = !.. OEý02 n.4
*
C ý3-5) i.; 4E-03 np
f (4-5) inF.
C (4-2) P. 20E--Ds nF
C U-S) i. O@E-OSnF
C Q-6) i. 24E-03 np
C ', 3--2) 1. ý30&-94 np
(, -2) S. 29E-OS mF
(1-0) i. 23F-03 np
0 C (2-0) S. SOE-94 nF
e 9 (2-5) 3. NE-02 3
(T =-L. 39F-02 ns
DejeLe '-ý1-3) (deteLing node) i Combine M-2) wiLh C(3--2) (U across nodes
Oelet. e R(2-5) R(S-2) C(2-S) , C(I-9) y r-(2-0)
? i

P. j, 4SE+qj

No. iF nodes
No. oF elements

q( i-3) = MOF. *Ol Ohm

b1 2c R (3-4) - 3.46E+02 ohm
R (5-2) - 2,93G+gi oh-ft
L ý4-0) = 7. ilE+99 nq
C (1-4) = 2XF-93 nF
C ý3-4) = I. IOE-02 nF
5 C 13-2) = i. 31F-gi nF
C (2-S) i. 90E-93 np
C ýi-2) i. 09F.-94 nF
'2-4)
9 ýT 2. GIE-02 S
=-1.26E-91 ns )
(V across nodes 31)

e
Mode i- Base
Mode 2- CalleCLDI
Node 0- Emittf-r

6.4 Modelling the Common Emitter Configuration (First Attempt)

Figure

A,
I2 HA [jrL

, W
ýýE MnDý-L
2COO

1130
2. ri

Cýr

-Z ')
- ;
i1Z
'j
4,
- rr
" ýIlz
10
IC'..

6.5 Gra-phs S12 forModel in Figure 6.4

Figure of
 102

the models for this configuration were on sj, and the graphs in Figure

6.5 show S12 for the second model in Figure 6.4 in blue, together with

the measured values in red and the values for the PRL general model

connected in the common emitter configuration (with the negative-valued

capacitor) in black. It can be seen that while the PRL model gives

good agreement with the measurements at low frequencies the model is very

poor at high frequencies, but the reduced model gives similar errors over

the entire frequency range.

Although it should be feasible to build up a model from this

reduced model, this attempt to produce a model for the commonemitter

configuration was abandoned. At the same time as this first attempt was
being made, the second attempt, using an initial model based on the final

model developed for the commoncollector case, was also being made.

Thus, it was already known that the second attempt had produced, after

Stage 1, a model with more nodes and elements which gave a considerably

lower value of F- It was therefore decided that the second attempt

showed more promise and should be continued in preference to the first

attempt.

6.2.2. Second Atteýt

The initial model used in the second attempt is shown in Figure

6.6 and it can be seen that this model is identical to the final model

produced for the common collector configuration but with nodes 2 and 0

interchanged. This model, in particular, lacked the substrate elements,

C and R7-0 in Figure 6.1, that were present in the PRL general model,
3-7

and gave a value of F=4.95E + 05 with correspondingly high absolute

values of the individual errors.

However, in Stage 1 only one element was deleted together with one

capacitor being replaced by a resistor, resulting in the second model

 103
P-4.9SE+9S

No. oF modes w7
No. oF elements = IS

R (3-6) = 2.21E+02 ohm

bI R (4-S) ft 2.22E+gl ohm
L (1-3) - i, 28F491 rJ4
L (S-9) - 7.72E+09 nH
C (3-2) x 1.41E-93 nF
C (3-4) = 2, k2E-04 nP
C (4-G) = 1.94E-92 nF
C (6-2) 2 L. HE-ai nF
9 (2-4) x S. 71E-93 S
ýT w 7.29E-92 ns;
(V across nodes
C (1-4) it 2. SGE-03rF
C (2-9) - 8.25F-gi rF
R (I-G) x i. 41E+02ohm
R (3-2) xI 19F494ohm
0 4L .
C (S-G) x 8.9SE-94 Ap
C (S-2) x i. 93E-94 nr-
DeleLe C(3-i)
Replace C(S-G) by R(S-G)

r -, 1.77E#93
No. oF modesjr 7
No. oF elements a 14

bi c- R (3-G) x 9, ISE+92 ohm

R (4-S) a 2.22E+gl ohm
L (1-3) w 1.17E+02nH
L (S-0) x 3.33E+09 nH
C (3-2) N 2. i2E-94 Mp
C (4-6) a 1.23E-92 nF
C (6-2) x 3. GBE-95nP
g (2-4) x 4.31E-92 S
(T a-S. 7GE-92ns
(U across nodes
C (1-i) - 3.3iE-23 nF
C (9-2) - 9.90E-94 nF
R (1-6) x 8.2, E+91 ohm
R (3-2) x 3. G3E+94ohm
R (S-6) a 1.916+93 ohm
0e C ýS-2) ar i. 19E-93 nP

Add R(I-i) Delete R(S-6)

Add R(I-S) Delete R(I-i)
Add R(1-0) Delete R(I-S)
FIdd R(2-i) Delete C(6-2)
.Add R(2-0)
Add new node with L(2-7)

r -r 3.94E+22

tic. oF modesw9
No. oF elements x IS

cR (3-6) 1.21E+03 Ohio

R (4-S) 3.32E+91 ohm
L (1-3) 1ASE-o,
92 rJ4
L (S-9) 2.37E*29 nH
C (3-7) 1.97E-94 np
C (4-G) 2.94E-02 np
9 (7-1) 7.72E--02 S
(T -LUE-92 ns
(U across nodes
C (1-4) 4.29E-93 rF
R (i-6) a. ! OF-Ql ohm
R (3-2) !. SGE+gs ohm
C (S-7) I. G3F--93 nF
Q (1-0) 8.9SE+92 ohm
R (7-1) 2. SGE+93 ohm
o A- R (7-9) - 1.18e+9i ohm
L (2-7) a 3.49G+93 nH ,
Add R(3--S)

Node I- Base
Node 2- C-illector
Made 0- Elnitter

Figure 6.6 Modelling the Common Emitter Configuration (Second Attempt

 104

shown in Figure 6.6 which had a value of F=1.77E + 03.

In Stage 2a number of additions and deletions were made resulting

in a model consisting of 7 nodes and 15 elements giving a value of

1.45E + 03.

On entry to Stage 3a new node was added by inserting an inductor

at node 2, the collector terminal. This was an interesting development

since PRL had stated that they had wanted an inductor at the collector

terminal of their model but had been unable to insert one successfully.

This new addition gave the last model shown in Figure 6.6 which had a value

of F=9.84E + 02.

Re-entry to Stage 2 resulted in one further element addition giving

the final model shown in Figure 6.7. This final model consisted of 8 nodes

and 16 elements and gave a value of*F = 8.81E + 02. The r. m. s. errors were

0.30 dB and 3.0* and the maximum absolute errors were 1.56 dB and 4.1*.

r=9.916+02

No. oF nodes -, 8
No. oF elements a iG

R (3-6) - i. M-493 ohm

R (i-S) a 3.22E+01 ohm
L (1-3) a 1.39E.+22 r.H
L (S-9) - 2.3SE400 nH
C (3-7) - 2. OSE-04 nF
C (i-S) 3.11E-02 np
(7-1) 8.29E--92 S
(T =--3.OGE--92ns )
(V across nodes 'I)
C (1-i) 4.41E-03 np
R (1-6) 8.19E+gl ohm
Q (3-7) 7.! SE-+OSohm
C (S-7) i. GSE-33 nF
R (1-9) = i. 41E+23 ohm
R (7-4) = 3.23E+93 ohm
0e Q (2-0) = S. 99ý+33 ohm
L (2-7) = 3. S3E+22 MH
p (3-S) - 3.32E+93 ohm

Node I- Base
Node 2- Collector
Node 9- EmMar

Figure 6.7 Final Model for Common Emitter Configuration

 los

* MI-ASUREMENTS

*PRLM ODFL

* NEW MODIL

Mr-in
Li j0 C
iiI

5 50

J,]riz
G -RES GHZ

r
-5 0

-100

2,0
-

C.
5 DI-
MrLD. -J PHASE'Df-ýG 521

5 2 00

0 FREQ GHZ i5o

r i 1-110

n ,-o
0
E!ý Gri Z
-15
r
-50
-20

Figure 6.8a s and S of Final Model for Common Emitter Configuration

11 21
 106

MEA5UREMEN15
PRL M0Fj L
t'
INL-ý MODEL

MSlri 'i i2 PHA'ýE DýG ý '12

20-5

20
i 50

-Pli i fl%
r,
ju

-30

Griz

--40
rý LL5G tl Z
IF
b9
10,
10

D, OlI., 322 PWA. DFG ý22

11,00

10

0 F-rlIE5 Griz

0 I GHZ

1r1lo

-10

1 r5c

-13

Figure 6.8b s and S of Final Model for Common Emitter Configuration

12 22
 107

The absolute error of 1. S6 dR occurred on s 2-1 at a frequency of 0.9 GHz

where the value of the modulus was -17.52 dB. Thus, although this

error appears comparatively large, it might not be considered to be of

great significance. The next highest absolute errors in the moduli were

0.9 dH in dB and 0.73 dB in -13.23 dB.

-11.54

The s parameters of this final model are shown in blue in the

graphs in Figure 6.8a and 6.8b. It can be seen that they are a consid-

erable improvement over the values given by the PRL general model.

The computing time taken to generate the final model in the second

attempt was approximately 1900 c. p. u. seconds on the UMRCCCDC 7600.

6.3. Common Base Configuration

The model for the commonbase configuration based on the PRL general

model with the negative-valued capacitor gave a value of F=1.67E + 05.

Changing the value of this capacitor from pF to 0.1 pF, changed the.
-0.8

value of F to 2.12E + 05. Using this model consisting of 8 nodes and

17 elements as an initial model, Stage 1 failed to find a local minimum.

The attempt was terminated after indications that either the voltage con-

trolled current generator should be removed or the nodes across which the

controlling voltage was taken should be made short-circuit.

In view of the results described in Section 6.2, an initial model

based on the author's final model for the common collector configuration

was then used. This is the first model shown in Figure 6.9 and consisted

of 7 nodes and 15 elements giving a value of F=8.65E + 05. As with the

initial model used in the second attempt for the commonemitter configura-

tion, this initial model lacked the substrate elements that were present
in the PRL general model.

In Stage 1, three elements were removed, resulting in the second

model shoArn in Figure 6.9, which gave a value of F=3.09E + 03.

 108

EMU
No. oF nodes -7
No. oF elements = iS
R (3-6) u 2. OIE+02 ohm
e1 c R (i-S) x 2.22E+01 ohm
L (0-3) 2 1.99E+91 nH
L (5-1) a 7.70G+29 MH
C (3-2) a 1.41E-03 nF
C (3-4) 2 2.12E-gi nF
C (1-6) 321.9iE-02 nF
,C (6-2) 1.8GE-8i nF
9 (2-1) 5.71E-02 S
(T = 7. DQE-02 ns )
W across nodes '4)
C (0-i) 2. SGE-03 np
C (1-2) p
9.2SE-04 1-,
R (0-6) 1.41E+02 ohm
R (3-2) 1.49e+04 ohm
1) C (S-G) z 8.95E-94 nP
C (S-2) J: 1.83E-94 nF

DeleLe RO-G) i C(4-6) t C(S-2)

Replace C(I-2) by R(I-2)
3.29E+03
No. oF nodes -7
No. oF elements a 12

R (4-S) x 4.9ge+20 ohm

t
e1 L (0-3) x B. SOE+20nH
L (S-1) a 4.24E-+Q9nH
C 0-2) u I. SSE-03 np
C (3-1) x 4.48F-01 nF
C (6-2) a 1.92E-04 nP
(2-1) S, 2.90E-02 S
(T a-1.39E-01 ns
(V across nodes
C (0-4) 8.7? E-Di nF
R (1-2) 4.29E+04 ohm
R (0-G) 7. G4E+gl ohm
R (3-2) 1.3sE+23 ohm
C (S-G) 1.36E-02 nF
b

Add C(1-3) M-0)

Add C(2-0) Delete Rý1-2) P Wi-S) (deleting node) r- = I. SSE+93
No. oF nodes c6
No. oF eleitemts = 13
i L (0-3) - 1.78E+21 nH
e1 L (4-1) - 1.31E+31 nH
C (3--2) 7.3SE-91 np
(3-1) 1.49E+03 ohm
IR
C (S-2) 8.31E-OS nP
9 (2-1) 2. SIE-22 S
(T a-9.2? E-03 ns
(V across nodes
C (9-1) 1.77E-03 nF
R (0-S) 9.3SE+Ial ohm
R (3-2) 1.81E-+03 Ohm
C (4-S) 8. GiE-93 np
C (1-3) -1.32E-Oi Mp
C (1-12) ;., K-03 np
C (2-2) !. ZGF--03nF
b0

Add C( 1-2 )i C( 5-9 )

Add R(I-1) Oelete C(O-1) 9(3-4)
Add C(1-2) C(4-0)
Add new node with R(G-2)

Made 1- EmlLter
Mode 2- Collector
Mode a- Base

Figure 6.9 Modelling the Common Base Configuration

 1,09

In Stage 2, duting the addition and deletion of elements, one node was

also deleted. The model produced after deletion of this node is the

third model in Figure 6.9. This model gave a value of F=1.88E + 03.

Continuing in Stage 2, further elements were added and deleted, C

1-2
being added, then removed_and then added again, until a model having 6

nodes and 15 elements giving a value of F=6.92E + 02 was produced.

Although sufficient accuracy had already been obtained, an entry

into Stage 3 resulted in the addition of a new node which created the

substrate elements in a similar form to those in the PRL general model.

The value of F was also reduced to 3.63E + 02. This final model,

which is shown in Figure 6.10 gave r. m. s. errors of 0.2 dB and 1.9*.

The maximum absolute errors were 0.7 dB and 3.40.

F=3. G3E+02
No. oF nodes a7
No. oF elements = IG
L (9-3) 1.99S+91nH
c.
e' L (i-I ) 3.03F+91nH
C (3-2? G.77E-gi rf-
c (S-2) 1.99E-04 nF
9 (2-1) 2. G2E--02S
(T -1.98F-01 ns
(U across nodes
6 R (9-S) a 3.40E+01 ohm
R (3-2) a !. i7E-+9I ohm
C (4-S) - IME-02 nr-
C (1-3) = 3.91E-gi np
C (G-0) a 1.42E-03 nF
C (S-9) = Lile-92 nF
R (1-4) 9. nE+gl aN,a
R (3-S) 9.99F:492 nhm
0b C (1-2) S. 97E--GS nP
C (4-3) i. 2Sý-03 nF-
R (G-2) 1.9sr-+91 ohth

Node I- GniLter
Node 2- Collector
Node 9- Sme

Figure 6.10 Final Model for Common Base Configuration

 110

VEýSupEwllýý -Is
PRL MOD, ý-L.
F'ý ML
'N

morl, 5 EI nFilýC*)Eri lu

ESGHZ
15Cl

If'
-

- rHZ

-20

_2r5 -1%J

r0k,
'15 '121 PHASE DEG 521

5 0 FQEIIJGriz
0s
0 HZ

-i,,,

-15

-20

Figure 6.11a S and S in Common Base Configuration

11 21
 MEA5UREMENTS
PRLM CCU'
L
N

m0'n-.5EI' 'i j2 Fill-G

PLIA'JE- 12

1OL
0
2ri

25
ýjrjz

35

-31i

-40
-15C

-45 ýýHZ
10

nki
nj !-jb, 'J22 rj,,. 322
-G

10

FRE5 CjHZ
Dr"
0 Griz

IC'
-I

-13
-200

Figure 6.11b S and S2 in Common Base Configuration

12
 112

As can be seen from the graphs in Figures 6.11a and 6.11b, this

final model gives excellent agreement with the measured s parameters

and is a considerable improvement over the PRL general model.

The computing time taken to generate this model was approximately

1120 c. p. u. seconds on the UMRCCCDC 7600.

6.4. General Model

The total computing time taken to generate the three individual

models for the different configurations of the vertical n-p-n transistor

was approximately 3620 c. p. u. seconds. If all the available data was

to be used in an attempt to produce a general model, then it was only to be

expected that the analysis of the errors at any particular stage would take

at least three times as long as for the individual models. Furthermore,

as the number of nodes and elements in a general model would be expected to

be greater than for the individual models, this could contribute to a

considerable increase in the required computing'time. The amount of com-

putation in the analysis of a model increases by the order of the square

of the number of nodes for each additional node.

The simplest way to reduce the computing time would be to fit

the measured S parameter data at fewer frequencies. However, the

author decided to proceed using all the available data.

The initial model used in the first attempt to develop an improved

general model was the PRL general model. In the second attempt, the

author produced an initial model based on the final models generated for

each of the common collector, common emitter and common base configurations.

6.4.1. First AtteýTt

The PRL general model shown in Figure 6.1 gave an overall error

function value of F=3.11E + 05. The r. m. s. errors were 3.3 dB and

 113

Fa3. IIE+05
No. oF nodes -9
No. oF elements a 22

5 3 R (4-S) z 1.13G+92 ohm

b R (S-G) a 2. SIE+03 ohm
R (3-6) a 9.43E+94 ohm
R (6-8) - 1.99[+99 ohm
R (7-0) - 6.23E499 ohm
1. ( 1-0 it 4.72E+09 nH
L (9-2) 4. i GE+9QnH
C (4-G) I. VIG-03 np
7 C (5-6) x I. OSE-02 nP
C (S-3) u 8.23G-95 nF
C (3-6) i. 20E-OS np
C (3-7) 1.31E-93 mr-
C (4-3) I. G2E.-94 nF

2 jE C (1-3)
(1-0)
SME-as nF
1.23E-03 np

I0 (3-0)
(2-G)
S. SQF-94 np
3.92E-02 S
(T w-L. DOE-92 ns )
(U across nodes 5g)
C (2-9) x 1.22E-93 nF
C (1-2) a S. WE-QS nF
C (2-3) x SME-gs nF

No. oF nodes w7
No. oF elements a IS
R (1-4) I. O?E+92 ohm
___ ___ C. G (4-S) 1.38E-93 S
R (6-0) i. 9SE+QIohm
L (5-2) x 1.48E+21nH
C (1-5) 2.! IF--03 nF
C (4-5) 7.8ae-93 MF
C (1-3) 7. ise-gs np
R (3-S) 1.47E+03 ohm
C (2-6) 2.43E-03 nF.
C (1-3) 1.62e-04 nF
C (1-0) - 1.83E-93 nF
R (3-0) 0 3.63E-+03ohm
9 (3-S) a 2.97E-92 S
e2 (T 2 1.7SE-92 ns )
(U across nodes 34)
C (2-0) wI G7E--93nF
C (1-2) u SAIE-94 nF-
0

x G.Vt+24
No. OF modes a8
No. OF elements a 21

3 R (7-4) x 9.3SC-+03ohm
R (G-2) a 1.31E+91 ohm
L (S-2) 2.32E-+9l nH
C (I-S) 2.3SE-03 nF
C (4-S) 3.32E-03 nF
C (4-3) S. ISEAS MF
C (3-G) 2.2SE--93nP
R (3-9) a 8.47E-+23ohm
9 ý3-9) & 2. SK-02 S
(T w-4. S.)G-92 ns )
(V Qcross nodes S)
C (I-G) 2.33E-93 Mp
C (2-G) i . 48ri-.23 nr-
R (1-2) 2.91E+02 ohm
R (1-3) !. D! E+OS ohm
R (2-S) G. 20E+92 ohn
C (1-4) ?. S7E--94 np
L U-1) I. I7E+02 nH
0 C (3-7) I. SGE-04 nF:
R (S-7) 1.71E+97 ohm
C (S-9) 2.92E-OG nF
Made i- Base R (7-9) = 2.99E+93 ohm
Node 2- Emitter R (1-4) 2 1.9? E+92 ohm
hoe 3- Collector
Node 0- Ground

Figure 6.12 General Model - First Attempt

 114

N
32.8". As can be seen from the graphs shown earlier, the model gave a

good fit at low frequencies but there were gross errors at higher

frequencies.

on using this model as the initial model for the general case,

Stage I caused the reduction of the model from 9 nodes and 20 elements

to 7 nodes and 15 elements. These are the first two models shown in

Figure 6.12. The two node deletions were attained by (i) the deletion

of the base terminal inductor and (ii) the deletion of the resistor

in series with the emitter terminal inductor. The value of F after

Stage 1 was 1.67E + 05. The individual errors now became more evenly

distributed although they were still smallest for the common collector

configuration.

After one entry each to Stages 2 and 3, during the latter of which

the base terminal inductor was re-inserted, and one further entry to

Stage 2, the final model shown in Figure 6.12 had been produced. This

model consisted of 8 nodes and 21 elements and gave a value of F =6.07E +-04.

The r. m. s. errors of this model were 1.45 dB and 14.50, but the maximum

absolute errors were 1.5 dB and 380 in the common collector configuration,

6.3 dB and 530 in the commonemitter configuration and 8.0 dB and 59* in

the commonbase configuration. Thus, although a large improvement had

been achieved, the errors were still rather large.

