Upload
dangthien
View
218
Download
0
Embed Size (px)
Citation preview
RF Power Amplifiers and MEMS Varactors
by Sareh Mahdavi
Department of Electrical & Computer Engineering McGill University, Montréal
November 2007
A thesis submitted to the Faculty of Graduate Studies and Research in partial fulfillment of the requirements for the degree of Master of Engineering.
© Sareh Mahdavi, 2007
1+1 Library and Archives Canada
Bibliothèque et Archives Canada
Published Heritage Bran ch
Direction du Patrimoine de l'édition
395 Wellington Street Ottawa ON K1A ON4 Canada
395, rue Wellington Ottawa ON K1A ON4 Canada
NOTICE: The author has granted a nonexclusive license allowing Library and Archives Canada to reproduce, publish, archive, preserve, conserve, communicate to the public by telecommunication or on the Internet, loan, distribute and sell theses worldwide, for commercial or noncommercial purposes, in microform, paper, electronic and/or any other formats.
The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
ln compliance with the Canadian Privacy Act some supporting forms may have been removed from this thesis.
While these forms may be included in the document page count, their removal does not represent any loss of content from the thesis.
• •• Canada
AVIS:
Your file Votre référence ISBN: 978-0-494-51467-2 Our file Notre référence ISBN: 978-0-494-51467-2
L'auteur a accordé une licence non exclusive permettant à la Bibliothèque et Archives Canada de reproduire, publier, archiver, sauvegarder, conserver, transmettre au public par télécommunication ou par l'Internet, prêter, distribuer et vendre des thèses partout dans le monde, à des fins commerciales ou autres, sur support microforme, papier, électronique et/ou autres formats.
L'auteur conserve la propriété du droit d'auteur et des droits moraux qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.
Conformément à la loi canadienne sur la protection de la vie privée, quelques formulaires secondaires ont été enlevés de cette thèse.
Bien que ces formulaires aient inclus dans la pagination, il n'y aura aucun contenu manquant.
Abstract
This thesis is concerned with the design and implementation of radio frequency (RF)
power amplifiers and micro-electromechanical systems - namely MEMS varactors. This
is driven by the many wireless communication systems which are constantly moving
towards increased integration, better signal quality, and longer battery life.
The power amplifier consumes most of the power in a receiver/transmitter system
(transceiver), and its output signal is directly transmitted by the antenna without further
modification. Thus, optimizing the PA for low power consumption, increased linearity,
and compact integration is highly desirable.
Micro-electromechanical systems enable new levels of performance in radio-frequency
integrated circuits, which are not readily available via conventional IC technologies.
They are good candidates to replace lossy, low Q-factor off-chip components, which have
traditionally been used to implement matching networks or output resonator tanks in class
AB, class F, or classE power amplifiers. The MEMS technologies also make possible the
use of new architectures, with the possibility of flexible re-configurability and tunability
for multi-band and/or multi-standard applications.
The major effort of this thesis is focused on the design and fabrication of an RF
frequency class AB power amplifier in the SiGe BiCMOS 5HP technology, with the
capability of being tuned with extemal MEMS varactors. The latter necessitated the
exploration of wide-tuning range MEMS variable capacitors, with prototypes designed
and fabricated in the Metal-MUMPs process.
An attempt is made to integrate the power amplifier chip and the MEMS die in the same
package to provide active tuning of the power amplifier matching network, in order to
keep the efficiency of the PA constant for different input power levels and Joad
conditions.
Detailed simulation and measurement results for ail circuits and MEMS deviees are
reported and discussed.
Il
Résumé
Cette thèse est concernée par la conception et l'exécution des amplificateurs de puissance
de la radiofréquence (RF) et des systèmes micro-électromécaniques - comme les
varactors de MEMS. Ceux-ci sont conduits par de nombreux systèmes de communication
sans-fil qui se déplacent constamment vers une croissance en intégration, une meilleure
qualité de signal, et une plus longue vie de batterie.
L'amplificateur de puissance consomme la majeure partie de la puissance dans un
récepteur/système émetteur (émetteur récepteur), et son signal de sortie est directement
transmis par une antenne sans modifier davantage. Ainsi, pour optimiser la PA pour une
consommation de puissance basse, une linéarité accrue et une intégration compacte sont
fortement souhaitables.
Les systèmes micro-électromécaniques (MEMS) permettent de nouveaux niveaux
d'exécution dans des circuits intégrés de radiofréquence, qui ne sont pas facilement
disponibles par l'intermédiaire des technologies conventionnelles d'IC. Ils sont de bons
candidats pour remplacer les composants hors puce de bas facteur-Q, qui ont été
traditionnellement employés pour mettre en application les réseaux ou les réservoirs
assortis de résonateur de rendement dans des amplificateurs de puissance de la classe AB,
de la classe F, ou de la classe E. Les technologies de MEMS rendent également
lll
l'utilisation possible de nouvelles architectures, avec la possibilité de re-configuration et
de tunabilité flexibles pour des applications multibandes et/ou multistandardes.
L'effort principal de cette thèse est concentré sur la conception et la fabrication d'un
amplificateur de puissance de la classe AB de fréquence de RF en technologie de SiGe
BiCMOS 5HP, avec les possibilités de l'accord avec les varactors externes de MEMS. Le
dernier a rendu nécessaire l'exploration des condensateurs variables de large-accord de la
gamme MEMS, avec des prototypes conçus et fabriqués dans le processus de Métal
MUMPS.
Une tentative est d'intégrer le morceau d'amplificateur de puissance et la matrice de
MEMS dans le même paquet pour fournir l'accord actif du réseau assorti d'amplificateur
de puissance, afin de garder l'efficacité de la PA constante pour différents niveaux de
puissance d'entrée et conditions de charge.
Des résultats détaillés de simulation et de mesure pour tous les circuits et dispositifs de
MEMS sont rapportés et discutés.
lV
Acknowledgments
First and foremost, I would like to thank my supervisor Professor Mourad El-Gamal for
his patience, invaluable guidance, encouragements and for giving me the opportunity to
pursue a masters' degree at the McGili Radio-Frequency Integrated Circuits (RFIC) lab.
1 would also like to thank ali the members of the RFIC lab especialiy Nicolas Constantin
and Tommy Tsang, who were always there to help and patiently answered my questions.
1 would like to thank Laurent Mouden from École Polytechnique de Montréal who helped
with the wire-binding of my chips. 1 am also grateful to Michele Perucic for her help with
the setting up SiGe environment in Cadence.
Many thanks to other members of the RFIC group Mohamed Shaheen, Jane Yu, David
Hong, Frederic Nabki, Kuan-Yu Lin, Hanzhen Zhang, Barry Zhao, and Erica Kwizak.
My deep gratitude to my good friends Y asmin Ahmad, Patricia Lee, and Hung Pao Yang.
1 sincerely thank my sister Sajedeh, and my brothers Saied and Sadegh for their
understanding and support.
This thesis is dedicated to my parents Ra'na Shahintabe and Gholamreza Mahdavi for ali
their love, support, and encouragement.
v
Table of Contents
1 INTRODUCTION ..................................................................................................................................... 1
1.1 MOTIVATION ........................................................................................................................................ 2
1.2 LITERA TURE REVIEW ......... ························ ........................................................................................... 4
1.3 OBJECTIVES .......................................................................................................................................... 7 1.4 THESIS ÜUTLINE ................................................................................................................................... 7
1.5 CONTRIBUTIONS ................................................................................................................................... 9
2 POWER AMPLIFIERS .......................................................................................................................... 10
2.1 PA ME TRIC DEFINITIONS .................................................................................................................... 11 2.1.1 Output Power and Power Gain .................................................................................................. 11 2.1.2 Efficiency ................................................................................................................................... 12
2.1.2.1 Gollector Efficiency .......................................................................................................................... 12 2.1.2.2 Power-Added-Efficiency ................................................................................................................. 12
2.1.3 Linearity ..................................................................................................................................... 13 2.1.3.1 Sources of Nonlinearity .................................................................................................................. 14 2.1.3.2 Measures of Linearity ..................................................................................................................... 15
2.1. 4 Stability ............. ......................................................................................................................... 15 2.2 CHOICE OF ARCHITECTURE ................................................................................................................. 16
2. 2.1 Linear Amplijiers ..................................... .................................................................................. 17 2.2.1.1 Glass A ............................................................................................................................................. 17 2.2.1.2 Glass B ............................................................................................................................................. 19 2.2.1.3 Glass AB ........................................................................................................................................... 20
2.2.2 Nonlinear Amplifier ................................................................................................................... 20 2.2.2.1 Glass G ............................................................................................................................................. 20 2.2.2.2 Glass F .............................................................................................................................................. 21 2.2.2.3 Glass E ............................................................................................................................................. 22
2.2. 3 High-Linearity and High-Efficiency RF PA - Tradeoffi ............. ............................................... 24 2.2.4 Analysis ofClass AB Power Amplijiers ..................................................................................... 24
2.2.4.1 Harmonie Termination .................................................................................................................... 27 2.2.4.2 Tunable Load Resistance .............................................................................................................. 27
2.3 PA DESIGN TECHNIQUES .................................................................................................................... 28 2.3.1 Power Match vs. Corifugate Match ............................................................................................ 29 2.3.2 Load-Pull Technique .................................................................................................................. 31
3 DESIGN OF POWER AMPLIFIERS ................................................................................................... 33
3.1 PA DESIGN IN A SIGE TECHNOLOGY .................................................................................................. 33 3.1.1 Choice ofTechnology ................................................................................................................. 33
VI
3.1.2 Power Amplifier Specifications .................................................................................................. 34 3.1.3 Single Transistor Design ............................................................................................................ 34
3.1.3.1 Biasing Circuit Design .................................................................................................................... 35 3. 1.3.2 Power Core Design ......................................................................................................................... 38 3.1.3.3 Input Matching Network .................................................................................................................. 42 3.1.3.4 Output Impedance Transformation ............................................................................................... 42 3.1.3.5 Simulation Results .......................................................................................................................... 44
3.1.4 Multiple Transistor Design ........................................................................................................ 45 3.1.4.1 ESD Protection ................................................................................................................................ 48
3.2PALAYOUT ........................................................................................................................................ 49
3.3 EXTRACTED PA SIMULATION RESULTS .............................................................................................. 51
3.4 PACKAGE, BONDWIRE, AND PCB MO DELS ......................................................................................... 53
3.4.1 Bondwire Mode! ......................................................................................................................... 54 3.4.2 Package Mode! ........................................................................................................................... 54 3.4.3 PCB Track Mode! ...................................................................................................................... 55 3.4. 4 Final Simulation Results ............................................................................................................ 55
4 MEMS TUNABLE POWER AMPLIFIER ........................................................................................... 56
4.1 MEMS TUNABLE CAPACITORS .......................................................................................................... 57 4.1.1 Overview .................................................................................................................................... 57 4.1.2 Principle ofOperation ............................................................................................................... 58
4.1.2.1 Parallei-Piate MEMS Varactor ....................................................................................................... 58 4.1.2.1(a) Type II Vertical Varactor ........................................................................................................... 60 4.1.2.1(b) Type III Vertical Varactor ......................................................................................................... 61
4.1.2.2 Laterallnterdigitated MEMS Varactors ........................................................................................ 61 4.1.2.3 Semi-Fractal Varactor ..................................................................................................................... 62
4.1.3 The Metal-MUMPS Process ...................................................................................................... 63 4.1.4 MEMS Varactor Design in Metal-MUMPS ............................................................................... 64
4.1.4.1 Suspension Design ......................................................................................................................... 66 4.1.4.2 Signal Pad Design ........................................................................................................................... 68
5 MEASUREMENT RE SUL TS ................................................................................................................ 69
5.1 THE TEST SETUP ................................................................................................................................. 69 5.2 STAND-ALONEPOWERAMPLIFIER TESTING ....................................................................................... 71
5.3 MEMS VARACTORS TESTING ............................................................................................................. 74
5.4 MEMS-TUNABLE POWER AMPLIFIER TES TING .................................................................................. 78
6 CONCLUSION ........................................................................................................................................ 81
6.1 SUMMARY .......................................................................................................................................... 81
6.2 TOPICS FOR FUTURE RESEARCH ......................................................................................................... 82
7 REFERENCES ........................................................................................................................................ 84
APPENDIX A: ADS MO DEL FITTING ................................................................................................. 92
vii
List of Figures
FIGURE 1.1: SIMPLIFIED TRANSMITIER ARCHITECTURE ...................•.........•..........•...•.....•••.....•..•.....•••......... 1
FIGURE 1.2: GENERIC SINGLE-ENDED PA HIGHLIGHTING SECTIONS THAT GAN BE INTEGRATED AS MEMS-
BASED COMPONENTS .......•..............•...........................•.......•.............•................•............•............••.•..... 2
FIGURE 1.3: POWER AMPLIFIER OUTPUT POWER PROBABILITY DENSITY FUNCTION FOR IS-95 URBAN AND
SUBURBAN ENVIRONMENTS [5], [6] .....................................•.........................................•...................... 3 FIGURE 1.4: THE POWER-CONTROLLABLE CLASS-E PA IMPLEMENTED IN [9] WITH TUNABLE OUTPUT
CAPACITORS ................................•..............................................•.........•......•......................•...............•.. 4
FIGURE 1.5: SIMULATED RESUL TS OF POWER-ADDED-EFFICIENCY VERSUS OUTPUT POWER FOR THE
POWER-CONTROLLABLE PA PROPOSED IN [9] ..................................................................................... 5
FIGURE 1.6: SIMPLIFIED CIRCUIT OF ELECTRONICALL Y TUNABLE GLASS E PA DEVELOPED IN [1 0] ............ 6
FIGURE 1.7: SCHEMA TIC OF THE OUTPUT MATCHING NETWORK WITH FOUR MEMS SWITCHES (M1-M4) AND ONE (SEMICONDUCTOR] VARACTOR. C1-C8 ARE FIXED CAPACITORS. MUNIS A MICROSTRIP
TRANSMISSION UNE (14] ....................................................................................................................... 6
FIGURE 1.8: RESULTS OBTAINED BY HAVING TUNABLE MATCHING NETWORKS IN [14]. THE DOTS
REPRESENT THE OPTIMIZED OUTPUT POWER AS A FUNCTION OF FREQUENCY. EACH CURVE
REPRESENTS THE FREQUENCY RESPONSE OF THE AMPLIFIER WHEN IT IS OPTIMALL Y CONFIGURED
FOR ONE SPECIFIC FREQUENCY [14]. ................................................................................................... 7
FIGURE 2.9: HARMONIC DISTORTION AND CARRIER-TO-INTERMODULATION RATIO ................................... 14
FIGURE 2.1 0: GENERIC SINGLE-ENDED PA. ............................................................................................... 17
FIGURE 2.11: CLASSA COLLECTOR CURRENT AND VOLTAGE WAVEFORMS .............................................. 18 FIGURE 2.12: CLASS B COLLECTOR CURRENT AND VOL TAGE WAVEFORMS .............................................. 19
FIGURE 2.13: (A) TRANSFORMER-COUPLED PUSH-PULL PA, (B) COMPLEMENTARY PA ...............•........... 20
FIGURE 2.14: CLASS C COLLECTOR CURRENT AND VOL TAGE WAVEFORMS .............................................. 21
FIGURE 2.15: CLASS F COLLECTOR CURRENT AND VOL TAGE WAVEFORMS ....................•.......................... 22
FIGURE 2.16: CLASS E PA CIRCUIT. ··························································································•·············•·· 23 FIGURE 2.17: CLASSE COLLECTOR CURRENT AND VOL TAGE WAVEFORMS .............................................. 23
FIGURE 2.18: GLASS AB COLLECTOR CURRENT WAVEFORM ...........................•.........................••......•.•..... 25 FIGURE 2.19: CLASS AB COLLECTOR VOL TAGE WAVEFORM .....................................•....•....•..................•... 25 FIGURE 2.20: THE GENERAL MATCHING T-NETWORK. ...............................•..........................••..................... 30
FIGURE 2.21: EXAMPLE OF LOAD-PULL CONTOURS FOR A GIVEN PA. ....................................................... 32
FIGURE 3.22: ONE-STAGE POWER AMPLIFIER WITH ACTIVE BIASING ......................................................... 35
FIGURE 3.23: ACTIVE BIASING CIRCUIT ....................................................................................................... 36
FIGURE 3.24: BIAS NETWORK WITH THE POSSIBILITY OF A THERMAL RUNAWAY ........................................ 37
FIGURE 3.25: CRITICAL RF PATHS IN THE PA ..................................................................................•.....•... 39
FIGURE 3.26: INITIAL PACKAGE PIN MODEL. ..........................•.•.•..........................••.........•.......•.......•........... 41
