134
Microwave Balanced Oscillators and Frequency Doublers Nipapon Siripon Submitted for the Degree of Doctor of Philosophy from the University of Surrey Uni Advanced Technology Institute School of Electronics and Physical Sciences University of Surrey Guildford, Surrey GU2 7XH, UK August 2002 © Nipapon Siripon

Uni - University of Surreyepubs.surrey.ac.uk/771938/1/250885.pdfMicrowave Balanced Oscillators and Frequency Doublers Nipapon Siripon Submitted for the Degree of Doctor of Philosophy

  • Upload
    vomien

  • View
    215

  • Download
    1

Embed Size (px)

Citation preview

Microwave Balanced Oscillators and

Frequency Doublers

Nipapon Siripon

Submitted for the Degree of Doctor of Philosophy

from the University of Surrey

Uni

Advanced Technology Institute School of Electronics and Physical Sciences

University of Surrey Guildford, Surrey GU2 7XH, UK

August 2002

© Nipapon Siripon

Abstract

The research presented in this thesis is on the application of the injection-locked oscillator

technique to microwave balanced oscillators. The balanced oscillator design is primarily

analysed using the extended resonance technique. A transmission line is connected

between the two active devices, so that the active device resonate each other. The

electrical length of the transmission line is also analysed for the balanced oscillation

condition.

The balanced oscillator can be viewed with the negative resistance model and the

feedback model. The former model is characterised at a circuit plane where the feedback

network is cut. By using both the negative-resistance oscillator model and the feedback

model, the locking range of the oscillator is analysed by extending Kurokawa's theory.

This analysis demonstrates the locking range of the injection phenomenon, where the

injection frequency is either close to the free-running frequency, close to (lin) x free­

running frequency or close to n x the free-running frequency. It also reveals the effect of

different injection power levels on the locking range. Injection-locked balanced

oscillators for subharmonic and fundamental modes are constructed. When the balanced

oscillator is in the locking state, it is clearly shown that the output signal is better

stabilised and the phase noise is attenuated. The experimental results agree with the

analysis. Furthermore, the spurious signal suppression in a cascaded oscillator is

investigated.

The other focus of this research is on the design of frequency doublers. A balanced

douber is designed and integrated with a balanced injection-locked oscillator. The

experimental result shows that the output signal is clean and stabilised. The other

important frequency doubler design technique studied is the use of the feedforward

technique to significantly eliminate the fundamental frequency component. The design

and the experiment show that the fundamental component can be suppressed to better

than 50 dBc.

Key words: injection-locked oscillator, locking range, phase noise, frequency doubler,

feedforward technique.

11

Email:

WWW:

[email protected]

http://www.eim.surrey.ac. uk/

III

Acknowledgements

I am indebted to my family, especially my mother, for their encouragement.

I wish to express my gratitude to my supervisor, Prof. I D. Robertson, for guidance during

doing this work. I am really grateful to Prof. M.J.Underhill for all his suggestion,

guidance and also valuable discussions. Also, I would like to thank Dr.S.Lucyszyn and

Prof. C.S.Aitchison for their help and suggestions.

I would like to express my appreciation to Dr.K.S.Ang for the great support, suggestions

and discussion on balanced oscillator design.

I am grateful to the Thai government for financial support. I would like to acknowledge

Mr.D.Granger and the Electronics Workshop and the Teaching Lab staff for their

supports.

I would like to knowledge the support of my friends: Dr.P.Asadamongkon,

Dr.S.Chuichu1cherm, Dr.S.Panyametheekul, Miss S.Daungkeaw, Miss K.Dankhonsakul,

Miss W.Usaha, Miss A.Pungnim, Mr.A.Butsayakul, Mr.A.Sangthanavanit and Miss

T. Tangmahasuk.

Finally, I would like to thank Mrs.L.Tumilty for her secretarial support and I would like

to acknowledge to Dr.S.Nam, Dr.D.McPherson, Dr.M.Chongcheawchamnan,

Mr.C.Chrisostomidis, Mr.S.Bunjaweht, Mr.M.Aftanasar, Mr.C.Y.Ng, Mr.J.Wong for

their helpful support.

IV

Contents

Contents

Abstract ........................................................................................................................... ii

Acknowledgements ......................................................................................................... iv

Contents .......................................................................................................................... v

L· fP' ... 1st 0 19ures .............................................................................................................. Vll1

List of Tables ................................................................................................................. xii

Chapter 1 ......................................................................................................................... 1

1 Introduction ............................................................................ : ..................................... 1

Chapter 2 ......................................................................................................................... 4

2 Balanced Oscillators ..................................................................................................... 4

2.1 Introduction .......................................................................................................... 4

2.2 Oscillation conditions ........................................................................................... 5

2.3 Balanced oscillator analysis .................................................................................. 8

2.4 A self-oscillating balanced mixer ....................................................................... 24

2.4.1 Design considerations for a balanced mixer .................................................. 25

2.4.2 Circuit realisation ......................................................................................... 26

2.4.3 Results and discussion .................................................................................. 29

2.5 Conclusion ......................................................................................................... 31

Chapter 3 ....................................................................................................................... 33

3 Injection-Locked Balanced Oscillators ........................................................................ 33

3.1 Introduction ........................................................................................................ 33

v

Contents

3.2 Phase noise theory .............................................................................................. 34

3.3 Injection-locked oscillator theory ....................................................................... 40

3.4 Design considerations for the injection-locked balanced oscillator ..................... .49

3.5 Circuit Realisations ............................................................................................ 50

3.6 Results and discussion ........................................................................................ 51

3.6.1 Fundamental injection-locked balanced oscillator. ........................................ 51

3.6.2 Sub-harmonic injection-locked balanced oscillator ....................................... 58

3.7 Conclusion ......................................................................................................... 61

Chapter 4 ....................................................................................................................... 64

4 Injection-locking applied to a cascaded oscillator. ....................................................... 64

4.1 Introduction ........................................................................................................ 64

4.2 Design considerations for the injection-locked cascaded oscillator ..................... 65

4.3 Circuit realisation ............................................................................................... 66

4.4 Results and discussion ........................................................................................ 67

4.5 Conclusion ......................................................................................................... 74

Chapter 5 ....................................................................................................................... 75

5 Frequency Doubler Design Considerations ................................................................. 75

5.1 Introduction ........................................................................................................ 75

5.2 General Doubler Design Considerations ............................................................. 76

5.2.1 Biasing considerations .................................................................................. 77

5.2.1.1 V gs near pinch-off (Class B) .................................................................... 80

5.1.0.2 V gs in the vicinity of forward conduction and between 0 volt and pinch-

off ................................................................................................................ 81

5.2.2 Frequency doubler design topologies ........................................................... 82

5.2.2.1 Single-ended frequency doubler ................................................................ 82

5.2.2.2 Balanced frequency doublers .................................................................... 83

5.3 Practical example of an injection-locked balanced oscillator and doubler ........... 84

5.3.1 Circuit realisation ......................................................................................... 84

5.3.2 Results and Discussion ................................................................................. 87

5.4 Conclusion ......................................................................................................... 90

VI

Contents

Chapter 6 ....................................................................................................................... 91

6 Frequency Doubler using Feedforward Technique ...................................................... 91

6.1 Introduction ........................................................................................................ 91

6.2 Design ................................................................................................................ 93

6.3 Circuit Realisations ............................................................................................ 95

6.3.1 Single-ended Doubler Design ....................................................................... 95

6.3.2 Reflection-type analogue phase shifter and directional coupler ..................... 98

6.4 Results and discussion ........................................................................................ 99

6.5 Conclusion ....................................................................................................... 104

Chapter 7 ..................................................................................................................... 105

7 Conclusions and Suggestions for Future Work .......................................................... 105

References ................................................................................................................... 110

VII

List of Figures

List of Figures

Figure 2- 1: A simple negative-feedback oscillator model ................................................ 5

Figure 2- 2: a) A microwave oscillator circuit, b) its signal flow graph ............................ 6

Figure 2- 3: a) Negative resistance oscillator, b) Two-port negative resistance approach .. 7

Figure 2- 4: a) The balanced oscillator architecture, b) its equivalent circuit.. ................... 8

Figure 2- 5: The Smith chart shows load and input impedance for the extended resonance

technique ......................................................................................................................... 9

Figure 2- 6: a) A negative-resistance balanced oscillator, b) its transformed feedback

oscillator ........................................................................................................................ 11

Figure 2- 7: An equivalent circuit of the negative-resistance technique .......................... 12

Figure 2- 8: The block diagram of the balanced oscillator .............................................. 19

Figure 2- 9: The signal flow graph of the balanced oscillator ......................................... 21

Figure 2- 10: The basic building block of a singly balanced FET mixer ......................... 26

Figure 2- 11: A block diagram of the Balanced Self-Oscillating Mixer .......................... 27

Figure 2- 12: A circuit diagram of the Balanced Self-Oscillating Mixer ......................... 28

Figure 2- 13: Simulated performance of return loss at RF .............................................. 28

Figure 2- 14: The anti-phase IF signals at the two outputs .............................................. 29

Figure 2- 15: Measurement set-up .................................................................................. 30

Figure 2- 16: Conversion gain vs. input power .............................................................. 30

Figure 3- 1: AM and FM components ............................................................................ 35

Figure 3- 2: Vector diagram of a carrier simultaneously phase and amplitude modulated

by the nth noise voltage interference .............................................................................. 36

Figure 3- 3: Spectrum of source in frequency domain .................................................... 37

Figure 3- 4: Leeson's model .......................................................................................... 37

Figure 3- 5: a) A equivalent circuit for the oscillator, b) equivalent circuit characterised in

both negative and feedback models and c) equivalent oscillator on right-hand side of the

balanced oscillator ......................................................................................................... 43

Figure 3- 6: The proposed injection-locked balanced oscillator configuration ................ 50

Figure 3- 7: The fundamental injection-locked balanced oscillator circuit ...................... 51

viii

List of Figures

Figure 3- 8: The sub-harmonic injection-locked balanced oscillator circuit .................... 51

Figure 3- 9: The measured free-running signal compared to the injection locking

frequency signals ........................................................................................................... 52

Figure 3- 10: The locking range of the fundamental injection-locked balanced oscillator

as function of the injection power level. ......................................................................... 53

Figure 3- 11: Experimental set -up for phase and magnitude measurement of the injection-

locked balanced oscillator output ................................................................................... 54

Figure 3- 12: The amplitude and phase response between two output ports of the ILBO as

locking on polar diagram ............................................................................................... 55

Figure 3- 13: a) Magnitude and b) Phase response of the fundamental injection-locked

balanced oscillator ......................................................................................................... 56

Figure 3- 14: The output signals between fundamental and harmonic injection-locked

balanced oscillator ......................................................................................................... 57

Figure 3- 15: The measured sub-harmonic injection locked frequency signal ................. 59

Figure 3- 16: The locking range as a function of the injection power .............................. 59

Figure 3- 17: Phase measurement test bench setup for sub-harmonic injection-locked

balanced oscillator ......................................................................................................... 60

Figure 3- 18: Measured phase difference between the two outputs of the sub-harmonic

injection-locked balanced oscillator ............................................................................... 61

Figure 4- 1: A block diagram of a cascaded oscillator .................................................... 65

Figure 4- 2: A single-ended oscillator circuit ................................................................. 66

Figure 4- 3: The cascaded oscillator circuit.. .................................................................. 67

Figure 4- 4: Free-running cascaded oscillator output without injection signal. ................ 68

Figure 4- 5: The output signal from the injection-locked cascaded oscillator: the injection

frequency is close to lx, (l/2)x and (l/3)x free-running frequency ................................. 68

Figure 4- 6: The output signal from the injection-locked single-ended oscillator: the

injection frequency is close to lx, (l/2)x and (1/3)x free-running frequency .................. 69

Figure 4- 7: The output spectrum from the single-ended oscillator during locking state

with respect to the injection signal at frequency of (1/2) x free-running oscillating signal

...................................................................................................................................... 70

IX

List of Figures

Figure 4- 8: The output spectrum from the single-ended oscillator during locking state

with respect to the injection signal at frequency of (1/3) x free-running oscillating signal

...................................................................................................................................... 70

Figure 4- 9: The output spectrum from the cascaded oscillator during locking state with

respect to the injection signal at frequency of (l/2)x free-running oscillating signal ....... 72

Figure 4- 10: The output spectrum from the cascaded oscillator during locking state with

respect to the injection signal at frequency of (l/3)x free-running oscillating signal ....... 73

Figure 5- 1: a) The physical form and b) the nonlinear model of the GaAs MESPETs .... 78

Figure 5- 2: PET output characteristics showing the region of operation for frequency

doubler operation ........................................................................................................... 80

Figure 5- 3: Waveforms and signal trajectory for class B multiplier ............................... 81

Figure 5- 4: The frequency doubler employing A / 4 transmission lines ......................... 82

Figure 5- 5: A balanced (push-push) doubler ................................................................. 83

Figure 5- 6: The circuit diagram of the novel sub-harmonic injection-locked balanced

oscillator-doubler configuration ..................................................................................... 85

Figure 5- 7: Single-ended doubler circuit configuration ................................................. 86

Figure 5- 8: The Wilkinson power divider ...................................................................... 87

Figure 5- 9: The output spectrum of the balanced oscillator-doubler circuit (with 3-dB

attenuator and without sub-harmonic injection-locking signal 0.5-dB cable loss at the

output) ........................................................................................................................... 88

Figure 5- 10: The comparison between the output of the balanced oscillator-doubler, with

and ................................................................................................................................. 89

Figure 5- 11: The measured locking range of the subharmonic injection-locked oscillator-

doubler as a function of injection power level ................................................................ 89

Figure 6- 1: Feedforward technique for a linearised power amplifier. ............................. 92

Figure 6- 2: Feedforward technique for fundamental signal suppression in a frequency

doubler .......................................................................................................................... 94

Figure 6- 3: Single-ended doubler circuit configuration ................................................. 95

Figure 6- 4: Single-ended doubler circuit ....................................................................... 96

Figure 6- 5: Filter configuration ..................................................................................... 97

Figure 6- 6: Simulation result of the filter ...................................................................... 97

x

List of Figures

Figure 6- 7: The reflection-type phase shifter configuration ........................................... 98

Figure 6- 8: The layout of reflection-type phase shifter at 10Hz .................................... 98

Figure 6- 9: The measurement result of the filter from the network analyzer .................. 99

Figure 6- 10: Output spectrum of the I-to-20Hz doubler without the feedforward

technique (with a 3-dB attenuator and 0.5-dB cable loss at the ouput) .......................... 100

Figure 6- 11: Phase and insertion loss of phase shifter ................................................. 101

Figure 6- 12: Output of the 1-2 GHz doubler with feedforward (with a 3-dB attenuator

and 0.5-dB cable loss at the ouput) ............................................................................... 102

Figure 6- 13: Measured output of the frequency doubler as a function of the input power

level. ............................................................................................................................ 102

Figure 6- 14: Feedforward technique for odd harmonic component suppression in

frequency doubler ........................................................................................................ 103

Figure 7- 1: Concept of the power dissipation and phase noise ..................................... 108

Figure 7- 2: A distributed oscillator ............................................................................. 108

Xl

List of Tables

List of Tables

Table 2- 1: The input and load coefficients with respect to different frequencies ............ 23

Table 2- 2: The length of the transmission line by simulation and calculation for different

frequencies .................................................................................................................... 24

Table 3- 1: The locking range with respect to the fundamental and harmonic injection-

locked balanced oscillator at power levels of -7.5 and -20 dBm .................................... 57

Table 4- 1: The locking range with respect to the injection power levels, where the

injection frequency is close to (l/2)x free-running frequency ......................................... 71

Table 4- 2: The locking range with respect to the injection power level, where the

injection frequency is close to (l/3)x free-running frequency ......................................... 71

Table 4- 3: The locking range for the single-ended oscillator with respect to the injection

power levels, where the injection frequency is close to (l/2)x free-running frequency ... 73

Table 4- 4: The locking range for the cascaded oscillator with respect to the injection

power levels, where the injection frequency is close to (l/3)x of free-running frequency74

Xll

Glossary of temlS

Glossary of Terms

N out

L«())m)

(S I N)in

(SIN)out

~8 nns,total

Ppavai/,nns

~8nns

Noise power applied at the output of the amplifier

Reactive energy going between Land C

Lowpass transfer function

Signal-to-noise ratio at the input of the amplifier

Signal-to-noise ratio at the output of the amplifier

Total phase deviation

Arbitrary phase angle

Available rms signal power

Equivalent available noise voltage due to the noise figure in the

amplifier

Angle of SIt

Capacitor charge on cpapcitor

External quality factor

Locking range

Noise power applied at the input of the amplifier

Phase deviation

Phase noise

Power dissipated in the resonator

Spectral density at the frequency at an offset frequency

Carrier voltage

Xll

F

11

81

82

Qc

lIN(jCO)

IL(jCO)

8T

IT

A(jco)

L

AM

Cds

Cgd

cgs

D

Frequency from the carrier

Modulation index

Noise figure

Noise voltage

Frequency in radian

Subtracting symbol

Phase of current in a balanced oscillator

Phase of current in a balanced oscillator

Carrier frequency in radian

Input coefficient

Load coefficient

Glossary of tenns

Electrical length of the transmission line in the terminating network

Terminating coefficient

Gain of the amplifier

Adding symbol

A growing signal

Amplitude modulation

A small noise signal generated in the circuit

Transfer function of resonator element

Capacitance

Drain-source capacitance

External capacitance

Gate-source capacitance

Gate-drian capacitance

Drain

xiii

dB

dBc

dBm

DC

e

fo

FET

PM

G

G

g

GaAs

GHz

HP

I(t)

IBLO

Id

IF

K

K

L

LMDS

Glossary oj temlS

Decibel

Decibel relative to carrier

Decibel relative to 1 mW

Direct current

Electrical length of the transmission line in a balanced oscillator

Operating frequency

Corner frequency

Field effect transistor

Offset frequency

Frequency Modulation

Gate

gain of the amplifier

Nonlinear function of active device

Gallium Arsenide

Drain-source transconductance

Gigahertz

Transconductance

Hewlett Packard

Current

Injection-locked balanced oscillator

Drain current

Intermediate frequency

Stability factor

Boltzmann's constant

Indactance

Local Multi-point Distribution Systems

xiv

LO

MESFET

MHz

q

Q

R

Rd

RF

Rg

Ri

RIN

RL

Rs

S

S

SIBLO

SSB

t

T

Vet)

Vd

Vds

Vg

Vgs

Glossary of tenns

Local oscillator

Metal Semiconductor Field Effect Transistor

Megahertz

Charge in Coulomb

Loaded quality factor

Resistance

Drain resistance

Radio frequency

Gate resistance

Resistance of the semiconductor region under the gate

Input resistance

Load resistance

Source resistance

Scattering parameters

Source

Subharmonically injection-locked balanced oscillator

Single sideband

Time

Temperature in Kelvin

Voltage

Small excitation voltage

Bias point of VI and VI­

Drain voltage

Drain-source voltage

Gate voltage

Gate-source voltage

xv

Vi

A

X

Zout

Glossary of terms

Input voltage

Wave length

Reactance

Input reactance

Load reactance

Impedance

50-ohm characteristic impedance in a transmission line

Characteristic impedance in a transmission line

Input impedance

Load impedance

Output impedance

XVI

Chapter I Introduction

Chapter 1

1 Introduction

An oscillator is often required to generate the local oscillator (LO) signal for many circuit

building blocks such as mixers and modulators. An ideal oscillator would produce a

signal at a fundamental frequency with an infinite quality factor. However, the oscillator

practically produces noise due to the noise sources in passive and active components.

Therefore, the performance of these circuits is not only dependent on the circuit blocks

but also depends on the quality of the LO signal. One way to characterise the oscillator

quality is to quantify the residual noise energy at a specified offset from the carrier [1].

This noise is called phase noise.

In order to achieve the purity in oscillator, many techniques used to improve the phase

noise in oscillators have been studied. One proposed technique applied to stabilise the

oscillating signal is called the injection-locked oscillator [20]-[39]. This thesis studies the

injection-locking technique used in a balanced oscillator. Moreover, using the frequency

multiplier one can extend the fixed frequency generated by oscillator. The other important

research point is to study the frequency doubler considerations. There are several

techniques to achieve gain and to obtain a high purity signal for frequency doublers [54]­

[61], [64]-[72]. In this research, a new technique to suppress the unwanted signal at the

frequency doubler output has been introduced.

In chapter 2, the general procedure for microwave oscillator design is introduced. The

common-source PET configuration with series feedback is used to generate the negative

resistance. The terminating network is determined in order to satisfy the oscillation

conditions. In the steady state, a sinusoidal signal is generated. By using the negative­

resistance technique and the extended resonance technique, a balanced oscillator is

produced. The concept of the balanced oscillator is that two active devices resonate each

Chapter 1 Introduction

other by means of an inter-connecting transmission line to achieve the oscillation

conditions. Furthermore, a self-oscillating balanced mixer is chosen as one of the possible

balanced oscillator applications. The experimental results are also demonstrated.

Chapter 3 presents the phase noise in the oscillators by using Leeson's model. To achieve

low phase noise, many important parameters are considered such as temperature, noise in

the active device and the tank circuit or resonator. However, there is another technique to

improve phase noise in oscillators. This technique is known as the injection-locked

oscillator. The main concept for this technique is to apply an external stabilizing signal

into the oscillator. Once the oscillator is in the locking state, the oscillating signal

becomes more stable and the phase noise is improved. The injection lock technique was

firstly studied by R.Adler [20]. Later, the injection-locking phenomenon was extended by

K.Kurokawa [21]-[22]. This technique is used in the balanced oscillator to improve the

stability of the oscillating signal. The injection signal where the injection frequency is

close to free-running frequency and close to (lin) x free-running frequency is injected

into the balanced oscillator at the center of the transmission line. By using Kurokawa's

theory, the locking range is analyzed. The injection-locked balanced oscillator in both

fundamental and subharmonic modes was observed. The experimental results show the

phase noise improvement in both fundamental and subharmonic injection-locked

balanced oscillators. Furthermore, the locking range was measured. The locking range

and phase noise suppression corresponding to the injection power level and the injection

frequency agree with the analysis.

