Gated tunnel diode pulse generator

Gated tunnel diode pulse generator
abstract book
GigaHertz Symposium 2008
5-6 March 2008
Chalmers University of Technology
Göteborg
Sweden
CENTRE
GigaHertz Symposium 2008
5-6 March 2008
Chalmers University of Technology
Göteborg
Sweden
www.ghz2008.se
Chalmers University of Technology
Department of Microtechnology and Nanoscience - MC2
GigaHertz Centre
Microwave Electronics Laboratory
SE 412 96 Göteborg, Sweden
Editor: Jan Grahn, Chalmers University of Technology
Technical Report MC2-125
ISSN 1652-0769
This Abstract Book is reported in Chalmers Publication Library:
http://publications.lib.chalmers.se/cpl/
This Abstract Bok is possible to download in pdf format at www.chalmers.se/ghz
CENTRE
www.chalmers.se/ghz
www.microwaveroad.se
www.chalmers.se/mc2
www.ieee.com
i
www.radiovetenskap.kva.se
Exhibitors & Sponsors GHz Symposium 2008
Platinum sponsor:
Swedish Governmental Agency
for Innovation Systems
(VINNOVA)
www.vinnova.se
Gold sponsor:
Agilent technologies
www.agilent.com
Silver sponsors:
Ageto MTT
www.agetomtt.se
Amska
AMSKA
www.amska.se
A m e r i k a n s k a Te l e p r o d u k t e r A B
Amtele
www.amtele.se
Anritsu
www.anritsu.se
Applied Wave Research
http://web.appwave.com
MicroComp Nordic AB
www.mcnab.se
National Instruments
www.ni.com
Ranatec Instruments
www.ranatec.se
Rohde Schwarz
www.rohde-schwarz.se
Wasa Millimeter Wave
www.wmmw.se
ii
GigaHertz Symposium 5-6 March 2008 at Chalmers
www.ghz2008.se
Chalmers Conference Center, Chalmers University of Technology
Wednesday 5 March 2008
0830-1000 Registration
Coffee + Sandwich
SILVER SPONSORS: Wasa Millimeter Wave AB
Applied Wave Research Inc.
1000-1200 SESSION I Runan
Chairman: Jan Grahn, Chalmers
1000 Welcome
Jan Grahn, Chalmers; General Chairman GHz Symposium 2008
Stefan Bengtsson, Vice President, Chalmers
1010 Plenary invited speaker
Intelligent Transmitter Technology for Next Generation Wireless Transceivers
Larry Larson
Univ. California San Diego
1050 Invited speaker
RF/DSP co-designed power amplifiers/transmitters for advanced wireless and satellite applications
Fadhel Ghannouchi
Univ. Calgary
1120
Tuneable technologies for agile microwave systems
S. Gevorgian
Chalmers, Ericsson
Design and Verification of a GaN S-band high power amplifier
J. Nilsson
Saab Microwave Systems
Self-Oscillating RF amplifiers
P. Reynaert, W. Laflere, M. Steyaert, J. Craninckx
KU Leuven, IMEC, Leuven
1200-1300 Lunch and Exhibition
GOLD SPONSOR: Agilent
1300-1500 Workshops Wednesday 5 March 2008
Agile Microwave Systems RF Power Amplifiers (1) Microwave
Components
Ascom/Catella
Runan
Scania
Moderator:
Hans-Olof Vickes
Saab Microwave Systems
Moderator:
Bo G. Berglund
Ericsson
Moderator:
Sven Mattisson
Ericsson Mobile Platforms
A method for switchable
rejection filters
The Frequency Spectrum
of Bandpass Pulse Width
Modulated Signals
Highly Integrated
MMICs for mm-wave
system application
T. Blocher, P. Singerl, A.
Wiesbauer, F. Dielacher
Graz Univ., Infineon
H. Zirath, S.E. Gunnarsson,
M. Ferndahl, R.
Kozhuharov, C. Kärnfelt
Chalmers, Ericsson
The potential of active
load and source tuning on
base station power
amplifiers
An Ultra Wide Band
LNA in 90 nm CMOS
N. Meissner
Saab Avitronics
60 GHz λ/8 Phase-Shifter
in EFFA Technology
X. Rottenberg, P. Ekkels, B.
Nauwelaers, W. De Raedt
Imec, KU Leuven
Tuneable Filters for Agile
Microwave Systems
A. Deleniv, S. Gevorgian
Chalmers, Ericsson
T. Lejon
Ericsson
Comparing Polar
Transmitter Architectures
using GaN HEMT Power
Amplifier
E. Cijvat, K. Tom, M. Faulkner,
H. Sjöland
Lund Univ., Victoria Univ.,
Melbourne
iii
THz Technology
Valdemar/Ledning
Moderator:
Staffan Rudner
Swedish Defence
Research Agency - FOI
Invited WS speaker:
An introduction to the
T4000 terahertz imager
C. Mann
Thruvision Ltd., Abingdon,
UK
W. Ahmad, A. Axholt, H.
Sjöland
Lund Univ
Flip Chip for
High Frequency
K. Boustedt
Ericsson.
Novel 220 GHz SlotSquare Substrate Lens
Feed Antenna
Integrated on MMIC
J. Svedin, S. Leijon, N.
Wadefalk, S.
Cherednichenko, B.
Hansson, S. Gunnarsson, I.
Kalfass, A. Leuther, A.
Emrich
FOI, Chalmers, FraunhoferIAF, Omnisys Instruments
Agile Microwave Systems
(cont')
Microwave MEMS
activities at the Royal
Institute of Technology
RF Power Amplifiers (1)
(cont')
J. Oberhammer, N. Somjit, M.
Sterner, F. Saharil, S. Braun, G.
Stemme
KTH
Phase-Comparison
Monopulse Direction
Measurement Antenna
Array for 6-18 GHz
Invited WS speaker:
Microwave
Components (cont')
Cryogenic X-band
Low Noise Amplifiers
THz Technology
(cont')
Planar antennas for
terahertz frequencies
Class MTM Power
Amplifier Linearization
N. Goia, M. Kelly, A.
Malmros, N. Wadefalk, J. P.
Starski
Chalmers
S. Cherednichenko
Chalmers
Low-Noise Cryogenic
Amplifier built using
MMIC-like /TRL
Technique
Geosynchronous Earth
Orbit Atmospheric
Sounder
D. E. Kelly
PulseWave RF, Austin
C. Johansson, T. Eriksson, J.
Grabs, T. Windahl
Saab Avitronics
MEMS Phase Shifters for
an Affordable Low-Power
Ka-band Multifunctional
ESA on a small UAV
R. Malmqvist, C. Samuelsson, A.
Gustafsson, T. Boman, S.
Björklund, B. Carlegrim, R.
Erickson, T. Vähä-Heikkilä, P.
Rantakari
FOI, Millilab-VTT
An adjustable broadband
MMIC equalizer
Different Classes of Power
Amplifiers using SiC
MESFET
S. Azam, R. Jonsson, Q. Wahab
Linköping Univ., FOI
Modeling of dual-input
power amplifiers
T. Eriksson, C. Fager, H. Cao, A.
Soltani, U. Gustavsson, H.
Nemati, H. Zirath
Chalmers
Invited WS speaker:
J. Grabs, U. Öhman, N. Meissner
Saab Avitronics
Tunable Impedance
Matching Network
M. R. Rafique, T. A. Ohki, P.
Linnér, A. Herr
Chalmers
Recent Advances in GaN
HEMT Power Amplifier
Technology for
Telecommunication
Applications
R. Pengelly, S. Wood, D. Farrell,
B. Pribble, J. Crescenzi
Cree Inc., Central Coast
Microwave Design, US
O. Nyström, E. Sundin, D.
Dochev, V. Desmaris, V.
Vassilev, V. Belitsky
Chalmers, Onsala Space
Observatory
Small-Signal Modeling
of Narrow bandgap
InAs/AlSb HEMTs
M. Malmkvist, E. Lefebvre,
L. Desplanque, X. Wallart,
G. Dambrine, S. Bollaert, J.
Grahn
Chalmers, IEMN Lille
Low-Noise, HighSpeed Strained
Channel Silicon
MOSFET Technology
for RF-Applications
B.G. Malm, J. Hållstedt, P.E. Hellström, M. Östling
Royal Institute of
Technology
Wideband Microstrip
90° 3-dB Two-Branch
Coupler with
Minimum Amplitude
and Phase Imbalance
D. Wang, M. Li, A. Huynh,
P. Håkansson, S. Gong
Nanjing Electronic
Equipment Institute,
Linköping University
Coded OFDM in Hybrid
Radio Over Fibre Links
J.F. Miranda, M. Gidlund
Univ. Gävle, Nera Networks
Equivalent Circuit of
Metamaterials with a
Negative Permeability
A. Rumberg, M. Berroth
Univ. Stuttgart
Design Consideration for
Varactor-Based Dynamic
Load Modulation
Networks
U. Gustavsson, B. Almgren, H.
Nemati
Ericsson, Chalmers
iv
S. Andersson, J. Embretsén,
A. Emrich, M. Ericson, M.
Hjort, J. Riesbeck, C.
Tegnander
Omnisys Instruments
Back-End Module
Demonstrator for
radio-astronomy
applications
J.L. Cano, B. Aja, E. Villa,
L. de la Fuente, E. Artal
Univ. Cantabria, Santander
ALMA Band 5 (163211 GHz) Sideband
Separating Mixer
B. Billade, I. Lapkin, A.
Pavolotsky, R. Monje, J.
Kooi, V. Belitsky
Chalmers, California
Institute of Technology
High Power Photonic
MW/THz Generation
Using UTC-PD
B. Banik, J. Vukusic, H.
Hjelmgren, H. Sunnerud, A.
Wiberg, J. Stake
Chalmers
An Ultra-Wideband
Six-Port transceiver
Covering from 3.1 to
4.8 GHz
Towards a THz
Sideband Separating
Subharmonic Schottky
Mixer
P. Håkansson, S. Gong
Linköping Univ
P. Sobis, J. Stake, A. Emrich
Chalmers, Omnisys
Instruments
Gated tunnel diode
pulse generator
HIFAS: HighPerformance fullcustom
Autocorrelation
Spectrometer ASIC
M. Nilsson, M. Ärelid, E.
Lind, G. Astromskas, L.-E.
Wernersson
Lund Univ.
A. Emrich, S. Andersson, M.
Hjort
Omnisys Instruments
1500-1530 Coffee and Exhibition
SILVER SPONSORS:
Anritsu
Ageto MTT
1530-1730 SESSION II Runan
Chairman: Piotr Starski, Chalmers
1530 Invited speaker
Extremely Low-Noise Amplification with Cryogenic FET’s and HFET’s: 1970-2006 (Where do we go from
here?)
Marian W. Pospieszalski
National Radio Astronomy Observatory, Charlottesville, VA
1600
560 GHz ft, fmax operation of a refractory emitter metal InP DHBT
E. Lind, A.M. Crook, Z. Griffith, M.J. Rodwell
Lund Univ., Univ. California Santa Barbara
Low phase-noise balanced Colpitt InGaP-GaAs HBT VCOs with wide frequency tuning range and small
VCO-gain variation
H. Zirath
GHz Centre, Chalmers, Ericsson
Feasibility of Filter-Less RF Receiver Front-End
S. Ahmad, N. Ahsan, A. Blad, R. Ramzan, T. Sundström, H. Johansson, J. Dabrowski, C. Svensson
Linköping University
Small-Size 2-10 GHz Radar Receiver Si-RFIC
H. Berg, H. Thieses, M. Hertz, F. Norling
Saab Microwave Systems
High frequency, current tunable spin torque oscillators: experimental characterization
S. Bonetti, J. Garcia, J. Persson, J. Åkerman
Royal Institute of Technology
N-coupling the capacity of wireless communication using electromagnetic angular momentum
B. Thidé
Swedish Institute of Space Physics, Uppsala
1730-1830 Visit (optional) MC2 Cleanroom or Microwave Labs, Chalmers (www.chalmers.se/mc2)
1900 Conference Dinner at Universeum (www.universeum.se)
Thursday 6 March 2008
0830- 1000 SESSION III
Runan
0830 Plenary invited speaker
The Next Wireless Wave is a Millimeter Wave
Chairman: Herbert Zirath, Chalmers
Joy Laskar
GeorgiaTech
0910 Invited speaker
High Frequency and Mixed Signal Design for Communication and Remote Sensing applications in
advanced technologies
Mehran Mokhtari
Teledyne Scientific
0940
MMIC design at G-band (140-220 GHz) including a 220 GHz Single-Chip Receiver MMIC with Integrated
Antenna
S.E. Gunnarsson, N. Wadefalk, M. Abbasi, C. Kärnfelt, R. Kozhuharov, J. Svedin, B.M. Motlagh, B. Hansson, S. Cherednichenko, I.
Angelov, D. Kuylenstierna, H. Zirath, S. Rudner, I. Kalfass, A. Leuther
Chalmers, FOI, Ericsson, Fraunhofer-IAF
A Quad-Core 130-nm CMOS 57-64 GHz VCO
V. P. Goluguri, J. Wernehag, H. Sjöland, N. Troedsson
Cambridge Silicon Radio Sweden, Lund University
SILVER SPONSORS:
1000-1030 Coffee and Exhibition
Amtele
AMSKA
v
MicroComp Nordic
1030-1210 SESSION IV Runan
Chairman: Niklas Rorsman, Chalmers
1030 Invited speaker
GaN HEMT development for microwave power applications- Current status and trends
Masaaki Kuzuhara
Univ. Fukui
1100
Paving the road for integrated gallium nitride transceivers
K. Andersson, M. Thorsell, N. Billström, J. Nilsson, J. Holmkvist, A-M. Andersson, M. Südow, M. Fagerlind, P.-Å. Nilsson, A.
Malmros, H. Hjelmgren, N. Rorsman,
GHz Centre, Chalmers, Saab
Demonstrator of Class-S Power Amplifier
A. Samulak, G. Fischer, R. Weigel
Univ. Erlangen-Nürnberg, Alcatel-Lucent
30/20 GHz Balanced Sub-harmonic MMIC Mixer for Space Applications
D. Kleén, J. Thelberg
Saab Space
Water Vapour Radiometer for ALMA
A. Emrich, M. Wannerbratt
Omnisys Instruments
European Radio & Microwave Interest Group (EuRaMIG) An initiative from GHz Centre: Status and Coming Activities
P. Olanders, Ericsson, Chairman GHz Centre,
J. Grahn, Chalmers, Director GHz Centre
1210-1300 Lunch and Exhibition
PLATINUM SPONSOR: VINNOVA
Antennas
Ascom/Catella
1300-1430 Workshops
RF Power Amplifiers (2)
Runan
Thursday 6 March 2008
Measurement - Modeling The GHz Entrepreneur
Scania
Valdemar/Ledning
Moderator:
Per Sjöstrand
Saab Avitronics
Moderator:
Johan Ståhl
Saab Microwave Systems
Moderator:
Niclas Keskitalo
Ericsson
Moderator:
Peter Wahlberg
Microwave Road
Integrated Antennas
for RF MEMS Routes
Output Power Density and
Breakdown Voltage in
Field-Plated Buried Gate
Microwave SiC MESFETs
Model-Based Predistortion for Signal
Generators
C. Luque, N. Björsell
Univ. Gävle
The WS provides some
personal reflections on
doing business from
innovations and IP in
RF/Microwave from three
small companies, one
global company and one
venture company. The WS is
concluded by a discussion
A Comparison of Antenna
Diversity Characterization
Methods using
Reverberation Chambers
and Drive Tests
Mikael Reimers, CEO
A. Rydberg, S. Cheng, P.
Hallbjörner, S. Ogden,
K. Hjort
Uppsala Univ., SP, Borås
Presented by C. Karlsson,
SP
Microstrip patch
antenna for wireless
applications
N.A. Touhami, B. Aja, A.