This model had taken approximately 6000 c. p. u. seconds on the UNIRCC

CDC 7600 to develop. Because of the size of the model each further

element addition was now taking an excessive amount of time. It was

therefore decided to make another attempt to produce a general model using

a different initial model.

 lis

r- = 9.78F+GS
No. oF nodes -9
No. oF elements i3

I L (1-4) a 3.20E421 nw
lo q (4-S) x 2.30E+92 Ohm
R (I-S) a 8.29E+91 ohm
C (1-3) x 7.90E-94 nF
R (4-3) it I. OQE+giohm
C (S-6) x 2.30E-92 nr-
9 (3-G) 2 SME-92 S
(T a-1.00E-92 ns )
(V across nodes 11)
L (7-2) a S.02EQ9 01
C (1-6) N 3.20E-03 nF
C (S-3) - 2.20E-94 rF
R (6-7) a 3. ME-+Ql ohm
C (3-7) a S. 09E-04 nr-
C (3-2) a L.QOE-OS Mr.

r- - 3.0SE+0S
No. OF nodes -7
No. OF elements a9
L (1-4) 2 S.SPE+99nH
b R (4-S) a 9.96E+91 ohm
C (4-3) - 2. S7E-94 nF
C (S-6) a i. ISE-92 rf
9 (3-6) a 2.37E-02 S
(T. -SAGE-02 ns
(U across nodes
L (G-2) x S. 9GE+99nH
C (1-6) a 2.25E-93 nP
C (3-6) u 4.37F-OGnp
R (3-2) u S.?SF+03ohm

r- a 4.78F+94
No. OF modes-9
No. OF elements = 19
L( 1-4) I. M+01 rAi
b1 C (4-3) G. GIF-OS nP
U. -G) 2.026-02 S
(T --L. GOE-91ns
(U across nodes
L (6-2) 3.23E-+QlnH
C (1-6) 3.97E-93 np
C (1-9) i. 3SE-03 MF
C (2-1) 1.21E-03 nF
C (2-9) i. i2F-03 np
9 ý4-5) x G.Sir=+33ohm
R (2-4) - 2.; ýIZE+02ohm
C U-3) 2.23E..93 pip
(S-? ) 1.49R+gi ohm
(7-9) 3. ýDE401 ohm
R (1-7) 1.3ýE+93 ohm
C (6-7) 9.22E-24 np
R ý2-3) 2.33E-+24 ohm
0 L (9-9) 2.17E+90 nH
C (1-3) i. ý2E-gi Mp

Mode I ease
Mode 2 Emitter
Mode 3 Collector
Mode 9 Ground

Figure 6.13 General Model - Second AtteTpt

 116

6.4.2. Second AttepRt

For the second attempt, the first model shown in Figure 6.13 was

used-as the initial model. This model was derived by coinbining those

elements of the three individual models developed by the author that

appeared in any two models. This model consisted of 8 nodes and 13

elements and gave a value of F=9.78E + 05. After Stage 1 the model

was reduced to 7 nodes and 9 elements giving a value of F-3.05E + 05.

This simple model, which is the second model in Figure 6.13, gave r. m. s.

errors of 3.3 dB and 32.5* and maximum errors of 11.0 dB and 1520.

After two entries to each of Stages 2 and 3, the final model

shown in Figure 6.13 had been obtained. There are two points about

this model that provoke comment. The first point is that node 5, at

which the controlling voltage of the current generator is taken, can be

seen to be merely splitting an element of resistance between nodes-1 and

7. However, this situation was developed by a complicated series of

additions and deletions and until a late stage other elements had been

connected to node S. The second point to be made concerns the inductor

L This element was added in the second entry to Stage 3 to generate

0-8*
a new node. In commonbase configuration, there is thus one inductor at

each of the emitter, collector and base terminals, but in common emitter

configurations, a rather complicated structure is

and common collector

formed in the region of the ground node. However, the addition of this

and node caused a significant reduction in the objective function

element

from 7.23E + 04 immediately prior to their inclusion to 4.78E + 04.

This model therefore gives a better overall result than that

obtained in the first attempt. The r. m. s. errors of this model are 1.3 dB

and 12.9* and the maximum absolute errors are 4.8 dB and 28.1* in the

commoncollector configuration, 2.3 dB and 26.2* in the common emitter

configuration and 6.2 dB and 36.50 in the commonbase configuration. Iýt

 117

MEASUREMENTý
PRL MODý-L
NLW MODF-L

MODD6 311 PHASE DEG 311

50

FýEG GHZ 0 -,\EG GHZ

)0
-1ý0

-100

-150

-20 -200

MOD DB S21 PHASE DEG S21

10 IN

5 50

0 FREQ GHZ 0 GHZ

)0
-5 -ýo

-10 -100

-15 -150

Figure 6.14a s and S for General Model in Common Collector Configuration

1 1-ý----21
 118

* MEASUREMENTS
w PRL MODEL

* NEL-W
MODEL

MOD DB S12 PHýSF.DEG S12

150
5
1100
0 ý-REQGHZ
0
50

0 EQ GHZ
-10 10-1
-50 -
-15
'10
-1 Ili
-20

MOD DB 522 PHýSF S22

-DFG
150

0 FREQ GHZ
100
)0

50

0 GHZ

-50

-20 -100

-25

6.14b s and S 22 for General Model in Common Collector Configuration

Figure 12
 119

PRL MODEL

moo IDB 511 PH,

A3E- rj'EG 311

0 FREU GHZ
FREQ GH7
1;
)0

-10

-15

-20
-200

MODOB 521 PHASE DEG 321

5 200

0 FREQ GHZ 150

D0
5) 100

-10 50

-15 0 GHZ

-20 -50

Figure 6.15a s and S for General Model in Common Emitter Configuration

11- 21
 120

* MEASUREMENTS
PRL MODEL
L

MODDB S12 PHASE DEG S12

-15
200

-20
150

-25
100

-30
50

-35
FREQ GHZ
D0
-40
",--REGGHZ -130
1 10 0
10'

MOD DB 322 PHAK DEG S22

100
10
50
5
0 FREQ GHZ
. 4,4
0 Q GHZ 10- )0
-50 -
-5
-100
-10
-150
-15

Figure 6. lSb s and S 22 for General Model in Common Emitter Configurat ion
12
_
 121

MEASUREMENTS
P.RL M'uDf-L.
NEWMODEL

li jI FIHýCJF-
MOD
FF-fý,
EQ GrIz 1K

r oro

-ý'EQGHZ

rý0
.0

25

menD OL" PHASE-

L) J21 ý"L-ýG

rý 0 'IHZ

F L-ý Gmz
50
.

1fil0

150
-10

20 ri

? 50

6.16a S and S2 General Model in Common Base Configuration

Figure 11 ___for
 122

MEASUREMENTS
PR' '--L
-0
ý41EW mn DL

PHýSE DEG 512

MO DE

,0 lý
50

C) -REQ GHZ

30 10

-50
-35
-100
40
-150
45 IKI
Griz
--
7S9
10-,'

522 PHASE DEG 522

MOD DB

10
50

5
0 FREQ GHZ
ýE:Q GHZ )0
0
-50

-5
-100

-10
-150

-15
-200

6.16b S and S for General Model in Common Base Configuration

Figure 12 22
 123

can be seen that not only are the maximum errors lower than those for

the final model obtained in-the first attempt but the distribution of the

errors is also different. This demonstrates that the model chosen for use

as the initial Model can have a significant effect on the results and that

it can be worthwhile to use more than one initial model if a satisfactory

model is not obtained after one attempt.

The S parameters of this general model are shown for each config-

uration in Figures 6.14,6.15 and 6.16. As expected, these graphs show

that the new general model is not as accurate as the individual models that

were developed. The new general model has removed the gross errors that

were present in the ýRL model at high frequencies but has lost some of the

accuracy the PRL model exhibited at low frequencies.

The time taken to develop this general model was 8300 c. p. u. sec-

onds on the CDC

UMRCC 7600 and it was decided that no further improvements

should be attempted.

6.4.3. Suggested Alternative Approaches

With the benefit of hindsight, the author now feels that the

approach used to produce the two general models was too ambitious.
Knowing that both the size of the problem and the computing time

required would be large, the first priority should have been to ensure

that the algorithm operated as efficiently as possible even if this meant

that some accuracy was lost. Therefore, in the analysis routine single

precision arithmetic could have been used and, as shown in Section 3.1.5,

this would have achieved a 50% reduction in the computing time required

for the analysis of the models. Furthermore, as mentioned in Section

5.2.2, the calculation of the approximation to the Jacobian matrix could

have employed forward differences rather than central differences, and

although this, too, involves a loss of accuracy, it would have halved the
 124

number of function evaluations required in that calculation.

Further savings could also have been made by using less of the

s parameter data that was provided. For example, instead of using

the measured s parameters at every frequency given, they could have

been taken at every alternate frequency given. This certainly might

have been a better approach using the initial model used in the second

attempt. However, in the first attempt, since the PRL model gave good

results at low frequencies, it might have been better to have attempted

to improve the model in stages by initially matching the S parameters

at low frequencies and gradually increasing the number of frequencies

and the maximum frequency at which the S parameters were matched.

 125

CHAPTER 7

A BIAS DEPENDENTMODEL FOR TWO SIMILAR BIPOLAR TRANSISTORS

The third set of data provided by PRL15 consisted of the sets of

s parameter measurements, given in the appendix, for two similar bipolar

transistors in commonemitter configuration. These transistors had the

reference numbers lE2M and 3E2M and were isolated n-p-n transistors made

in silicon with implanted arsenic emitters and diffused baron bases.

The JE2M had one emitter stripe and the 3E2M had three emitter stripes.
The data provided consisted of the s parameter measurements for

the two devices at 16 frequencies in the range 0.1 GHz to 2GHz with

collector-emitter voltages, V of 0.5 V and 3V and with emitter

ce,

currents, Ie, in the ranges 0.5 mA to lOmA for the lE2M and 0.5 mA to

27 mA for the 3E2M. The graphs in Figure 7.1 showing fT and the modulus
I fT
(GHz)
3.0-
1TM
2-0-
: ýýI 1M
E2
.01

Mod
S, I
a?lGHZ
(dB)

le(mA)
Figure 7.1 Transistor Characteristics for Vce ý0*5V
 126

of s2l at I GHz plotted against emitter current for a collector-emitter

voltage of 0.5 V, lead to the assumption that the small-signal operating

ranges of. the transistors up to IGHz were 0.5 mA to ImA for the lE2M

and 0.6 mA to 8mA for the 3E2M.

A bias dependent model for these devices had been developed by

16,17
Slatter This model was designed to represent the small signal

behaviour of the transistors up to a frequency of 1GHz for emitter

currents such that fT was still less than its peak value. The model

was based on the physical structure of the device and contained a number

of bias dependent elements. Some of the values of the model elements

were chosen using the construction of the devices as a guide, while

others were chosen using optimisation techniques. The relationships for

the bias dependent elements were based on theotetical considerations.

The results published by Slatter17 for this model show that in

general a good fit to the measured s parameters had been obtained for

frequencies up to 1GHz with the bias variation well simulated for

emitter currents of up to 2 mA for the lE2M and up to 4 mA for the 3E2M.

Slatter16 states that the model for the 3E2Mfits the measured s para-

meters more accurately than that for the lE2M and that s1 for the lE2M

shows larger errors than expected. He also states that the higher fre-

quency limits for the models might be 1GHz for the lE2M and 1.6 GlIz for

the 3E2M.

The author decided that an interesting experiment would be to

apply the modelling algorithm described earlier, to this problem. By

producing optimised models for each of the bias conditions, an alternative

method of studying the bias dependence of the model elements could be

achieved. Some of the results of this experiment were published by the

76 in 1981.
author
 127

7.1. Optimisation of the Model

The author started by producing a model for the lE2M using the

modelling algorithm described earlier and using the same weightings on

the individual errors in the modulus and phase as were used when modelling

the vertical n-p-n transistor described in the previous chapter. The

initial model used is shown in Figure 7.2. This model, which consisted

of 8 nodes and 12 elements, and the initial element values shown, were

based on the PRL bias dependent model. A new model was produced for

the IMI over the frequency range 0.1 GHz to 2GHz for V = 0.5 V and
ce
Ie `2 4 mA. This model consisted of 6 nodes and 14 elements and gave an

overall error function, F, of 5.15E + 02. However, this model was at

the limit of the linear region of operation of the lE2M and when it was

used as the initial model for the same transistor with V =3V and
ce
Ie = 0.5 mA, further changes were made resulting in the less complicated

model shown in Figure 7.3. This model consisted of 6 nodes and 9

elements and gave a value of F=4.35E + 02 for the lE2M with v = 3V

ce
and Ie `ý 0. S mA. The r. m. s. errors of this model were 0.2 dB and 1.8"

and the maximum absolute individual errors were 0.7 dB and 3.8". The

graphs in Figure 7.4 show the s parameters of this model compared with

the measured s parameters. The scales of the graphs cover the same

absolute ranges as those in the previous chapter. This very accurate

model was developed in approximately 2200 c. p. u. seconds on the UMRCC

CDC 7600. This time includes the two stages in the development of the

model, from the initial model in Figure 7.2, through the first model

developed for V = 0.5 V, to this final model for V = 3V

ce ce .
Now, using this latest model as the initial model for the lE2M

with Vce 'ý 0.5 V and Ie=0.5 mA, it was found that excellent results

were obtained without any further topology changes. Furthermore, it was

found that this model, with the element values optimised, also gave good
 128

No. oF nodes =8
No. oF elements = 12

bI c. R (1-3) = I. OOE+01ohm
R (3-4) = 2. OOE+01ohm
C (4-S) = 3ME-05 nF
R (5-2) = I. OOE+01ohm
C (3-6) = IME-03 nF
R (1-6) = S. OOE+01ohm
C (4-6) = IME-0i nF
g (S-G) = I. OOE-02S
(T =-2 SOE-03 ns )
.
(V across nodes 4)
R (6-0) = I. OOE+00ohm
C (2-7) = 1,00E-03 nF
R (7-0) 2. OOE+02ohm
C (1-2) I. OOE-OSnF

Mode I- Base
Node 2- Collector
Node 0- Emitter

Figure 7.2 Initial Model Based on PRL Bias Dependent Model

P= V30+02

No. oF nodes =6
No. oF etements
bI
C (1-2) = 1.33E-Oi nF
9 (2-5) = 2.5SE-02 S
(T =-2. GGE-02ns
(V across nodes s
C (i-S) 8. ISE--Oinp
C (3-5) I. We-03 nF
R (3-1) i. giE+01 ohm
R (1-2) 1.9! E+02 ohm
R (1-5) I. Sep+03ohm
R (O-S) 1. 90"E+01oh-n
C (2-3) 9.07E--OSnP
0

Node I- Base
Node 2- Collector
Node 0- Emitter

Figure 7.3 Final Model for 1E211 for V ý3V and I. = 0.5 mA
ce
 129

* MEA5UREMENTS

,k NEW MODEL

DF 1ýiI PH L-'

: - -,
ýEI-,
I, -- -jFlZ F-E riZ

i 5c

MOD DE '121 PýAl-j ýjHlý '121

15
25ý
1I"
IL, -, , r,

ý- F\ t ýJ 'i Fl Z
0 I '\'

It

ý rýll-ý li'OZ

7.4a s and S of Optimised Model for V =3V and 10=0.5 mA

Figure 11 21 ce
 130

MEASUREMENTS

NEWMODEL

ti -12

2CIG

-25

r
--3,
IE IJ-1
'I--1ý Iýriz
39 10
-

-40 EQ 1ý z
10

MC,DD In' 22 PýW)f: D,-,G ý22

I (\(
1C, I JLI

-1Ei. JdZ C F-1ý1ý0GHZ

I 'r

-15C

7.4b S and S of Optimised Model for V =3V and 10=O. S mA

Figure 12 22 ce
 131

results for the MM with Ie=0.. 5 mA and with both of the values of V
ce
Following this, each of the four new models was used as the

initial model for the same transistor and the samevalue of Vce with the

next higher value of Ie* Excellent results were again obtained without

further changes in the topology and in every case the optimisation of the

model element values took less than 200 c. p. u. seconds on the Leicester

University CDCCyber 73. Therefore the process was repeated for still

higher values of Ie , until the results indicated that a topology change

was necessary to achieve similar accuracy. The values of the objective

function for each of the models having the same topology as that shown in
Figure 7.3 are given in Table 7.1.

Objective Function, F

mA lE2M 3E2?4
e
V = 0. sV V = 3V V = O. SV V = 3V
ce ce ce ce

O. S 3.33E + 02 4.35E + 02 4.12E + 02 S. 82E + 02

1.0 6.62E + 02 8.94E + 02 4.04E + 02 6.96E + 02

2.0 6.80E + 02 1.22E + 03 3.7SE + 02 3.99E + 02

4.0 3.80E + 03 1.6SE + 03 3.94E + 02 6.27E + 02

8.0 1.46E + 03 1.11E + 03--

Table 7.1. Values of the Objective Function for Optimised Models of the
lE2M and 3EN Transistors

It was felt that satisfactory accuracy had been obtained at values

Of Ie up to 4mA for the lE2M and up to 8mA for the 3E2M. Although the

value of F=3.80E + 03 for the 1EN with V = 0.5 V and Ie=4.0 mA,
ce

giving r. m. s. errors of 0.5 dB and S. 4% was relatively high, it was still

considered acceptable since the maximum absolute errors of 1.7 dB and 15.2*

occurred at 2GHz, the highest frequency, where the moduli of the s para-

meters ranged from -8.9 dB to -IS. 28 dB. The effective ranges of Ie of

 132

this model for the two types of transistor agree with the suggested small-

signal operating ranges of the devices given earlier. Thus a single

accurate model for the small-signal operation of the two similar bipolar

transistors covering the frequency range 0.1 GHz to 2.0 GHz had been

developed.
Although the modelling process had gone through a number of diff-

erent stages, this new model was quite similar to the PRL bias dependent

model. - Looking at the elements in the final model in Figure 7.3 those

which have obvious equivalents in the PRL model in Figure 7.2 are C
1-2"
C and g2-5' Those having a slightly weaker resemblance to elements
3-2
in the PRL model are R, C4-, and R The remaining elements: R
-4. 5-6' 1-32
R and C seem to form a 11 arrangement where in the PRL model there
1-5 3-5 ,

was aT arrangement. The well-known T-Hor star-delta transformation

is described by Weinberg77 and the author suggests that there is the possi-

bility that the n arrangement in the final model is a transformation of

the T arrangement in the PRL model, with some of the extra elements that

would be present by this transformation rendered unnecessary by the optimi-

sation process. Thus, by the T-11 transformation, R in the new model

1-3
would be equal to R, +R 3-4 in the PRL model and should also have a
_,
series inductor. The 11 arm between nodes 1 and 5 should consist of a

resistor, equal to Rl_, in the PRL model, in series with a capacitor, and

the n arm between nodes 3 and 5 should consist of a resistor and a

capacitor in series. Thus it could be conjectured that in the new model

Rl_, is equivalent to R, +R in the PRL model, R1_5 is equivalent to

_, 3-4
R in the PRL model and C in the new model is due to C4-6 plus some
1-3 3-5
additional capacitance from the T-H transformation.

Having developed the new model, the next step was to examine the

variation in the values of the elements in the model with the bias condi-

tions. The results of this exercise are given in the next section.
At values of Ie greater than 4mA for the lE2M transistor and
 133

8mA for the 3E2M, topology changes were indicated. Details of some

models developed for these cases are given in Section 7.4.

7.2. Examination of the Bias Dependent Elements

Taking each of the model elements in turn, graphs were drawn of

the element values against the emitter current. These graphs are shown

in Figures 7.5a and 7.5b. In each of these graphs the points-are joined

by straight lines. The red lines refer to the lE2M transistor and the

blue lines to the 3E2M. The solid-colour lines are for V = 0.5 V and
ce
the dotted-colour lines are for V = 3v . The black dotted lines are
ce
the functions that were fitted to these points and will be described

later.
These graphs were extremely encouraging as most of the elements

showed a distinct pattern of variation with the emitter current and the

collector-emitter voltage.

Although the model was different from the PRL bias dependent model,

some of the elements in the two models could be compared. The bias

relationships in the PRL model involved only VVce ,V ce and I, these

e
being based on theoretical considerations. The author decided to use a

curve fitting routine from the NAG library, E02ADF78, to examine the fits

obtained to each of the sets of points. First, only the relationships

Iwere studied, allowing of functions VT 12 and reciprocals

with eeee ,I,
thereof. Then the effect of Vce was taken into consideration. The bias

dependence of each element is discussed below with reference being made to

the model shown in Figure 7.2 for the comparisons with the PRL bias depen-

dent model.

C1-2 - This element had a direct equivalent in the PRL model which

they suggested was not current dependent, with values of 0.33/VV- pF for
ce
the lE2M and 0.621VV pF for the 3E2DI. In the author's model the
ce
 134

C' (1
-2)
N' GM "3

ý
30,39
---:

311)
ýý61 0, -ýý X

[iJrN M

NU-C c (4-51 N,

/
/ /
0.
/
-- -- -----------

Li I ',
_) c_)

C kj Q
t?,, _-1, j -ý !JA

C (3 -T

- lE2M Oý.5V

1E2M 3, OV

3lý2M OAV

3
ýUM jý 3v
=
1.0'1 -w=.
ý-L1
iNk'7f C,1,-)

CURQt:NT, NA

Figure 7.5a Variation of Model Element Values with Emitter Current

 13.5

(3 -1) 0 1-11ý R (4- 2) C,

20
-- --------

40 40
3 ri
2O
.3r

"r

0-
2'7e

MIA !IA

R (i -ý'jý 0`11'ý R"0, -,J:ý 011N

22'3
IA
14
12
140
203

!JA
(j ý r, ý- KI -, !I
ý

C (2--3) N'

-1 E2M 0.5V

1E2M 3. OV

- ýE2M 0.5V

7 ?m 11 ýv
,
----------

----------- F-u INý' -1 cl,Ni1ý

- -- 11

5' 6

Figure 7.5b Variation of Model Element Values with Emitter Current

 136

value of C appeared to reach a limit where the maximum values obtained

1-2
in the ranges of le und.er consideration were

0.47 pF for IE2M at 0.5 V giving 0.33/YV PF

ce
0.24 pF for lE2M at 3V giving 0.42/VV PF
ce
0.99 pF for 3E2M at 0.5 V giving 0.701VrV PF
ce
and 0.45 pF for 3E2M at 0.5 V giving 0.78/Y'V pF
ce

However, the value of C1-2 was definitely varying with the value of Ie

at lower values of Ie. The model was also sufficiently sensitive to

the value of C1-2 that a 5% change in the optimum value Of CI-2 made a

significant change in the value of the objective function, F. Thus the

functions chosen for C1-2 were functions of both Ie and Vce as shown

in Table 7.2. ThL- black dotted lines on the graphs of C1-2 in Figure

7.5a show the function evaluated at the appropriate values of V and Ie.
ce

g2-5 - PRL suggested that the value of the current generator should be

proportional to Ie A further
relationship stated by PRL was that
1
where hfe re was element R4-6 in Figure 7.2 and h fe 0 so
r.
e
It is possible that R4-6 in Figure 7.2 could be equivalent to element

R in Figure 7.3. If this were so then g could be equal to SO/R

1-5 2-5 1-5,
Checking the optimised values of g and R led to the conclusion that
2-5 1-5
although this type of relationship was possible, the value of 50 was not

quite right. The average values were 44.4 for lE2M and 46.3 for 3E2M.