Spurious signals in the injection-locked oscillator are investigated. A single-ended

oscillator shows the unwanted signals occurring at the output under the locking state

where the injection frequency is close to (1/2) x and (1/3) x the free-running frequency.

Chapter 4 shows the unwanted signals are dramatically reduced in the injection-locked

cascaded oscillator. A cascaded oscillator was designed by employing the power

combining technique. Moreover, the locking range is also observed. It is showed that the

locking range in the single-ended oscillator is higher than that in the cascaded oscillator.

2

Chapter 1 Introduction

Chapter 5 introduces the frequency doubler design considerations. In order to achieve the

high gain in the FET frequency doublers, the biasing point is considered. The advantages

of single-ended and the balanced configurations are also presented. In addition, the

integrated subhannonic injection-locked balanced oscillator-balanced frequency doubler

is demonstrated as an example application. Due to the injection-locked technique the

phase noise of the frequency doubler is improved. The locking range as a function of the

injection power level was also measured.

Frequency multipliers are widely used in order to extend the frequency limit of fixed or

variable frequency low phase-noise oscillators. Chapter 6 presents a novel technique to

eliminate the fundamental frequency component at the output of the frequency doubler.

To obtain the high purity signal from frequency doubler, the feedforward technique often

used in amplifiers [73]-[78] is used in the frequency doubler. The important concept of

this technique is the phase and magnitude balance in the reference and frequency doubler

paths. The experimental results show that the frequency doubler provides 0.5-dB gain.

Additionally, the fundamental component at the output is suppressed by more than 50

dBc.

Finally, the research works covered by this thesis is concluded in chapter 7. A summary

of achievements is given and the future work areas are suggested. One possible topic is

the power combining concept to obtain high oscillating power and the proportion between

the power of the signal from the carrier and the power of the carrier signal is reduced,

giving low phase noise. The other feasible idea is to suppress the odd harmonics in a

frequency doubler by using the feedforward technique [40]. This can be extended from

the work mentioned in chapter 6 by employing an error loop. This loop contains the odd

harmonic components. With the suitable magnitude and phase between the output from

the first loop and the output from the error amplifier in the error loop, the odd harmonics

can be cancelled by using a subtractor or combiner.

3

Chapter 2 Balanced Oscillators

Chapter 2

2 Balanced Oscillators

2.1 Introduction

A balanced oscillator can be used in many microwave applications such as balanced

mixers, phase detectors, frequency doublers, differential frequency dividers and so on.

The balanced oscillator generates two symmetrical outputs of equal amplitudes and 1800

phase difference [2]. These properties are inherent in the balanced oscillator. However,

the degree of the balance relies on how identical the active devices can be made.

For oscillation to occur either the Barkhausen feedback criterion [3] for the open-loop

gain, or the negative resistance condition must hold. For the second of these, the

oscillation condition can be determined in terms of input and load impedances. This can

also predict the approximate oscillation frequency, but it cannot predict the output power

level due to the non-linearity of the active devices in the oscillator. As a result of the non­

linearity, the operating frequency can also shift.

In this chapter, the balanced oscillator principle and topology are introduced. The

principle is that two active devices resonate each other by means of an inter-connecting

transmission line to achieve the oscillation condition. To design the balanced oscillator,

the device s-parameters are used. The analysis shows what length of the lossless

transmission line that is needed to achieve the oscillation condition. Also, the output

signals of the balanced oscillator in terms of phase and amplitude are examined. For

identical active devices, the analysis shows that the balanced properties will occur for the

oscillating signal no matter what the characteristic impedance of the transmission line is.

4

Chapter 2 Balanced Oscillators

To investigate the balanced properties of the balanced oscillator, a self-oscillating

balanced mixer is chosen as one of the possible balanced oscillator applications. Just as

for an ordinary balanced mixer, the main self-oscillating balanced mixer parameters are

considered. The other important key factor for the self-oscillating mixer is the biasing

point. This must be chosen to ensure the mixing mechanism occurs in order to obtain

conversion gain. Therefore, the experimental results are also demonstrated.

2.2 Oscillation conditions

The basic oscillation condition is that the circuit should be unstable. With a simple

negative feedback oscillator model [3] as shown in figure 2-1, an instability condition for

generation of the periodic signal is given by the Nyquist theory [4]. The open-loop gain is

expressed as:

A(jm)B(jm) -1 = 0 (2.1)

or A(jm)B(jm) = 1

Where A(jm) represents the gain of the amplifier and B(jm) represents resonator element.

.-------1A(j m) >------,

Figure 2- 1: A simple negative-feedback oscillator model

The interpretation is that the periodic signal occurs when the open-loop gain is at least

equal to unity, and this is the expression of the oscillation condition. However, the non­

linear amplifier also can effect the oscillation condition. Then the value of the open-loop

gain may no longer be unity. To solve this problem, the amplifier gain must be sufficient

to guarantee the oscillation condition. The resonator also ensures such that the oscillation

can occur at the desired frequency.

5

Chapter 2 Balanced Oscillators

In general microwave circuits, the oscillation condition can be viewed as the open-loop

function [4]. Figure 2-2 a) and b) introduce a microwave circuit and its signal flow graph,

respectively. The closed-loop and open-loop gain can be determined from the signal flow

graph as follows:-

anrIN (jm) a = for the closed-loop gain

L 1-rIN (jm)rL (jm) (2.2)

(2.3)

for the open-loop gain

Where r/N(jw) r L(jW) represent the input and load coefficients, an and aL represent a

small noise signal generated in the circuit and a growing signal.

a) b)

Figure 2- 2: a) A microwave oscillator circuit, b) its signal flow graph

The periodic signal is generated when the open-loop gain is equal to unity. From this, the

circuit can be viewed as the input and load impedance. The magnitude of the input

impedance is negative so it is easier to model as the negative resistance concept shown in

figure 2-3 a). In practice, the amplifier and the terminating network form the negative

resistance part demonstrated in figure 2-3 b). This approach is called the two-port

6

Chapter 2 Balanced Oscillators

negative resistance. It is necessary to examine the stability factor, K, for the active device.

The stability factor should be less than unity to guarantee the possibility of oscillation

condition [4]-[5]. The linear oscillation condition becomes:

(2.4)

Where ZIN(A,w) is the input impedance, depending on the amplitude and frequency and

ZL( w) is the load impedance, depending on the frequency.

From this, the prediction of the oscillating frequency is obtained by separating the

negative resistance and the resonant parts. This analysis is accomplished by using the

small-signal model; however, this analysis cannot determine the accurate operating

frequency and output power due to the non-linear behaviour of the active device. The

reactances within the device change significantly during the oscillation, resulting in a

shifting frequency.

i(t)--.

+

XL«({)) v(t)

XlN«({))

RL«({)) RlN«({))

ZL«({)) ZIN«({))

(a)

Input port Terminating port

• Load Transistor Terminating

network (S) +- -. network .- r+

(b)

Figure 2- 3: a) Negative resistance oscillator, b) Two-port negative resistance approach

7

Chapter 2 Balanced Oscillators

2.3 Balanced oscillator analysis

Figure 2-4 a) shows the proposed resonant technique used in balanced oscillator

configuration, and figure 2-4 b) shows its equivalent circuit using the negative resistance

oscillator model. Here, we present two MESFET oscillators using a single layer

microstrip circuit. By using the extended resonance technique [6], the length of the

transmission line is chosen in order that the each device resonates one another. Also, in

the steady state, 1800 phase difference of the free running oscillation signals is achieved

from the output ports [7].

portl' port 2 and port 2'

Zo,e

Zo Zo

[s] [s]

a)

1 -.iX(A,ro) 1

-jX(A,ro)

-R(A,ro) -R(A,ro}

b)

Figure 2- 4: a) The balanced oscillator architecture, b) its equivalent circuit

8

Chapter 2 Balanced Oscillators

An easier way to understand the oscillation condition is to view the balanced oscillator as

an impedance on the Smith chart and this has been studied by K.S.Ang [7]. A two-port

negative resistance approach [4] is used to generate the input impedance, ZlN] or ZlN2.

Either ZLl or ZLZ represents the load impedance. The input impedance and the load

impedance in this case are voltage dependent and frequency dependent. By using a

transmission line, the input impedance of the first device, ZlNlA, OJ), is transformed to the

load impedance ZL2(A, OJ), seen by looking at the other end of the transmission line. It is

conjugate. The input impedance and the load impedance can be expressed as:

(2.5)

ZLl(A,OJ) = ZL2(A,OJ) = -R + jX (2.6)

The impedance transformation is clearly depicted on the Smith chart as shown in figure

2-5. In addition, in the steady state the oscillation condition becomes:

ZINl (A,OJ) + ZLI (A, OJ) = 0 or vice versa (2.7)

Figure 2- 5: The Smith chart shows load and input impedance for the extended resonance technique

9

Chapter 2 Balanced Oscillators

Also, the balanced oscillator can be analysed with the feedback model. Figure 2-6 a) and

b) show the balanced oscillator viewed as the feedback model [10] and its equivalent

circuit. Here, the active device functions as the amplifier, which is a voltage-controlled

current source. The output current is fed back to the resonant network where the negative­

resistance model is charaterised as presented in figure 2-4. In this case, the resonant

network consists of the linear and nonlinear components. It is clearly shown that the two

active devices in the balanced configuration resonate with each other.

10

a

Terminating network

~r T

Chapter 2 Balanced Oscillators

a)

b)

Figure 2- 6: a) A negative-resistance balanced oscillator, b) its transformed feedback oscillator

11

Chapter 2 Balanced Oscillators

Next, we consider the negative resistance looking into the gate and source terminals.

Figure 2-7 b) shows an equivalent circuit of the negative-resistance technique used in the

oscillator. An external capacitance, Cf, is used to connect at the source of the field effect

transistor (PET). To evaluate the value of the resistance looking at the gate-source

terminal, it is assumed that the impedance of the gate-drain capacitance, Cgd, is much less

than that of the gate-source capacitance, cgs , so at high frequency the gate-drain

capacitance is negligible. In other words, it is assumed that no current passes through this

capacitance at high frequency.

a)

Gate Rg Cgd Rd Drain

NW +1

I V Cgs i d (Vg) = g m Vg + g

Cds Vd g ds

iii Ri

r v. I

Zin Rs

Cf T Source Source

b)

Figure 2- 7: An equivalent circuit of the negative-resistance technique

12

Chapter 2 Balanced Oscillators

From the equivalent circuit, we obtain

(2.8)

(2.9)

(2.10)

(2.11)

Where

dId I g m = -- V g = v go and S = j OJ dVg

From equation (2.8) to (2.11), we get

(2.12)

Finally, the impedance looking at the gate-source terminal can be expressed as follow:

V· Z. =R +_1 In g

1i

Vg +SCgsRiVg +-l-vg(gm + scgJ+Rsvg(gm +scgJ = R + _______ s_c..::....f ___________ _

g

(2.13)

13

- Chapter 2 Balanced Oscillators

Since

Z. = Re[Z. ]+ Im[Z. ] In In In

(2.14)

(2.15)

Equation (2.14) and (2.15) show the real and imaginary parts of the impedance looking at

the gate-source terminal of the active device, respectively.

Due to the nonlinearities in the MESFET equivalent circuit, the eN or QN characteristic

of an element has been examined [9]. A nonlinear capacitance can be presented as the

nonlinear dependence of charge on voltage: -

(2.16)

Where Qc is the capacitor charge. By usmg a Taylor senes, the charge function is

expanded. Then, subtracting the dc component of the charge, the small-signal component

of charge is:

(2.17)

14

Chapter 2 Balanced Oscillators

The small-signal current is the time derivative of the charge and is given as

. dq l=-

dt

(2.18)

For simplicity, the small-signal current can be expressed as

(2.19)

The MESFET's channel current represents the controlled current, id, and the gate-drain

conductance, gds. The controlled current is gate voltage and drain voltage dependent;

therefore, a two-dimensional Taylor series is used to expand the function. Then the dc

current component is subtracted from the function and the controlled current is given as

follows:

. af af 1 (a 2 f 2 a 2 f a 2 f 2 J l =-v +--v +- --v +2 vv +--v d av 1 av 2 2 av2 1 avav 1 2 av2 2

1 2 1 1 2 2

(2.20)

where v1 and v

2 are the small excitation voltages. V1•0 V2•0 are the bias point of Viand

Since function f is represented by I d (Vg , Vd ) , where the capital letters represent the large

signal voltages and currents while the lower letters represent incremental ones. It is

assumed that the magnitude of any product between V1 and V2 is small, so it is negligible.

Then, we obtain

15

Chapter 2 Balanced Oscillators

(2.21)

For the simplicity, we can express this as follows:

(2.22)

Where g m and g d are the transconductance and the drain conductance.

From the equivalent circuit, we obtain:

(2.23)

(2.24)

(2.25)

Substituting equation (2.23), (2.24) and (2.25) into equation (2.9), we get

16

Chapter 2 Balanced Oscillators

_ (() 2 3 )dVg Vi - Vg + Ri Cgs1 Vg + Cgs2 (Vg )Vg + Cgs3 (Vg )Vg + CgS4 (Vg )Vg dt

[( 2 3 2 3 J + R g ml V g + g m2 V g + g m3 V g + g dl V d + g d2 V d + g d3 V d) +

s ( () 2 3 )dV g C gsl Vg + C gs2 (Vg )V g + C gs3 (Vg )V g + C gs4 (Vg )V g ---:it

(2.26)

Finally, the impedance looking at the gate-source terminal can be expressed as follow:

+ Vg

( () 2 3 )dV g

C gsl Vg + C gs2 (Vg )V g + C gs3 (Vg )V g + C gs4 (Vg )V g -­dt

( () 2 3 )dV g

C gsl Vg + C gs2 (Vg )V g + C gs3 (Vg )V g + C gs4 (Vg )V g dt

( () 2 3 )dV g

C gsl Vg + C gs2 (Vg )V g + C gs3 (Vg )V g + C gs4 (Vg )V g dt

(2.27)

17

Chapter 2 Balanced Oscillators

Comparing equation (2.27) to (2.13), the real and imaginary parts of the input impedance

looking into the gate and source terminals are expressed as:

and

1 Im[ZJ=-SCI

+

( () 2 3) dv g C gsl Vg + C gs2 (Vg )V g + C gs3 (Vg )V g + C gs4 (Vg )V g ----;Jt

( () 2 3 )dV g

C gsl Vg + C gs2 (Vg )V g + C gs3 (Vg )V g + C gs4 (Vg )V g ----;Jt

( () 2 3 )dV g

C gsl Vg + C gs2 (Vg )V g + C gs3 (Vg )V g + C gs4 (Vg )V g -­dt

(2.28)

(2.29)

The value of the resistance is very dependent upon the values of transconductance, gm,

gate-source capacitance, cgs, and external capacitance, Cf. The transconductance and gate­

source capacitance are nonlinear elements and their values also relies on the biasing. In

order to get the negative resistance by this technique; therefore, the external capacitance

and dc biasing must be considered carefully.

Now, the phase difference between the outputs of the balanced oscillator is determined.

The use of the z parameters [4], [8] of the transmission line is considered. Figure 2-8

illustrates the block diagram of the balanced oscillator.

18

Two-port

Negative

Resistance :::J

f---I-+----

Z _ parameters _ of _ the

transmissi on _line, Zo,f)

Chapter 2 Balanced Oscillators

Two-port

Negative

Resistance

Figure 2- 8: The block diagram of the balanced oscillator

By using the Kirchhoff's law, we get

(2.30)

Thus, equation (2.30) can be rewritten as

(2.31)

(2.32)

Where Zll, Z12, Z21 and Z22 are z-parameters of the transmission line connected between the

two active devices. The values of these parameters are Zl1 = Z22 = -j Zo cot(), Z12 = Z21 = -j Zo csc(), and IlJl=lhl=I. Multiplying by lie jeJ in equations (2.31) and (2.32). Then

equations (2.31) and (2.32) become:

_ Z j(B2-BI) = Z j(B2-BI) + Z IN2e lie 12

(2.33)

Z - Z j(B2-BI) Z INI - 21 e + 22

(2.34)

19

Chapter 2 Balanced Oscillators

Due to the symmetrical configuration, the value of ZlNI is equal to value of ZlN2. If one

subtracts equation (2.34) from equation (2.33) and then substituting the values of the z­

parameters, we obtain:

( Z ·Z t8 ·Z 8) }(82-8l) Z ·Z 8 . - IN2 +} 0 co -} 0 csc e = - INl +} 0 cot - }Zo csc 8

(2.35)

e}(82-81) = 1 (2.36)

82 = 81 or 81 ± 2n7Z" (2.37)

where e}(82-81) = cos(82 - 81) + j sin(82 - 81) and n = 1, 2, ... , n

With the directions of the two currents as depicted in figure 2-8, the equation (2.37)

shows that 82 is equal to either 810r 81 ± 2n7Z". It can be described that no matter what

the value of 81 is, the value of 82 is equal to either 810r 81 ± 2n7Z". Thus, the net

direction of the two currents flowing through the transmission line still maintains anti

phase. It is clearly seen that under the steady state the phases of the currents in this circuit

as shown in figure 2-8 agree with the analysis.

However, the characteristic impedance of the transmission line could not be equal to 20,

50 ohm. Zc denotes the characteristic impedance of the transmission line, which is not

equal to Zoo The z parameters of the transmission line become Zll = Z22 = -j Zc cote, Z\2 = Z21 = -j Zc csc8, where the angle 8 is the electrical length of the transmission line. Then,

the equation (2.35) is rewritten as

( Z ·Z 8·Z 8) }(82-8l) - Z ·Z t 8·Z 8 - IN2 +} c cot -} c csc e - - INI +} c co -} c csc (2.38)

By comparing equation (2.38) to equation (2.35) we can see that the current directions are

out of phase no matter what the characteristic impedance of the transmission line is.

To find the required electrical length of the transmission line such that the oscillation

conditions for the balanced oscillator are satisfied, a different plane parameter [4] is used.

20

Chapter 2 Balanced Oscillators

It is assumed that the characteristic impedance of the transmission line is Zo and the angle

e is the electrical length of the transmission line between port 1 and port 1'. The

scattering matrix at the reference planes at port 1 and port 2 is expressed as:

(2.39)

Assuming travelling waves on a loss-less transmission line, the scattering matrix at the

reference plane at port 1 'and port 2' is written as:

S e-j8

] 12

S22

(2.40)

To analyse the electrical length of the transmission line connected between the two active

devices, the oscillation condition is considered. A convenient technique to present and

analyse the oscillation condition is to use the knowledge of the signal flow graph. The

use of Mason's rule [4]-[8] can analyse the signal flow graph of the balanced oscillator as

shown in Figure 2-9.

a' 1

h' 1

h' 2

I a' S12 2

Figure 2- 9: The signal flow graph of the balanced oscillator

21

Chapter 2 Balanced Oscillators

The input and load coefficients are given:

(2.41)

(2.42)

Since rn = rT2 = rT • We assume that the characteristic impedance of the transmission

line connecting between the drain of the active device and the load, 20, is equal to Zoo

Then the value of rT is as shown below:

r. - 0 T-(2.43)

For the balanced oscillator the oscillation condition must satisfy equation (2.5) and (2.6),

and we obtain:

(2.44)

Where IlN2 = IrlN21LIIN2 and r L2 = IIIN21L - r 1N2 · Substituting the value of IlN2 and

r L2 ' the equation (2.44) then becomes:

1 12 j(2LSll-2B)_1 21

Sl1 e - Sl1 (2.45)

2Lsl1

- 28 = +2N1l, where N = 0,1,2, ... (2.46)

22

Chapter 2 Balanced Oscillators

Thus, the electrical length of the transmission line is calculated as given as:

e = LS ll + 1l (radian) , where LSll is the angle of Sll (2.47)

Due to the balanced structure, the oscillation condition of the other side of the oscillator

can be analysed in the same fashion as described above. The electrical length of the

transmission line connected between the two active devices is equal to 1l + Ls 11 radians,

but the transmission line, in general, has loss due to finite metal conductivity and

dielectric tan J. The actual electrical length might not necessarily be the same as the

length given by equation (2.47). In order to consider an accurate value for the length of

the transmission line, the loss due to the finite conductivity and the dielectric material has

to be determined.

Using the Infineon CFY30 GaAs MESFET as an active device and FR4 as a substrate,

series feedback is used at the source in order to achieve the negative input impedance.

The input and load coefficients are given in Table 2-1. Table 2-2 shows an example of the

electrical length of the transmission line from simulation and calculation corresponding to

various different frequencies. The simulation results confirm the accuracy of the analysis

shown in equation (2.47).

Frequency (GHz) rIN rL

1.8 1.018L-35.861 ° 1.018L35.861 °

1.9 1.042L-42.561 ° 1.042L42.561 °

2.0 1.070L-49.077° 1.070L49.077°

2.1 1.100L-55.451 ° 1.100L55.451 °

2.2 1.134L-61.722° 1.134L61.722°

Table 2- 1: The input and load coefficients with respect to different frequencies

Chapter 2 Balanced Oscillators

Length of the Length of the

Frequency (GHz) transmission line by transmission line by

simulation (J.Lm) calculation (J.Lm)

1.8 36160 36215

1.9 32645 32702

2.0 29500 29583

2.1 26740 26792

2.2 24230 24277

Table 2- 2: The length of the transmission line by simulation and calculation for different frequencies

2.4 A self-oscillating balanced mixer

As a practical application of the balanced oscillator a balanced self-oscillating mixer is

now presented. A mixer is an important block in communication systems. The balanced

mixer has many advantages over the single-ended mixer such as the RF-LO isolation [5],

[9]. There have been many studies on balanced mixers and self-oscillating mixers [12]­

[19]. The balanced self-oscillating mixer has inherent oscillating signals and functions as

the mixer at the same time by the non-linear mechanisms in the device. The other

advantage of the self-oscillating mixer is that the RF signal is down converted to the IF

signal without a balun or external oscillator signals.