FIGURE 3.27: TwO-COMPONENT MATCHING NETWORKS: (A) LOW-PASS AND (B) HIGH-PASS ...................•...•. 42
V Ill
FIGURE 3.28: SINGLE-TRANSISTOR PA LOAD-PULL CIRCLES FOR MAXIMUM Pour······································ 43 FIGURE 3.29: SINGLE-TRANSISTOR PA OUTPUT POWER CHARACTERISTICS VS. INPUT POWER ............... 44 FIGURE 3.30: SINGLE-TRANSISTOR PA POWER GAIN .........................................•.•................•.................... 45 FIGURE 3.31: BIASING SCHEMES FOR PARALLEL TRANSISTORS ................................................................. 46
FIGURE 3.32: 0FF-CHIP POWER-COMBINING .............................................................................................. 47 FIGURE 3.33: ELECTROSTATIC DISCHARGE PROTECTION CIRCUIT ...............................•....••....................... 48 FIGURE 3.34: SEPARATION AND ISOLATION OF RF SIGNALS TO MINIMIZE INTERFERENCE ........................ 49 FIGURE 3.35: LAYOUT VIEW OF A BOND-PAO WITH ESD PROTECTION ......................................•••.•............ 50 FIGURE 3.36: FINAL PA LAYOUT .............................................................•.......................•..............•............ 51 FIGURE 3.37: EXTRACTED PA GAIN AT 10 = 10 MA. .................•.............................•..............................•.... 51
FIGURE 3.38: EXTRACTED PA POWER-ADDED-EFFICIENCY, PAE, AT 10 = 10 MA ....•...........•...•......•.•.•..... 52 FIGURE 3.39: THE SETUP FOR DEVELOPING EMPIRICAL MODELS FOR THE PCB TRACK, THE BONDWIRE,
AND THE PACKAGE PIN ........................................................................................................................ 53 FIGURE 3.40: BONDWIRE MODEL FOR A 1 MM OF BONDWIRE .......................................................•............. 54
FIGURE 3.41: A SINGLE PACKAGE PIN MODEL. .........................................•......•....•..............•••................•.... 54
FIGURE 3.42: PCB LINE MODEL PER 50 MILS (0.127 CM) ..............................................................••.......... 55 FIGURE 3.43: S21 RESPONSE OF THE PA WITH THE NEW AND MORE ACCURATE MODELS IN PLACE ........ 55 FIGURE 4.44: TWO-PLATE ELECTROSTATICALLY ACTUATED MEMS VARACTOR ....................................... 58 FIGURE 4.45: TYPE Il PARALLEL-PLATE MEMS VARACTOR WITH WIDE TUNING RANGE ............................ 60 FIGURE4.46: TYPE Ill PARALLEL-PLATE MEMS VARACTORWITH VERYWIDE TUNING RANGE ..•.......••..... 61 FIGURE 4.47: INTERDIGITATED LATERAL MEMS VARACTOR. ..................................................................... 62 FIGURE 4.48: TOP VIEW OF THE SEMI-FRACTAL MEMS VARACTOR .......................................................... 63 FIGURE 4.49: 0VERVIEW OF THE METAL-MUMPS STRUCTURAL LAYERS ...............•....•............................ 64 FIGURE 4.50: TOP VIEW OF THE VARACTOR DESIGNED IN METAL-MUMPS .............................................. 64 FIGURE 4.51: FUNCTIONAL MODEL OF THE DESIGNED TWO-PLATE WIDE RANGE VARACTORS .................. 65 FIGURE 4.52: T-TYPE SUSPENSION AND THE EQUIVALENT SPRING MODEL. ............................................... 67
FIGURE 4.53: THE GSG SIGNAL PAO FOR ON-CHIP PROBING .....................•.•.........•..........•.....•..•.......••...... 68
FIGURE 4.54: THE CROSS SECTION AL VIEW OF THE RF PAO IN METAL-MUMPS .......................................... 68 FIGURE 5.55: THE PA DIE ........................................................•.................................................................. 69 FIGURE 5.56: (A) THE PA DIE PLACEMENT IN THE PACKAGE, (B) THE TOP VIEW OF THE PCB CONTAINING
THE STAND-ALONE PA ........................................................................................................................ 70 FIGURE 5.57: THE MEASURED S11 AND S21 FOR THE UN-MATCHED STAND-ALONE PA. ............................ 71 FIGURE 5.58: SIMULATED S11 AND S21 FOR THE UN-MATCHED STAND-ALONE PA .................................... 72 FIGURE 5.59: MEASURED S11 AND S21 FOR THE STAND-ALONE PA ........................................................... 73 FIGURE 5.60: THE MEMS VARACTORS DIE ................................................................................................ 74 FIGURE 5.61: THE MEMS SPIRAL-ARMS 1 PF NOMINAL CAPACITOR ......................................................... 75 FIGURE 5.62: CAPACITANCE OF SPIRAL-ARMS 1 PF NOMINAL CAPACITOR AS A FUNCTION OF FREQUENCY
AT0VAND35V .....................................................................•........•.................•....•.•..•••.•.................. 75
FIGURE 5.63: CAPACITANCE OF THE SPIRAL-ARMS 1 PF NOMINAL CAPACITOR AS A FUNCTION OF
ACTUATION VOLTAGES AT 630 MHz AND 1.88 GHz ...........•...•........•.•....•....................•.......•............ 76
FIGURE 5.64: FUNCTIONAL MODEL OF THE TWO-PLATE WIDE RANGE VARACTORS WITH TUNING STEPS .. 77 FIGURE 5.65: THE PACKAGED MEMS-TUNABLE POWER AMPLIFIER. ......................................................... 78
FIGURE 5.66: THE MEASURED S11 AND S21 FOR THE UN-MATCHED MEMS-TUNABLE PA. .....•........•........ 79 FIGURE 5.67: THE MEASURED 811 AND 821 FOR THE MATCHED MEMS-TUNABLE PAS .................................. 80 FIGURE A.68: THE BLACK BOX AND THE EMPIRICAL RLC MODEL FOR CHARACTERIZING THE PACKAGE PIN,
THE BONDWIRES AND THE PCB TRACES ............................................................................................ 93 FIGUREA.69: THE SIMULATION SETUP FOR THE MODEL FITTING SCHEME ................................................. 93
IX
List of Tables
TABLE 2.1: EFFICIENCY OF GLASS F PA WITH RESPECT TO THE NUMBER OF HARMONICS [15]. ............... 22
TABLE 2.2: PA LINEARITY AND EFFICIENCY TRADEOFFS .....................................•...........•........................... 24
TABLE 3.3: THE CURRENT CAPABILITY OF SIGE TRANSISTORS .................................................................. 39
TABLE 3.4 POWER GAIN AND PAE AS FUNCTION OF la··············································································· 52 TABLE 4.5: CAPACITOR PLATE DIMENSIONS ................................................................................................ 66
TABLE 4.6: CAPACITOR SUSPENDING ARM DIMENSIONS ............................................................................. 67
TABLE 5.7: MEASUREMENT RESULTS FOR MEMS VARACTORS AT 1.88 GHz ........................................... 77
x
/"'"',
Chapter 1 : Introduction
1 Introduction
A power amplifier converts the radio frequency, RF, input power and DC power from the
supply into RF or microwave output power. Figure 1.1 shows a simplified wireless
transmitter architecture. The precise frequency of the output signal is set by the voltage
controlled oscillator, VCO, and the phase-locked loop, PLL. The power amplifier, PA,
then amplifies the desired radio frequency signal to a specified power level before it is
radiated by the antenna. A matching network, MN, is used to set the output impedance of
the ,pA for maximum efficiency or power transfer. The application of power amplifiers
are not limited to wireless communications; they also find widespread use in radar, RF
heating, plasma generation, laser driving circuiting, magnetic resonance imaging, and
miniature DC-DC conversion [1], [2].
r---------, 1 1
1 1
1 1
1 1 1
1
1
Figure 1.1: Simplified transmitter architecture.
1~
Chapter 1: Introduction
Biasing Circuit
V cc
RFC Output
Matching Network
Transistor
Figure 1.2: Generic single-ended PA highlighting sections that can be integrated as MEMS-based components.
The emerging micro-electromechanical systems technology, MEMS, offers the
possibility of improving the RF integrated circuit performance in terms of size, linearity,
and power consumption through simple surface micromachining techniques. This
technology thus has the potential of overcoming the limitations that are conventionally
inherent with the power amplifier design.
1.1 Motivation
The trend in wireless communication systems design in general is towards more
integration, lower power consumption, and better signal quality. In terms of power
amplifier design, similar trends are also seen. Figure 1.2 demonstrates a generic single
ended power amplifier. Practical PA design involves a great deal of optimization; for this
reason the input and output matching networks are often realized through discrete
components to allow for flexibility. The se components are good candidates to be
implemented by MEMS-based tunable capacitors or inductors. Another possibility is to
use high quality factor MEMS resonators as tho se in [3] and [ 4] in the output resonator
tank in class AB, class For classE amplifiers (to be discussed in detail in chapter 2).
2
Chapter 1: Introduction
5% 1 1
1 ' Urban PDF ------· if ' Suburban PDF -4% _,.....
.tl' \"" ~~~ ,. '\
~ 3% ' ~ ' ;~' ' '\. ::c \ ca
J:l \.
"' v ~ e " ; 2% Il.
;/' / \. 1 ... .;' / ... \ 1 ~ '" ..... 1%
/ ....... ~/ ""~
/ ....... ..........
. 20 ·10 0 10 20 30 Pout (dBm)
Figure 1.3: Power amplifier output power probability density function for IS-95 urban and suburban environments [5], [6].
Since the power amplifier is the most power-consuming deviee in the whole
receiver/transmitter chain improving the efficiency of the PA can significantly increase
the battery life and therefore the talk time. In the power amplifiers used in wireless
communication, the required level of output power varies depending on the distance from
the base station. The efficiency of a power amplifier is usually optimized to be maximum
at the peak output level and decreases significantly for lower power levels. However, as
seen in Figure 1.3, the wireless transmitter is required to transmit at the peak level only
during a short period of the operating time. On average, most of the transmitted signais
are at much lower power levels [5], [6]. Consequently, to preserve the battery life and
increase the talk time, it is desirable to be able to keep the efficiency constant for various
output levels [7]. One way to achieve this is by using variable matching networks to
obtain a better average power consumption.
3
Chapter 1 : Introduction
1.2 Litera tu re Review
The output power level in a PA can be adjusted by varying either the lev el of the RF
drive or the quiescent bias point of the power transistor. The idea of actively tuning the
output and, to a lesser degree, the input impedances of a power amplifier is not new.
However, only a few attempts have been made in this regard. Semiconductor-based
junction varactors have been used in the output matching network to maintain either
constant efficiency over a given output power range, or constant power over a limited
frequency range [8], [9], [10], [11]. Other variations include schemes where the
efficiency is kept constant by switching arrays of variable transformers or FET transistors
[12], [13].
In [9] a power-adaptive Class E power amplifier, based on high-Q semiconductor
varactors, is implemented in 0.6-flm CMOS technology. Figure 1.4 shows the proposed
class-E PA with the MOSFET transistor acting as a switch which tums on and off at the
input frequency. L 0 and Co are set to re sonate at the input frequency th us passing a
sinusoïdal current to the Joad RL. When the Joad value is Jess than the optimal Joad value
required for maximum output power, the output power and Power-added-efficiency
decrease sharply (see section 2.31).
L=r=---T RL
I Cp Junction Va~ ! 1 Vc1 Vc2 -="
Figure 1.4: The power-controllable class-E PA implemented in [9] with tunable output
capacitors.
4
Chapter 1 : Introduction
50
0+-------~----~~~~--~--------~ 467 ~~8 38<' 3'15 309 276 249 224 203 16$ 170 T$6 14) 132 123 114 106 il8.
6
Figure 1.5: Simulated results of power-added-efficiency versus output power for the power-controllable PA proposed in [9].
To overcome this problem, a three-element n-matching network is proposed in [9]
consisting of two tunable varactors and an inductor to transfer the 50 n extemalload into
an optimal load required at the drain of the power transistor for maximum power-added
efficiency; thus ensuring maximum power transfer. The tunable capacitors are
implemented with semiconductor junction varactors. In order to ensure that the varactors
are reverse-biased the voltages Vc1 and Vc2 should be very high, for which purpose
"charge pumps" are needed. The simulated results presented in [9] are shown in Figure
1.5: It is clear that the power-added-efficiency has been kept relatively constant over an
output power range of 370 mW. It should be noted that the frequency of operation is not
specified by the authors in [9].
In [10], F. H. Raab presents a 20-W classE power amplifier at 31 MHz with MOSFET
transistors operating form 20 V power supplies. As will be discussed in detaillater, class
E operation requires a certain drain-shunt susceptance and load-series reactance at every
given frequency for correct operation. As shown in Figure 1.6, the output tuning network
developed in [10] employs fixed inductors, L2B and L3, and high-voltage MOSFET
varactors, C4 and C7 to obtain constant output power over the 19 to 31 MHz frequency
('.. range.
5
Chapter 1: Introduction
PATENT PDI>ING
L2A ltF T2 OUTPUT
flF DRIVE
-..
T1 f 1
Il
Figure 1.6: Simplified circuit of electronically tunable classE PA developed in [10].
In recent years, sorne attempts have been made to utilize the MEMS technology in the PA
design context. In [14] MEMS switches are used in the output matching network of a
class AB power amplifier to select one of the 16 output matching networks in the 8- 12
GHz frequency range, Ml-M4 in Figure 1.7. However, the design depends on
semiconductor-based voltage-controlled varactors to fine tune the impedance. By tuming
the MEMS switches on and off and varying the bias point of the varactor, the frequency
response of the amplifier is optimally configured for one specifie frequency as seen in
Figure 1.8. The design, however, requires extra circuitry to determine which switches
should be tumed on at any given time.
mlin
aractor
Figure 1.7: Schematic of the output matching network with four MEMS switches (M1-M4) and one [semiconductor] varactor. C1-C8 are fixed capacitors. mlin is a microstrip
transmission line [14].
6
Chapter 1: Introduction
30T-------------------------------------~
25
.. ... '
-20 e Dl '-8.5GH:z :E.. 15
8 0.. 10
' . 5 ' ••
' 1
1 , /
\.· 0+-~----~r-------~-----'---T----~--~
- -9.5GHz .... •10.5GHz
- • 11.5GHz • Envelope
B 9 10 Frequency (GHz)
11 12
Figure 1.8: Results obtained by having tunable matching networks in [14]. The dots represent the optimized output power as a function of frequency. Each curve represents the frequency response of the amplifier wh en it is optimally configured for one specifie
frequency [14].
1.3 Objectives
As seen in the literature review, no work has been done so far to implement tuning of
power amplifiers by means of MEMS varactors. The goal of this research is to use
MEMS varactors designed in the Metal-MUMPS process to tune the output impedance of
a class AB SiGe power amplifier such that the efficiency is kept constant over a range of
power levels. The best class of power amplifiers that best demonstrates the relationship
between input and output power, as well as between the efficiency and linearity, is the
class AB PA (re fer to chapter 2). A variable biasing network is included in the design to
vary the quiescent current of the power transistor. The designed MEM tunable capacitors
can be reconfigured through a control voltage and offer a very wide tuning range over a
wide frequency spectrum.
1.4 Thesis Outline
The thesis begins with an overview of wireless transmitter architectures with special
attention to the power amplifier and its applications. The need and motivation for power
amplifiers with constant efficiency and/or linearity are presented. An overview of the
7
Chapter 1 : Introduction
micro-electromechanical deviees and their possible applications m power amplifier
design follows.
In chapter 2, the principles of power amplifier operation are presented. The metrics that
are conventionally used to characterize the be havi or of the PAs, such as linearity and
efficiency, are defined. The inherent tradeoffs between the linearity and efficiency are
also discussed. Next, the different power amplifier operating classes, namely class A
through class E, are presented. Relevant to the topic of this thesis, class AB architecture
is studied in more detail. Design techniques and issues specifie to high power amplifier
design such as large-signal S-parameters, power matching, and load-pull techniques are
presented in detail.
Chapter 3 deals with the design of power amplifiers. This includes the design of the
power core, biasing circuits, and input and output matching networks. Paralleling
multiple transistors in order to obtain higher gain and related problems such as power
combining and splitting are discussed next. Load-pull technique is revisited in the context
of practical PA design. Empirical models are developed for the printed circuit board
(PCB) tracks, the bond-wires, and the package ali of which contribute losses to the
circuit.
Chapter 4 is dedicated to radio frequency micro-electromechanical capacitors. The
operating princip les of different varactor structures such as parallel-plate and lateral inter
digitated are studied. Next, an overview of the Metal-MUMPS process is given. Several
MEMS varactors are designed taking advantage of the two structural layers available in
this technology. Related issues such as the suspension arms and signal pads are visited.
In chapter 5, the fabricated deviees are tested. First, the stand-alone power amplifier is
tested and the required input and output matching networks for maximum output power
are designed. Secondly, the behaviors of the MEMS varactors are established through on
probe testing. Next, the SiGe power amplifier and the MEMS varactors are mounted in
8
Chapter 1 : Introduction
the same package. The overall PA is tested and the measured results are compared with
the simulation results.
Chapter 6 consists of a summary of the work done in this research and future work that
can be done to improve upon this project.
1.5 Contributions
The contributions ofthis thesis are summarized as follows:
1. This thesis, in addition to offering a summary of different classes of power amplifiers,
provides a detailed analysis of class AB power amplifier operation. More specifically, the
chip implementation of a 2.2 V class AB PA in a SiGe BiCMOS 5HP technology. Due to
the parasitic !osses associated with the package and the bonding wire, the measured gain
is lower than designed at the PCS band center frequency of 1.88 GHz.
2. Methodically describing the different modes of operation of MEMS variable
capacitors, and providing a design methodology for implementing wide tuning-range
varactors. Based on the mentioned methodology, various MEMS tunable capacitors are
successfully implemented in the Metal-MUMPS process. The wide- and very wide
tuning-range capacitors are successfully tested from DC up to 6 GHz.
3. An attempt is made to integrate the power amplifier chip and the MEMS die in the
same package to pro vide active tuning of the power amplifier matching network in order
to keep the efficiency of the PA constant for different input power levels and load
conditions.
9
Chapter 2: Power Amplifier
2 Power Amplifiers
In this chapter, the basic principles of power amplifier operation are discussed. The
power amplifier is one of the most important components in transmitters design. In a
conventional transmitter, the digital bit stream, after going through the required digital
signal processing, is converted to an analog signal which is then up-converted to the
desired frequency. The radio frequency, RF, signal is then amplified to the specified
power level through the power amplifier to be radiated by the antenna. Thus, the power
amplifier, PA, can simply be defined as the circuit that converts the DC input power into
RF (or microwave) output power. The applications of PA are not limited to wireless
communication; they also finds widespread use in radar, RF heating, plasma generation,
laser driving, magnetic resonance imaging, and miniature DC-DC conversion [15], [16].
Power amplifiers can be incorporated into transmitters in different architectures including
linear, Kahn, envelope tracking, outphasing, Doherty, and so on [16]. The output power
lev el of the PA is determined by the communications system specifications. The outgoing
signal should be high enough for the receiver to sense and recover it after ali the losses
encountered in the propagation path. For base-station applications, the transmitted power
needs to be in the order of hundreds of watts. For mobile wireless communications, this
value varies between hundreds ofmilliwatts, mW, to a few watts. In closer-range wireless
applications, such as the Bluetooth, powers in orders of tens to hundreds of milliwatts
r" suffice.
10
Chapter 2: Power Amplifier
In section 2.1 the metrics that are used to characterize and quantify the performance of a
PA, such as efficiency and linearity, are defined. Unless the process of converting the DC
power from the battery to the RF power delivered to the load is lossless, the PA itself
becomes a large consumer of power, often more than what it delivers to the load. This
results in power amplifiers being the primary consumers of DC power in any transmitter
chain, hence the great emphasis on improving their efficiencies [17].
Next, different power amplifier architectures and operating classes, e.g. A through E, are
briefly discussed; since a class AB power amplifier is investigated in this research, more
emphasis is put on this specifie mode of operation. Finally, design issues that are specifie
to PA design su ch as large-signal S-parameters and the load-pull technique are presented
in detail.
2.1 PA Metric Definitions
In arder to be able to discuss the characteristics of power amplifiers comprehensively, to
classify the PA operating modes, and to make meaningful comparisons, a set of uniform
parameters must be defined. The basic set of parameters used for characterizing power
amplifiers are the output power, power gain, and efficiency. The linearity and stability of
the amplifier should also be taken into consideration. Each of these parameters is defined
in this section along with guidelines on measuring them in a PA circuit.