However, taking these average values with the optimised values of Rl_,,,

the values calculated for g2-, were not sufficiently accurate. Thus the

relationship chosen for 92-5 was a piecewise linear function dependent only

on I The sudden change in the value of this element for the 3EN at
e*
Ie=8 mA with V = 0.5 V was taken as a sign that the device was becoming
ce
non-linear. This effect can be seen in several of the model elements

and in each case it occurs at the highest value of Ie for which an acceptable
 137

value of F had been obtained with V "" O'S V- No attempt was made to
ce
model this effect and the fits of the piecewise linear functions chosen

can be seen as the black dotted lines on the graphs of g2-5 in Figure 7.5a.

-r - The time constant applied to the current generator in the PRL model was

given the independent values of -25 ps for the lE2M and -26 ps for the

3E2M. It can be seen that the optimised values Of T in the author's

model vary with the emitter current and that the values of T for the

lE2M also vary with the collector-emitter voltage. However, as there

was no apparent pattern in the variations in the values of T and bearing

in mind the PRL values for T, it was decided that constant values

should be chosen based on the optimum values at low emitter currents since

these were the values when the devices were most certainly in the linear

region of operation. The values chosen were therefore -28 ps for the

JUDI, the emphasis being on the values for the higher voltage, and -32 ps

for the MM.

C4-5 - The function given for the equivalent element in the PRL model

was A/F where A took a constant value of 1.12 pF. Vi for the lE2M and

1.16 PF. V1 for the 3E2M and F was a function of the collector-substrate

voltage V such that F= YV + 0.75. This element was one

c-sub c-sub
where the optimised values changed drastically at high values of emitter

current while at lower values the element values were almost constant.

Therefore, it was decided that, ignoring the high current effects, an

inverse square root function of V as given in Table 7.2 was appropriate.
ce

C - The optimised values of this element, as with C underwent

3-5 4-5'
large changes at higher values of emitter current and again, these effects

were not modelled. C could be considered equivalent to element C4-6

3-5
in the PRL model in Figure 7.2 perhaps with some additional capacitance

due to the T-TItransformation discussed earlier. C in the PRL model

4-6
 138

was dependent on the value of the current generator. Investigating the

possibilities of this type of relationship showed that similar smooth

curves were obtained for both transistors on plotting the optimised values

of C3-51g2-5 against emitter current. The function dependent on the

.
value of g given in Table 7.2, was found to give a best fit for this
2-5'

element.

R3-1 - This element could be taken as equivalent to R and R3-4

1-3
in the PRL model in Figure 7.2. These resistors together with C in
3-6
Figure 7.2 were used by PRL. to model some of the high frequency effects

of the transistors and the combined constant values of these two resistors

in the PRL model were 66 Q for the lE2M and 23 Q for the 3MI. In the

author's model, R appeared to depend on both Ie and V as can be seen

3-1 ce
from the graphs in Figure 7.5b. The maximum values this element attained

for were 54SI with V

the UZI `ý 0.5 V and 56 0 with V =3V and for the
ce ce
3EMI were 22 SI with V =O. SV and 26a withV = 3V
. These values
ce ce
were quite similar to the constant values used in the PRL model. However

the sensitivity of the model to the value of R3-1 meant that the effect of

changing current was significant. Therefore, it was decided that both the current

and voltage effects should Ve included. The effect of increasing voltage

appeared to be to increase the value of R by a constant amount over the

3-1
entire range of Ie plotted in Figure 7.5b. The function developed for

this element is given in Table 7.2.

R - The equivalent element in the PRL model was given constant values
4-2
of 209LI for the lE2M and 185sl for the 3EN. It was felt that constant

values were the most appropriate and the values chosen by the author were

1890 for the lE2M and 14S Q for the MM. It can be seen from the

graphs in Figuie 7.5b that these values were based on the optimised values

of this element at Ie=0.5 mA.

 139

R1_5 - This element has already been mentioned during the discussion of

g2-5 . No relationship between these two elements was discovered. it

was also decided that there was no dependence on Vce exhibited by R

Functions of Ie were therefore fitted to the average element values for

the two values of V for each transistor. The functions giving the best
ce
fit are given in Table 7.2 and are plotted in Figure 7.5b.

R - This element could be equivalent to R in the PRL model in

0-5 0-8
Figure 7.2. PRL gave R constant values of l. lSI for the lE2M and
0-8
0.8SI for the 3E2M. These values did not agree with the values of R
0-5
in the author's model. This element appeared to depend on Ie particul-

arly for the lE2M for which the value of R0-5 at Ie00.5 mA was as high

as 18 Q, reducing to 4SI at Ie=2 mA. Functions of Ie were therefore

developed for this element.

C - In the PRL model, the equivalent element, C was defined as

2-3 4-5'

a function of the voltage between nodes 5 and 8 in Figure 7.2. This

voltage was calculated as (V -I R2-5)*. In the author's model there

ce e*
is no element equivalent to PRL's R However, the optimised values
2-5*
C
of 2-3 seemed to depend on both Ie and V Fitting functions to the
ce'
element values proved difficult because the element values at V =3V
ce
became zero. The final compromise was to fit a function of Ie to the

element values for V = 0.5 V first and then to modify this function as
ce
necessary for the higher value of V The final complicated function
ce'
given in Table 7.2 is not entirely satisfactory and the fits obtained

using the function are shown in Figure 7.5b. It can be seen that the

values with V `2 0.5 V are considerably more accurate than those with
ce
V=3V. Fortunately, the model was not highly sensitive to these
ce
low values of C2-3'
 140

Bias Dependent Constants Constants

Element Value of Element for for
(I in mA, V in V) lE2M 3E2M
e ce
(A
C +IB A = 2.84 A=6.22
1-2
e, 10
nF B =-1.12 B=-1.81
ce
L

2 A= - 1.12 A= - 0.47
9 (A + B. I ). 10- 1s1 mA B 1s1 mA
2-5 e B= 7 21 e = 4 . 64 e
.
S A= 3.4 A= 1.5
1>1 mA B= 3.0 >1 mA
B= 3.0 e e

T ns A A = 0.028 A=-0.032

C4-5 I- 3
A+ 10- A = 0.54 A=0.34
i+ V17-
nF ce

[A A = 0.024 A=0.024
C B
3-5 + I 92-5
nF e B = 0.018 B=0.055

R A = 62.9 A= 22.9
3-1 B
A V B = 43.1 B= 17.2
+I+C. ce C = 1.7 C=1.2
e

R
4-2 A A = 189 = 145
0

R A = 8.385 A=7.371
1-5 105
A+ B. I . .
Q e B = 108.32 B= 76.19

R A = 1.0 A=0.1
6-5 A+B
C
SI B = 8.0 B=0.8

C 5 A = 3.27 A=8.2
2-3 +BIeV lo-
ce B = 5.95 B=9.7
nF

Table 7.2. Bias Dependent Relationships for the Model in Figure 6.3
 141

7.3. Resultant Bias Dependent Model

The effect of the different bias conditions on the values of the

model elements was discussed in Section 7.2 and functions were developed

for each of the elements. This exercise could be seen as a feasibility

study for the determination of the effects of different bias conditions

using optimised models of transistors. Although the effects of differ-

ent emitter currents on the model element values could generally be seen,

the number of values of collector-emitter voltage, V for which data

ce ,
was provided was too small to give confidence in the proposed functions

of Vce' However, a bias dependent model had been developed.

It was known that some of the functions would not give the

optimised values of certain elements for V 0 0.5 V when Ie was

ce
greater than 2mA for the lE2M and greater than 4mA for the 3E2M. As

a check on the effectiveness of the new bias dependent model, graphs were

drawn of th6 calculated and measured values of the modulus of s at

21
lGHz against emitter current. These graphs are shown in Figure 7.6.

As expected they show that the model is accurate for values of modulus S21

up to its peak value. This appears to be slightly better than the PRL

model which, in the paper by Slatter17 is shown to lose accuracy for the
,
3E211 before the peak value of modulus s is reached. Graphs of the
21
S parameters calculated using the author's bias dependent model, together

with the measured s parameters, against frequency are shown in Figures

7.7 and 7.8 for the lE2M and 3E2M respectively, at the arbitrary bias

values of V =3V and Ie=2 mA . Looking at these graphs the author

ce
has some doubts about the accuracy of the measurements of s for the
12
IEV. However, the graphs show good agreement between the measured and

calculated values of all the s parameters and would appear to offer an

improvement on those shown by Slatter17 especially as the agreement con-
,
tinues up to a frequency of 2GHz while Slatter's model was only valid
 142

o MEASUREDVALUES
MOD'321 lE2M 0.5V ,, MODELVALUES

0 CURRENTMA
ýD 246 10 Je

-10
-15 -0

MODS21 lE2M 3. OV

10
5
n CURRENT
MA
To
-5
-10
-15

MOD521 3E2M 0.5V (Vre)

lo
5
0- ýQ CURRENT
MA
0m5--- 20 25
-5 ----

-15 - -0

MOD321 3E2M 3. OV (Vet)

10 -
5
0 CURRENT
MA
5 10 Is 20 25 le
-5
-10
L
-15

Figure 7.6 Modulus S21 at 1GHz for the Bias Dependent Model
 14

MEASUREMENT5

NEWMODEL

MIU DHPE: iE iI I

C:
5
'1
_rd HZ

"I
-

I "

0DD bi c,21 PLI'Vi. ýI

J2
ý .1 1-

20

1,r3 IC

1c

1 (r
rz

HZ

5C r
_________________________
-; -' 'ilL
-\

10' 10

Figure 7.7a s and S for lE2M for V ý3V and Ie=2

11 21 ce mA
 144

MEASUREMENTS

NEWMODEL

M'O5DF, i2 Dý'G '; 'i

r%r
ýQ

715

40,
Gilz U - -
!-Lý'L
ý ill-iz
-'

PC 1o"

MO3 D[ ýý22 ")22

If'

L
-J

0 0 ýAZ

4r rlc
-

15
-

Figure 7.7b s and S for IE2M for V =3V and Ie=2

12 22 ce mA
 14 S

A5UREMENTS
ME:

NEWMODEL

rulDLAI,

r
ic
-
0 HZ

-I 1ý11
C

-I 5c
'Az
10,

MODOD 21 PHTh ¶21

2C
15c
113

C -J

50
-

10,

Figure 7.8a s and S for 3E2M for V =3V and Ie=2 mA

11 21 ce
 146

MEASUREMENTS

NEW MODFL

MCC, CIE, 2 DHA)'F JHC

1DC

czc,
,

-09 1ý
"z
.

--

10 1O

MC.C. ID-11 ?
"--'2 PHVjH' Uý-G "'ý22

i rý

5
rr CjrZ
0 ltdz

ri
- 11.1,

-15

Figure 7.9b s and S for 3E2M for V =3V and Ie=2 mA

12 22 ce
 147

up to 1GHZ.

The conclusions drawn from this were that accurate bias dependent

models could be developed using the author's modelling algorithm and that

it was possible to develop quite simple functions describing the depen-

dence of the model elements on the bias conditions. Although some theo-

retical indications of likely relationships could be useful, they were not

absolutely necessary. Examination of the variation of the model element

values at different bias conditions could indicate the possible forms the

function could take. Then the use of a curve fitting technique with a

number of different functions of the bias variables could assist in the

final choice of the most suitable function.

7.4. Models for High Emitter Currents

-

After the bias dependent, model for the linear region of operation

had been successfully developed, attempts were made to develop models fot

the transistors at the highest values of emitter current for which data

was provided.
For each transistor, using the s parameter data at the highest

values of Ie given and with both values ofV given, the initial
ce
model used was the appropriate optimised model for Ie=4 mA or Iea8 mA

from the earlier work. Applying the modelling algorithm to these four

cases produced four different models. However, re-optimising each of

these four models for each of the four cases, one of the models proved to

be the most effective with V= '2 0.5 V and I 10 mA and 28 m.A for the
ce e
lE2M and 3E2M respectively, while another model was the most effective for

both transistors with V =3V and their respective highest values of I

ce e
These models are shown in Figure 7.9. The value of the objective

function in each case was less than 4.3-E + 03 and the maximum absolute

individual errors were 1.5 dB and 12*. The models were again effective
 148

No. oF nodes =2
No. oF elemenLs

Node I- Base
Node 2- Collector
Node 0- ErMter

0
IE2M O.SV lOmFl 3E2M O.SV 27mA

F=i. 26E+03 F --- I SSE-+03

9 =. 1.80E-01 S (6-S) = i. 97E.-Ol S

=-I. O?E-01 ns (T =--3.09E-02 ns )
(V across nodes s (V across nodes 3)
C (3-5) = G. iOE-02 nF C (3--S) 1.20E-01 nP
R (3-1) = 3.61E+01 ohm R (3-1. ) 1.36E+gl ohm
R (4-6) = 1.92E+01 ohm R (1-6) 2.36E+00 ohm
L (0-S) = 1.13E+00 nH L (0-5) S. iOE--Ol nH
C 1,1-4) = 1.18e-03 nF 3.06E-03 nF
C (3-4) = i. SOE--03 nF C (3-4) 9.04F.-03 nF
C (1-5) = 6. SGE-04 nF C (I-S) 4.03e.-03 nF
C (6-0) = 1.30C--03 nF (6-0) 1.90E-03 nF
R (1-5) = 1.33C-+02 ohm R (I-S) ý. GH+01 ohm
R (2-6) = ]. ý3E+01 ohm R (2-6) J. ISE+01ohm

bi c

lio. oF modes m6
No. oF eletents a 12
Node I- Base
Node 2- Collectir
Node 0- ErAitter

0
IE211 3.9V lOmA 3E2M 3.2V 2?mA

F=i. 2SE+23 P=1.42E+93

9 (2-9) =I SOE--g
IS (2-0) = ý. S4e-gl S
ýT =-i. 3SE-22 ns (T ::-3.2iE--02 ns
(v across nodes 2 (V acrass nodes 3) 9)
(1-0) = i. 90E-03 nF (i-0) = 3.16E-03 np
C (3-0) = !. SýE-02 nF C (3--0) = 3.97ý-02 nF
q O-S) = S. 93E+91 ohm q (3---) = 2. Z8F+31 ohm
R (1-2) = 2.91E+02 ohm R (1-2) = 2JOE+02 ohm
9 (5-0) = j. S@E+02ohm p (S-0) - G.S7E+el ohm
C (5-3ý = 9.42E-04 nF C=2. SSS-03 nP
1.98E+Oi ohm = !. ý7E-101ohm
(2-0) 1.39E-01 nP (2-0) i. s3e-oi n"-
L (I-S) 1.9ýE+00 nH (1-5) MH-01 nH
C (1-2) 2,72F-V nP C (1-2) S. iSE-Oi nF-
C (i-3) i. 02E-04 nP C (1-3) 9. ýH-04 nF

Figure 7.9 Models for High Emitter Currents

 149

MEASUREMENTS

NEV MODEL

IU
MOD DE lE2NI 0.5V I OmA ý?, IF 2M 3%, 1OmA

10

0 `CýE SHZ

0 ý rl, 1, -17
1- ýI

r
MOD DE, 3E2M 0.5V 28mA Iuj
__J 31,2M 3V 28mA
:

2:

' r\"--r-! li HZ
15

10

r5
C

20
-

Figure 7.10 Modulus S Against Frequency for the High Current Models
21
 iso

up to 2GHz and graphs of modulus s against frequency for each case

21

are shmm in Figure 7.10.

The author found this result interesting as it seems to confirm

the similarity Of the two transistors. The effect of the extra

emitter stripes in the 3E2M would appear to enable the transistor to

operate in a similar manner to the lE2M at higher emitter currents for

the same values of collector-emitter voltages. The fact that there

were different models at the two values of collector-emitter vol. tage

demonstrates that there are major effects on changing V in this region

ce
of operation of the transistors.
 151

CHAPTER 8

CONCLUSIONSAND SUGGESTIONSFOR FURTHERWORK

In Chapter 2, the standard models for bipolar"transistors were

described together with methods by which the values of the elements in a

model of any particular device could be determined. Where the model

chosen gives adequate accuracy there are no problems. However, the

author felt that there was insufficient guidance on the choice of alter-

native models and on the determination of their element values.

Chapters 3 and 4 provided some background information relevant to

the author's modelling algorithm which was developed for those cases

where a suitable model of a particular device was not readily available.

Chapter 3 described transistor parameter measurcments and gave details of

how models could be analysed in order to compare their performance with a

device for which parameter measurements were available. Chapter 4 then

described a number of methods of numerical optimisation.

The development of the author's modelling algorithm was described

in Chapter S. The algorithm started as one to optimise the element

values of particular models in order to reduce the errors between the s

parameter measurements of a transistor and the s parameters of the model.

From this, techniques were developed whereby, if necessary, the topology

of the model could also be optimised. The development of a model for a

lateral p-n-p transistor was used to illustrate the development of the

algorithm. Then, after a complete description of the modelling algorithm,

details were given of the development of a model for the lateral p-n-p
transistor using the final version of the modelling algorithm.

As a further demonstration of the efficacy of the modelling

algorithm, in Chapter 6 details were given of the models produced for a

vertical n-p-n transistor in each of commonbase, common emitter and

 152

common collector configurations. The models produced were highly

accurate, but attempts to produce a single general model for all three

configurations was not so successful. This was mainly because of the

size of the problem and the computing time required.

The final example of use of the modelling algorithm was in

Chapter 7, which described the development of a'bias dependent model of

two similar bipolar transistors. The modelling algorithm was first

used to develop a suitable model and then to optimise the element values

of the model for different bias conditions of the transistors. Using

these optimised element values as a guide, suitable expressions for the

model elements as functions of the bias conditions were suggested by the

author.

The modelling algorithm was developed with the main aim of

producing accurate models of particular transistors. The emphasis,

therefore, was on the achievement of results rather than on producing a

small, fast computer program. The author's view is that computers are

ever becoming faster and larger; therefore, although Computer programs

should be written efficiently, they should not take short-cuts that could

cause the failure to find a solution to a particular problem. However,

the development of the general model for the vertical n-p-n transistor

demonstrated that, with the facilities available to the author, some

sacrifices of accuracy might be necessary in order to achieve satisfactory

models for some large problems. However, these sacrifices might not

have a detrimental effect on the efficacy of the modelling algorithm and

tests could be conducted to check the effect of a number of ways in which

the modelling algorithm could be made faster. These include

the use of single precision arithmetic in the analysis

of the models,
 153

(ii) the use of forward differences in the calculation of the

approximation to the Jacobian matrix.

The use of a smaller increment in the calculation of the approximation to

the Jacobian matrix could improve the accuracy of this calculation.

This could offset the loss of accuracy due to (ii) above, and would not

cause any increase in the required computing time.

In Stage 2 of the modelling algorithm, the author decided that

each successful addition should remain in the model. One effect of this

was that sometimes elements were added and then removed again during the

addition of-a different element. Cases have been seen where a parti-

cular element is added and removed several times. There are two alter-

native techniques that the author considers would be worth evaluating.

The first of these would be to test each possible addition and to

select only the best of these. Then a suitable technique for further

additions in Stage 2 could be, either, to attempt to add the remaining

successful elements, one at a time, in order of effectiveness, or, to re-

evaluate all the possible additions, taking only the best again. The

former suggestion for further additions might seem to be the most effic-

ient method, but changing a model by the addition of only one element

could cause such differences that elements which could be successfully

added prior to the change, could no longer be inserted.

A second, alternative technique for Stage 2 could be to evaluate

each possible element addition and then to insert all the successful

additions together at their optimum values. The model could then be

re-optimised in a similar manner to Stage 1 so that any elements that were

not necessary could be removed. In this way a number of element

additions might be achieved in one operation.

For large problems such as the general model of the vertical n-p-n

transistor, even these suggestions might not reduce the required computing
 154

time sufficiently. Therefore, in view of the large amount of data

available for that particular problem, as suggested in Chapter 6, it

might be feasible to use only part of the available data.

In the development of the bias dependent model described in

Chapter 7, the modelling algorithm was used to calculate the optimum

element values at each of the different bias conditions. Instead of

choosing the constant values in the rather arbitrary manner used by the

author it might be possible to use an optimisation technique both to

choose which elements should take constant values and to ascertain their

optimum values. This would, like the general model of the vertical

n-p-n transistor, create a very large scale optimisation problem and it

might be necessary to use only a selection of the available S parameter

data.

After the constant element values have been ascertained, more

accurate functions for the bias dependent elements might be obtained if

the models are then re-optimised with the constant element values fixed,

The development of bias dependent models valid beyond the linear

region of operation could be possible using the author's modelling

algorithm. The author's approach to this problem would be to develop new

models for progressively higher values of Ie and V Hopefully the

ce *
topology of the models would only change slowly and these changes could,

perhaps, be incorporated into the models by the use of mathematical

functions which would allow, for example, the gradual reduction of a

resistor to zero, after which it would remain as a short-circuit. Al-

ternatively, elements or groups of elements could be made part of the.

model under certain bias conditions by the use of diodes in much the same

way as in the Ebers Moll models described in Chapter 2.