Typically, the singly balanced mixer requires an external LO signal with the external

balun or 180-degree hybrid. There are many important parameters in the balanced mixer.

The design considerations for the balanced mixer are examined. Similar to the balanced

mixer, the self-oscillating mixer requires the same performances. In this case, the biasing

point is carefully considered in order to obtain the conversion gain.

24

Chapter 2 Balanced Oscillators

The design of the self-oscillating mixer is concerned with the inherent oscillating and

mixing mechanism. The balanced oscillator configuration and the gate mixer topologies

are used to accomplish the purpose. Furthermore, this design topology is also expected to

maintain the balanced properties under the mixing operation by employing the

symmetrical configuration. The return loss at the RF port has been taken into account so

that the maximum RF signal can be taken into the circuit.

Additionally, the phase and amplitude of the IF signals at the drains of PETs also have

been investigated. Then, the IF signal was subtracted by using an external balun. The

measured results demonstrate the conversion gain as a function of the RF power levels

with different biasing points. The advantages of this technique are the small size of the

circuit, reduced power consumption and simple design.

2.4.1 Design considerations for a balanced mixer

The basic building block of a singly balanced PET mixer is typically composed of a 180-

degree hybrid for the RF and LO inputs, PET mixers and an other 180-degree hybrid to

combine the IF signals. Figure 2-10 shows a 180-degree hybrid mixer. The RF and LO

signals are applied to the identical mixers via the hybrid. The LO signals injected to the

PETs have 180-degree phase difference whereas the RF signals are in-phase. The 180-

degree out-of -phase IF signals are subtracted by using either a balun or hybrid. The

advantage of this configuration is that it gives RF-to-LO isolation due to the use of the

hybrid.

Although the balanced mixer topology provides the two mutual isolated ports of the 180-

degree hybrid, there are the other properties to be carefully considered. First is the AM

noise-rejection properties. In an ideal balanced mixer, this property is limited by the

balance of the hybrid and the AM noise in the LO signal [5], [Ill The other is the RF-to­

IF isolation. This can be achieved by using either the IF balun or hybrid, where the two

out-of-phase IF signals are subtracted.

25

Chapter 2 Balanced Oscillators

FET MIXER

~ IF RF

180-degree 180-degree HYBRID HYBRID

LO

FETMIXER

Figure 2- 10: The basic building block of a singly balanced FET mixer

A self-oscillating balanced mixer functions as both the balanced mixer and balanced

oscillator simultaneously. Similar to the typical balanced mixer, a self-oscillating

balanced mixer requires the same considerations. However, the other important

consideration for the self-oscillating mixer is that the balanced oscillator is not locked to

the injected RF signal. That means the circuit must perform as the self-oscillating mixer

rather than the injection-locked oscillator. The mixing operation can be achieved by

considering the device non-linearity. The important key controlling operation is the

device non-linearity, which has been studied in many previous works [5], [11]. The self­

oscillating mixer can provide the conversion gain at a bias condition where the non­

linearity is strong. The dc-bias operating point for a PET mixer is considered [13]-[15].

Furthermore, the conversion gain/loss is also dependent on the circuit topology of the

mixer [5], [11].

2.4.2 Circuit realisation

A balanced self-oscillating mixer, shown in Figure 2-11, consists of two important parts,

namely, a common-source feedback oscillator and a gate mixer. Series feedback is at the

source of the PET so that the instability for the active device is achieved. Then, the use of

the two-port negative resistance approach is applied to build the oscillator. From equation

(2.48), the circuit meets the oscillation conditions, which are generally described in many

literatures [4], [5]:

26

Chapter 2 Balanced Oscillators

Re[ZTNI (A,m)] + Re[ZLl (A,m)] < 0 or vice versa (2.48)

Im[ZTNI (A,m)] + Im[ZLl (A,m)] = 0

Balanced Oscillator

0 180 degree + degree

Gate Mixer

Figure 2- 11: A block diagram of the Balanced Self-Oscillating Mixer

To design the self-oscillating balanced mixer as illustrated in figure 2-12, each part of the

circuit is examined. Firstly, the balanced oscillator is designed. The design method as

described in section 2.3 is used. The Infineon CFY30 GaAs PETs are used as active

devices. The circuit is fabricated on FR4 substrate.

Next, the LO and RF rejection is determined. In order to eliminate the LO and RF signals

at the drains of the active devices, a low pass filter is required. This filter is designed to

function as the matching network for the oscillation condition as well as to perform as the

low pass filter for the IF signals.

Finally, the RF port is considered. The RF matching is examined in order that the

maximum RF signal can couple into the circuit. In this case, a coupled transmission line is

used. This coupled line functions as the band pass filter for the RF signal. It is connected

at the mid-point of the transmission line connected between the gates so that the circuit

can maintain the balanced configuration during oscillation and mixing operations. The

27

Chapter 2 Balanced Oscillators

return loss of the RF port is shown in figure 2-13. This is the simulated result when 50-

ohm loads are used to terminate the output ports of the circuit.

Z --R-jX V I~ GS..=.

~ -

IF -IF

RF

Figure 2- 12: A circuit diagram of the Balanced Self-Oscillating Mixer

0.0 (

-5.0 , , '

- - - - - - - -- - ---- - - - ---~ - - - ---- - ------ - - -- -;-- ----- - ----- - ------ -- ~ ---- -----------------

I 1 1

, ' ' --------------------1---------------- ----:--------------------1---------------------, , ' , , ' , ' ' , , ' , ' ' , ' ' : : : , , ' , ' ' , ' , ' , ' ' --------------------1---------------- ----:--------------------1------------------ ---, : '

-10.0

-15.0

-20.0 0.0 4.0

Frequency 1.0 GHzlDIV

Figure 2- 13: Simulated performance of return loss at RF

28

Chapter 2 Balanced Oscillators

2.4.3 Results and discussion

The self-oscillating balanced mixer was built and tested. The DC bias was provided by

bias-tees. The biasing point was considered in order to get the significant effect on the

mixing process due to the non-linearity of the devices, where a mixing operating point

close to class B is preferred [13]-[15]. This could enhance the conversion gain. The

circuit was first terminated at the RF port with 50-ohm load. The circuit then provided

1.85-GHz oscillating frequency with -5.2-dBm amplitude, which was corresponding to

the designed oscillation frequency, and the fact that the LO output is suppressed due to

the low-pass filter at the output of the circuit. The leakage of oscillating signal at the RF

port was measured by terminating with 50-ohm loads at the two outputs. The value of the

leakage signal was -20 dBm.

Next, the waveform of the IF signal from the drains of the devices was investigated. A

1.78-GHz RF signal was fed into the RF port. Figure 2-14 illustrates the waveform of the

IF signals. It is noticed that the amplitude and phase of the IF signals from the drains are

equal and 180-degree out-of phase.

10

20

;; 30 e

'-" QI "0 0 = ..... '8 ell -10 = ~

-20

-30

120.0mV 220.0mV

'I . :1

1

r

" '. ~ ..... '::: .. -

; ....... Ii .. . I:

.l-O.OOs S.OOns/

,.. ,

· . , ! .. , ~" ....... : . " . · . · .

1.. ' ..... ,' . A .". ,!,~".,,, i." i,.,.> .... ,-,.,

~

t

Vp-p(I)=8S.62mV Vp-p(2)=83.12mV Freq(2)=70.0SMHz

Time(ns) S.OOns/

Figure 2- 14: The anti-phase IF signals at the two outputs

29

Chapter 2 Balanced Oscillators

The conversion gain of the balanced self-oscillating mixer was observed. Figure 2-15

shows the experimental set-up. In this case an external IF balun was used to subtract the

two IF signals in order to obtain the conversion gain for the down-conversion. The

conversion gain as a function of the RF input power levels for difference biasing points

was measured. Figure 2-16 indicates the measured conversion gain. The mixer conversion

gain is flat up to an RF power level of -15 dBm. The maximum conversion gain is

approximately 2.5dB.

A Balanced IF signal signal .... IF Spectrum RF

~

~ "'0 '-" ~ ..... ro 01)

~ 0 ..... CI.l

""" ~ ;> ~ 0

U

...,... .... Self-Oscillating

... ..... ....... analyser .... Balun

Mixer -...

Figure 2- 15: Measurement set-up

5

.J.. * ~ 0

-5

-10

0" ~u v U V~ ~ ~ " A. A

~~ \ "

\~ ~ (r~ i * ~

\ \

-15

-20

-25

-30

-35

\

\ \\ \ ~

-*- Vds = O.9V , Vgs = -O.05V 1\1 -e- Vds = 1.0V ,Vgs = o.ov \\ ---*-- V ds = 1.2V ,v gs = o.ov

\ -40

-30 -28 -26 -24 -22 -20 -18 -16 -14 -12 -1 0

Input power (dBm)

Figure 2- 16: Conversion gain vs. input power

30

Chapter 2 Balanced Oscillators

As seen from the experimental results, this designed circuit topology gIves many

advantages. One of these is the good RF-to-LO isolation due to the small amount of the

LO leakage power level at the RF port, but the RF matching is narrow band. In addition,

the designed circuit is simple: using the balanced configuration to maintain the balanced

properties of the balanced mixer. From this it can be stated that the directions of the RF

signal looking at the two gates of the PETs are in-phase, mixing with the ISO-degree LO

signals, since the IF signals obtained from the drains are ISO-degree out of phase. The

other point is that the circuit consumes less power and the circuit size is smaller compared

to the typical balanced mixer since the external oscillator and external balun for the lS0-

degree for the LO input are not required.

2.5 Conclusion

This chapter has introduced the balanced oscillator technique. To do this, the oscillation

condition was considered. This condition can both be viewed as the simple negative­

feedback oscillator and negative resistance oscillator models. In microwave design, the

two-port negative resistance is used to ensure the instability of the device. Next, the

negative-resistance approach is described. By utilising the external capacitance

connecting at the source of PET, the negative resistance is produced. Its value is

dependent on nonlinear and linear components.

The balanced oscillator can be viewed as the feedback model, where the active device

behaves as an amplifier generating the harmonic signals and the feedback network part

can be considered for the oscillation condition. This model is explained more in next

chapter. Additionally, some analyses have been presented in order to demonstrate the

balanced properties of the circuit where the two active devices are identical and to

calculate the length of the transmission line used in the balanced oscillator design

topology. The simulated electrical length well agrees with the calculated one.

A self-oscillating balanced mixer was introduced to demonstrate the balanced oscillator in

a practical example. The biasing point in the self-oscillating mixer circuit was determined

since it ensures the mixing operation. As the results of the experiment, the self-oscillating

balanced mixer provided a 2.S-dB maximum conversion gain. The balanced properties

31

Chapter 2 Balanced Oscillators

were investigated and the results agree with the analysis. Furthermore, the LO signal

power level at the RF port was -20 dBm, which shows the good RF-to-LO isolation.

32

Chapter 3 Injection-locked Balanced Oscillators

Chapter 3

3 Injection-Locked Balanced Oscillators

3.1 Introduction

Local oscillator phase noise is a key performance parameter in a communication system,

since, for example, it effects the rejection of adjacent channel interference (AeI) and the

ability to detect weak signals, called the receiver sensitivity. Therefore the low phase

noise oscillator signal is needed in the next generation of millimetre wave communication

system. A technique popularly used to stabilise the free-running frequency is the

injection-locked oscillator. This also introduces an improvement of the phase noise in the

oscillator. The prospect of using the synchronous oscillator to a lock signal at the same

frequency (free-running frequency) or a higher frequency (harmonic frequency) have

been widely studied in the literature [20]-[39]. The fundamental and sub-harmonic

injection-locked technique has been also studied as a particular technique for optical

synchronisation of the remote local oscillator in microwave and millimetre-wave

applications [23]-[24]. Recently, an injection-locked push-pull or balanced oscillator was

proposed and applied to a spatial power combining array antenna [25]. Though the

structure proposed in [26] can be applied with external fundamental and sub-harmonic

injection signals, its main disadvantage is a large circuit area due to a use of a

transmission line to achieve 1800 phase difference outputs.

In this chapter, the explanation of phase noise theory is introduced. Leeson's model [47]

is generally used to describe phase noise. This reveals the important parameters which

effect the phase noise performance in oscillator. Therefore, these parameters, such as the

Q factor of the tank circuit, are determined for the oscillator design. Also, the injection­

locked balanced oscillator is analysed in section 3.3. By characterising the balanced

oscillator with the negative-resistance and feedback models, the locking phenomenon is

33

Chapter 3 Injection-locked Balanced Oscillators

determined. Kurokawa's theory [21]-[22] is used to explain the locking range and phase

noise improvement. We extend Kurokawa's work by using the feedback model, which

has a feedback network characterised as a negative-resistance approach.

For more understanding about the injection-locked oscillator technique, we propose a

novel structure for a fundamental and sub-harmonic injection-locked balanced oscillator

(SILBO). The injection-locked balanced oscillator design approach is based on the

extended resonance technique [6]-[7], which is described in Section 3.4. Sections 3.5 and

3.6 describe the implementation and measurement results, respectively. The stabilised

injection-locked oscillating signal and the locking range with respect to the level of the

injection power are investigated. It also shows the balanced amplitUde and phase

difference between the two outputs of the fundamental and sub-harmonic injection-locked

oscillators under the locking state. These experiments were set-up in order to verify the

amplitude and phase output properties of the balanced oscillator. The advantage of the

injection-locked balanced oscillator technique is that it is simple to design and the circuit

consumes low power. In addition, the circuit size is relatively small because a circulator

and balun at the injection port and input/output ports, respectively, are not required.

3.2 Phase noise theory

An oscillator generates a periodic signal from its own circuit noise [40]. The oscillation

may be started by the noise generator [41]. It induces an exponentially growing

oscillation until the stable oscillation is achieved by some limiting mechanism. This

mechanism limiting the growing signal has been observed by many researchers [40]-[42].

The magnitude grows until the system reaches the sufficiently positive nonlinear

resistance [42]. Then, the oscillating signal becomes a periodic signal with an operating

point at which the positive and the negative resistances cancel, as described in chapter 2.

The electronic nOIse sources that provide the nOIse III any oscillator system can be

categorised into two main groups. The first group is the so-called intrinsic noise sources

and it consists of thermal, shot, and flicker noise. These noise sources arise from random

fluctuations within the physical systems [44]. The other is interference, which is

34

Chapter 3 Injection-locked Balanced Oscillators

determined as the undesirable effect of noise from outside. The supply nOIse IS an

example of the latter group.

Even for a steady-state oscillation, noise is still present. Noise affects the oscillating

signal, causing fluctuations in the operating frequency. It appears as sidebands of the

signal. This noise is separated into AM and PM components. For the amplitude

modulation mechanism, the noise voltage in the spectrum can be added in a power or

root-sum-square basis, so the AM components can represented as a pair of symmetrical

sidebands. The phase deviation is defined as the PM modulation index l1¢max. The

frequency spectrum for the two sidebands is theoretically presented above and below the

carrier frequency, but a pair of these sidebands is anti-symmetrical. Therefore, the noise

voltage due to the AM and PM components is represented as sidebands of the carrier as

shown in figure 3-1. The noise voltage at the lower noise sideband can be cancelled out.

\b~age \b~age \b~age

input input input

Vc Vc Vc

L L 1 \h1

I I VnJ2

\ VnJ2 + VnJ2

> Rl.dian I I > :~~~ncy

Rl.dian frequency

" 1/

(J) i frequency i (J) i c c

((J) +(J) ) ((J) -(J) ) ((J) +(J) ) -VnJ2 c n c n c n

i ((J) +(J) ) c n

((J) -(J) ) c n

Rv1

Figure 3- 1: AM and FM components

For a better understanding, I.G.Ondria [45] has shown that the vector diagram of a carrier

simultaneously phase and amplitude modulated by the nth noise voltage interference as

depicted in figure 3-2. The noise component is located at OJn above the carrier. It is

assumed that Vn «V" where Vn and V, are defined as the noise and carrier voltages,

respectively. The vector diagram shows that a carrier at radian frequency OJe has been

35

Chapter 3 Injection-locked Balanced Oscillators

amplitude and phase modulated at the radian frequency (j)n. From this, the instantaneous

resultant carrier voltage Vet) can be expressed as

Where en is an arbitrary phase angle.

carner

voltage V c

4

3i

CD c

V (rrn s) = nth noise n

voltage

2

instantaneous resultant /" carner voltage

~r ~'" = instantaneous

-- ~ - n phase deviation

(3.1)

Figure 3- 2: Vector diagram of a carrier simultaneously phase and amplitude modulated by the nth

noise voltage interference

For AM and PM noise in a microwave oscillator, all noise components are found by

summing the magnitude square of the components [46]. Figure 3-3 illustrates the signal

spectrum in frequency domain. If we assume that the AM noise is practically very low in

comparison with the PM noise, the PM noise components are defined as phase noise.

Phase noise is the standard deviation of phase fluctuation in oscillators and is also

quantified by its short-term frequency stability. It is defined as the ratio of the single­

sideband (SSB) power, in aI-Hz bandwidth fm Hz away from the carrier, to the total

signal power [47].

36

Chapter 3 Injection-locked Balanced Oscillators

Ps

IPssb/Psl

fo

Figure 3- 3: Spectrum of source in frequency domain

To broaden the understanding of the phase noise in oscillators, an oscillator model is

introduced. By using Leeson's model as the simplest possible model of the oscillator

circuit [47]-[49], an oscillator is viewed as a feedback system. The noise sources are

considered as an input, shown in figure 3-4. The phase noise can be modelled as a noise­

free amplifier with phase modulation at the input. The transfer function of the modulated

RF signal passing through the bandpass filter equals the transfer function of the

modulating signal passing through an equivalent lowpass filter [47]. The transfer

function in this model can be explained as follows:

Phase modulator

>--~------i Reso nator

Noise-free

Amplifier

Figure 3- 4: Leeson's model

37

Chapter 3 Injection-locked Balanced Oscillators

As mentioned in figure 2-1 in Chapter 2, the oscillator consists of the amplifier and the

resonator. The noise figure is determined as the noise added to an amplifier and it can be

defined as

F = (S I N)in

(S I N)out (3.2)

Where F is the noise figure, (S I N)in and (S I N)out are the signal-to-noise ratio at the

input and output of the amplifier, N in and N out are the noise power at the input and output

of the amplifier.

The noise figure is determined as the input phase noise in a I-Hz bandwidth at frequency

offo+fm from the carrier, causing the phase deviation. This phase deviation is given by

v ~e = n,rms

peak Vpavail,rms

FkT (3.3)

In rms terms, we get

~e =_1_ FkT rms Ji P

avail

(3.4)

Where Vn,rms and Ppavail,rms are represent an equivalent available noise voltage due to the

noise figure of the amplifier and the available rms signal power. G, T and K represent the

gain of the amplifier, temperature in Kelvin, and the Boltzmann's constant equal to1.38x

10-23 W.S/oK, respectively.

With a random noise phase relation to the noise signal atfo+fm. the same phase fluctuation

of the signal exists at the frequency of fo-fm from the carrier. If the values of ~e rms are

added, then the total phase deviation is:

38

Chapter 3 Injection-locked Balanced Oscillators

i18 = rrns,total (3,5)

Transforming into the spectral density:

s . = i18 2 = FkT 8m rrns p

avail

(3.6)

Also, a flicker noise or lIf noise component is described by the comer frequency fe, which

is low-frequency device noise in the transistors. The signal purity can be degraded by the

flicker noise at the frequency close to the carrier due to phase modulation. From this, the

spectral phase noise becomes

(3.7)

As mentioned above the transfer function of a modulating signal in Leeson's model can

be characterised as the transfer function of the modulating signal passing through an

equivalent lowpass. The tank circuit can be viewed as a bandpass resonance resulting in

the lowpass transfer function given by:-

(3.8)

Where mm andmo are the offset frequency and carrier frequency in radians, respectively,

The QL represents the loaded Q factor. Equation (3.8) shows that the phase noise is

transformed through the resonator up to the half of bandwidth. The resonator attenuates

the passing phase noise with 6 dB per octave when the modulation rate increases further

[47]-[ 49]. Then the closed loop response of the phase feedback loop is determined by

1 i18 (m) = i18. (m )

out m 1 '(2Q / ) m m + ] LOJm OJo

(3,9)

39

Chapter 3 Injection-locked Balanced Oscillators

The power transfer function is transformed into the phase spectral density as expressed as:

(3.10)

The phase noise at the output of the amplifier with the RF modulating input is given by:

(3.11)

Substituting equation (3.7) into (3.11), then the phase noise becomes:

(3.12)

Equation (3.12) shows the phase noise in the oscillator due to phase modulation and noise

source in the oscillation system. Not only the noise figure, F, and the lIf noise, which are

inherent in the active device, are the main important parameters to determine phase noise,

but the loaded Q factor due to the tank circuit is also crucial. Hence, in order to reduce

phase noise in the oscillator these parameters are considered. For example, the selected

active device should provide a small value of the noise figure. Also, the tank circuit or

resonator should be selected in order to maintain the maximum loaded Q value.