2. 1.1 Output Power and Power Gain
The power amplifier output power is generally specified in units of dBm, which is the
output power in dB with respect to 1 rn W. In other words,
P = lOlog Pw dBm 0.001 Watt
(Equation 2.1)
where Pw is the power in watts. So 1 W is equivalent to 30dBm, 100 rn W is equivalent
to 20 dBm, and so on.
11
Chapter 2: Power Amplifier
Another metric often considered in PA design is the power output capability (or transistor
utilization factor) which is defined as the output power per transistor, normalized for peak
collector voltage and current of 1 V and 1 A, respectively [18].
The power gain is defined as the ratio of the source, or input, power to the load, or output,
power. When expressed in dB it becomes:
Gain (dB) = Pout (dB) - P;n (dB) (Equation 2.2).
For sorne applications, the gain is specified at the 1-dB compression point and is referred
to as the "compressed" gain [3].
2.1.2 Efficiency
2.1.2.1 Collector Efficiency
The DC power in portable applications cornes from the battery, a source with a finite
available power; bence, the power consumed in the PA is the main source of degrading
the battery life. Collector efficiency, 1f, is one of the metrics that quantify how efficiently
the DC power is converted to RF power:
(Equation 2.3),
where PoutRF is the RF output power and Pvc is the input DC power. The limitation of
collector efficiency (or in the case of MOS transistors drain efficiency) as a metric is that
it does not take into account the RF power delivered to the deviee from the source. As
most high frequency power amplifiers have rather low gain, the collector efficiency can
overrate the actual efficiency of the power module [ 17].
2.1.2.2 Power-Added-Efficiency
The power-added-efficiency, P AE, is a better indicator of the power amplifier' s
efficiency, since it takes into account the RF input power, P;nRF, by subtracting it from
the output power. The PAE is defined as:
p -P p AE = out RF inRF (Equation 2.4).
PDC
12
Chapter 2: Power Amplifier
Rearranging the components, the above relation can be rewritten in terms of the power
gain of the amplifier, G, and the collector efficiency, 11 [17]:
1 PAE = (1- G)IJ (Equation 2.5).
The power-added-efficiency is a reasonable measure of PA performance wh en the gain is
high; it can, however, become negative when the gain of the amplifier is low.
An even more accurate indicator of efficiency that can be used under ali circumstances is
the overall efficiency, defined as:
IJoverall = p pout; (Equation 2.6). DC + inRF
The overall efficiency can be adjusted to take into account the power consumed by the
driver stage and the biasing circuitry as weiL In order to conserve the battery power and
avoid interference with the signais from other transmitters in the same frequency band,
the peak amplitude of the transmitter signais need to be kept below the rated "peak output
power" of the transmitter by about 10 to 20 dB; The peak output power is only needed
for the worst-case links [15].
2. 1.3 Linearity
Linearity, like efficiency, is a key factor in power amplifier design. Nonlinearities can
impair the replication of the amplified signal, resulting in distortion. Depending on the
specifie application, linearity may be the most crucial parameter that defines the way a
certain power amplifier is designed. In short, the constraints on how linear or undistorted
the output signal should be can very weil define which class of amplifier is employed.
Linear amplification is required when the input RF signal contains both amplitude and
phase modulation. Examples include single side band (SSB) voice, modem shaped-pulse
data modulation schemes (QAM, QPSK, CDMA) and multiple carriers (OFDM or
orthogonal frequency-division mulitplex). Signais such as continuous wave (CW), FM,
classical FSK and GMSK, on the other hand, have constant envelopes (amplitudes) and
therefore do not require linear amplifications [15], [19].
13
Chapter 2: Power Amplifier
2.1.3.1 Sources of Nonlinearity
Intermodulation distortion is a major source of distortion in power amplifiers and can
cause the output amplitude to differ in shape from the input signal. When two or more
signais with different frequencies are applied to a nonlinear amplifier, the output will
contain additional frequency components called intermodulation products. The
generation of spurious frequencies and intermodulation distortion can lead to serious
problems in multi-carrier wireless transmitter systems, where the spurious signais can
appear in the adjacent channels and contaminate them. This sort of nonlinearity is mainly
caused by the variable gain or saturation of the RF transistors [16], [17].
Another source of signal distortion is the nonlinear phase characteristics of the amplifier.
Ideally, in arder to have no distortion, the power gain transfer function should be constant
as a function of frequency. In practical applications however, both linear and nonlinear
phase shifts are present. A linear phase shift produces a constant time delay at the signal
frequency, whereas a nonlinear phase shift produces different time de1ays for different
frequencies, thus introducing unwanted phase modulation. Amplitude-ta-phase
conversion can arise from a voltage-controlled capacitance that the designer has no
control over [16], [20], [21].
~ Il. .. .,..
' ' ' ' : IMD
' ' ' Second Harmonie ' --
Third Harmonie
f 2f1-f2 f1 f2 2f2-f1 Frequency
F1gure 2.9: Harmomc d1stort1on and carner-to-mtermodulat1on rat1o.
14
Chapter 2: Power Amplifier
2.1.3.2 Measures of Linearity
Based on the specifie signal and application, linearity is characterized and measured by
various techniques. Sorne of such techniques are carrier-to-intermodulation ratio, noise
power ratio, and adjacent-channel power ratio.
Carrier-to-intermodulation ratio, C/I, is the traditional measure of linearity and compares
the amplitude output at the desired frequency to the intermodulation-distortion products.
For this purpose, the power amplifier is driven with two tones of equal amplitude placed
at± ~f around the fundamental frequency. Nonlinearities create intermodulation products
at frequencies corresponding to the sums and differences of multiples of the carrier
frequency. The amplitude of the third order intermodulation distortion product, IMD, is
compared to the signal amplitude at the fundamental to obtain the C/I (Figure 2.9). A
typical IMD value for linear PA's is -30 dBc or better.
Noise-power ratio, NPR, is a method of measuring the linearity of a power amplifier for
noise-like and broadband signais. In this case, the PA is driven with Gaussian noise with
a notch in one part of its spectrum. Nonlinearities cause power to appear in the notch;
NPR is then defined as the ratio of the notch power to the total power.
Adjacent-channel power ratio, ACPR, is the most widely used measure of linearity in
shaped-pulse digital signais such as the NADC (North American digital cellular) and the
CDMA (code-division multiple access). ACPR characterizes how the nonlinearities affect
the adjacent channel by comparing the power in a specified band outside the signal
bandwidth to the power in the carrier signal [15], [16], [19], [22], [23].
2. 1.4 Stability
The stability ofPA's should also be considered. As will be seen in chapter 3, the parasitic
inductances and capacitances may form a resonant tank and cause undesired oscillations.
In order to investigate the small-signal stability of an amplifier, the whole circuit,
including the input and output matching networks, is considered as one black box and its
15
'
/~ 1
Chapter 2: Power Amplifier
stability is verified for a 50 .n input and output load. Based on the S-parameters, the k
factor is defined; if the k-factor is larger than unity, at the specifie bias and frequency, the
amplifier is stable:
k __ 1-jS11j
2 -jS22j
2 + D2
21821Jjst
2J (Equation 2.7),
and
D = S11S22-S12S21 (Equation 2.8).
In order to verify the large-signal stability of a PA, the circuit is excited with an RF pulse
and the transient response of the PA is verified. If the response is free of ringing, the
power amplifier is stable [1], [17], [18].
2.2 Choice of Architecture
Power amplifiers are divided into different classes of operation, based on where the
deviee is biased and whether the deviee is operated as a switch or not. The design
constraints on linearity and efficiency de fine the PA architecture employed. A rough
grouping of power amplifiers can be done based on the linearity of the output signal of
the power module. According to this classification, class A, B, and AB fall under linear
amplifiers, whereas class C, F, and E fall un der nonlinear power amplifiers. A variety of
other modes of operation also exist that are a combination of the above mentioned modes.
In this chapter, the operation of each of the above classes are described. Since the goal of
this research is to design a linear power amplifier with tuning capabilities, class AB
operation is investigated in more detail.
16
V cc
RFC
Chapter 2: Power Amplifier
Output Matching Network
Transistor
Figure 2.10: Generic single-ended PA.
2.2.1 Linear Amplifiers
2.2.1.1 Class A
In class A operation, the power transistor is in the active region during the whole RF
cycle. This is achieved through biasing the transistor such that the quiescent current is
half of the maximum current of the transistor. In other words,
1 _]max Q-
2 (Equation 2.9),
where IQ is the quiescent current and IMAX is the maximum collector current that the
transistor can sustain. Due to the biasing, both the positive and negative swings of the
input signal affect the collector current, and therefore class A provides the highest gain of
ali types of power amplifiers.
As shown in Figure 2.11, the collector current and voltage waveforms are perfect
sinusoids, hence the linear amplification in this mode. The output power delivered to the
load is vo:t 1 2R, where Vout is the voltage on the load and R is the load resistance.
17
Chapter 2: Power Amplifier
The transistor acts as a voltage-controlled current source: the collector current is
controlled by the base-emitter voltage. Linear amplification implies that increasing the
quiescent current or decreasing the RF input signal level will decrease the harmonies and
the intermodulation distortion. The low harmonie content of this mode of operation
means that it can be used at frequencies close to the maximum operating frequency,
F MAX, of the transistor.
The maximum theoretical efficiency of class A is 50 %. The efficiency of practical class
A amplifiers is even lower due to the presence of the on-resistance of the transistor, the
saturation voltage of the transistor, and the parasitic losses. It is further degraded by the
fact that practical loads are not solely resistive and contain reactive components too; in
essence, more output current or voltage needs to be extracted from the PA to de li ver the
same output power to the load. In short, due to the inherent characteristics of class A
amplifiers, they are often used in high frequency applications where high linearity and
gainarerequired [2], [15], [17], [18], [20], [24], [25].
Figure 2.11: Class A collector eurre nt and voltage waveforms.
18
Chapter 2: Power Amplifier
Figure 2.12: Class B collector current and voltage waveforms.
2.2.1.2 Class B
In class B operation, the base of the transistor is set at the threshold of cutoff such that
transistor is on during half of the RF cycle, resulting in a collector current that is a half
sine wave. For this purpose, the quiescent current is set to
J _]max Q- ;r (Equation 2.1 0).
The collector current and voltage waveforms are shown in Figure 2.12. Since the
amplitude of the collector current is proportional to the amplitude of the RF input signal,
the shape of the collector current waveform is fixed and the amplification is linear. The
output power is th us controlled by the RF drive lev el and varies as vo:, 1 2R . The
efficiency varies linearly with the RF output voltage, and can ideally reach ;r 14 ( ~
78.5%) at the peak output power.
Class B is generally used in a push-pull configuration to provide amplification over the
entire cycle. Due to the limitations of RF p-type power transistors, the use of
complementary topologies is limited to audio, low, and medium frequency. At HF and
VHF frequencies, a transformer coupled push-pull configuration is used to allow
broadband operation with minimum filtering. Transformer-coupled push-pull and
19
.r---\
Chapter 2: Power Amplifier
complementary PA configurations are shown in Figure 2.13 [2], [15], [18], [20], [24],
[25], [26].
V cc
V cc
Q1 Output Filter
RL ~ '"'v
Q2 RL
(b) -
- -Figure 2.13: (a) Transformer-coupled push-pull PA, (b) complementary PA.
2.2.1.3 Class AB
The ideal class AB amplifier is biased between class A and class B, as a re suit of which
the collector current can swing between zero and IMAX· The efficiency is between that of
class A and class AB and increases as the quiescent current is decreased. The class AB
operation is discussed in detail in section 2.2.4.
2.2.2 Nonlinear Amplifier
2.2.2.1 Class C
The base of the transistor in a class-C PA is biased near the cutoff for more than half of
the RF cycle. The transistor is cutoff un til the RF signal applied between the base and the
emitter makes it conduct. A high-Q parallel-tuned resonant circuit is usually used at the
output stage to recover the signal at the fundamental frequency and suppress the
harmonies. Class C is extremely nonlinear, but its efficiency can theoretically be
increased toward 100% by decreasing the conduction angle toward zero. This, however,
20
Chapter 2: Power Amplifier
causes the output power to decrease toward zero and the input drive to increase toward
infinity. A typical compromise is a conduction angle of 150° and ideal efficiency of 85%.
Examples of collector current and voltage waveforms are shown in Figure 2.14.
In case of MOS transistors, when the transistor is driven into saturation, the output
voltage is locked to the supply voltage and the efficiency is stabilized, thus allowing
linear high-level amplitude modulation. Class C is widely used in vacuum tube
transmitters but is impractical for sol id state PAs. The main reason is that class C requires
low collector (drain) resistance, making the implementation of parallel-tuned output filter
difficult. lt is also difficult, especially in case of bi polar transistors, to set up the bias and
the drive to produce a true class C waveform [2], [15], [18], [20], [24], [27].
Figure 2.14: Class C collector current and voltage waveforms.
2.2.2.2 Class F
Class F takes advantage of harmonie resonators at the output to boost both efficiency and
output power by reshaping the collector waveforms. This is done through a series and a
shunt resonator tank. The series tank is tuned to the third harmonie, while the shunt tank
is tuned to the first harmonie. The resultant output voltage contains one or more odd
order harmonies and approximates a square wave, while the output current contains even-
21
Chapter 2: Power Amplifier
arder harmonies and approximates a half sine wave. As seen in Table 2.1, as the number
of the harmonies increases, the efficiency increases from 50% to unity. The class F
current and voltage waveforms are shawn in Figure 2.9 [2], [15], [18], [27], [28], [29],
[30].
Table 2.1: Efficiency of class F PA with respect to the number of harmonies [15].
Number of Harmonies Efficiency (%)
2 70.7
3 81.65
4 86.56
5 90.45
3~~----~------~----~----~~
2.5 ------+-···········
i Vc 2 ------ ------------
Figure 2.15: Class F collector current and voltage waveforms.
2.2.2.3 Class E
In class E, a single transistor is operated as a switch that tums on and off at the input
frequency. The series resonator tank is tuned to the first harmonie of the input frequency,
and the collecter voltage waveform is the result of the sum of the DC and RF currents
charging the shunt parasitic capacitance, Cp in Figure 2.16. In the ideal of class E
operation, as the transistor tums on, the collector voltage drops to zero with a zero slope;
the result is an ideal efficiency of 100%, due to reduction of the switching lasses and
r---. elimination of los ses incurred when the parasitic collector capacitance is charged.
22
Chapter 2: Power Amplifier
Optimum class E operation requires a drain susceptance of 0.18361 R and a drain series
resistance of 1.15R, and delivers an output power of 0.577V};D 1 R. The design of power
modules operating in class E requires a driver stage, usually a class F, to generate the
square waves required for the main stage to operate in this mode. Class E operation is
efficient even when a significant parasitic capacitance is present at the drain, which
makes it useful in many applications. Class E high frequency PAs with power levels
close to 1 kW can be implemented using low cost MOSFET technologies. The drain
current and voltage waveforms for a class E PA is presented in Figure 2.17 [2], [27], [28],
[29], [31], [32], [33].
VDD
Figure 2.16: ClassE PA circuit.
'
1 ' 1 • -- - • -1~-- ·- ------- ·r ·-- ·----·---- r ·----- ·------r··--------· ·
' ' ' . • ' 1 • • ' 1 • • ' 1 • 3 1 ' 1 • • ' 1 ' ' ' 1 • ' 1 1 1
• •-- • .1.- •--- • •-- • • .L •-- • •- ·•- •-- • .L- •---- • • • • -••'-•••••••••• • •
' 1 1 1 1 1 1 1 ' 1 1 1 ' 1 • ' ' 1 1 ' 1 1 ' 1
' ' ' ' ' ' ' ' ' • • • • • -,- • • • • •• ·- • • ·-- ---- • • -- • ·- ·- T ·-- • • • • • • • ---r---------2 • 1 • 1
' 1 • ' 1 1 1 1 1 1 ' ' 1 1 1 1
' ' ' ' ' . ' ' 1 1 ' 1 ----- -·------------- ... --- --------"---------- --- -----------' ' ' ' . ' ' . ' ' ' ' . . ' ' . ' ' ' ' . ' ' . . . -.- ... -,-.--.---.--- .. ,. - ... --- ---.- ---r------------' ' ' ' ' ' ' ' ' ' ' ' . '
0 pif2 3*pif2 2*pi
Figure 2.17: Class E collector current and voltage waveforms.
23
Chapter 2: Power Amplifier
2.2.3 High-Linearity and High-Efficiency RF PA - Tradeoffs
In power amplifier design, linearity and efficiency are inversely related to each other.
ClassA operation offers extremely high linearity at the expense of low efficiency. Class
B has low distortion and the potential for good efficiency. Class AB enjoys the good
efficiency of class B with the low distortion of class A.
Switched-transistor PAs, such as class F and E, are very good in terms of efficiency but
can be extremely nonlinear. The linearity-efficiency tradeoffs are summarized in Table
2.2.
Table 2.2: PA linearity and efficiency tradeo ff s.
Mode of operation Efficiency Linearity
ClassA Low Very high
Class AB and B Medium High
Class C High Low
Class F High Low
ClassE Very High Very Low
2.2.4 Analysis of C/ass AB Power Amplifiers
As discussed before, in linear power amplifiers the quiescent current determines the
method of operation, and thus the efficiency and power output capability. Class AB has
higher efficiency than ClassA and higher linearity than Class B.
24
Chapter 2: Power Amplifier
Io
····i···----~---···1·····
1 : ' J • 1 ... ,----·--t-·····1··· 1 : 1
••••..• '.J •••••. ,.:. ..•••• J.. : 1 : 1
1 1 1 ... , .......... J.. .. ... : ... ............. l, ......... ...
1 1 : 1 ' 1 : 1 • 1
...... :..... -------~------l- ··--t--··· .t ..... : 1 1 1
·····!······ ··-···+······~ ···--j-···· ~--··· : t : ••••••••••• , .. 1 --- ....... •• • • •• -- - ••••• : 1 1
1 1 ' : 1
... "' ...... - .. : .. "' .......... J •• ' 1
1
1 1 • 1 ï ··---~---····r 1 : 1 1 • 1 1
3rr/2 2'Jr
Figure 2.18: Class AB collecter current waveform.
1 1 1 1 --------- __ ... _____________ , ______________ ,___ _ ________ .__________ --1 1 1 1
' ' ' ' ' ' ' ' ' 1 1 1 1
••••••••••• •T•••••••••••••.,•••••••••••••-o• •••••••••••r"••••••••••• •
' ' ' ' ' '
VQ +----+---P------1---.....,,__---l Vt ' ' _, ______ ------- _._ ____ ---------
' ' ' ' ' --------- ... --------- ___ , ______________ .. ____________ _
' ' ' ' ' ' ' ' ' '
0 rr/2 r:t12 1T 3rrl2 21f
Figure 2.19: Class AB collecter voltage waveform.
