Evaluation of broadband MMIC oscillators with external tank401135/FULLTEXT01.pdfThe cross-coupled topology was chosen to be used in the broadband VCO core due to its broadband performance

Evaluation of broadband MMIC

oscillators with external tank

MSc in Microelectronics

A n e t t e B r a n d t

Master of Science Thesis

Stockholm, Sweden 2010

TRITA-ICT-EX-2010:159

EVALUATION OF BROADBAND MMIC

OSCILLATORS WITH EXTERNAL TANK

KTH Royal Institute of Technology, Stockholm

Sivers IMA AB

Chalmers University of Technology, Gothenburg

EVALUATION OF BROADBAND MMIC

OSCILLATORS WITH EXTERNAL TANK

Anette Brandt

Civilingenjörsutbildning i mikroelektronik

KTH Royal Institute of Technology

Supervisors:

Dr Sten Gunnarsson, Sivers IMA AB

Christer Stoij, Sivers IMA AB

Dr Dan Kuylenstierna, Chalmers University of Technology

Examiner:

Ass. Prof. Urban Westergren,

Department of Microelectronics and Applied Physics, MAP

KTH Royal Institute of Technology

i

ABSTRACT

This thesis evaluates broadband Voltage Controlled Oscillators, VCO´s, using different VCO

topologies and external tanks. The work has been done at Sivers IMA AB within the

“INTOSC” project in the framework of the VINNEX Centre “GHZ CENTRE”.

Different VCO topologies are investigated and their performance is compared and evaluated.

The active part of the VCO, the core, is implemented in a GaAs HBT MMIC technology. The

main emphasis of this work has been on the design of the frequency selective off-chip

resonance circuit, the “tank”, which consists of varactor diodes and bond wires as capacitors

and inductors, respectively. Using only two unique MMIC VCO cores, one fundamental and

one harmonic MMIC, a family of VCOs covering a frequency range of 2-27 GHz is designed,

(in bands) . Most of these VCO`s demonstrates more than octave (2:1) bandwidth. Besides the

design and assembly work, analysis of how the tolerances of the critical components in the

tank affects the performance of the VCOs are also done within this work.

SAMMANFATTNING

Arbetet utvärderar bredbandiga “Voltage Controlled Oscillators”, VCOer, med olika VCO

topologier och externa tankar. Arbetet är utfört på Sivers IMA AB inom “INTOSC” projektet

i ramen för VINNEX centret “GHZ CENTRE”.

Olika VCO topologier undersöks och deras kapacitet jämförs och utvärderas. Den aktiva

delen av VCO kärnan är implementerad i GaAs HBT MMIC teknologi. Tyngdpunkten för

arbetet är design av frekvensselektiva resonanskretsar som består av varaktordioder och

bondtrådar vilka utgör kapacitanser respektive induktanser. Med två unika MMIC VCO

kärnor, en fundamental samt en harmonisk MMIC, designas en VCO familj som täcker ett

frekvensområde på 2-27 GHz, (i frekvensband). Flertalet av dessa VCOer visar mer än en

oktav i bandbredd (2:1). Förutom design och monterings arbete analyseras toleransen för hur

de kritiska komponenterna i tanken påverkar kapaciteten för VCOerna.

ii

ABBREVIATIONS

𝑙 Alumina

Tantalum Nitride

AuSn Gold tin

BW Bandwidth

CAD program Computer-aided design program

FSUP Signal Source Analyzer, Rhode & Schwarz

GaAs Gallium Arsenide

HBT Heterojunction Bipolar Transistor

InGaP Indium Gallium Phosphide

InP/GaAs Indium Phosphide Gallium Arsenide

LP filter, LPF Low Pass filter

MMIC Monolithic Microwave Integrated Circuit

RF Radio Frequency

SMA SubMiniature version A (Contact)

SMD Surface mounted device

U300 Liquid under fill encapsulant

VCO Voltage Controlled Oscillator

iii

NOTATIONS

Emitter capacitance

Parallel tank capacitance

Varactor diode capacitance

Parallel tank inductance

Parasitic inductance in the tank

Quality factor for a parallel resonance circuit

Quality factor for a series resonance circuit

Quality factor for the tank

Negative resistance

Parallel tank resistance

Parasitic resistance in the tank

Base bias

Collector bias

Characteristic impedance for the tank

Impedance for the resonance circuit

𝑓 Resonant frequency

𝑔 Tranceconductance

Phase noise

Empirical fitting parameter

Complex transfer function

𝑓 Transfer function

Current

Inductance for a bond wire

Temperature

iv

Voltage

𝑑 Diameter of a bond wire

𝑘 Boltzmann´s constant

𝑙 Length of the bond wire

Feedback element

Angular frequency

v

TABLE OF CONTENTS

Abstract ...................................................................................................................................... i

Sammanfattning ........................................................................................................................ i

Abbreviations ............................................................................................................................ ii

Notations .................................................................................................................................. iii

Table of contents ....................................................................................................................... v

1 Introduction ...................................................................................................................... 1

1.1 Background and motivation ......................................................................................... 1

2 Oscillator theory ............................................................................................................... 2

2.1 General considerations for an oscillator ...................................................................... 2

2.2 The cross-coupled oscillator topology ......................................................................... 3

2.3 Cross-coupled differential oscillator: Fundamental and harmonic output .................. 5

3 Resonance circuit, tank, theory ....................................................................................... 7

4 Measurement setup, implementation ........................................................................... 11

5 Fundamental MMIC VCO ............................................................................................ 15

5.1 Different topologies for the capitols fundamental VCOs .......................................... 15

5.2 Investigation of chip-capacitor value in collector bias lpf ........................................ 19

5.3 Comparison of bonded vs. flipped varactor diodes ................................................... 22

5.4 Investigate the effect of a feedback emitter capacitance ........................................... 26

5.5 Fundamental 2-4 GHz oscillator with a varactor diode of 2.7pF@4V ...................... 29

5.6 Height analysis of the inductor bond wires in the tank ............................................. 34

5.6.1 Analysis based on measured capacitance of the varactors ................................. 34

5.6.2 Analysis based on calculated capacitance of the tank ........................................ 36

5.7 Analysis of tank capacitance ..................................................................................... 38

5.7.1 Analysis based on measured capacitance of the tank ......................................... 38

5.7.2 Analysis based on calculated capacitance of the tank ........................................ 39

5.8 Performance versus temperature ............................................................................... 40

6 Harmonic 4-8 GHz and 8-16 GHz oscillator with a varactor diode of 2.7 pF@4V.. 43

6.1 Suppression ................................................................................................................ 44

6.2 Performance as function of temperature for a harmonic 4-8 GHz Oscillator............ 45

6.3 Performance as function of temperature for a harmonic 8-16 GHz oscillator .......... 47

7 Summary and conclusions ............................................................................................. 49

8 Future work .................................................................................................................... 51

9 Acknowledgment ............................................................................................................ 52

vi

10 References ....................................................................................................................... 53

Appendix A ............................................................................................................................. 54

Alumina design .................................................................................................................... 54

Appendix B .............................................................................................................................. 59

Measurement details ............................................................................................................. 59

Chapter 1. Introduction

1

1 INTRODUCTION

1.1 BACKGROUND AND MOTIVATION

This thesis deals with the evaluation of generic broadband VCO cores implemented in a GaAs

HBT MMIC technology. In order to facilitate a generic design, suitable for many different

frequency bands, the tank circuitry was not integrated on the MMIC itself but is instead

implemented with external components, connected to the MMIC by bond wires.