3.3 Injection-locked oscillator theory

There has been great interest in the injection locked oscillator technique [20]-[39]. This

technique is used to stabilise the oscillating signal by injecting a stable reference signal

into the free-running oscillator. Alder [20] described the case of synchronisation where

the external signal is impressed on an oscillator. When the external signal is modulated

with the free-running signal, the oscillator signal contains strong harmonics

/).OJ = OJo - OJi

if the oscillator is almost locked [20]. His work also showed that the speed

40

Chapter 3 Injection-locked Balanced Oscillators

of the locking process is proportional to the ratio of the injecting voltage and to the

oscillator voltage and the bandwidth of the tuned circuit. Later, the noise improvement of

the oscillator by the synchronisation approach was analysed by Kurokawa [21]-[22]. The

negative resistance oscillator model was used and a graphical approach in terms of device

and circuit impedance loci was employed to describe the injection phenomenon. A

stability criterion was developed. The intersection of the two impedance loci depending

on the amplitude and frequency change indicated a steady-state operating point. With a

small injecting signal whose frequency is close to the free-running frequency, the

relationship between the injection vector, impedance locus and device line are considered.

The locking condition is dependent on the ratio of the magnitude of the injection signal

and the magnitude of the free-running frequency signal. He also explained the noise

reduction in a synchronised microwave oscillator. The FM noise can be greatly improved.

However, the PM noise improvement depends on the injection signal purity.

However, both Adler's and Kurokawa 's works can only explain the injection locking

phenomenon in the case when the injection frequency is close to the free-running

frequency. Other works associated with the injection locked oscillator have been reported,

and many different methods of injection locking have been studied. There are many

injection-locked oscillator studies using sub-harmonic and super-harmonic injection [22]­

[39]. The latter case is of less importance in a microwave circuit system since it is not

easy to create the stabilised external signal at high frequency in practice.

In order to examine the injection-locked oscillator technique for fundamental and

subharmonic modes, an oscillator model is needed. Figure 3-5 a) shows the equivalent

balanced oscillator circuit where the injection signal is applied at the centre of the

transmission line. From equation (2.13) and (2.27), the simplest equivalent oscillator on

either left- or right-hand side of the balanced oscillator can be modelled as shown in

figure 3-5 b). Under the free- running oscillation condition, e(t) is equal to zero.

Considering a current loop as shown in figure 3-5 c), the equation for the current i is

expressed as:

41

Chapter 3 Injection-locked Balanced Oscillators

L di +(-R-R)i+-1 Jidt+~Jidt=O

dt C1 C

woL and Q --'--

loaded - R + R (3.13)

B • 1

- R(A,w)

ZIN\////

e(t)

[ Injection

: Zo,8 A

i 1------<1 ·---~l C(A,wJ

- R(A,w)

a)

.....

Z Vd d

+

b)

(3.14)

.+2

Chapter 3 Injection-locked Balanced Oscillators

e(t)

1 i(t)

- R(A,m)

c)

Figure 3- 5: a) A equivalent circuit for the oscillator, b) equivalent circuit characterised in both

negative and feedback models and c) equivalent oscillator on right-hand side of the balanced

oscillator

By using Kurokawa's theory [21], neglecting harmonic terms, then equation (3.14) can be

written as:

(3.15)

Where m = = --, C = ,and then: ~CI +C fzE CIC LeI C LC x x CI + C

di(t) (d9 J dA (t) --=-AI(t) m+-I sin(ox+91(t))+ I cos(ux+91(t)) dt dt dt

(3.16)

fi(t)dt = (AI (t) _ Al (t) d91 JSin(OX + 91 (t)) + ~ dAI (t) cos(ux + 91 (t)) m m 2 dt m dt

(3.17)

Substituting equation (3.15) to (3.17) into equation (3.13), then multiplying by either

cos(mt + 9(t)l) or sin(mt + 9(1)1) and integrating with respect t from t - To to t, where To

Chapter 3 Injection-locked Balanced Oscillators

is a period of the free-running oscillation, Kurokawa showed the resultant equation is

expressed as:

(3.18)

( 1 JdA - 2 t L+-2 - --+(Rj-R)A=- fe(t)cos(OX+¢)dt

m ex dt ATa t-To

(3.19)

Where R is an average resistance over one cycle and is a function of A. The value of it is

given by:

2 I

R = -- f RA cos 2 (OJ{ + ¢ )dt ATo I-To

(3.20)

Now considering the synchronised oscillator mechanism, Kurokawa explained the

injection-locked phenomenon for the fundamental injecting signal, where the injection

frequency is close to the free-running frequency. He assumed that the external injecting

signal is e(t) = ao cosmJ, where ao is small. When the oscillator is synchronised, the

oscillating frequency must be ms. The right-hand side of equation (3.18) is replaced by

~ sin ¢. For the steady state, d¢ = O. Then the difference frequency between the Ao ~

synchronising and free-running frequency is given as

-a a !::..mo = __ 0_ sin ¢o or I!::..mo I ::; __ 0 - sin ¢o

2LAa 2LAa (3.21)

To use Kurokawa's theory [21] in the subharmonically injection-locked case, the

feedback oscillator is used. The equivalent circuit characterised in both negative and

Chapter 3 Injection-locked Balanced Oscillators

feedback models is introduced in figure 3-5 b). From this, we assume that the injecting

signal is fed into the amplifier. The output current at the output of the amplifier is fed

back into the feedback network, where the negative resistance oscillator model is

characterised. The nth harmonic signal is generated. This is due to the nonlinearity in the

amplifier. As a result, e(t) is replaced by:

where mo is close to nms.

Then the right-hand side of equation (3.18) is replaced by ~ 0 sin ¢. In the locking state,

the locking range becomes:

(3.22)

For the harmonic case, an external injecting frequency is close to n times the free-running

frequency, mo :::::: ms / n . The locking range is denoted as tlmO(lIn). In the locking state, the

right-hand side of equation (3.18) is replaced by bn sin ¢, where the absolute value of Ao

bn is higher than that of ao and the absolute value of ao is higher than that of an.

Comparing the locking range between the fundamental and the harmonic injection locked

phenomena, we found that the locking range for the harmonic is higher than that for the

fundamental in which the injection signals are the same. However the locking range for

the fundamental case is higher than that for the subharmonic case. Therefore, it can be

noted that at the same injecting power level the relation of locking bandwidth for

harmonic, fundamental and subharmonical injection-locked oscillator is:

(3.23)

Where n=1,2, ...

From equation (3.21), Kurokawa described the locking range in terms of power. He

assumed that the available power of the injection signal source is denoted by Ps and the

45

Chapter 3 Injection-locked Balanced Oscillators

output of the free-running oscillator is defined by Po. The values of Ps

and Po are given

as follows:

(3.24)

P = R (~J2 = RoAg-o 0.fi 2

(3.25)

Substituting (3.24) and (3.25) into (3.21), then:

(3.25)

( !J..mo J2 < _1_ a;

- 2 2 mo Qext ~

(3.25)

mL. . Where Qext = _0 - and Ro 1S the load res1stance.

Ro

The external stabilised power is injected into the oscillator. The harmonic power

components are generated due to the nonlinearity of the active device. As a result, the

power is passed through the feedback network. Here, the injection-locked phenomenon is

considered at the feedback network, where the negative-resistance model is operative.

The locking range for the harmonic and the subharmonic locking oscillator can also

described in the power form by considering only the fundamental component of power

fed back through the feedback network.

Next, the phase nOIse degradation in the injection-locked oscillator is determined.

Leeson's equation is used to explain the phase noise. Focusing on one half of the

resonator [47], equation (3.12) can be expanded as follows:

46

Chapter 3 Injection-locked Balanced Oscillators

(3.26)

Where QL can be defined as the ratio of reactive power to total dissipated power [47]:

Q L = 27(0 We 27(0 We

P diss (total) ~n + Pres + P sig

(3.27)

Where We represents the reactive energy going between Land C and Pres is the power

dissipated in the resonator. ~n and P sig are input power and signal power, respectively.

Putting (3.27) into (3.26) gives:-

(3.28)

Then we consider the phase noise suppression factor of an injection-locked oscillator,

which is defined as the phase noise ratio between the injection-locked oscillator and the

free-running oscillator:

The phase noise suppression factor = [. (f m ) injection-locked / [. (f m ) free-running (3.29)

[ J

2

We(free-rUnning)

- We(injection-locked)

(3.30)

Where W = ~ CV 2. We transfonn the voltage fonn into the power fonn. It is assumed e 2

that the power is due to the power of the free-running oscillator and that available from

the injection signal. Then, equation (3.28) becomes:

47

Chapter 3 Injection-locked Balanced Oscillators

The phase noise suppression factor = (CIRo~ J2 C2RoP2

Where ~ = Po and P2 = Po + Ps • From equation (3.25), P2 is given as:

(3.31)

(3.32)

Substituting (3.31) into (3.32), and assuming that there is no capacitance change in the

injection-locked oscillator, then the phase noise suppression is found as follows:

The phase noise suppression factor", [1 + ( "': J Q:" J (3.33)

Equation (3.33) can be rewritten as the phase noise of the injection oscillator as shown

below:

(3.34)

As shown in equation (3.34), the phase noise suppression factor is dependent on the

locking range and the external quality factor. In the locking state, it is shown that the

nois'e far away from the carrier will be rerduced by the phase noise suppression factor.

Additionally, the degree of phase noise improvement not only relies upon the injection

frequency, which can be a harmonic, fundamental or subharrnonic frequency, but also

depends on the power level of the injection signal as described above.

48

Chapter 3 Injection-locked Balanced Oscillators

3.4 Design considerations for the injection-locked balanced oscillator

Highly stabilised oscillators with low phase noise are in high demand for many electronic

systems. The injection locking technique has been proposed [20]-[22] amongst other

techniques to provide a stable oscillation signal. An injection locked push-pull oscillator,

alternatively called an injection-locked balanced oscillator, was first proposed [26] for

spatial power combining, where the oscillators are synchronised using a single patch

resonator or antenna [25]. The circuit consists of two active devices and two 1800 phase­

shifting transmission lines at both input and output of the push-pull oscillator. Those two

phase-shifting components consume a large circuit area in RFIC and MMIC realisations.

Figure 3-6 shows the proposed injection-locked balanced oscillator configuration. The

transmission line between the two devices is used such that 180-degree phase difference

free running oscillation signals are achieved from the output ports [7]. The matching

network, M2, is used to couple the injection signal power into the centre of this

transmission line. Due to the injection signal power, the non-linear devices are

perturbed. In the case of the fundamental frequency, the system will be locked to the

injecting signal. If the applied injection signal power is small compared with the free

running signals, the condition of 1800 phase difference is still maintained. It is very clear

that to apply the injection signal at the middle port requires only a matching component.

In addition, a circulator is not required at the middle port due to the balanced topology,

which allows perfect LO isolation at the injection port. Similar to the fundamental

injection-locked balanced oscillator, the sub-harmonic injection-locked oscillator

employs the same injection method, but the injection network port, M2, is considered for

the sub-harmonic frequency, (free-running frequency/2). Therefore, the injection network

port, M2, is designed corresponding to the injection frequency signal in the circuit.

49

Chapter 3 Injection-locked Balanced Oscillators

Terminating

network

Ml

12ej82

......................................... : I lej8l

t----~ ... ~ Zo,8 ~ ! +~_--.....j

M2

Z Z Z Injection Z Z Z T2 IN2 L2 L1 INl T1 signal

Terminating

network

Ml

Figure 3- 6: The proposed injection-locked balanced oscillator configuration

3.5 Circuit Realisations

The fundamental and sub-harmonic injection-locked balanced oscillators were designed

and constructed on FR4 using Infineon CFY30 GaAs FET devices. Series feedback was

used at the two sources to obtain the negative resistance and the transmission line was

connected between the gates. This causes the input impedance of one active device to be

the load impedance for the second active device, and vice versa. As described in chapter

2, in the steady state the oscillation condition is given as:

where ZLX(A, m) is the input impedance, ZIN(A, m) is the load impedance and X is 1 or 2.

Figure 3-6 shows the fundamental injection-locked balanced oscillator circuit. For the

fundamental mode, a coupled line section was used to form the injection network port,

M2. From this, the oscillation condition is taken into account and the return loss at the

injection port is also determined. The return loss should be as low as possible so that the

maximum injection power can pass into the circuit. Due to the symmetrical configuration,

the gate bias for the two active devices is common at the centre of the transmission line.

Figures 3-7 and 3-8 show the photographs of the fundamental and subharmonic injection­

locked balanced oscillators. In steady state, these two circuits are expected to lock at the

injection frequency and twice injection frequency for the fundamental and sub-harmonic

injection-locked balanced oscillators, respectively.

50

Chapter 3 Injection-locked Balanced Oscillator

Injection port

Figure 3- 7: The fundamental injection-locked balanced oscillator circuit

+- Injec1ion port

Figure 3- 8: The sub-harmonic injection-locked balanced oscillator circuit

3.6 Results and discussion

3.6.1 Fundamental injection-locked balanced oscillator

Two oscillating signals were observed at zero gate bias and 2.S V drain bias. The total

power consumption is around lSmW. The LO leakage at the injection port is ob erved.

Chapter 3 Injection-locked Balanced Oscillators

The value of the LO leakage is -19dBm. This indicates that the two LO signals cancel out

each other at the mid-point of the transmission line.

Figure 3-9 shows a comparison of output spectra for two different injection power levels

compared with the free running signal from one of the output ports. The obtained free­

running output signals have a power level of 3.33dBm. As the injection power level

decreases the phase noise of the output signal is increased but is still better than that of

the free-running signal. The locking range with different injection power levels is

observed. Figure 3-10 shows the measured locking range corresponding to the injection

power level. As expected, a higher injection power level provides a wider injection

locking range. The maximum locking range is 6 MHz at an injection power of -5dBm.

o -10

-- -20 e ~ "0 -30 '-"

-40

-50

-60

-70

-80

ATTEN20dB RL 10.0dBm

RBW 3.0kHz VBW 3.0kHz SWP 140ms

CENTER 1.8896500GHz RBW3.0kHz

10dBI MRK 3.33dBm 1.8896500G Hz

Free-running

VBW3.0kHz SPAN 500.0kHz SWP 140ms

Figure 3- 9: The measured free-running signal compared to the injection locking frequency signals

52

Chapter 3 Injection-locked Balanced Oscillators

6000 ..-. 5000 N

~ '-" 4000 ~ ~ = 3000 ~ ~ ~

2000 = :E2 ~ 0 1000 ~

0 -17.5 -15.0 -12.5 -10.0 -7.5 -5.0

Injection power level (dBm)

Figure 3- 10: The locking range of the fundamental injection-locked balanced oscillator as function of

the injection power level

By increasing the injection power level, the measured locking range was higher. The

locking range of the fundamental injection-locked balanced oscillator is a function of the

injection power level. These experimental results agree with Kurokawa's theory [21] as

well as with the Leeson's model [82], which is explained in the previous section.

Therefore, it is noted that Kurokawa's theory can be used for the balanced oscillator.

To investigate the phase and amplitude response between the two output ports, the test

bench is set up as follows: A -14 dBm external signal is obtained from an HP8510C

network analyser system and fed into the injection locking port of the balanced oscillator.

The first output port is connected to the second port of the network analyzer via a 6-dB

attenuator to protect the network analyzer. The second output port of the circuit is

connected to an HP8563B spectrum analyzer to monitor the locking state. The signal

synthesizer from the network analyzer is swept throughout the whole locking range of the

oscillator. This measurement technique is used to observe the locking range with one

power level.

Figure 3-11 shows experimental set-up for phase and magnitude measurement of the

injection-locked balanced oscillator outputs. The phase difference between the output

signals can be examined by investigating the S21 on the polar display [36]. First, the

53

Chapter 3 Injection-locked Balanced Oscillators

transmission (S21) of output 1 and injection ports was measured. Then, the network

analyser measured the phase of the transmission (S21) at output 2 and injection ports.

Figure 3-12 compares the polar diagram of S21 for the first and the second output port of

the ILBO. The result of the transmission locking polar diagram at each port presents the

phase difference of the two output signals. It is shown that the injection gain of each

output port is not constant for the whole locking range. As shown in Figure 3-12, marker

I at each port on the polar diagram shows the beginning of the locking state while the

marker 2 represents the end of the locking state. Also, figures 3-13 a) and b) show the

phase and magnitude responses of the two output signals for the whole locking

bandwidth. The amplitude imbalance between the two output ports is better than 0.8 dB.

The locking bandwidth of the two output signals are relatively the same while the phase

difference between these two outputs under the locking state is 180-degrees.

Vector Network Analyzer

S_Parameter Test Set

Port 1

50 _ohm load

Port 2

Synthesised Oscillator

Spectrum Analyzer

Figure 3- 11: Experimental set-up for phase and magnitude measurement of the injection-locked

balanced oscillator output

54

Chapter 3 Injection-locked Balanced Oscillators

start 1.870000000GHz stop 1.890000000GHz

"fr 5 21 at the output 1 and injection ports

1)"- 5 21 at the output 2 and injection ports

Figure 3- 12: The amplitude and phase response between two output ports of the ILBO as locking on

polar diagram

-m 'tJ

- 0 or-N en

10.0 dBI

r---u A R1{~R 2-1

I--3. 197MHz

MARKER 2*-1* 1*1 2 2 'Sl f--

~

3. 197MHz W I~ V ~

--=:;: --

START 1.870000000GHz STOP 1.890000000GHz

Frequency( GHz)

a)

f'-..- - I-- _

55

Chapter 3 Injection-locked Balanced Oscillators

360

~}*

"-- 180 Q)

~ C) Q) "C '-' 0 Q)

~} '---"\

CI) C1l

.r:::: a.

-180

-360

~ ........

---- -

START 1.87780300GHz STOP 1.88100000GHz

Frequency(GHz)

b)

~

2

L-::' 2*

Figure 3- 13: a) Magnitude and b) Phase response of the fundamental injection-locked balanced

oscillator

Next, the harmonic injection-locked balanced oscillator was measured. An external

injection signal at twice the free-running frequency and with magnitude of -20 dBm was

applied to the injection port. As the locking state was reached, the output was monitored

with a spectrum analyzer. Figure 3-14 shows the output signals of the fundamental and

harmonic injection-locked balanced oscillator. Furthermore, the locking range was

investigated for power levels between -7.5 and -20 dBm. Table 3-1 shows the measured

locking range of both the harmonic and fundamental locked oscillators. However, the

harmonic frequency signal was attenuated since the matching network was designed for

the fundamental frequency signal. The measured locking range for the harmonic case was

still higher than that for the fundamental case. This agrees with the analysis described

above. The harmonic component is transformed into the fundamental frequency term,

resulting in higher magnitude compared with the magnitude of the fundamental injection

locked case.

56

Chapter 3 Injection-locked Balanced Oscillators

0

-10

-. -20 9 = -30 "0 '-' ~

-40 "0 .a '8 -50 ell eu ~ -60

-70

-80

ATTEN20dB RL 10.0dBm

. -.'»;.- '.~;'J.';r:~~:;' .. :'."

'.:: : ..... ,.

?

~0' ',.h.' i'::.h.

Injection@fo

'\ '"'

.11:\ .n..

~ ~t.'-J. ~A~!~ ~tJ p~

/, III

ff~ -, .

I~~~l j I . .J.J~, JI

M~ rr'l~\ ~1 ~'1 I

CENTER 1.8882007GHz RBW3.0kHz

10dB!

~ .. ~: "'

...

I i , if.;,.. ~~~. II " \ rr~.l"'J

\,II , rl lj~ It:/W , I

VBW3.0kHz

MRK 3.S0dBm 1.888007GHz

Injection@2fo ~

/ ~~, ~ ~ ~l~ .. d l}, .. "1', II • ~'t~r ~r~'\1V;; ~ L~ " U'>L ~ ~,

prMr~'IT~j ~,,~

SPAN SOO.OkHz SWP 140ms

Figure 3- 14: The output signals between fundamental and harmonic injection-locked balanced

oscillator

Injection Locking range (MHz) Locking range (MHz)

Frequency( GHz) with -7.5dBm with -20dBm injection

injection power power

fundamental 3.86 0.6

frequency

Second harmonic 16.9 3.2 Frequency

Table 3- 1: The locking range with respect to the fundamental and harmonic injection-locked

balanced oscillator at power levels of -7.5 and -20 dBm

57

Chapter 3 Injection-locked Balanced Oscillators

3.6.2 Sub-harmonic injection-locked balanced oscillator

The drain biasing is provided through a bias tee at each output port. The circuit was

biased at V DS = 2.5 V and V GS = -0.4 V. The free-running frequency of this balanced

oscillator was measured by terminating a 50-ohm load at the injection port. The oscillator

provided a free-running frequency of 1.9539 GHz. with a power of 4.5 dBm. The circuit

consumed 23.3 mWatt.

The subharmonic injection-locked signal was investigated. The coupled transmission line

for the matching network is used to couple the subharmonically injection signal into the

oscillator. With a -2.5-dBm external injection frequency of 977.86 MHz applied to the

injection port, figure 3-15 shows the output spectrum of the sub-harmonic injection

locked balanced oscillator. Compared to the free-running case, the sub-harmonic injection

locked oscillator provides better oscillation frequency stability and reduces the phase

noise. In addition, the locking range with respect to the injection power level was

measured and shown in figure 3-16. For -5 dBm and 2.5 dBm injection power levels, the

locking ranges were 110 kHz and 600 kHz, respectively.

Similar to the fundamental case, the locking range for the subharmonic injection-locked

balanced oscillator obtained from the experiment shows that the locking range IS

amplitude dependent. By extending Kurokawa's work, both negative and feedback

oscillator models are defined in the balanced oscillator. Due to the nonlinearity of the

amplifier the harmonic components are fed back into the feedback network, resulting in

locking as described in the previous section. However, the amplitUde of the second

harmonic term is less than the fundamental term. To achieve the locking state, the

sub harmonic injection signal power must be lower compared to the fundamental and

harmonic injection signals. The measurement of the locking range of the harmonic,

fundamental and subharmonic injection-locked oscillators from the experimental results

are 16.9 MHz, 3.86 MHz and less than 100 kHz, respectively. These measurement results

agree with equation (3.22) in the previous section.