25
Chapter 2: Power Amplifier
Based on Figure 2.18, the collector RF current can be written as
l.c --{1Q +1pkcos(8), if -a/2(8(+al2
0 , else (Equation 2.11 ),
where a is the conduction angle. The DC component of the current is obtained by:
1 ___ 1_ af12
1 cos(&)- cos( a 1 2) dn
u (Equation 2.12), DC 27! -a/
2 max 1-COS(a /2)
where ]MAXis the maximum current in the power transistor. Based on Equation 2.12, the
quiescent DC current for class A is 1 max 1 2, and that for class B is 1 max 1 :TT • As a result,
the DC current for class AB operation becomes:
J max (J ( J max
:TT Q 2 (Equation 2.13).
Similar to Equation 2.12, the magnitude of the nth harmonie is given by:
In = .!._ aT 1 max cos(&)- cos( a 1 2) cos( nO) dB :TT -a/ 2 1-cos(a/2)
(Equation 2.14).
Referring to Figure 2.19, VQ is defined as the bias voltage level at the base of the RF
transistor. As the transistor is biased closer to cutoff, higher drive levels are required to
maintain a peak current of ]MAX· For example, as the operation moves from class A
towards class B, the drive level has to be increased by a factor of two in order to get the
same peak current. In other words, a 6 dB increase of input power level is required,
which translates into class B gain reduction [1], [34], [35].
A common solution is to "under-drive" the amplifier or operate the amplifier at lower
drive levels than is needed, in order to keep efficient operation at higher gains (through
IMAX). For example, instead of increasing the input drive by 6 dB from class A to class B,
it can be increased by 3 dB. In this case, at zero bias, the current maximum will
26
Chapter 2: Power Amplifier
be 1m~ instead of IMAX· To offset this reduction in maximum linear power, the load
resistance can be increased by the factor of J2 . Overall, efficiency of 78.5% can be
achieved with only 1.5 dB reduction of gain [34], [36], [37].
2.2.4.1 Harmonie Termination
One of the important elements in a class AB design is the resonator tank at the output. A
truncated sine wave has high harmonie contents; the tank provides a harmonie short that
prevents the generation of harmonie voltages at the output. The shunt capacitor chosen
for the harmonie tank must be large enough to short ali harmonies except the fundamental
to ground. The final output signal will be a sine wave whose amplitude depends on the
RF input signal level, P;n, and load resistance, RL. In practical applications RL is chosen
such that, under the maximum drive, the voltage swing makes use of the full range, i.e.
the amplitude approaches the DC supply.
2.2.4.2 Tunable Load Resistance
Two important issues in RF power amplifier are the rapid drop in efficiency as the input
drive level decreases, and the need to control the power over a wide dynamic range. Both
of which can be solved by using a variable load.
The concept of under-drive mentioned in section 2.2.4 can be applied in this context. In
class AB, the under-drive efficiency is not as high as the fully driven case because the
input drive, P;n, is not as high, according to the power-added-efficiency relation
(Equation 2.4). If the load resistance, RL, can be changed dynamically with P;n, thus the
drop in efficiency, due to the drop in P;n, can be counteracted, enabling control over the
output power over a wide dynamic range.
The voltage and current that appear at RL, at the fundamental frequency, are related
through Ohm's law:
(Equation 2.15),
27
Chapter 2: Power Amplifier
where h is the output current at the fundamental as frequency defined by Equation 2.14.
If RL is made variable as Ro , then as h goes from zero to 1 max , v0 remains constant:
vs 2
v,~(~}· = Ro (/max .vs) (Equation 2.16).
vs 2
=VDC
Assuming an ideal harmonie short, the output power becomes po = VDC 1 max vs .
2 2
Note that, according to the above equation, the output power depends on Vs and 1s
independent of P;n. In which case the efficiency also becomes independent of the input
drive [34]:
pout RF tJ=-
PDC
7&' (VDC )(/max) = VDCVslmax l Z Vs
7&' =
4
2.3 PA Design Techniques
(Equation 2.17).
Power amplifier design differs from other amplifier designs due to the mere presence of
large collector (or drain) signais. The small-signal S-parameters are not sufficient
anymore for defining and characterizing the behavior of the amplifier. Similarly, the
conjugate matching does not provide the best power transistor for large output signais.
These parameters have led to the introduction of large-signal S-parameters and the
development of different load-line matching approaches specifie to PA design.
Small-signal analysis is based on the assumption that the deviee currents and voltages
undergo small fluctuations about constant DC bias conditions. Under this assumption,
approximate linear relationships between deviee currents and voltages can be derived. In
28
Chapter 2: Power Amplifier
large-signal operation, on the other hand, there are large variations in currents and
voltages. Meaning that linear 1-V relationships are not adequate to characterize the
deviees.
2.3.1 Power Match vs. Conjugate Match
Conjugate matching is the matching technique used in ali law-power amplifier designs.
The conjugate match states that, for maximum power transfer, a given load should be
matched to the conjugate of the source impedance. In other words, if the source and load
impedances are given by zsource = Rs + xs and zload =RI+ xl' then
(Equation 2.18).
The conjugate matching technique however, neglects the fact that the voltage source may
have limited physical conditions, especially in terms of the voltage that it can sustain
across its terminais. Conjugate matching for a source of 50 n requires a load of 50 n. If
the voltage source in question is the output of a previous stage amplifier with a current of
1 A, a load of 50 n means that a voltage of 25 V will appear across the output of the
previous stage transistor, which may be well over its physicallimits [1].
In power amplifier design, the load impedance is selected based on the maximum
physical current and voltage of the transistor. In other words, the value of the load
impedance that extracts the maximum power from the transistor without exceeding the
RF voltage swing limit of the transistor or the de supply. The value referred to as Zopt can
be approximated to first arder as:
z = vmax apl J
max
(Equation 2.19).
In actual PA design, the value of Zopt is determined through load-pull techniques to be
discussed in section 2.3.2. It has been shawn that power match gives at least 2 dB
improvement in output power compared to conjugate match [1].
29
Chapter 2: Power Amplifier
Port1 Port2
Figure 2.20: The general matching t-network.
Assuming that the value of Zopt is already known for a given PA, a matching network can
be designed to power match the output impedance of the PA, Zout, to this optimal value
[38]. The matching network is assumed to be a t-network with the reactances shawn in
Figure 2.20.
Referring to Figure 2.20, if Port 2 is terminated with an impedance ZB, the equivalent
impedance looking into Port 1 simply becomes:
Z X x,;
1 = J 11 + -.-_.:::;'----jX22 +Zs
(Equation 2.20).
Similarly, if Port 1 is terminated with an impedance ZA, the equivalent impedance
looking into Port 2 is derived to be:
Z X X(2
2 = J 22 + -.--=-]Xli +ZA
(Equation 2.21).
Separating the resistive and reactive components, Equation 2.20 and Equation 2.21 can
be written as:
(Equation 2.22),
(Equation 2.23),
30
Chapter 2: Power Amplifier
(Equation 2.24),
(Equation 2.25).
Solving Equations 2.22 through 2.24 yields:
(Equation 2.26),
(Equation 2.27),
(Equation 2.28),
where
These equations can be programmed into Excel, where the power-match network can be
easily created once the values of Zopt , Zout , and the frequency are specified.
2.3.2 Load-Pu/1 Technique
The load-pull technique consists of loading the PA with different impedances and
verifying the output power level as function of the impedance to determine the best
output load. A load-pull system consists of a test fixture and a pair of low-loss,
accurately resettable precision mechanical tuners. Data are measured for a large number
of impedances and plotted on a Smith chart. The accuracy of the plots depends on the
accurate calibration of the tuners, both in terms of impedance and losses [1], [34], [39],
[40].
The load impedances that deliver a given RF power level with a specified collector
voltage lie along parallel-resistance lines on the Smith chart. Similarly, the impedances
31
Chapter 2: Power Amplifier
for a specified current level follow series-resistance line. The resultant is that constant
power contours are elliptically shaped rather than being circular as seen in Figure 2.21.
The transformation of the collector impedance through the parasitic collector capacitance,
and the bondwire and package inductances causes the constant-power contours to become
distorted and rotated [15], [17]. It should be noted that the power and efficiency contours
are not necessarily aligned. Similarly, maximum power and maximum efficiency do not
necessarily always occur for the same load impedance. This is especially true at higher
power levels and can be seen in Figure 2.14 as weiL
Maximum PAE Contours
Maximum Pout Contours
Figure 2.21: Example of load-pull contours for a given PA.
32
Chapter 3: Design of Power Amplifier
3 Design of Power Amplifiers
The goal of this research is to design and fabricate a MEMS-tunable RF power amplifier
based on a SiGe technology. In this chapter, the PA design methodology is presented:
First, a power amplifier utilizing a single power transistor is designed along with the
required biasing circuitry. Design aspects related to both operation at RF frequency, such
as bondwire and package parasitics, and specifie to PA design, like the load-pull
technique, are presented in the design context. In order to obtain higher gain, the multi
transistor PA design and the issues related to this topic such as power-splitting and -
combining and layouts are investigated. The be havi or of the complete design is presented
in terms of output power, gain, PAE, and IMD. In the last part on this chapter, empirical
models are developed for the bondwires, the package, and the PCB tracks and the PA
design is revisited based on these models.
3.1 PA Design in a SiGe Technology
3. 1.1 Choice of Tech no/ogy
The RF power amplifier in this research is designed using a SiGe hetrojunction bipolar,
HBT, transistor. Commercial power amplifiers are currently designed and manufactured
in two competing technologies: Gallium Arsenide, GaAs, HBTs and Silicon Germanium,
SiGe, HBTs.
33
Chapter 3: Design of Power Amplifier
Gallium Arsenide is the technology of choice for linear power amplifiers for commercial
handheld wireless communications. GaAs HBTs offer higher power gain, efficiency,
linearity, and breakdown voltages compared to SiGe HBTs. They however have a lower
base-to-emitter tum-on voltage, or threshold voltage, than their SiGe counterparts. This
makes the SiGe a better choice for low-voltage applications, where the supply is limited
to below 3 V. Moreover, SiGe offers a better cost benefit: SiGe dies are fabricated on 8-
inch wafers where as GaAs wafers are usually 4- and 6-inch large. Considering the yield
benefits of larger wafers, the average cost of a SiGe power module can be one half or
even one quarter that of a comparable GaAs module [ 41] - [ 48].
3. 1.2 Power Amplifier Specifications
The goal of this section is to design a linear power amplifier to investigate the possibility
of tuning it using MEM deviees. Since the matching is done off-chip, the linearity and
efficiency of the PA can be adjusted as needed. The PA is designed to used for the
Personal Communications Service, PCS, band. The North American PCS band is defined
in the range of 1.85 GHz to 1.91 GHz for mobile to base applications. The design
frequency in this work is thus set to 1.88 GHz, i.e. in the middle of the PCS band. The
objective is to extract a maximum of 25 dBm of power from the PA with an IMD of -25
dBc or better, and the maximum attainable efficiency. The ultimate goal is to be able to
tune the output load with respect to the input drive level to keep the efficiency at
maximum values.
3.1.3 Single Transistor Design
The initial design goal is to meet the design specifications using a single power transistor.
A one stage amplifier with the biasing network is shown in Figure 3.22.
34
r·\
Chapter 3: Design of Power Amplifier
VcoNr V cc
V cc
RFC
Coecoupling
Figure 3.22: One-stage power amplifier with active biasing.
3.1.3.1 Biasing Circuit Design
The design of the biasing circuit is an important part of the PA design. The performance
of the amplifier cannat be sacrificed by having a poor DC biasing network; the specifie
DC parameter set by this network should be stable over transistor parameters and
temperature variations. In applications where the temperature changes are not significant,
resistor biasing networks are used. However, when the temperature variations become
significant, active biasing becomes essential [20].
The transistor parameters that are affected most by temperature variations are: lcBo
(reverse current), hFE, and VBE· IcBo doubles with every 10°C increase in temperature.
VBE has a negative temperature coefficient; i.e.
(Equation 3.1).
35
' .~
Chapter 3: Design ofPower Amplifier
VcoNT Vcc
RFC
Figure 3.23: Active biasing circuit.
The DC current gain, hFE, is defined as the collector-to-base current at constant VCE; hFE
increases linearly with temperature at a rate of 0.5%/°C [20].
The biasing network shown in Figure 3.23 acts as a variable voltage source with low
resistance. The two back-to-back diodes are used to prevent thermal runaway. In the
absence of the diodes a voltage of approximately 2 VBE, due to Qhias and QRF, appears
across R3 (Figure 3.24). As the temperature increases, 2VBE decreases (according to
Equation 3.1), which translates into an increased IBJ and therefore an increased IB2· The
increase in the base currents increases the temperature even further and ultimately results
in a thermal runaway. The presence of diodes D1 and D2 means that the voltage-drop
across R3, ignoring the drop ac ross R1, is approximately 2 VBE- 2 VBEdiode, which remains
constant with temperature variations, hence preventing thermal runaway.
Another measure that ensures thermal stability is the presence of the resistor R1 in the
emitter of the bias transistor. Any increase in IB2 results in an increase in the voltage-drop
across Rb and therefore reduces IB2 through negative feedback.
36
Chapter 3: Design of Power Amplifier
VcoNr V cc
• • •
RFC
Figure 3.24: Bias network with the possibility of a thermal runaway.
The resistor divider, R:z and R3, is used to present a fraction of the control voltage, VcoNT,
to the base of the bias transistor, thus enabling finer tuning of the voltage. The voltage set
by the bias network, VB in Figure 3.23, appears at the base of the RF power transistor
through the extemal RF choke. The role of the RF choke is to prevent the RF signal from
the source to leak into the biasing circuit. The reactance of the RF choke should be at
least 10 times the value of R in order to have a minimum loading effect on the power
circuit, i.e.
R x <Lo- 10 (Equation 3.2),
where R is 50 n and Xw is the reactance of the choke. Furthermore, the following
relation also holds:
(Equation 3.3),
where Q is the quality factor of the choke. For a Q of 10, the minimum value of the
choke, for it to have negligible effect on the circuit, is 0.42 nH at 1.88 GHz and 0.88 nH
at 900 MHz [49]. In the design presented here a 1 JlH off-chip choke with a self-resonant
frequency of 10 GHz is used.
37
Chapter 3: Design of Power Amplifier
The required resistor values are determined by simulating the bias network in Cadence.
The objective is to provide a collecter current range of 1.25 mA to 80 mA to the power
transistor by extemally varying the control voltage VcoNT· The resistors are realized using
the on-chip resistors available in the SiGe technology. R1 and R3 are both PBD resistors
with values of 50 n and 100 n, respective! y; R2 is a 2 kQ FRD resistor. Both are
polysilicon resistors placed over a deep trench in order to reduce their parasitic
capacitances. The difference between the two is that PBD resistors have a sheet resistance
of 220 Q/o while the sheet resistance of the FRD resistors is higher, namely 1500 Q/o
[2], [35], [50].
3.1.3.2 Power Core Design
The quiescent current in class AB operation is set between that of classA and class B [1],
namely,
Jmax (1 (Jmax re Q 2
(Equation 3.4).
By performing simulations to characterize the DC behavior of the various transistors
available in the BiCMOS 5HP technology, the maximum collecter current for each
transistor is obtained and summarized in Table 3.3. Based on these values and according
to the above relation, the range of the quiescent current for class AB operation is
determined. This technology offers vertical HBT NPN transistors with the emitter width
fixed at 0.5 j..lm, while the length of the emitter can be changed from 2.5 1-1m to 20 j..lm.
The higher emitter length allows the transistor to handle more current, thus the transistor
has a higher individual power gain. The high breakdown transistors, NPNHB, have
higher collector-to-emitter breakdown voltages, but lower stand-alone gain. This trade..:
off becomes cruciallater on when deciding which transistor type is to be used as the main
power transistor.
At low frequencies, an emitter resistor, RE, may be used in the power transistor to provide
quiescent current stability. At higher frequencies, however, the presence of RE not only
lowers the gain but may also cause oscillation by forming a parallel resonant tank with
38
Chapter 3: Design ofPower Amplifier
the parasitic capacitances. As a result, the emitter of a power transistor, especially at GHz
frequencies, is grounded [20].
a e . : T bi 3 3 Th e curren
Transistor- Emitter width
x length
NPN2-0.5 x 5
NPN2- 0.5 x 10
NPN2 - 0.5 x 20
NPNHB2 - 0.5 x 20
Biasing Circuit
t capa 11ty o 1 e bTt fS"G t . t rans1s ors.
Irated Class AB IQ Range
(mA) (mA)
45 0.64-1
88 1.9-3
160 7.6- 12
120 7.6- 12
V cc
Figure 3.25: Critical RF paths in the PA.
The package model and the bondwires parasitic inductances should be taken into account
from the early design stages. At high frequencies, the bondwires that connect the on-chip
bond-pads to the package pins exhibit significant inductance. The bond wire parasitic
inductances become especially important when in one of the critical paths of the design
shawn in Figure 3.25. The inductance of the bondwire in the input and output RF paths,
critical path 1 and 2, can be absorbed into the input and/or the output matching networks.
The parasitic inductance in path 3 however has a direct degrading effect on the gain of
the PA.
39
Chapter 3: Design of Power Amplifier
The inductance ofthe bondwire can be approximated by [51]:
(Equation 3.5),
where 1 and r are the length and radius of the bondwire in meters, respectively; 11 is the
permeability of the wire in Hlm. The rule ofthumb often used in RF designs assigns 1 nH
of parasitic inductance for every 1 mm of bondwire.
Great care must be taken to minimize the parasitic inductances in the grounded emitter of
the RF power transistor, especially since every dB of gain at RF frequencies counts. One
way to address this issue is to place the power transistor at the layout stage, with respect
to the chip border, and to keep in mind the placement of the package pins, such that the
bondwires for the critical paths will be as straight and as short as possible. Another
important solution is to assign two or more pins to a given grounded emitter. Ideally, two
identical electrically-parallel inductors have an overall inductance of one half. This is true
only if the two inductances are placed physically at a 90° angle to each other, otherwise
the equivalent inductance will be higher than 0.5 due to mutual coupling. In case of two
bondwires that are both physically and electrically parallel, the overall inductance is
about 7110 of the initial inductance [52].