Tazón, E. Artal
Univ. Cantabria, Santander
P.Å. Nilsson, F. Allerstam, K,
Andersson, M. Fagerlind, H.
Hjelmgren, A. Malmros, M.
Südow, E.Ö. Sveinbjörnsson, H.
Zirath, N. Rorsman, Chalmers
Silicon-on-SiC hybrid
substrate with low RFlosses and improved
thermal performance
J. Olsson, Ö. Vallin, D. Martin,
L. Vestling, U. Smith, H.
Norström
Uppsala Univ., Infineon
Small Microstrip
Fractal Antenna for
RFID Tag
A review of validation
criteria for behavioral
power amplifier models
P. Enoksson, M. Rusu, A.
Curutiu, H. Rahimi, C.
Rusu
Chalmers, Bucharest Univ.,
Bonn Univ., Imego
P. Landin, M. Isaksson,
Univ. Gävle
D. Nyberg, M. Franzén, P.S.
Kildal
Chalmers, Bluetest AB
Measuring Relative
Receiver Sensitivity of
Wireless Terminals in One
Minute in a Reverbaration
Chamber
M. Andersson, C. Orlenius, M.
Franzén
Bluetest AB
vi
Foodradar Systems AB
www.foodradar.com
Tomas Ornstein, CEO
Ranatec Instrument AB
www.ranatec.se
Antennas
(cont')
Dual-band choke horn
Eleven Feed
A. Yasin, J. Yang,
T. Östling
Chalmers, Arkivator AB
RF Power Amplifiers (2)
(cont')
CMOS for micro- and
millimeter-wave power
applications
Measurement - Modeling
(cont')
Modeling of SiGe HBT
Operation in Extreme
Temperature Environment
The GHz Entrepreneur
(cont')
M. Ferndahl, H. Nemati,
H. Zirath
Chalmers
P. Sakalas, M. Ramonas, M.
Schroter, A. Kittlaus, H. Geisler,
C. Jungemann, A. Shimukovitch
Qamcom Technology
AB
www.qamcom.se
Comparative analysis of
the complexity/accuracy
tradeoff for power
amplifier behavior models
A. Tehrani, H. Cao, T. Eriksson,
C. Fager, Chalmers
Optimization of 200800 MHz Eleven
Feed for Use in
Reflector Antennas of
GMRT
Y.B. Karandikar, P.S.
Kildal
Chalmers
Method for Circuit
Based Optimization
of Radiation
Characteristics of
Multi-port Antennas
K. Karlsson, J. Carlsson
SP Borås, Chalmers
Circular Monopole &
Dipole Antennas for
UWB Radio Utilizing
a Flex-Rigid Structure
M. Karlsson, S. Gong
Linköping Univ.
Design, Manufacture
and Test of Eleven
Feed for 1-13 GHz
J. Yang, I. Karlsson, X.
Chen, P.S. Kildal
Chalmers
A Computational LoadPull Investigation of
Harmonic Loading Effects
on AM-PM conversion
O. Bengtsson, L. Vestling, J.
Olsson
Univ. Gävle, Uppsala Univ.
Johan Lassing, CEO
Dresden Univ.; Semiconductor
Physics Institute, Vilnius; SUSS
Microtec, Bundeswehr Univ.
Neubiberg; Univ. California San
Diego
On-wafer network
analyser uncertainty
estimation
J. Stenarson, K. Andersson, C.
Fager, K. Yhland
SP Borås, Chalmers
Peter Olanders,
Technology Strategist
Ericsson AB
www.ericsson.com
Identification of
Distortions in a RF
Measurement System
Mm-wave device testing
using wideband coplanar
transitions
Bengt Gustafsson, CEO
H. Cao, A. Soltani, C. Fager, T.
Eriksson
Chalmers
E. Villa, B. Aja, L. de al Fuente,
E. Artal
Univ. Cantabria, Santander
Microwave Technologies
AB
Further enhancement of
Load pull simulation
technique to study
non linear effects of
LDMOS in TCAD
High Efficiency using
Optimized SOI Substrates
L. Vestling, O. Bengtsson, K.-H.
Eklund, J. Olsson
Uppsala Univ., Univ. Gävle
A. Kashif, T. Johansson, C.
Svensson, T. Arnborg, Q. Wahab
Linköping Univ., Infineon, FOI
GaN device and MMIC
development at Chalmers
M. Fagerlind, M. Südow, K.
Andersson, A. Malmros, P.Å.
Nilsson, H. Zirath, N. Rorsman
Chalmers
Open debate and
discussion with
auditorium
Spin Torque Oscillator
Simulations and Circuit
Design
Y. Zhou, S. Srinivasan, J.
Persson, J. Åkerman
Royal Institute of Technology
1430-1500 Coffee
SILVER SPONSORS: ROHDE & SCHWARZ
National Instruments
1500-1600 SESSION V
Runan
Chairman: Arne Alping, Ericsson
1500 Invited speaker
Industrial aspects of 100 Gb/s optical communication
Bengt-Erik Olsson
Ericsson Research
1530
All-Optical Waveform Sampling with TeraHertz Capacity
M. Westlund, P.A. Andrekson, H. Sunnerud
Chalmers, Picosolve Inc.
High Speed 1.3 μm VCSELs for FTTH and RoF
P. Westbergh, E. Söderberg, J.S. Gustavsson, P. Modh, A. Larsson, Z.Z. Zhang, J. Berggren, M. Hammar
Chalmers, Royal Institute of Technology
1600
Closing Remarks
Jan Grahn, Chalmers; General Chairman GHz Symposium 2008
Henrik Sjöland, Lund University: Next GHz Symposium arranger
1630-1730 Visit (optional) MC2 Cleanroom or Microwave Labs, Chalmers (www.chalmers.se/mc2)
vii
Welcome to GigaHertz Symposium 2008
The GigaHertz Symposium is the Nordic meeting place for presenting new
findings in GHz technologies: Components, circuits and sub-systems for wireless /
wireline communication, and sensing. It is now time for the 9th GigaHertz
Symposium. This time, the event is carried out on 5-6 March 2008 at Chalmers
campus, Göteborg, Sweden. The arranger is Chalmers University of Technology.
At the GHz Symposium 2008, you can find presentations from a vast range of
R&D players extending from 1 GHz to 1 THz and beyond. This time, we are also
happy to present outstanding invited speakers who will present RF/microwave
communication and sensor technologies for applications ranging from intelligent
transmitters for wireless communication to cryogenic cooled receivers for radio
astronomy.
The Program Committee for GHz Symposium 2008 decided at its first meeting to
test some new concepts. The Proceedings have been replaced by the Abstract book
you now have in your hand. Furthermore, we have omitted the poster session and
introduced shorter presentations and parallel workshops. Finally, we involved
industry by letting distinguished company representatives lead the workshops with
dedicated "hot" GHz themes.
Chalmers is proud to announce that this probably is the largest GHz Symposium so
far with almost 270 delegates at the time of this writing. Around half of the
delegates are from industry representing around 50 companies. Almost 20% of the
delegates come from outside Sweden. The aim of the Program Committee to create
the GHz mixing zone between academia, research institutes, and companies seems
to have been fulfilled at GHz Symposium 2008.
I would like to thank in particular the invited speakers, and all contributors, to
make this GigaHertz Symposium 2008 happen. Many thanks to the Program
Committee, industrial workshop moderators, Chairmen, and Chalmers employees
for keeping this conference together.* I am particularly grateful to the GHz Task
Force at MC2 for all practical assistance. Finally, I thank our exhibitors for their
sponsoring; Take the time to meet them at the exhibition!
I hope you enjoy GigaHertz Symposium 2008 in Göteborg!
Jan Grahn
General Chairman
GigaHertz Symposium 2008
*
Chalmers employees carrying an orange badge are willing to answer questions and/or help our
visitors.
viii
Statistics GHz Symposium 2008
Participants at GHz Symposium 2008
25
4
15
129
Companies
Universites
Research institutes
Public Sector
Master students
94
Organisations at GHz Symposium 2008
2
8
Companies
Universites
Research institutes
Public Sector
19
52
ix
Program Committee GHz Symposium 2008
Dr. Jan Grahn
(General Chairman)
Prof. Anders Rydberg
Dr. Anders Sjölund
Dr. Gunnar Malm
Prof. Hans-Olof Vickes
Dr. Henrik Sjöland
Dr. Klas Yhland
Dr. Maria van Zijl
Prof. Niclas Keskitalo
Prof. Sheila Galt
Prof. Spartak Gevorgian
Prof. Staffan Rudner
Dr. Thomas Lewin
Chalmers University of Technology
Uppsala University
Swedish Foundation for Strategic
Research (SSF)
Royal Institute of Technology (KTH)
Saab Microwave Systems
Lund University and Ericsson Mobile
Platforms
SP Technical Research Institute of
Sweden
Business Security
Ericsson, and University of Gävle
Chalmers University of Technology
Chalmers University of Technology, and
Ericsson Research
Swedish Defence Research Agency
(FOI)
Ericsson Research, Ericsson
GHz Task Force at MC2, Chalmers
Jan Grahn
Jeanette Träff
Eva Hellberg
x
Invited Speakers GHz Symposium 2008
•
Prof. Larry Larson
Center for Wireless Communications
University of California at San Diego
Intelligent Transmitter Technology for Next Generation Wireless
Transceivers
•
Prof. Joy Laskar
Schlumberger Chair in Microelectronics,
Director Georgia Electronic Design Center
Georgia Tech, Atlanta
The Next Wireless Wave is a Millimeter Wave
•
Prof. Fadhel Ghannouchi
iRadio Laboratory
University of Calgary
RF/DSP co-designed power amplifiers/transmitters for advanced wireless
and satellite applications
•
Dr. Marian W. Pospieszalski
National Radio Astronomy Observatory
Charlottesville, VA
Extremely Low-Noise Amplification with Cryogenic FET’s and HFET’s: 19702006 (Where do we go from here?)
•
Dr. Mehran Mokhtari
Teledyne Scientific and Imaging Company
Thousand Oaks, CA.
High Frequency and Mixed Signal Design for Communication and Remote
Sensing applications in advanced technologies
•
Dr. Bengt-Erik Olsson
Ericsson AB
Ericsson Research, Mölndal, Sweden
Industrial aspects of 100 Gb/s optical communication
•
Prof. Masaaki Kuzuhara
Department of Electrical and Electronics Engineering University
of Fukui, Japan
GaN HEMT development for microwave power applications- Current status
and trends
•
Invited Workshop speakers:
THz Technology: Dr. C. Mann, Thruvision Ltd.: An introduction to the T4000
terahertz imager
RF PA(1): D. Kelly: Pulsewave RF, Austin: Class M Power Amplifier
Linearization
RF PA(1): Dr. R. Pengelly, Cree Inc.: Recent Advances in GaN HEMT PAs for
Telecom Applications
xi
Abstract Book Contents GHz Symposium 2008
Wednesday 5 March 2008
SESSION I 1000-1200
1
Workshops 1300-1500
Agile Microwave Systems
RF Power Amplifiers (1)
Microwave Components
THz Technology
7
18
27
38
SESSION II 1530-1730
48
Thursday 6 March 2008
SESSION III 0830-1000
56
SESSION IV 1030-1210
61
Workshops 1300-1430
Antennas
RF Power Amplifiers (2)
Measurement - Modeling
The GHz Entrepreneur
68
77
87
96
SESSION V 1500-1600
102
xii
Session I
1000-1200 Wednesday 5 March 2008
1
GHz Symposium 5-6 March 2008
Intelligent Transmitter Technology for
Next Generation Wireless Transceivers
Professor Larry Larson
Department of Electrical and Computer Engineering
University of California, San Diego
Abstract: Future generations of wireless communications are expected to place increasing
burdens on the efficiency and linearity of power amplifiers, due to the use of more complex
waveforms and wider bandwidths. This has implications for the design of transmitters in
portable/mobile devices as well as base stations and access points. Although semiconductor
device technology has made rapid progress in fundamental power amplifier transistor
performance, the use of digital signal processing techniques to optimize the linearity and
efficiency provides enormous potential for further improvement. This paper will summarize
the digital techniques that can be employed to improve the performance of wireless power
amplifiers.
2
GHz Symposium 5-6 March 2008
RF/DSP co-designed power amplifiers/transmitters for
advanced wireless and satellite applications
F. M. Ghannouchi, Fellow IEEE
Professor and iCORE/CRC Chair, Electrical and Computer Engineering Department
Director iRadio laboratory, University of Calgary, Calgary Canada
E: mail: fghannouchi@ieee.org
Abstract:
The wireless and satellite communications communities have always been looking for
power and -spectrum efficient amplification systems. The design of such power
amplifiers has to be considered closely together with the system architecture in order to
ensure optimal system level performances in term of linearity and power efficiency. This
implies the use of adequate transmitter’s architectures that convert the analog base
band information to architecture dependent amplifier driving signals such as sigma-delta,
EE&R, and LINC architectures. This talk layouts the principles behind software enabled
linear and highly efficient power amplifiers/transmitters sub-systems for multi-standard
and multi-band applications. Recent advances/trends and practical realizations will also
be presented and discussed.
3
GHz Symposium 5-6 March 2008
Tuneable technologies for agile microwave systems
S. Gevorgian
Department of Microtechnology and Nanoscience MC2, Chalmers University of Technology
spartak.gevorgian@mc2.chalmers.se
A brief review of tuneable technologies useful for development of components for agile
microwave systems will be presented. It includes a comparative analysis of mechanical,
electric, magnetic and optical tuning methods. Apart semiconductors (p-i-n diodes, varactors,
field effect transistors etc.) the report will focus on technologies not yet compatible with
standard semiconductor fabrication processes, such as micromechanical and nanomechanical
components (carbon nanotube, nanowires etc.), ferrite/ferromagnetic components, liquid
crystals, plasma, ferroelectric and piezoelectric (varactors, thin film bulk acoustic wave
resonators) components etc.
4
GHz Symposium 5-6 March 2008
Design and verification of a GaN S-band high power amplifier
Joakim Nilsson, joakim.u.nilsson@saabgroup.com
Saab Microwave System, SE-412 89 Göteborg
Summary:
This paper describes the design of a 2.7-3.3 GHz GaN power amplifier. For the design
two GaN-power bars with a total gate width of 10 mm each were used. The GaN
power bars has been developed by QinetiQ, UK, within a European co-project called
Korrigan [1]. The power bars were fitted into packages and characterized at Saab
Microwave Systems. For the device characterization, pulsed Load-Pull and
S-parameter-measurements were performed on the packaged power bars. The
matching networks were then designed using the measured data in Agilent ADS 2006.
Intro:
Future radar systems require more efficient power-amplifier-realizations. Meeting
these new requirements, is to use microwave power transistors in WBG-technology.
The use of microwave power transistors in WBG-material for power amplifiers in
radar systems, is a possible way. With the possibility of high supply-voltage, the
system power distribution will be more efficient, resulting in lower losses in voltageconversion. The high feeding-voltage for the power amplifiers leads to higher output
impedance and therefore simpler matching networks with lower losses, which
contributes to a higher system-efficiency. Other advantages with microwave power
transistors in WBG material are heat-hardiness and high power density. Typical
power- density values are 4 W/mm-gatewidth for a 30 W GaN-chip in comparison
with below 1 W/mm with Si or GaAs. That gives, for instance, greater possibilities to
design smaller MMIC-power amplifiers than with common technologies.
Results:
The measurements were performed under pulsed condition with gate-voltage and RFsignal pulsed with 50-200 µs pulse-length at 5-20 % duty cycle. The drain-voltage, at
the measurements, was varied between 25 and 35 V.