58

..-e ~ "C '-" QI

"C .a ·a OJ.) cu ~

0

-10

-20

-30

-40

-50

-60

-70

-80

ATTEN20dB RL lO.OdBm

l-

SPAN 100.0kHz

'"

}~

~j ~J\MJ6 IN. l~ b .. ) rItIt\~~~ I)\tV" rf t"1 , ·rw'

CENTER 1.9557266GHz RBW 10kHz

Chapter 3 Injection-locked Balanced Oscillators

10dBI

1\ I

1\ I \ / \ / \

/ \l 1/1'

VBW 1.0kHz

IWrt ~) II I ,

MRK 4.50dBm 1.9557266GHz

&UI IM1IA JJ I'.\dA~JI f1 • lV' " 'If'l

SPAN 100.0kHz SWP250ms

Figure 3- 15: The measured sub-harmonic injection locked frequency signal

-N :z: .¥: -(1) C) £: C'CS ... C) £: ~ (,) 0 ..J

700

600

500

400

300

200

100 v::: o -7.5

/ /

/ V

k. ~

-5.0 -2.5 o Injection power (dBm)

Figure 3- 16: The locking range as a function of the injection power

2.5

59

Chapter 3 Injection-locked Balanced Oscillators

In order to observe the amplitude balance and phase difference between the two outputs

of the sub-harmonic ILBO, the test bench was setup as shown in Fig. 3-17. An RF signal

is equally split by a Wilkinson divider. Two mixers are used for down-converting the LO

signals to IF. Two 21.37 dB directional couplers were constructed on FR4. For more

accuracy, the phase difference between the direct ports of these two couplers is measured

and the value of the phase difference is approximately 0.35°. The first coupler is used for

monitoring the injection-locked spectrum at the first port while the second one is applied

for compensating a phase delay. Therefore, it is noted that the phase delay of these two

couplers is almost exactly equal.

spectrum analyzer

signal generator signal generator

D CH1 CH2

Figure 3- 17: Phase measurement test bench setup for sub-harmonic injection-locked balanced

oscillator

An external sub-harmonic signal of 981.055 MHz at 0 dBm was injected into the

balanced oscillator which was biased at 2.63 V drain-source and -0.9 V gate-source. A

1.9-GHz RF signal with O-dBm was applied at the Wilkinson divider input. In the locking

state, the circuit provided the locked signal of 1.96211 GHz, which is twice the injection

60

Chapter 3 Injection-locked Balanced Oscillators

signal and close to the fundamental frequency signal of the oscillator, and then the RF

signal is down-converted, providing the Intermediate Frequency (IF) signal. Fig. 3-18

shows the 62.11 MHz-IF signals from the two mixers monitoring the phase difference

response corresponding to the subharmonic injection-locked oscillator outputs. Thus, it

can be stated that the balanced oscillating signals still maintain 1800 phase difference in

the locking state.

30

20 -.. ;,

10 a '-' Q,j

"0 0 .a 'a OJ) -10 ~

~ -20

-30

. ~: : . l. ... _ . __ ... ... __ . _. _. - - ........ .

Freq(2)=82.11MHz TIME (ns)

Figure 3- 18: Measured phase difference between the two outputs of the sub-harmonic injection­

locked balanced oscillator

3.7 Conclusion

Oscillators are widely used in communication systems for mixers, modulators and phase­

locked loops in the receiver. Generally, there is no ideal oscillator which provides a very

narrow bandwidth at the oscillating frequency without any noise. Due to the noise effect,

the purity of the oscillating signal is degraded. The frequency and magnitude fluctuations

typically appear as sidebands of the oscillating frequency. These can be described by a

standard modulation mechanism.

61

Chapter 3 Injection-locked Balanced Oscillators

The frequency fluctuation in the oscillators is known as phase noise. In Leeson's model,

h · h " 1 t e mput p ase nOIse IS - FKT where F is noise figure The available noise voltage is 2 .

fed back to a tank circuit. Phase noise can be modelled as the modulating signal pass

through the tank circuit. In addition, this model is show that the purity of the signal can

be degraded by the flicker noise at the frequency close to the carrier due to the phase

modulation. In order to obtain low phase noise in the oscillator, the components such as

the tank circuit and active device must be chosen carefully. The tank circuit must have a

high (loaded) Q factor.

There is another technique to improve the stability of the oscillating signal: This

technique is called injection locking. The injection-locked oscillator was first described

by Adler [20]. He explained the case of synchronisation where the external signal is

impressed on an oscillator. Later Kurokawa [21]-[22] showed that the noise of the

oscillator was improved by the synchronised approach. However, Kurokawa's work

introduced an injection-locked oscillator where the injection frequency was close to the

free-running frequency.

To describe the injection-locked phenomenon where the injection frequency is close to

either a harmonic or subharmonic of the free-running frequency, Kurokawa's work [21]

is extended. The negative-resistance and feedback oscillator models are used to

characterise the balanced oscillator. As an external signal is applied into the oscillator,

the harmonic signals are generated due to the nonliearity of the amplifier. Then all

harmonics are fed back to the feedback network, which can be viewed as the negative­

resistance approach. Here, the locking phenomenon is described by Kurokawa's theory.

From this, the analysis shows that with the same injection power level the locking range

for the harmonic injection case is higher than that for fundamental injection and is also

higher than that for subhamonic injection.

A novel injection-locked balanced oscillator has been presented. The fundamental and

subharmonically injection-locked balanced oscillators are introduced. The oscillators

provide the out-of-phase oscillating outputs. By employing the symmetrical

configuration, the properly coupled transmission line is used as the matching network at

62

Chapter 3 Injection-locked Balanced Oscillators

the center of the circuit. In the fundamental mode, the injecting signal is applied to the

circuit passing through the matching network. Similar to the fundamental injection­

locked circuit, the subharmonic one utilises the same technique but the matching network

for the injection port is designed for the injection frequency which is approximately half

the free-running frequency. As a result, the locked oscillating signals are obtained. In the

locking state, both oscillators provide the locking frequency which is approximately

equal to the injection frequency for the fundamental locked phenomenon and is close to

twice the injection frequency for the subharmonic injection-locked case. As the circuit

reaches the locking state, the output spectrum shows a significant phase noise

improvement. The measured locking ranges with different injection power levels have

been investigated. With the same injection power level, the locking ranges as a function

of the injection frequencies, fundamental, harmonic and subharmonic, have also been

investigated. These measurement results show an agreement with the analysis. In

addition, the phase difference between the two outputs of the balanced oscillator under

the locking state for both different injection frequencies have been investigated by using

different techniques. The network analyser and the mixing technique are used to monitor

the balanced properties for the fundamental and the subharmonic locking states. As

shown in the experimental results, the two outputs maintain 1800 phase difference with

the same magnitude. The advantage of this injection-locked balanced oscillator technique

is that it is simple to design since the design eliminates the need for a circulator for the

injection signal. Furthermore, this design technique can improve the phase noise

degradation in the balanced oscillator, which can cause many difficulties in the

communication applications. Also, the injection-locked balanced oscillator technique

can be applied for many applications, such as balanced mixers and modulators.

63

chapter 4 Injection-locking applied to a cascaded oscillator

Chapter 4

4 Injection-locking applied to a cascaded

oscillator

4.1 Introduction

Many power combining approaches in the microwave and millimetre-wave frequency

range have been studied over the years [50]-[53]. In 1980, S.Mizushina [50] presented the

power combining of oscillators by using a short-slot coupler. The three identical

oscillators and matched load were connected to a four-port hybrid coupler. By using the

injection-locking technique, he applied the external injection signal to the first oscillator,

then the locking signal was injected to the other two oscillators at port 2 and port 3 of the

coupler. From this he got the stabilised power combined oscillation at port 4. Later in

1982, M.Madihian and S.Mizushina [51] explained the operation of the combiner circuit

using a directional coupler by employing the phase-locked theory to control the

frequency. The conditions for the optimum operation were also analysed.

Since the hybrid combiner provides good isolation between sources, the device

interaction and instability problems associated with multidevice operation are minimised

[52]. The input is applied at port 1 and the reflected waves are added at port 4. However,

K.Chang and K.Sun [52] revealed that as the number of input sources increased it was

more difficult to operate the combiner at the optimum point with the same frequency. To

achieve this, they suggested that for a large number of input sources using the hybrid

method, it was better to use injection-locked operation [53]. For simplicity, this can be

viewed as that each successive oscillator is connected in a chain by using a coupling

network and the injection-locking signal is applied. The oscillating frequency is

64

chapter 4 Injection-locking applied to a cascaded oscillator

controlled by the locking signal. Furthennore, the phase noise could be significantly

reduced by making the first stage oscillator a low phase noise source [46].

For the power combining technique with the injection-locked oscillator approach, this

chapter studies the additional sidebands which are the result of the synchronised effect in

the locking state, rather than focusing on the maximum combined power from the

oscillator. Unlike the nonnal power combining technique, a -20dB directional coupler is

used to connect between the two identical oscillators in the cascaded configuration. The

subhannonical injection-locking signal is applied at the first or master oscillator. After the

two oscillators reach the locking state, the second oscillator provides a stabilised

oscillating signal as an output. We observe the sideband signals in a single oscillator and

a cascaded oscillator during the locking state. The experiment shows that the sidebands in

the cascaded oscillator are reduced by more than 30dB compared to the single oscillator.

Also, the locking range with the injection frequency close to (l/2)x and (l/3)x the free­

running frequency is measured.

4.2 Design considerations for the injection-locked cascaded oscillator

The concept of a cascaded oscillator is that two single-ended oscillators are connected in

cascade configuration by using a directional coupler to couple the output signal from the

first oscillator to the second oscillator. Figure 4-1 shows a block diagram of the cascaded

oscillator concept. In order to investigate the injection-locked behaviour for the cascaded

oscillator, an injection signal close to the free-running-frequency divided by n is applied

at the first single-ended oscillator and a -20-dB directional is used to connect between the

first and the second oscillators. The output signal of the first oscillator not only consists of

the free-running frequency, but is also composed of its hannonics. We assume that the

stabilised output signal from the first oscillator is attenuated by the -20dB directional

coupler, resulting in a reduced magnitude of the fundamental component. Finally, the

output signal of the second single-ended oscillator will be observed.

ction Inje si

at fa, gnal ~ fo/2, fo/3

-20dB A single-ended --'0.. directional ""

A single-ended oscillator

, , oscillator

coupler

Figure 4- 1: A block diagram of a cascaded oscillator

Outpu t

"" r

65

L-nUpler 4 Injection-locking applied to a cascaded oscillator

4.3 Circuit realisation

According to the block diagram of the cascaded oscillator as depicted in figure 4-1, two

singled-ended oscillators and a - 20dB directional coupler were designed. These were

fabricated on FR4. The Infineon CFY30 MESFET was used as the active device in the

single-ended oscillator. The two-port negative resistance technique was used. The series

feedback was connected to the source of the active device in order to get the negative

input impedance as described in chapter 2. Then the matching circuit was designed. From

this, the oscillation condition was considered so that the free-running oscillation was at

about 2 GHz. Figure 4-2 shows the single-ended oscillator used in the cascaded oscillator.

By connecting all the two oscillators and the directional coupler, the cascaded oscillator

circuit is as shown in figure 4-3.

Figure 4- 2: A single-ended oscillator circuit

66

Chapter 4 Injection-locking applied to a cascaded oscillator

Figure 4- 3: The cascaded oscillator circuit

4.4 Results and discussion

First, the cascaded oscillator with no injection signal was observed. The two oscillators

were biased at Vgs= 0 V and Vds= 3.4 V, where the free-running frequencies of the two

oscillators were the same. As the two oscillators synchronised, the output signal at the

second oscillator was measured as shown in figure 4-4. The synchronised frequency was

about 1.967GHz with a magnitude of 2.50dBm. Then, the cascaded oscillator was

injected with the external stabilised signal at different frequencies : close to the free­

running frequency, free-running frequency/2 , and free-running frequency/3, respectively.

Figure 4-5 illustrates the output signals in the locking state. It is shown that phase noise

in the injection-locked cascaded oscillator output improves. However, the amplitude noise

is slightly degraded when the injection frequency is close to the free-running frequency/2

and is more degraded when the injection frequency is close to the free-running

frequency/3.

67

---a = "CI '-' ~

"CI ::I -'a ell ~

~

0

-10

--- -20 a = -30 "CI '-' ~

-40 "CI S 'a -50 ell ~

~ -60

-70

-80

Chapter 4 Injection-locking applied to a cascaded oscillator

0

-10

-20

-30

-40

-50

-60

-70

-80

ATTEN20dB RL 10.0dBm

CENTER 1.968224GHz RBW30kHz

10dBI

VBW30kHz

MRK 2.5OdBm 1.968231 G Hz

SPAN 2.000MHz SWP50ms

Figure 4- 4: Free-running cascaded oscillator output without injection signal

ATTEN20dB RL 10.0dBm

~ I ,~, ,~f~

~ ~~ ~~~ I

~ .J.kr~ ~ II ~rrr \ rVV\JI I

J1 J\I ~ ~I~~~ I,~ ~f, 'I~~

\J ,. r I

CENTER 1.9650002GHz RBW 1.0kHz

10dBI - - -

I \ I \ j \

~~A r ~ (i \.y

~, " I~

VBW 1.0kHz

- - -

MRK 2.83Bm 1.9650002G Hz

-- - - -

Injection@fol3

/

7 j~ ~

~A

Injection@fol2 -

I I

Injecron@fo

J;1 ~k ~ h, t ~

I'V' ;1fo ~ ~ ~r I,Mj 11"

~:Hr II: I

)1 ~\~ AI. .r~/~1 IF r r~~y I 'i ~r~~ij

SPAN 100.0kHz SWP250ms

Figure 4- 5: The output signal from the injection-locked cascaded oscillator: the injection frequency

is close to lx, (l/2)x and (l/3)x free-running frequency

68

Chapter 4 Injection-locking applied to a cascaded oscillator

Next, a single-ended oscillator was investigated. The oscillator was biased at the same

voltage where the two oscillators in cascade configuration were synchronised. Figure 4-6

illustrates the output signal of the injection locked oscillator circuit at different injection

frequencies: close to (l/2)x and (l/3)x free-running frequency. As for the locking state,

we then observed the harmonic output signals. The injection signal at frequencies of

(l/2)x and (l/3)x free-running frequency with the power of -10dBm was injected into the

oscillator. In the locking state, the output signal contains the harmonic components,

whose frequencies are n times the injection frequency. Figure 4.7 and figure 4.8 show the

output spectra of the injection-locked single-ended oscillator where the injection

frequencies are close to free-running frequency/2 and close to free-running frequency/3,

respectively. Moreover, the locking range was measured as a function of the injection

power levels. The locking ranges for these two different injection frequencies are shown

in Tables 4-1 and 4-2.

o -10

-- -20 S = "0 -30 --~ "0 -40 .a ·S -50 ~ ~

~ -60

-70

-80

ATTEN20dB RL 10.0dBm 10dBI

MRK 3.33Bm 1.9636800G Hz

/ Injection@fol3 ~----+---I-------+----t----+--t J-t

l-- I . I,In

jection@f

ol2

/ \.11 , / I /Injlection~fO ~ k .f ,jJ 11J ~I} I

CENTER 1.9636800G Hz RBW 1.0kHz VBW 1.0kHz

SPAN 100.0kHz SWP250ms

Figure 4- 6: The output signal from the injection-locked single-ended oscillator: the injection

frequency is close to lx, (1I2)x and (1/3)x free-running frequency

69

--e = "C --~ "C = ...... '8 ~ ~

~

0

-10

-20

-30

-40

-50

-60

-70

-80

Chapter 4 Injection-locking applied to a cascaded oscillator

ATTEN20dB RL 10.0dBm

CENTER 1.957GHz RBW1.0MHz

10dBI

VBW 1.0MHz

MRK 2.OdBm 1.957GHz

SPAN 4.087GHz SWP50.0ms

Figure 4- 7: The output spectrum from the single-ended oscillator during locking state with respect to

the injection signal at frequency of (1/2) x free-running oscillating signal

o -10

-- -20 e f§ -30 --~ "C -40 .a '8 -50 ~ ~

~ -60

-70

-80

ATTEN 20dB RL 10.0dBm

CENTER 1.965GHz RBW 1.0MHz

10dBI

VBW 1.0MHz

MRK 1.50dBm 1.965GHz

SPAN 4.13GHz SWP50.0ms

Figure 4- 8: The output spectrum from the single-ended oscillator during locking state with respect to

the injection signal at frequency of (113) x free-running oscillating signal

70

cnapter 4 Injection-locking applied to a cascaded oscillator

Input power level Locking range

(dBm) (kHz)

-4 107.71

-6 86.95

-8 60.10

-10 36.16

-12 23.32

Table 4- 1: The locking range with respect to the injection power levels, where the injection frequency

is close to (1/2)x free-running frequency

Input power level Locking range

(dBm) (kHz)

-8 406.92

-10 261.61

-12 157.10

-14 114.73

-16 64.58

Table 4- 2: The locking range with respect to the injection power level, where the injection frequency

is close to (1/3)x free-running frequency

Next, the cascaded oscillator output signal with its harmonic components was observed.

By applying the injection frequency where the frequency is close to the free-running

frequency/2 and close to the free-running frequency/3 with the injection power level of -

10 dBm, the spectrum analyser shows the output signal from the cascaded oscillator as

71

cnapter 4 Injection-locking applied to a cascaded oscillator

shown in figure 4-9 and 4-10. The output signal consists of the hannonic components of

the injection signal. However, it is noted that the magnitude of the harmonic components

close to a wanted signal is less than that of the single-ended oscillator. Furthermore, the

locking ranges for this cascaded configuration were also investigated and the

measurement results are shown in table 4-3 and 4-4. For the different injection

frequencies, the first one is close to the free-running frequency/2 and the other is close to

the free-running frequency/3. It is shown that the locking range for the first case is higher

than that for the latter case. This can be described by the analysis as mentioned in the

previous chapter. Compared to the single-ended oscillator case, the locking range of the

cascaded configuration is lower that of the single-ended case. It could be because of the

coupled oscillator. The two oscillators in the cascaded configuration have to maintain the

internal synchronisation mechanism and perform a locked state due to the external

injection signal simultaneously, resulting in lower locking range.

o -10

-20

-30

-40

-50

-60

-70

-80

ATTEN20dB RL lO.OdBm

CENTER 1.971GHz RBW 1.0MHz

lOdBI

VBW 1.0MHz

MRK 1.50dBm 1.971GHz

SPAN 4.087GHz SWP50.0ms

Figure 4- 9: The output spectrum from the cascaded oscillator during locking state with respect to the

injection signal at frequency of (1/2)x free-running oscillating signal

72

-.. e ~ "C --Q,j

"C .a 'a CJl ~

~

0

-10

-20

-30

-40

-50

-60

-70

-80

Chapter 4 Injection-locking applied to a cascaded oscillator

ATTEN20dB RL 10.0dBm

CENTER 1.971GHz RBW 1.0MHz

10dBI

VBW 1.0MHz

MRK 1.5OdBm 1.969GHz

SPAN 4. 139GHz SWP50.0ms

Figure 4- 10: The output spectrum from the cascaded oscillator during locking state with respect to

the injection signal at frequency of (1/3)x free-running oscillating signal

Input power level Locking range

(dBm) (kHz)

-4 84.12

-6 42.47

-8 36.80

-10 32.23

-12 2l.2

Table 4- 3: The locking range for the single-ended oscillator with respect to the injection power levels, where the injection frequency is close to (1I2)x free-running frequency

73

Chapter 4 Injection-locking applied to a cascaded oscillator

Input power level Locking range

(dBm) (kHz)

-8 630.8

-10 460.2

-12 251.0

-14 151.1

-16 71.3

Table 4- 4: The locking range for the cascaded oscillator with respect to the injection power levels,

where the injection frequency is close to (1/3)x of free-running frequency

4.5 Conclusion

An injection-locked cascaded oscillator is presented in this chapter. By using the power

combining approach, the directional coupler is connected between the two identical

oscillators. The advantage for this configuration is that a) it is simple in construction and

b) the coupler gives high port-to-port isolation, resulting in minimal interaction among the

devices. Furthermore, the circuit provides the phase noise improvement as shown in the

experimental results. However, instead of focusing on the optimized power combining

this work is concerned with the sideband suppression due to the subharmonically

injection-locked phenomena in the single-ended oscillator. Then, the investigation of the

injection-locked cascaded oscillator is considered, where the injection frequency is close

to (1I2)x and (l/3)x the free-running frequency. As a result, the sidebands are

significantly reduced by more than 30 dB for both cases. Moreover, the locking ranges of

both single-ended and cascaded oscillators were measured. The locking range of the

single-ended oscillator is higher than that of the cascaded oscillator. This might be due to

the limitation of the coupled configuration.

7.+

Chapter 5 Frequency Doubler Design Configurations

Chapter 5

5 Frequency Doubler Design Considerations

5.1 Introduction

Frequency doublers can be used to extend the operation range of an oscillator since there

is the limitation on the fundamental oscillators of a given technology. Also, there is a high

demand for RF sources with excellent frequency stability and low phase noise in

communication systems. Such a local oscillator might consist of a low phase-noise

fundamental source followed by a chain of frequency doublers and amplifiers. For the

multiplication process, a nonlinear transfer characteristic is required [54] which involves

the nonlinear relationship between output current and applied voltage to the device.

Suitable nonlinear devices can be diodes, bipolar transistors, MESFETs, and HEMTs.

This chapter studies the design and operation of MESFET frequency doublers. The use of

the MESFET as the frequency doubler is attractive since the MESFET frequency doublers

have the advantage of providing conversion gain and also have some isolation between

input and output ports [55]-[56]. MESFET frequency doubler design is focused on the

consideration of operating point. For optimum nonlinear operation, the device can be

biased either in the vicinity of pinch-off or in the vicinity of forward conduction where

the gate-source voltage swings between pinch-off and zero volts [54]-[60].