The usual practice is to assign two ground pins to each emitter. Having more pins means
a slightly lower equivalent inductance, but it also requires a larger package with a larger
cavity, which calls for longer metal paths on chip, and longer bondwires. At this design
stage, two package pins are assigned to ground the emitter of the power transistor. Given
the dimensions of the Si Ge chip to be manufactured and a rough number of the required
pins, a 44-pin ceramic quad flat package, CQFP, is chosen. Taking into account the
dimensions of the package, each bondwire is modeled by a 3 nH inductor.
The package too exhibits losses at high frequency. The most important of these are the
series-inductive losses and shunt capacitive losses. A simple model for the pins of a 24-
pin CQFP is provided by the Canadian Microelectronics Corporation (CMC) which take
into account the coupling between adjacent pins as shown in Figure 3.26. As will become
40
Chapter 3: Design of Power Amplifier
apparent later, it was found that this model is not accurate for the 44-pin CQFP at a GHz
range. The losses associated with the package are inevitable; the power amplifier should
be designed around these losses.
254.7 pH 2 n
2.54.7 pH 2n
-----------
Pinn
Coupling Cap 3.75 fF
Pin n+l
Figure 3.26: Initial package pin model.
Similar to the RF choke, the reactance of the decoupling capacitor should also be at least
10 times 50 n in order for it to have negligible loading effect on the main circuit. The
following relation should also be satisfied:
R Xco = Q (Equation 3.6)
where Ris 50.Q and Q is the quality factor of the decoupling capacitor. For a Q of 10, the
minimum value of the decoupling capacitor is 17 pF at 1.88 GHz and 35 pF at 900 MHz.
The decoupling capacitor used here is an off-chip 0.01 JlF capacitor with a self-resonant
frequency of 40 GHz.
As discussed in section 2.2.4, the shunt-tuned output resonant tank is set to resonate at the
fundamental frequency and bypass the harmonies contents; i.e. z1o = oo, z12 = 0, Zp = 0,
and so on. In essence, in the absence of a complementary or transformer-based
architecture, the tank recovers the sinusoïdal shape of the signal. The resonant frequency,
W 0 , of a parallel resonant circuit is given by:
41
Chapter 3: Design of Power Amplifier
1 OJ = -- (Equation 3.7)
0 JLC
where L and C are the inductance and the capacitance of the tank, respectively [17]. The
capacitance, C, should be chosen to be large enough to short all higher order harmonies
to ground. A reasonable value at 1.88 GHz would beL = 1 nH and C = 7.2 pF.
3.1.3.3 Input Matching Network
After supplying the RF transistor with the correct quiescent current, the input of the PA
bas to be conjugate matched to the source, in order to transfer the maximum power from
the source to the PA. Using the small-signal S-parameter analysis in Cadence, the input
impedance of the power amplifier under a specifie biasing at a given frequency is
obtained. The matching is done based on the equations presented in section 2.3.1. There
are two possible two-component matching networks: low-pass (Figure 3.27 (a)) and high
pass (Figure 3.27(b )). In order to prevent node X to be DC ground, the low-pass matching
network is used in the design here. Other options in elude using a T- or a n-network.
3.1.3.4 Output Impedance Transformation
As discussed earlier, the matching requirements in power amplifier are quite different
from that for other types of amplifiers. Conjugate matching does not often yield the best
1oad impedance for maximum output power or efficiency. To complicate the matter
further, the Joad-pull circles for maximum output power and maximum efficiency do not
always coincide, this is especially true at higher power levels.
(a) (b)
Figure 3.27: Two-component matching networks: (a) low-pass and (b) high-pass.
42
Chapter 3: Design of Power Amplifier
Be fore performing the load-pull procedure, the output impedance of the PA, Zout, under a
certain biasing condition is determined by S-parameter analysis. Next, the load-pull
analysis is performed to determine the optimum load impedance, Zopt, either for
maximum output power or maximum efficiency. The load-pull contours in Cadence are
obtained through Spectre's Periodic Steady-State analysis, PSS, and by using the
PortAdapter from the standard "rfExamples" library. At a given drive lev el and
frequency, both the phase and magnitude of the output reflection coefficient, r, are swept
and constant-power contours on the Smith chart are obtained [53]. Using the general
matching technique detailed in chapter 2, Zout is matched to Zopt through a T -network.
The load-pull contours showing the normalized optimal output impedance are presented
in Figure 3.28. The load-pull simulations in Cadence take a long time and do not always
converge. Simulating with more points results in closed contours but may also cause the
simulator to fail to converge. It should be noted that the Cadence analysis tool is capable
of providing the maximum power contours and not the maximum efficiency ones. ADS
on the other hand is a much more powerful tool.
Figure 3.28: Single-transistor PA Joad-pull circles for maximum Pout·
43
20
~ 10 E m "0
0.0
-10
Chapter 3: Design of Power Amplifier
Periodic Steody State Response
Pin (dBm)
Figure 3.29: Single-transistor PA output power characteristics vs. input power.
3.1.3.5 Simulation Results
The input drive power is swept with the output matching network in place. The resultant
output power characteristics and power gain are shawn in Figure 3.29 and Figure 3.30,
respectively. The maximum obtainable gain with one transistor is only 12 dB; the gain is
not enough for the targeted application. The most obvious way to increase the gain is to
use a larger size power transistor. Since the largest transistor size in this technology, with
an emitter length of 20 J.lm, has already been employed, this option was exhausted. More
than one transistor can be put in parallel to increase the overall gain. As will be shawn
later, this presents challenges with regard to power distribution and combining and the
lasses associated with it. The ground pins for the emitter can also be increased to sorne
degree; nonetheless, this did not help increase the gain by much.
Another important issue is the collecter breakdown. The voltages that appear across the
collecter of a power transistor are inherently large. There is a certain amount of voltage
the transistor can handle before the collector-emitter junction gives away. For this reason,
the high breakdown transistors are used in the power core which, even though have a
higher junction breakdown, still cannat handle power levels higher than the given level.
44
Chapter 3: Design of Power Amplifier
P~ri~di' St111~d<; ~t~Jte A:iii-~PQ"I'f~
12 .,~. <)
l1
Figure 3.30: Single-transistor PA power gain.
3. 1.4 Multiple Transistor Design
Putting multiple transistors in parallel increases the overall output power while
decreasing the output impedance. In addition, when too many transistors are paralleled,
the output parasitic impedances ( e.g. the collector-to-substrate parasitic capacitance) will
overcome the transistors output impedances resulting in gain reduction rather than gain
enhancement. Performing simulations with different numbers of transistors showed that
the optimal number of transistors in parallel for this specifie application is four. With
more than four transistors, the output power and gain do not exhibit a significant increase.
The biasing network designed in section 3 .1.3 .1 for the single transistor can be used with
multiple transistors design as well, since the base biasing voltage is transferred through
the RF choke. The issue that arises is the splitting of the input RF power and the
combination of the output power from the different transistors. Different matching
schemes are also possible. Either the matching can be done at one stage for all the
transistors, Figure 3.31 (a), or each transistor can be primarily matched to a value
between 10 n and 25 n and then matched to 50 nin a second stage, Figure 3.31 (b) [54],
[55], [56].
45
Chapter 3: Design of Power Amplifier
Figure 3.31: Biasing schemes for parallel transistors.
In the initial design, the output powers of the four transistors were added on-chip and
then taken off-chip to the load; in ether words, only the transistor multiplicity was
increased. Extensive simulation showed that under this condition, the individual
transistors broke down at about 0 dBm of input drive since, as mentioned before,
paralleling the transistors directly reduces the output impedance. Using high breakdown
NPN transistors with the largest emitter length, NPNHB 0.5 x 20 J..lm2, improves the
situation but not significantly. The problem is alleviated to a great extent by combining
the power from individual transistors off-chipas demonstrated in Figure 3.32. In fact, by
introducing the package parasitics to the signal path, the load-line has moved to a new
location to prevent the breakdown of the transistors. The new load-line presents the
transistors with a lower collector voltage which implies that the individual transistors are
not driven at their full power capabilities, and the larger output power is achieved
basically by the power-additive nature of the parallel transistors.
Similarly, at the input side there can be severa} ways of delivering the RF drive to
individual transistors. One way is to have one path for each transistor, i.e. four input
paths in total. Even though there are enough pins on the package for this purpose, the
bonding wires will have different length for each one. Another way is to have one single
line bringing in the RF drive into the chip and then splitting on-chip. The third method is
to have two input RF pins, each splitting into two once inside the chip. In this case the
bondwires corresponding to these pins will have comparable sizes, thus minimum phase
delay between the inputs and lower distortion at the output.
46
Chapter 3: Design of Power Amplifier
Reference Plane
RF in RF out MN MN
Reference Plane
Figure 3.32: Off-chip power-combining.
Simulations were performed under all the mentioned conditions. The optimum option for
this specifie design is the third solution (Figure 3.32). It should be kept in mind that each
of these methods presents the input of the PA with a different impedance: The input
impedance is not decoupled from the output impedance, meaning that, in this design,
using two pins for the RF input presents the PA with the best load-line at the input and
the output reference planes.
The input impedance of the multi-transistor PA is also conjugate matched to the source
impedance of 50 n. The reference plane for input matching is before the power is split up
in Figure 3.32. Likewise at the output side, the matching reference plane is after the
power combining node, same node as the output resonant tank. So, this power amplifier
in reality consists of four integrated power modules that are matched and combined off
chip. Once the input power-splitting and output power-combining methods are
established, iterative simulations are run to find the optimum number of transistors.
Under these circumstances too, the optimal number of transistors are found to be four -
having more than four transistors does not contribute significantly to the output power.
47
(~
Chapter 3: Design ofPower Amplifier
Bond-pad
To Vww To VHIGH
Figure 3.33: Electrostatic discharge protection circuit.
3.1.4.1 ESD Protection
Care must be taken to protect the PA against electrostatic discharge, ESD, that can
destroy or damage the chip. Bipolar transistors are more rugged with respect to
electrostatic discharge, compared with CMOS transistors where the gate oxide can be
very sensitive to this phenomenon. The ESD protection circuit used in this design
consists of four back-to-back diodes that are placed close to critical bond-pads. The
diodes are reverse-biased under normal operating conditions through Vww of 0 V and
VHIGH of 2.2 V. When an electrostatic discharge occurs, the ESD protection network acts
like a capacitor and directs the built-up charge to ground. A schematic view of the ESD
protection network is shown in Figure 3.33.
The ESD protection circuit should be designed with small size deviees and be placed
close to the bond-pads to reduce its effect on the circuit in normal conditions [57].
Nevertheless, since it acts like a capacitor, it can charge and discharge during the
operation of the PA. The most pronounced effect of ESD protection is on the gain of the
amplifier; the priee of protecting the PA from electrostatic discharge is a gain reduction
of 2 - 2.5 dB. The presence of the ESD protection also causes serious complications with
load-pull simulations in Cadence to the extent that they hardly converge [58].
48
Chapter 3: Design of Power Amplifier
3.2 PA Layout
The important issue of the layout stage is floor planning, with the objective of
minimizing the lengths of the critical RF paths and the RF signal interferences. This is
achieved by placing the RF input, RF output, and RF ground bond-pads at 90° with
respect to each other; with the DC signais occupy the remaining part of the chip as shown
in Figure 3.34. As mentioned earlier, since a maximum of 44 pins are available in the
package, two ground pins are assigned to the grounded emitter RF transistors. A
grounded bond-pad is also placed between each RF bond-pad to shield the active signais.
The supply voltage, VCC, for the bias network is also separated from the VCC of the
power core to minimize any interference through the supply.
In order to create a low resistance path to ground for surge currents and protect the chip
from ionie contamination during fabrication, a guard ring is placed along the chip border.
For specifie details on guard ring requirements for this specifie technology, please refer
to the SiGe Design manual [59].
ICChip
Figure 3.34: Separation and isolation of RF signais to minimize interference.
49
Chapter 3: Design of Power Amplifier
Protection
Figure 3.35: Layout view of a bond-pad with ESD protection.
The SiGe technology contains standard bond-pads that are placed over the n+ region in
the p- silicon substrate to shield it against noise. The dimensions of the standard bond
pads are 109 x 109 11m2• The bond-pads can also be protected by local guard rings to
provide more signal isolation; in this case, the local guard ring is to be tied to the lowest
potential, VSS. A bond-pad with the accompanying ESD protection circuitry is shown in
Figure 3.35. Note the close placement of the ESD protection to the bond-pad and the
biasing of the diodes.
The final PA layout is presented in Figure 3.36. The thicknesses of the metal paths differ
based on the average current they carry and the current ratings of the specifie metal layer.
One last detail to note is the serpentine shape of the RF input lin es to ens ure comparable
phase delays in the four paths.
50
Chapter 3: Design of Power Amplifier
Figure 3.36: Final PA layout.
3.3 Extracted PA Simulation Results
The power amplifier layout is extracted in Cadence. The extraction will take into account
ali the losses associated with the metal paths, the bond-pads, and the parasitic
components. The extracted PA gain and P AE for a quiescent current of 10 mA are shown
in Figure 3.37 and Figure 3.38, respectively. Large-signal stability simulations, based on
the guidelines in chapter 2, show the PA to be stable under the operating conditions.
13.12!
12.12!
11.0
~ 10.0
9.12!0
8.12!0
7.12!0 -212!.00 -11.25 -2.500
Figure 3.37: Extracted PA gain at 10 = 10 mA.
51
~ ..
Chapter 3: Design of Power Amplifier
20
10
Figure 3.38: Extracted PA power-added-efficiency, PAE, at 10 = 10 mA.
The quiescent current is swept, through VcoNT, and the resultant gain, power-added
efficiency, and linearity results are tabulated in Table 3.4. In order to have a better view
of the real capabilities of the power transistors, the input drive power under which the
breakdown occurs under each bias condition is also reported. Note that the given
quiescent current is per transistor. The trend in moving from the class B behavior towards
the class A behavior is clearly seen. At low quiescent currents, the efficiency is high but
the linearity is also low. As the quiescent current is increased, and the PA moves towards
class A, the efficiency drops but the deviee becomes more linear.
T bi 34 P a e . owergam an d PAE f as fi unct1on o 112-
IQ Gain (dB) Maximum Breakdown IMD~
(mA) (for P;0 = -10 dBm) PAE(%) Power (dBm) (dBc)
5 13 60 5
10 12.8 23 6
20 12 14 8 -17
30 10.6 12 10 -40
40 9.8 9 10 -44
51 8.8 2 12 -46
* IMD: Intermodulation Distortion.
52
1
Chapter 3: Design of Power Amplifier
3.4 Package, Bondwire, and PCB Models
The models used in the initial design for the package and bondwires were found later on
to be not accurate enough. Moreover, the printed circuit board (PCB) path losses should
have been accounted for in the design. For this purpose, empirical models were
developed. Utilizing the 44-pin QCFP package, an input pin is bonded and grounded as
depicted in Figure 3.39. This setup allows the designer to devise a model for the
components that will have the most degrading effects on the PA output, namely the
bondwires, the package, and the PCB tracks. One-port S-parameters of this deviee are
measured using a vector network analyzer, VNA, and imported into the Advanced Design
System, ADS, environment. There are several advantages in using ADS for developing
empirical models. The most important is the easy interaction between ADS and the
majority of measurement instruments which facilitates the import and export of data. The
significance of easy interactivity becomes evident when compared with the Cadence
simulating tool where it is very difficult to import arrays of data into the system.
Moreover, ADS has user-friendly pre-defined tools for model generation and
performance optimization. For this purpose, initial rough models are setup for each
component and ADS tools are used to fit the S-parameters to the RLC networks. For a
description of the model fitting scheme that is used in ADS please refer to Appendix A.
Figure 3.39: The setup for developing empirical models for the PCB track, the bondwire, and the package pin.
53
Chapter 3: Design ofPower Amplifier
3.4.1 Bondwire Mode/
The bondwire is modeled by series inductive and resistive components. In the original
design, it was modeled with an inductor with value of 1 nH/mm. In the new model, the
inductive component is kept the same but resistive components have been added to model
the losses that are independent of frequency. The fitted bondwire model for a 1 mm of
bondwire is shown in Figure 3.40.
213 fH 1 nH 213 fH
80.3 mn 80.3 mn Figure 3.40: Bondwire model for a 1 mm of bondwire.
3.4.2 Package Mode/
The fitted package model is far more complicated than the one used before. The initial
package pin model, shown in Figure 3.26, was a single RLC network. The behavior of
the actual package pins however turned out to have more of a distributed nature.
Moreover, the series inductive losses and the shunt capacitive losses add up to far greater
values than what the previous model had. The empirical model for a single package pin is
presented in Figure 3.41; the sub networks are designed to be identical.
Figure 3.41: A single package pin model.
54
Chapter 3: Design of Power Amplifier
3.4.3 PCB Track Mode/
Modeling the PCB tracks is more complicated than the bondwire and package. In the
initial design, a PCB line was modeled as a 0.5 nH inductor. For the empirical madel,
every 50 mils (0.127 cm) of the PCB track were modeled with an RLC network as seen in
Figure 3.42. When the PCB lasses are accumulated, they become significant. This is
especially true in the output path where four branches are used, with the best case length
of 400 mils (1.0 16 cm) and worst case 1ength of 500 mils (1.27 cm).
827 pH 315 mO
T 2191F
Figure 3.42: PCB line mode! per 50 mils (0.127 cm).
3.4.4 Final Simulation Results
After incorporating the new models into the design, the PA simulations are rerun. The
response shows that the off-chip losses are so high that the output of the PA is degraded
to a great extent; the S21 response is shawn in Figure 3.43. In chapter 5, these results are
compared with the actual measured results and solutions for improving the gain are
discussed.
••
Figure 3.43: S21 response of the PA with the new and more accu rate models in place.
55
Chapter 4: MEMS Tunable Power Amplifier
4 MEMS Tunable Power Amplifier
Micro-electromechanical systems (MEMS) enable new levels of performance in radio
frequency integrated circuits which are not readily available via conventional IC
technologies. The low quality factor of IC passive elements, such as inductors,
capacitors, and filters, force the designers to use off-chip elements. Off-chip components
are undesirable due to the parasitic losses associated with the component itself, the
soldering, and the PCB tracks. Moreover, large value extemal components often have
low self-resonance frequencies making them unsuitable for RF applications. RF Micro
electromechanical deviees offer the opportunity to integrate many RF subcomponents
that are traditionally implemented off-chip. RF MEMS have entered the
commercialization phase in 2003 offering compact, low power, low loss, highly linear,
and high Q solutions. They also make possible new architectures, with the possibility of
reconfigurability and tunability for multi-band or multi-standard operation [60], [61].
In this part of the research, MEMS variable capacitors are designed and implemented in
the Metal-MUMPs process. First, the mechanical and electrical operations of MEMS
tunable capacitors are viewed, and different mechanisms for achieving tunability are
discussed. Next, an overview of the Metal-MUMPs process is presented. Several varactor
designs based on the capabilities of the Metal-MUMPs are proposed. Finally, the
designed varactors are visited in the context of power amplifier matching and load-pull
analysis.