It is the author's opinion that the aims of the research stated

in the Introduction have been achieved. A modelling algorithm has been
 iss

developed which will optimise model element values and topologies in

order to produce accurate models of bipolar transistors using the s

parameter measurements of the devices. -The effectiveness of this

algorithm has been demonstrated by the modelling of several different

transistors. Although suggested model topologies were provided for

these cases, the author feels that an initial model based on the modified

hybrid pi model described in Section 2.2.2 would be suitable for the

modelling of most bipolar transistors. The author has also shown that

the modelling algorithm can be used in the development of bias dependent

models. A model for the linear region of operation of two similar

bipolar transistors using simple functions for the bias dependent elements

was developed by the author.

It was hoped that this regearch might indicate how the standard

models of bipolar transistors could be improved especially at high fre-

quencies. Although a number of different accurate models have been

described here, the author is of the opinion that a much larger selection

of devices must be modelled before any conclusions can be drawn concerning

the general applicability of the non-standard elements. However, by the

development of the new modelling algorithm, the author has provided a

means by which this can now be achieved.

 156

APPENDIX 1

S PAWIETER DATA PROVIDED BY PHILIPS RESEARCHLABORATORIES

 157

(L) CN CVI LM

41 10 1 1 1 1

Z
0

I
. 10 0 0 0

V
ca

$4
4) 10 C) rý 0
%. cel
41 W '0 r-4 P-4 r4
Aj
19

cn LM KM Co 0
PQ cz
"4
cz
P-4
4
r-i P.4
10

r4
Cd
34
41
93 4)
m 10

41

c4
0 c4
41 10 %ID LM tn

(Li
cn
c4
Cd W 0
cý f-4
. 92
43 M 1 1 1 1 Cli

rn

cr
C)

z
 158

cl r-
CIJ rij 0
) ..
i-,
.1.,:, 'Cl

Cj
rx..
r F...
r, Ln-.--j.-T. c-i
rr.. ---v
I..) r,
C.
")
- 4. ;Z7-
........... . c

-.1 r c. .-.i o cur.- on

Z
c
C I.C.
IT E
r=
c
1%

0
.001.7") -4 ...p f.'IJ CIJ 1=1

z.co:r, cr.r,.
,Z. C. i: ý

41

tw
C.j
ce
N 41
m
CT. Q
C) -4,0 C."
r,
. YJ k 4. d
0 "r-I 0
z 0 E
U, U
CD 0
0 Ln *a) 9:.

C, co C, 1=1 1 0 P-4
cl) .:: 1=11=11=1.: ý
C) Z ;Z
N
:=I tz
Cl Ci
Z IM
14:ýI..'- Z. i.'..-z' .:, Z,..Z, IZ,, Zl
 159

S PA;;, V--.E7ERS GF VERTICAL UP14 TRAi. -SISTOR Vi CU'-"! O-*4 CULLEClUR CO-SIGJRATION

FREQUENCY Sil 512 S21 S22

MOD PHASE v r)to PHASE fj D PHASE 1"Up PHASE
GHZ Do ULG Drs DU, L)d DF-5 U6 OEG
58.30 -. 11 qu 73.50
. 1,1 -. 73 -20.30 -11.57 -1.14
-22.
-29.40
-10.63ht)
-M 9 5 3. t, u
-20.70 -
-34.30 -7. b9 3zl. iu -2.18 -J. -49 315SOU
29 7u eb
-39. QU -t. 90 -3.15 -3h. '4u -h*44 ,90
: 30 -3.36 "
-42. 30 -0.: 58 -3.84 -37.2ký -o. hl 13,30
: .4.41 uo : 1.934s h
-. ý, .4U -64'Ad 14* uv
4; j
. 5ý -4o. -5.40
ti L ?:
bl2b -Sýi 0
IV ud4 au ') . -4 u tt5 17 plu
.
-0.04 -63. eo -4 Pf -7.0u 3u -5.7b -4d 0 - 9: 50
. 00
7. ) -0.65 -13. eu -4 73
-lb. -6.11 -4ý)., lij t) I -SI. 60
1 4.9 ,1
eu -7.19 -IJ14 SU
- 2,:) . !)u -b. 43 -51. Iu -7 o ut) -44 UO
. 90 -o . 62 -a 7 .! >0
. -7.58 -9b: 70 -ý) .ae -3b 9N -7 " hu -t)t3 uu
I uci -8.13 -114.10 -6.4. j -43: 9u -7.77 -05, Qu -d. t) 9 -74: 40

S PARAMETERS OF VERTICAL wt., TRANSISToR IN COMMON EMITTER CONFIGURATION

FREQUktiCy sit $12 S21 S22

MOD PHASE *v'UD PHASE 1400 PHASE '4UD PHASE
GHZ D8 OEG Ob DtG Ob DtG DO DEG

to -2.29 -29-80 : 3h. 48 67,20 4 19 124,40 : : 17 -q. 00

: 15 -36. uo 34.42 h'4.40 2: 3-3 1u7. hto 33 -13.10
-1.73 95.3u
a. -4. bS -43. UU -. 53.15 61.30 . (,
)s -. Sd -Ine4v
25 -5.75 -47.70 -32.04 5q. 50 -. 79 Bs 1 80 '42 -19.110
30 -50 9u -31.37 00.50 -2. Ob 77.00 -. 41 -d! 2. UO
. 4() -6.54 Sb 66 1. 14(i
-1.90 -SO: 420 -30.4u 11 -4*IU b #-. t) I -dd. 10
, O. IQ -el) bu -30.7b 61,40 -5.88 44.50 -967 -35. qo
., 30 70 -31.70 71.7u -7.43 -, 046 -44940
bO
. 70 -10.37 -74: .5b00u
-84.50 -32.40 94.140 -9.? 7 25.20 -1.22 : 54.90
. 80 -11.34 IeH. -30 b4 12.9u hi. 70
. 90 : 12 .52 114 hO -30.4b
140.90
-1l,
I. uu
-1.72
12: 96 -105: 80 -e5olq -17452 -? 024 -17, ov
. -19. bb 1,5u, ou -b. 60 -102, #
1.00 -11. d!3 -120.20 -21.63 -90*70

S PARAMETERS OF VERTICAL NPN TRANSISTOR IN COMMUN BASE CONFIGURATION

FREQUENCY sit s12 S21 S29!

MOD PHASE mon PHASE MOD PHASE HUD PHASE
GHZ Da bEG Ds DEG 06 UEG ud GEG
10 -14.38 (?I. qu 4414. 414 69140 29 -31.50 -8.90
. 15 41.94 t>5. bo . 46 102
. 20 12.69 t)9 . 14ý
4(1. (42 bFi su
-. -413.130 -. 13 -13.5o
. 25 -11.57 5 3. o () ,
bb 90
-1 .. e4 -50.3o .I ts -1 1.40
II*06 4G . 2k) -40 uu 2.01 -7u eo
. 30 30.70 51: 20 -eee -dI.340u0
. 40 -10.78 : 40.00 -4e. 91 -(iO. 6U
10.72 15 . '10 4v. 19? 4*42 -99 . 7(o %) 7u
. 50 - s(I 90
-11,12 d4. 4;u -5.90 -117.0v :
. 60 11.67 -4.70 7.07 -1311.20 -194 4n: 30
. -b. 21
.IcIýu : 12 25 -1u -149. %)o -1.42
2.0s
-5A*30
12: h2 9.45 -163. týv -ftt). ly u
, 40 175.3kk
0 12. otl -I u1- -2.0)9 -19.80
J. uo 12*22 -34.10 91 177.2u 34ý'Ie -yel3v
 160

1ý01.
CJSEF 117,r-
I-Efill LEFT F',-'ti ! ':', CiUl7E! FPIPS
IE FF,F:F.I-,EIEF'C'

IEEMI I EIATIEP; EMMEP HIDIH 2 I,iICFr. -Il'.:*)

QCiL TS=
FF.EC! I . C_
I CIO.C.9-710 -. 27 -8. z -Sro. 94 C16._7 4.42 1E.ZI-1.
A
1"'C'. CE,Cl
CCICI
- I"? - C",
-- ol -E
I -I(.
.s
I'.
'. 00. Ci(;
tj 21.t-14 --:.
-. 42
fiýcot.) . C,C:C, 21-:
1 67 .
aI I

I-, C101-1
IICICI.
6100.r-jc:cl -1.67 -SUL
7 C!C. C,QC-1 13IE-7 S3C
It'
7
It,. rl-, 2.
SOO. OCIC, 4 -
C. 12

ý-E!0. CjC,C-I 14 E15

-3.27 Sa E.
.Z!. 22 .

I F-IElCI r.,0Q 1-?I. 7 3-":,;? 3,'.ý,

Z; .,--
. Oti. 4
17
IaC.tcl . 1)(10 S4

1400. clc:cl _17'.7 12. ý_5 CI 2

.1600. CICIC-1
-12.85 -1224, t6
82.
-2-Se. .4
e -C'
.

.
74 3.1

1C-:
CIO C,EIC, cil 2 e.:1-o -IE. 25 E.S. -, Fl. E. ",

1
ý4 -9 q, 9 t4; *.:I
13 2 ,:., -I . E,
12 -1
. 'Ile
1.7 C.

-SOTH 1101.1EI-TEP
11SM LEFD P-11 ISURTED FRIPS
1,
)=t,lcc I=Ic S FPFf-.t-IEIEFS lllý'ISYIIEG

C
I' EMITIEFI, EhITTEP l,:IIITH 2 MICF.C46)
L,C,LTS= CL.,
F,F:Etil' I-,.
Fi=
FREQ IDA I I*IQ
s2 i
00.001 - 12. IS " C, IS'C,
. S4.5 I C, C-;ý4 I fin 5.0
. cl
Cl. C.QCi - 17 . .0 -E5.46 7,7 1ro- ý8. 15C 1 . -at.
-2
76.5 .
1C.; ý5 151. S
ES7 1 ;-. 2 56 14 1. cl
00. C'CCI 7 C,
I! CO. Octri I Cl. E. IE 5 57.5 E
CO. 11(qu -, C!.; E. - 17 54
. . -, ýC17'.I ":2
C0. CJ Z:,47. .
.6.1. 5.1:
llý 9 "'Ci. 2 E E.
clIt . C-1 -:
",:, I. IE 1%-- 4? ý- 0...
- I C-)1:1 ::. *.:I ..
-, . 10pl: 7
15-C,Q. C,E,C, SCI .a -
I OF-10.0 CIO -7 CIO
',
-5. '74 -4. SE; ';:17. -
I r'-C,C,. C.,
C,cl - 'C
el. 71 -1r. ý- -C.?. 9 5 C%-.
cl
.I 4 f. . -, I . c. .
"ý- #I. E.
.

71 E7 I 0-11
a-ItF. I--. .
E2.6 - 2-. 11 -52. ID

parameters o4 iE2M and 3E2M tran5jetors

as a ; unction ai 4"reciuency ancl bias.

TabLe I Sht. I o; IG
 161

*A.
711ý1
BEFIN LEFT F'.,*tl HiCILAIE!, PH F.C-
I , 7-il 9DEC
ý- F-FF.f: t-:El E F'I-- Ur

IES'l.,, I Et-:171EF, ES:IITEP I-III17H 2 r-IICFCitl$)

65 C.L.IFF'EI;T 1,, Fl= 2
F PEC! $11 !.- 12
C-C-1. 3 5 OIL.
7
i cl Cl.
Oct. Lter.,
0C.I. (IC-0 -56. " -Ea. E3 53.5
C0. C,C.CD 2,94.5 -Z. I. 4S 51.0-. C'4 11 c.
28 -C. ", !. E, 51.0
C-00. Oclu -7.22 2el E,
-al.?.

119.7?
6E U. LIC-10 -SI
1.1

. 13 _26ý. 9
.
2 -f "5 ;?
7 OCf.000 -$. Ze .
C cli -EEC. 11: 1
1 57 4.69 7.2
SOO.CIE-1f) -1 08 ZE 2. C. -IS. 12 4.11 W"
62.1
I C.ecl. CIC-C, 256.9 17. 4 ýo 51
- e5 . 31. 10 79.2 4
12 0 (1. (1E'Cf I
C-50.2 - 16. 'ý9 46.4 I. I.B.5 7%;,.C!
140cf. OCICI -12. SE, 2:45.7 47. E. -7.64 -44.1
-16.18 7ý 6 8. z, 7., c:.C 3145-.0
clcici - 1,ý IE.I 24E. v . 41
. ý -15.71 -312.7 -. _296.4 40 312.3
14.06 I"%- 4'".
9CK4. CICý-4
- C-,:.7 ri
228.2
-
14.e-0 ý C, S-3 59.1
-14.48 -14. E7 45., + 1-1 5S.
.4 -el. -73 -50.3

,201h MOUMBEP
ISAM LEAD P,,ti I-, n.LFilEI, PAIP,",
1.1=1)cE: I=IE: S F,FF.F.t-IETEF.,,:IN
1,19DEG
1EM(l EPIMEP; VAITER WIDTH 2 MIUCINS)
S CUPPENT NA= 4
FREG sli S12 sel
1 EiEi C,Cir_I C4 '2 2
. -62.4 -2-7.95 46.4 S9 1F.O. 5
CE. _:3.. jZ -20 9
-75. C; -;+7. S 10.1, C, ice. 2
-4 4-ki
. _-Z
. g
.

21,00. -1,3.57 264.2 4 C1

-24. 4.76 9 1. ID -! 5. $Cf _Z4 .
-; EIEi. -1 4. SE, 258.1 - Z4.17 46.9 a. ý_53
ICOCI.ÜC]k--l -15. 'Ire*? M. 7 -6 . 2ý4
1.
EI ü. Cl 56 M4
clCi
cl 59 11C',
7.
. -6.1? 1 33-14
7 C*ü.CICIC, ý4
7
0 ei. C,Cý
E., EI -EC. 44 Es 6 7.0 ''
,9c-tC-. C,C,f2 J C-4
-
1,:, ":, "# .
C4 7' 75 ,:,
27.4
. - .
ý.
%. , . -4.05 C-3.2 --. tF.f,
1£EICi.0 OL-4 a, '3 Z 1 ID5 SE 2 _ 5.0 16 E.1.6
1 ,
56 1
1 E', CtC,
C,. 1-- C5. .
1 EAD C-Ci
L', -7.10 0 .:,?
"1
C'13' 7
. -6 . 87 Cjý
4
.
1 1EI.üC.,13 -15.74
C. 5. ;7 ct 1 71
-E:.,: 5 47. 't-
1S-59. -1 C Ze. 5.
1c ., .
49 .5 -8.7f - 0:,,(17
.
-E . 87 45.5 -9.11 --T.

Sht. 2 o; IG
 162

LýU::, F-11 ISOLPIE!, PRIF.I.-F.

I. FF.F:F:.',!ETEPS' Itt : 11:ýI, EC
-ILE
IE.Ef: il E1.111EP; ENATITEF HIDIH 2 M1C:Fr..,11',
kJOL aS=: ZI CLIFFEM NFi= -; )
E.
F FEC Sal
1CO. GE-0 -10.62 -9,3-. 7 a64 rj
IzE,. C.,
CIO -12.6c. I Ict. E, -2f : 5.5
.3 7.54 S4.1ý ý4
t.) Cf C,
-I
.QC. E-8
GCf C,(, -ý' -1 4.4-:, ' Z26.1 25. S6 S91 71Z, 77.77
. -7.14 4
0,ILI - 14. ?7 219.58 2.4 5 46. C-1C,
.Qc; - -7. E.
c1c1c, -15.24 C-1V.. 56 50. -G. 7 6ý1",
1.0cf.
600 Cie* 2
-2.7 .6 -7. ?6% `*2e.. t
- -22-ii-Ec. -307.2 -4., '6 -Z-01. ý,. 1 ý4 Z2
00.0613 -15.14 207, -21-, . 58 5a. e, -5.72 Z%-.
.1", E*7
-c C.E. 2 4ý4. J.
-7.11 14 z2c,
900. C-100 -14.61 16- 19.91 E. 77. ý41 Z. 7- 11
.6- 316.1
1000.000 198.7 - 19 .t21! 3.5
-14.55 -8.9,: 1 44.5. -9.67
1E (I C., El0 U, -14.06 19 9. ":; - IS. 7 Fic, FJ 39.6
. -I-I - -9.74 -54.1
1400. C1,00 -1 L". 7*,ri c'100.7 -17.66 52.1 ý4 E) -
2100.17.1
-11.41 . -9 . 89 3C-12
.2
1660.000 -14.05 E. _308 -12. da _321.1 I Cf. L-s- 29.. 3
c-;11 .
1 ISC,cl -1.. t6 P.?7.2_ -12 Fi 51 rJ 62 39. El I cl 2z, -ý14.6
.c . .
1 9ý49.9 S-9 -1 3.06 19,". E4 - 15.62 "56.0 13.0ý4 Fj 7,9
. .

20TH NOM-11EP IS75

BERM LEAD P-11 ISOLATED PAIRS
kl=QCE I=IE: S PRPRIETERS IN 1I1:y1!EG

IEEMI1* ENITIER; EMITTEP WIDTH 2 MCF-011S)

VOLTS= CUPF-EhT P,, Fi= 8
FFEE, S11 sla S21
100.000 - -11.51 - 117.3 2,
- G1.03.235 9ý4. a
-6.,. 9 -23. 'c',
15c.i.. 000 -12.54 2,27.2 - 217.1 42. CS4.4
0012, 218.1 -7.41 -24 . It
-cic'.
co. - 12. IS8 - 26. S,ci 4 cl. 67
000 -1.1B. 1*? 3.17
1 42.6 -24. ?
4--oo. 0.7 27,
400. CICO -1: ý. 41 204.1 4 el. . . -ES4.6
-Z4.43
Z!F)o. 000 44 201.1 1
15
48.0
. -2.46 61.5 - e,. ect -1,
4.1
-1 -22.3 '54.7
-S. 12
1: 1: 21
600.000 -1*:...; 3 197.4 _22.16 _Z,1121.6 _E. 06 _S11.4 ". 4. "1 3.17. G
7110.000 - 1,::. 20 V?5.2 1. CZ
800. cicit., - 13.41 1? 1-5 45. cl
900. (100 I I?Ell S-C. O. 01 Icl.
_12.72 4S4.5 ý,ý 21
- lo. 45 05. C5
I c,c0cI cl -12.7Z, 8-16: 8 !51.1 :ýa . a. -I I cfeýz I 15:
-11.1-ý E;
I Eric.,- CIEW -1 81 47.? -Ia. E-5 .
1400. C:EIO -12.21 193. (1 1 C, 49.7 r
ro
1ECO. 0(10 -1 a. 51 1 c-1 -1; -. ': 'I Co.7
1S-00. CIC0 -1 Z'. E44 16- C-I
1.61 11-2-17.4 - It.. 23 7 I- 11 -- cl
1. -IC! 3 . ý4

Sht. 3 o4 1rc
 163

1101.1EI-II-EP
1.S7
'FEAN LEP:, F-1; ISCUIEL PAIPS
I=lr S PFFE-METEPSIN DI! I
YIIEG
IE'r-l'ttl Ef-II11EFI; EMI7TEF. WIIIIH 2 MICRUG)
VÜLTE- 51 CUFF.EliT MR= 10
FFEC' .
5111
C, C, c) ,'ý21 ""22
1C--, . -10.79 -144.2
1to. 1. -;icici -! 1. C-11 ITDc .c Ll ý4
n. cl
-11.01 ISC, 10
EIci k.1ür;1 -11. Ell
1 1 S4. 37.9
15: . -2.2E. 61.2 11
4c.le. -11.16 7 4. S.O 50. IS
-11.12 -11.78 %306.1
-C-. 87, 4E. 2 12.07 2? S..,
-11.06 3'19. C, 5, Cj 12 r'.7 an. I
-
-IC.1.93 -10.1 :3. E9.9
C- I S6.6 -11.62 U. _7
i,e5.1 -19.2E 41. c- - 12'. 64 2-4.1
1Ociel.WO Efo 27C.. 4
-11?. 21 3.91 24.7 1%S 19
-ý42
12190.ccici Is ":,C-1 2w. 3 12. i e.
-Ic, I
1-4EI0. C1C1C1 .
6.09
CiEICI 187'. e. -12.64
16.67 6 12.7?
-I Ci.(q.-. 42.7 6.25 a -
74 16.4 4 42.8 -16.13 41.2 0* -117.7

Sht. .4 of IG
TEXT CUT

OFF IN

ORIGINAL
 164

ýVlf 1 0. U.
I: EF,I-l LEFI! P.-It
it: I, EC

-IF-EV1:" El-:.;IIE. F; ENT11EF, t-jjC. FC.j-.S)

L,C!Ll c C.L;F.F,Etil NF,=
F FEC: Sl I
00. OU,
C. ,7 2.
-1 Cf.
Q. CA-1.1 -1.14
C, I
%:
_C,
clso.COO -j: 7 .4
-1 .17 17
0 -1" Iz cl 2. IT 1 IC ;
is 4's' cf
7,00. C-00 -2. t_.e- 77S.. rr
SOO.(100 --
-, . 07 2' 70.1 -11. 6
'.:, (164. 1"" 1. C14
, .
COO . o:4 ý1621. E, - 11
.E '26 . -0 cv c
1(Alcl. QCIL" 11 C,.,
E. 3. ec: 72.
,7
C-515.E- . C- - 1.04
12: 00. cl (10 -4.14 E46.1E -I12 16.3 7'. 6
-1
14 C'El.C,Oto -4.15 e. -11 55 15 E, EE, r,,S. 4
.7 . - 51
1 E.('[1. C,(Q -4.27 E. -II. t- 12,7e '4 5,
McAl. clrýcl 0-II. E-4 I C. 5- Cl
65
S4'ý"?, ý49?
63 220. C. -II. E.S e. 9 S4 ý Cs
-4.65 .6 4

2 CITH hCASPSEP
1EN-1 LEAD P.11 ISOLF17H, PAIRS
I.-Qrx I=IrS FFRENVERIS M Dl: yl; EC

"E21"? EN
AT1EF; EMITTEP WIDTH 2 MICF.C111S)
MOO Sf CLIPPENT MA= I
FFE0 C11 S162-1 S21
C1
OC' 77 -22.4 IC
-ac". 05 76.5 11 C,21EI. C.
C1 . 1 -. act
2CC1. C10c,
-1.031:a -31.? -21.03,1 C. 72 1A. I
-1. 0-13 C.
ý4.
-1

a7 -2 "4
500. C1 CH -2. C17 iF" t. . e- -Iz.. F7 ec.l. 1 _711110 11 !. E.
c,c1c:. 29 . -,:: - 7-1 1.1C. 11
22f-,E7 14. Z
44 .
VC.. c
700.0 0_1 -3.71 C,17, -1.;. UE. C, 45 -z .
ý-CICI. CICIC, -4. IL 249Z, _1* -, e-,4 4, CIO. co :02.
C,0. (10 C1 26 2_iE. C, 67 -C ?
*C.
-1 25.4 2
C,C C' ro C-0 61 co IS ZS 5. E.
6. E12.;,. 4 E.
. Z
cl 0. C;l C.,Q 72 222. c? I,: ý-S* 1:2. E5
.5I. C.
C-0 M I.:,;

c 1 -7.70
C.