The power amplifier shows, with a drain-voltage at 25 V, an output power of more
than 50 W over the whole frequency band with PAE > 35 % and Gain > 10 dB. A
peak output power at 75 W was obtained at 3 GHz with a drain-voltage at 35 V.
That means a power density at nearly 4 W/mm-total gate width measured at power
amplifier level.
Conclusion:
With a peak output power at 75 W and a power density at about 4 W/mm, the power
amplifier shows very promising results for further progress within the Korrigan [1]
project and for European GaN-component development.
References:
[1]
Project number RTP 102.052 within EDA
5
GHz Symposium 5-6 March 2008
Self-oscillating RF amplifiers
P. Reynaert, W. Laflere, M. Steyaert and J. Craninckx*
Katholieke Universiteit Leuven, Belgium
Department of Electrical Engineering, ESAT-MICAS
*Imec, Leuven, Belgium
patrick.reynaert@esat.kuleuven.be
Although CMOS RF transceivers for wireless communications have become very
common, it somehow seems that CMOS RF power amplifiers have been left behind.
Achieving sufficient output power from a low-voltage CMOS technology is for sure a
challenge. Furthermore, the high peak-to-average power ratio of today's wireless
standards further impedes the design of CMOS PAs. Not to mention the challenges
that arises when the PA is integrated together with other RF and analog circuitry.
Several research groups have already demonstrated the feasibility of constantenvelope GHz CMOS PAs. These PAs do not allow linear amplification and can only
transmit one single output power level. But when a constant envelope PA is turned on
and turned off, the average output power at the fundamental frequency can be
modulated according to the duty-cycle. Such a burst-mode operation, driven by a
PWM or SD modulator, results in spectral tones at the PWM or SD clock frequency,
and these tones obviously need to be suppressed by the antenna filter. The major
advantage is that the constant envelope PA is either on or off, and thus always
operates at a high efficiency. As such, this technique allows the amplification of
amplitude modulated signals at high efficiency.
In this work, we propose a self-oscillating delta modulator to achieve high linearity
and to lower the spurious tone of the modulator. To demonstrate this technique, the
topology was implemented in 0.18um CMOS. The implemented solution can serve as
a power efficient CMOS pre-driver as it achieves up to 8.26dBm with an efficiency of
35%.
6
GHz Symposium 5-6 March 2008
Workshop
Agile Microwave Systems
1300-1500 Wednesday 5 March 2008
7
GHz Symposium 5-6 March 2008
A method for switchable rejection filters
Niklas Meissner (niklas.meissner@saabgroup.com) and Jan Grabs (jan.grabs@saabgroup.com)
Saab Avitronics, SE-175 88 Järfälla, Sweden.
A method is proposed for designing rejection filters that can be switched off. The usual way
for switching between passband and rejection band is by using switches and dual channels. With
this method a second mode for the filter is introduced. The two modes of the rejection filter can be
changed by a diode. With this dual mode filter the frequency passband will be very broad and
insertion loss will be kept low without the need for active components.
The development of communication and radar systems using high carrier frequency band for
signals is increasing. For applications operating at these frequencies there is a need for
suppression of these unwanted signals. Suppression is sometimes the only option for avoiding
interference when the power level is high. By introducing rejection filter in the receiver channel
they can be rejected. In some cases there are signals within the rejection band that you need to see.
This paper gives an example of a design where a method has been used for designing rejection
filters that can be transformed into transmission lines. Microstrip technology was used to realize
the filter.
Normally when designing rejection filters at microwave frequency stubs of λ/4 length are
used, where λ is the wavelength at the rejection frequency. Here resonators are used instead of
stubs for filtering so that the passband will be broadened. To enable the use of the same channel
for passband and rejection, low coupling between
filter resonators and the main waveguide is
needed for the passband mode. The filter
consists of a coupling line and a resonator
part. The method used introduces a diode
between the coupling line and the resonator.
The diode is wire bonded into the filter and
coupled through a capacitance to ground so it
can be DC biased into different modes. When
the diode is conducting the microstrip filter
will transform into a coplanar transmission
line.
In the example in this paper a microstrip
rejection filter with eight segments were
designed. The insertion loss parameter was
measured for the two modes. Here the signals
were rejected more than 45 dB in a 200 MHz
band around 4 GHz and the passband loss
was about 1 dB for the rejection mode and
2.5 dB for the passband mode from 1 to 19
GHz.
S21
0
-5
-10
-15
-20
-25
-30
-35
-40
-45
-50
1
2
3
4
5
6
7
8 9 10 11 12 13 14 15 16 17 18 19
Frequency (GHz)
A broadband microwave dual mode rejection filter was designed with promising results.
8
GHz Symposium 5-6 March 2008
60 GHz λ/8 Phase-Shifter in EFFA Technology
X. Rottenberg 1,2, P. Ekkels 1,2, B. Nauwelaers 2 and W. De Raedt 1
1
2
IMEC v.z.w., Division MCP, Kapeldreef 75, B3001 Leuven, Belgium
K.U.Leuven, ESAT, Kasteelpark Arenberg 10, B3001 Leuven, Belgium
xrottenb@imec.be
RF-MEMS technology is widely accepted as a key enabling technology for current and future telecom
applications. RF-MEMS devices indeed allow defining high quality, cost effective, highly integrated,
low power consumption and tunable passive components. These devices however often rely on fairly
complex and/or arguably expensive technologies. Recently, we developed a cheap and extremely low
complexity electrostatic RF-MEMS technology involving only two lithographic steps, i.e. EFFA
(Electrostatic Fringing-Field Actuator)[1]. The EFFA’s rely indeed on the fringing E-field lines for
their actuation and not on parallel-plate E-field as conventional devices.
In this paper, we present the design, realization in EFFA technology and characterization of λ/8 phase
shifters for 60GHz application as presented in Figure 1. As conventional EFFA’s [1] are too large to
be considered lumped at 60GHz, we designed a distributed phase-shifter taking advantage of their
electrical length. It consists of a tunable 4mm long multi-stage CPW (CoPlanar Waveguide) realized
in a 1µm thick Al layer partially suspended 3µm above an AF45 glass substrate. The RF-ground
consists of fixed semi-infinite planes and suspended strips defined 10µm from the signal line. This
narrow slot allows actuating the CPW by DC-biasing its signal line vs. its ground. The strip width is
designed so that the characteristic impedance of the line, Z0, remains close to 50Ω in idle and
actuated states while the total electrical length of the line is strongly modified [2].
The measurement results in Figure 2 show insertion and return loss (IL and RL) respectively better
than 2.3dB and 10dB at 60GHz. This IL corresponds to that of a fixed 4mm long CPW. The RL in
idle state is slightly above its design goal. The CPW is then too inductive. We attribute this
discrepancy to fabrication tolerances of the 10µm suspended slots. Finally, the λ/8 phase shift at
60GHz is perfectly defined. Linear in frequency, the phase shift is expected to reach λ/4 at 120GHz.
Figure 1 Top view microphotograph of the 4mm long phase shifter and one of its building blocks, along with schematic
cross-section of idle and actuated states.
(a)
(b)
(c)
Figure 2 Measurement results of the phase shifter of Figure 1 – (a) and (b) insertion and return loss in idle (grey) and
actuated (black) states (c) phase shift between idle and actuated states.
References
[1] X. Rottenberg et al., "An electrostatic fringing-field actuator (EFFA): application towards a lowcomplexity thin-film RF-MEMS technology", J. Micromech. Microeng. 17 (2007) S204-S210.
[2] D. M. Pozar, "Microwave Engineering", Addison-Wesley, 1993.
9
GHz Symposium 5-6 March 2008
Tuneable Filters for Agile Microwave Systems
A.Deleniv1, S.Gevorgian1,2
Department of Microtechnology and Nanoscience, Chalmers University of Technology, 41296 Gothenburg,
Sweden
2
Microwave and High Speed Electronics Research Canter, Ericsson AB, Mölndal, Sweden
1
A general approach is developed that allows optimisation of the performance of
the tuneable filters by trading between the tuning range and the losses. Tuneable resonators
based on tuneable ceramics, semiconductor and ferroelectric varactors are considered as the
main enabling technologies.
Tuneable filters based on ferroelectrics and semiconductor varactors will be
reported. As an example, Fig.1a shows the measured performance of a five-pole tuneable
filter using GaAs semiconductor diodes. This is developed for the application at 2GHz and
allows 200MHz (10%) tuning range. Another practical example of a tuneable dielectricwaveguide resonator is demonstrated at 15GHz. This uses thin film ferroelectric varactors,
which at these frequencies are distinctly better than their semiconductor competitors. The
measurement of the resonator revealed ∼120MHz tuneability and unloaded Q=530. To
demonstrate the potential of the ferroelectric varactors a conceptual design of tuneable dualmode dielectric-waveguide filter is developed with 15GHz centre frequency. The performance
of the filter, Fig.1b, obtained using HFSS demonstrates ∼240MHz tuneability and ∼3dB
insertion loss.
a)
b)
Fig.1 Performances of tuneable filters based on semiconductor (a) and ferroelectric (b)
varactors.
10
GHz Symposium 5-6 March 2008
GHz Symposium, Chalmers, March 5-6, 2008, ABSTRACT
Microwave MEMS activities at the Royal Institute of Technology
J. Oberhammer, N. Somjit, M. Sterner, F. Saharil, S. Braun, G. Stemme
Microsystem Technology Lab, School of Electrical Engineering
KTH-Royal Institute of Technology, 10044 Stockholm, Sweden
contact: joachim.oberhammer@ee.kth.se
This paper gives an overview on the ongoing RF and microwave MEMS activities in the Microsystem Technology
Lab (MST) at KTH, School of Electrical Engineering.
RF MEMS activities in this group are carried out since 2001, when the development of a special, S-shaped filmactuator based switch design was carried out, which combines the features of large DC and RF isolation in the offstate and very low actuation voltages of 12-15 V. This switch design has been adapted for constructing a switch array,
consisting of 400 double-pole-single-through microrelays integrated and encapsulated on a single chip of a size of
10x14 mm2, for the application of re-configurable telecommunication networks, to be employed in main distribution
frames of the copper-wire network infrastructure. A SEM picture and a schematic overview of an array reduced to a
2x2 DPST configuration, is shown in Figure 4.
A further RF MEMS switch project comprises a mechanically bistable switch mechanism, which is based on two
interlocking cantilevers (Figure 1), resulting in a switch with true zero-power consumption in the on-state and in the
off-state. Figure 1 also shows a SEM picture of a special variant of this switch, where the complete switch mechanism
is embedded inside the signal line of a coplanar waveguide, which, in contrast to conventional switch designs where
the actuator is built ontop of the waveguide and therefore disturbing the wave propagation, results in an extremely low
reflections and loss microwave switch of <0.1 dB switch insertion loss up to 20 GHz. Furthermore, this switch
concept is based on a 3D micromachined, 30µm thick coplanar waveguide, which results in extremely low insertion
loss, since the major part of the field lines is concentrated above, and not in, the substrate.
Within a NORDITE Scandinavian ICT programme project, novel RF MEMS components for the RF front-end of
(automotive) radar beam-steering units at 76-81 GHz are investigated. Figure 2 shows a passive, microelectromechanically tunable phase-shifter concept based on a moving dielectric block ontop of a transmission line,
suitable for handling signals of much higher power as compared to switched delay lines. Figure 3, finally, shows a
method of direct beam-steering by a micro-electromechanically tunable high-impedance surface, a tunable metamaterial structure developed in collaboration with Helsiniki University of Technology.
V1
signal
in
V1
V1
V0
signal
out
V0
V0
V 0=OFF V 1=OF F
V 0=ON V 1 =ON
1st mechanically
stable state:
switch is OFF
closing
opening
V 0=OFF V 1= OF F
2nd mechanically
stable state:
switch is ON
Figure 1
Figure 2
Figure 3
Figure 4
11
GHz Symposium 5-6 March 2008
Phase-Comparing Direction Measurement Antenna Array for 6-18 GHz
Christer Johansson (christer.v.johansson@saabgroup.com), Thomas Eriksson
(thomas.a.eriksson@saabgroup.com) , Jan Grabs (jan.grabs@saabgroup.com),Tomas Windahl
(tomas.windahl@saabgroup.com), all at Saab Avitronics, SE-175 88 Järfälla, SWEDEN.
Introduction A direction measurement antenna array aimed at high
accuracy measurements of the direction to an emitter by using a
single emitter pulse is designed. The antenna array uses 4+4 densely
Σ
Σ
packed spiral antennas for the direction measurement (see the figure
S1
S2
to the right) and another 4+4 spiral antennas for the ambiguity
resolution. One spiral antenna is used as guard antenna. The signal from each spiral antenna is
calibrated in phase as well as in amplitude. The direction of the main antenna lobe is electronically
steerable by means of phase shifters. The combined signals from 4 spiral antennas form the signal
S1 and the combined signals from the 4 neighbouring spiral antennas form the signal S 2 . From S1
and S 2 the phase-comparing monopulse method [1] is used to calculate the direction.
Theory The direction is calculated by the formula
(1)
⎛
θ measured = arcsin⎜⎜ λcarrier
sin(θ calibrated ) ⎞
arctan(Im(Δ / Σ ))
⎟,
+ λcarrier
Mπa
λcalibrated ⎟⎠
⎝
where Im(Δ / Σ ) is the imaginary part of the monopulse quotient, θ calibrated is the direction of the
main lobe for the calibration frequency c / λcalibrated , M is the number of antenna elements
combined in each channel (M=4 in this case), a is the spiral antenna element distance (a=33mm in
this case) and λcarrier is the carrier wavelength of the emitter. By using perturbation theory and
basic probability theory the expected standard deviation of the measurement error due to the phase
shift discretization can be derived from Eq (1) as
2
2
σ phase
shifter ⋅ λ carrier
2
σ θ = Var[θ measured ] = 2 2 3
(2)
,
2π a M ⋅ cos 2 (θ true )
where σ phase shifter is the standard deviation of the phase shift error and σ θ is the standard deviation
(due to the phase shift error) of the measured direction. Since 5-bit phase shifters are used the
phase error is uniformly distributed within ± 5.625°, giving σ phase shifter = 11.25 o / 12 ≈ 3.25 o . Eq.
(2) can be seen as the theoretical limit of the performace of a monopulse antenna array that is
phase calibrated and directionally steered with phase shifters.
Measurements The plots to the right show
Calibrated at f =7GHz. B =1 GHz.
Calibrated at f =17GHz . B =1 GHz .
45
measured angle
measured angle - true angle
meas ured angle - true angle
measured angle
50
examples of direction measurements when the main
antenna beam is pointing 40° off boresight and
40
40
when the polarization of the incomming wave is
30
35
known. The antenna was calibrated at the direction
30
40
50
35
40
45
40° and for this polarization. The upper row in the
true angle
true angle
1
1
figure to the right shows two plots of the measured
0.5
0.5
direction versus true direction, and the lower row
shows measurement error versus true direction. The
0
0
-0.5
-0.5
two leftmost plots show the result for f∈[6.5, 7.5]
GHz, and the two rightmost for f∈[16.5, 17.5] GHz.
-1
-1
35
40
45
38
39
40
41
42
true angle
true angle
Eq (2) predicts the measurement error σ θ to be
0.15° (at 7 GHz) and 0.06° (at 17 GHz). Hence, from the lower row it is seen that the
measurement error of the antenna array is close to the theoretical limit at 7 GHz, and about twice
the theoretical limit at 17 GHz.
Acknowledgements This work was financed by
Administration.