The other basic and important design consideration is the circuit topology. Since the

unwanted components appear at the drain, the required frequency component can be

obtained by utilising either a single-ended or balanced doubler topology. However, the

input and output matching networks are also considered. The input network should be

designed to achieve a maximum power to the device at the fundamental frequency whist

75

Chapter 5 Frequency Doubler Design Configurations

the output of the device is matched for maximum output power at the second harmonic

frequency. These design considerations are described in this chapter.

Additionally, the practical example of an injection-locked balanced oscillator and doubler

is demonstrated. An idea of this example is the integration of the balanced oscillator [61]

and balanced doubler circuits. By using the injection-locked balanced oscillator

technique, the entire circuit gives a low-phase noise signal. Also, the external balun for

the balanced configuration is eliminated, resulting in small circuit size.

5.2 General Doubler Design Considerations

In the microwave region, the Schottky barrier diode, the bipolar transistor and both

single-and dual-gate MESFETs can be used for frequency doublers due to their nonlinear

device characteristics [54]. The nonlinear transfer characteristic of the devices is the

important key mechanism for the frequency doublers. This transfer characteristic can be

described by an expansion of a Taylor series. From many applications, the MESFET is

considered because of its potential for providing gain [54].

The aim of this section is to consider the MESFET operation as a frequency doubler and

investigate the gate-source bias conditions. MESFET frequency doublers can be classified

into two bias conditions: operating at gate-source voltage in the vicinity of pinch-off and

between pinch-off and zero voltage. To achieve the frequency doubler design, the

MESFET must be kept stable. Similar to amplifier design, the stability factor K is

considered [57].

Also, various doubler design topologies are introduced. One is the single-ended frequency

doubler. From many previous studies, many techniques to obtain good performance for

the single-ended doublers are considered. The other topology is the balanced frequency

doubler. This design topology basically provides the wider bandwidth compared to the

single-ended topology.

76

Chapter 5 Frequency Doubler Design Configurations

5.2.1 Biasing considerations

As described by R.Gilmore [54], a basic concept of the multiplier process can be

introduced by considering a nonlinear device such as a diode. The transfer characteristic

of the diode can be explained by using a Taylor series as follows:

The transfer characteristic is given by

(5.1)

Where g is the nonlinear function of the device, io is the output current and Vi is the

applied input voltage.

By using a Taylor series, an expansion of g(Vi) is given by:

(5.2)

Substituting Vi = AcosUljt, then we get

(5.3)

From equation (5.3), the output current consists of dc, fundamental frequency and

harmonic frequency components. The principal objective of the frequency doubler is to

obtain only the second harmonic component and to minimize other current output

components. To achieve this objective, a bandpass filter is ideally required in order to

eliminate all unwanted current output components

From many applications [9], [54]-[68], the MESFET has been used for the nonlinear

device in a frequency doubler. The main principle of the MESFET frequency doubler is

the half-wave rectification in the gate-channel Schottky junction, with a relatively high

reverse breakdown voltage [9]. A half-wave cosine waveform is obtained at the drain

77

Chapter 5 Frequency Doubler Design Configurations

terminal. Thus, the efficiency and the output power of the frequency doubler's signal is

dependent upon the biasing point and the output resonator to eliminate all unwanted

voltage components except the second harmonic component. This can be clearly

explained by introducing the MESFET model. Figure 5-1 a) and b) show the physical

form and nonlinear model of the GaAs MESFET, respectively.

Epitaxial layer

Buffer layer

Semi­insulating substrate

s

Source Gate

a)

Cd

b)

Drain

Depletion

region

s

Figure 5- 1: a) The physical form and b) the nonlinear model of the GaAs MESFETs

78

Chapter 5 Frequency Doubler Design Configurations

By applying the gate-source voltage, V gs, and the drain-source voltage V ds, there is a

depletion region under the gate. As the drain-source voltage is increased, the voltage

across the region is greater at the drain end, resulting in the wider depletion region at the

drain end. The current channel becomes narrower due to the longitudinal electric field.

This causes the electrons to move through the channel faster, so the drain current is

increased. However, the velocity of the electron is limited, to approximately 1.3xl07 cmls

in GaAs, and the current does not increase after the drain-source voltage is increased

beyond the value that causes velocity saturation [11].

It is noted that MESFET operation is dependent on the charge mechanism due to the

MESFET's structure and the biasing voltage. This causes the Schottky-barrier formation.

In the MESFET nonlinear model, there exist two capacitances: gate-source capacitance,

Cgs, and gate-drain capacitance, Cgd, associated with the Schottky barrier. These

capacitances are generally nonlinear and are gate-source and drain-source voltage

dependent. Also, the Schottky barrier gives the two diodes [57]. One is drain-source

diode, Dgs, and the other is gate-drain diode, Dgd. Cds is the drain-source capacitance,

which is the metalization capacitance. It is treated as a constant. The other parameters in

the nonlinear model are the gate, source and drain ohmic contact resistances, which are

represented by Rg, Rs, and Rct, respectively. Ri is the resistance of the semiconductor

region under the gate. Moreover, the controlled drain current source !ct, is the important

parameter in the designed circuit and its value relies on the voltage across the gate-source

capacitance.

In the nonlinear model, the input capacitor Cgs and series resistance Ri of the gate-source

junction can be analysed as a lossy diode [59]. The value of the gate-drain capacitance

Cgd also contributes to device nonlinearity. However, this capacitor is, in general, not an

important factor in harmonic generation since the gate-drain capacitance becomes small

as the drain-source voltage increases [69]. Figure 5-2 shows the regions of operation for

frequency doubler operation [54]. In region A, where the bias is near pinch-off, the

MESFET's channel conducts only during the positive half cycle of the input sinusoidal

signal. Thus, the drain current is given as a form of a half-wave rectified signal. Also.

when the gate-source bias is near zero voltage below the forward conduction point (in

region B), the voltage waveform is clipped

79

Chapter 5 Frequency Doubler Design Configurations

<' E -

40 ,-------------------------------------....... _. __ ..... _ .... -..... ----·----1Vgs=OV --- .. -~

-------_ ... -----_._----_. __ .... -._--_._. .-.. _ ..... -_ .. Vgs=-O.222V

i / / ;' ____ .- -' - Vgs=-o.444V

,f // ___ .. ______ -.-- ._--_.--! / //-

/ /;/ _.------. _ _ .... _ .. ___ ..... _ ... _.- -- Vgs=-O.667V Iii /_._._ .. Ii / // .. ~ ... --.---.- Vgs=-1.111 V ;i /' /1 ____ .... ---.. - ••. ---... -~ _.-_ .... _ ... -/?//==-__ . __ .......... _ ......... _- .... _ .. _ .. -._ .. _ ... _. __ -- ~9S=-~ .:::~

f:~::.~:-' - ..... --.-~::::.: ..... -- ...................... --- '. ~ ••........ -- -. V~:::2:000V -2

o 5 Vds(V)

Figure 5- 2: FET output characteristics showing the region of operation for frequency doubler

operation

5.2.1.1 V gs near pinch-off (Class B)

The active device in a class B multiplier is biased in the vicinity of pinch-off. Figure 5-3

shows waveforms and signal trajectory for a class B multiplier [57]. Ideally, the active

device conducts when the input voltage is more than pinch-off. With proper output

matching at the second harmonic frequency, the class B doubler provides the second

harmonic output power given by the product of the second harmonic current component

and the voltage component. The rms average power is obtained [57] as:

(5.4)

80

Chapter 5 Frequency Doubler Design Configurations

IDS IDS IDS(t) .,. . '

IF +~" ,"\

\

1\ I,'

\ I~, " I!

\

I ' Ii ! " "

I'

l :i

'~ i "~

I I

VGS ~) VDS I I ~ I ..

==--::> .. _.---

t t

Figure 5- 3: Waveforms and signal trajectory for class B multiplier

5.2.1.2 V gs in the vicinity of forward conduction and between 0 volt and

pinch-off

In this mode, the device is biased between 0 volt and pinch-off and if the input swing is

large enough, the output at the drain is clipped at both ends. If the output wavefonn is

symmetrical, the even harmonic components will be smaller than the odd hannonic

components [65]. When the PET is biased close to 0 V, the voltage wavefonn across the

gate-source capacitance is clipped due to the junction conduction [59] and will be half­

wave rectified due to the conduction cycles operated by the gate-source diode [65]. This

rectified waveform is transferred to Ids through the device transfer characteristic. By

Fourier analysis, the half-wave rectified waveform consists of a fundamental and a second

harmonic voltage of V /2 and 2V / 31l [59], respectively. Where V represents the drain­

source voltage.

81

t

Chapter 5 Frequency Doubler Design Configurations

5.2.2 Frequency doubler design topologies

5.2.2.1 Single-ended frequency doubler

Figure 5-4 shows a basic single-ended frequency doubler configuration. A quarter­

wavelength open-circuited stub is often used in the single-ended frequency doublers [11],

[57] to supress the fundamental frequency component at the output. The rest of the design

is to consider the input and output matching networks corresponding to fundamental and

second harmonic frequencies, respectively.

Input at fo

r·································~

: i I JJ4 :

~------~ i ! i at fo i Input

Matching atfo

! ! i i : i i ! :

I , i :

:. ................................ J

f····························j i ! i ! ! i j ! ! ir-------~

: : Output i ~

: ! 1J4 i at fo i L ............................ ~

Matching at 2fo

Output at 2fo

Figure 5- 4: The frequency doubler employing A / 4 transmission lines

For high performance single-ended PET frequency doublers, there are many previous

studies examining the input and output networks. Gilmore [54] stated that the optimum

output load at fo is an open circuit shifted by the conjugate of the output capacitance of

the PET. Furthermore, the reflection network has been used in the frequency doubler.

Rauscher [55]-[56] and Borge [60] presented a design method of the second harmonic

reflector type at the output PET frequency doubler. Rauscher's study [55]-[56] explained

that the optimum output load at the fundamental frequency was a short-circuited stub at

an optimal distance from the PET drain. Borge used the same technique in the Darlington

doubler employing bipolar transistors. In addition, other studies are based on the use of

reflector networks. There appear that the input and output reflector networks used in the

PET and HEMT frequency doublers by Iyama [67] and Thomas [68], respectively. For

the latter work, Thomas stated that these reflection networks function as both filtering and

82

Chapter 5 Frequency Doubler Design Configurations

matching for the input and output of the doubler. The second harmonic component is

reflected back into the gate at the proper phase angle as well as the fundamental signal

being properly phased into the drain. As a result of the effect of proper reflection phase

angle, the optimum conversion gain can be obtained.

5.2.2.2 Balanced frequency doublers

The other popular doubler design is a balanced (push-push) doubler [57] which is shown

in figure 5-5. In a balanced frequency doubler, the quarter-wavelength open-circuited stub

is not required. The balanced circuit consists of a pair of nonlinear elements and a 180-

degree input balun [11], [57]. The applied input power is divied by the balun. These anti­

phase inputs are then applied to the two FETs. As a result, the two FETs conduct on

alternating half-wave cycles. Due to the balanced configuration, the second harmonic

components from each drain device are combined in phase whilst the fudamental and

other odd harmonic frequency components are subtracted, combining out of phase.

Therefore, the second harmonic components are summed in power.

Input Matching

/). Output

Input Matching Output

at fo L at 2fo

Input Matching

Figure 5- 5: A balanced (push-push) doubler

The advantage of the balanced frequency doubler is that it gives the elimination of the

fundamental and odd-harmonic frequency components. Also, the wide bandwidth is

obtained [11], [57], althougth the bandwidth of the balanced frequency doubler is limited

by the balun and the imbalance of devices which are ideally identical. Another important

83

Chapter 5 Frequency Doubler Design Configurations

point is that the input and output power can be attenuated due to the insertion loss in the

balun and the power combiner, respectively. This can degrade the conversion gain of the

balanced frequency doubler.

5.3 Practical example of an injection-locked balanced oscillator and

doubler

The demand for high performance oscillators is rapidly increasing in microwave and

millimetre-wave applications. In an advanced millimetre-wave communication system,

such as LMDS (Local Multipoint Distribution System), it has been reported [62] that the

phase-noise of local oscillators in the upconversion and downconversion chain is the main

noise contribution, degrading the bit-error-rate of the communication system. For

collision avoidance radar systems, signals with low phase-noise are also required [63]. To

achieve a highly stabilised signal in the millimetre-wave and microwave bands, several

techniques have been proposed, including the phase-locked loop, injection-locked

oscillators with frequency multipliers, dielectric resonator oscillators, and a number of

other techniques [4]-[5].

The injection-locking technique with frequency multiplication is attractive for generating

a stabilised source in many applications due to its simplicity and low component count. In

Chapter 3, the injection-locked balanced oscillator was described [61]. The technique is

simple and consumes low DC power. Also, it provides balanced stabilised signals with no

external balun, resulting in a small circuit area. In this section, the sub-harmonic

injection-locked balanced oscillator signals directly apply to a balanced frequency

doubler as input. This eliminates the high performance balun, which is required in a

conventional circuit. The technique is demonstrated experimentally in the S-band.

5.3.1 Circuit realisation

The circuit diagram of the proposed balanced oscillator-doubler is shown in figure 5-6.

The sub-harmonic injection-locked balanced oscillator provides a pair of balanced

outputs, at half the final output frequency. These outputs have previously been shown to

have a precise 180-degree phase difference and amplitude balance; more precisely

Chapter 5 Frequency Doubler Design Configurations

balanced, in fact, than is normally obtained with a balun. As mentioned in previous

chapters, the balanced oscillator is based on the extended resonance technique, with a

transmission line connected between the gates of the two active devices, so that the two

active devices resonate with each other [7], [61]. Series feedback, using a short-circuited

stub, is applied at the source of each active device to produce negative resistance. A sub­

harmonic injection-locking signal is applied to the balanced oscillator through the

matching network (M}) at half the free-running frequency. This matching network is

connected at the mid-point of the transmission line between the gates of the active devices

in order to maintain the balanced configuration. The outputs of the balanced oscillator are

immediately fed into a balanced doubler.

I···························································· .. ···r······························································T····················· ············1

i I I I : ~

:

Subharmonic Ml

Injection

Signal, fol2

Wilkinson

Balanced oscillator ! i Doubler combiner !

i !

Balanced

! 1 ! ! : ~ .......................................................................................... .J ..•..••.•.•.•.....•.•••.....•.•..•..•...••.•••..•....•..•..•••..•.. _ ...................... .-..-.. _ .. _ ................. _._ •.•••. _.

Output at2fo

Figure 5- 6: The circuit diagram of the novel sub-harmonic injection-locked balanced oscillator­

doubler configuration

85

Chapter 5 Frequency Doubler Design Configurations

A balanced doubler consists of two identical doublers and a Wilkinson divider. These two

frequency doublers operate at 2-to-4 GHz. The doubler configuration is illustrated in

figure 5-7. The doubler design concept is based on GaAs PETs biased near zero gate­

source voltage [64]-[65]. The optimal matching networks at the fundamental frequency

and second harmonic frequency are used to terminate the input and output of the active

device, respectively. The input and output ports of the doubler are conjugately matched in

order to achieve the maximum power into the input port and to obtain the maximum

power at the output port. In this case, external feedback is used in this doubler design to

compensate for the transistor internal feedback. The external feedback is for

counteracting or complementing the parasitic feedback effects [55], resulting in the

elimination of the parasitic feedback at the frequency of the second harmonic [56]. The

external feedback in the doubler is found to increase the conversion gain, but also

degrades the bandwidth.

r--------------------------------------------------------------------------------------, , ' , '

! (Optimal fo and 2fo external feedback) ! ! ! b <5-------------------------------------------------------~ ~

: : ' I

Input Matching Network

Output Matching Network

Figure 5- 7: Single-ended doubler circuit configuration

The outputs of the two devices in the balanced doubler are combined in a Wilkinson

divider, designed at twice the frequency of the oscillator. At the output of the balanced

doubler the even frequency components are summed, whereas the odd frequency

components - including the fundamental frequency from the oscillator - are cancelled out.

86

Chapter 5 Frequency Doubler Design Configurations

Basically, the Wilkinson power divider functioning as a power combiner consists of t 0

wave sections [8] as shown in figure 5-8. The characteristic impedances of these two

sections - connecting in parallel wi th the input line - are Z2 and Z3, respectively. Since the

Wilkinson power divider operates as a 3-dB power divider. The characteristic impedance

of Z2 and Z3 is equal to .fi Zo, where Zo is the input impedance.

Figure 5- 8: The Wilkinson power divider

5.3.2 Results and Discussion

The sub-hannonic balanced oscillator was designed and fabricated using FR4 board and

Infineon CFY30 OaAs PETs as active devices. The oscillator was designed to operate at

about 20Hz, so that the output of the complete circuit was at almost 40Hz. The balanced

oscillator and the doubler circuits were biased at V ds = 2.2 V, V gs = -0.97 V and V ds = 3.0

V, Vgs = -0.3 V, respectively. From this, the balanced oscillator gives a free-running

frequency of 1.9460Hz. Also at the output of a power combiner the second harmonic

frequency component is -0.67dBm, whereas the fundamental and the third harmonic

components are -18.67dBm and -23 .67dBm, respectively. Figure. 5-9 shows the output

spectrum of the balanced oscillator-doubler. These measured results include a 3-dB

attenuator and 0.5-dB cable loss at the output port.

7

-..

= = "'0 --Qi "'0 ::s ...... 'a OJ) ec ~

0

-10

-20

-30

-40

-50

-60

-70

-80

ATTEN20dB RL lO.OdBm

CENTER 3.893GHz RBW 1.0MHz

Chapter 5 Frequency Doubler Design Configurations

lOdBI

VBW 1.0MHz

MRK -O.67dBm 3.901GHz

SPAN 5.000GHz SWP IOO.Oms

Figure 5- 9: The output spectrum of the balanced oscillator-doubler circuit (with 3-dB attenuator and

without sub-harmonic injection-locking signal O.5-dB cable loss at the output)

To investigate the use of the injection-locked oscillator with the balanced doubler, a

subharmonic locking signal of 974.8 MHz with a power level of -2.0dBm was injected

into the balanced oscillator. It was noticed that with the integrated balanced oscillator and

balanced doubler the oscillator still maintained balanced characteristics. In the locking

state, the magnitude and phase of the balanced oscillator are equal with 180-degree phase

difference. The power output of the balanced oscillator-doubler is slightly degraded, but

the phase noise of the output signal is improved. Figure 5-10 illustrates the measured

output in the locking state compared to that output without injection signal. The phase

noise is reduced more than 15 dB at the offset frequency of 50 kHz. Moreover, the

locking range as a function of the injection power level was investigated. Figure 5-11

shows the measured locking range of the sub-harmonic injection-locked oscillator-

doubler for injection power levels from -4 dBm to 4 dBm.

88

Chapter 5 Frequency Doubler Design Conjigul"uJ,v)d

o -10

-20

-30

~ -40 .a 'c -50 ?I ~ -60

-70

-80

ATTEN20dB RL 10.0dBm

Output with injection-locking

[email protected]

10dBI

CENTER 3.8936640G Hz RBW3.0kHz VBW3.0kHz

MRK -1.67dBm 3.8936640GHz

Output without injection signal

SPAN 500.0kHz SWP 140ms

Figure 5- 10: The comparison between the output of the balanced oscillator-doubler, with and

without subharmonic injection-locking signal (0.5dB cable loss at the output and 3dB attentuator)

"........, 1600

~ 1400 '-'

~ 1200 "'0 1000 '~ "'0 800 = ~ .0. 600 ell = 400

:£2 ~ 200 0 ~

0

-4 -2 0 2 4

injection power (dBm)

Figure 5- 11: The measured locking range of the subharmonic injection-locked oscillator-doubler as a

function of injection power level

89

Chapter 5 Frequency Doubler Design Configurations

5.4 Conclusion

In design of frequency doublers , the input signal is fed into nonlinear devices, such as

diodes and PETs. The input and output matching circuits are used to short circuit at all

unwanted harmonics of the excitation frequency at the input and output, respectively.

Practically, the input impedance is conjugately matched to the active device so that the

maximum energy can transfer from the source to the device. In addition, the resonant

circuit is employed to provide the output matching at the desired harmonic.

Since the frequency doublers exploit the MESFET's nonlinearity [11], [57], the

performance of the doubler is dependent on bias point and the design topology. To

achieve a good doubler design, the nonlinear device is used in the stable condition, like

the design procedure for amplifiers. For the biasing consideration, the PET biased in the

vicinity of pinch-off generates a half-wave sinusoidal current being rich III even

harmonics [57]. The device functions as a half-wave rectifier, resulting III good

efficiency. Another biasing case is near zero voltage producing the waveform close to the

half-wave rectified waveform. In this region, the device can give a high output voltage.

However, there can be an unacceptable conversion loss and high dc power dissipation in

the active device. The other important design consideration is the choice of topology: a

single-ended or a balanced frequency doubler. The basic design principles for these

doublers have been explained. Also, the design techniques to improve the single-ended

doubler have been summarised.

As the applicable example, an injection-locked balanced oscillator and doubler has been

introduced. By using the reflector network type [55]-[56], the balanced frequency doubler

gave good conversion gain. The injection-locked oscillator is also used in order to

stabilise the oscillating signal. This example shows the feasibility of the integrated

balanced oscillator and balanced doubler. The advantage of this work illustrates the low

phase noise output signal and small circuit size because the external balun is not required.