Broad band VCO´s is one of the main commercial products that Sivers IMA manufactures

and the field of applications is wide.

Today, Sivers IMA has a family of VCOs covering 2-25 GHz. The VCOs are all fundamental

frequency designs and have built-in regulators, buffer amplifiers, and output filters.

These VCOs are based on thin film technology with discrete components such as varactor

diodes, transistors, capacitors and MMIC integrated amplifiers. Due to the technology and

manual manufacturing necessary, the costs are high. In order to lower the cost, a new

generation of VCO´s based on a broadband generic MMICs is evaluated in this work.

The MMICs evaluated in this thesis are produced in an InGaP/GaAs HBT process at WIN

semiconductor (www.winfoundry.com).

The thesis work consists of several parts. First, different MMIC designs with somewhat

different topology are investigated and the overall best performing MMIC is identified. The

MMICs are designed by Sten Gunnarsson at Sivers IMA. Secondly, the surrounding circuitry

such as DC filters needs to designed and evaluated. Finally, different tank combinations are

designed and the performance versus frequency and temperature is characterized.

In order to make these investigations possible, evaluation boards on thin film substrate

( 𝑙 i.e. alumina) are designed.

Chapter 2. Oscillator Theory

2

2 OSCILLATOR THEORY

2.1 GENERAL CONSIDERATIONS FOR AN OSCILLATOR

An oscillator generates a periodic output and the circuit must entail self-sustaining

mechanisms that allow its own noise to grow and become a periodic signal.

Most RF oscillators can be viewed as feedback circuits with following transfer function.

Figure 2.1 shows a two-port model though the feedback loop closed around a two-port

network. [1]

Figure 2.1 A feedback oscillator system.

where and are complex transfer function and feedback element respectively.

A self-sustaining mechanism arises at the frequency if and the oscillation

amplitude remains constant if is purely imaginary, i.e. 𝑗 .

For steady oscillation two conditions must occur simultaneously at . [2]

+


3

Barkhausen´s criteria.

i. The loop gain must be equal to, or larger than, unity.

𝑗

ii. The total phase shift around the loop must be equal to zero.

𝑗

A resonator circuit or a frequency-selective network is included in this loop in order to

stabilize the frequency. This frequency-selective network is commonly referred to as the

resonant tank when discussing VCOs.

2.2 THE CROSS-COUPLED OSCILLATOR TOPOLOGY

The cross-coupled topology was chosen to be used in the broadband VCO core due to its

broadband performance and its “willingness” to start to oscillate compared to other common

topologies such as the Colpitt-type oscillator [3].

The active VCO core consists of a cross-coupled HBT transistor pair according to Figure 2.2.

In order to compensate for the loss in the external resonance network the active part of the

balanced cross-coupled oscillator provides a negative resistance .

In order to achieve a broadband generic VCO core, the VCO core should provide a negative

resistance over a broad frequency range and this negative resistance should be larger than the

resistance in the tank over the full bandwidth.

The power in the tank resonator itself will decay with time since the energy that oscillates

between the capacitor and inductor is lost in the form of heat in the unwanted parasitic

resistor.

In order to derive the negative resistance of a cross-coupled VCO topology, the transistors are

replaced with a simplest form of a small signal model, the ideal voltage-driven current source

[2]. This gives a good understanding of the fundamental operation of the cross-coupled

topology and proves that a negative resistance is obtained.


4

Figure 2.2 Balanced cross coupled topology and an equivalent small signal circuit.

𝑔 and 𝑔 are the transconductances for the equivalent small signal circuit.

𝑔 𝑔

𝑔

𝑔

𝑔

𝑔

𝑔

𝑔

and if 𝑔 𝑔 𝑔 , then

𝑔

Thus, the active part of the balanced cross-coupled oscillator provides a negative resistance.

1 2 1 2

𝑝

2 1 𝑐𝑒2 𝑐𝑒1

+ + 𝑔𝑚1 𝑐𝑒2 𝑔𝑚2 𝑐𝑒1

𝑐 𝑏

𝑒

𝑏 𝑐

𝑒


5

2.3 CROSS-COUPLED DIFFERENTIAL OSCILLATOR: FUNDAMENTAL AND

HARMONIC OUTPUT

Figure 2.3 Schematic of fundamental and harmonic MMIC VCO cores respectively.

In this work, a harmonic oscillator is defined as an oscillator where the fundamental RF

outputs are combined and destructive interference of the fundamental and odd tones will

occur due to the balanced topology. This suppression will be optimal if the resonance circuit

is well balanced. The even harmonics will instead be added in phase, thus, constructive

interference will occur, Figure 2.4. [1]

0 0

𝑏𝑏 𝑏𝑏

𝑐𝑐 + 𝑛𝑘 𝑐𝑐 + 𝑛𝑘 𝑐𝑐 + 𝑛𝑘 𝑐𝑐 + 𝑛𝑘

1𝑜𝑢𝑡 2𝑜𝑢𝑡

𝑜𝑢𝑡


6

Figure 2.4 Output signal for fundamental and harmonic MMIC VCO oscillators.

𝑡

𝑡

𝜃 = 𝜋

Fundamental output

Fundamental output

Harmonic output, X2

Harmonic output, X2

Harmonic output, X2

𝑡

2𝑜𝑢𝑡 𝑡

𝑜𝑢𝑡 𝑡

1𝑜𝑢𝑡 𝑡

Chapter 3. Resonance Circuit, Tank, Theory

7

3 RESONANCE CIRCUIT, TANK, THEORY

The simplest form of a resonance circuit, a tank suitable for a VCO application, sets up a

frequency selective network by using a capacitor and an inductor in parallel.

For biasing purposes in the used MMIC VCOs, it is convenient to split the resonance circuit

into two, identical parts. The inductor will then be made up of two bond wires, and , and

the capacitor is represented by two reverse-biased diodes, varactor diodes, and . In this

work, the tank capacitances, varactors, tune over a range of 0-20 V.

The tank configuration can be seen as a parallel resonance circuit.

Figure 3.1 Ideal resonance circuit and equivalent circuit.

The tank capacitance and inductance is given by

where is the varactor capacitor.

And the tank inductance is given by

2 1 1 2 1 2 1 2

1 2

𝑝

𝑝 ⟹


8

As mentioned in section 2.2, the tank demonstrates a decaying oscillatory behavior since the

energy that oscillates between the inductor and capacitor is lost in the unwanted parasitic

resistor found in all actual components. In order to model this imperfection, an additional

resistor is introduced in the model of the tank, Figure 3.2.

Figure 3.2 Resonance circuit with additional resistor to model the inherent loss in the circuit.

The impedance for the resonance circuit is given by

𝑗 𝑗

The resonant frequency 𝑓 is defined to be the frequency at which the impedance is purely

resistive, i.e. the total reactance for the circuit is zero. The impedance of the inductance

should equal the impedance of the capacitance in magnitude i.e. [4]

𝑓

𝜋

The negative resistance in the active part of the balanced cross coupled oscillator must be

larger than or equal to in order to sustain oscillation. The and of the tank provide the

resonance frequency of interest 𝑓 .