12the Swedish Defence
GHz Materiel
Symposium
5-6 March 2008
References [1]: Skolnik: Radar Handbook, Second edition, McGraw-Hill, 1990
MEMS Phase Shifters for an Affordable Low-Power Ka-Band Multifunctional ESA on a Small UAV
1
Robert Malmqvist, 1*Carl Samuelsson, 1Andreas Gustafsson, 1Tomas Boman, 1Svante Björklund, 1 Börje
Carlegrim, 1Roland Erickson, 2Tauno Vähä-Heikkilä, 2Pekka Rantakari
1
Dept. of Sensor Technologies, FOI Swedish Defence Research Agency, Linköping, Sweden
2
Millilab, VTT Technical Research Institute of Finland, Espoo, Finland
(*Presently with Saab Communication AB, Linköping, Sweden)
costly to realize since it requires an LNA and PA
behind every antenna element which will increase
the cost, especially for large arrays. For a more
affordable approach we consider instead a passive
ESA where a single LNA and PA are used behind a
certain subarray. Since commercially available Kaband phase shifters have too high losses we study
here the possibility to use MEMS phase shifters in a
Ka-band multifunctional ESA on a small UAV.
Abstract—We report on a study of a Ka-band
multi-functional electronically steerable antenna
(ESA) on a small UAV that is based on using
sub-arrays with low-loss RF MEMS phase
shifters. Our power estimations show that low
phase shifter losses are critical if the dissipated
radar hardware DC power should fit within the
given system requirements. Simulated results
further indicate it can be possible to achieve
adequate RF performance with a 4-bits 35 GHz
MEMS phase shifter design made on quartz.
SYSTEM ARCHITECTURE AND REQUIREMENTS
Beam steering will be needed for SAR/GMTI
(spot mode) as well as for S&A. Electronic scanning
is more flexible and adaptable to sharing the
apertures between multiple functions compared to
mechanical scanning. A conformal ESA could also
more easily fulfill the demands on low weight and
compact size for a small UAV. Meeting the power
requirements for the SAR function is found to be
limited by the available DC power for the radar
hardware (50 W is assumed for a small UAV). Lowloss phase shifters are then key elements in a passive
ESA where the number of LNAs and PAs can be
minimized using sub-arrays (and thus the cost).
II.
INTRODUCTION
Unmanned Airborne Vehicles (UAVs) can be
deployed to perform various missions such as
surveillance, reconnaissance and monitoring by the
use of imaging systems. Compared with optical
sensors, a radar sensor provides all-weather and allnight capabilities as it can function almost
independent of rain, snow, fog and the time of the
day. A small UAV equipped with a multifunctional
RF system needs to support different radar functions
as well as a data link. High resolution images of the
ground are obtained with synthetic aperture radar
(SAR). The ground-moving-target-indicator (GMTI)
mode makes it possible to also detect moving
objects on the ground. An on-board all-weather
Sense-and-avoid (S&A) capability will be needed
for UAVs to avoid collisions with flying objects.
This is also a requirement for UAVs to be allowed
to operate autonomously in non-segregated airspace.
It is also important to achieve low cost, size, weight
and DC power for the radar hardware. With the
technology of today, the easiest realization would
probably include mechanically scanned antennas.
However, for the future it is believed that advances
in phased array technology can facilitate multifunction and conformal antennas which are
integrated in the fuselage. An active electronically
scanned antenna is flexible but also complex and
I.
RF MEMS PHASE SHIFTERS
We have designed a 4 bits phase shifter which has
been sent to be fabricated at ICT Trento in Italy
using their RF MEMS process The two largest bits
were simulated using an EM-simulation tool.
According to simulations, the loss for the shortest
and longest delays (reference state and close to
360°) equals 1 dB and 3 dB at 35 GHz, respectively.
The average loss equals 2 dB. This compares
relatively well with a previously reported Ka-band
MEMS phase shifter, especially since that the phase
shifter design presented in this paper involves one
more bit. Measured results of the fabricated MEMS
phase shifter will be reported on at the conference.
III.
13
GHz Symposium 5-6 March 2008
An adjustable broadband MMIC equalizer
Jan Grabs (jan.grabs@saabgroup.com), Ulf Öhman (ulf.ohman@saabgroup.com)
and Niklas Meissner (niklas.meissner@saabgroup.com)
Saab Avitronics, SE-175 88 Järfälla, Sweden.
Introduction. Broadband GaAs amplifiers that are used in many microwave circuits have an
inherent drawback, the gain changes with temperature. This gain level variation can be handled by
temperature controlled attenuators but there is also a change in the frequency slope when the
temperature varies. An amplifier covering the 2 – 18 GHz frequency range can have a frequency
slope of about 2 dB within the temperature range of -55°C to +120°C. This can cause problems
when a number of amplifiers are used in the gain chain. E.g. a circuit with five amplifiers will
cause a frequency slope change of about 10 dB. A solution to this problem is to use an equalizer
with a positive frequency slope that changes with temperature.
The realisation of the equalizer is an MMIC chip made on GaAs. The circuit consists of a resistive
attenuator and two reactive elements, a capacitor and an inductor. These elements are used to
bypass the high frequencies and thereby creating a frequency slope. To adjust this slope two FETs
are used. These work as tuneable resistors where the resistance is changed by adjusting the bias
voltage. The circuit also consists of DC blocking capacitances and bias circuits.
Equivalent circuit of the equalizer
Results. The measured performance of the
broadband MMIC equalizer show that the
attenuation is adjustable between 4.5 dB and
9 dB at 2 GHz resulting in a tuneable
frequency slope change of about 4.5 dB. The
attenuation at 18 GHz is about 2.5 dB. The
picture to the right shows the normalized
gain with different bias voltage settings.
Normalized attenuation
0
S21-Ref (dB)
-1
-2
-3
-4
-5
-6
2
4
6
8
10
12
Frequency (GHz)
14
16
18
Data showing the improvement using this equalizer chip in a microwave circuit will also be
presented at the conference.
Acknowledgements. This work was partly financed by the Swedish Defence Materiel
Administration.
14
GHz Symposium 5-6 March 2008
GHz Symposium, March, 2008
Tunable Impedance Matching Network
M. R. Rafique, T. Ohki, P. Linner and A. Herr
Abstract—We present superconducting SQUID based tunable
impedance matching networks designed for highly miniaturized
filters. The performance of miniaturized filters is very sensitive to
parasitic and fabrication related spread. The presented
experimental results show that using tunable impedance matching
networks, the degraded performances of these types of filters is
improved.
-1-1
-2-2
Quasi lumped
C
Zf
|S 21| dB
-3
L
Zk
-4-4
Lumped
without
ground
-5
-6-6
-7
C
-8-8
Lumped with
ground
-9
I. INTRODUCTION
-10
-10
00
T
YC (YC + YL + Y f ) + YL (YC + Y f )
Filter type
Lumped filter
With ground
Lumped filter
Without ground
Quasi-lumped
filter
5
6
6
7
Passive
Tunable
Passive
Tunable
Passive
Tunable
L
pH
161
170
163
275
190
207
BW
MHz
5
3.5
28
18
110
75
ripple
dB
0.15
-
2.38
4.3
-
Losses
dB
8.7
5.9
2
1.46
4.75
1.12
f0
GHz
2.11
2.15
1.96
1.94
4.45
4.42
IV. CONCLUSION
SQUID based superconducting tunable impedance matching
networks for miniaturized microwave high-Q filters are
successfully demonstrated. The presented circuits are one of
the first examples that demonstrate the SQUID with a highly
miniaturized monolithic passive filter to overcome parasitic
and fabrication related spreads dependent performance
degradation.
(2)
where YL=1/(jωL), YC=jωC, Yf=1/Zf. From (1) and (2), at an
optimum matching of Zk =Zp at the filter's passband, f0, the
small shift in Zf, ∆Zf, can be compensated by tuning Icon
as ∆I con
5
TABLE I
COMPARISON OF FILTERS' PERFORMANCES
(1)
,
4
4
The developed tunable matching network has been
connected to three different superconducting filters presented
in [3]. The effect of the matching network tuning has been
tested by measuring transmission characteristic of the filters, at
4.2 K using an Agilent E8364B network analyzer. Fig. 2(b)
shows the measured maximum transmission, |S21|, at the
passband of all the filters. The measured current,
corresponding to the maximum inductance tunability is 4.1
mA. Table 1 summarizes the measured performance of the
filters with passive and tunable impedance matching networks.
where A, B and D are SQUID's parameter dependant constants.
From (1), for a required small ∆L, the needed ∆Icon is ∝ ∆L .
In the tunable impedance matching network, the integration
of series SQUIDs can be efficiently done by replacing the
inductance, L, of the lumped π-network (Fig. 1(a)). For a filter
impedance of Zf, standard microwave load impedance, Zk=50
Ω, and shunt capacitors of the π-network, C, the port
impedance, Zp, of the network can be expressed as follows
YC + YL + Y f
33
III. EXPERIMENTAL RESULTS
The total inductance of N SQUIDs in series with a common
Icon can be expressed as
Zp =
22
(a)
(b)
Fig. 1: Equivalent model (a) and measured passband loss (b) of the filters
as a function of control current of tunable impedance matching networks.
II. THEORY
N
,
L∝
A + B cos(DI con )
11
Icon mA
UNABLE impedance matching network using serial
SQUIDs as a tunable inductance is presented. A tunable
impedance matching network is an essential component of
high-Q miniaturized superconducting filters. A tunable
impedance matching network compensates for the inevitable
reflection losses at ports of the filter due to the parasitic and
fabrication related spread of parameters. Comparing the other
methods of inductance/capacitance tuning, SQUID based
tuning is easily achievable at low power applications, as it
needs no add-on layers or mechanical setups. From [1], the
overall inductance of a dc-SQUID can be varied by control
current, Icon. The SQUID based tunable impedance matching
networks, designed for Hypres 4.5 kA cm-2 process [2], are
theoretically and experimentally verified.
REFERENCES
∝ ∆Z f .
[1]
[2]
[3]
Raihan Rafique, Thomas Ohki, Peter Linner and Anna Herr are with the
department of Microtechnology and Nanoscience, Chalmers University of
Technology (corresponding author, phone: +46317721875; fax:
+46317723622; e-mail: raihan.rafique@mc2.chalmers.se).
15
T. A. Fulton, L. N. Dunkleberger, and R. C. Dynes, “Quantum
interference properties of double Josephson junctions,” Phys. Rev. B,
vol. 6, 1972.
Hypres Nb Design Rules. [Online]. Available: http://www.hypres.com/
R. Rafique, Towards Superconducting Monolithic Microwave
Integrated Circuits. Chalmers: PhD dissertation, 2008.
GHz Symposium 5-6 March 2008
CODED OFDM IN HYBRID RADIO OVER FIBRE LINKS
Juan F. Miranda M. and Mikael Gidlund*
Center for RF Measurement Technology, University of Gavle, Sweden
and Nera Networks AS, Norway, e-mail: juan_mm24@hotmail.com
Distributed antenna systems based on Radio over Fiber (RoF) has proofed to be the most
efficient solution to achieve in-building coverage in 3G systems [1]. Today 3G uses an
air interface based on WCDMA. However, Coded Orthogonal Frequency Division
Multiplexing (COFDM), has been selected in IEEE802.11a, WiMax and the long term
evolution of 3G (LTE) due to its flexibility and robustness in fading environments [2].
The objectives of the present investigation are therefore to examine the OFDM reliability
in the nonlinearities and multipath posed by the RoF link, and how its performance can
be improved by using representative linear block codes.
As a measure of reliability, the Bit Error Rate (BER) of the uncoded and coded OFDMRoF system for Uplink (UL) and Downlink (DL) was simulated. The system modeling
considers multipath fading from radio transmission [3] and Amplitude-to-Amplitude
(AM/AM) and Amplitude-to-Phase (AM/PM) distortions produced by the Electrical-toOptical (E/O) conversion and amplifying stages of the fiber link [4]. Several BCH and
Reed-Solomon codes were studied with different code rates [5], complexity and
interleaving depth. Their properties are related to system parameters such as Bit Energy
per Noise Density, Input Back-Off (IBO) and modulation order.
The Results show that using 4-QAM modulation with a moderate Input Back-Off (i.e. 410dB), BCH and small Reed-Solomon codes can reduce BER by a factor even greater
than 10 depending on the IBO. However, using 16-QAM modulation BER reduction is
poor even if Reed-Solomon is used with large number of bits per symbol. Future work
would consider using a more complex model for the E/O conversion and power amplifier,
more sophisticated codes such as Turbo or LDPC, and the use of nonlinear channel
estimation to further enhance system performance.
References
[1]
Hans Beijner, The importance of in-building solutions in third-generation
networks: Ericsson Review, Page(s): 90 – 97 2004
[2]
R. Prasad, OFDM for Wireless Communication Systems. New Jersey: Artech
House, 2004.
[3]
N. Chayat and Breezecom, "IEEE P802.11 Wireless LANs, Updated Submission
Template for TGa - revision 2, doc.:IEEE802.11-98/156r2," March 1998.
[4]
X. N. Fernando and A. B. Sesay, "Higher order adaptive filter based predistortion
for nonlinear distortion compensation of radio over fiber links," IEEE
International Conference on Communications ICC'00, New Orleans, June 20
[5]
S. B. Wicker, Error Control Systems for Digital Communication and Storage.
New Jersey: Prentice Hall, 1995.
This work was sponsored by Fiber Optic Valley and EU regional strategic fund under grant
X2.304-1135-06.
16
GHz Symposium 5-6 March 2008
Equivalent Circuit of Metamaterials with a Negative Permeability
A. Rumberg and M. Berroth
Institute of Electrical and Optical Communications Engineering, Universität Stuttgart,
Pfaffenwaldring 47, 70550 Stuttgart, Germany, axel.rumberg@int.uni-stuttgart.de
Abstract
An equivalent circuit of cut-wire pairs providing a negative permeability is presented. The working
frequency is around 10 GHz. The equivalent circuit precisely describes the behavior of the real
objects including the magnetic resonance.
Introduction
Veselago [1] investigated theoretically the properties of materials having a negative permittivity and
a negative permeability simultaneously describing them with a negative refractive index. In 2001
periodic split ring resonators (SRR) demonstrated the negative index in a metamaterial structure [2].
There are now several modifications of the SRR structure. One of the simplest is the so called cutwire pair (Fig.1) [3]. It can provide a negative permeability at microwave and optical frequencies.
Simulation, Experiment and Results
The wire pairs work with an LC-resonance. The incident electromagnetic field (Fig.1) excites a
current in one wire which is coupled to the other wire by the displacement field. A current loop is
generated. The excited magnetic field counteracts the incident magnetic field and can change its
direction. The dimensions are as follows: l = 8.5 mm, w1 = 0.7 mm, h = 10.16 mm, p = 5.715 mm,
t = 0.5 mm. The equivalent circuit describes the resonance of the single wires (LEL and CEL ) and the
magnetic resonance. The substrate RO4003C is modeled with LTL and CTL. The wire pair is coupled
to this transmission line by the mutual inductance M = F·LTL. F is the fraction of the unit cell
occupied by the wire pair. Fig. 2 shows the results. The simulation is done with the FEM solver
HFSS at normal incidence and in a waveguide with oblique incidence. The equivalent circuit
simulation shows good agreement with the normal incidence simulation. The permeability is
retrieved from the S-Parameters [4]. The measurement is done in a waveguide as here only small
pieces of substrate are needed. The HFSS simulation agrees well with the measurement. The overall
transmission is lower because of the oblique incidence. The magnetic resonance can be seen clearly.