E.
i C. e

isou. C-10C) Zia. 1 1:1 . r-C 17'. C14

"J ý4 ý4 9ý
ý'5 206. e-1 2'! E
-4. 17.7 -I. 7
CE: 4e.. C, E4
Q*c-

Sht. 5 0-F1(
 165

crlh I: r, l.,E.I-'"-"EF' 11-7,'

TEFIN LEF.11 P, *t
I =.I E: *I- FF-F*f-T!
E.IF-F.S :11.: SPI-F-C

Et- MEF; EIATTEP wiriH E: I-'ICF: C-tl$)

C.L:F:FEI. " t-:Fi=
FFEC
Zz. 72. jr, C. 4

-; L1
ýA` -7
Els,.
54
ZOO. C10-1 . 67
C-4'
1
11
"Cit.D. Lic-0
T. 1
ZEE
(f
-17.42
E;7
' :a I
1-4
Cit. 2 S7.7
60cl. C'(10 ý4. Cie
- -c. 61.7 &':
33
C.. .
7 CO. C,0.1 I;?. (10 -26?. 110 c "C? C'.69.
*:-,. I Iz. Cl Z, 2-1-Bl. C.
MCI. CIC-Cl C7 CI-Irl-,
E. E2 871.v '. (-,'I C".5o.
222.4 E7. --C.
? Ocl. 00C, 65 ý
-'S .t0 CU2. C,
1COO. CIVEJ 21C... 4 -lE . 75 27.1 10 7 ?. 9.11 -IS 1. C,
I MI. C:C.r-! 211.7 It. 7*,% E5. 77 `72. C.
12 _E116t
14 C,Q. C,(10 20 2 E-7 6f
E. ". .E '67 CVC. (
E-CO.CICICI 1 -29ý4.1
-9.7.0 6.7
20cl. 1 15. CIS 28, CIE
E;Co. C, .1 . 73 56. G -? . 9--
2'
E . ý-'
ES 1S5. So 14.87'
-4. rl_,
c",.I -. 09 51 .S -I (I. CIS, -ý-? . E.

*Cfll-i 11CII.
CE.. Sti-EP IS7!
TEAM LEAD P-41 I"-OLF!TE! i PPIPO
ý'=Vrx lhý'N: ýDEG
I=IE S F"FSTIZIERS

*,M, ':,,- Et-IITI EP; EMIITER WIDTH Z hjCF. C--tr:

-.)
C
IUOLIS!'- . CUFFEIIIT t-:P,= a
FF.EQ Sil S12 0--^C. I c llý ý.%
64. w G.Am
I Cie. Cleo -011.5 -211:
1.?I 146.7
Owe.ClOcl 3 -22. I'S.
-S
I I S". cl
0 0. C'
CC, aq. C ý4 a Ile
C,
0
.z:
C,O. CIOCI cl-It"a.6 -. a e.
55.
17
4C,Cl. CIOQ 2138.7 --4
1 CIS. *6
c"CIC'. ZEE:.Cl C.
220.7 r
7 Co..C,(10 -4. 2 14. -ý-
',-"CC'. C,C,C, .7C, -z'-Z . et. 11 C.2. "':- A. 'r-Z: ý.67 7
204.7 7 7e 7' IQ.
';:(i Ci. Cl(ic'
I s. '!I. t
2C. E. 21 7E.
L clcl. 1.10. Cl 11:16.7 17 S 37 C, 71.1
-: 00. C'00 I -, Z, 7,2 65.71
6 E,C,. C,C,C' ZO 5
1E"Co.C*C' IS-9. -10.1 2,; 7.
i99. . "9 40.4
-I It .

Sht. G a; fG
 166

! Efl-l LE.F.Ii P,.,Ii - FF; IPS

ITSCILVIE:
IEX IN : -!., I-E.G

V Ef, IIIEF; ENI 11 EFk1 : 11H2 1-11

CTCttS
CL.,
PF.E.111 1-:
Fl=
FFE C ': "
I-
0 (.0 -I. 4" l-
.;.
". 4
"-
f4.
£. J....
r -- t.
. ....
re
s's'.
C,10.
f-I -0
27 2 I. 1: 1C?... -.:; "t..
2C.O. C,00 1 1:.?.. jIC., -1':. -7
CIO
cir-1.11 27 I..
JI., _'. _. *
CICI.C.CO P rL
_(...
_" .. -
I'
-

Ir ._II _. _i
Z.7,r,
27 f' E.
ell
ý4 .
-" . 4::
Zt. 2.7
.-. I-- 7-1
-.. -. ., 0 -. , 9. %-

I rjclo. 0 Ou _c _J.
IC-16.rl
_', _p
-4.S3 *-13.77
Z.C0r: QCl -..
1.4 ict'. 7 -1.03 te. "i -1 ýV 11:. Q,
-. . g- %0,

COO ,
-.
11 Ir$11.e -I ...
-2. ict E-e. 6
I C.00. (I Cis) -5. C-48 184.2 -17.37 - S.CIO.5 07 -4 te
i COC-1. -4.91 I El. 71-j Ci1 1.
-16.1- r S:

201ý; f-r.'l.'El-TEP 1ý-7,9,

':.EF,I-l LEFT PJs ISOLPTU. FPIPS
I=IE: S FFFFJiE.-iEP,,'-* lh llil-EC

ZEU! (I*ý EMMEP; EMITTEP V11,714 2 f-'.ICF. C-IIS)

14
FREC C 1 c..
12
0.0Ci -4.
1'.--5.
14.
76 cf. C:
C. I S.I. E. -El' "I"::
T4 8 .2 .
EiC, 112:
1.2 :. 9 ES 72. (
cicl cl1.1 1:. C, .:.4., -I. EQ
(1e.171 1E'..i. 2 -"; E...4: -:: j1.
. 1,10
V-&-. "
4S. 17.
-
ý-üü. 0 (ICI 4.4
- 12. *ý-2- ra-Zr; . so

C11
NO. I
I I'D 28
.
1Sc,0. C,C,CI S 17.
S1?9.911.:
-. 1.1.:-: Oto
1 49
-4. 5111-71. -16.78 ' S-4.8 12. CEO
.1

Sht. 7 o; iG
 167

IEN-1 LE.F.T! P/rl : C-CILFIIEII FF!IF-.'-:

Eil U!k:ý' Et,.IIIE. F*- EMIIIEF. WIBIH -Cl

I S- -
C.LiF-.
r-Et, l t,,Fi= 21
.5
FFEC! 11 w* ' 17w
,. C G.
CO. C'00 -4.11 1?. 4
C,7 I* t-:' C. -27 ".::- E6. C.
. .
J.c 4.4
14.7
ClCf.Ctc-ct
.Z:
97
500. CICIL,
CC-Cl.I: Clci I C -2': *. E-7 _3 11 .7 -i-. 74 SI 42.1
cl
C. Cl
0 .: El -E2. 20
f 4o'
. .5
SC.Q. 00
9 ClCl. ClCl 17 - 101."0 EI 'r.. -;-
- 10. E'4 ? OE:-T
.
I CIOC-1. CICIC, c1c, -21 E5
1Lee. 00cl 77 Z7 tie. 2 -I T7
11;el0. cl0 ci 7Z 17'c':.E. C'5' ot. -11 , E6 S6. I
ri .0 (1C-1 178. C, -15.415 -320.0 7 77
0. I.DC,C-1 55 176.9 -,5.6 41,
-17.95 -1!,. 74 C- "o
r-
5: 15.9 23
-17. C- 5,E.. 1 C2 C..

ECITH I-ILAIEWT-EP197,c,
BEAN LEFill P/*Ii PEGLAIED PAIFS
1=lr- S FPFRETEF-C III YI,EC
-'t'!
SEEP ',';.'-Et: ITTER; ENITTER WIDTH 2
QOLTS= .5 CUFFEW 1.: Fi= 27
FFECi sle C'21 I;
1 c-Ci.c 111Q-1 V5 -29.3 1 '.
" -. -"
(". .--
.; '(
.-

I llý.0. ClQC' IE5.4 -Eo,. aL;

'4
41 -27. E_ý,
E. cl
. 116.
.; cc'. C,C, C' Z: Z 1 42. s
-L. IIEI ý,1.*- I C. ale
-25.2E. 0'.2.5 10
500.000 54
-2-;. t' 1
EIDO. CCU
7 M. CiOC, -E2. S1 47.5
I"
,:.C, Cl. C,(I cl 179.2 - 22. Q!! as. C,
'Kic'. CICIO Z.. '.. 1"FE.4 -a I. Ic,15 "s 47, i S. E3 271.
rJE16.C,CICI Z:. 45 Is 50. rD E4 E
Z.
12,CIO. C4

14C.C. ClOct I "",7.6 C', ot.

. Z.- 1. .,
1ticir-1. CtQcl -,:. Z,I 1777.1 1 :-!. E,S fit.. C. 31
ri 0. (I (I C-1
1'12: 176.1 1e'. C-12 55.2p C13 1,17 C
1999.91: 11il 1 17. -; C, 55. E. 56-4 -JE"
- .6

sht. 8 oý f C-D
 168

r2 NCIU MBSP 197!,

ISAWLEA D P, N ISCILME1, PAIPS -CITH
I= Ir- cl: PARRIETEPS IN

IELM fl EPIIIIE P; Ef-1.11TEPWIM H2 MIU MS)

C,LIS CL,,
F.F:EtiT t,,,
ri=
FREC! .5
s CE1
I Ocl. I 2:2 -6.4 -`(-7 . 162 '?I. 3 171.8 U,
. 1:1,o . -4.0
150.000 -. -% 4 I
-S, IS _31 167.4 06 6,
2 00. 0.7 6 - Z-fS. 4 -6. w
4.27 ic
-27'. E.E GI. 8. 74;;
iE2 -16.8 .
-. 72 -21.2 -2-I. S3 -79.7 S. E-7 14$. Ci C-0 -IC.
500. CIC10 86 ý35. ,- `2.7 C,
-C. 7. 52 143. E -. 02
".
iticl. C1110 -1.06 Z ,;-1.5 -C-1.12 24 1
Z,2 (1
-ES6.6 %. -1 - 08 C,.S
00.000 -1.42 -'C. Cl. I '.:. -290.2 1: CIO -229.1 -1.42 S3ý4. L)
--'ecl. 000 -2. Elo :322.7 - 19. Cis 6.6.6 :S -2 3-S.
-3. -1 . 96
? 00.000 -2.17' 31S.. Fj - le. ": 1 E.S. 9 2.39 11
-1214 -1
Cl(10
CIE110. -3. cis 14.0 - Vcý.Efa 61.6 1.70 114.1
a C, C,C171 -31.6-5
If-: - -2. S2 -2E .E
-53.5 -1-7.12 IS. 1. C13 105.3 -2. C-5 6
14rjo. rýfCA. Ci -4.34 Ze,
L"tI. 4 -16., " Ci "I'll CIO.6 E4
C,C,cf. 1:1(1Cl -Z; 2.0
-5. al
C- 2:9 -f. El - 16.2 5 --, t I A- 5 -. 46 -268.5 -3 . 96 3C1.6 2
00. C,(10
1-12 CS14.1 46 47.5 .
5.5 ( -. 68 8
.5-2 -3.83
1999. 289.4 45.6
-6.73 -1.31 1 79.8-

2cilFi rio(Erir-Ep
LF-All P/fi ISOLPIED PFIIPS
V=tIrr I=I rS F'FiPf4-IETEF.
S IN DI-9DEG

IE'r-'Mfl EI-IITTEP; EMITTER WIDTH 2 MIC KtfIS)

VOLIS= 3 CUFPENT 1-1Fi= 1
FF:EL, Si l s12 sei .E22
100. EieIC-1 -. 59 _G. s _Z,9. c16 C,1 'G 11.03 169.5 c13
1ZEI. Elclei -. 71 -22.45 Ei 1 Cl. e15 164.4 .
NE3 - . 106
ZOO. cilDcl -. SI -: i1C1.
E.CI Vw.
.7 1 e. 1-4 159.5
7 7. ES
400. cie-Z1 -i. -ýS -, l'ý 25. Z:' 75.9 ?. 71
,'
5C1C1.
660. C.,
CIC, 31ý4. Z, -22. EIZ, E. G3 -21%Ct.7
-1. "-4 3,3 7, cl
C .
CO -Z'. 75 `:OS. 6 29 62.7 7.4: 3 -24: 3.2 E6
9 ei E.4 60.2 7 .
1EiC,C, Ctf.t 17 290.6 -a. c
.0 -119.: 37 E.e. 1 6.31 1077.0
120ýt. CICII7i -5.95 -69.3 -le. ES 51-1.0 5.36 '98-9 -a. 81
14E;,rj. CCIL -E--. 24 1 -1 :D. 02 92.6
16eci. C'Eic, A - -c. 25 2,:,Ei. 1-c. ',.
': 1 Z' e
', 't
-44
ýý-33 -273.5
-. s. 2,re
-4. ýtý4
1SKI. Ctoci 42 a( 3 E. E-9
119919. 271.4 8- 3
-16. CZ 76 6
. - -Z; . 29

Sht. IG
.
 169

cklh 11MEITER 1ý-7!;

ltfifi LF-Fill P, *tf AS-CILFiTEI*PFITPS
I=lr- FF:PRAIEFIE. IN : tltl, EC

IEEI,! I l El-IIIIEN E1,111TER,t-47iTH 2 MICPCIIIS)

CUFT-.,EhT tiFi= a
FF.EQ 12 se l
I CIO.0 Oo -1.14 -12.1 E. I "'.. 64 IE6.7 -. 01
cl. clC11
1.05 -I. L-, -1 -7.(I - 1:ýCI.ID9 - -,"e..5 1C-.ýý7 160.3 -. 23. -9. a.,
ac"Cl. Cica-I -1.54 -22.2 1. :8 79.7 1"';,
5.1 154.2 31 .
. ý
-,cAl. CICIO -2. it-n - -,: 2. S, - E,8.5 E. 70.3 14.43 14':
. -. 7 '-;, -17.
0. ClC,Ci 67 1119. C, -2-6. CI-2 -(E. S 19 1 ae I i. I
-,,,
500. clc;cl Z,. Cal
5 CE.: C, 1 1 C-7.0 6. Fj
. -1
6cici. C-160 95 S06. C, -2S15. 12.1 Q, -2-io. 1 -Z.;?
7 C,C, ElCi -4-70, 299. z -22.6-7 6-4.1 1
800. CIOL-1 91 a S4,
-:. 4
-21.72 62. e Ic. 2 2 21. V. I. Q
-6.41 2E'7. E -21.12 E1.0 ý4.79 101.8. 5 Lo. L;
C,Oci -7.60 201.8 -EO. 96 61.8 S-77 9e. 9 32 a
EICIO -8.45 -06. c -F-Ci. 17 r.-5.8 -. 59 91
-9 -4.72 s
10C., C! c0 -9.135 26 7. 1"ý.42- L:-7.3 44 C6.7 4
. 261 . E, -Ic
160CIAIC-10 -10.37 3 :Z5 E,
C. ClOci 259. C-' -1.:-,. CI2 11,56.5 4. ý .
JER. I.
Q9 ý4. '.: I ý4
-IlAnc. ' E-6 76.6 -5.70 '32.6.0
2:50.9 -17.49 Z6.6 3.78 772.

KAM LEAD P/M I-SOLPIED PAIRS

Q=Vr.r. I=I ES PPPFPETEFS IN I'LiDEG

(I EMITIEP; EMITTER WIDI H2 MCF011S)

IMI
VOLTS= 3 CUPPEfil NA= 4
FREC, Si l sla C^
C.j S22
lrjrj. ciciri . -2.1. -:; -is. I -sq. -7-s- ý40-9 1.9.73 Q 161.7 1 -e.. o
150. CICICI -2.4t-:, -25.3 - '2v
3. E.S 76.2 1,53.4
200. C100 -2. -129 -11*1121 -:
1. '. -ý: 2.
e 11:,r.
0 77.5 11S. 84 14'1ý.6 -. 67
cl CIO - 73.1 -14.5
1,100.
zo -, ". 92 -4 (. E -'310.24 1?. 66 Is". 3 C.
84 ': 02.7 E 7( Cc.
.
500.000 -5.74 2,?4.7 -Ee. EE: EC. 3 IZ. 2-6 110%. 1 c
6rdo. C, 6 u, . C,c. 3 1
700-C -7. 2,80.3 E4. E, '7
3 65- E. IZ*- 09 -E57 .4
4 " E.6 2 . ot
Z3.51 E-6.:21 11.78
800.cicicl -el.96 Ej 22.7,9 65. E.
r4
":.
I _7
": 0.7.
C-65. 11.113 Sa. E.- 4,
25?. 5 66. - ý
-22.3-3 90.5 -g,. 11, - Z'?. 6
12 ClCi.Cl(If'l -11.1; 2 C'nij. 4 61.7 S. E-7 C-4.6
14 C'rj C,EtC, 11 &D 245.2 64.0 a
. - ..:.; SO 6-1 76
16C.'el. cloo -12.7c, 241.1 1-?. 0 -2? 5. 6.04 0 .
1E 5. I IS. 79 6-31.6 77
I ý-S4ý4. S199
63.1 ,I 0
45 L;F;. 7 -e. 5.--, Z:. 2
1.