𝑝

1 2

𝑝

𝑝


9

Thus

Bonding wires that connect the external tank to the MMIC core circuit contribute as parasitic

inductance, illustrated in Figure 3.3 below as with an additional .

Figure 3.3 A parallel resonance circuit with parasitic inductance.

The quality factor for a parallel resonance circuit is defined as

𝑒𝑛𝑒 𝑔 𝑙𝑜 𝑝𝑒 𝑒𝑐𝑜𝑛𝑑 𝑖𝑛 𝑡𝑒𝑚

For a parallel resonance circuit without parasitic inductance the quality factor is

And for the series parasitic inductances the quality factor is

1 21

2 11 2

𝑝 𝑝 𝑝

𝑆

𝑆 𝑆

𝑆

𝑡 𝑛𝑘


10

These series parasitic inductances and will also affect the resonance circuit together

with parasitic conductance in the MMIC VCO core.

Another parameter of importance is the characteristic impedance of the tank. The

characteristic impedance should match the negative resistance of the VCO core.

The characteristic impedance for a tank is defined as [5]

In this work, the inductance in the tank is implemented with bond wires; the inductance of a

bond wire in free space is given as

𝑙 𝑙𝑛

𝑛

where the length of the bond wire 𝑙 and the diameter 𝑑 of the bond wire are given in cm. [6]

Leeson´s equation is an empirical model for phase noise in active devices, based on

experimental data. [7]

𝑘

The Leeson equation identifies the most significant causes of phase noise in oscillators.

Therefore it is possible to highlight the main causes in order to be able to minimize them.

Two conclusions can immediately be drawn by examining Leeson´s equation:

The Q-value in the tank should be as large as possible.

The power in the tank ( ) should be as large as possible, thus, the current and/or

the voltage swing should be maximized.

Chapter 4. Measurement Setup, Implementation

11

4 MEASUREMENT SETUP, IMPLEMENTATION

The thin film (alumina, 𝑙 ) carrier used for the evaluation is manufactured with gold

plated via holes and resistance (45 /square) consisting of tantalum nitride ( ). The

conductor strip lines are made of plated gold.

Figure 4.1 Alumina, 13.3 X 16.6 𝑚𝑚 .

The MMIC VCO core and resonance circuit are soldered directly on the alumina carrier with

an appropriate solder paste (AuSn 80/20).

RF outputRF outputOptional LC-filters

Connection pads for bias

Via hole

3 X 250 Ohm

Tank

MMIC


12

Figure 4.2 MMIC and resonance circuit on the alumina.

The reactive part of the low pass (LP) filter used at tune and collector bias constitute of

standard surface mounted devices (SMDs), size 0402 and 0603, and were determined

experimentally. Series resistors are used to further suppress spurioses on the bias lines. These

components are glued with a two component silver epoxy. The LP- filters is shown in Figure

4.3 and its simulated performance is found in Figure 4.4 and Figure 4.5.

Figure 4.3 Low pass filter at collector and tune bias respectively.

Varactor diodes assembled adjacent to MMIC

Vbb without LP-filter on-chip

Vbb with LP-filter on-chip

Series resistor to suppress spurioses

C1

68pF

R1

750

C2

1pF

R1

220

L1

47nH

1 2

C1

68pF

0

C4

33pF

0

C3

10nF

C3

10nF

L1

47nH

1 2

𝑜𝑢𝑡 𝑜𝑢𝑡


13

The transfer function is given by

𝑓

And the value for the cut off frequency 𝑓 is given by

𝑓 𝑑

Figure 4.4 Bode diagram for LP-filter at collector bias.

Figure 4.5 Bode diagram for LP-filter at tune bias.

Frequency

10KHz 30KHz 100KHz 300KHz 1.0MHz 3.0MHz 10MHz

DB(V(OUT)/V(IN))

-30

-20

-10

-0M

a

g

n

i

t

u

d

e

d

B

Frequency

10KHz 30KHz 100KHz 300KHz 1.0MHz 3.0MHz 10MHz

DB(V(OUT)/V(IN))

-30

-20

-10

0M

a

g

n

i

t

u

d

e

d

B

𝑓 kHz @ -3 dB

𝑓 kHz @ -3 dB


14

The LP-filter at the collector and tune bias have cut off frequencies 𝑓 kHz and

𝑓 kHz, respectively.

The alumina substrate is placed in a package with SMA connectors where bias and RF output

connects with 75 μm gold wire and 250 μm gold ribbon, respectively.

Figure 4.6 Package with SMA connectors, 275x375 𝑚𝑚 , with bias connections and with

alumina assembled respectively.

Power supply connects directly to the package with SMA connectors in which the substrate is

mounted. RF output connects directly on the FSUP signal source analyzer and measurements

to consider for this thesis is output power, frequency, spurious, suppressions (in the case for

harmonic oscillator) and phase noise. For the harmonic oscillators, the output power was very

low and an additional amplifier was necessary for the phase noise measurement. A description

of the operation of the FSUP is found in appendix B.

Figure 4.7 Schematic setup of the measurement.

Vtune

Vbb

Vcc

Vbb

Vtune

RF out

Power supply

Package with SMA

connectors

OptionalPower amplifier

FSUP Signal Source

Analyzer

Chapter 5. Fundamental MMIC VCO

15

5 FUNDAMENTAL MMIC VCO

5.1 DIFFERENT TOPOLOGIES FOR THE CAPITOLS FUNDAMENTAL VCOS

Eight different MMIC VCO cores (four fundamental and four harmonic VCOs) have been

evaluated and considerations of their performance and behavior will be discussed. The

fundamental and harmonic VCO cores uses the same circuit variations and conclusion about

the best version for both the fundamental and harmonic oscillators can be drawn from an

investigation of the former. The four different fundamental MMIC VCO cores are:

i. MMIC VCO core with the varactor diodes flipped on top of the MMIC device. The

main advantage with this version is the smaller parasitic inductance (Figure 3.3). The

main drawback is the need of flipped varactors which is troublesome to use in volume

production.

Figure 5.1 Drawing, MMIC design and schematic for (i).

MMIC VCO core Varactor diodes

Fundamental

RF output

𝑐𝑐 𝑡𝑢𝑛𝑒

33 pF

0

𝑐𝑐 + 𝑛𝑘 𝑐𝑐 + 𝑛𝑘


𝑏𝑏


16

ii. MMIC VCO core with the varactor diodes flipped on top of the MMIC device and a

feedback emitter capacitance in the MMIC VCO core. The main pros and cons with

this MMIC is the same as for version (i). However, unlike the MMIC in version (i),

additional capacitors to ground were introduced at the emitters of the HBTs. In

simulations, these extra capacitors improved the gain and phase margins of the VCO.

Although the simulations didn’t show any failures in this respect, the start up to

oscillate could still be an issue in higher temperatures (+85 deg.) The HBT model used

in the simulations were not verified at these high temperatures and a VCO version

with additional emitter capacitors was therefore included as a precaution.

Figure 5.2 Drawing, MMIC design and schematic for (ii).


Fundamental

RF output

𝑐𝑐 𝑡𝑢𝑛𝑒

33 pF

00 0


𝑐𝑐 𝑐𝑐

𝑏𝑏


17

iii. MMIC VCO core with the varactor diodes assembled next to the MMIC and

connected with bond wires. The main advantage of this version is the use of bonded

varactors next to the MMIC which is easy to use in production. The main drawback is

the introduction of parasitic inductance (Figure 3.3) through the bond wires that

connects the MMIC to the varactors.