90

Chapter 6 Frequency Doubler using Feedforward Technique

Chapter 6

6 Frequency Doubler • uSing Feedforward

Technique

6.1 Introduction

There is a great demand for high spectral purity signals in microwave and millimetre­

wave communication and radar systems. Frequency multipliers are widely used in a

variety of applications, especially in frequency translation circuits, in order to extend the

frequency limit of fixed or variable frequency low phase-noise oscillators. Several

doubler techniques have been studied in the literature [55]-[56], [70]-[72]. Active

multipliers, in general, provide conversion gain and wider bandwidth compared to the

varactor diode type. The nonlinear transconductance is the most important mechanism for

a harmonic generator. Active doublers, such as FET doublers, are normally used at a bias

point in the vicinity of pinch-off, allowing the active devices to generate a high level of

even harmonics. There are two main doubler circuit design techniques as described in the

previous chapter. One is the use of quarter-wavelength open-circuit stub. This technique

is often used in a single-ended doubler to suppress the fundamental frequency at the

output [67]. Another popular doubler design is a balanced doubler technique. In this

topology the fundamental frequency is cancelled out due to the summation of odd

harmonics which are out-of-phase at the output of the circuit, whist the second harmonic

is added [63], [66], [70]. However, these doubler design techniques generally provide

only 25dB of fundamental signal suppression due to the practical limitations of open-stub

and balun circuits in the single-ended and balanced doublers, respectively. For fully

monolithic transceivers it is highly desirable to increase the fundamental suppression.

91

Chapter 6 Frequency Doubler using Feedforward Technique

In this chapter, a new method of fundamental frequency suppression in a frequency

doubler is proposed. The feedforward technique has been widely applied for distortion

cancellation in amplifiers [40], [73]-[78]. The principle of the feedforward technique in a

power amplifier is viewed as the sum of a replica of the input and error signals as shown

in figure 6-1. The degree of distortion cancellation in the power amplifier is dependent on

the carrier suppression in the signal cancellation loop, and having good phase and

magnitude balance in the error cancellation loop [40].

Main power amplifier

Vin delay 2 + Vout

(jerror c::Oc;'/atJon _

attenuator

signal cancel/atlo

(j de/ay1

Error amplifier

Figure 6- 1: Feedforward technique for a linearised power amplifier

The use of the feedforward technique for substantially eliminating the fundamental

frequency component of the frequency doubler is described. To do this, a simplified

representation of using the feedforward technique for fundamental component

suppression in a I-to-2GHz frequency doubler is demonstrated. Unlike the use of the

feedforward technique in power amplifiers, the output of the frequency doubler itself has

to be well estimated in order to ensure the magnitude balance between the fundamental

components of the doubler output path and the reference path. Also, the phase balance

between these two paths is determined in advance.

92

Chapter 6 Frequency Doubler using Feedforward Technique

To demonstrate the fundamental suppression by using the feedforward technique, the

design of the 1-to-2GHz doubler and the 1GHz directional couplers functioning as a

splitter and subtractor are presented. In the doubler design, a filter is required to first

suppress the fundamental component at the doubler output; otherwise, the fundamental

power level of this path will be larger than that in the reference path. The fundamental

component cancellation may not occur at the output of the second coupler. Additionally, a

reflection-type analogue phase shifter is also designed at 1 GHz. This phase shifter is used

to adjust phase balance of the fundamental frequency components between the reference

and the doubler paths.

Due to the amplitude and phase imbalance between the doubler and the reference paths,

the measurement results show the degree of fundamental suppression at the output of the

second directional coupler: More than 50dBc fundamental frequency suppression can be

achieved. The measured second harmonic output and the fundamental frequency power

levels as a function of the input power levels are also demonstrated.

6.2 Design

Figure 6-2 illustrates the feedforward technique for fundamental signal suppression in the

frequency doubler. The circuit consists of a 1-to-2 GHz frequency doubler, phase shifter,

and input and output couplers. The input power is split into two paths by using coupler 1.

One goes to the reference path, consisting of a phase shifter, and the other goes to the

main path, consisting of the frequency doubler. The reference path is generally chosen to

have low loss and is also used to adjust the phase balance. Substantial fundamental

component cancellation occurs when the fundamental frequency power of the two paths is

added in anti-phase by employing coupler 2. Therefore, the coupler 2 functions as a

subtractor for the fundamental component.

93

Chapter 6 Frequency Doubler using Feedforward Technique

IGHz Input

Coupler 1 Coupler 2 I-to-2GHz

Multiplier

<I>

Variable phase Variable gain/attenuator

(may be needed)

2GHz Output

Figure 6- 2: Feedforward technique for fundamental signal suppression in a frequency doubler

In order to achieve the maximum fundamental frequency cancellation, the magnitude and

phase of the outputs of frequency doubler and phase shifter are determined. Ideally, the

magnitude and phase of the fundamental components between these two outputs must be

equal and I80-degree out-of-phase. Therefore, the degree of cancellation relies on the

fundamental power level of the doubler output and the two couplers. Also, it is dependent

on the phase adjustment of the phase shifter.

Frequency doubler path consists of frequency doubler and filter. With the filter connected

at the output of the frequency doubler, the power level of the fundamental component is

less than the power level of the second harmonic component at the output of the

frequency doubler path. Compared to the reference path, the fundamental power level at

the output of the filter in the doubler path should be the same as the power level in the

reference path, so that the fundamental component can be cancelled out by using coupler

2, which functions as the subtractor.

In the feedforward system, the higher input power generally occurs at the doubler to

produce the high power level of the second harmonic output. The small coupling factor is

94

Chapter 6 Frequency Doubler using Feedfonvard Technique

considered. Like the coupler!, the coupler 2 should also have small value of the coupling

factor. If the coupling factor of coupler 2 is high, the second harmonic component will be

degraded at the output of the coupler 2. The other reason for small coupling factor is that

the insertion loss is smaller.

A reflection-type analogue phase shifter is used for phase adjustment in the reference

path. The designed circuit is fabricated on FR4 board and the BB833 diodes are used as

the varactor diodes in order to control the tuning phase. Also, both the measured phase

and insertion loss are tested.

6.3 Circuit Realisations

6.3.1 Single-ended Doubler Design

The doubler design is based on the PET active device biased in the vicinity of pinch-off.

This leads to an output signal rich in even harmonics. Figure 6-3 shows the doubler

circuit configuration [55]-[56]. The output of the active device is optimally matched so

that maximum output power can be obtained at the designed harmonic, 2fo. This output

matching network also functions as a bandpass filter, providing some suppression of the

fundamental and other harmonic frequencies.

,.----_.-----------_ ... -------------------------------------------------------------------, , , , , i (Optimal fo and 2fo external feedback) i : I , , , , , , I I : .--------------------------------------------------------¢ ¢

<> ¢ , , : : I I

Input Matching Network

ZIN

Output Matching Network

Figure 6- 3: Single-ended doubler circuit configuration

95

Chapter 6 Frequency Doubler using Feedforward Technique

The input matching network is used to achieve matching into the gate of the device at the

fundamental frequency. Practically, the input port and output port of the doubler are

conjugately matched at the fundamental frequency and second harmonic frequenc .

respectively. An important aspect of this circuit is that this doubler configuration utilises

external feedback. Due to the transistor internal feedback, the incorporated external

feedback is for counteracting the parasitic feedback effects [55]. This eliminates the

second harmonic parasitic feedback loss [56] . However, the use of the second harmonic

feedback involves a direct trade-off between the conversion gain and bandwidth. In this

case the external feedback is employed in order to increase the conversion gain for the 1-

to-2GHz doubler. The doubler design in this work focuses on using the feedforward

technique to dramatically reduce the fundamental signal at the output.

The I-to-2GHz frequency doubler was designed and constructed on FR4 using Infineon

CFY30 GaAs FETs as active devices. A transmission line with coupling capacitors is

used to implement the external feedback. To investigate the effect of the external

feedback, the input and output matching networks are designed to match 50-ohm for the

fundamental and second harmonic frequency, respectively. This was achieved by using

harmonic-balance analysis in the Libra program. Figure 6-4 shows the doubler circuit

board using the external feedback, based on FR4.

~VDS

Input,

Figure 6- 4: Single-ended doubler circuit

Output. f· ut= ~fin

,/

Chapter 6 Frequency Doubler using F eedforward Technique

In order to achieve the fundamental frequency cancellation the magnitude of the

fundamental frequency between the doubler and the reference paths are carefullv

examined. The magnitude of the two paths should be accurately estimated. A filter is

introduced in the doubler path. In this case, the coupled line is used to function as the

lumped elements [78]. A coupled line can be modelled as an equivalent lumped element

circuit. The filter configuration is shown in figure 6-5. Also, figure 6-6 illustrates the

simulation result. From this, the signal can be suppressed at the frequency of 1 GHz whist

the second harmonic is passed at the frequency of 2GHz.

Port 1

Port 2

Figure 6- 5: Filter configuration

+ 511 )( 512 .522

0.0 '--'---Ci;::::=:p~,----r::l

iii' -10.0 "C -Q)

-g -20.0 I--+l+--H+---ti--r--t--t--i ~ r::: C) CIS :E -30.0 1---lI---J---+-\t----t--t----"1

-40.0 0.0

Frequency 1.0 GHzlDIV

Figure 6- 6: Simulation result of the filter

6.0

97

Chapter 6 Frequency Doubler using Feedfonvard Technique

6.3.2 Reflection-type analogue phase shifter and directional coupler

A phase shifter is represented as the variable phase block shown in figure 6-2. The

function of the phase shifter is steady control of the relative phase between the input and

output with a flat group delay frequency response [5]. A reflection-type analogue phase

shifter consists of a 3-dB Lange coupler and terminating impedances, ZT, reali sed with

varactor diodes. Figure 6-7 and figure 6-8 show the reflection-type phase shifter

configuration and its layout. The desired phase shift can be obtained by changing the bias

of the varactor diodes. This phase shifter was designed for 1 GHz based on FR4. The

BB833s are used as the varactor diodes. Also, a pair of lOdB directional couplers was

constructed on FR4.

3dB quadrature directional coupler

~ / identical reflection terminations

Figure 6- 7: The reflection-type phase shifter configuration

..- Tuning voltage

Figure 6- 8: The layout of reflection-type phase shifter at IGHz

9

Chapter 6 Frequency Doubler using Feedfonvard Technique

6.4 Results and discussion

The filter performance was first investigated. Figure 6-9 shows the measurement result

from the network analyser. This result offers a good agreement with the simulation result

as shown in figure 6-9. With the filter connected to the doubler, the complete doubler

circuit was biased at VDS = 3.0 V and VGS = -2.2 V, relative to a pinch-off voltage of -2.0

V. An input power level of -ldBm was applied to the doubler. Figure 6-10 shows the

output spectrum of the frequency doubler without applying the feedforward. It shows that

the power level of the fundamental component is -26.17 dBm. This measured result is

taken from an HP8563E spectrum analyzer, which is connected via a 3-dB attenuator and

a cable with 0.5-dB loss. Therefore, this doubler provides 0.5 dB gain. Compared to the

doubler output without connecting the filter, the power level of the fundamental

component is reduced by approximately 23 dB. However, the second harmonic

component is slightly degraded due to the insertion loss in the filter.

+ 511 0521 .522

m -1 O.O---ir---~f----f----\+---+-------j

" -Q)

" :l ~ C 0) n:s :E ·30. Of--------fJf--t-------tt---Jf------i----tH

-40. 0 L.-__ J.. __ ---L ___ --L-__ --'

0.0 4.0 Frequency 1.0 GHzlDIV

Figure 6. 9: The measurement result of the filter from the network analyzer

99

-10

-20

-30

-40

~ -50 .a -a -60 Oil ~ ~ -70

-80

-90

Chapter 6 Frequency Doubler using Feedforward Technique

ATTEN 10dB RL OdBm 10dBI

MRK -4.00dBm 2.000GHz

r-~---+---+--~--4---~-----, -

CENTER 2.000GHz RL 1.0MHz VBW 1.0MHz

SPAN 2.2S0GHz SPW SO. Oms

Figure 6- 10: Output spectrum of the 1-to-2GHz doubler without the feedforward technique (with a

3-dB attenuator and O.S-dB cable loss at the ouput)

To investigate the fundamental frequency suppression technique, the two couplers and

reflection-type phase shifter were connected as shown in Figure 6-2. The phase of the

phase shifter was tuned by varying the voltage applied to the varactor diodes. The phase

is variable between _35 0 and 1400 while the insertion loss varies between -1.9 and 0 dB

as shown in figure 6-11.

100

Chapter 6 Frequency Doubler using Feedforward Technique

140

120

100

- 80 ~ ;..

60 OJ) Q,I

"'0 -Q,I 40 rIl ~ .c 20 ~

0

-20

,,--r---r- - i"",

~J--J- __ ~ _____ ~J_~~_~~ , • , I I ,', . I • I ,

+----------~---- ------~-------~~----~~-~--i , I ,i I I ii, ' ,

+------------ ----------~--------------~-~ I' ,. " I I I I 'I + ~ -.. _1- •• _'_ _ _ _ _ __ ' ___ ' _ ~ __ ~. _' ___ ' ______ ~ __ .• _: _ ~ __

. I , I •

t I • •• I I h (d '\ __ .. __ .. __ .. _________ .. __ .. __ -+-p ase eg ree J

.... ' I Ii, 'I .• - -

•• , I • , 'I ----s (dB) I I _~_ ___ _ ~~~2~1~~~~~

..., ", - - - - -. - - -. - i - - i - ,--= -,- - - - - - i - - - - - i -I-

I ~'illlllll'IIII",

I I I I I I I I I I I I I .~ I ' I I I I I I I , I I I' I

: - - - ,- : - ,- ~ - - ~ ..... j -if • -.- .. ----.. ----- ~ --

0

-0.2

-0.4

-0.6

-0.8 -~ -1.0

"'0 --N

-1.2 rF1

-1.4

-1.6

-1.8 -40 , I I I I I I I I -2.0

Tuning voltage (Volts)

Figure 6- 11: Phase and insertion loss of phase shifter

After adjusting the phase of the fundamental frequency component from the feedforward

path and the doubler path, the signals from the two paths are combined using the second

coupler. Consequently, the output of the second coupler is observed. Figure 6-12 shows

the output spectrum of the doubler using the feedforward technique. It is clearly shown

that the power level of the second harmonic component is -5.17 dBm whereas that of the

fundamental frequency component is -65.17 dBm, where the phase of the feedforward

path is set to about 85°. A conversion loss ofO.67dB was obtained from this experiment.

Comparing the fundamental frequency power level of the doubler with and without the

feedforward branch, the fundamental frequency component is substantially improved,

with more than 50 dBc achieved at an input power level of -0.82 dBm. The second­

harmonic power level is attenuated by 1.17 dB due to the loss in the additional

components. Finally, the proposed frequency doubler with the feedforward technique is

also measured as a function of input power. Figure 6-13 shows the graph of the output

power at the fundamental and second harmonic frequencies vs. the input power.

101

-10

-20

-30

-40

~ -50

2 '8 -60 ~ = ~ -70

-80

-90

ATTEN 10dB RL OdBm

CENTER 2.000GHz RL 1.0MHz

Chapter 6 Frequency Doubler using Feedfonvard Technique

10dBI

VBW 1.0MHz

~IRK -5.17dBm 2.000GHz

SPAN 2.250GHz SPW 50.0ms

Figure 6- 12: Output of the 1-2 GHz doubler with feedforward (with a 3-dB attenuator and 0.5-dB

cable loss at the ouput)

0

-1

IS -2 G -3 f"l

~ -4 .... = -5 .e-= 0 -6 ... ~

~ -7 0 ~ -8

-9

-10

I~ I I :

'J:_-::-~ .. :~:..=.-=-~....-...::...:.i:....:..:::~-~~~-~-:...:....:..:.i.:....::...:-"i~~--~--~--

N QC)

N I

, I I I j I I I I I I

__ - .., - - - - r ... - - i - ...... -1- - - - ,. - - - -,- - - - T - - - ., - - - - .... - - - , - - - ... r - - -I I I I I I I I

I I I I I I I I I I t ___ ~ ____ ~ ___ J ____ I ____ l ____ I ____ l ___ ~ ____ L ___ J- ___ L __ -

I , I I I I I I I , I , , ,. : I t I I I I I ___ .., ____ r ___ i ____ I ____ ~----,----r---.,----r---~----r---

I I I I I'

I I I I I I ____ , ____ L ___ J ____ 1 ____ .J. ____ , ____ 1 ___ J - - -

I I I I I I I I

, , ---~----~---~ ___ I_---~-- .

I I I I , , I I I I

---~----r---'-- -,--- ,----, , , ,

___ J ____ ~ ___ ~ ________ ~ ___ .

I I I I I

~ N ~ ~ N QC)

rt'l QC) ~ QC) ~ ~

N ~ ~ Q Q Q I I I I I

-J:-

QC) \C Q

at_2_GHz

at_l_GHz

Q() QC) Q() ~ '>C ~

~ ~ N

Input power (dBm)

Q() '>C N

0

-10 IS

== -20 ~ ...-4 .... eo:

-30 .... = ~ ....

-40 = 0 ... ~

-50 ~ 0 ~

-60

,.,1\ QC) ~

~

Figure 6- 13: Measured output of the frequency doubler as a function of the input power le\'el

102

Chapter 6 Frequency Doubler using Feedforward Technique

As it is seen in the experimental results, the fun dam an tal component can be suppressed to

better than 50 dBc. Using the feedforward technique in the doubler gi yes an excellent

fundamental component cancellation. However, using this technique suffers from several

shortcomings [40], similarly to the power amplifier case. First, the phase shifter in the

implementation requires passive components, such as microstrip, giving power loss. This

could affect the unequal magnitude cancellation at the second combiner which is the

second coupler. In addition, these components can possibly degrade the second harmonic

component performance. Secondly, the amount of fundamental cancellation is dependent

on the magnitude and phase matching of signals sensed by the combiner and the phase

shifter.

In order to obtain improved purity of the doubler output signal, the feedforward technique

can be applied as depicted in figure 6-14 [40]. The error amplifier is rich in odd harmonic

frequency components whereas the output of the doubler is rich in even hamonics,

especially the second harmonic component. Then, with appropriate phase and magnitude

match between the two paths, the odd harmonics can be removed by the second combiner.

Compared to the previous topology shown in figure 6.2, the the purity of the doubler

output might be improved by eliminating all odd hamonic components.

t r i r I r i 1 f f 2f 3f f 2f 3f + 2f o 0 0

0 000 0

x2 + Input

Error Output cancellation

Single loop cancellation loop

I i f 3f

Av1 0 0

Av

Amplifier Attenuator Error amplifier

Figure 6- 14: Feedforward technique for odd harmonic component suppression in frequency doubler

103

Chapter 6 Frequency Doubler using Feedforward Technique

6.5 Conclusion

The broad conclusion of this chapter is that there is more to be learnt about the subject of

fundamental frequency component elimination in the doubler by using the feedforward

technique. This chapter presents a technique to eliminate the fundamental frequency

component in a frequency doubler. The use of the feedforward technique is proposed to

be a convenient way to solve the problem associated with the purity of the output signal

in the frequency doubler. The concept of this technique is dependent upon the amplitude

and phase balance of the fundamental frequency components between the doubler and

reference paths.

The measured results demonstrate the effectiveness of this proposed technique. It is

shown that the fundamental frequency component at the output of the second coupler is

tremendously improved to better than 50dBc. In addition, this design technique is simple

and can be wide-band, depending on the doubler design topology. However, the drawback

in the use of the feedforward technique in doubler is the sensitivity and the additional loss

in the passive components, which can degrade the conversion gain of the doubler.

For the improved purity of the doubler output, the feedforward technique is also

introduced to eliminate the odd harmonics. The topology needs the error amplifier to

generate the odd harmonic components in the error cancellation loop. The purity of the

output signal is dependent on the phase and magnitude imbalance between the output of

the signal cancellation loop and the output of the error amplifier. However, this topology

can degrade the conversion gain of the doubler due to the passive elements such as

couplers.

10.+

Chapter 7 Conclusions and Suggestions for Future work

Chapter 7

7 Conclusions and Suggestions for Future

Work

The aim of this research is to improve the spectral purity performance of microwave

balanced oscillators. This is to achieve the purity improvement in the balanced oscillator

where the stabilisation and phase noise are considered. The injection-locking technique is

focused on getting a better-stabilised oscillating signal and obtaining low phase noise.

The oscillator frequency is also restricted to the limitations of the active devices. But the

restriction can be solved by using frequency doublers. A new technique to suppress the

unwanted output signals is proposed in this thesis. All the circuit designs are based on a

frequency of 20Hz for ease of fabrication.

The main achievement of this research can be summarised as follows:

• Analysis of the electrical length of the transmission line used in the balanced

oscillator, where the extended resonance technique is used to determine the

oscillator conditions for the balanced oscillator. A self-oscillating mIXer IS

illustrated as an example application of the balanced oscillator.

• Using the injection-locking technique in a balanced oscillator. This is achieved by

using the matching network connecting at the centre of the transmission line.

where it is connected between the two active devices.

• By extending Kurokawa's work, the equivalent circuit of the oscillator has been

characterised as both negative-resistance and feedback models. For a harmonic and

subharmonic injection signal, both models are used to analyse the locking range.

The harmonic signals are generated due to the nonlinearity of the active device and

these signals are fed back into the feedback network, where the negative-resistance

model is characterised. At this point the locking range is considered by using

Kurokawa's theory [21]-[22].

lOS

Chapter 7 Conclusions and Suggestions/or Future H'ork

The agreement between experiment results and the analysis shows that the locking

range of the injection-locked balanced oscillator for a harmonic frequency is higher

than that for the fundamental, and also higher than that for the subharmonic

frequency, when the injection power level is the same.

• The power combining technique is used for the cascaded oscillator. Rather than

focusing on the optimised combined power, the additional sidebands, which are the

effect of the subharmonic injection-locking phenomenon, are investigated. The

experiment shows that the effective sidebands are further reduced in the cascaded

oscillator compared to the single-ended oscillator. The degree of the sideband

suppression is about 30 dB.

• The PET frequency doubler design considerations are considered. To achieve gain,

the biasing and the PET frequency doubler configurations are taken into account.

An integrated subharmonic injection-locked balanced oscillator-balanced doubler

is introduced to demonstrate an example application of the frequency doubler.