Top View Cross Section
w1
LEL
E
h
LTL
t
l
p
k
CEL
H
CTL
M
LP
LEL
CEL
LTL
CTL
M
LP
CP
LEL
CTL
CEL
CP
=
=
=
=
=
=
=
3.3 nH
31.5 fF
0.625 nH
7.5 fF
0.625 nH
2.475 nH
100 fF
0
8
-5
6
0
HFSS
Circuit
-25
/ dB
HFSS
-20
0
10
11
Frequency / GHz
12
-10
-15
Sim.
Meas.
-2
-4
9
-5
2
Circuit
21
-15
4
S
-10
Permeability
S
21
/ dB
Fig. 1. Top view, cross section, measured piece and equivalent circuit of the cut-wire pair.
-20
9
10
11
Frequency / GHz
12
9
10
11
12
Frequency / GHz
Fig. 2. Transmission at normal incidence, retrieved permeability and waveguide measurement.
Conclusion
The equivalent circuit describes the transmission behavior of the cut-wire pair very well. Also the
retrieved permeability matches. The equivalent circuit provides a good understanding of the structure
and allows in this way a simple optimization.
[1] V. Veselago, Sov. Phys. Usp., vol. 10, pp. 509-514, 1968.
[2] R.A. Shelby, D.R. Smith, and S. Schultz, Science, vol.292, pp. 77-79, 2001.
[3] J. Zhou, L. Zhang, G. Tuttle, T. Koschny, C.M. Soukoulis, Phys. Rev. B, vol.73, p. 41101, 2006.
[4] X. Chen, T.M. Grzegorczyk, B.I. Wu, J. Pacheco, J.A. Kong, Ph. Rev. E, vol.70, p. 16608, 2004.
17
GHz Symposium 5-6 March 2008
Workshop
RF Power Amplifiers (1)
1300-1500 Wednesday 5 March 2008
18
GHz Symposium 5-6 March 2008
The Frequency Spectrum of Bandpass Pulse Width
Modulated Signals
Thomas Blocher1 , Peter Singerl2 , Andreas Wiesbauer2 and Franz Dielacher2
1
Christian Doppler Laboratory for Nonlinear Signal Processing
Graz University of Technology, Austria, E-mail: Thomas.Blocher@tugraz.at
2
Infineon Technologies Austria AG
xP (t)
V
I. I NTRODUCTION
Third generation radio transmitters generally uses bandwidth efficient modulation techniques to provide the required
high data rates. Because such modern modulation formats
generates signals with high peak-to-average power ratios,
linear power amplifiers (PAs) are needed to meet the spectral
requirements. Unfortunately, such linear power amplifiers have
very low efficiencies which directly affects the operational
costs of a radio basestation. Switched mode PAs have the
potential to increase the efficiency considerably. Because such
switched mode PAs can only be operated with a limited
number of different input signal amplitudes, the modulated
(amplitude and phase) RF signal must be modified accordingly.
One method is to encode the envelope and the phase of the
RF signal into a binary signal with different pulse widths and
pulse positions where the original spectrum can be recovered
with a simple bandpass filter.
F. H. Raab presented in [1] the bandpass pulse width
modulation (BP-PWM), which generates such binary signals.
The pulse train xp (t) in Fig. 1 depicts a BP-PWM encoded
signal. The binary signal possess exactly one pulse per carrier
period, where the widths of the pulses are proportional to the
envelope a(t) and the positions of the pulses depend on the
phase φ(t). The block diagram of the BP-PWM proposed by
Raab is shown at the top of Fig. 2, where the envelope is
compared with the 90◦ phase shifted and rectified carrier. A
pre-distortion with the arcsin function of the reference signal URef and the input signal does not change the modulator
output. If the trigonometric function arcsin is applied to the
rectified sine-wave reference signal, we obtain a triangular
signal. The BP-PWM block diagram with pre-distortion (as
used in [2]) is shown at the bottom of Fig. 2. If the reference
signal is not phase-modulated, the BP-PWM corresponds to a
PWM with pre-distorted input [3].
II.
RESULTS
We derive the BP-PWM output spectrum for sinusoidal
input signals with a two-dimensional Fourier transform. This
analytically derived spectrum is identical to the spectrum of
the PWM, where the input is pre-distorted and the PWM
reference frequency is replaced by the phase modulated reference signal. Numerical simulations have shown that BPPWM spectra contains the original signal plus an additional
noise. Fortunately, most of the distortion is distributed close
19
t
xRF (t) = a(t) sin(ωc t + φ(t)
a(t)
Fig. 1.
.
RF signal, its envelope and the corresponding BP-PWM output
a(t)
>0
|cos(ωc t + φ(t)|
URef
a(t)
arcsin
xP (t)
>0
φ(t)
URef
Fig. 2. Block diagram BP-PWM (top), BP-PWM with pre-distortion (bottom)
to the carrier harmonics far away from the frequency band
of interest. Due to the discrete-time nature of the proposed
modulator architecture we introduce some additional magnitude and phase noise which is primary caused by the time
discretization process of the binary signal. Therefore we show
the fundamental relationships between the sampling frequency,
carrier frequency and the dynamic range.
III.
CONCLUSION
Switched mode power amplifiers are suitable to increase the
efficiency of new generation radio transmitters considerably.
To drive such PAs, we need new modulation schemes which
are able to encode the amplitude and phase information e.g.
into binary signals. Bandpass PWM generates such signals
where the desired information can be recovered with simple
low-order bandpass filters. The drawback of bandpass PWM
is the short average pulse duration of the PA driving signals.
R EFERENCES
[1] F. H. Raab, “Radio frequency pulsewidth modulation,” Communications,
IEEE Transactions, vol. 21, no. 8, pp. 958–966, 1973.
[2] S. Rosnell and J. Varis, “Bandpass pulse-width modulation,” Microwave
Symposium Digest, 2005 IEEE MTT-S International, p. 4, 2005.
[3] Z. Song and D. V. Sarwate, “The frequency spectrum of pulse width
modulated signals,” Signal Process., vol. 83, no. 10, pp. 2227–2258, 2003.
GHz Symposium 5-6 March 2008
The potential of active load and source tuning on
base stations power amplifiers
Thomas Lejon, Ericsson AB
thomas.lejon@ericsson.com
Summary
This paper addresses the potential in active impedance tuning network for power
amplifiers used in base stations for cellular networks.
Abstract
The cellular networks nowadays need to operate in several frequency bands, this threatens
to generate a product portfolio with many variants because of the narrow bandwidth of the
power amplifier used in the transmitters. There is also a need for high efficiency to reduce
the operating cost (OPEX) of the network and large bandwidth for high data rates or multi
carrier scenarios. Because of this it is interesting to try to find solutions for an electrically
controlled low loss and low Q impedance transformation network that can handle the RF
voltage swing created by the amplifier. If we can create such a network it would open
possibilities to retune the amplifiers both in frequency and/or increase the efficiency in
back off at high modulation bandwidth. This paper will address the need for such a
network, the requirements of it, some challenges in creating it and also the potential if it
could be created.
Figure 1 Load modulated class F amplifier
20
GHz Symposium 5-6 March 2008
Comparing Polar Transmitter Architectures
using a GaN HEMT Power Amplifier
Ellie Cijvat1 , Kevin Tom2 , Mike Faulkner2 and Henrik Sjöland1
1
2
Dept. of Electrical and Information Technology, Lund University, Sweden
P.O. Box 118, SE-221 00 Lund, Sweden. e-mail: ellie.cijvat@eit.lth.se
Centre for Telecommunications and Micro-Electronics (CTME), Victoria University, Australia
P.O. Box 14428 MCMC, Melbourne 8001, Australia
Short summary - A power amplifier (PA)
with variable gate bias is compared to an
Envelope Elimination and Restoration
(EER) configuration. Each use the lowfrequency envelope and high-frequency
phase component of the signal. The test
circuit is implemented using a discrete GaN
HEMT power device. Measurements show
that the EER architecture maintains a
relatively high drain efficiency for a wide
output power range, while the PA with
variable gate bias shows a significant drop
in efficiency for lower output powers.
I. INTRODUCTION
The power amplifier is a critical component in
the transmitter. Many architectural solutions have
been investigated and implemented, such as
Envelope Elimination and Restoration (EER) and
Envelope Tracking (ET) [1]. On the circuit design
level, an increased interest in switched-mode
power amplifiers (PA) has been shown.
Power amplifier architectures suitable for
transmission of separated envelope (lowfrequency) and phase (high-frequency) signals
are compared. First, we discuss a power amplifier
with variable gate bias [3]. The PA is switchmode, and both the low- and high-frequency
signal operate on the gate of the device; The
phase signal thus causes the transistor to be
switched on for a shorter or longer period of time,
resulting in a pulse-width modulated signal at the
drain of the device, which is then filtered by the
output network, giving a variation in output
power. Unfortunately, this amplifier structure is
not linear, so linearization will be required.
This architecture is compared to a more
conventional polar transmitter architecture, that is,
EER. Both EER and ET architectures need a
supply modulator [1], commonly implemented as
a DC-DC converter or class-S amplifer followed
by a low-pass filter. However, this modulator is
not included in the design, nor in the efficiency.
II. RESULTS
A power amplifier was implemented with a
discrete GaN HEMT device [2]. In Fig. 1 the
output power and efficiency for the variable gate
21
bias and EER architecture, respectively, are
shown. The variable gate bias shows a maximum
output power of 29dBm, with a maximum drain
efficiency of 59%. However, the efficiency
decreases quite rapidly for decreasing output
power. For the EER architecture the efficiency
stays high over a large range of output power.
Efficiency
Efficiency
Pout
Pout
(a)
(b)
Fig. 1. Measurement results, (a). Output power and drain
efficiency as a function of gate bias voltage, with Vdd = 10V
and fin = 360MHz, (b). Output power and efficiency as a
function of varying supply voltage (EER), with Vbias = -2.5V
and fin = 360MHz.
III. CONCLUSIONS
The variable gate bias and EER transmitter
architectures were compared, using a power
amplifier implemented with a discrete GaN
HEMT device.
Measurements show an efficiency of 66 to
50% for the ideal EER architecture, for an output
power of 14 to 34 dBm, while the variable gate
bias PA has a slightly larger range of output
power but shows a significant drop in efficiency
for lower output power: from 6 to 59% for an
output power range of 3 to 29 dBm.
REFERENCES
[1] F. Wang et al., “An Improved Power-Added
Efficiency 19-dBm Hybrid Envelope Elimination and Restoration Power Amplifier for
802.11g WLAN Applications”, IEEE Trans.
on Microwave Theory and Techniques, Vol. 54
no. 12, pp. 4086-4099, December 2006.
[ 2 ] Cree Inc, CGH40010 product specification.
Available: http:// www.cree.com/products/
pdf/CGH40010-Rev1_4.pdf
[ 3 ] E. Cijvat, et al., “A GaN HEMT Power
Amplifier with Variable Gate Bias for
Envelope and Phase Signals”, in Proc. of the
25th Norchip Conference, November 2007.
GHz Symposium 5-6 March 2008
Class MTM Power Amplifier Linearization
David E. Kelly
PulseWave RF, Austin, TX 78746 USA
Vice President of Engineering – dkelly@pwrf.com – 001 (512) 437-2600; fax 001 (512) 437-2601
Abstract - An asynchronous noise-shaping modulator is
employed to convert high crest factor analog signals into a
data stream. This digital data stream can be directly applied
to a broad-band switch mode power amplifier wherein
linearity is accomplished via real-time RF feedback. The
same Class MTM system can also be applied in a plurality of
linear control systems such that it processes distortion signals
only. These methods can offer advantages in coding
efficiency and device utilization.
Traditional linear RF PA design begins with a class AB
output stage biased for the peak signal power it must
deliver. In order to improve PAE one, or more, of the
preceding methods is employed to facilitate operating the
device deeper into compression without sacrificing
linearity in the process. The limit case approach is hard
switching at the carrier frequency, as in the LINC method.
Both of the envelope methods listed involve amplitude
modulation of the output stage and are special cases of
switch-mode operation.
Index Terms — Class MTM is a proprietary linearization
technology developed by PulseWave RF.
I. INTRODUCTION
One subject of this paper is hard switching at the carrier
frequency up to, and including, the output stage. A noise
shaping modulator is employed to frequency translate and
encode a modulated analog carrier. The amplitude and
phase information from the modulated signal are
conveyed in pulse density fashion to a switching
amplifier. Non idealities in the switching amplifier are
mitigated through loop feedback. A sample of the output
is routed back through the modulator, in real-time,
wherein the effects of distortion and loop delay are
minimized. Note the simplified block diagram in figure 1.
This is a digital amplifier that encodes an analog input
signal and compares it to the output signal on a bit by bit
basis. The modulator provides correction signals to
maintain linearity as necessary.
For the past decade, perhaps longer, considerable effort
has been channeled into improving the power added
efficiency, PAE, of linear power amplifiers. Capital and
operating expenses, reliability, and radio talk time are but
a few of the motivations to improve PAE.
The modern digital cellular communications standards
are optimized to extract best utilization from the available
spectrum. Since spectrum is the scarce resource,
bandwidth efficiency is the first item on a large list of
challenges to address. The obvious solution for the
systems engineer is to increase modulation coding. Higher
data rate calls for higher spectral efficiency, which
translates to signals with higher peak to average ratios or
PAR. This in turn drives higher linearity requirements on
the transmit link, especially, the output power amplifier,
or PA. Several techniques have evolved, including novel
hardware design and digital signal processing, which
reduce the burden of high PAR on the PA. The short
summary list of these techniques reads something like;
1. Crest Factor Reduction
2. Analog Predistortion
3. Digital Predistortion
4. Linear Feed-forward Amplifier
5. Doherty Amplifier
6. Envelope Tracking
7. Envelope Elimination and Restoration (EER)
8. LINC (linearization w/ non linear components) or
out-phasing
Figure 1 – Class M Amplifier System
Additional examples of CLASS MTM amplifier
embodiments, along with their supporting laboratory
measurements, will be discussed in the full workshop
presentation.
22
GHz Symposium 5-6 March 2008
Different Classes (A, AB, C & D) of Power Amplifiers using SiC
MESFET
Sher Azam1, R. Jonsson2, Q. Wahab1, 2
1
Department of Physics (IFM), Linköping University, Linköping, Sweden
2
Swedish Defense Research Agency (FOI), Linköping, Sweden
ABSTRACT:
In this abstract we are presenting wide bandgape semiconductor transistors (SiC MESFET) in different
classes of amplifiers. The results are based on designing and fabrication of class-AB power amplifiers using
ADS and simulation of physical transistor structure (after optimization) in TCAD for class-A and switching
response in class -C & D power amplifiers.
1): Class-A Power Amplifier results
In class-A, The limiting frontier of DC voltage at the drain of an enhanced version of previously fabricated
and tested SiC microwave power transistor is studied at 1 GHz. An increase in the power dissipation
(W/mm) and decrease in PAE (%) with the increase in DC drain voltage is observed. The PAE (%) at
1 GHz, at Vdc= 50 V is 25.6 %, reduced to 16.8 % at Vdc= 70 V (34.4 % decrease). The power dissipation is
increased from 8.41 W/mm at Vdc= 50 V to 11.25 W/mm at Vdc= 70 V (36.36 % increase). Thus DC voltage
applied at the drain of SiC MESFET greater than 50 V is disadvantage in class-A power amplifier.
2): Class-AB Power Amplifier results
In class-AB, We designed and fabricated three power amplifiers (30-100 MHz, 200-500 MHz and 0.61.8 GHz).The designs are based on measured S-parameters. The 30-100 MHz amplifier showed 45.6 dBm
(~36 W) power at 1 dB compression (P1dB), at 50 MHz . The power added efficiency (PAE) is 48 % together
with 21 dB of gain. The maximum output power at 2 dB gain compression was 46.1 dBm (~41 W). One
power sweep was performed at a drain bias of 60 V, Vg= -8.5 V, and at this bias point the P1dB was 46.7
dBm (~47 W).