Figure 5.3 Drawing, MMIC design and schematic for (iii).

𝑡𝑢𝑛𝑒


𝑐𝑐

Fundamental

RF output 33 pF

0



𝑏𝑏


18

iv. MMIC VCO core with the varactor diodes assembled next to the MMIC and

connected with bond wires as well as a feedback emitter capacitance in the MMIC

VCO core.

Figure 5.4 Drawing, MMIC design and schematic for (iv).

Section 5.2, 5.3 and 5.4 will compare different MMIC and tank topologies and eliminate those

of no interest for further evaluation. Section 5.5, 5.6 and 5.7 discusses the chosen fundamental

MMIC VCO core with associated resonance circuit.

𝑡𝑢𝑛𝑒


𝑐𝑐

Fundamental

RF output 33 pF

00 0


𝑐𝑐 𝑐𝑐

𝑏𝑏


19

5.2 INVESTIGATION OF CHIP-CAPACITOR VALUE IN COLLECTOR BIAS LPF

A chip capacitor of 33pF is used in the LPF for the collector bias. This capacitor also serves

as the common node for the two bond wires that serves as inductors in the tank. In theory with

a perfectly balanced tank, the common node of the two bond wires is a virtual ground for the

frequency of oscillation and the chip capacitor will therefore not affect the performance of the

VCO. However, in an actual, somewhat unsymmetrical tank, the chip capacitor will influence

performance of the VCO.

The default chip capacitor of 33 pF was replaced with 10 and 100 pF capacitors in order to

investigate how the value of the chip capacitor affects the performance of the VCO.

Figure 5.5 MMIC, varactor diodes and the chip capacitor that serves as common node for the

bond wires in the tank.

Frequency, phase noise, and output power versus tuning voltage was measured. During all

measurements in this thesis, the tuning voltage is applied RELATIVE to the collector voltage.

What is denoted as 0 V tuning voltage is actually the absolute value of collector voltage of the

VCO, typically 5-6 V. This somewhat impractical usage of the tuning voltage origins from the

topology where the collector of the HBTs are biased through the tank, thus the tuning voltage

applied to the varactors will bias the varactor relative to the collector voltage.

In this case 75 𝑚 ribbon bands are used in the resonance circuit as inductors.

MMIC VCO core Varactor diodes 33 pF


20

Figure 5.6 Frequency versus tune voltage for different capacitors that serves as common node

for the bond wires in the tank.

𝑓 [GHz]

𝑓

[GHz)

[GHz] 𝑓

[%]

10 pF 6.11 10.1 3.99 49.2

33 pF 5.57 9.72 4.15 54.3

100 pF 5.54 9.63 4.09 53.9

Table 5.1 Measured frequencies for different capacitors.

The variation in bandwidth is very small between 100 pF and 33 pF whereas it is somewhat

larger between 33 pF and 10 pF. The highest bandwidth is obtained with 33 pF. The

displacement of the frequency range for the 10 pF case depends on the length of the bond

wires in the resonance circuit which happened to be somewhat shorter compared to when the

other chip capacitors were used.

The effect of the chip capacitor on the frequency range of the VCO is thus considered to be

small.

Also output power and phase noise was measured for the VCOs using three different chip

capacitors, Figure 5.7 – 5.8.

5

6

7

8

9

10

11

0 5 10 15 20

Fre

qu

en

cy [

GH

z]

Tune [V]

Frequency vs. tune

10 pF

33 pF

100 pF


21

Figure 5.7 Power versus tune voltage for different reconciliation capacitors.

Figure 5.8 Phase noise versus tune voltage for different reconciliation capacitors.

In conclusion, no significant change in the performance of the VCO is found when the 33 pF

chip capacitor was replaced with other values. A 33 pF chip capacitor will be used for the

remaining work.

-30

-25

-20

-15

-10

-5

0

0 5 10 15 20 25

Po

we

r [d

Bm

]

Tune [V]

Power vs. tune

10 pF

33 pF

100 pF

-90

-85

-80

-75

-70

-65

-60

0 5 10 15 20

Ph

ase

No

ise

@1

00

kHz

[d

Bc/

Hz]

Tune [V]

PN @100kHz vs. tune

10 pF

33 pF

100 pF


22

5.3 COMPARISON OF BONDED VS. FLIPPED VARACTOR DIODES

The sought tank-resonance in the chosen oscillator topology is parallel. Reactive elements

such as in Figure 3.3 in series with the tank may introduce unwanted series resonances.

This is a potential problem for all VCOs using long connecting lines between the VCO core

and the tank. These series inductors may create unwanted series resonances in version (iii)

and (iv) of the VCOs in this thesis, section 5.1. In these versions, the varactors are mounted

next to the MMIC and connected with bond wires. One option to avoid this is to flip the

varactor diodes and mount the varactor mesa directly on the MMIC VCO core with solder

paste as in version (i) and (ii).

For the cases when the varactor diodes are flipped on the MMIC device: the varactors are

soldered on a gold plated Kovar carrier with solder paste for easier handling, and the varactors

are then mounted upside down with the mesa of the varactors soldered directly on the MMIC

VCO core. Varactor cathode (and voltage tune) connects the Kovar carrier with ribbon band

directly to the tuning voltage circuitry on the alumina. An appropriate under-fill (U300) was

used on the flipped varactor diodes for enhanced mechanical stability.

This comparison is to compare the performance of the two cases when the varactor are either

flipped upside down on the MMIC (Figure 5.9) or mounted with more traditional methods,

i.e. with bond wires to the varactor soldered next to the MMIC on the alumina substrate

shown in Figure 5.10.

Figure 5.9 Varactor diodes flipped on the MMIC.


23

Figure 5.10 Varactor diodes assembled beside MMIC VCO core.

During this investigation, both VCOs employs identical varactor diodes (1.2 pF@4V) and the

same bias applied. Again, frequency, phase noise, and output power was measured versus

tuning voltage.

Figure 5.11 Frequency versus tune voltage.

𝑓 [GHz]

𝑓

[GHz)

[GHz] 𝑓

[%]

Flipped diodes on MMIC

3.76 7.88 4.12 70.8

Diodes adjacent to MMIC

4.75 8.88 4.13 60.6

Table 5.2 Measured frequencies for flipped and bonded varactors in the tank.

3

4

5

6

7

8

9

10

0 5 10 15 20

Fre

qu

en

cy [

GH

z]

Tune [V]

Frequency vs. tune

Varactor flipped on MMIC

Varactor connected with bondwire


24

-30

-25

-20

-15

-10

-5

0

5

0 5 10 15 20

Po

we

r [d

Bm

]

Tune [V]

Power vs. tune



-110

-100

-90

-80

-70

-60

0 5 10 15 20

Ph

ase

No

ise

@1

00

kHz

[d

Bc/

Hz]

Tune [V]

PN @100kHz vs. tune



A noticeable difference in bandwidth of the VCOs is observed. Again, the frequency range of

the two VCOs has a displacement of 1 GHz due to the assembly method. From this

comparison, it is clear that the version with diodes flipped directly on the MMIC has the

possibility to obtain broader bandwidths.

Figure 5.12 Power output versus tune voltage.