- Sh-t. 10 o; iG
 170

ca-Clti
! 'EFit-I LEAD P, 11 I SCILAIED PAIPS
Q=VCE I=I E: S FKPI-VEPI-ý S 1h lil-pliEG

lEr---h.( l EMMER; EMITTER NIDTH 2 MICT ONS)

QOLTS- 21- 3 CUFFEW MR= E. .
F PE C! Si l sle I
I CIO.CIC-10 -"::. 55 -2-7.6 -14-8.819 Els. C. 1 Zý 3
I "o. C,CIO -4. EA -,: 7.77 -'Z4. Ci1 718.4 210.63 142.9 -. 62 -13.4
71.. '.!' 1ý,. 7174 112A. a
11,00.01110 -6.44 7 E. 1El. 1-710 121.4 -1.77
.6 68 e. f.
4010.000 -7.64 'C-E'6.e -, E9. C.S. - I 1 Ill. w
. 27,c:. 5? 71-6 14. 1(14.5 Z;37 . C,
000 -:?. 71 4 -. -64 -: 4.01
50cl. 1, -. 48 r-6E.. 7
600. CICAL-11-9- 811 2-71.4 C4
6
700.000 -1 El. 58 12.037 -26,51.8
eloo. ciclo -11.78 255.8 -24.47 71.5 1c'. 96 90.7 -4.49 SZ;E,
248.1 70.9 1G. 2:6 85.9 .
2-334.4
,9(jo. L)(jO -23.40 -4.45
lrjrdFj. ciclrj -12.: E.r- 241.4 -C. 3. C'5 7G--.7 9. [19 84.2 -5.06 -cl. gl:. 9
1? c1cl.OCIO Q 234.7 -22.15 66.9 7.71 79 .I EG" 5
"29.1: 20
C,.0 El0 - 3.,. o c-
1.1ED
4 - 8,9
- 68. r.71 6-, S'9 7Z. 2 49 2
1600. CICIO - 190 226.3 -Eo. 19 -292.0 C12 -288.3 -6.06
18-00. Cleo -14.13 220.0 -19.01 67.7 4.40 6-1.5 -5.83 ;.":;o . Erf
IS-199.999 -14.09 211.0 66. -( 2, IEA. 6 -6.10 -'30.4
.45

20TH tMEMBER IS75

BEAM LEAD P/11 IS OLATEr PAIPS
t)=Lpcr- I=j r6 SU RRIEUPS * Itl DEPPEG

JEEM11 EMITTER; EMITTER WIDTH 2 MICUMS)

VOLTSý :3 CUFREliT MA= 8-
FFEQ I 11
1%- sla S21 22
93, 65 20.27 140.2 %CWI
70
leo. cic-0 -5.59 -42.4 -SS. - -3' -. -V.. 8
50. ElCIO -7.14 -54.7 -36.22 64.1 1S. 66 7. c-
1&---, -1.14 -11.4
(lei. ci CIO 1.4 16. ý43 11: 8.4
-8.49 -6,5.1 -34.37 -1.47 - 1;2.
ri cl. o Fj0
L-, - 10.77 2 'S,1.1 ":'---I1 64. Et 14.14 1C17.9 14. a:
400. C-1 le-O -12.171ý4 23.4
227', -SO. 75 71. Q. 11.74 1C, -16. ()
5rjo. rjecl 267.2 - 28. 'c8 71-f. 8 1El. ('s 96.1
600. C,CIO -14.26 2-61.4 -26. S6 -283.1 S. 56 -267.1
700. ClCIO -14.68 254. S4 - 26. C,C -2: S2. S 7.40 -, -'-'7Ei E, Cs, :107
.
E.
'ecl. -15.51 2:47.4 -24. E!2 777.6 6.06 817.2 57
CtEll)
19rdc-o. -15.44 240.0 -23. E-9 7 6.2 5- "e. 3 -3-116
1Cleo. etclo -15.76 2 '33.3 -22.94 77.1 4. SO M. 1 -114.S9 C,
ý2:.'4-.5 8 -10.2 '%,
ý. 02 77- 9 7
EE10.0 ciCl -15-15 -4.133
226.3 -C-0.89 71.4 1 C;L, 74.8. -4., 37
1400.000 -15-01 -E7.9
I C.I&C. C,(10 -15.27, 223.8 -20.07 -ESS4.6 E-2 -288.4 -4.9E, -: rj :-

1999.999 -14.78 209.1 -18.25 61r. 8 -10 65,4 02

.3

sht. I ia
 171

ca-0-THNOI.ENBEP IIS-75
HAM LEfir, P/h I,'-.CUiIE: PPIPS
1)=I, lr-E: I=IF- ',-' FF.P.PPIVERSIN rjivr: EG

IEZEM1 EMITIEF; EMITTEF WITj7H 2 MICRONS)

VOLTSf CUF,.
F.EliT NFI= 10
FPEL' sit S12 Sel
Co. C. 1Co -9.42 -Q. 37 ow-7.4 16.97 124.2- --

150. CIC-ICI -11.40 -W. 04 63.31 14.43 113.1

2:ciF). 0 cl0 -52.7.11 67. 1 3 106.4
";. Oe
.3 C--*
64. a 10 91?. cl 1. ý-2
400. CICK-1 6. 7S e. '.74
6. E-4 94.1
- -I I. 'r-,
. 2.17 -13.6
2 CICIO -c-'4 239. -Z'e. 62 M2 4.99 90.5 '144. S4
.15EIC-f.
E GEt. ric-tri a0 3-.S -27'. S'9 -28 1. 'S 3.59 -2 72'. 4
7e Cl. CiCI:71. 228.2 -L6.79 -279.6 2.47
-275.7
ec-fEl.ricio -I 1--,.28 221.8 25.4 1 79.9 Lao f32.3 Z:7 41.2
90 ct. CICICI -15. il .0 2105.5 24.10 -19.6 58 M4 -*'. IS. 1: ý4 .',
ZOO -15. 56 ' 21 9.3 . .
1CIEICI. -23.53 48. 7.2
142,00.Elcle -1 A. 71 210.7 -22.61 (2.8 -1.63 72.2 3.75 -25. S
14C0. CCIL-, -14.40 -2.70 68.6 -3. ?a -27.3
16e0. ecict -14.64 2'0 7.2 -20059 -287.6 -1. w 3,t 0.6
1eiDe. 000 -14.14 202.0 -19.37 72.1 E8.9 -4.27 3';: Cl. 4
159te. 999 -1: 3.70 1515.3 -le. 57 70.6 -5.06 57.9 -4.54 -31.1

5ht. fZ o; IG
 172
Z*
iffit-i LEfill FVI'. I'SCUIElf FATFS
Q=Qct I=Irr. 'I-* PF.F.F.t-EcIEFS M M: 91EC

;SE2E, ' EMME. F.; EhIITEF WIDIH 2: tilCfC. tl,. )

I f. t. ".
FFE(l '1.

I C,Cl. cl1) 4.c c:CIE., - 5. C.

ccl. (to 17'. Z7.; 4. OLD
-. 12 -E-1.i.
-211. Ic. _Z5. 7,G -., SCI -IC:. r.,
cloo S C,
S5 (I S4 SI. -41 I Z.
-1.1'5 -1.10 3S7
(101) 40 17 Z. 64 -as 7.2 .0
IE-CICI.
7c Cf.clcis) 17.2-a -244. e
-1 ell 31.
(jr-10
ECIC-1. a ý;
2 C,
S-clo.C'Eic, 277.0 -16.1? al. -Ic". 1.12 icli. C!
i CIC10.ricle, c'-70.3 G
-It., . C-, 18 96. C.
IS, .9
200. CIM -100. ý- -15.61L S.2. ":
ý4
14 G0.0 EJ-71 -: 3.57 05I. C. -I. 1XI 79.6 3 17 . 3-
1E-ClEl.C.CIO -2.1-7 -F-87. I-E- 31S. 9
1elCo. 0 CIO -4. -2b 77 C-7 3,12.9
IS-199.999 -4. $4 -15.4: 3 24.7 4.
-3.49 -!; Cl. I

SC!TH tIMENISP 197Z.

BERI LEAD P-41 ISCIPTED PPIFS
kl=i)c:r. I=I ES PFFRETEP S IN ITYLEG

SEEII (S. EI-1171EF.; EMITTER WIDTH2 MCF-Cillq)

VOLTS'2 3 CLIFREIIT MR= I
FREQ Si l s la 14,121 S22 0: f.
160. ocio 64 -16.1 196 77.9 11 . ,;ý; 166.5
15 0. cl(%--1 -. 81 -22.6 -a-t. 45 ell-0 11.11 16(!. 1 -7 . 3
--1? -11.2
"OCI.
C- C-00 -.? 3 -29.8 -21E-.42 74. '1 I Cf.S6 ý4
-. 319
",
4. W. CEO -1.4c, -43.9 -2, -. 26 E-6.9 1Cf.Sa
ý 143.6 -. 79
4CIO. CICIO S-1. 0
IC
1W..
" 1:, 0
1. -1.28 ' -24. e.
. roo. C-100 -2. Ell 2ý45. t5. C; C!.s4re 126. C,
_ý' -I-7 `5 E
a. 1:15 11E:6. -13 - 18.5,8 -31 C-1 1
700.0 CC -2.60 277.0 _IEZ1.61to 46. a 7. t9 -241're..s
0. C,C,CI
C,ED -3.43 ' E68.5 , '5 42.6 6.45
'9010.000 -3.53 EEO.
'ý` 5 -I 7e.E. E. S:
9.5 6.06
5.1ý4
5%31.E - 17.7E 28'. 4
C-6. 6.6
1-400 CEO -4.42 6.1 13 331. -4. -43
C,OCI -20.1
le-GO. CICICI
-4.75 224. 16. S-9 I r_'3 67. S
I?
1,999.999 S44 ale:. 1 -16.96 S3.1 Sa 6,:,
W- ý- -E.. set

5ht. Q3 c-F fG
,
 173
01
&. F: I
H. F:h LEAD F',,h J'C(-U-.-iE1' FAIPS
FF-PENEIEF-S. ill 111-ji, EG

3ELMIZ. - EI-1117EP; EMITIEP 1-.IDIH E.

3 CLIPPEhl t-)A-- 2
FF EC.' '12
-,:: a
cic' (to 1 z:. 11 _76. Cl 16. . LIE, 163.7
I ID -1.211: -21 .2 -It. C.: -. 11 -9.,,:
-1.41 -29. S. 15 76? I M6.2.
c,C1.C10c, ii
-I . E-E, 26 . C, 69. E. E--Q I -cr'ýC- I-' -- EE. -17 S
'6. C. 14 .
CIO.(, (:Cf E. E. -!
'C'ý4 CI. ý, -aa. 1: 1
E
. E.
-2.
-c r-6 r,15.5 I 56 1E. kil.
I 52
I, .
12' .9z:.
nc, cf. CIE-Cf -3.05 -21.56 -2 .75 11 S. ro E.ý 7
ti OCIO
le-cf. -;?.. 46 2E'?. 1 -'213.2 11.76 Z:2.5
"Sti. e59. a? 4.7 l'
7C-IcI.OCIO -c'-cl. I ; -E:155.57! -4.1 C.1 S2c:
2*51. 41.9 9. E-3 'e ý4 ý4 'c', .6
:BC-10.oroo -4.42 -EO. -;. . -. .31 .I
S.CfcfI.oe cj 24S. 5 -15.61 Q. 77 9. C15 9ý1:1 C.
-4.41 . 1 -4. ? 71 316.6
C,cf. CfE; 237. E-3,40.8 C..0:3
-4.92 .I-I?.
222. E. - 19.29 S,7.4
91.1 cf
MOO. Octo -4.9E. E 64 . c-
CIO.CfC10 -5. Ell 22-1.4 -IE. - e. 67
1W. c- SE19.7-
.s6 ý'ýS4.6
1600. CIE-cl -5.24 211.5 -Ie.. 78 -319.1 4.51 -ne'. -7. c.c. -ýri
CIO.0 CIO
I CO. -5.16 212 `5 41.1 :3. E.S. 67.9 C16.
.3-. -7.12 -;,
1 -5.25 2-07.0 - is. CIE. 42. Z. 2.79 6:3. c- -7.47 -5-5.

E BEP 1? 75
SEAM LEFID Peff ISOLMED PAIPS
Ll=',Icc I=IE S F'FiF:
f-.tIETEF.S IN Dr-illEG

ZEEM-3-PlITTEP; ENIITER WID7H 2 MICF.C.11SI

VOLTS= 3 CURF. Elil MA= 4 sc.,
FFEQ $ 11 S12 I
2
1
Cie. Octc, -2.37 -29.9 -,.'*. 64 74.1 20. E:ý-, 1t'.
-;: ý4 "C -
1 to. rj('r_'
.
- -arc, -I'e'.. 4
-2.6-cZ, -42.5 -E,?. a(, 74.0 Ect-:ý6
200. c1c: 0
-54.7 -f.?
54 284.3 -2r.. 76 t4.0 1CR1ýt'14
. 1: 77 i
e,
, cio. ocio - :
11, eici 26,9. ,: - 26+. -14 I- E", 16. lc';
c IIe. 0
500.001-71 -4.1 257.1 68 -",
910. IS -7Q II re. -*: -,Qý, ýs
62 247. w
Elio. o0c, -22.78 -012.2 1-4-41 _Esi. 1 ft :31 C.
7,00. CIO&.
) 87' 2 ' ýCec_ -22. E0 46.4.
-CUE'. (1
23 1.. C, -15.7?
eco. C'E"D 27 -22.41 45.10 11. '?6 ILI'
rdC,C, -, 5.12 Z24. 1.79 4Z. 9
1 c! Ci0.0 CIO -5.44 219.2 'r '22, 8. C, 1Cl. 19 "X.. 7
212. t _1.709 45.9 Ic
1200.00c, -5. -21 . C,
. 4S4. ý'
D. Cicj 7,16.Z. C E.
*'ý,:
16,00. Cicit) -5.51 ' .
4at
1600. Ell-k-i _r
-19.2,2 52.3 5. co ý7.1
99SI. ý4 9'
196. 53.2 -9. C-C,
4 56 6:3. E.

Sht. 14 o; IG
 174

Ufit-l LEfilf Pet, HULP-tEll PRIPS

V=QLE: 1=1E FF.PRUERS '11h I, I: vIEC
ý.-
I
ZM, 11 EN11TEF; El"IllEP WIDIH 2 I-11CF.
C1,S)
V ClLT CUFFEtil NA= IS
FREC! Sil
I CIO.fj Erj -4.42 -.; 7.6 -Z-4.71 71.7 25. CIE 15 1.
O j ý4. r_-,4 C-5. IS 24.11 14(-.
- C -E; -.;. 7*0 -65.1 -; 4 -1 -10 -27A.
C,-7
C-10.c1clCt 11 Z'15.6 -LE'. 0c, 49.4, E.I. C-13 11 ý4 7f ': Q C.
clclcl -5.2 7 241.6 -27. lv, 5 1. cl 19. CIE 1ole. a
.4clo. -5. C,6 317. c.
C-clo. r2ccl -5.27 23 1 el 5 C1,3.3 17.14 IC 1.
cl clQ 22*2'.4 -25.5t, -S0 1'62' 16-05 -6. '07 1,-,
.q
700- CIC-1c, ý17 2-16. "0 - Etý. I C. C. 14.05 631a. S
0' ricir OE-4 -11.2
C- -",a 4.7'(' C5.0 I ". 37 S e. I z,6 :
-o.
-5.30 206.1 - 23, ! 6.1 1 Z. 5S Soe. ý4
(I Cl -0.1. C,-1 202.
1,57.5 '3
a cicf. C,C,C-1 -, r-..Z-Z, 1SIC: I. -22.70 10.05 ( ý. 1,!,
C, CIE'Cl 21
S&. .
1.746.2 E., 61.1 4
-9.9-: ý
C... ::C,
-ZE. -
. -5. -21.6c, C1.4
GO.C,rX, 45 193-9 -297.7.. 0 -290.5 C.3 :! C,2.2
19. ý40
C,1). cfcl0 -5.20 191.1 -
62. "P 6. C-2 6". 6
ISS. 1- 1S..1ID 63.3 64 6Q. C, -6c. 9

207K tIOVERBER 1975

BEFIN LEAD PAi V-20LATED PRIPS
tl---J)r-E: I=lc S PFFRIETEPS IN YJ-r;
tI, Er,

3EEMS EMITTEP; EMITTER WID7H 2 PIICFOIiS)

IJOLTC-- el CUPPENT PiFi= 14
FREG sli S12 S21 E-22
10Ei. ü 19171 C.5 -71.3 6. r'.ri 65. EI 2,7'. 17 143. 't' -1.27
190.10CAD 27 -9:3.1 45 E-6.6 2;. 410 131.0
-6- 01 251-1 -: X. 27 '.57 4.
. C'.4. ZZ,4 121.5
-5.7-*B r, 0.5 21. I'sC11
2-10. E. 54.
cio,
-5.68 -29.48 19. Cj 101.2
2 121. 1. -90 Z- 17'.
50 la. clEiri -5.48 - 2-I-E
-. 1 "ES
-1: 61.7 17 "4
..: 95.1 7 S7 t. 16.1 .
-27.12 .
-S69.2 E1 i0 1
203. -26. E2 E2.6 .C 7( *It.
-E! 4
s -5.5-4 IS-9. - 2E.. C, .2 131.as S10
e Eic"21 -5. ir- 1ý-r*.6 -E-1-QI 65.7 12.6 -7 cl z., 9.7
C,Eik--, -! 51.Za 1913.6 C.S. 2 'o, .
ID -24.2-4 11. r75 7e. 9
lýac-Iri.Cicict 13 191.1 -23. ZZ 65. ict. 10 7-. I Cl. 70 -t 17
14r-ý0.C.Eiii -5. lei le-9. *ý" -22.1 a 61ý. 4 C:;
2 -7cl * -( I Cf.1104 S:-U7,1, L"
.,
11, o Ei. e,Eto 48 QW.. -11.10 SCIV.. 4
-
9. ý49*?
-2-,0.27 E9. I E. 59 6
1:;. -I I. CJý4 so-ý. 6
I 1:
-? 69.2 "S.E-0 C.C. 6

Sht. Is0; 1G
 175

I-OkIENTER 1$75
c*-C-17ý,
MEN UK' P.-*tl ISOLME! FAIPS
S FPFF.t-!EIEF",-. IN : 11'tI;EC

Ef-ITTEF;
12121,1Q. El-,
1177E.P wirilH 2 ttiuota)
CLTF.EhT fiFi= 21
FPEC! Si I S21 SC112
C.O. I-Xict -104.7 - 37,
C,:
9 2:4.44 11 -2.11? _C"erZ.

Oct.ciclu I 11 C.). 9
(I(-C, S"', 2 1 64 10-1.4
4-c-cl.
204.0
.4
. -, C. r.,-. ,4
4-
C;c1c, -5.74
17.11'
S44.7
r 00. (10.1 ECK,. 0-: I e". ; Cl 8-9.?
,, 19.2
600. c1clu 47 IS,7.0 -, r -ES2. -Ial 179 - 273.
i cl.
c, (10-1 12.46 -2777.3
EOct. ciCIO -In. IC16 -2E. 71 69.4 11 cl 1
- IS,
0.8
SC,0. C,CIE, -5.13 18,9.5 - 25 . 25 4-2.S' 1 C-6r--0 77.1
12.8.2 -
oc,C'.C,CIO -5.27 -24.74 -72.1 9- 08 76.1
MOO. CICIC, -r , 1. Elt, 7*. E.4 71 'ri
.
1; cl0. ED CIEl -5. EICI 126.1 -22.22 73.4 66 3
1i CIO.Q00 1S'5- C'- - al, 3,3 -287.1 5. Ci-'-- -295.2 .
-5.13 .
1'1:110. C,CIO 183.4 27 72.7 el i. i
-4.78 -F-Cf. -4.24 6(1.9
19199.9 95. -4.7 1 -19., 0 72.2 3.25 501.1
0 -53.7

2U-17H MIXI-SEP 1? 7.5

"ISKI LEAP. Netf ISOLATED PPIPS
I=Ir;. C; F'PF'FtIEIEP'--' IN DP.oI:EG

3ECEW3 MITTEN EMITTEP WIDTH Z., I-IjCF:Ctj<,)

CLTPEIIT- t, Fi= 7
C'.
FREC, I ZýI a: S21
EICl0 '-22
0-9 .5 2.2 4 117.4
-
, 1 C. c Z-7 ý4. 10 C
12 I. .4 l _4" 7 a:
z 6.7E3 7 IV 1 0: 3
-5.79 -4 101 ý
-4.47 11E
0 00 , 1S. I-E-1.2 Ei, 7 l
57 3. S,-cl; E6.7 11. Er"
00 ci -C,,. 44 I. eq Cl. 2? 75.4 ý4. C,.
Iý
','3.7. -Fl . 6
E.Cle-.0-C-10 I S9. F, _218.E8 -E '--7. Cl zIE:-274.4 -17
-284.6
": -a. -,
i - C:
f 7. "E'. oldQ 1: 1 4
I S_
E. 74. C: et.
IU k . C. 49 75.4 _t Szlo I
. -77. .
icicic-i. -Cl. 10 184.4 4-1.79 77.77 2,
-. Ir
'25.7,
-e 76. C*7 ýo
142. C.,0.
IE-13. IS, - Z. 9 71 E. 14 72.1 _ -z.' 7
75 I E.S. E -a. 4S 70W 1. Z15 -'r. 6
.5 t -IS. -7-V9
SO 7 -El 67- -EIE-4.9 1.4
. _C'95. lfk -7 . -: 4 S1 E 2
E:C.Ci.O(-iC1 1S.1.4 -EC. 47 7 5. C, C.7 . .
179.12 0'. 75. Co It M . I
%

-1.4's 57.77 -ý.. cý:'?

IG
 176

APPENDIX 2

RELEVANT PUBLICATIONS BY THE AUTHOR

 177

International Conference on

ComputerAided Design
and Pilanulacture.
of ElectLrnnic
Components,Circuits and Systems

3-6July 1979

Organisedbythe
Electronics Division of the
Institution of Electrical Engineers

in association with the

British Computer Society
Institute of Electrical and Electronics Engineers
(Circuits and Systems Society)
Institute of Electrical and Electronics Engineers
(Region 8)
Institute of Mathematics and its Applications
Institution of Electronic and Radio Engineers

with the support of the

Convention of National Societies of Electrical Engineers of Western Europe (EUREL)

Venue
Universityof Sussex
 178

EXPERIENCE IN MODELLING INTEGRATED CIRCUIT TRANSISTORS

O. P. D. Cutteridge and Monica Dowson

University of Leicester, U. K.

INTRODUCTION than did the original.

Industry is paying increasing attention to RESULTS OBTAINED
the linear modelling of integrated circuit
transistors. The authors have been fortun. The S-parameter data shown in Table I refers
ate in receiving data from Philips Research to the lateral pnp transistor mentioned
Laboratories (1) on two such Javices, firstly, earlier. Table 2 contains the errors in the
a lateral pnp transistor, and secondly, a S-parameters of the models shown in Figures
vertical npn transistor. The data 1,2 and 3.
provided in the first case is shown in Table
I and consisted of measurements of the The element values for the initial model'
S-parameters of the lateral pnp transistor shown In Figure I were those obtained using
at four frequencies in the range O. S MHz to the global search (21). The value
method of
10.0 MHz. the overall error function at this stage was
384.2. After Gauss Newton optiaksation of
A brief account of the method used by the this model during which three elements ware
for Jelling linear devices improved
authors .... active removed, the model shown in Figurg
is given in the following section. Figures 2. which gave an overall error function value
1.2 and 3 show the models obtained for the of 114.9 was obtained. Elements
, ware
lateral pnp transistor at three different then added until the final model in
shown
stages in the modelling procedure. with the Figure 3, having an overall error function
corresponding errors tabulated in Table 2. value of 6.1 was produced. As can be
.
seen from Table 2, the errors were reduced
mODELLING METHOD from maximum errors of O-S do It
and
degrees in the initial model to 0.1 do and
Construction of Error Function I degree in the final model.
The modulus and phase of each S-parameter The order of element addition was found to be
matched at every real frequency for
were quite critical, especially that of thv second
measured values were given. Every Although
which generator g2 . this element w4s
modulus error was given the same weighting. present in the originally suggested modoi,
as was every phase error, although these the optimisatLon ttchniquo
different. An overall would not tolurato
weightings were error its insertion un-til the final It
by taking stage.
function was constructed the sum of might be noted that the direction of the
the squares of the individual weighted errors. current is reversed
generator in the ftnal
model as compared with the initial model.
General Strategy
ACK%OWLEDCEMENT
The method rea uires an initial model with
* to be Although The authors
element values provided. a would like to express their
by the donors grati-
topological model was suggested tude to K. W. 4oulding and P. J. Rankin of
of the data. no element values were given. Philips Research Laboratories for providing
in order to obtain suitable starting values them with the accurately moasured S-paramoter
for the element'. the g! obal search method data for tl%e two tYpes of transistor.
described by Price 12) was used. With these
starting values the tain optimisatLon proce. REFERENCES
dure, the modified Gauss Newton section of a
computer program descrioed by Cutter. dge (3), Moulding, K. W. gn J Rankin, P. J., 19? 7.
was then utilised. Privato communication.
By running this optimisation algorithm in the Price. W. L., 1977. "A Contrall0#1 Random
domain of the logarithms of the Independent Search Procedure for Global OPtimL4atavn".
(repre5entznq the element ealue. s) COmn- J.,
variables 20,367-370.
it is possL 'ble to determine from the vaiiies
ef the corru,;, ions generated at successi. ve Cutteridge, O. P. D.. 1974, "Poarejal
Newton if
(; auts iterat. ons which, any, I.. part Program for Solution sunlinvar
be After
()f
elements should rLmuved. climina- Simultaneous Equations... C. tact I. Ott
ting all such elements the 3tgartthm conver. 10,182-L84. .
ges on to a Loca! minimum of the ovural'.
function. From such a point ftirther
error
improvement Can OnIf `)C Obt3LrkCd bV tht)
of new altoments. 3, Y 3d'JLnq now
addition
Une a time Lt is no-4%. hle P#% P. -II
elements -it
immcJi ate! je w:, e tý.,! r it I, not FIR
thev
I,..
Aru efPvvt t Fe
I"IMn ro VIn'! Ph.. , ), I, - -C4 ý

are : len-ifici - ith no%LtIvu vuL, ji:, i ,k it

a* -&I.
the nvcra! '. u-)r ý'un,:? ios ar,,
rejucing
re taIAFt, - rj -- P. 'ý'i ts trý It -S fn
later iivUcts liaving 4 greater nu-ý')er oi noL! cs
 179

TARLr 1. Measured S-parameters of a lateral pnp tratishttor.