The typical results obtained in 200-500 MHz are; at 60 V drain bias the P1dB is 43.85 dBm (24 W) except at
300 MHz where only 41.8 dBm was obtained. The maximum out put power was 44.15 dBm (26 W) at 500
MHz corresponding to a power density of 5.2 W/mm. The PAE @ P1dB [%] at 500 MHz is 66 %.
Preliminary measured results obtained in 0.6-1.8 GHz amplifier are; S11 is below -4 dB, S22 below -9 dB
and gain is above 9 dB upto 1.2 GHz and above 7 dB for the rest of the band.
3): Class-C & D Switching Power Amplifier results
In class-C, the switching response of SiC MESFET is studied by applying square pulses of 5 % duty cycle
at the gate. The results are; efficiency of 71.4 %, power density of 1.0 W/mm with a power gain of 31 dB
and switching loss was 0.424 W/mm.
In class-D, a square pulse of 30% duty cycle is applied at the gate and at the same time a square pulse of
same duty cycle is applied at the drain (instead of applying sin wave) at 1 GHz. A Vdc voltage of 10 V is
applied at the drain to keep the transistor above turn ON. The results are; an efficiency of 49.3 %, power
density of 2.1 W/mm with a power gain of 29.3 dB is obtained. The switching loss is 2.15 W/mm.
The device shows resistive loss behavior at the crossing point of the drain current and voltage waveforms in
both amplifiers. It is due to the sudden increase in drain current during the turn off. The space charge
accumulation at the channel and buffer layer interface and near the gate on drain side and electron current
density in the channel at gate voltage pulse off and drain voltage pulse rise time (which is observed using
Tech plot software) is also found to be responsible for the switching losses and resulting reduction in the
efficiency.
23
GHz Symposium 5-6 March 2008
Modeling of dual-input power amplifiers
Thomas Eriksson†, Christian Fager, Haiying Cao, Ali Soltani†, Ulf Gustavsson, Hossein Nemati and Herbert Zirath
†Department of Signals and Systems
Department of Microtechnology and Nanoscience
Chalmers University of Technology, Göteborg, Sweden
I. I NTRODUCTION
constant. Now, by feeding the baseband path with an arbitrary
signal that can fully excite the amplifier, we can identify the
baseband path model. Second, we use a constant input at the
baseband path, and feed the RF path with the desired signal,
making identification of the RF model possible.
In many communications systems, most of the electrical
power is consumed by the final RF power amplifier (PA)
stage of the transmitter. Maximizing efficiency is therefore a
fundamental concern in design of such PAs. There has recently
been an increased interest in polar modulation architectures,
where the envelope of the communication signal is extracted
and separately used to amplitude-modulate the RF amplifier,
for example based on well-known techniques such as Envelope
Elimination and Restoration (EER) [1] or Envelope Tracking
(ET) [2]. Load modulation (LM), where the output signal
is modulated by dynamically altering some parameter in the
amplifier network, can also be used [3]. Polar modulation and
load modulation has the potential advantage of allowing the
RF amplifier to work in its most efficient mode (close to
saturation) at all times.
The above techniques have in common that they divide
the communication signal into two parts, an RF part and a
baseband modulation part, and feed the signals to different
inputs of the power amplifier circuitry. To acheive the highest
efficiency, the RF and modulation signals should be jointly
optimized. Such optimization requires good amplifier models
including two paths, i.e. dual-input amplifier models [4], which
is the main topic of this paper.
B. Approach 2: Combination by a static nonlinearity
The model above is, albeit simple, able to model all nonlinear and memory effects of both the RF path and the baseband
path. However, it cannot model conceivable nonlinearities in
the combination circuitry. To overcome this difficulty, one
must include nonlinearity in the combination of the two paths,
see Figure 2. With this approach, the output is given by
Baseband
input
Nonlinearity
with memory
(real)
Static nonlinearity
(complex)
RF input
Nonlinearity
with memory
(complex)
yn =
In our research, we study dual-input black-box amplifier
models of different accuracy and complexity. In this paper,
we study two approaches, as described in the following
subsections.
RF input
Nonlinearity
with memory
(complex)
xn
K X
L
X
ak,l |xn |2k xn rnl .
(1)
The identification is more involved compared to the previous
case. A working approach is to start similar to the identification above, by a simple initialization of the combination
model and separately identify the two paths, followed by
identification of the combination model. This must then be
iterated until convergence.
The simplest dual-input model consists of two single-input
models combined by multiplication, see Figure 1. With this
rn
xn
k=0 l=0
A. Approach 1: A simple combiner
Nonlinearity
with memory
(real)
yn
Fig. 2. A dual-input model with two single-input models combined with a
static nonlinearity.
II. O UR WORK
Baseband
input
rn
III. S UMMARY
In the final version of our paper, the performance of different
approaches will be compared, together with description of
identification and complexity of the models.
R EFERENCES
yn
[1] L. R. Kahn, “Single-sideband transmission by envelope elimination and
restoration,” Proceedings of the IRE, vol. 40, no. 7, pp. 803–806, 1952.
[2] A. A. M. Saleh and D. C. Cox, “Improving the power-added efficiency
of fet amplifiers operating with varying-envelope signals,” IEEE Trans.
on Microwave Theory and Techniques, vol. 83, no. 1, pp. 51–56, 1982.
[3] F. H. Raab, “High-efficiency linear amplification by dynamic load modulation,” Microwave Symposium Digest, 2003 IEEE MTT-S International,
vol. 3, pp. 1717–1720, 2003.
[4] J. C. Pedro and S. A. Maas, “A comparative overview of microwave
and wireless power-amplifier behavioral modeling approaches,” IEEE
Transactions on Microwave Theory and Techniques, vol. 53, no. 4, pp.
1150–1163, 2005.
Fig. 1. A simple dual-input model with two single-input models combined
with a multiplier.
model, the output is given by the simple equation yn = rn xn .
The identification is simple, since the two paths can be
identified separately. First, the RF path of the amplifier is fed
with a pure sinusoidal input, leading to xn being a complex
24
GHz Symposium 5-6 March 2008
Recent Advances in GaN HEMT Power Amplifier Technology for
Telecommunication Applications
Raymond Pengelly1, Simon Wood1, Donald Farrell1, Bill Pribble1
and James Crescenzi2
1 Cree Inc.
2 Central Coast Microwave Design LLC.
Abstract
The recent evolution of GaN HEMT transistors has opened up a range of power
amplifier architectures for the various telecommunication bands that have, in the
past, been limited to lower frequencies. Following a description of the
characteristics of wide bandgap technologies, the unique advantages of GaN
HEMTs applied to Class A/B, Doherty, drain modulated and switch mode
amplifiers will be detailed particularly in respect to efficiency and bandwidth.
Practical examples of Class A/B amplifiers will be compared with a number of
Doherty amplifiers employing digital pre-distortion with peak power handling up to
150 watts in the 2500 to 3500 MHz frequency range. The Doherty amplifiers
feature hundreds of megahertz of RF bandwidth with wide video bandwidths
accommodating multiple simultaneous signals.
Efficiency improvements with waveform engineered amplifiers including Class J,
Class E, Class F and inverse Class F, covering frequencies from 900 to 3500
MHz, will also be described.
The specific design approach for envelope tracking (drain modulated) power
amplifiers in the 2100 to 2700 MHz frequency range will be described where
peak power levels from 10 to 150 watts can be produced with average
efficiencies of greater than 60%.
Finally, the development of wideband efficient amplifier “modules”, covering 700
to 900 MHz, 1800 to 2200 MHz and 2100 to 2700 MHz, using harmonic matching
and novel power combining approaches will be described. These same modules
can be employed in a variety of configurations to further boost efficiency.
25
GHz Symposium 5-6 March 2008
Design Considerations for Varactor-Based Dynamic Load Modulation Networks
Ulf Gustavsson#*, Björn Almgren#, Hossein Mashad Nemati*
#
Ericsson AB, *Chalmers university of technology, Department of Microtechnology and nanoscience
I. INTRODUCTION
In modern wireless communication systems, one
of the key component regarding both performance
and energy efficiency, is the RF power amplifier.
Measures are therefore taken in order to meet the
requirements of high efficiency, high bandwidth
and re-configurability that these systems need.
II. OUR WORK
A high linearity varactor network suitable for
dynamic load modulation is presented in [1]. A way
of implementing this is suggested in [2] and [3].
While studying the capacitance voltage dependence
in combination with the applied RF voltage from
the amplifier, one quickly realises the problematic
situation. The problematic scenario will be
−M
illustrated using C (V ) = C0 ⋅ (1 − Vc φ ) to describe
the C-V characteristics of the varactor stack. By
applying the voltage of the amplifier as shown in
Figure 1, across the selected varactor topology with
C-V characteristics shown in Figure 2, one can
easily discover that the hyper-abrupt behaviour of
the C-V curve will not contribute to the decrease of
the said amplifier output voltage needed, thus
during parts of each RF cycle, one of the varactors
will be biased in a forward direction.
Figure 2 - Example of a hyper-abrupt varactor C-V
characteristic and the "wanted" C-V characteristic
(dashed line). Points of 1 and 3 pF are marked out.
Using a variable capacitor from Johansson tech.,
the output of the amplifier can be statically altered.
This method of verifying the controllability of the
amplifier is described in [2]. Simulations have
shown the need for linear varactors in order to
successfully implementing a dynamic load
modulation network.
III. CONCLUSIONS
As shown during this work and the work referred
to here, dynamic load modulation shows high
potential, but is in need of variable capacitive
elements with linear characteristics. This will also
decrease the non-linearity for which one needs to
compensate for as well as the amplitude needed to
supply the control voltage to the said varactor
network.
Varactor Voltage
22
20
18
16
REFERENCES
[1] K. Buisman, et al., "'Distortion free' varactor diode
topologies for RF adaptivity," IMS 2005, Long Beach,
CA, Jun. 2005, pp. 157-160.
[2] Lepine, Fabiene; Zirath, Herbert; Jos, Rik: “’A Load
Modulated High Efficiency Power Amplifier.”’
European Microwave Conference 2006 Manchester
[3] Almgren, Björn; “’Dynamic Load Modulation’”, Master
thesis in microwave engineering at Gävle university
14
1
2
3
4
5
6
7
Varactor Capacitance (pF)
Figure 1 – RF PA output voltage dependence of the
capacitive load.
26
GHz Symposium 5-6 March 2008
Workshop
Microwave Components
1300-1500 Wednesday 5 March 2008
27
GHz Symposium 5-6 March 2008
Highly integrated MMICs for millimeterwave system
application
Herbert Zirath1, 2 Sten E. Gunnarsson1, Mattias Ferndahl1, Rumen Kozhuharov1, Camilla Kärnfelt1
1
Chalmers University of Technology, Department of Microtechnology and Nanoscience
Microwave Electronics Laboratory, Göteborg, Sweden.
2
Ericsson AB, Microwave and High Speed Electronics Research Centre, Mölndal, Sweden.
Abstract — In order to reduce the manufacturing cost
for future 60 GHz products, a high integration level is
necessary. Recent results on mHEMT and pHEMT
multifunctional receiver/transmitters realized at Chalmers
University are reported. Multifunctional MMICs utilizing
single ended, subharmonically pumped, balanced and
single sideband mixers are reported. Other chipsets for
additional applications such as radiometers are reported as
well. Most recently, a 220 GHz integrated receiver utilizing
mHEMT technology was demonstrated.
Various receiver and a transmitter chips were recently
developed based on pHEMT and mHEMT processes
from WIN Semiconductors, with following applications
in mind
1
2
3
Fig. 1 Block diagram of the receiver MMIC
60 GHz Wireless LAN
53 GHz radiometer for earth observation
24 GHz radar chip
A block diagram of a typical receiver MMIC is shown
in Fig 1, with the corresponding chip photo in Fig. 2.
The chip measures 5.7 × 5 mm. The RX chip possesses
a 3 dB RF bandwidth of 8 GHz between 55 and 63 GHz
with an optimal conversion gain, GC,of 8.6 dB at 58
GHz. The image reject ratio is larger than 20 dB
between 59.5 and 64.5 GHz. Transmitter chip were also
developed and will be reported. Due to the general
architecture of the chipset any modulation format can be
used. A test bench for system tests was setup where
general modulation signals can be calculated and loaded
to the ESG. Due to the limitation in the measurement
setup, we used ASK for higher bitrates than 200Mbit/s.
This chipset was implemented in both a pHEMT and an
mHEMT version with similar RF-performance. The
power consumption was cut to 420 mW and 450 mW
for the mHEMT-based receiver and transmitter
respectively. A modified version of the 60 GHz receiver
was designed for a 53 GHz radiometer application. The
mixer has I-Q wideband intermediate frequency output
which is sampled by a high speed spectrometer
processor. The measured 3-dB bandwidth is 7 GHz.
Fig. 2 Photo of the receiver chip. (5.7 × 5.0 mm2)
Integrated receiver for 220 GHz radiometer
Multifunctional MMICs for frequencies 118, 183, and
220 GHz with a similar approach as the 60 GHz chipset,
are being developed at Chalmers University at the
moment, utilizing an mHEMT-process from IAF. In a
recent 220 GHz receiver design, the antenna was
integrated on ‘chip’. The receiver noise figure of the
MMIC-frontend is less than 10 dB. To the best
knowledge of the authors, this represents the highest
frequency of any multifuntional active MMIC.
28
GHz Symposium 5-6 March 2008
An Ultra Wide Band LNA in 90nm CMOS
Waqas Ahmad1, Andreas Axholt2, Henrik Sjöland2
Master’s student at the faculty of engineering, Lund university, sx06wa7@student.lth.se
2
Dept. of electrical and information technology, Lund university, Sweden
{Andreas.Axholt,Henrik.Sjoland}@eit.lth.se
1
Abstract— This paper presents a design of a
two-stage LNA in 90nm CMOS targeted for the
ultra wide band frequency range, 3.1GHz 10.6GHz. Post-layout simulations show a gain of
23dB±1dB, a noise figure of 4dB, input reflection
S11 below -10dB, and an IIP3 above -10.4dBm
over the entire band of interest, with a total
power consumption of 14.4mW from a supply
voltage of 1.2V. The circuit measures just
310x410µm2 including pads.
Fig. 1. Two-stage UWB low noise amplifier
III. RESULTS
The circuit was simulated using SpectreRF with
BSIM4 transistor models. The inductors were
modeled using Indentro [1]. The circuit consumes
14.4mW and measures 310x410µm2 including pads.
Simulated results are presented in Fig. 2. The circuit
is currently under fabrication.
I. INTRODUCTION
FCC allocated a 7.5GHz frequency band
Ifor2002
ultra wide band applications with restricted
N
transmission masks. Since then UWB has gained a
large momentum in both academia and industry.
This new wide frequency span allows for low cost,
high data rate, and low power short range
communications.
IV. CONCLUSION
A two-stage UWB low noise amplifier implemented
in UMC 90nm CMOS process has been presented
together with simulation results. This paper is the
result from a master’s thesis performed by Waqas
Ahmad at Lund University, Sweden [2].
II. CIRCUIT TOPOLOGY
The schematic of the UWB LNA is shown in Fig 1.
The input stage is realized using a common-gate
amplifier to achieve broadband input matching to
50Ω. The load resistor value R1 is chosen such that
the DC gate voltage of transistor M2 ensures that it
operates in the saturation region. Both noise figure
and input impedance are functions of gm1, hence a
compromise must be made in noise figure to fulfill
the S11 <- 10 dB requirement.
The second stage is a cascode common-source
stage with inductive load. To achieve wideband gain
in the second stage, a shunt feedback technique has
been adopted, implemented by capacitor C1 and
resistor R2 in Fig. 1.