Figure 5.13 Phase Noise at 100 kHz offset.

The phase noise shows an enhanced result at higher frequencies in favor of the device with

flipped varactor diodes.

The measurements in the 0-3 V tuning range have an uncertainty due to the low output power

and the bias voltage is not optimized for best performance.


25

One also needs to consider the methods of assembly before the choice of tank configuration

may be taken. From assembly perspective, the tank configuration with diodes assembled next

to the MMIC VCO core is better and more stable. Flipping the varactors introduces an

uncertainty regarding how to place the solder paste on the varactor diodes in volume

production without affecting the mesa. This is especially crucial for the smaller varactor since

the size of the mesa scales with varactor size.

In conclusion, the arrangement where the diodes are assembled next to the MMIC has the

most stable performance in volume production without degrading performance too much and

will therefore be used.


26

5.4 INVESTIGATE THE EFFECT OF A FEEDBACK EMITTER CAPACITANCE

Figure 5.14 Schematic MMIC VCO core with and without feedback emitter capacitance

respectively.

During this test, both VCOs use identical varactor diodes (1.2 pF@4V) and the same bias

applied.

The length of each bond wire is approximately 450 (without emitter capacitance) and 430

(with emitter capacitance).

00 0 0


𝑏𝑏



𝑏𝑏



27

4

5

6

7

8

9

10

0 5 10 15 20

Fre

qu

en

cy [

GH

z]

Tune [V]

Frequency vs. tune

With emitter capacitance

Without emitter capacitance

Figure 5.15 Frequency range.

The diameter of the bond wire is 25.4 𝑚 and the inductance for each bond wire calculated

from equation (5.5) and evaluated @4V tune since the varactor capacitance is defined at this

value.

𝑓 [GHz]

𝑓

[GHz)

[GHz] 𝑓

[%]

With 4.77 8.35 3.58 42.9

Without

4.75 8.88 4.13 46.5

Table 5.3 Measured and calculated values for the setup of comparison.

From Table 5.3, it is clear that the MMIC without feedback emitter capacitance has a

somewhat larger bandwidth.

Figure 5.16 Output power versus tune voltage.

-30

-25

-20

-15

-10

-5

0

5

0 5 10 15 20

Po

we

r [d

Bm

]

Tune [V]

Power vs. tune




28

Output power differ about 5-6 dB at higher frequencies between the two devices but there is a

difference present for all tuning voltages.

Figure 5.17 Phase Noise versus tune at 100 kHz offset.

Phase noise performance with and without emitter capacitance are very similar. The

discrepancy between the two for the lowest tuning voltages is probably related to

measurement inaccuracy due to the low output power of the VCO at these points.

During the design of the MMICs, it was noted that adding emitter capacitances, shown to the

left in Figure 5.14, could improve the start up of the VCO core to start oscillate. MMIC VCO

cores with these additional emitter capacitances were also manufactured in case the

“ordinary” MMIC VCO cores would fail to oscillate; this was especially a concern in high

temperature. However, the additional emitter capacitors decreased the tuning range somewhat

in simulations.

The conclusion from the simulations of the MMIC VCO cores has been shown to be true also

in actual measurements. The MMIC VCO cores without the additional emitter capacitance

have somewhat better performance. Unless this oscillator will fail to oscillate under some

other test conditions, the MMIC VCO core without the additional emitter capacitance will be

used.

-110

-100

-90

-80

-70

-60

0 5 10 15 20

Ph

ase

No

ise

@1

00

kHz

[d

Bc/

Hz]

Tune [V]

PN @100kHz vs. tune




29

5.5 FUNDAMENTAL 2-4 GHZ OSCILLATOR WITH A VARACTOR DIODE OF

2.7PF@4V

The choice of MMIC and chip capacitor has now been determined and the next step is to

finalize a series of VCOs using the chosen MMIC.

A fundamental MMIC VCO core assembled with two varactor diodes of 2.7 pF @4V tune

was assembled, Figure 5.18, and the tank was optimized to gain the optimal performance. The

output frequency versus tuning voltage is found in Figure 5.19. The VCO demonstrates more

than octave bandwidth.

Figure 5.18 MMIC with resonance circuit for 2-4 GHz oscillator.

Figure 5.19 Measured frequency versus tune voltage for a fundamental MMIC VCO core.

1,5

2

2,5

3

3,5

4

4,5

0 5 10 15 20

Fre

qu

en

cy [

GH

z]

Tune [V]

Frequency vs. tune


30

The lengths of the bond wires in the tank are estimated by Pythagoras' theorem in accordance

with Figure 5.20.

Figure 5.20 Estimation of the bond wire.

The parameters, , , 𝑏 and 𝑏 is measured in a microscope and Pythagoras' theorem gives

an estimation of the length according to

𝑙 𝑙 𝑙 𝑏

𝑏

Measured lengths and heights for the bond wires in the used tank are listed in the Table 5.4

below. Ideally, they should be equal but in practice, there is always a slight difference

between the two wires. The differences will not affect the bandwidth for the VCO but will

affect the suppression of harmonics due to an unbalanced resonance tank.

𝑚 𝑏

𝑚 𝑙

𝑚

𝑚 𝑏

𝑚 𝑙

𝑚 𝑙 𝑙 𝑙

𝑚

𝑛

Wire 1 210 392 445 250 1150 1177 1622 1.51

Wire 2 170 224 281 140 1093 1102 1383 1.25

Table 5.4 Measured lengths and heights for the bond wires.

The length of each wire is estimated to 1622 𝑚 and 1383 𝑚 respectively due to equation

(5.1) Pythagoras' theorem. The total length is approximately 3000 𝑚 and hence a total

inductance for the resonance circuit of 2.76 nH is calculated from equation (3.11).

𝑙 𝑙

𝑏 𝑏


31

From table 5.4, measured and calculated values for oscillator are listed in the Table 5.5 below.

For the calculations, the nominal value of 2.7 pF was chosen for the varactor capacitance.

𝑓 [GHz]

𝑓

[GHz)

[GHz] 𝑓

[%]

[pF]

[nH]

𝑓 @4V tune

[GHz]

𝑓 @4V tune

[GHz]

+25˚C 1.87 4.07 2.2 54.1 1.35 2.76 2.61 2.80

Table 5.5 Measured and calculated values for the oscillator in room temperature.

Due to the equation (3.4) and (3.11) the calculated center frequency is 𝑓 = 2.61 GHz @4V

tune, and show a difference of 7.3 % compared with the measured value. Given the crude

inductance model and that the actual value of the capacitance is unknown, this is considered

to be a good agreement.

Furthermore, the capacitive part of an ideal parallel tank can be calculated from the output

frequency and the total inductance found in the same tank according to:

𝜋𝑓

The actual capacitance value of the “2.7 pF @ 4V” varactor as a function of tuning voltage

was measured with a capacitance-meter. This measured capacitance value is plotted in Figure

5.21 together with the calculated capacitance value from equation (5.2).

Figure 5.21 Difference of measured and calculated frequency.

0 2 4 6 8 10 12 14 16 18 200

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Tune [V]

Capacitance [

pF

]

Mesaured capacitance

Calculated capacitance


32

From Figure 5.21, it is clear that the difference between the calculated C (i.e., the effective C

that determines the oscillating frequency) is not a simple constant parasitic element since the

calculated C is lower than the measured C at low tuning voltages and larger than the measured

at high tuning voltages. The differences between measured and calculated capacitance of the

tank are instead due to both parasitic capacitive and inductive elements in the overall VCO.