S1 S12 S21 S.
.2

Frequ 0 ncy modulus Phase Modu I us Pha s 0 %todulus Ph*% : Modulus Phu::
MHI dB degrees dB degri ýcs (III dog re ' 45 egr ,

0.5 -0.10 -t. 40 --- --- -12. SO 179.10 0.00 -0.2 0

. 5.40 88.70 -12. SO 169.00 0.00
2.0 . 0.20 -64.70 . 0.30
94.00 ISS. 00 0.00
5.0 -0.40 --11.20 -S7.20 -12.80 -0.5 0
10.0 -1 . 20 -19.40 -52.00 76.3n -14.00 t32.90 0.00 . 0.90

A measurement of S12 at O. S M11: was not avaLlable.

TABLE 2. Errors in the_S-parameters of models of a lateral onp transistor.

S S12 S21 3
22

Fr*qu e ncy modulus Phase Modulus Phas: modblus Pha : : Modulus Phase
MH dB degrees da deg re s dB degr , dB degrees

O. S -0.10 -0.37 0.00 0.00 0.27 0.17 0.03 -0.19

initial 2.0 -0.14 -1.33 O. SS -0.17 0.34 -1.89 0.03 -0.26
model S. 0 -0.02 -1.60 0.1.0 -3.41 0.46 -S. 12 0.03 -0.41
10.0 OAS . 3.31 -0.36 -9.66 0.46 -11.02 0.03 . 0.71

O. S -0.09 -0.21 0.00 0.00 -0.30 0.52 0.03 -0.19

0.08 2.19 0.03
Improved 2.0 . 0.11 -0.71 -0.19 -0.49 . 0.27
Model S. 0 0.10 -0.36 -O. OS 1.97 0.07 -2.00 0.03 -0.42
10.0 0.38 -2.21 -0.04 . 1.43 0.43 -6. S2 0.03 -0.74
O. S 0.00 -0.21 0.00 0.0(1 -O. OS 0.67 0.00 -0.14
Final 2.0 -0.66 0.10 1.03 0.01 9.13 0.00
-0.04 -0.07
model S. 0 0.04 0.09 -0.13 0.04 (). a1 0.06 0.00 0.07
10.0 0.00 0.08 0.02 -0.19 -0.03 -0.05 0.00 0.24

A measurement S O. S MHz was not avaklable.

of 12 at
 180

G,a 6.12xlO-IOS C, a 8.68 x 10-S PF G2 z 4.67 x 10' SS C, al. OSX10-4 ILF

C22 5.06 X 10,7 ILF o3a 6.63 x 10'3s CZa3.90xlO- I
G2a 5.22x 10-SS pF
G] a 6.77x 10-3S C] z 3.71 x 10-9 gF g, a 3.94 x 10'3 s
gIz 3.53x 10-3s
g2x 2.07x 10-5S
Flgxe 1 Initica model Figure 2 Improved model

01 a 4.43 x 10- 5
G3s 2. X x 10-Is 223 2.72 x 10-'S C2z6.19x 10 gF LZz 0.16gH
C,,. 1.1S A 110"S 0- 2 C4 xV uF
Uj-;
C4,-3 3.7 3xV'gF

Figure 3 Final modei

 181

Algorithms supplement

riots bYthe EdRor The pair of equations A-0 and A-0 have three solutions, the
AlSonthm No. 107 is publisW i out AjFCCI t with the arproximate locations at (0,0), (1.5,3.1) and (1.5, -2.5) having
Institute of Mathemtscs &M IU Appliations. and is snociated been obtained by a global search procedure. The WS procedure
with the authon' article in the JIMA. Volum 24. Number 1. was used to obtain a refinement of the solution for which the initial
Aug= 1979. approximation is xt - 1.5. xt - 3.5. A zone size z-I was chosen
so as to exclude the neighbourhoods of the other two solution
AlPrithm 107 points. The precision was specified by ACC - 10-4. The WS
A WEJG117ZD S1.04PLEXPROCEDURE FOR THE SOLLMON procedure achieves the exact solution at (Z4) in six iterations. the
OFS114ULTANEOUSNONLLVEAX EQUA71ONS same number as is required by the NR procedure from the same
W. I- Prke and. W. Dowsm starting point and with the sameprecision.

Audo", Not" Rtferewe

The weighted simplex (WS) procedux provides an altaustive to die Nim W. L (1979).A Weighted Simplex Procedure for the Solution
t4ewwn-Raphsoo CNR) procedure for the solution of Y simui. of Simultaneous Nonlinear Equations, JIMA, Vol. 24 no. 1,
tancous nonlinear equations in N real vwubk-& rapid convergence pp. 1-9).
being obtained from a suffiactitty good itutuil appro,, unution. in
contrast on " NR the US bEIGNTED SlkPLEI PROGRAM
peac ure does not involve the compu- hNITTIN BY W.L. PRICS
tation Of dcrivatnres. The principle of dw %%Smethod. totether TUhSLAM INTO FORTRANBY h. 00b.SON
VMh the rewilts o( comparative performance two. are Publisheil
elsewhm (Price. 191,n DIMMUN V(3.4), U(3.4). 1(2), F(2)
Given the set of equationsfe (zj... xx) 0.1 m I.... JV. the
, = SM-RICUS UZ*UT &.3 MLLOb3
Procedure OPCI tes on the data -esoci-ted with a S1161PLEX of
N+I Points in N-sp- cc Prior to each iteration the co-ordinates; baM &. NO. Or VAAIABLL3 AbU nhCTIONS
of dme Points. So (%hem I-1. j-1. V + 1). and the
.. -V; .. -, U-T 1/0 MEAKS
COr. CSPOndinji function values. f, #. are held in store. For each point I
of the simpkx a WEIGHT. wj, is computed try "ving iby Gauss
CliwilutiOU) the set Of. V +I linear equations Zwj ft 1; Zwjfj m 0.
1101
The Xwjýzq j I&PUt N "D STIATING VALUES Or Vk&lkblAS l(k)
JV. co-ordinates x, . of t'he weighted centroid.
X. of dw simpleit are then dctermijxý ILArmth. loom
wid each Of tiw v functions
is evaluated at X. If. at X. ttw magiutude of every function is less ULC(lb. 1001)(X(I). I. I'M)
thin ACCý & user-supplied hPl. "I
mejoun of tbc mquimd ccurscy. tlien LP2. b. 2
the Procedure UMMIes. Otheraise one of ON* simplex Points is 42-02
discarded. its place being taken by X and tise
moddied amplex
Which MUlts form CALCULAIR IPUhCTIONVALLE3 AND FRUT ThEh
the haw of the next ites tion. The discs point
is chosen to be 1, that Point
Of the ample% vnth the least positive CALL FIACT(I. P. L)
(MIDU nCg1tivC) %Vghl4 2000)
unless L harrew to be the X mnt of the kalml"T.
Previous iteration in ii hich can the discard I-AM6(louT. 2001) Q.
point is chosen randomly
ftm the simplex but excluding the point w, that Point with the JbPVT ZOAL SIZI AND ACCURACI ILQUIkED
onost Positive "ght (Prwlý 1979 for full explanation). The procedure
is initialised by lim"atIng AUZ(lb, lool)Z. Acc
a ample% o( A-I points randomly
Positioned 'kidlin A hYPercL; be of linear 4imenston kAlT&%la%lT.230a)Z. ACc
zt the zom an)
the hypercube bang entred on 11w initial 4rMximabon supplied CALL SD%pLn KUTIhs
by tbC user- The chOICT
Of: a not critical. but ideally the hvDa=be
obould be IUP enough to embrwe the exact solution sought yet CALL bS"U(I. I. I. V. U. bPl. bP2. N2, Z. ACC)
small tmugh 90 gIckids other rosw bit solutions to the given system CJJ'tft? 3LLUIXON VALUES
0(equations.
T'le Procedure has been %&lT6(IOt, T. 2OC3)
PCOVInuned to ANM FORTRAN so as to
be gimerally applicable
and to require the mmunum of um coding a T-4
10 run a sPCc& rroWelit. All am%s
used " the program am
'kriared in the "win MOVIns so that the uw tias only to change c fOrASTC12)
the DIMENSION sute'llent to &Jopi t1w program kir use vnth any
1000
10&1 rchhWatio. 3)
zcoo SIMPLEX PhCGAAh///16H STAAIIKG VALUES//
Particular maximum %Cut of V. Tim us" must surply & submtne
FUNCT(-r., F. N) aluch I 51. ýLVAIIAbLES, ISX. 9hFLhC-,I,; t4S/)
c&kutstc, the values *(his set o(functions 2001 POPPATO-h X(. 12.4h) IPEIO-3)
F- corresponding to the Z002 iýW. ATG'/13t. =14. SIZE - IeLIO. 3121h AC&. (,FACT hLQ41RED
variables X -two Jr and F are "th of
dimension V. lie I -,FEIO. 3)
must 11-) Suvrly. as data. the value of N. the 2003 f"ihAT(//16h 50"TION VALLLS//54,9hVARIABLI: S. 191.9hiUNCTIOr. S/)
C40"Ord'notts Of the U-sal at)rroxtmanon.
the zone size -, and the ZKD
required 4ccUrsC-y ACC. T,-w SUMOINIL SMPLI(A. F. A. ý. 6. NPI, NP2,42. Z. ACC)
user must code a random number
llenctatOr apPrOOnAtt to %is is a standard
function On the CYbcf 72 coir. puter-RANF *91"10 SMOLEX SUEROUTILE
Used bv the authors. IAITIEk 81 I.. L. PRICE
As an example Of the MANSLATED Ib-O rOATRAb DI H. C44SCh
OMralien of the proV= a printout is
surP"ed (Or a run 1`0311119
to it,* sgm=fk mo-vanable problem in Ob ENTRY . SIARTING VALUES
16.AARAT Of VAKALL&S.
%hich tM (", -txws am CK LILT . SOLUTION VALUES
F& - ZZ, 3 So #*A*mAl Cf ktkCTIQN VALUES CURALSeChOlhG 10 1
- #2 N. &UkW Of K64.11CLS ANL, VAIlAbLý
A- 641 - ills + X1

Th*ComputerJournal Volumt22 Nurnber3

 182

cV AnD U ARE ARMS USLO 61 6SMPLX

00 70 12I. N
c hpl. h. 1 IPI*I*l
c NP2. N. 2
KKKaNPI+lPi
G N2. b*2
DO 50
F. ZxZOhE SIZE
K-Lkk-kk
c ACC. RLGUIREO ACCURACI It GO TO 50
c E* 40 J-1, N?2
DikENSION V(A.Pl, h2), U(NPl. NP2). X(N)d(Ns b*U(9. J)
FWPI. FLCAI(NPI)
c
c GENLRATL IhITIAL SlhPLEX 40 cohilfuh
so CONTINUE
CALL FUNCUM, k) 00 60 Kal, N
00 10 W, N KPInK*l
Do 60 J. I. NPI
JPIOJ+l

10 CONTIMA 60 CONIII#UE
DO 40 Jal. b 70 CONTME
jpl. j*l U(NPI. k? 2)su(bpl, l; p2)/U(Npl, Npl)
40 20 K. I, N L-kpl
ti. Np,r
DO*90 Ktal, N
AW IS RMON hUht: 01 UFMMICR hIlk Mhhl AKG4hihT
TO
90API-Kk
NANKING (A-IrChOU %4&hLF,. Wtu VAL6LS IN hAhQ& 0.0 1.0
1.0.0
KFI. K*l
V(Jpl, K)21(K) DO 90 JoKpl. tpt
20 COI-TlhtV fiRB*U(K. J)*U(J, NFf)
CALL fbhCI(X, F,; -) so CONTME
DO 30 Ksl, V VK, I.P2)s(U(L, NP2)-l0/Q(I. K)
KfhýK+N
VCJPI, LFN)SI(k) FIbD k AND L
30 CONTINTL
40 CONTINUE It (U(e. K?2). LT. U(L, NP2)) LsK
IDII If (U(K, KP2). GT. U(h, N?2)) K-K
go CON11hum
#;L*;? t;Tb, WEIGhIED MUM, I RETURN
END
50 CALL 3UBhOUTINE fUhCT(X. F. N)
00 60 ksl, N MIkSION X(N). F(h)
l(K)SO-0 l0)s2-Q*l(l)'*3*X(2)-X(2)*113
Cri 60 JOI, Npl F(2)-6.0*X0)-1(2)0*2+1(2)
X(K)@l(K)+b(jdlp2)bv(j, K) MORN
60 COMILUE ED
k6VALUAIL &I X AND ILbl tch COt.VLhGbJ.CL

CkLL FUCI(l. t. 0 Algoridim 108

00 TO L. I, h
If (AUMM. M. ACE) Q IG 80 EFFICIENTSOLU77ONOF TRIDIAGONAL LINEAR S Y=SfS
TO CChTIt. UL Ole Osterby
ComputerScienceDepartment
Ull It C&hTLj%GEC
AarhusUniversity
RiT4R%
Author'sNote
ChOGSL DISCARD POINT The solution of a tridiagonalsystemof linear equationsof the form
60 It (L. NL. 10 GO TO 90
LaUINTUNP16RANEW. 0)) bi ct x, di
go IWL at bi c3 XI dt
RLPLACL DISCARD POW 61 1

00 100till'k dm
an b* 6xsý

IGO LONlIbUL
is carried out efficiently by meansor Gaussianelimination. In the
-COTO 50
following we assumethat pivoting is not necessary. The idea behind
LND the algorithm is not new ýSprague,1960;Leavenworth. 1960).but
SUEbOt,Tlbl&
wehavesuppliedthe procedurefor easyreference.
#A*.PUTrb 11LIGMS AhL FIND MUST POSITIVi 6EIGNI, k The operation count is: 2n -I divisions, 3n -3 multiplications
AWLoLEAST P03111VE VALIGH1, L and 3n -3 additions. The only new feature in our procedureis a
i; ImLNSIC#; V(NPl, h2), U(hPl, kP2)
slightly more efficient use or simply subscripted variables. The
number of such referencesis lOn -5 compared to 13n -9 in
IkITIALISik U ARNAI Leavenworth(1960)and Sprague(1960).The solution is returned
in array D and array B is destroyed.
00 10 tal, h?2 The reasonfor this very detailedoperation count is that on most
u0,10.1.0
10 cohTINUE computersdivision is slowerthan multiplication. and floating point
DO 20 K-2, týPl addition is not so much faster that it makesthe time ;ns:gnif cant.
u(g. 6P2). 0 a Subscriptedvariablesare countednot only becauseof the subscnpt
20 CONT111. Ii2
DO 10 J. I. t. handlingbut alsobecausetheyimply memoryreferences. In cor.trast,
jvl. j*l the simple variables can (and -should, if possible) stay in Ust
jph. J. h
to 0 Kal, Npl
registers, of which most modern computers have sufficiently
many.
30 If severalsystemswith the samecoefficientmatrix are to besolved
subsequently,we can store the intermediateresults from the LU-
Ck)h*lijlk. ý.ElGhTS decomr)ositionin array A for later use when the right hand sides
become available. The operat-on count for subsequcntsystems

The Computer Journal Volume22 Number3 283 -

 183

IEE

DIVISION
ELECTRONICS

COLLOQUIUMON

FORCIRCUIT
PARAMETERS
'MODEL
IS
ANALYS
ORGANISED BY

PROFESSIONALGROUPE10
(CIRCUIT THEORYAND DESIGN)
AND PROFESSIONALGROUP E3
(MICROELECTRONICSAND SEMICONDUCTOR
DEVICES)
FRIDAY 27 MARCH1981
I
DIGEST No: 1981/25

" ""-" . "- r ' _. " "- "

 184

ON MDEL MODIFICATION
CONSIDERATIONS

Monica Dowson

1. Introduction

A number of equivalent circuit models of electronic devices exist

which can be optimised to match the measured characteristics of an actual
device. Vher; the model does not give a su-E-ficiently accurate
agreement with the measured characteristics then mod4fications have to
be made to the model. This paper discusses some aspects of model
modification in order to match the S-parameter measurements of high
frequency integrited circuit transistors.

Determination of original model

For any given device a model based on the physical aspects of the
device is normally available where perhaps some of the elements are
given fixed values but most of the element values are unhnown. These
element values must therefore be optimised to provide a fit to the
required characteristics.

Given a set of S-parameter measurements of a device at a range of

frequencies together with a suggested model, the optimisation method,
favoured by the author involves setting up an eiýror function given by the
sum of the weighted absolute differences between the measured S-parameters
and those calculated for the model. This error function can then be
minimised using the Gauss Newton method. By running this optimisation
algorithm in the domain of the logarithms of the variables (i. e. the
element values), the element values are constrained to be positive.
Vhen convergence of the algorithm onto a minimum is obtained all the
corrections to the element values are very close to zero and this is a
positive indication that the model is a suitable one.

3. Modifications to the model

If the convergence of the gauss Newton algorithm is not obtained this

can be an indication that the model is not suitable. By looking at the
values oi the corrections to the individual elements it is possible to
determine what action should be taken.

3.1. Removal of elements from the model

Since the algorithm is operating in the logaritimic domain it is

possible to say that a large negitive correction to an element could
indicate that the element should be : ero. If, after applying that
correction to the element, the correction for that same element is still
large and negative, this acts as confi=ation. Thus, if for example,
the element was a resistance then this should be short circuited. A

lttnica Dowson is with the Department of Engineering, University

of Leicester.

.:).. I.
 iss

similar argument may be used for corrections that are large and positive.
Fig. 1 shows a suggested low frequency physical model for a vertical npn
transistor operating at frequencies of 0.1 to 1.0 CH., After applica-
.
tion of the technique described here the model given in Fig. 2 gave coývergence
in Guass Newton over the entire range of frequencies with an attendant
improvement in the high frequency response of S n particular, as shown
12
in Fig. 3.

3.2. Addition of elements

Just as this technique can be used to remove unwanted elements, so it can

be used to determine whether a particular element could be added to the model.

The placing of additional elements is the cause of some disagreement.

One view is that all elements in a model must have a physical meaning within
the construction of the device. To comply with this requirement, attempts
should only be made to place elements in certain predetermined positions in
the model. However, as sho%mby Cutteridge and Dowsonl, it should not be
a*ssumedthat because the addition of a particular element has failed once, it
will not be acceptable at a later stage.

Where a model is required that gives close agreement with the measured
characteristics, the requirement of adding only physically based elements can
be an unnecessary restriction particularly if the view is taken that any model
is only an approximation to an infinite connection of elements. In order to
be certain that each element added is the best one at that point in the model
growth all possible placements must be attempted. This can be most time
consuming and some restrictions are usually necessary. Thus, to add elements
without creating a new node one can either attempt to add elements in parallel
with elements already present in the model Can approach most physicists would
not find too unorthodox) or attempt to add elements between each combination
of nodes. To add a new node, the simplest way is to attempt to add one
element in series with an existing element although, of course, this does not
cover all possibilities.

Results

Fig. 4 shows an example of a model grown from the reduced model in Fig. 2.
The frequency response of modulus S. for this model is shown in Fij. 5 and
1
in Fig. 6. - As can be seen this model now gives
of modulus and phase S,,
good agreement with the 'required characteristics.
References

Cutteridge, O. P. D. and Dawson 1!. "Experience in Modelling Integrated

,
Circuit Transistors", I. E. E. Conierence on Computer Aided Design and
Manufacture of Electronic Components, Circuits and Systems, July 1979.