56
Chapter 4: MEMS Tunable Power Amplifier
4.1 MEMS Tunable Capacitors
4. 1. 1 Overview
Variable capacitors are used in a number of RF circuits, such as voltage controlled
oscillators, matching networks for power amplifiers, and loaded-line phase shifters. In all
of which, the performance of the circuit is directly related to the quality of the variable
capacitor. Depending on the application, the requirements on MEMS capacitors differ. In
voltage-controlled oscillators, for example, the quality factor of the MEMS varactor
should be as high as possible. For matching purposes, on the other hand, a wide tuning
range is important.
Traditionally, variable capacitors m microelectronic systems have been implemented
using p-n junction diodes or MOS capacitors. The se deviees have a relative} y poor tuning
range, are extremely non-linear, and have significant parasitic losses through the
substrate. MEMS varactors can handle large voltage swings. In p-n junction capacitors,
the voltage swing is limited in order to ensure the proper biasing of the p-n junction, in
other words to always keep the junction reverse-biased.
The most important drawback of semiconductor varactors is the dependence of the
capacitance on the RF signal level, which makes the component behavior highly
nonlinear. In case of MEMS capacitors on the other hand, good linearity is inherent.
Since the mechanical resonant frequencies of micro-machined tunable capacitors usually
lie in the 1 0-1 00 kHz range, these deviees do not respond to RF frequencies and therefore
do not produce a significant amount of harmonie distortion [62], [63], [64].
There are severa} issues associated with MEMS varactors. One drawback is the need for
large actuation voltages, sometimes in ex cess of 40 V. Another issue is the series resistive
losses resulting in a low unloaded Q.
57
Chapter 4: MEMS Tunable Power Amplifier
Suspended Plat
Fixed Plate
Figure 4.44: Two-plate electrostatically actuated MEMS varactor.
4. 1.2 Principle of Operation
MEMS-based tunable capacitors have been implemented using a variety of techniques
and mechanisms. The most commonly used structures rely on varying the overlap area of
the capacitor by changing the distance of either vertical plates or lateral fingers meant to
increase the fringing field. These two methods of tunability are presented here in detail.
Other modes of operating MEMS varactors have been explored in [65], [66], [67]. These
tunable capacitors require customized manufacturing capabilities, such as the ability to
slide the dielectric material and/or the presence of piezoelectric materials, and do not
demonstrate a significant advantage in terms of performance and will not be considered
here.
4.1.2.1 Parallel-Piate MEMS Varactor
A functional madel of a two-plate electrostatcially actuated MEMS tunable capacitor is
shawn in Figure 4.44. The top plate is suspended by arms having an overall spring
constant of k wh ile the bottom plate is fixed. The capacitance of this structure, assuming
air as the dielectric, to first order is given by:
C = &oA 0 d (Equation 4.1)
58
Chapter 4: MEMS Tunable Power Amplifier
where A is the overlap area of the two plates, d is the gap between the two plates under
the no-bias condition, and 80 is the permittivity of air. Applying a bias voltage of V(t)
results in an electrostatic force that pulls down the top plate towards the bottom plate,
thus varying the inter-plate capacitance according to Equation 4.2:
C= &OA d-x
(Equation 4.2)
where x is the displacement of the top plate due to the applied bias, V(t). The resto ring
force of the spring however opposes the electrostatic pull-down. As long as the gap
between the two layers is larger than a third of the original gap, the two forces are in
equilibrium. If the gap displacement becomes larger than this critical value, the
electrostatic force overcomes the spring restoring force and the top structure collapses
completely on the lower electrode. This phenomenon, known as pull-in, limits the
displacement of the movable plate, x, to d/3; thus the maximum value of capacitance
before failure, CMAX, is limited to i &oA. Based on Equation 4.3, the tuning range of the 2 d
simple two-plate electrostatically-actuated MEMS varactor is limited to 50% [62], [68],
[69]:
C -C Tuning Range = MAX
0
Co (Equation 4.3)
The pull-in voltage is dependant on the spring constant of the suspension arms, k, the
distance between the plates, d, and the plate area, A, as described by the following
equation:
vpull-in = (Equation 4.4)
The vertical electrostatially-acutated capacitor is the simplest form of a MEMS varactor
to implement by surface micromachining; the only drawbacks are the need for large
actuation voltages and the performance limitations imposed by the pull-in effect.
Different MEMS-based varactor designs have been implemented which through the
addition of an extra structural layer, take advantage of the readily available surface
micromachining techniques while enjoying much wider tuning ranges. Two of these
generic structures, referred to as Type II and Type III, are discussed here.
59
Chapter 4: MEMS Tunable Power Amplifier
k
V(t)
Figure 4.45: Type Il parallel-plate MEMS varactor with wide tuning range.
4.1.2.1(a) Type II Vertical Varactor
Figure 4.45 illustrates the schematic for a type II parallel-plate MEMS varactor. At }east
three structural layers are required to construct the two gaps and separate the actuation
electrodes E1 from the main capacitor plate E2 . For extended tuning range, the un-biased
capacitor gap d1 should be larger than one third of the electrode gap d:z, i.e. dz 2: d:z/3.
Upon the application of actuation voltage, V(t), the movable top plate is pulled down
towards the bottom plate, in this case, the distance that defines the onset of pull-in is d2/3,
as opposed to the shorter distance of d1/3, hence the maximum achievable capacitor,
CMAX, becomes ~ &oA. Therefore the tuning range, TR, becomes: 2 d 2
TR = 3d! type!I 3d - d
1 2
(Equation 4.5)
which can be as high as 84%, depending on the technology specifications that define the
available gaps [70], [71], [72], [73].
60
Chapter 4: MEMS Tunable Power Amplifier
Figure 4.46: Type Ill parallel-plate MEMS varactor with very wide tuning range.
4.1.2.1(b) Type III Vertical Varactor
The tuning range of the parallel-plate varactor can be further extended by using a two-gap
type III structure, as illustrated in Figure 4.46.
In this class of variable capacitors, the actuation electrodes E1 are kept separate from the
main capacitor plate E2 and the gap between the suspended membrane and the signal
electrode, d1, is smaller than a third of the distance between the membrane and the
control electrodes, d2, i.e. d1 :::; d2/3. As a result, the pull-in of the control electrodes does
not occur before the suspended membrane touches the signal electrode. Thus, the
maximum achievable capacitor is determined by the dielectric material that is on top of
the signal electrode E2• The tuning range can theoretically go to infinity; in practical
applications however tuning ranges as high as 280% have been reported [69], [72], [73],
[74].
4.1.2.2 Lateral lnterdigitated MEMS Varactors
Figure 4.47 shows a laterally-driven interdigitated tunable capacitor where one set of the
combs is fixed while the other set is free to move. Upon the application of a voltage, the
electrostatic force at the fringes of the fingers causes in-plane actuation of the movable
61
Chapter 4: MEMS Tunable Power Amplifier
combs. The tunability is thus achieved by varying the overlap area of the capacitor. The
tuning range of a lateral capacitor is not limited by a pull-in voltage; tuning ranges of
200% have been achieved with this structure [75], [76], [77]. The only real limitations are
the mechanical characteristics of the comb fingers and the suspension spring. Long
fingers increase the fringing field capacitor and allow larger tuning, but are also more
susceptible to breaking during the manufacturing and testing phases. Also, the greater the
number of fingers, the larger the gravitational force, thus the higher the chances of
suspension arm failure. Compared with vertical MEMS capacitors, the interdigitated
varactors usually have lower Q-factors and low self-resonance frequencies, which makes
them not suitable for RF applications [78], [79], [80].
4.1.2.3 Semi-Fractal Varactor
In the microelectronics industry, fractal capacitors have been investigated. They make
use of both vertical and lateral electric fields to increase the capacitance per unit area,
which can result in considerable reduction in the chip space consumed by the capacitors
[81]. An ideal case in the MEMS-based varactor design would be a fractal MEMS
capacitor that benefits from the increased capacitance density offered by fractal structures
and the tunability offered by MEMS varactors. A similar structure is proposed in [82],
however, no analytical expression for the capacitance is offered.
Figure 4.47: lnterdigitated lateral MEMS varactor.
62
/"'~ 1 ',
Chapter 4: MEMS Tunable Power Amplifier
Figure 4.48: Top view of the semi-fractal MEMS varactor.
A top view of the semi-fractal MEMS varactors is shown in Figure 4.48. The semi-fractal
capacitor is similar to a simple MEMS varactor, with the difference that the top plate is
made of interdigitated fingers to increase the capacitance per unit area.
4.1.3 The Metai-MUMPS Process
The Metal Multi-User MEMS process, Metal-MUMPS, surface micromachining
technology is used to implement the MEMS varactors designed in this research. As seen
in Figure 4.49, the important feature of this process is the presence of a 20 11m thick
nickel structural layer covered by 0.5 11m of gold. The second structural layer is a 0.7 11m
thick doped polysilicon which can be suspended over the silicon substrate. A 0.35 11m
thick ni tri de layer is deposited on top of the polysilicon; this layer acts as a dielectric and
prevents short circuits in case the metal layer collapses on the poly. The gap between the
metal and the polysilicon is 1.1 11m and the gap between the suspended polysilicon and
the un-etched substrate is 0.5 Jlm. A 25 Jlm cavity is etched in the silicon substrate over
which the polysilicon structural layer can be suspended. A combination of platinum and
chrome are used to anchor the metal layer to the substrate [83]. It has been shown that
removing the substrate results in higher self-resonance frequency and quality factors in
general due to the reduction of parasitic capacitive losses to the silicon substrate [84].
63
' l
Chapter 4: MEMS Tunable Power Amplifier
Figure 4.49: Overview of the Metai-MUMPS structurallayers.
The presence of two structurallayers, where both can be made movable, and two gaps,
makes the implementation of wide tuning range varactors with this technology possible.
Also, the low resistivity of the metal structural layer lowers the series lasses in the
capacitive plate, the suspending arms, and the signal pads to a great extent. However,
since the other structural layer is polysilicon, high values of Q cannat be expected [85].
4.1.4 MEMS Varactor Design in Metai-MUMPS
Figure 4.50 shows a top-view schematic of the MEMS capacitors designed in the Metal
MUMPS process.
Figure 4.50: Top view of the varactor designed in Metal-MUMPS.
64
t
t
•
Chapter 4: MEMS Tunable Power Amplifier
Figure 4.51: Functional model of the designed two-plate wide range varactors.
In order to understand the operation mechanism of this particular capacitor, the cross
section along the AA' line is given in Figure 4.51. Both the polysilicon and the metal
layers are suspended. The polysilicon layer is set to ground, while a DC voltage is
applied to the metal layer. The electrostatic force initially pulls down the metal
membrane towards the polysilicon, varying C1, until it collapses onto the polysilicon
membrane. Due to the presence of the thin nitride layer on top of the poly, short circuit
does not occur. Since the poly membrane is also free to move, the collapsed structure
moves toward the silicon substrate this time varying the C2. The two phase tunability of
the total capacitance results in an extended tuning range. Note that C3 denotes the
parasitic capacitance of the metal layer to the substrate which is not very significant due
to the presence of the trench, but still can deter the overall performance.
Using Equation 4.1 and assuming an initial gap of 1.1 Jlm, the top plate dimensions for
different capacitances are calculated and tabulated in Table 4.5. The dimensions of the
bottom plate are the same as that of the top plate. Uniformly spaced ho les are etched in
the bottom and top plates to facilitate the removal of the sacrificial oxide between the
polysilicon and the substrate and the polysilicon and metal. As discussed in section
4.1.2.3, the structures that combine the fringing capacitance of interdigitated capacitors
with the vertical plate varactor concept are of interest. For each of the simple varactor
structure, an equivalent semi-fractal structure is also included. The semi-fractal designs
65
Chapter 4: MEMS Tunable Power Amplifier
are expected to have higher initial capacitances, but may be more susceptible to failure
due to gravitational forces because the weight of the plate is not uniformly distributed.
Table 4.5: Capacitor plate dimensions.
Capacitance Length
0.05 pF 80 f.1m
0.1 pF 112 f..lffi
0.5 pF 250f.1m
1 pF 353 f..lffi
2pF 500 f..lffi
5pF 800 f.1m
4.1.4.1 Suspension Design
The suspension structures that are most often used in MEMS varactors are of the T -type.
As seen in Figure 4.52, the T-type suspension can be modeled to a first degree by a series
and parallel combination of springs. The equivalent spring constant of each suspending
arm is given by [62]:
k = 2k,k2 k, +2k2
(Equation 4.6)
where k1 and k2 are the elasticity constants corresponding to L1 and L2, respectively, and
are determined through:
(Equation 4. 7)
where W;, L;, and T; are the width, length and thickness of the bearn as indicated in
Figure 4.52, respectively. E is the Young's modulus of the structural layer, 200 GPa for
the nickel structure and 158 GPa for the polysilicon layer. The overall spring constant is
4k for a plate supported by 4-arms.
There is a large temperature-dependent stress in metal structures. The stress is caused by
the difference in the coefficient of thermal expansion of the metal and the substrate, as
well as between the metal and the polysilicon [69]. A curved suspension design, which is
66
Chapter 4: MEMS Tunable Power Amplifier
basically a quarter of a circle with a given width, fixes the metal plate in the plane
parallel to the substrate while allowing horizontal rotational movements in the metal
structure; thus partially absorbing the stress [73], [86].
Figure 4.52: T -type suspension and the equivalent spring model.
Using the plate dimensions given in Table 4.5 and limiting the pull-in voltage to 10 V,
the required dimensions of the arms can be calculated based on Equations 4.4, 4.6 and
4.7. Low actuation voltages require long arms, which increases the chance of failure due
to gravity since the spring is also suspend and is thin. The calculated values for the arm
dimensions are summarized in Table 4.6. The longer part of the arms, L 1, can be either
realized as a straight cantilever or a serpentine arrangement. For comparison purposes,
bath cases were included in our designs.
T bi 4 6 C a e . : apac1 or suspen mg arm "t d" d" amensaons.
Capacitance Width Length
0.1 pF 10 f.!m 1300 f.!ID
0.5pF 20 f.!ID 970 f.!ID
1 pF 30f.!m 880f.!m
2pF 40f.!m 772 f.!m
SpF 50 f.!ID 612 f.!m
67
Chapter 4: MEMS Tunable Power Amplifier
.. 100 1..1 ~
Figure 4.53: The GSG signal pad for on-chip probing.
4.1.4.2 Signal Pad Design
The signal pads are designed for on-chip probing with Ground-Signal-Ground, GSG,
probes with a pitch of 150 Jlm (Figure 4.53). The ground pads are 100 x 100 J.Lm2 and the
signal pad is 70 x 70 Jlffi2 to minimize capacitive losses to substrate. As seen in Figure
4.54, the signal pads are implemented using the 20.5 J.Lm thick nickel-gold layer which is
anchored to the substrate by a platinum-chrome layer, two nitride layers with total
thickness of 0.7 Jlm, and the 2 Jlm thick isolation oxide. The theoretical pad resistance is
given by:
sA C pad =- (Equation 4.8)
x
where A is the area of the signal pad, i.e. 70 x 70 J.Lm2, and x is the thickness of the ni tri de
and oxide layers, which amounts to 2.7 J.Lm. The calculated value of Cpad is 0.055 pF. The
above calculation does not take into account the fringing field capacitance of the signal
pad which results in the actual value of the capacitance being larger. In [69] it was shown
that removing the substrate from under the signal pads can increase the Q from 3.5 to 182
at 1 GHz and to 119 at 2 GHz.
0.7
Figure 4.54: The cross sectional view of the RF pad in Metal-MUMPS.
68
Chapter 5: Measurement Results
5 Measurement Results
The packaged power amplifier module is tested by means of a printed circuit board, PCB.
The MEMS capacitors are tested separately to determine their capacitances and tuning
ranges. The measurement results are presented in this chapter and are compared with the
previously obtained simulation results.
5.1 The Test Setup
The picture of the manufactured power amplifier die is shown in Figure 5.55. Initially,
the die is mounted on a 44-pin gold-plated Ceramic Quad Flat Package, CQFP, with a
0.56 x 0.56 cm2 cavity. The placement of the die in the package is shown in Figure
5.56(a). The placement is such that the critical RF paths are minimized as much as
possible.
69
Chapter 5: Measurement Results
Figure 5.56: {a) The PA die placement in the package, (b) The top view of the PCB containing the stand-alone PA.
A printed circuit board, PCB, is designed to test the packaged PA employing off-chip
matching networks. In the PCB layout severa} issues are taken into account to minimize
the parasitic losses as much as possible. The lengths of the PCB traces carrying RF
signais, whether input or output, and the RF ground should be minimized. The use of a
bottom ground plane ensures the minimization of the RF flux loop area, and thus the
parasitic inductances. Also, a separate PCB trace is allocated to each RF ground pin,
which reduces the effect of ground bounce [87], [88]. For the off-chip components that
are in the input/output paths, small footprints are used to minimize the parasitics. By
simulating the coplanar structure of the PCB traces in ADS, the width of the PCB routing
paths are set to 0.508 mm, 20.0 mils, with 0.305 mm, 12.0 mils, of clearance to achieve
50 n lines. Wide traces have lower inductance per length and are especially suitable for
lines that carry large currents since the heat dissipation is more easily achieved [89], [90].
End launch gold-plated SMA connectors are used for RF input and output; these SMAs
have lower loss and better voltage standing-wave ratio, VSWR, at RF frequencies
compared with vertical and horizontal SMAs (Figure 5.56(b)).
70
•
Chapter 5: Measurement Results
5.2 Stand-alone Power Amplifier Testing
The PCB with the packaged power amplifier die is tested for DC characteristics. The
testing includes measuring the DC voltage at the accessible nodes, as well as the DC
current flowing through the biasing network and the main PA core. Once the correct DC
behavior is established, the RF characteristics of the PA are measured using the Agitent
8753D 60Hz Vector Network Analyzer, VNA. Initially, the input and output matching
networks are shorted in order to obtain the behavior of the PA core along with the
bondwires, the package, and the PCB traces. The measured input reflection coefficient,
Su, and the gain, S21, for the un-matched PA are presented in Figure 5.57.
10
0 --...... ...... èïi -10 .._.. ca "0
-20
0
1.62 GHz -24.4 dB
2 3
freq, GHz
4 5
40~----------------------~--------------~
20
--...... 0 N' .._.. (/) ar -2o "0
-40
-60-t--,--,---,---,--t-,----rl
0
1.48 GHz -29 dB
2 3
freq, GHz
4 5
Figure 5.57: The measured 811 and 5 21 for the un-matched stand-alone PA.