Output power and phase noise of the VCO versus tuning voltage is also measured and are

found in Figure 5.22 and Figure 5.23, respectively.

Figure 5.22 Output power versus tune voltage.

Figure 5.23 Phase noise versus tune voltage.

-23

-18

-13

-80 5 10 15 20

Po

we

r [d

Bm

]

Tune [V]

Power vs. tune

-105

-100

-95

-90

-85

-80

0 5 10 15 20

Ph

ase

No

ise

@1

00

kHz

[d

Bc/

Hz]

Tune [V]

PN @100kHz vs. tune


33

A typical phase noise spectrum is found in Figure 5.24. The local increase of phase noise

around 50 kHz is a disturbance from the lab environment and can thus be neglected.

Figure 5.24 Phase noise in room temperature @10V tune voltage.

Measurement A borted

R &S F S U P S i gna l S ourc e A na l y ze r L O C K E D

S e t t i ngs R e s i dua l N o i s e [T 1 w/o s purs ] P ha s e D e te c to r +0 dB

Signal Frequency: 3.683636 GHz Int PHN (1 .0 k .. 3 .0 M) -15.8 dBc

Signal Level: -5 .46 dBm Res idual PM 13.201 °

P LL Mode Harmonic 1 Res idual FM 4.083 kHz

Internal Ref Tuned Internal P hase Det RMS Jitter 9 .9549 ps

P hase Noise [dBc/Hz] Marker 1 [T1] Marker 2 [T1] Marker 3 [T1]

RF A tten 0 dB 10 kHz 100 kHz 1 MHz

Top -30 dBc/Hz -73.58 dBc/Hz -95.33 dBc/Hz -115.39 dBc/Hz

10 kHz 100 kHz 1 MHz1 kHz 3 MHz

-130

-120

-110

-100

-90

-80

-70

-60

-50

-40LoopBW

1 CLRW R

SMTH 1%

2 CLRW R

*

A

PA

SPR OFF

TH 0dB

Frequency Offset

1

2

3

Date: 27.NOV.2009 15:58:16


34

5.6 HEIGHT ANALYSIS OF THE INDUCTOR BOND WIRES IN THE TANK

5.6.1 ANALYSIS BASED ON MEASURED CAPACITANCE OF THE VARACTORS

The capacitance of one varactor diode has been measured over the tuning range, 0-20 V, in

order to evaluate the effects on the bandwidth and center frequency for the oscillator by

letting the heights for the inductor bond wires vary 50%. The varactor capacitance @4V

tune is measured 2.9 pF instead of 2.7 pF and hence the center frequency will decrease from

2.61 GHz to 2.5 GHz as a consequence. The length and heights for the bonding wires are

assumed to be equal to the ones listed in Table 5.4. The frequency of oscillation is calculated

according to equation (3.4).

Figure 5.25 Different heights of the bonding wires based on measured tank capacitance.

Figure 5.26 shows the data from Figure 5.25 once again, but this time, frequency is plotted

versus height difference for different tuning voltages with a voltage range of 0-20 V in steps

of 1 V.

0 2 4 6 8 10 12 14 16 18 201

1.5

2

2.5

3

3.5

4

4.5

5

Tune [V]

Fre

quency [

GH

z]

Frequency vs. tune with different heights of the bond wire

- 50%

- 40%

- 30%

- 20%

- 10%

0%

+ 10%

+20%

+ 30%

+40%

+ 50%


35

Figure 5.26 Frequency vs. different heights of the bond wires in percent with a voltage range

of 0-20 V in steps of 1 V.

In conclusion, the height of the bond wire can differ with 50% without too much effect on

the frequency range in the case of an ideal tank configuration without parasitic capacitances in

the MMIC VCO core.

-50 -40 -30 -20 -10 0 10 20 30 40 501

1.5

2

2.5

3

3.5

4

4.5

5

Different heights of the bond wires [%]

Fre

quency [

GH

z]

Different heights of the bond wires


36

5.6.2 ANALYSIS BASED ON CALCULATED CAPACITANCE OF THE TANK

The analysis in section 5.5.1 was repeated but instead of using the measured value of the

varactor capacitance, the calculated values of the effective tank capacitor according to

equation (5.2) were used and plotted in Figure 5.27 and 5.28 below.

Figure 5.27 Different heights of bond wires with a calculated tank capacitance.

Figure 5.28 Frequency vs. different heights of the bond wires in percent with a voltage range

of 0-20 V in steps of 1 V and calculated tank capacitance.

0 2 4 6 8 10 12 14 16 18 201.5

2

2.5

3

3.5

4

4.5

Tune [V]

Fre

quency [

GH

z]

Frequency vs. tune with different heights of the bond wire

- 50%

- 40%

- 30%

- 20%

- 10%

0%

+ 10%

+20%

+ 30%

+40%

+ 50%

-50 -40 -30 -20 -10 0 10 20 30 40 501.5

2

2.5

3

3.5

4

4.5

Different height of the bond wires [%]

Fre

quency [

GH

z]

Different height of the bond wires


37

From Figure 5.27 and Figure 5.28, it is clear that the bandwidth of the VCO is more sensitive

to changes in the height of the bond wires when using the effective tank capacitance

compared to when using the measured stand-alone varactor capacitance. This is explained

from the simple fact that the capacitance range is smaller in the former case. Thus, the LC

product which determines the frequency in equation (3.4) will vary less in the former case and

the bandwidth will hence decrease.


38

5.7 ANALYSIS OF TANK CAPACITANCE

In analogue with chapter 5.5, the tank capacitance is also varied and the effect on the

bandwidth and center frequency for the oscillator from this deviation is also evaluated. The

tank capacitance is varied 10% which is a typical tolerance for commercial varactors. The

length and heights for the bonding wires are assumed to be the equal to the ones listed in

Table 5.4.

5.7.1 ANALYSIS BASED ON MEASURED CAPACITANCE OF THE TANK

An analysis based on the capacitance values measured for one varactor diode over the tuning

range 0-20 V in steps of 1 V is found in Figure 5.29.

Figure 5.29 Different values of the tank capacitance that been measured.

In conclusion, the varactor capacitance can differ 10% with a minor effect on the frequency

range when using the measured capacitance values of the varactor.

0 2 4 6 8 10 12 14 16 18 201

1.5

2

2.5

3

3.5

4

4.5

5

Tune [V]

Fre

quency [

GH

z]

- 10%

- 5%

0%

+ 5%

+ 10%


39

5.7.2 ANALYSIS BASED ON CALCULATED CAPACITANCE OF THE TANK

Same analysis done for calculated values according to equation (5.2) based on the measured

device is shown in Figure 5.30 below.

Figure 5.30 Different values of the tank capacitance with calculated values.

As in the case for the height analysis of the bond wires, the bandwidth decreases when using

the effective calculated capacitance in the tank. The origin of this is once again the smaller

variation in effective calculated capacitance compared to when using the measured values.