• A novel technique for the fundamental frequency component supression In a

frequency doubler is introduced. The principle of feedforward, as applied to

distortion cancellation in amplifiers, is used in the frequency doubler. The

experimental result shows that the fundamental component is significantly

suppressed to better than 50 dBc.

Papers published on this research include 2 papers in the European Microwave

conference [79]-[80], one at the the IEEE Microwave MTI-S Int. Symp. Dig.[81], two at

the European Frequency and Time Forum [82]-[83] and two publications in lEE

Electronics Letters [61], [84].

There are two interesting areas as which can be pursued further. The first is to further

improve the phase noise in the oscillator: The injection-locked balanced oscillator can be

viewed as a distributed oscillator. By using the concept of the power dissipation and phase

noise [40], figure 7-1 illustrates the addition of the output voltages of N identical

oscillators. It is assumed that the noise sources of different oscillators are uncorrelated

whist the relationship between the carriers is correlated [85]. For simplicity, the addition

of two output voltages from the two identical oscillators is given as an example. Let us

106

Chapter 7 Conclusions and Suggestions/or Future work

assume that the carner voltages of these two identical oscillators are Vc\ and Vec

'

respectively. Thus, the result of the summation of two carrier voltages can be expressed as

(7.1)

However, the total noise power related to the carrier is calculated [85] as follows:

(7.2)

Where Vn1 and Vn2 represent the noise voltage related to the carrier with the offset

frequency (fm)' Intuitively, the output voltages of N identical oscillators are added [40],

thus, the proportion of the total carrier to the total phase noise related to the carrier

increases. Therefore, the phase noise related to the carrier decreases.

However, it might be difficult to combine the N identical oscillators, which generate

exactly the same carrier frequency. To overcome this difficulty, the injection-locking

technique is introduced. Figure 7-2 shows a distributed oscillator by using the concept of

the power dissipation and phase noise as well as the injection-locked technique. The

major focus here is to obtain the second harmonic of the oscillating signal. The balanced

oscillator is used in this approach. As a result, the output voltages can increase. With the

uncorrelated relationship between the noise sources of the N identical oscillators, the total

phase noise will decrease. It may be concluded that one of the advantages of the injection­

locked distributed oscillator topology is to provide the more stabilised output signal with

far lower phase noise. The other is that the injection signal can be either at the

fundamental or a sub-harmonic frequency of the signal. However, this topology might

consume more power and the size of the circuit might be large.

107

Chapter 7 Conclusions and Suggestions for Future work

0)0 oscillator 1 .......-----

\ oscillator 2\1-------• • •

oscillator N !--____ ...J

Figure 7- 1: Concept of the power dissipation and phase noise

subharmonic

injection~ signal, fo

2

power splitter •

Balanced oscillator

• •

Balanced oscillator

Wilkinson combiner

Wilkinson combiner

Figure 7- 2: A distributed oscillator

,--------, output,

power combiner

2fo

108

Chapter 7 Conclusions and Suggestions for Future )'.·ork

The other possible idea is to eliminate the unwanted signals in the frequency doubler. By

extending the work in chapter 6, this could be done by using the feedforward technique.

Referring to figure 6-14, the second loop, known as the error loop, is introduced. Here,

the error amplifier will produce a signal which is rich in odd harmonic components. For

the correct magnitude and phase, the signals from the first loop and the output of the error

amplifier at the second loop will be subtracted, resulting in the odd harmonic component

cancellation. However, the output can be degraded due the loss in the passive elements in

the circuit.

109

References

References

[1] M.K.Nezami, "Evaluate The Impact Of Phase Noise On Receiver Performance Phase­

noise analysis, Part I", Microwave&RF Journal, pp.165-172, May 1998

[2] B.Petrovic, "A Balanced RF Oscillator", RF Design, pp. 35-38, December 1989

[3] J.Millman and C.C.Halkias, " INTEGRATED ELECTRONICS: ANALOG AND

DIGITAL CIRCUITS AND SYSTEMS", McGRAW-HILL, Inc., 1972

[4] G.Gonzales, "Microwave Transistor Amplifier Analysis and Design", 2nd edition,

1996

[5] LD.Robertson, "MMIC Design", The institution of Electrical Engineer, London UK,

1995

[6] A.Mortazawi, and B.C.De Loach, Jr., "Spatial Power Combining Oscillators Based on

Extended Resonance Technique", IEEE Transactions on Microwave Theory and

Techniques, vol. 42, No. 12, pp.2222-2228, December 1994

[7] K.S.Ang, S.Nam, and LD.Robertson, "A Balanced Monolithic Oscillator at K-Band",

IEEE Radio Frequency Integrated Circuits (RFIC) Symposium, pp.163, 1999

[8] R.E.Collin, "Foundations of Microwave engineering", 2nd edition, McGraw-Hill, 1992

[9] S.A.Maas, "Nonlinear Microwave Circuits", Arttech House Inc., 1988

[10] LD.Robertson, and S.Lucyszyn, "RFIC and MMIC design and technology", The

Institution of Electrical Engineer, chapter 6 Oscillators, K.K.M.Cheng, pp. 257-276,

London, UK, 2001

[11] S.A. Maas, "Microwave Mixers", 2nd edition, Arttech House Inc., 1993

[12] V.D.Hwang and T.Itoh , "A Quasi-Optical HEMT Self-Oscillating Mixer", pp.1093-

1096, IEEE Int' MTT-s Digest, 1988

[13] X.Zhou and A.S.Daryoush, "A Push-pull Self-oscillating mixer for optically FED

Phased Array", IEEE Int' MTT-s Digest, pp.321-324, 1993

110

References

[14] X.Zhou, and A.S.Daryoush, "An efficient self-oscillating mixer for communication",

IEEE Transactions on Microwave Theory and Techniques, vo1.42, pp.1858-1862,

October 1994

[15] X.Zhou, X.Zhang, and A.S.Daryoush, "A New Approach for a Phase Controlled

Self-oscillating Mixer", IEEE Transactions on Microwave Theory and Techniques.

vol.45, pp.196-204, February 1997

[16] C.M.Montiel, L.Fan, and K.Chang, "An X-Band Self-Mixing Oscillator Antenna for

Transceiver and Spiral Power-Combining Applications", IEEE Transactions on

Microwave Theory and Techniques, vol.46, pp.1546-1551, October 1998

[17] MJ.Roberts, S.Iezekiel, and C.M.Snowden, "A W -Band Self-Oscillating

Subharmonic MMIC Mixer", IEEE Transactions on Microwave Theory and

Techniques, vol. 12, pp.2104-2108, December 1998

[18] M.Claassen, and U.Guttich, "Conversion Matrix and Gain of Self-Oscillating

Mixers", IEEE Transactions on Microwave Theory and Techniques, vol.1, pp.25-30,

January 1991.

[19] K.Kamozaki, N.Kurita, T.Tanimoto, H.Ohta, T.Nakamura, and H.Kondoh, "50-100

GHz Octave Band MMIC Mixers", IEEE Radio Frequency Integrated Circuit

Symposium, pp.95-98, 1997

[20] R.Adler, " A Study of Locking Phenomena in Oscillators", Proceedings of the 1.R.E.

and Waves and Electrons, pp.351-357, June 1946.

[21] K.Kurokawa, " Noise in Synchronized Oscillators", IEEE Transactions on

Microwave Theory and Techniques, vol. MTT-16, No.4, pp.234-240, April 1968

[22] K.Kurokawa, " Injection Locking of Microwave Solid-State Oscillators",

Proceedings of the IEEE, vol. 61, No. 10, pp. 1386-1410, October 1973

[23] A.S.Daryoush, "Optical synchronization of millimeter wave oscillator for distributed

architecture (invited paper), IEEE Transactions on Microwave Theory and

Techniques, vol. 38, pp.467-476, May 1990

[24] A.S Daryoush, "Comparison of Indirect Optical Injection-locking Technique of

Multiple X-band Oscillators," IEEE Trans. Microwave Theory and Techniques. MTT-

34, vol. 12, pp. 1363-1370, December 1986

111

References

[25] J.Birkeland, T. Itoh, "Spatial Power Combining Using Push-Pull PET Oscillators

with Microstrip Patch Resonators", International Microwave Symposium Digest, pp.

1217-1219, 1990

[26] X.Zhou and A.S.Daryoush, "An Injection Locked Push-Pull Oscillator at Ku-Band".

IEEE Microwave and Guide Wave Letters, Vol.3, No.8, August 1993

[27] J.Birekeland and TatsuoItah, " A 16 Elementary Quasi-Optical PET Oscillator Power

Combining Array with External Injection Locking", IEEE Transactions on

Microwave Theory and Techniques, vol. 40, No.3, ppA75-481, March 1992

[28] X.zhang, X.Zhou and A.S.Daryoush, "A Theoretical Study of the Noise Behavior of

Subharmonically Injection Locked Oscillator", IEEE Transactions on Microwave

Theory and Techniques, vol. 40, No.5, pp.895-902, May 1992

[29] X.Zhang, X.Zhou, B.Aliener and A.S.Daryoush, "A Study of Subharmonic Injection

Locking for Local Oscillators", IEEE Transactions on Microwave Theory and

Techniques, vol. 2, No.3, pp.97-99, March 1992

[30] P.R.Herczfeld, A.S.Daryoush, A.Rosen, A.K.Sharma and V.M.Contarino, " Indirect

Subharmonic Optical Injection Locking of a Millimeter-Wave IMPATT Oscillator",

IEEE Transactions on Microwave Theory and Techniques, vol. MTT-34, No.12,

pp.1371-1375, December 1986

[31] C-C.Huang and T-H.Chu, " Analysis of MESFET Injection-Locked Oscillator in

Fundamental Mode of Operation", IEEE Transactions on Microwave Theory and

Techniques, vol.42, No.10, pp.1851-1857, October 1994

[32] H.P.Moyer and A.S.Daryoush, " A Unified Analytical Model and Experimental

Validations of Injection-locking Processes", IEEE Transactions on Microwave Theory

and Techniques, vol.48, NoA, ppA93-499, April 2000

[33] M.E.Hines, J-C.R.Collinet and lG.Ondria, " PM Noise Suppression of an Injection

Phase-Locked Oscillator", IEEE Transactions on Microwave Theory and Techniques,

vol. MTT-16, No.9, pp.738-742, September 1968

[34] M.Odyniec, "Nonlinear Synchrnized LC Oscillators: Theory and Simulation", IEEE

Transactions on Microwave Theory and Techniques, vol. 41, No.5, pp.77..t-780, May

1993

112

References

[35] K.F.Schtinemann and K.Behm, "Nonlinear Noise Theory for Synchnized

Oscillators", IEEE Transactions on Microwave Theory and Techniques, vol. MTT-27.

No.5, pp.452-458, May 1979

[36] A.P.S.Khanna, " Direct Measurement of the Nonlinear M.Le. Oscillator

Characteristics Using Injection Locking Polar Diagram", 1983 IEEE MTT-S ITIt.

Microwave Symp. Dig., pp. 501-503, 1983

[37] G.H.Bertil Hansson and K.Ingemar Lundstrom, " Stability Criteria for Phase-Locked

Oscillators", IEEE Transactions on Microwave Theory and Techniques, vol. MIT-20,

No.10, pp.641-645, October 1972

[38] D.Sommer and N.J.Gomes, "Wide-Locking Bandwidth Optically Injection-Locked

Oscillators: S-Parameter Design and Modulation Effects", IEEE Transactions on

Microwave Theory and Techniques, vol. 43, No.7, pp.1424-1432, July 1995

[39] J.Troger, P.-A.Nicati, L.Th6venaz, and Ph.A.Robert, " Novel Measurement Scheme

for Injection-Locking Experiments", IEEE journal of Quantum Electronics, vol. 35,

No.1, pp.32-38, January 1999

[40] B.Razavi, "RF Microelectronics", Prentice Hall, Inc., pp. 314-315,1998

[41] L.R.Malling, "Phase-Stable Oscillators for Space Communications, Including the

Relationship Between the Phase Noise, the Spectrum, the Short-Term Stability, and

the Q of the Oscillator", PROCEEDINGS OF THE IRE, pp.1656-1664, July 1962

[42] J.Meluin, "Background Noise in Nonlinear Oscillators", PROCEEDINGS OF THE

IRE, pp.1960-1473, August 1960

[43] C.Schiebold, "Getting Back to the Basics of Oscillator Design", Microwave Journal,

pp.336-339, May 1998

[44] H.E.Ott, "Noise Reduction Techniques in Electronic Systems", Wiley-Interscience

Publication, 1976.

[45] J.G.Ondria, "A Microwave System for Measurements of AM and PM noise Spectra",

IEEE Transactions on Microwave Theory and Techniques, vol. MIT-16, No.9, pp.

767-781, September 1968.

113

References

[46] H.C.Chang, X.Cao, U.K.Mishra and R.A.York, "Phase Noise in Coupled Oscillators:

Theory and Experiments", IEEE Transactions on Microwave Theory and Techniques,

vol. 45, No.5, pp.604-615, May 1997

[47] D.Schere, "Today's Leeson-Learn About Low-Noise Design Microwave, Part I",

Microwave Journal, pp.l16-122, April 1979

[48] G.D.Vendelin, A.M.Pavio, and U.L.Rhode, "Mirowave Circuit Design Using Linear

and Nonlinear Techniques", pp.418-437, New York, Chichester:Wiley, 1990

[49] M.Underhill, "Fundamantals of oscillator performance", Electronics &

Communication Engineering Journal, pp.185-193, August 1992

[50] S.Mizushina, "Corporate and Tandem Structures for Combining Power from 3N and

2N+1 Oscillators", IEEE Transactions on Microwave Theory and Techniques, vol.

MIT-28, No. 12, pp. 1428-1432, December 1980.

[51] M.Madihian and S.Mizushina, "Combining the Powers from Multiple-Device

Oscillators", IEEE Transactions on Microwave Theory and Techniques, vol. MTI-30,

No.8, pp. 1228-1233, August 1982.

[52] K.Chang and C.Sun, "Millimeter-Wave Power-Combining Techniques", IEEE

Transactions on Microwave Theory and Techniques, vol. MTI-31, No.2, pp. 91-107,

February 1983.

[53] S.Kudszus, T.Berceli, A.Tessmann, M.Neumann and W.H.Haydl, "W-Band HEMT­

Oscillator MMICs Using Suharmonic Injection Locking", IEEE Transactions on

Microwave Theory and Techniques, vol. MIT 48, No. 12, pp. 2526-2532, December

2000

[54] R.Gilmore, " Concepts in the Design of Frequency Multipliers", Microwave Journal,

pp.129-139, March 1987

[55] C.Rauscher, "Frequency Doublers with GaAs PET's", 1982 IEEE MTJ-S Int.

Microwave Symp. Dig., pp. 280-282, 1982

[56] C.Rauscher, "High-Frequency Doubler Operation of GaAs Field-Effect Transistors",

1983 IEEE Transactions on Microwave Theory and Techniques, vol. MTI-31, pp.

462-472, June 1983

ll-l

References

[57] E.Camargo, "Design of FET Frequency Multipliers & Harmonic Oscillators".

Arttech House, 1998

[58] E.Camargo, " Sources of Non-Linearity in GaAs MESFET Frequency Multipliers",

1982 IEEE MTT-S Int. Microwave Symp. Dig., pp. 343-345, 1983

[59] A.Gopinath and J.B.Rankin, "Single-Gate MESFET Frequency Doublers", IEEE

Transactions on Microwave Theory and Techniques, vol. MTI-30, No.6, pp. 869-875.

June 1982

[60] M.Borg and G.R.Branner, " Novel MIC Bipolar Frequency Doublers Having High

Gain, Wide Bandwidth and Good Spectral Performance", IEEE Transactions on

Microwave Theory and Techniques, vol. 39, No.12, pp. 1936-1946, December 1991

[61] N.Siripon, M.Chongcheawchamnan and LD.Robertson, "Injection-Locked Balanced

oscillator", Electronics Letters, Vol. 36, No. 22, pp. 1854-1855, 26th October 2000

[62] J.L Jimenez, 1. Gismero, M.B. Notario, and F. Lopez, "A High Perfromance PLDRO

for a 28 GHz LMDS System," in 2000 European Microwave Conference, Paris,

September, 2000

[63] Y. Campos-Roca, L. Verweyen, etal, "Coplanar pHEMT MMIC Frequency

Multipliers for 76 GHz Automotive Radar," IEEE Microwave and Guided Wave

Letters, Vol. 9, No.6, June 1999

[64] T.Hirota, and H.Ogawa, "Uniplanar Monilithic Frequency Doublers", IEEE

Transactions on Microwave Theory and Techniques, Vol.37, No.8, August 1989.

[65] D.G.Thomas, Jr. and G.R.Branner, "Optimization of Active Microwave Frequency

Multiplier Performance Utilizing Harmonic Terminating Impedance", IEEE

Transactions on Microwave Theory and Techniques, vol. 44, No.12, pp. 2617-2624,

December 1996

[66] R.Stancliff, " Balanced Dual Gate GaAs FET Frequency Doublers", 1981 IEEE,

pp.143-145,1981.

[67] Y.Iyama, A.Iida, T.Takagi, and S.Urasaki, " Second-Harmonic Reflector Type High­

Gain Frequency Doubler Operating in K-Band", 1989 IEEE MTT-S Int. Microwave

Symp. Dig., pp. 1291-1294, 1989

115

References

[68] D.G.Thomas, Jr. and G.R.Branner, " Single-ended HEMT Multiplier Design Using

Reflector Networks", IEEE Transactions on Microwave Theory and Techniques, vol.

49, No.5 pp. 990-993, May 2001

[69] H.A.Willing, C.Rauscher and P.D.Santis, "A Technique for Predicting Large-Signal

Performance of a GaAs MESFET", IEEE Transactions on Microwave Theory and

Techniques, vol. MTT-26, No.12, pp. 1017-1023, December 1978

[70] Y.Campos-Roca, L.Verweyen, M.Neumanm, M.Femandez, M.C.Curras-Francos ,

E.Sanchez-Sanchez, A.Htilsmann, and M.Schlechtweg, "Coplanr pHEMT MMIC

Frequency Multiplier for 76-GHz Automotive Radar", IEEE Transactions on

Microwave Theory and Techniques, vol. 9,No. 6, pp. 242-244, June 1999

[71] I.Angelov, H.zirath, N.Rorsman, and H.Gronqvist, "A Balanced millimeter wave

doubler based on pseudomorphic HEMTs", 1992 IEEE MTT-S Int. Microwave Symp.

Dig., pp. 353-356, 1992

[72] lGrajal, V.Krozer, E.Gonalez, F.Maladonado, and lGismero, "Modeling and

Design Aspects of Millimeter-Wave and Submillimeter-Wave Schottky Diode

Varactor Frequency Multipliers", IEEE Transactions on Microwave Theory and

Techniques, vol. 48, pp. 700-711, April 2000

[73] Y.K.G.Hau, V.Postoyalko, and J.R.Richardson, "Effect of impedance mismatch and

imperfect isolation on distortion cancellation in feedforward amplifiers", lEE Proc.­

Circuits Devices Syst., vol. 145, pp. 35-39, February 1998

[74] Y.K.G.Hau, V.Postoyalko, and lR.Richardson, "Compensation of amplifier

nonlinear phase response to improve wideband distortion cancellation of feedforward

amplifiers", Electronics Letters., vol. 33, pp. 500-502, March 1997

[75] S.Kang, I.Lee and K.S.Yoo, " Analysis and Design of the Feedforward Power

Amplifier", 1997 IEEE MTT-S Int. Microwave Symp. Dig., pp. 1519-1522, 1997.

[76] C.McNeilage, E.N.Ivanov, P.R. Stockwell, and J.R.Searls, "Review of Feedback and

Feedforward Noise Reduction Techniques", 1998 IEEE International Frequency

Control Symp., pp.146-153, 1998

[77] N.Pothecary, "Feedforward Linear Power Amplifier", Artech House, Inc .. 1999

[78] J.Hogerheiden, M.Cimihera, and GJue, "Improved Planar Spiral Transformer

Theory Applied to a Muniature Lumped Element Quadrature Hybrid", IEEE

116

References

Transactions on Microwave Theory and Techniques, vol. 45,No. 4, pp. 462-472,

April, 1997

[79] N.Siripon, K.S.Ang, M.Chongcheawchamnan and I.D.Robertson, "Design and

Performance of a novel balanced self-oscillating mixer", 3rJh European Microwave

Con!, Paris, France, 2-6 October 2000

[80] N.Siripon, M.Chongcheawchamnan and I.D.Robertson, "Novel Sub-hannonic

Injection-Locked Balanced Oscillator", 3th European Microwave Conj, London,

UK, 24-28 September 2001

[81] N.Siripon, M.Chongcheawchamnan and I.D.Robertson, "Novel Frequency Doubler

using Feedforward for Fundamnetla Frequency Component Suppression", 2001 IEEE

Microwave MTT-S Int. Symp. Dig., Phoenix, Arozona, 20-25 May 2001

[82] M.J.Underhill, N.Siripon and N.Christorou, "Determination of Capture Range and

Locking Range of Injection Locked Oscillators", 15th European Frequency and Time

Forum, Switzerland, March 2001

[83] M.J.Underhill and N.Siripon, "AN INVESTIGATION OF A CASCADED PAIR OF

INJECTION-LOCKED OSCILLATORS AT 2GHz", 16th European Frequency and

Time Forum, Switzerland, March 2002

[84] N.Siripon, M.Chongcheawchamnan and I.D.Robertson, "Injection-Locked Balanced

oscillator-Doubler", lEE Electronic Letters, Vol. 37, No. 15, pp. 958-959, 19th

July

2001.

[85] A.V.Derziel, "Noise", Prentice-Hall Electronical Engineering Series, pp. 1-7,

Prentice-Hall, Inc., 1956

117