71
~ ..
Chapter 5: Measurement Results
The corresponding Su and S21 obtained by simulating the stand-alone PA are presented
in Figure 5.58. According to the simulation, the gain is expected to drop to zero at 910
MHz. The simulation also predicts a gain of 4.3 dB at 630 MHz. The measured results
show that the gain drops to zero at 660 MHz and the gain at 630 MHz is 0.92 dB. In both
cases, the Sn and S21 responses show a dip. The simulation predicts the Sn dip at 1.49
GHz and the measured response shows the dip at 1.62 GHz. Similarly, the S21 is
predicted to have its lowest value at 1.8 GHz while the actual circuits exhibits the notch
at 1.5 GHz. So, even though the numerical values do not fully agree, the shapes of the
simulated and measured responses are similar. Referring back to Figure 5.57, at
frequencies above 660 MHz, as seen in chapter 3, the losses associated with the
bondwires, the package, and the PCB traces become so high that they reduce the gain to a
great extent.
,.. ,..
.... M
-f8 -
t
.. , .... f
f 1
_, J., L ~····, ..
10
~St
"li)
-20
->•
--•ht
-~ec
630MHz -1.45 dB
'"···-,--~ .. _./_ ·1'!---.........
11 630MHz 4.3 dB
'"""""' .. .,._
1.49 GHz -2.3 dB
1.88 GHz -38 dB
1.8 GHz -62 dB
.,
'"' '"'"
Figure 5.58: 5imulated 511 and 8 21 for the un-matched stand-alone PA.
72
/~
Chapter 5: Measurement Results
In the next step, the S-parameters of the input-matched PA are measured and imported
into the ADS and load-pull simulations are performed to determine the optimum load for
maximum power transfer. The optimum load is determined to be 1.4- j 13.5 Q. Based on
the fact that the output impedance of the input-matched PAis 0.1 + j 15.5 Q; according to
the general matching equations of section 2.3.1, the output matching network for
maximum output power, consisting of series capacitors and a shunt inductor, is designed.
The measured S 11 and S:u for the input/output matched PA are shown in Figure 5.59. It is
evident that, due to the input conjugate matching and output power matching, the gain
has increased from 0.92 dB at 630 MHz to 9.44 dB at the same frequency. ln other
words, the power matching results in about 8.5 dB increase in output power.
0
-5 ........ ........ ...--...- -10 _.. en _.. co "U
-15
-20
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
freq, GHz
20
0 ........ ........ ...-N - -20 en -co "U
-40
-60 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.
freq, GHz
Figure 5.59: Measured 511 and 821 for the stand-alone PA.
73
Chapter 5: Measurement Results
Figure 5.60: The MEMS varactors die.
5.3 MEMS Varactors Testing
Figure 5.60 shows a microscopie view of the MEMS varactors die. The fabricated
MEMS varactors are tested on-die using a probe station. Two DC sources are put in
series to achieve actuation voltages up to 35 V. The actuation voltage is applied to the
MEMS deviees through the built-in biasing-tee of the Agilent 8753D 6GHz Vector
Network Analyzer. The DC voltage that the Network Analyzer can tolerate is specified as
40 V. In order to ensure the safety of the test equipment, the maximum applied voltage is
kept at 35 V. Most of the fabricated MEMS structures are single-ended and are tested
using a single GSG probe. Two probes are used to test the differentiai capacitors. Figure
5.61 shows a closer view of one of the MEMS deviees, namely the nominal 1 pF
capacitor with spiral arms.
74
Chapter 5: Measurement Results
Figure 5.61: The MEMS spiral-arms 1 pF nominal capacitor.
Figure 5.62 compares the impedance characteristics of the spiral-arms 1 pF nominal
capacitor as a function of frequency at actuation conditions of 0 V and 35 V; the obtained
tuning range is 260% at 630 MHz and 355% at 1.8 GHz for this capacitor.
630MHz 5.7pFat35V
1.88 GHz 1.43 pF at 0 V
1.88 GHz 6.5 pF at 35 V
630MHz 1.58 pF at 0 V
freq (SO.OOMHz to 2.000GHz)
Figure 5.62: Capacitance of spiral-arms 1 pF nominal capacitor as a function of frequency at 0 V and 35 V.
75
Chapter 5: Measurement Results
Capacitance vs. Voltage
7
6 lt.ss GHz l /
5 r -v· ·;
Cl) Il\ CJ 4 c
di630MHz J9 "ëj
[3 Cil r ()
2 ... 1
0
<::> ~ <o "~).. "fo ~ 1' ~ ~1).. ~~)..~ ~~ <t Voltage
Figure 5.63: Capacitance of the spiral-arms 1 pF nominal capacitor as a function of actuation voltages at 630 MHz and 1.88 GHz.
As explained earlier, the varactors exhibit a two-stage tuning behavior (Figure 5.63). At
voltages lower than 25 V, the capacitance increases slowly with increasing actuation
voltages. At voltages between 25 and 32 V, depending on the size of the varactor under
test, a sudden jump in the capacitance is observed. The observed behavior is a direct
consequence of the two-gap structure of the Metal-MUMPs process which is repeated in
Figure 5.64 for convenience. The first tuning is due to the actuation of the metal
structural layer, i.e. C1. When the actuation voltage reaches the pull-in voltage of the
metal-polysilicon gap, the metal collapses on to the polysilicon. However, due to the
presence of the 0.35 ~-tm-thick nitride dielectric layer which is deposited on top of the
polysilicon, a short circuit does not occur. The second tuning is then due to the actuation
of the combined metal-polysilicon layer over the etched substrate, i.e. C2• The pull-in for
this structure occurs at voltages higher than 35 V. However, due to the limitations of the
VNA, the maximum directly-applied DC voltage is kept around 35 V.
76
Chapter 5: Measurement Results
Figure 5.64: Functional model of the two-plate wide range varactors with tuning steps.
The measurements are done at 1.88 GHz and 630 MHz for four dies. The average results
at 1.88 GHz for wide-tuning range MEMS varactors are summarized in Table 5.7.
Comparing the results based on the plate structure, normal vs. semi-fractal, presents
interesting points. Even though semi fractal plates have higher capacitances under zero
actuation voltages compared with simple plate structures, they exhibit lower capacitances
at higher actuation voltages. This can be due to the un-uniform plate surface that reduces
the effective area when the actuation voltage is applied. Another issue is the arm
structure. It seems that the spiral arm structures (Figure 5.61) have the best elasticity
behavior and therefore yield the best results.
Table 5.7: Measurement results for MEMS varactors at 1.88 GHz.
Nominal Arm Plate Capacitance Capacitance Tuning
Capacitance Structure Structure atOV at35V Range
0.1 pF Spiral Normal 0.69 pF 1.42 pF 102%
0.1 pF Spiral Semi-Fractal 0.67 pF 1.23 pF 81.5%
0.1 pF Straight Normal 0.69 pF 1.59 pF 129%
0.5pF Spiral Normal 1.67 pF 7.7pF 358%
0.5pF Spiral Semi-Fractal 1.6 pF 4.53 pF 181%
0.5pF Straight Normal 1.02 pF 4.02 pF 294%
0.5pF Curved Semi-Fractal 1.14 pF 1.98 pF 71.6%
1 pF Spiral Normal 2.34 pF 11.2 pF 375%
1 pF Spiral Semi-Fractal 2.38 pF 10.8 pF 350%
2 pF Straight Semi-Fractal 3.78 pF 13.87 pF 266%
77
Chapter 5: Measurement Results
5.4 MEMS-Tunable Power Amplifier Testing
As the last stage of testing, the power amplifier die and the MEMS die are mounted
inside a 68-pin CQFP package with a 1.02 x 1.02 cm2 cavity. Two of the MEMS
capacitors with wide tuning ranges are differentially bonded to the package. The selection
of the MEMS capacitor is done on the PCB by soldering a short circuit path to either of
the capacitors. The overall circuit is initially tested for DC behavior, then the RF
characteristics are measured with the VNA. Similar to the stand-alone PA, the input and
output matching networks are first shorted in order to investigate the behavior of the
packaged power amplifier along with the effect of the PCB traces. The losses associated
with the bondwires and package are expected to be higher for this new setup, due to
severa} factors. First, the package pins, unlike the 44-pin package, are aluminum and not
/.-.., gold-plated. Second, due to the larger size of the cavity, the bondwires are also longer.
78
Chapter 5: Measurement Results
Even though care has been taken to keep the critical paths such as the RF grounds and RF
input to minimum, the RF output is much longer than ideal due to the geometry and the
position of the dies inside the package. Moreover, the low Q-factor of the MEMS deviees
also contributes to the output tosses. The measured Sn and S2z characteristics of the un
matched MEMS-tunable PA are given in Figure 5.66. As expected, the S21 under the un
matched condition is lower than that of the stand-alone PA, -8.9 dB vs. 0.92 dB at 630
MHz. Also, no gain at any frequency is seen.
--...... ...... -- -10 en --al "'C
-15
-10
~ -15 ...... ~ -20 en --al "'C -25
-30
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.
freq, GHz
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
fre , GHz
Figure 5.66: The measured 5 11 and 5 21 for the un-matched MEM5-tunable PA.
79
~ i '
Chapter 5: Measurement Results
After conjugate-matching the input, the output impedance is found to be 94.7 + j14.8 n.
Load-pull simulations are performed to determine the optimum load for maximum output
power to be 110 - j 17.5 .Q. Matching the output impedance to the optimum load through
a T -network, the S 21 of the full y matched PA is found to be -11.5 dB at 630 MHz as seen
in Figure 5.67; the power amplifier still exhibits no gain over the frequency spectrum.
The main reason for absence of gain is the lasses incurred in the RF output path. The
main contributing factors to which are the long bondwires (as evident from Figure 5.65),
the lassy package pins, and the low Q-factor of the MEMS deviees. Currently, work is
being done to optimize the PCB mainly through on-board bonding to minimize the
effects of package losses and reduce the lengths of the bondwires, which due to time
constraints are not included in this thesis.
-2
:::::::::: -4 ...... :s -6 Cl)
m "C -8
-10
-10
--...... -20 ~ Cl) m -3o "C
-40
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
freq, GHz
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
fre , GHz Figure 5.67: The measured S11 and S21 for the matched MEMS-tunable PAs.
80
Chapter 6: Conclusion
6 Conclusion
6.1 Summary
The growing demand for wireless communication systems IS demanding increased
integration, better signal quality, and longer battery life. The power amplifier consumes
the most power in the whole receiver/transmitter system, and its output signal is directly
transmitted by the antenna without further modification. Th us, optimizing the PA for
lower power consumption, increased linearity and integration is highly desired. One way
to ac hi eve this objective is to actively vary the output matching network of the PA for
maximum output power, as the input RF drive or the output load conditions vary. MEMS
variable capacitors were designed to be used as part of an active matching network. The
major effort of the thesis was focused on the design of MEMS tunable class AB power
amplifier in the SiGe BiCMOS 5HP technology, and the designed implementation of
MEMS variable capacitors in the Metal-MUMPS process.
In chapter 1, the motivations and objectives for power amplifiers design, in general, and
tunable PAs, in particular, for RF applications were introduced. Chapter 2 dealt with the
power amplifier operation principles, the methods of characterizing and quantifying the
PA performance, and the classification of power amplifiers based on their modes of
operations with special attention to class AB operation. Design techniques specifie to
81
Chapter 6: Conclusion
high power amplifiers including large-signal S-parameters, power matching (as opposed
to conjugate match), and the load-pull technique were also addressed in detail.
Chapter 3 was dedicated to the design of the SiGe BiCMOS power amplifier. The issues
regarding the design of the power core, biasing circuitry, input and output matching
networks, and power splitting/combining were addressed. The load-pull technique was
employed to determine the output matching network for maximum output power. In this
chapter, empirical models were developed for parasitic losses resulting in the PCB paths,
the package pins, and the bond-wires.
In chapter 4 MEMS variable capacitors for RF applications were designed. MEMS
tunable capacitors operating principles and different varactor structures were
investigated. Very-wide tuning range tunable capacitors were designed taking advantage
of the two structural layers offered in the Metal-MUMPS process. The measurement
results for the fabricated varactors were given in chapter 5, where it was shown that the
capacitors covered a frequency range from DC to 6 GHz, with tuning ranges as high as
375%.
Chapter 5 summarized the measurement results for the stand-alone power amplifier and
the MEMS-tunable power amplifiers. In the first case, the packaged, input/output
matched PA exhibits a gain of 9.44 dB at 630 MHz. It was shown that the gain at higher
frequencies is low due to the parasitic losses of the bond-wires, the package pins, and the
PCB tracks.
6.2 Topics for Future Research
The idea of actively tuning the matching network of a power amplifier by means of
MEMS variable capacitors for better efficiency and output power was explored in this
thesis. MEMS capacitors have the potential of offering greater degree of integration as
well as better performance, reduced power consumption and better linearity, for the
design of power amplifiers. This thesis highlighted the areas that needed further research
82
Chapter 6: Conclusion
in order to attain the specified goals, as weil as ether areas that can benefit from the
integration with MEMS.
a) In order for a power amplifier design to be of any practical use, it should be fully
packaged. This leads to the issue of characterizing and developing accurate models for
the packages. In practice, the losses incurred due to the package and bond-wires can have
devastating effects on the performance of the circuit, and thus have to be taken into
account from the very early stages of design.
b) In this thesis, the emphasis was on class AB power amplifier and its tunability, since it
offers the best tradeoff between linearity and efficiency. However, switched-mode power
amplifiers, such as class E and class F, can also be good candidates for integration with
MEMS. One area where MEMS can be used is for the output resonant tank, which is
cri ti cal to the operation of the se PAs. The resonant tanks are conventionally implemented
as LC networks which are either on-chip, thus having lower parasitics but generally
invariable, or are off-chip, hence suffering from large degrees of parasitic losses in favor
of being tunable. These tanks can be replaced by in-package high-Q MEMS resonators
such as the ones designed in [3] and [4].
c) In this thesis micro-electromechanical variable capacitors with very wide tuning ranges
were implemented successfully. However, the designs were based on first order
simplified equations of the spring constants, pull-in voltages ... ect. If simulation models
are developed and are available for a given MEMS technology, the potential of the
technology can be explored to a fuller extend, and even better structures can be designed
and fabricated.
83
Chapter 7: References
7 References
[1] Cripps, S.C., RF Power Amplifiers for Wireless Communications, Artech House, 1999
[2] Sokal, N.O., "RF power amplifiers, classes A through S - How they operate, and wh en to use each", Professional Pro gram Proceedings of Electronics Industries Forum ofNew England, May1997, pp. 179-252
[3] Clark, J.R.; Bannon, F.D.; Ark-Chew Wong; Nguyen, C.T.-C., "Parallel-resonator HF micromechanical bandpass filters", IEEE International Conference on Solid State Sensors and Actuators, Vol. 2, June 1997, pp. 1161 - 1164
[4] Kun Wang: Ark-Chew Wong; Nguyen, C.T.-C., "VHF free-free bearn high-Q micromechanical resonators", Journal of Microelectromechanical Systems, Vol. 9, No. 3, Sept. 2000, pp. 347-360
[5] Kipnis, I., "Refining CDMA mobile-phone power control", Microwaves and RF, Vol. 39, No. 6, June 2000, pp. 71 -76
[6] Fowler, T.; Burger, K.; Cheng, N.-S.; Samelis, A.; Enobakhare, E.; Rohlfing, S., "Efficiency improvement techniques at low power levels for linear CDMA and WCDMA power amplifiers", IEEE Radio Frequency Integrated Circuits
Symposium, June 2002, pp. 41 - 44
[7] Raab, F.H.; Asbeck, P.; Cripps, S.C.; Kenington, P.B.; Popovic, Z.B.; Pothecary, N.; Sevie, J.F.; Sokal, N.O., "RF and microwave power amplifier and transmitter technologies- part 5", High Frequency Electronics, January 2004, pp. 46-54
84
Chapter 7: References
[8] Leuzzi, G. and Micheli, C., "Variable-load constant efficiency power amplifier for
mobile communications applications", 33rd European Microwave Conference, Vol.