0 2 4 6 8 10 12 14 16 18 201.5

2

2.5

3

3.5

4

4.5

Tune [V]

Fre

quency [

GH

z]

- 10%

- 8%

- 5%

- 4%

- 2%

0%

+ 2%

+ 4%

+ 6%

+ 8%

+ 10%


40

5.8 PERFORMANCE VERSUS TEMPERATURE

The VCOs evaluated in this thesis are intended to be operated over the standard temperature

range for similar products, i.e. -40 to +85 deg. For these measurements, a temperature

chamber was used to change the range of temperature. The temperature chamber is cooled by

gas and the temperature is measured with a wire sensor type K connected to the device

inside the chamber and to a multimeter outside the chamber.

Figure 5.31 Schematic of the measurement setup at measurements over temperature.

The 2-4 GHz VCO was evaluated for varying temperature and frequency. Output power and

phase noise were measured and plotted in Figure 5.32, 5.33 and 5.34.

Figure 5.32 Frequency versus tune voltage for different temperatures.

Temperature chamber

OptionalPower amplifier

FSUP Signal Source

Analyzer

Package with SMA

connectors

Power supply

1,5

2

2,5

3

3,5

4

4,5

0 5 10 15 20

Fre

qu

en

cy [

GH

z]

Tune [V]

Frequency vs. tune

Freq[GHz] -40˚C

Freq[GHz] +85˚C

Freq[GHz] +25˚C


41

𝑓 [GHz]

𝑓

[GHz)

[GHz] 𝑓

[%]

[pF]

[nH]

𝑓 @4V tune

[GHz]

𝑓 @4V tune

[GHz]

1.84 4.13 2.29 55.4 1.35 2.76 2.61 2.92

1.84 4.07 2.23 54.8 1.35 2.76 2.61 2.79

1.87 4.07 2.2 54.1 1.35 2.76 2.61 2.80

Table 5.6 Measured and calculated values for different temperatures.

As can be seen in Table 5.6, the difference in bandwidth over frequency is very small.

The alternation of frequency due to the temperatures gives a temperature drift of 1.5 MHz / .

This frequency shift is however mainly due to changing collector-bias current versus

temperature. And, as already discussed, the tuning voltage over the varactors is dependent on

the absolute collector voltage. And since the collector current varies in temperature, the

collector voltage varies and so does the tuning voltage. This will result in a slight frequency

shift over temperature. In a product, the VCOs will instead be biased with constant current

and the frequency shift over temperature will be smaller.

Figure 5.33 Output power versus tune voltage at different temperatures.

Power vs. temperature behavior has a maximum of 2.2 dB difference @0V tune.

-23

-18

-13

-8

0 5 10 15 20

Po

we

r [d

Bm

]

Tune [V]

Power vs. tune

-40˚C

+85˚C

+25˚C


42

Figure 5.34 Phase noise versus tune voltage at different temperatures.

The phase noise of the VCO is different over temperature. The difference is roughly

maximum 5 dB which is considered to be acceptable.

In summary, the 2-4 GHz VCO behaves well over temperature. Even better performance is

expected in a product where the VCO is biased with a constant current rather than a constant

voltage.

-105

-100

-95

-90

-85

-80

0 5 10 15 20

Ph

ase

No

ise

@1

00

kHz

[d

Bc/

Hz]

Tune [V]

PN @100kHz vs. tune

-40˚C

+85˚C

+25˚C

Chapter 6. Harmonic 4-8 GHz And 8-16 GHz Oscillator

With A Varactor Diode Of 2.7 pF@4V

43

6 HARMONIC 4-8 GHZ AND 8-16 GHZ OSCILLATOR WITH A

VARACTOR DIODE OF 2.7 PF@4V

Four different harmonic MMIC VCO cores with the same properties as for the fundamental

MMIC cores shown in chapter 5 have been evaluated with the same approach as for the

fundamental VCO. According to the conclusions made in sections 5.2 and 5.3 one harmonic

MMIC VCO core has been evaluated with the varactor diodes assembled next to the MMIC

and connected with bond wires.

Figure 6.1 Drawing, MMIC design and schematic for a harmonic VCO core with the diodes

assembled next to the MMIC.


X2 RF output

Fundamental RF output 𝑐𝑐

𝑡𝑢𝑛𝑒

33 pF

0


𝑜𝑢𝑡

𝑏𝑏

𝑏𝑏



44

6.1 SUPPRESSION

The different output tones were measured over tuning voltage, Figure 6.7.

Figure 6.2 Suppression of the odd harmonics in room temperature.

As expected, the even harmonics are dominant and the odd harmonics are suppressed due to

the balanced structure of the VCO, section 2.3. The suppression of the odd harmonics, labeled

P1, P3 and P5, depends on the symmetry of the tank. Optimally, they would cancel totally.

The measurements shows that the fourth harmonic is also of interest and such a VCO will be

measured and the results are shown in section 6.3.

-70

-65

-60

-55

-50

-45

-40

-35

-30

0 5 10 15 20

Po

wer [

dB

m]

Tune [V]

P1

P2

P3

P4

P5



45

6.2 PERFORMANCE AS FUNCTION OF TEMPERATURE FOR A HARMONIC 4-8

GHZ OSCILLATOR

The second harmonic output result for the harmonic MMIC VCO core are shown in Figure

6.3, 6.4 and 6.5 over temperature.



3

4

5

6

7

8

9

0 5 10 15 20

Fre

qu

en

cy [

GH

z]

Tune [V]

Frequency vs. tune

-40˚C

+85˚C

+25˚C

-48

-46

-44

-42

-40

-38

-36

-34

-32

-30

0 5 10 15 20

Po

we

r [

dB

m]

Tune [V]

Power vs. tune

-40˚C

+85˚C

+25˚C



46

The output power has a periodic behavior that is probably due to mismatch at the RF output,

i.e. the connection between the alumina and the SMA adapter which is not optimized for high

frequencies.


In summary, the harmonic VCO using the 2nd

harmonic of the VCO works very well.

-100

-95

-90

-85

-80

-75

-70

-65

-60

0 5 10 15 20

Ph

ase

No

ise

@1

00

kHz

[dB

c/H

z]

Tune [V]

PN @100kHz vs. tune

-40˚C

+85˚C

+25˚C



47

6.3 PERFORMANCE AS FUNCTION OF TEMPERATURE FOR A HARMONIC 8-16

GHZ OSCILLATOR

The fourth harmonic output result for the harmonic MMIC VCO core are shown in Figure 6.6,

6.7 and 6.8.



7

9

11

13

15

17

19

0 5 10 15 20

Fre

qu

en

cy [

GH

z]

Tune [V]

Frequency vs. tune

Freq[GHz] -40˚C

Freq[GHz] +85˚C

Freq[GHz] +25˚C

-44

-43

-42

-41

-40

-39

-38

-37

-36

0 5 10 15 20

Po

we

r [d

Bm

]

Tune [V]

Power vs. tune

-40˚C

+85˚C

+25˚C



48


In summary, the harmonic VCO using the 4th

harmonic of the VCO also works very well.

-90

-85

-80

-75

-70

-65

-60

0 5 10 15 20

Ph

ase

No

ise

@1

00

kHz

[dB

c/H

z]

Tune [V]

PN @100kHz vs. tune

-40˚C

+85˚C

+25˚C

Chapter 7. Summary And Conclusions

49

7 SUMMARY AND CONCLUSIONS

This work has been focused on the design, assembly, and evaluation of broadband VCOs

using external tanks. Different MMIC VCO cores have been evaluated and the best

performing topology has been used for the design of a family of octave (2:1) VCOs covering

2-27 GHz in bands.