REFERENCES
[1]
[2]
Niklas Troedsson, http://www.indentro.com
Waqas Ahmad, ”Front-End for Ultra Wide-Band (UWB) Wireless
Receivers in 90nm CMOS”
Fig. 2. Simulated gain, noise figure, and S11
29
GHz Symposium 5-6 March 2008
Flip Chip for High Frequency
Katarina Boustedt
Microwave and High Speed Research Center
Ericsson AB, Mölndal, Sweden
Katarina.boustedt@ericsson.com
Die interconnect through wire bonds is the most often employed process for high and lowfrequency chips. The die is soldered or adhesively joined to a carrier, and gold or aluminum wires
connect the die bond pads to corresponding pads on the carrier. When high-frequency operation
is considered, the wires cannot be disregarded electrically, since they cause significant parasitic
losses, primarily due to inductance, resulting in undesirable signal disturbances, unless
compensated for during the design phase. As a rule of thumb, one adds a 1-nH inductance for the
bond-wire effects, but this approximation is too coarse for sensitive cases.
Flip chip offers some very important benefits over wire bonding, for example better electrical
performance and a smoother assembly process. This facilitates reduction of production costs and
makes low-cost, high-volume production possible. Another considerable factor in choosing an
interconnect method for high frequency operation is the fact that the geometrical shape of flipchip interconnects is very predictable and repeatable. Bond wires vary in length, leading to
variations in parasitics, which makes compensation in the die design more difficult. The benefits
of flip-chip interconnect for lower frequency applications are well-known.
Microwave chips typically have much fewer interconnect pads than other dice. Therefore, the
need for flip chip does not typically come from a need for more I/Os per unit area, even though
the die surface area is normally smaller than typical silicon dice. An edge length of a few
millimeters is common. It has been stated that MMIC cannot be assembled on a chip carrier with
short interconnection lengths using an automated wire bond tool. Hence, a significant driver for
implementing flip-chip dice in high-frequency applications is the simplified chip assembly and
the improved performance stemming from the short and consistent lead lengths, i.e., the
interconnect bumps.
There are many variations deemed feasible when implementing flip-chip GaAs die for highfrequency applications. These are variations in materials and methods for fabricating carriers and
bumps, die design, techniques for simulation, and production.
The flip-chip interconnect builds on three fundamental elements, bumps on a chip, the chip
carrier or substrate, and the method of joining a die to a carrier. These building blocks are
interdependent, thus it is vital to take all of them into account, in order to select the optimal flipchip system for a given application. The possible use of an encapsulant will be another
consideration. Recent literature provides many variations for high-frequency flip-chip systems,
some of which are discussed in this presentation.
30
GHz Symposium 5-6 March 2008
CRYOGENIC X-BAND LOW NOISE AMPLIFIERS
Naiara Goia, Matthew Kelly, Anna Malmros, Niklas Wadefalk, and J. Piotr Starski
Chalmers University of Technology, Department of Microtechnology and Nanoscience
Microwave Electronics Laboratory, SE-412 96 Göteborg, Sweden
Abstract - This paper describes a number of cryogenic low noise amplifiers with very low noise for the
frequency band 8.4-8.5 GHz. The amplifiers have been designed with different inputs: standard coaxial and
waveguide
At 10 K ambient temperature the three-stage InP-based amplifiers have a gain of 33.0+/-0.5 dB and a noise
temperature of 5 -7K.
The InP transistors used in the amplifiers were processed at Chalmers clean room facility in our own
proprietary process.
INTRODUCTION
InP based High Electron Mobility Transistors (HEMTs) have extremely good low-noise performance and superior high frequency
performance. In this study we have designed, manufactured, and measured amplifiers with low noise and different input circuits.
Amplifiers with different input circuits are of interest for combinations with different types of isolators (coaxial/waveguide) to obtain
the lowest possible noise for a combination of the isolator/amplifier in cases where the input match of the amplifier is not sufficient.
The amplifiers are of a three-stage design in microstrip. The first stage is always equipped with an InP HEMT transistor processed at
Chalmers. The remaining transistors are commercially available GaAs based HEMTs from Mitsubishi.
HEMT FABRICATION AND CHARACTERISTICS
A lattice matched structure on InP grown by molecular beam epitaxy was used in the fabrication of the Chalmers HEMT. The
transistors have a 40Å AlInAs spacer, whose purpose is to separate the 300Å thick 53% InGaAs channel from the planar silicon
doping. We use a 200Å thick AlInAs Schottky barrier and a cap layer consisting of 50Å undoped InGaAs. The measured
transconductance at room temperature is close to 550mS/mm, which gives gm ≈110mS/200μm device.
AMPLIFIER DESIGN AND MEASUREMENTS
A three-stage, 8.4-8.5GHz, single-ended amplifiers with specified gain of 33dB were designed. We have designed, manufactured and
measured three amplifiers with different input circuits: standard coaxial, coaxial followed by a high directivity 30 dB microstrip
coupler, and waveguide. All amplifiers had coaxial output. Typical measured values for noise and gain at 15K are shown in Fig. 1,
the photograph of amplifier is shown in Fig.2.
Noise [K]
Gain [dB]
8.4-8.5 GHz LNA #1A @10K
With Chalmers MBE856 InP HEMT in first stage
Vd1=0.90V
Id1=7.00mA
Vg1=-0.40V 50
Vd2=1.50V
Id2=10mA
Vg2=-0.04V
Fig. 1. Measured noise and gain of
the amplifier with standard coaxial
input at 15K.
40
Vd3=2.00V
Id3=14mA
Vg3=-0.00V
Gain [dB], Noise [K]
30
20
10
Fig. 2. 8.4-8.5 GHz LNA
0
7
7,5
8
8,5
9
9,5
10
Frequency [GHz]
SUMMARY AND CONCLUSION
Recent work on cryogenically cooled HEMT-based amplifiers with different inputs has been presented. The lowest noise temperature
of around 5K has been achieved with an InP-based amplifier over the 8.4-8.5GHz band with an associated gain of 33+/-0.5dB. The
InP HEMTs have been manufactured at Chalmers.
31
GHz Symposium 5-6 March 2008
Low-Noise Cryogenic Amplifier
built using Hybrid MMIC-like / TRL Technique
O. Nyström, E. Sundin, D. Dochev, V. Desmaris, V. Vassilev, V. Belitsky
Group for Advanced Receiver Development (GARD), Department of Radio and Space Science
with Onsala Space Observatory, Chalmers University of Technology, Gothenburg, Sweden,
E-mail: olle.nystrom@chalmers.se
HEMT cryogenic low-noise amplifiers are an important part of instrumentation: the amplifiers
use as a front-end for different measurements and as IF amplifiers in heterodyne receivers.
During last few years the low-noise limit has reached as low level as approximately 0.5 K/GHz
for GaAs [1] and 0.25 K/GHz for InP HEMT [2]. However, besides electrical performance
improvement there were not many improvements on mass and dimension side of such
amplifiers as they were built based on standard TRL technology with discrete active and
passive components. Mass and dimensions are also very important for real applications. When
ultimate low-noise performance is placed in focus, pure MMIC technology seems to loose
against design using discrete components. With this in view, pioneered work by E. F. Lauria,
et. al. [3] have successfully demonstrated a design employing MMIC approach while using
discrete components and based on a microstrip on Cuflon with lumped bias network.
Encouraged by this work, we propose a compact design of a 4-8 GHz cryogenic low noise
amplifier using a combination of standard TRL and lumped element technology to achieve
both ultimate noise performance over the specified band and a very compact size. In our
design, the size reduction of the amplifier is realized by selecting an alumina substrate with a
high dielectric constant, (εr = 9.9), but also by taking advantage of the lumped networks in the
matching and bias circuitries. Avoiding quarter wave transformers and instead use a lumped
element design approach opens up for the possibilities to reach greater bandwidths and
simultaneously obtain a more compact design. In order to make optimum design, we have
performed extensive simulations. Each amplifier stage has been simulated in Agilent EMDS,
3D electromagnetic field simulation package, including the single layer capacitors, and then
implemented in the ADS circuit simulations as an S-parameter file. Over the 4-8 GHz band, the
simulations predict noise temperature, Taverage < 4.3 K, S11 < -12 dB, S22 < -15 dB, and a gain,
S21 > 35 dB. The transistors selected for the design are commercial InP HEMT (HRL) chosen
due to their excellent noise performance [2], but also for the very low power consumption,
which is of great importance at cryogenic temperatures. All the components used in the RFsignal path and in the bias circuits are mounted with conductive epoxy. Apart from the RFsignal path, all components are interconnected via bond-wires. Fine tuning is done by adjusting
the length and loop heights of the bond-wires. At the conference we plan to report results of
measurement and characterization of the prototype amplifier.
[1] C Risacher, et. al., “Low Noise and Low Power Consumption Cryogenic Amplifiers for
Onsala and Apex Telescopes”, Proceedings of Gaas 2004, October 2004, Amsterdam.
[2] N. Wadefalk, et. al., “Cryogenic Wide-Band Ultra-Low Noise IF Amplifier Operating at
Ultra-Low DC-Power”, IEEE Transactions on Microwave Theory and Techniques, vol. MTT51, no. 6 June 2003.
[3] E. F. Lauria, et. al., “A 200-300 GHz SIS Mixer-Preamplifier with 8 GHz IF Bandwidth”,
2001 IEEE International Microwave Symposium, Phoenix, AZ, May 2001.
32
GHz Symposium 5-6 March 2008
Small-signal modeling of narrow-bandgap InAs/AlSb HEMTs
Mikael Malmkvist1, Eric Lefebvre1, Ludovic Desplanque2, Xavier Wallart2, Gilles Dambrine2,
Sylvain Bollaert2, and Jan Grahn1
1
Microwave Electronics Laboratory, Department of Microtechnology and Nanoscience (MC2),
Chalmers University of Technology, SE-412 96 Göteborg, Sweden
2
Institute of Electronics, Microelectronics and Nanotechnology (IEMN),
P.O. Box 60069, 59652 Villeneuve d'Ascq, France
(mikael.malmkvist@chalmers.se)
The small-signal model (SSM) of 2×50 µm gate-width InAs/AlSb HEMTs with gate-length
variation from 225 nm to 335 nm has been investigated. To account for the relatively high gateleakage current IG in InAs/AlSb HEMTs, the conventional FET SSM has been extended by
shunting Cgs and Cgd with Rgs and Rgd, respectively. By utilizing this modeling approach the gatelength dependence of Cgs and Cgd was analyzed. When reducing the gate length from 335 nm to
225 nm, Cgs and Cgd were reduced by 40% and 35%, respectively.
Owing to its semiconductor properties and type-II energy-band alignment, the InAs/AlSb HEMT
exhibits superior electron mobility and carrier channel confinement. This makes the InAs/AlSb
HEMT suitable for high-frequency, low-noise and ultra-low power applications [1]. To be able to
evaluate and improve the device as well as to implement it in an MMIC, an accurate SSM is
required. Furthermore, to improve the HEMT figures-of-merit fT and fmax it is important to reduce
Cgs but also to keep a low Cgd/Cgs ratio.
On-wafer DC measurements revealed a transconductance gm as high as 1300 mS/mm at
VDS= 0.5 V and a gm above 900 mS/mm at VDS of only 0.3 V, as shown in Fig. 1(a). At VDS= 0.1 V
and VGS= -0.9 V an IG of 7 µA was measured, whereas at VDS= 0.5 V the value was as high as
210 µA. Due to this high IG the conventional SSM is insufficient. In Fig. 1(b), the measured and
modeled K is presented with and without shunting Cgs and Cgd with Rgs and Rgd, respectively. A
significant discrepancy is observed below 10 GHz when omitting Rgs and Rgd. By applying the
extended SSM, the Cgs and the Cgd behavior versus gate length were studied. Within the studied
gate-length interval, Cgs and Cgd were reduced by 40% and 35% respectively, with an increased
Cgd/Cgs ratio of 6%, see Fig 1(c). The different gate lengths were obtained through FIB-SEM
measurements, see inset in Fig. 1(c). If the gate length is further reduced, the improvement with
respect to fmax will not be as significant due to degradation in the Cgd/Cgs ratio.
InAs/AlSb HEMTs with gate-lengths ranging between 225 nm and 335 nm have been
investigated by extending the conventional SSM accounting for the elevated IG present in this
narrow-bandgap HEMT technology. As the gate length was reduced, Cgs was reduced by 40% and
the Cgd/Cgs ratio was only slightly increased warranting improved RF performance.
1400
0
2
120
Measurement
Model excluding Rgs & Rgd
Model including R & R
1200
−0.5
IG
−1.5
400
−2
Cgd
90
gd
1
60
gs
m
600
gs
C , Cgd [fF]
g
G
−1
800
I [mA/mm]
gm [mS/mm]
1000
Stability factor K
1.5
Cgs
0.5
225 nm
30
200
(a)
0
−1.2
−1
−0.8
V
GS
(b)
−0.6
[V]
−0.4
−2.5
−0.2
(c)
0
0
10
20
30
Frequency [GHz]
40
50
0
220
240
260 280 300
Gate length [nm]
320
340
Fig.1. 2×50 µm gate-width InAs/AlSb HEMT characteristics of (a) gm(VGS) and IG(VGS) with a 225 nm gate length at
VDS= 0.1, 0.3 and 0.5 V, (b) the influence on stability factor K with or without Rgd and Rgs shunting Cgd and Cgs,
respectively, and (c) the gate-length dependence on Cgd and Cgs (inset: FIB-SEM image of a 225 nm gate).
[1] B. Y. Ma, J. Bergman, P. Chen, J. B. Hacker, G. Sullivan, G. Nagy, and B. Brar, "InAs/AlSb HEMT and its
application to ultra-low-power wideband high-gain low-noise amplifiers," IEEE Transactions on Microw. Theory
Tech., vol. 54, no. 12, pp. 4448-4455, 2006.
33
GHz Symposium 5-6 March 2008
Low-Noise, High-Speed Strained Channel Silicon MOSFET Technology for
RF-Applications
B. Gunnar Malm, Julius Hållstedt, Per-Erik Hellström, and Mikael Östling
KTH - Royal Institute of Technology, School of ICT, Kista, Sweden, E-mail: gunta@kth.se
Strained channel silicon MOSFETs on virtual substrates (VS) are a promising technology for
high performance RF and analog applications. Significantly increased electron mobility can
be achieved by growing a tensile strained silicon layer on top of a relaxed Si1-xGex buffer
layer, where the composition x is in the range of 20-30%. The hole mobility is also improved
but to a smaller degree [1]. The high mobility improves transconductance as well as RFparameters such as cut-off frequency (fT) and power gain (fMAX), and the improvement is
maintained at deca-nanometer gate lengths. Furthermore, for high quality strained layers the
low-frequency (1/f) noise is preserved or lowered as compared to unstrained silicon
technology [1]. For devices operating at high current densities, such as a MOSFET biased for
maximum cut-off frequency, the poor thermal conductance of the SiGe buffer layer is a
potential drawback [2]. To solve this problem special epitaxial growth techniques have been
suggested to decrease the layer thickness and improve heat removal. We present results from
nMOSFETs fabricated using a sidewall transfer lithography technology [3], to achieve a
physical gate length of 50-80 nm, as shown in Fig. 1. Both thin and thick virtual substrates
have been investigated. Multifinger RF nMOSFETs demonstrate fT above 100 GHz, see Fig.
2. The high resistance of the 50 nm length/10 µm width gate fingers limits the fMAX, devices
with finger width of 5 µm showed fMAX close to 50 GHz. Devices fabricated with a so-called
supercritical thickness of the strained Si layer (110 nm) on thick 20% SiGe buffers show
excellent junction leakage [4] and the 1/f noise is almost identical to the silicon reference,
shown in Fig. 3.