2. D-
 186

-. S 3 i
". -
a 'Z cm
1A

6t

IS
I 'o
\ 06
S
S
.5

00
CY I'll
LM
PM
2 ,a 0
2
C2 'o -V cy

£
0

I, 010!..
.4S. a

-
-
OT
-ii
I I
I

2.31
 187

9 1
46 r. a

ý0 «so
-0

§ r4 Cm ml mi ýr 0800
. ýQ ýr tý4 CD
2
ýo ýr r14

1"
S
I
IU
h
Lcr-
A. J"

- *J
*

Ix

2-. Lm
 188

NC0MFECE
EUR0PE ko'-A..
on ELEC-7FO'MICDESIGN
AUTO'wý"IQATIOM,

1-4
September 1981

Organisedbythe

Electronics and Management and Design Divisions

of the Institution of Electrical Engineers

in association with the

British Computer Society (BCS)

Convention of National Societies of Electrical
Engineers of Western Europe
Institute of Electrical and Electronics Engineers Inc
(Circuits and Systems Society) (IEEE)
Institute of Electrical and Electronics Engineers Inc
(Region 8) (IEEE)
Institution of Electrical Enginers (IEE)
(Southern Centre)
Institution of Electronic and Radio Engineers (IERE)

Venue
University of Sussex,
Brighton, United Kingdom
 189

24

EQUIVALENT CIRCUIT MCDELLING or A SMALL SIGNAL HIGH FREQUENCY BIPOLAR TRANSISTOR AT DIFFERENT
BIAS CCNDITIONS

Monica Dowson

University of Leicester, UK

INTRODUCTION Details of this modelling technique have

been given previously by Cutteridge and Dawson
The purpose of this paper is to describe the (3) and Dawson (4).
progress made in the equivalent circuit
modelling of two similar small signal high The optimised model was then used as the ini.
frequency bipolar transistors uitder different tial model for type B and gave good results
bias conditions. without any further topology changes.

The transistors being modelled were bipolar Thus. taking the model as the basis of a bias
transistors designed to be used in beam lead dependent model, the optimisation algorithm
integrated circuits. S-parameter data for described above was then used to obtain suit-
the transistors was available in the fre- able element values at different values of
quency range 100 MHz to 2 GHz for several emitter current.
different emitter currents.
BIAS DEPENDENT MODEL
Using optimisation techniques, the aim was to
produce accurate models simulating the The model shown in Fig. 3 gave good results for
measured S parameters of the transistors. both type A and type B transistors over
with as few bias dependent elements as the entire range of frequencies at emitter
possible. currents within the normal operating regions
described above. However, at currents in
A brief description is given here of the opti- excess of the maxima stated further changes in
Mi33tiOn techniques used and the results ob- the topology of the model were required.
tained are illustrated by diagrams of the
model together with the S parameters of the The original hypothesis was that there would
model compared with those measured for the be a small number of linearly varying elements
transistors. * in the model that would describe the bias
variation. Within the normal operating.
DESCRIPTION OF THE TRANSISTORS regions this was found to be a valid assump.
tion for this particular model. For example.
The transistors being modelled were isolated as is suggested by the graphs in Fij. 4, R
n-p-n transistors made in silicon with im- is inversely proportional to the emitter
planted arsenic emitters and diffused boron current and C, can be approximated by a
bases (Slatter (1)). The first type (A) had function of the form
one emitter stripe and the second type (a)
b
had three emitter stripes. The data for
these transistors was supplied by Philips
Research Laboratories (Slatter (2)) and con-
S parameters where a and b are constants.
sisted of the of the transis-
tors in common emitter configuration at 12
frequencies 100 MHz to ; GHz Taking the model beyond the normal operating
in the range
at emitter currents of O. S MA to 10 MA for regions. evidence of the type of topology
O. S mA to 27 mA for type B. changes required is given by C, which in
type A and of
Fig. 4 shows a rapid increase ,
in'valut as the
From the graphs in Fig. 1 it can be deduced emitter current rites and which the optimisa-
A up tion algorithm deletes open circuit at higher
that the normal operating range of type
to I CHz was O. S mA to 2 mA and of type values of current. and by R2 which is later
B was I mA to 8 mA deleted short circuit.
.
RESULTS
MODELLING TECHNTQUE
Examples of the S parameters calculated for
The modelling process was started by working
the model compared with the measured parame-
on type A only. Using an initial model as A and type B transistors
ters of type are
shown in Fig. 2. an optimised model was ob- in Fils. 5 and 6. As can be seen these
for the whole of the given range of given
tained frequency
frequencies of emitter give good agreement over the entire
at a single value 100 Milz 2 Gil: is a
range to and there
current within the normal operating region of in cases
This shown in smoothing effect where there appear
the device. optimised model,
by using to be measurement errors.
Fig. 3, was obtained successive appli-
cations of Gauss Newton to minimise the
CONCLUSIONS
weighted errors between both the calculated
modLius and phase of each of the S par&-
The bias dependent model obtained shows sever-
meters of the model and the giYen measurements
to the one described by
The correction al similarities
of the device. terms calcu- Slatter (1) which by a different
lated by Gauss Newton were used to indicate was produced
were required, approach to the problem.
where topology changes conver.
gence in Gauss Newton being required at each
The model described here simulates the trans.
stage in the development of the model.
istors being modelled over a wide range of
 190

25

frequencies and emitter currents and with to-

pology changes gives good results at even
larger values of emitter current.

ACKNOWLEDGEMENT

The author would like to thank K. W. Moulding

and P. J. Rankin of Philips Research Labora-
tories for providing the necessary data.

REFERENCES

1. Slatter, 1977, "Optimisation of a Small

Signal High Frequency Bipolar Transistor
Model Using 9 Parameter Measurements at
Different Btas Conditions". Conference on
Modelling Semiconductor Devices, Ecole
Polytechnic do Lausanne, October 1977.

2. Slatter, J. A. G.. Internal Publication,

Philips Research Laboratories. 1977.

3. Cutteridge, O. P. D. and Dowson. M.,

"Experience in Modelling Integrated
Circuit Transistors". I. E. E. Conference
on Computer kided Design and Manufacture
of Electronic Components, Circuits and
Systems. July 1979.

4. Dowson, M., "Considerations on model

Modification", I. E. E. Colloquium on Model
Parameters for Circuit Analysis, March
1981.

FT(MHz)
3000
2000 Hill.
1000

niiii ---Fi
1IV. O-sv
2-- -
-7-
Mod 0-
S21
dB at
1 GHz -2,

\yp eA Type b

-T,
ilu 10.0 1000.0
Lig 1 T-cL.
nsis2r chcrc: "Mr-,:tics --t, 1GH le(mA)
 191

26

Cl

I 2
Ri RI
1 2

C4

fig. 2 Initiat Model fjg..] QptimisedModel

Rin
C,pF i-
2400 .-1-
.00
2000
1600 0-7-
1200 0.5
800
0-3
4w
4-
02468 le mA 02468 lemA

R20 C2nF
so 0.04.

0-03-
40
30 0,02-

20. 0.01

10 0-00
0 lemA lemA
-Type A
Type 8
Erg,. ýcmoesof bias de eriddent
L E? etementvalues
_p_
 192

27

ModdB Sil Phase11 ModdB S12 Phaseo

lu , lu
Freq GHz Freq GHz

ModdB S21 Phase* ModdB S22 Phase*

-1

-I

lu FreqGHz JU lu- FreqGHz

ModulusS- modelx measured lez 1mA

Phase S --model v 0.5v
o measured cez

Lg, 5 Measured and modet S parameters-TypeA

 193

28

ModdB S11 Pnasel ModdB s Phase

12

10 , lu lu, lu FreqGHz
Freq GHz

ModdB S21 Phase* ModdB S22 Phase*

I-

160

140

120

-100

80

rreq unz rveq unz

I, = 2mA
Modulus S- model x measured
Vce=O'5V
Phase S ----model o measured

6 Measured and model S parameters -Tyoe B

_fg,
 194

REFERENCES

1. MacLean, D. J. (_1977), "Problem 211, Electronics Circuits and Systems,

vol. 1, No. 2, p. 84.

2. Mathews, N. A. and Ajose, S. O. (1977), "Modelling Technique for

Microwave FETs and Bipolar Transistors", Conf. proc. 'Computer-

Aided-Design of Electronic and Microwave Circuits and Systems',

University of Hull, 12-14 July 1977.

3. diMambro, P. H. (1970), "An Investigation into the Uses of Coefficient

Matching in Network Synthesis", M. Sc. Thesis, University of Leicester.

4. diMambro, P. H. (1974), "A Study of the Evolutionary Approach to Network

Synthesis Using Coefficient Matching", Ph. D. Thesis, University of

Leicester.
S. Wright, D. J. (1974), "Studies in Optimisation as an Aid to Circuit

Synthesis and Design", Ph. D. Thesis, University of Leicester.

6. Krzeczkowski, A. J. (1976), "Theory and Design of Lumped Linear Three-

Terminal RC Networks". Ph. D. Thesis, University of Leicester.

7. Hegazi, O.M. H. O. (1977), "An Investigation into Current Possibilities

in Automated Network Design", Ph. D. Thesis, University of Leicester.

8. Savage, W.H. (1979), "Automated Synthesis of Lumped Linear Three-

Terminal Networks", Ph. D. Thesis, University of Leicester.

9. Dowson, M. (1976), "Multimodal Univariate Search Techniques and their

Application to Optimisation Problems", M. Phil. Thesiss, University

of Leicester.

10. Henderson, J. T. (1978), "Optimization Techniques and their Application",

Ph. D. Thesis, University of Leicester.

11. Dimmer, P. R. (1979), "The Use of Second Derivatives in Applied

Optimisation", Ph. D. Thesis, University of Leicester.

 195

12. Cutteridge, O. P. D. and diNfambro, P. H. (1970), "Simultaneous

Generation of the Coefficients of Network Polynomials and their

Partial Derivatives from the Nodal Admittance Matrix", Electron.

Lett., Vol. 6, No. 10, pp. 308-310.

13. Rankin, P. J. and Moulding, K. W. (1977), Private communication with

Dr. O. P. D. Cutteridge, 17 June 1977.

14. Rankin, P. J. and Moulding, K. W. (1977), Private communication with

Dr. O. P. D. Cutteridge, 25 July 1977.

15. Rankin, P. J. (1979), Private communication with Dr. O. P. D. Cutteridge

including extracts from Ref. (16), 24 July 1979.

16. Slatter, J. A. G. (1979), Internal Report, Philips Research Laboratories,

Redhill.

17. Slatter, J. A. G. (1977), "Optimisation of a Small Signal High Frequency

Bipolar Transistor Using S Parameter Measurements at Different

Bias Conditions", Conf. proc., "Modelling Semiconductor Devices! '

Ecole Polytechnique Federale de Lausanne, 18-20 October 1977,

Journ6es d'Electronique, pp. 183-193.

18. Barson, R. J., Bradly, D. P., Dimmer, P. R. and Young, A. G. (1977),

"Engineering Applications", course notes for "Interactive

Computer Graphics for Engineers", University of Leicester, pp. Hl-H20.

19. Electronics (1968), "The Transistor: Two Decades of Progress", Electronics,

Vol. 41,19 February 1968, pp. 77-130.

20. Getreu, I. E. (1978), "Modeling the Bipolar Transistor", Elsevier.

21. Navon, D. H. (1975), "Electronic Materials and Devices", Houghton

Mifflin Company.

22. U. S. Department of the Army (1963), "Basic Theory and Application of

Transistors", Dover Publications Inc.

23. Hamilton, D. J. and Howard, W.G. (1975), "Basic Integrated Circuit

Engineering", McGraw-Hill Inc.

 196

24. Chua, L. O. and Lin, P. -M. C1975), "Computer Aided Analysis of

Electronic Circuits", Prentice-Hall Inc.

25. Brown, R. W. C1969), "Device Models for Circuit Analysis Programs",

Computer Aided Design, Vol. 1, No. 2, pp. 33-42.

26. Brown, R. W. (1970), "Device Modelling" in "Computer Aided Desýgn

Techniques", ed. E. Ifolfendale, Butterworths.

27. Gray, P. E. and Searle, C. L. (1969), "Electronic Principles - Physics,

Models and Circuits", John Wiley & Sons Inc.

28. Brayden, R. L. C1967), "Simplified Characterisation of H. F. Transistors",

Electro-Technology, May, pp. 42-45.

29. Ebers, J. J. and Moll, J. L. (1954), "Large Signal Behaviour of Junction

Transistors", Proc. I. R. E., December, pp. 1761-1772.

30. Linvill, J. G. (1958), "Lumped Models of Transistors and Diodes",

Proc. I. R. E., Vol. 46, pp. 1141-1152.

31. Beaufoy, R. and Sparkes, J. J. (1957), "The Junction Transistor as a

Charge-Controlled Device", ATE Journal, Vol. 13, pp. 310-324.

32. Hamilton, D. J., Lindholm, P. A. and Narud, J. A. (1964), "Comparison of

Large Signal Models of Junction Transistors"s Proc. I. E. E. E.,

Vol. 52, No. 3, pp. 239-249.

33. Ruch, J. G. (1977), I'Modelling Bipolar Transistors", Conf. proc.

"Modelling Semiconductor Devices", Ecole Polytechnique Federale

de Lausanne, 18-20 October, Journ6es d'Electronique, pp. 89-118.

34. Agajanian, A. H. (1975), "A Bibliography on Semiconductor Device

Modelling", Solid State Electron., Vol. 18, pp. 917-929.

35. Orlik, A. (1979), "Parameters Identification Method for I. C. Bipolar

Transistor Charge Control Model", I. E. E. Conf. proc. "Computer

Aided Design and Manufacture of Electronic Components, Circuits

and Systems", University of Sussex, 3-6 July, pp. 145-149.

 197

36. Roulston, D. J. (1981), "Determination of CAD Model Parameters for

Bipolar Transistors from Fabrication Data Using the Bipole

Numerical Program", I. E. E. Colloquium I'Model Parameters for

Circuit Analysis", 27 March, Digest No. 1981/25, ppý 4.1 - 4.5.

37. Bassett, H. G. (1966), "The Derivation of Equivalent Circuits for

Transistors by Optimization Methods, Using a General Purpose

Computer Programme", Symposium "The Use of Computers as an Aid

to Circuit Design and Analysis", R. A. E. Farnborough, October,

Tech. Memo. RAD759, pp. 53-65.

38. Lucal, H. H. (1955), "Synthesis of Three-Terminal RC Networks", I. E. E.

Trans. Circuit Theory, Vol. CT-2, pp. 308-316.

39. diMambro, P. H. (1977), Private communication with O. P. D. Cutteridge,

6 November 1977.

40. Enden, A. W.M. v. d. and Groenendaal, G. C. (1977), "An Improved Complex-

Curve Fitting Method", Conf. proc. "Computer-Aided-Design of

Electronic and Microwave Circuits and Systems", University of Hull,

12-14 July 1977, pp. 53-58.

41. Calahan, D. A. (1965), "Computer Design of Linear Frequency Selective

Networks".. Proc. I. E. E. E., 53, No. 11, pp. 1701-1706.

42. Spence, R. (1970), "Linear Active Networks", Wiley-Interscience.

43. Aitken, A. C. (1956), "Determinants and Matrices", University

Mathematical Texts, 9th ed.

44. Traub, J. R. (1966), "Associated Polynomials and Uniform Methods for

the Solution of Linear Problems", SIMI Rev., 8, pp. 277-301.

45. Cutteridge, O. P. D., and diMambro, P. fl. (1971), "Simultaneous Generation

of the Partial Derivatives of Network Polynomial Coefficients:.

Further Details and Results", Electron. Lett., Vol. 7, No. 1, pp-3-4.

46. Fairbrother, L. R. and Bassett, H. G. (1965), "A Computer Program for

Analysing Networks Containing Three-Terminal Active Devices

Characterised by their Two-Port Parameters". Radio I

and EleCtTon.

Eng., Vol. 29, pp. 85-92.

 198

47. WOinert, F. (1966), "Scattering Parameters Speed Design of High-

Frequency Transistor Circuits"s Electronics, 5 September 1966,

pp. 78-88.

48. Hewlett Packard (1968), IIS-Parameters - Circuit Analysis and Design",

Hewlett Packard Application Note 95, September 1968.

49. Hewlett Packard (1967), "Transistor Parameter Measurements", Hewlett

Packard Application Note 77-1, February 1967.

50. Dixon, L. C. W., Gomulka, J. and SzegtF, G. P. (1975), "Towards a Global

Optimisation Technique" in "Towards Global Optimisation", eds.

Dixon, L. C. IV. and Szegd, G. P., North Holland/American Elsevier,

pp. 29-54.

51. Price, W.L. (1977), "A Controlled Random Search Procedure for Global

Optimisation", Computer J., Vol. 20, pp. 367-370.

52. Price, W.L. (1978), "A Controlled Random Search Procedure for Global

Optimisation", in "Towards Gl9bal Optimisation 2", eds. Dixon,

L. C. W. and Szegd, G. P., North Holland/American Elsevier, pp. 71-84.

53. Box, H. J. (1966), "A Comparison of Several Current Optimization Methods

and the Use of Transformations in Constrained Problems". Computer

J., Vol. 9, pp. 67-77.

54. Lootsma, F. A. (1972), "A Survey of 'Methods for Solving Constrained

Minimization Problems via Unconstrained Minimization", in "Numerical

Methods for Non-Linear Optimization", edited by Lootsma, F. A.,

Academic Press, pp. 313-348.

55. Knudsen, M. S. (1981), "Optimization Procedure for MOS - Transistor

Parameter Determination", I. E. E. Conf. proc. "Electronic Design

,
Automation", University of Sussex, 1-4 September 1981, pp. 34-38.

56. Nelder, J. A. and Mead, R. (1965), "A Simplex Method for Function

Minimisation", Computer J., Vol. 7, pp. 308-313.

57. Price, VI.L. (1979), "A Weighted Simplex Procedure for the Solution of
.
Simultaneous Non-Linear Equations", J. Inst. Maths. Applics. Vol. 24,

pp.
 199

58. Price, W. L. and Dowson, 11. (1979), "Algorithm 107 -A Weighted

Simplex Procedure for the Solution of Simultaneous Nonlinear

Equations", Computer J., Vol. 22, pp. 282-283,

59. Durbin, F., Montaron, J. and Heydemann, M. H. (1981), "A Tentative

Approach towards the Global Optimization of Electrical Circuits",

I. E. E. Conf. Proc., "Electronic Design Automation", University of

Sussex, 1-4 September 1981. pp. 127-130.

60. Cutteridge, O. P. D. Dowson, M. C1979), "Experience in Modelling

-and
Integrated Circuit Transistors", I. E. E. Conf. Proc., "Computer Aided

Design and Manufacture of Electronic Components, Circuits and

Systems", University of Sussex, 3-6 July 1979, ppý 138-140.

61. Agnew, D. G. (1979), "Minimax Optimization Techniques for Electronic

Circuits", ibid., pp. 12-14.

62. Cutteridge, O. P. D. (1974), "Powerful 2-Part Program for Solution of

Nonlinear Simultaneous Equations", Electron. Lett.,. Vol. 10, No. 10,

pp. 182-184.
63. Fletcher, R. and Reeves, C. M. (1964), "Function Minimisation by

Conjugate Gradients", Computer J., Vol. 7, pp. 149-154.

64. Davidon, W.C. (1959), "Variable Metric Method for Minimization",

A. E. C. Research and Development Report No. ANL-5990.

65. Fletcher, R. and Powell, M. J. D. (1963), "A Rapidly Convergent Descent

Method for Minimization", Computer J., Vol. 6, pp. 163-168.

66. Lanca, M. J. A. and Nichols, K. G. (1981), "Interactive Design to Time-

Domain Specifications", I. E. E. Conf. Proc., "Electronic Design

Automation", University of Sussex, 1-4 September 1981, pp. 122-126.

67. Levenberg, K. (1944), "A Method for the Solution of Certain Non-Linear

Problems in Least Squares", Quart. Appl. Math., Vol. 2, pp. 164-168.

68. Marquardt, D. N. C1963), "An Algorithm for Least Squares Estimation of
Non-Linear Parameters", S. I. A. M. J.., Vol. 11, pp. 431-441
 200

69. Broyden, C. G. (1965), "A Class of Methods for Solving Non-Linear

Simultaneous Equations", Math, Comp. Vol, 19, pp, 577-593.

70. Gill, P. E. and Murray, W. (1972), "Quasi-Newton Methods for Uncon-

strained Optimisation", JIMA, Vol. 9, PP. 91-108.

71. Brown, K. M. and Dennis Jr., J. R, (1972), "Derivative Free Analogues of

the Levenberg-Marquardt and Gauss Algorithms for Nonlinear Least

Squares Approximation", Num. Hath., Vol. 18, pp. 289-297.

72. Nabawi, A. A. -F. and Nichols, K. G. (1979), "Optimisation and Design of

Linear Networks, Including Nested Subnetworks to Some Frequency-

Domain Specifications", I. E. E. Conf. Proc. "Computer Aided Design

and 11anufacture of Electronic Components, Circuits and Systems",

University of Sussex, 3-6 July 1979, pp. 7-11.

73. Dowson, H. (1981), "Considerations on Model Modification", I. E. E.

Colloquium on "Model Parameters for Circuit Analysis", 27 March 1981,

Digest No. 1981/25, pp. 2.1 - 2.4.

74. Cutteridge, O. P. D. (1978), "Expressions for Optimum Values to Grow

Virtual Elements", I. E. E. J. on Electronic Circuits and Systems,

Vol. 2, No. 1, pp. 29-30.

75. Callahan, M. J. jr. (1972), I'Models for the Lateral P-N-P Transistor

Including Substrate Interaction", I. E. E. E. Trans. on Electron. Devices,

pp. 122-123.

76. Dowson, M. (1981), "Equivalent Circuit Modelling of a Small Signal

High Frequency Bipolar Transistor at Different Bias Conditions",

I. E. E. Conf. Proc., "Electronic Design Automation", University of

Sussex, 1-4 September 1981, pp. 24-28.

77. Wainberg, L. (1962), "Network Analysis and Synthesis", McGraw-Hill Inc.

78. Numerical Algorithms Group (1976), IIE02ADFII, NAG Fortran Library Mark S.