A tolerance analysis of the height of the bond wires used as inductors in the tank shows that

they are not critical for the performance of the VCOs. In addition, a tolerance analysis of the

varactor diodes was also done with similar result.

Due to practical limitations in the length of tank-inductance, bond wires and size of tank-

varactor diode it is difficult to obtain higher frequencies than approximately 10 GHz with a

fundamental VCO core. To extend the frequency range beyond 10 GHz, a harmonic MMIC

VCO core employing the 2nd

and 4th

harmonic is therefore designed and evaluated.

A significant part of this thesis includes design of the alumina carriers necessary to assembly

and test the MMIC core with suitable tank circuits.

The results assist the future design for a more integrated MMIC VCO cores with integrated

buffer amplifiers.

Two different MMIC VCO cores together with three different resonance circuits cover a

frequency range of 2-27 GHz and the result is shown in Figure 7.1.

Chapter 7. Summary And Conclusions

50

Figure 7.1 Result of different MMIC VCO core together with different resonance tank that

together covers a frequency range of 2-27GHz.

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

0 2 4 6 8 10 12 14 16 18 20

Fre

qu

en

cy (G

Hz)

Vtune (V)NOTE!! Absolute Vtune is roughly 5 V higher!!

Frequency vs. Vtune

2.7pF, X1, VCO7, 1.9-4.1 GHz

2.7pF, X2, VCO8, 3.9-8.4 GHz

2.7pF, X4, VCO8, 7.8-16.9 GHz

2.0pF, X1, VCO7, 2.4-5.2 GHz

2.0pF, X2, VCO8, 5.5-12.5 GHz

2.0pF, X4, VCO8, 11.1-24.9 GHz

1.2pF, X4, VCO8, 14.7-27 GHz

-100

-90

-80

-70

-60

-50

-40

0 2 4 6 8 10 12 14 16 18 20

Ph

ase

no

ise

@ 1

00

kH

z

Vtune (V)NOTE!! Absolute Vtune is roughly 5 V higher!!

Phase noise @ 100 kHz vs. Vtune

2.7pF, X1, VCO7, 1.9-4.1 GHz

2.7pF, X2, VCO8, 3.9-8.4 GHz

2.7pF, X4, VCO8, 7.8-16.9 GHz

2.0pF, X1, VCO7, 2.4-5.2 GHz

2.0pF, X2, VCO8, 5.5-12.5 GHz

2.0pF, X4, VCO8, 11.1-24.9 GHz

1.2pF, X4, VCO8, 14.7-27 GHz

Chapter 8. Future Work

51

8 FUTURE WORK

The next generation of MMIC designs will include buffer amplifiers for more constant output

power versus tuning voltage. The constant voltage bias-schema of the VCOs will also be

replaced by a constant current bias for improved VCO performance.

Another consideration is to replace the GaAs varactor diodes in the resonance circuit with Si

varactor diodes and thus lower the cost by a factor of ten for every varactor device.

Chapter 9. Acknowledgment

52

9 ACKNOWLEDGMENT

I would like to acknowledge my supervisors Sten Gunnarsson, Christer Stoij and Dan

Kuylenstierna for their big support, engagement, patience and advice during this thesis. I

would also like to acknowledge my examiner Ass. Prof. Urban Westergren for his support and

advice.

Especially thanks to Managing Director Olle Westblom for giving me opportunity to perform

this thesis project at the company Sivers IMA AB.

Big thanks to all of my colleagues at Sivers IMA AB for the support and encouragements.

Finally, I would like to give a big thanks to my family and friends for their support and

patience during my years at KTH Royal Institute of Technology.

This work has partly been carried out within the GHz centre in the INTOSC project financed

by Vinnova, Chalmers, Sivers IMA AB and Ericsson AB.

Chapter 10. References

53

10 REFERENCES

[1] Sten Gunnarsson, Christer Stoij and Dan Kuylenstierna. Private

communications, 2008-2010.

[2] Behzad Razavi, “Design of Analog CMOS integrated Circuits”.

[3] Herbert Zirath, Harald Jacobsson, M. Bao, Mattias Ferndahl, Rumen

Kozhuharov, “MMIC-Oscillator designs for ultra low phase noise”, Proc. of

2005, Compound Semiconductor Integrated Circuit Symposium (CSICS).

[4] Allan R. Hambley, “Electrical Engineering, Principles and Applications”

2nd

edition, Prentice Hall, Upper Saddle River, NJ 07458, USA,

ISBN 0-13-094349-5.

[5] Robert E. Collin,”Foundations for Microwave Engineering”,

ISBN 0-07112569-8.

[6] Allen Sweet, “MIC&MMIC amplifier and oscillator circuit design”, Artech

House Inc, 685 Canton Street, Norwood, MA 02062, ISBN 0-89006-305-2.

[7] Ali Hajimiri, Thomas H. Lee, ”A general Theory of Phase Noise in Electrical

Oscillators”, IEE Journal of Solid-State Circuits, Vol. 33, No 2, February 1998.

[8] R&S®FSUP Signal Source Analyzer, Operating Manual, © 2009 Rohde &

Schwarz GmbH & Co. KG, 81671 Munich, Germany, NJ 07458, USA, ISBN 0-

13-887571-5.

Appendix A. Alumina Design

54

APPENDIX A

ALUMINA DESIGN

Four different thin film (alumina) carriers have been designed in AutoCad to allow for easy

evaluation of the MMIC VCO cores together with different resonance tanks. The carriers also

includes features such as low pass filters for all three bias connections.

The alumina is manufactured with gold plated via holes and resistance (45 /square)

consisting of tantalum nitride ( ).


55

SG_XcVCO_1+3+12+13

Figure A1 Alumina design for varactor diodes flipped on a fundamental MMIC VCO core.

RF outputRF outputOptional LC-filters

Connection pads for bias

Via hole

Varactor diodes flipped on MMIC




56

SG_XcVCO_2+4+10

Figure A2 Alumina design for varactor diodes flipped on a harmonic MMIC VCO core.

Varactor diodes flipped on MMIC




57

SG_XcVCO_5+7

Figure A3 Alumina design for varactor diodes assembled beside fundamental MMIC VCO

core.

Varactordiodes assembled adjacent to MMIC




58

SG_XcVCO_6+8

Figure A4 Alumina design for varactor diodes assembled beside harmonic MMIC VCO core.

Varactordiodes assembled adjacent to MMIC



Appendix B. Measurement Details

59

APPENDIX B

MEASUREMENT DETAILS

Phase noise measurement for the fundamental oscillators is done with R&S®FSUP26 Signal

Source Analyzer.

The DUT signal is mixed with a signal from a reference source in the FSUP. When both

signals exhibit the same frequency, a DC voltage is obtained at the output of the mixer or

phase comparator that is superimposed by the noise from the DUT and the reference source.

The 90˚ offset is adjusted at the reference signal source and phase noise can be measured at

the output after a low pass filter.

Figure B1 Schematic for the phase comparator method for phase noise measurements. [8]

LP filter

PLL

Reference

source

Mixer

DUT

Phase noise

result

Documents

Evaluation of broadband MMIC oscillators with external tank401135/FULLTEXT01.pdfThe cross-coupled topology was chosen to be used in the broadband VCO core due to its broadband performance