1, Oct. 2003, pp.375- 377
[9] Tu, S.H.L., "A power-adaptive CMOS classE RF tuned power amplifier for wireless
communications", IEEE International Proceedings on SOC [Systems-on-Chip} Conference, Sept. 2003, pp. 365 - 368
[10] Raab, F.H., "Electronically tunable class-E power amplifier", IEEE MFT-S
International Microwave Symposium Digest, Vol.3, May 2001, pp.1513- 1516
[11] Neo, W.C.E.; Liu, X.; Lin, Y.; de Vreede, L.C.N.; Larson, L.E.; Spirito, S.;
Akhnoukh, A.; de Graauw, A.; Nanver, L.K., "Improved hybrid SiGe HBT class-AB
power amplifier efficiency using varactor-based tunable matching networks", IEEE
Proceedings of the Bipolar/BiCMOS Circuits and Technology Meeting, Oct. 2005,
pp. 108- 111
[12] Coder, A.C. and Brown, E.R., "The feasibility of a variable output matching circuit in a high-power SSPA", IEEE Radio and Wireless Conference, Aug. 2002, pp. 189-
191
[13] Tu, S.H.-L. and Toumazou, C., "Design of highly-efficient power-controllable
CMOS class E RF power amplifiers", Proceeding of the I999 IEEE International Symposium on Circuits and Systems (ISCAS '99), Vol. 2, May- June 1999, pp. 602
-605
[14] Qiao, D.; Molfino, R.; Lardizabal, S.M.; Pillans, B.; Asbeck, P.M.; Jerinic, G., "An
Intelligently Controlled RF Power Amplifier With a Reconfigurable MEMSVaractor Tuner", IEEE Transactions on Microwave Theory and Techniques, ,Vol. 53, No. 3, March 2005, pp.1089 -1095
[15] Raab, F.H.; Asbeck, P.; Cripps, S.C.; Kenington, P.B.; Popovic, Z.B.; Pothecary, N.;
Sevie, J.F.; Sokal, N.O., "Power amplifiers and transmitters for RF and microwave",
IEEE Transactions on Microwave Theory and Techniques, Vol. 50, Issue 3, March
2002, pp. 814 - 826
[16] Raab, F.H.; Asbeck, P.; Cripps, S.C.; Kenington, P.B.; Popovic, Z.B.; Pothecary, N.;
Sevie, J.F.; Sokal, N.O., "RF and microwave power amplifier and transmitter technologies- partI", High Frequency Electronics, May 2003, pp. 22-36
[17] Pozar, D.M., Microwave Engineering, Wiley, 2005
[18] Raab, F.H.; Asbeck, P.; Cripps, S.C.; Kenington, P.B.; Popovic, Z.B.; Pothecary, N.;
Sevie, J.F.; Sokal, N.O., "RF and microwave power amplifier and transmitter
technologies- part 2", High Frequency Electronics, May 2003
85
Chapter 7: References
[19] Duclercq, J., "GSM base station power amplifier module linearity", Microwave Journal, Vol. 42, No. 4, April1999, pp. 116- 127
[20] Gonzalez, G., Microwave Transistor Amplifiers: Analysis and Design, Prentice Hall, 1997
[21] van der Heijden, M.P.; Spirito, M.; Pelk, M.; de Vreede, L.C.N.; Burghartz, J.N., "On the optimum biasing and input out-of-band terminations of linear and power efficient class-AB bipolar RF amplifiers", Proceedings of the 2004 Meeting of
Bipolar/BiCMOS Circuits and Technology, Sept. 2004, pp. 44 - 4 7
[22] Larson, L.; Asbeck, P.; Hanington, G.; Chen, E.; Jayamaran, A.; Langridge, R.; Xuejun Zhang, "Deviee and circuit approaches for improved wireless communications transmitters", IEEE Persona! Communications, Vol. 6, No. 5, Oct.1999, pp. 18-23
[23] ITS Electronics Inc., "Multiband linear power amplifiers", Microwave Journal, Vol. 41, No. 10, Oct. 1998, pp. 154- 158
[24] Ortega-Gonzalez, F.J., "High efficiency power amplifier driving methods and circuits: Part 1", Microwave Journal, Vol. 47, No. 4, April2004, pp. 22-38
[25] Kushner, L.J., "Output performance of idealized microwave power ampli fiers", Microwave Journal, Vol. 32, No. 10, Oct. 1989, pp. 103- 116
[26] Paidi, V.; Xie, S.; Coffie, R.; Moran, B.; Heikman, S.; Keller, S; Chini, A.; DenBaars, S.P.; Mishra, U.K.; Long, S.; Rodwell, M.J.W., "High linearity and high efficiency of class-B power amplifiers in GaN HEMT technology", IEEE
Transactions on Microwave Theory and Techniques, Vol. 51, No. 2, Feb. 2003, pp. 643-652
[27] Raab, F .H., "Class-E, class-C, and class-F power amplifiers based upon a finite number of harmonies", IEEE Transactions on Microwave Theory and Techniques,
Vol. 49, No. 8, Aug. 2001, pp. 1462- 1468
[28] Sowlati, T.; Salama, A.T.; Stich, J.; Rabjohn, G.; Smith, D., "Low-voltage, high efficiency GaAs class E power amplifiers for wireless transmitters", IEEE Journal of Solid-State Circuits, Vol. 30, No. 10, Oct. 1995, pp. 1074- 1080
[29] Sowlati, T.; Greshishchev, Y.; Salama, A.T.; Rabjohn, G.; Stich, J., "Linear Transmitter design using high efficiency class E power amplifier ", IEEE
International Symposium on Persona!, Indoor and Mobile Radio Communications,
Vol. 3, Sept.l995, pp. 1233- 1237
86
Chapter 7: References
[30] Grebennikov, A. V., "Circuit design technique fro high efficiency class F amplifiers", IEEE MTT-S International Microwave Symposium Digest, Vol. 2, June 2000, pp. 771 -774
[31] Sokal, N.O., "Class-E RF power amplifiers", QEX, Jan.- Feb. 2001, Pp. 9-20
[32] Sokal, N.O., "Class E high-efficiency power amplifiers, from HF to microwave", IEEE MMT-S International Microwave Symposium Digest, Vol. 2, June 1998, pp. 1109- 1112
[33] Sokal, N.O. and Sokal A.D., "ClassE- a new class ofhigh-efficiency tuned singleended switching power amplifiers", IEEE Journal of Solid-State Circuits, Vol. 10, No. 3, June 1975, pp. 168-176
[34] Cripps, S.C., Advanced Techniques in RF Power Amplifier Design, Artech House, 2002
[35] Grillo, G. and Cristaudo, D., "Adaptive biasing for UMTS power amplifiers", Proceedings of the 2004 Meeting of Bipolar/BiCMOS Circuits and Technology, Sept. 2004, pp. 188- 191
[36] Larson, L.E., RF and Microwave Circuit Design for Wireless Communications, Artech House, 1996
[37] Zhang, X.; Saycocie, C.; Munro, S.; Henderson, G., "A SiGe HBT power amplifier with 40% PAE for PCS CDMA applications", IEEE MMT-S International
Microwave Symposium Digest, June 2000, pp. 857- 860
[38] Webb, J., EM Transmission and Radiation (course notes), McGill University
[39] Cripps, S.C., "A theory for the prediction of GaAs FET load-pull power contours", MTT-S International Microwave Symposium Digest, Vol. 83, No. 1, May 1983, pp. 221-223
[40] Geis, L.A. and Dunleavy, L.P., "Power contour plots using linear simulators", Microwave Journal, Vol. 39, No. 6, June 1996, pp. 60-70
[41] Nellis, K. and Zampardi, P., "A comparison of bipolar technologies for linear handset power amplifier applications", Proceedings of the Bipolar/BiCMOS Circuits
and Technology Meeting, Sept. 2003, pp. 3 - 6
[42] Nellis, K. and Zampardi, P.J., "A comparison of linear handset power amplifiers in different bipolar technologies", IEEE Journal of Solid-State Circuits, Vol. 39, Issue 10, Oct. 2004, pp. 1746- 1754
87
Chapter 7: References
[43] Moniz, J.M., "Is SiGe the future of GaAs for RF applications?", ]9th annual
Gallium Arsenide Integrated Circuit (GaAs IC) Symposium, 1997, 12-15 Oct. 1997,
pp. 229-231
[44] Ali, F.; Gupta, A.; Higgins, A., "Advances in GaAs HBT power amplifiers for
cellular phones and military applications", Digest of Microwave and Millimeterwave Monolithic Circuits Symposium, June 1996, pp. 61 - 66
[45] Beckmann, S.; Sommet, R.; Nebus, J.-M.; Jacquet, J.-C.; Floriot, D.; Auxemery, P.;
Quere, R., "Characterization and modeling of bias dependent breakdown and self
heating in GalnP/GaAs power HBT to improve high power amplifier design", IEEE
Transactions on Microwave Theory and Techniques, Vol. 50, No. 12, Dec. 2002, pp.
2811-2819
[46] Halchin, D. and Golio, M., "Trends for portable wireless applications", Microwave Journal, Vol. 40, No. 1, Jan. 1997, pp. 62-78
[47] Johnson, J.B.; Joseph, A.J.; Sheridan, D.C.; Maladi, R.M.; Brandt, P.-O.; Persson, J.; Andersson, J.; Bjorneklett, A.; Persson, U.; Abasi, F.; Tilly, L., "Silicon-Germanium
BiCMOS HBT technology for wireless power amplifier applications", IEEE Journal ofSolid-State Circuits, Vol. 39, No. 10, Oct. 2004, pp. 1605- 1614
[48] Johnson, J.B.; Joseph, A.J.; Sheridan, D.C.; Maladi, R.M., "SiGe BiCMOS
technologies for power amplifier applications", IEEE 251h Annual Technical Digest,
Gallium Arsenide Integrated Circuits (GaAs IC) Symposium, 2003, pp. 179 - 182
[49] Krauss, H.L., Bostian, C.W., and Raab, F.H., Solid State Radio Engineering, Wiley,
1980
[50] Pusl, J.; Sridharan, S.; Antognetti, P.; Helms, D.; Nigam, A.; Griffiths, J.; Louie, K.; Doherty, M., "SiGe power amplifier ICs with SWR protection for handset
applications", Microwave Journal, Vol. 44, No. 6, June 2001, pp. 100- 13
[51] Lee, T.H., Planar Microwave Engineering: a Practical Guide to Theory, Measurement, and Circuits, Cambridge University Press, 2004
[52] El-Gamal, M. N., RF Microelectronics (Course notes), McGill University
[53] "SpectreRF User Guide", Cadence Design Systems,, Version 5.0, June 2003
[54] Dye, N. and Granberg, H., Radio Frequency Transistors: Princip/es and Practical
Applications, Butterworth-Heinemann, 1993
[55] Kazimierczuk, M. and Sokal, N.O., "Cause of instability of power amplifier with
parallel-connected power transistors", IEEE Journal ofSolid-State Circuits, Vol. 19,
No. 4, Aug. 1984, pp. 541-542
88
Chapter 7: References
[56] Allison, R., "Silicon Bipolar Microwave Power Transistors", IEEE Transactions on Microwave Theory and Techniques, Vol. 27, No. 5, May 1979, pp. 415-422
[57] Ming-Dou Ker and Bing-Jye Kuo, "ESD protection design for broadband RF circuits with decreasing-size distributed protection scheme", IEEE Radio Frequency Integrated Circuits (RFIC) Symposium, 6-8 June 2004, pp. 383-386
[58] Ma, Y. and Li, G.P., "InGaP/GaAs HBT RF power amplifier with compact ESD protection circuit", IEEE MTT-S International Microwave Symposium Digest, Vol. 2, June 2004, pp. 1173 - 1176
[59] "SiGeHP (BiCMOS 5HP) Design Manual", SiGe Technology Development, IBM Microelectronics Division, May 2001
[60] Mansour, R.R.; Bakri-Kassem, M.; Daneshmand, M.; Messiha, N., "RF MEMS Deviees", Proceedings of International Conference on MEMS, NANO, and Smart Systems, July 2003, pp. 103 - 107
[61] Bouchaud, J. and Wicht, H., "RF MEMS: Statues of the Industry and Roadmaps", IEEE Radio Frequency Integrated Circuits Symposium, June 2005, pp. 379-384
[62] Dec, A. and Suyama, K., "Micromachined Electro-Mechanically Tunable Capacitors and Their Applications to RF IC's", IEEE Transactions on Microwave Theory and Techniques, Vol. 46, No. 12, Dec. 1998, pp. 2587-2596
[63] Dussopt, L. and Rebeiz G.M., "Intermodulation distortion and power handling in RF MEMS switches, varactors, and tunable filters", IEEE Transactions on Microwave Theory and Techniques, Vol. 51, No. 4, April2003, pp. 1247- 1
[64] Tilmans, H.A.; De Raedt, W.; Beyne, E., "MEMS for Wireless Communications: 'From RF-MEMS components to RF-MEMS-SiP'", Journal of Micromechanics and Microengineering, No. 13, 2003,pp. 139- 163
[65] Ionis, G.V.; Dec, A.; Suyama, K., "A zipper-action differentiai micro-mechanical tunable capacitor", Microelectromechanical Systems Conference, Aug. 2001, pp. 29 -32
[66] Yoon, J. and Nguyen, T.C., "A high-Q tunable micromechanical capacitor with movable dielectric for RF applications", Electron Deviees Meeting, Dec. 2000, pp. 489-492
[67] Park, J.Y.; Yee, Y.J.; Nam, H.J.; Bu, J.U., "Micromachined RF MEMS Tunable Capacitors Using Piezoelectric Actuators", Microwave Symposium Digest, May 2001, pp. 2111-2114
89
Chapter 7: References
[68] Dec, A. and Suyama, K., "A 1.9 GHz micromachined-based low-phase-noise CMOS
VCO", ISSCC Digest Technical Papers, February 1999, pp. 80-81
[69] Nieminen, H.; Ermolov, V.; Nybergh, K.; Silanto, S.; Ryhanen, "Microelectromechanical Capacitors for RF Applications", Journal Micromechanics and Microengineering, No. 12, 2002, pp. 177 - 186
T., of
[70] Zou, J.; Liu, C.; Schutt-Aine, J.E.; Chen, J.; Kang, S., "Development of a wide tuning range MEMS tunable capacitor for wireless communication systems", Electron Deviees Meeting, 2000, pp. 403 - 406
[71] Chen, J.; Zou, J.; Liu, C.; Schutt-Aine, J.E.; Kang, S., "Design and modeling of a
micromachined high-Q tunable capacitor with large tuning range and a vertical
planar spiral inductor", IEEE Transactions on Electron Deviees, Vol. 50, No. 3,
March 2003, pp. 730-739
[72] Tsang, T.K. and El-Gamal, M.N., "Very wide tuning range micro-electromechanical capacitors in the MUMPs process for RF applications", IEEE Symposium on VLSI Circuits, June 2003, pp. 33 - 36
[73] Tsang, T.K. and El-Gamal, M.N., "Micro-electromechanical variable capacitors for RF applications", 451
h Midwest Symposium on Circuits and Systems, Vol. 1, Aug.
2002, pp. 25 - 28
[74] Wong, W.M.Y.; Ping Shing Hui; Zhiheng Chen; Keqiang Shen; Lau, J.; Chan,
P.C.H.; Ping-Keung Ko, "A wide tuning range gated varactor," IEEE Journal of
Solid-State Circuits, May 2000, pp. 773 - 779
[75] Varadan, V.K.; Vinoy, K.J.; Jose, K.A., RF MEMS and their applications, Wiley,2003
[76] Rebeiz, G.M., RF MEMS: theory, design, and technology, Wiley, 2003
[77] De Los Santos, H., RF MEMS circuit design for wireless communications, Artech
House, 2002
[78] Yao, J.J., "RF MEMS from a Deviee Perspective", Journal of Micromechanics and Microengineering, Vol. 10, No. 4, Dec. 2000, pp. R9- R38
[79] Borwick, R. L.; Stupar, P. A.; DeNatale, J.; Anderson, R.; Tsai, C.; Garrett, K., "A high-Q, large tuning range, tunable capacitor for RF applications," ]5th IEEE MEMS International Conference, Jan. 20-24, 2002, pp. 669- 672
[80] Borwick, R. L.; Stupar, P. A.; DeNatale, J.; Anderson, R.; Erlandson, R., "Variable
MEMS capacitors implemented into RF filter systems," IEEE Transactions on
Microwave Theory and Techniques, Vol. 51, Jan. 2003, pp. 315-319
90
Chapter 7: References
[81] Samavati, H.; Hajimiri, A.; Shahani, A.R.; Nasserbakht, G.N.; Lee, T.H., "Fractal capacitors", IEEE Journal ofSolid-State Circuits, Vol. 33, No. 12, Dec. 1998, pp. 2035-2041
[82] Seok, S.; Choi, W.; Chun, K., "A Novel Linearly Tunable MEMS Variable Capacitor", Journal of Micromechanics and Microengineering, Vol. 12, No. 1, Jan. 2002, pp. 82 - 86
[83] Cowen, A.; Mahadevan, R.; Johnson, S.; Hardy, B., "MetalMUMPs Design Handbook, Revision 2.0", MEMScAP Inc.
[84] Sun, Y.; Van Zejl, H.; Tauritz, J.L.; Baets, R.G.F., "Suspended membrane inductors and capacitors for application in silicon MMIC's", Microwave and Millimeter-wave Monolithic Circuits Symposium, June 1996, pp. 99 - 102
[85] Bakri-Kassem, M. and Mansour, R.R., "Two Movable-Plate Nitride-Loaded MEMS Variable Capacitor", IEEE Transactions on Microwave Theory and Techniques, Vol. 52, No. 3, March 2004, pp. 831-837
[86] Dec, A. and Suyama, K., "Microwave MEMS-based voltage-controlled oscillators", IEEE Transactions on Microwave Theory and Techniques, Vol. 48, No. 11, Nov. 2000, pp. 1943 - 1949
[87] Montrose, M.l., EMC and the printed circuit board: design, theory, and layout made simple, New York, IEEE Press, 1999
[88] German, R.F.; Ott, H.W.; Paul, C.R., "Effect of an Image Plane on Printed Circuit Board Radiation", IEEE International Symposium on Electromagnetic Compatability, August 1990, pp. 284-291
[89] Sadiku, M.N.O.; Musa, S.M.; Sudrashan, R.N., "Comparison of Dispersion Formulas for Microstrip Lines", IEEE Southeast Conf Proceedings, March 2004, pp. 378-382
[90] Robertson, C.T., Printed circuit board designer 's reference: basics, Upper Saddle River, NJ. Prentice Hall Professional Technical Reference, 2004
91
AppendixA
Appendix A: ADS Model Fitting
In this thesis, model fitting simulations in ADS are used to develop empirical models for
chip the package pins, the bondwires, and the PCB traces. The deviee under test is
characterized by means of a vector network analyzer (VNA), and the measured data are
imported into ADS as a black box with a given number of input and output ports. Next
step is to create a rough RLC model of the deviee under test based on its know behavior.
Each R, L, and C elements are assigned a variable, which is defined to have a certain
initial value within a practical range (Figure A.68).
The measurement equation is defined as:
dmSll = (abs(S(l,l))- S((2,2)))) (Equation A.l ),
where S(l, 1) is the one-port S-parameter corresponding to the RLC network and S(2,2) is
the experimental S-parameters corresponding to the actual deviee. The Goal of the
simulation (Figure A.69) is to set this difference to zero. In other words, the aim is to
vary the value of the RLC network components until the two sets of S-parameters are
matched as much as possible.
92
, L,
, L5û ,
. L-t..w.nH .
. R,
L=LinH R• .
R37'
AppendixA
· File="E:\Pbwér ,IX.mpli1ief'IPA mèasLirm'ent"reSultSI.oct19lpaCkaQe.S1 p"
: b oiuiW.ir:es 'L' 'L' A · L:'l · L51 f't36' . l=A nH . . L-t..w.nH. R=Rw Ohm . R, . Rë
Figure A.68: The black box and the empirical RLC model for characterizing the package pin, the bondwires and the PCB traces.
~IIAR ~-··.
test7 'Lt='0.827 {o} · 'Ct=0219{o} · ·Rt=0.315 {o} . .Lw;:::8 524!3-4 .{o}. Rw=0.321 {o} cp~O 1.69 {o} 'Lp=0458 {o} · · rp=6 .3 · { o} ·
. . . . . . . . . .
1 @.1 : S•PA~A~·E"FÊR$;·>11 .. : 1 ..... !!< _.'? .. :-PA .. '. L .. ·.i.._ .. · 1 . Zir.1 SP1
. Zit:~1
. Zin1=zin(S11..PortZ1)
start=·so ~~Hz . Stop='2 GHz. · Step=20 MHz
l"t~·j M_easEqn Meas1
· dr'nS11 = (abs((S(1 ;1)-8(2;2)))')
OptimG0a1·1 Expr="dmS11" SimlnstMceName="SP 1" Min: · Max=O. Weight=1 RangeVar[1]= RangeMin[1]= · RangeMax[1]=·
Figure A.69: The simulation setup for the model fitting scheme.
93