This work was supported by SSF and Vinnova. We acknowledge our partners in the SiNano network (Univ. of Stuttgart and Warwick) for
epitaxial growth.
[1]
[2]
[3]
[4]
M. von Haartman, B. G. Malm, P. E. Hellström, M. Östling, et al, "Impact of strain and channel orientation on the low-frequency
noise performance of Si n- and pMOSFETs," Solid-State Electronics, vol. 51, pp. 771-777, 2007.
S. H. Olsen, E. Escobedo-Cousin, J. B. Varzgar, R. Agaiby, et al, "Control of self-heating in thin virtual substrate strained Si
MOSFETs," IEEE Transactions on Electron Devices, vol. 53, pp. 2296-305, 2006.
J. Hållstedt, P.-E. Hellström, Z. Zhang, B.G. Malm, et al "A robust spacer gate process for deca-nanometer high-frequency
MOSFETs," Microelectronic Engineering, vol. 83, pp. 434-439 2006.
J. Hållstedt, B. G. Malm, P.-E. Hellström, M. Östling, M. Oehme, J. Werner, K. Lyutovich, and E. Kasper, "Leakage current
reduction in 80 nm biaxially strained Si nMOSFETs on in-situ doped SiGe virtual substrates," presented at ESSDERC, Munich,
2007.
-16
1x10
Thin VS/ SiGe
buffer
fT Si ref.
60
fMAX VS1
40
fMAX Si ref.
-18
10
-19
10
strained
strained
strained
strained
Si REF
-20
10
-21
10
20
-22
1x10
0 -6
10
Fig. 1. XTEM image showing a
50 nm s-Si channel nMOSFET
on 200 nm thin SiGe virtual
substrate.
-17
10
2
NiSi
80
fT Si VS1
SID (A /Hz)
Strained
Si-channel
F requency (G H z)
100
-4
10
10
Drain current (A)
-2
Fig. 2. HF-characteristics of RF nMOSFET on a
silicon reference and thin VS, gate width (4 × 10
µm), VDS 1.0 V. The fT increase is about 40 %.
34
0.1
1
10
100
Drain current (µA)
Fig. 3. Drain current noise SID for strained channel
nMOSFETs LG 80 nm compared to unstrained
reference.
GHz Symposium 5-6 March 2008
Wideband Microstrip 90º 3-dB Two-Branch Coupler with Minimum Amplitude and Phase Imbalance
Duxiang Wang1, 2, Ming Li1, Allan Huynh2, Pär Håkansson2 and Shaofang Gong2
1
Nanjing Electronic Equipment Institute, P.O.Box 1610, Nanjing, 210007 Jiangsu, P.R.China
Phone: +86-25-84638544, Fax: +86-25-84498353, E-mail: wangduxiang@hotmail.com
2
Linköping University, Department of Science and Technology, SE-60174 Norrköping, Sweden
Phone: +46-11-363089, Fax: +46-11-363270, E-mail: allhu@itn.liu.se
A wideband 90º 3-dB directional-coupler is one of the key
components in a six-port transceiver circuit for wireless ultrawideband (UWB) or other communication systems. For the
Mode 1 band group of UWB communications, the frequency
band ranges from 3.1 to 4.8 GHz, i.e., a 43% relative
bandwidth. However, conventional two-branch 90º 3-dB
directional-couplers are typically limited to a relative
bandwidth of 10%, which does not fulfill the UWB
requirement. In this work, a new branch coupler has been
designed, manufactured and measured. Both simulation and
measurement results show that it has low insertion loss, and
small amplitude and phase imbalances in a relative bandwidth
of 59 % from 3.0 to 5.5 GHz.
0
Scale transmission S-parameters (dB)
-10
-20
-30
Measured S21
Measured S31
Simulated S21
Simulated S31
-40
-50
-60
-70
1
2
3
(c)
As shown in Fig. 1a this wideband microstrip coupler is
made of an ordinary two-branch coupler but with matching
networks at the 4 ports. The matching network, as shown in
Fig. 1b, consists of a half wavelength transformer and an open
stub to broaden the bandwidth of the conventional two-branch
coupler. This method was reported before by other authors, but
their focus has been on the amplitude balance. This work has
been focused on both amplitude and phase balances to cover
the 43% relative-bandwidth required by UWB. This twobranch coupler uses higher impedance line, with Z1 and Z2 to
be 1.36Z0 and 1.71Z0, respectively, to achieve the minimum
amplitude and phase imbalance in the frequency range of 3.05.5 GHz, as shown in Figs 1c and 1d, respectively.
4
Frequency (GHz)
5
6
7
200
Measured S21
Measured S31
Simulated S21
Simulated S31
150
Phase realtion(Degrees)
100
50
0
-50
-100
-150
-200
3
3.2
3.4
3.6
3.8
4
4.2
Frequency (GHz)
4.4
4.6
4.8
5
(d)
Port 1
Port 2
Port 4
Port 3
Fig 1. Branch coupler with match networks, having a relative
bandwidth of 59 % from 3.0 to 5.5 GHz.: (a) prototype, (b) matching
network, (c) amplitude balance, and (d) phase balance.
It is concluded that the conventional two-branch coupler has
a relative bandwidth of 10 %. However, with matching
networks connected to the four ports of the two-branch coupler
the bandwidth can be enlarged to a 59 % relative-bandwidth
within the frequency range of 3.0-5.5 GHz. The phase
imbalance must be considered when considering the amplitude
balance. This new coupler has low insertion loss, and both
small amplitude and phase imbalances in the frequency range
of 3.0-5.5 GHz, which exceeds the UWB requirement for the
Mode 1 band group 3.1-4.8 GHz. Therefore, this coupler can
be used in a UWB six-port transceiver design.
(a)
O p en
θ
Z2
θ
In p u t
Z0
To b ra n ch
co u p ler
Z1
(b)
35
GHz Symposium 5-6 March 2008
An Ultra-Wideband Six-port Transceiver Covering from 3.1 to 4.8 GHz
Pär Håkansson and Shaofang Gong
Linköping University, Department of Science and Technology, SE-60174 Norrköping, Sweden
Phone: +46-11-363089, Fax: +46-11-363270, E-mail: parha@itn.liu.se
Tx/Rx
Switch
SP4T
ZLx
I
Judgement
circuit
0,10
0,05
0,00
-0,05
-0,10
-0,15
-0,15
-0,10
-0,05
0,00
0,05
0,10
0,15
Q(V)
Fig. 3. Measured I/Q constellation diagram of a QPSK
signal at 3.96 GHz and a data rate of 1 Mbps.
TABLE I. MEASURED PROPERTIES OF THE TRANSCEIVER
flo = frf
3.432 GHz
3.96 GHz 4.488 GHz
Error Vector Mag (%RMS)
15.2
5.3
9.6
Mag. Error (%RMS)
8.6
4.6
8.6
Phase error(°)
7.6
1.56
2.3
Quadrature Error (deg)
-10
4,8
9.8
Gain imbalance (dB)
-2.1
0.4
0.5
RF
LP
Diode Filter
PA
Correlator
0,15
I (V)
This paper presents an ultra-wideband
transceiver based on the six-port technique. The
six-port transceiver covers the frequency spectrum
from 3.1 to 4.8 GHz, i.e., it covers the lower band
of the UWB spectrum. The transceiver has thus an
relative bandwidth of 43%. A key component in
the six-port transceiver is the correlator which
utilizes three ultra-wideband 3-dB 90º branch
couplers and one 3-dB 0º Wilkinson power divider.
It is manufactured utilizing microstrips on a
printed circuit board. It is shown that the
transceiver is able to demodulate high order
modulations such as 64-QAM and modulate QPSK
signals.
Q
LNA
Tx/Rx
Tx/Rx
LO
Fig. 3 shows the measured I/Q diagram when the
data is transmitted by the transceiver in the
transmit mode and received by the receiver
(another six-port receiver module). The power of
the transmitted signal is -12 dBm and the power of
the receiver LO is 3 dBm. Table I summarizes the
measured results at the center frequencies of the
three UWB sub-bands, i.e., 3.432, 3.96 and 4.448
GHz when the data is transmitted by the
transceiver shown in Fig. 2 and received by
another receiver. The quadrature error, i.e., the
phase error between the I- and Q- paths, ~10% at
the öower and higher bands. This can be improved
by further optimization of the six-port correlator.
The small phase and amplitude imbalances of
the ultra wideband correlator, i.e., 7° and 1.4 dB,
make it possible to produce a high quality RF
signal receiver and transmitter without using any
calibration technique in the UWB frequency band
3.1 – 4.8 GHz which has never been reported
before.
Transmit data
Fig. 1 Block diagram of the transceiver, the circuitry
within the dashed lines is implemented in this work.
Fig. 1 shows the block diagram of the six-port
transceiver. The received path consists of a low
noise amplifier (LNA), Tx/Rx switch, correlator,
SP4T switch, four radio frequency diodes, four
low-pass filters and a judgment circuit. The
judgement circuit is implemented utilizing two
instrumentation amplifiers. In the transmit mode
the SP4T switch is switched between the different
impedance loads ZLx shown in Fig. 1.
W1
W2
RF
Datain
SP4T Switches
LO
W3
W4
Fig. 2. Photo of the manufactured six-port transceiver.
36
GHz Symposium 5-6 March 2008
1
Gated tunnel diode pulse generator
M. Nilsson, M. Ärlelid, E. Lind, G. Astromskas, and L.-E. Wernersson.
Abstract
We demonstrate the function of a
gated tunnel diode as a high frequency pulse generator. The maximum oscillation frequency is 22
GHz and the shortest pulse length is 1 ns. The
output voltage swing delivered to a 50 ohm load is
100 mV. The technique is scalable and will be used
for ultra wide band communication experiments.
Collector current [mA]
18
16
V g=0.5V
14
12
10
8
V g=−1.5V
6
4
I. Introduction
2
HE negative dierential resistance (NDR)
property of resonant tunneling diodes
(RTDs) has been exploited in various high frequency applications such as oscillators and pulse
generators. These applications uses plain RTDs
integrated into an electrical circuit. Instead
we have integrated an RTD into a eld eect
transistor, the gated tunnel diode (GTD). The
device IV-characteristics is shown in gure 1,
for further details regarding the GTD refer to
[1]. By integrating the GTD in parallel with an
coplanar resonance circuit, an oscillator circuit
has been constructed. The possibility to tune the
resonance frequency with the gate and collector
bias to form a voltage controlled oscillator (VCO)
is one of the many features of the GTD oscillator.
0
0
T
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Collector voltage [V]
Fig. 1.
IV-characteristics of the GTD. The gate voltage is
increased in steps of 0.5 V.
25
600
20
Output voltage [mV]
10
400
5
300
0
200
−5
−10
Input voltage [mV]
500
15
100
−15
0
−20
−4
−3
−2
−1
0
time [ns]
1
2
3
4
II. Results
Fig. 2.
When biased in the NDR region the oscillator
generates a tunable -10 dBm signal, the highest
frequency being 22 GHz. The main application
of the circuit is not to use it as a continuous oscillator but rather as a pulse generator. This is
motivated by the recent development toward low
power ultra wide band (UWB) wireless communication, and especially impulse radio (IR) UWB.
Pulse generator operation is achieved by biasing
the device in the NDR region and then applying a
square wave to the gate. The GTD is then forced
out o its NDR region and instead enters a region
of positive dierential resistance. This switches
the oscillations o, gure 2.
Output from the pulse generator, the 16.4 GHz
output signal is down converted with a 13.1 GHz signal.
Due to the bias stabilization network the bias conditions
are 1.9 V and 377 mA.
III. Conclusions
We have constructed a pulse generator using a
gated tunnel diode. The pulse generator shows
promising features such as the possibility to generate very short pulses at a high frequency. So
far the output pulse length is limited by the input pulse length. The main goal is to construct a
wireless IR-UWB system for the 60 GHz regime
capable of transmitting at least 3 Gbps.
References
M. Nilsson, E. Lind, G. Astromskas and L.-E. Wernersson are
with the Department of Solid State Physics, Lund University,
Lund. M. Ärlelid is with the Department of Electroscience,
Lund University, Lund, (e-mail: Mikael.Nilsson@ftf.lth.se).
[1] E.Lind, P. Lindström, L.-E. Wernersson. Resonant tunneling permeable base transistor with high transconductance,
IEEE Electron Device Lett. 25, 678 (2004),
37
GHz Symposium 5-6 March 2008
Workshop
THz Technology
1300-1500 Wednesday 5 March 2008
38
GHz Symposium 5-6 March 2008
Introduction to the T4000 Passive Terahertz Imager
Chris Mann, Chief Technical Officer
ThruVision Ltd, 17-18 Central 127, Milton Park, Abingdon, Oxon OX14 4SA
chris.mann@thruvision.com, www.thruvision.com.
Passive millimetre wave imaging is now an established and accepted technology that is
finding viable commercial applications in many areas, particularly security and border
control. The upper frequency of operation has largely been governed by the availability
of solid state uncooled detectors to around 100GHz. Passive operation at higher
frequencies potentially offers some unique features such as higher optical resolution for a
given system size, increased depth of focus and improved material contrast. However, the
technological challenges involved in realising arrays of terahertz detectors with the
required sensitivity, packing density, repeatability and reliability are considerable.
ThruVision’s T4000 passive imager incorporates such arrays which are combined with
proprietary optics and scanning. The imager has some unique capabilities which make it
highly attractive for emerging security and civil applications such as the detection of
concealed contraband and firearms at a remote distance. This talk provides an
introduction to ThruVision’s products and associated technology.
39
GHz Symposium 5-6 March 2008
A Novel 220 GHz Slot-Square Substrate Lens Feed Antenna
Integrated on MMIC
J. Svedin1, S. Leijon1, N. Wadefalk2, S. Cherednichenko2, B. Hansson2, S. Gunnarson2,
I. Kalfass3, A. Leuther3, and, A. Emrich4
1
Dept. of Microwave Technology, Swedish Defense Research Agency (FOI), SE-581 11 Linköping, Sweden
Dept. of Microtechnology and Nanoscience, Chalmers Univ. of Technology, SE-412 96 Göteborg, Sweden
3
Fraunhofer Institute for Applied Solid State Physics (IAF), Freiburg, Germany
4
Omnisys Instruments AB, SE-421 30 Västra Frölunda, Sweden
Email: jan.svedin@foi.se
2
radiation and feeds the slot-square on-chip antenna through a
square opening in the ground plane of the MMIC. The
antenna was derived from a co-planar waveguide (CPW) fed
slot-square antenna by moving the ground plane to the
backside of the chip, shorting it and using the CPW signal line
as the MS feed line, see a 3D schematic view in Fig. 2.
The measured impedance of some manufactured 220 GHz
antenna break-out circuits was found in good agreement with
simulations.
Noise factor measurements (using the hot/cold load Yfactor method) on an MMIC with the novel slot-square feed
antenna, an LNA and a mixer positioned on top of a 12 mm
Si lens showed a total DSB NF < 10 dB at 220 GHz [1].
I. ABSTRACT
F
or several decades the interest for applications at
millimetre wave frequencies has been increasing. This
mainly concerns sensor- and communication systems for
military use and the corresponding for space based systems.
More recently emerging applications such as e.g. imaging
systems for detection of concealed weapons has suggested the
use of the atmospheric windows, e.g. at 220 GHz.
To reduce the cost, weight and volume of such systems,
MMIC based solution are preferred compared to discrete
solutions. In terms of price per unit and manufacturing
repeatability it would be highly advantageous to have a singlechip solution. By also integrating the antenna on-chip, only
low-frequency inputs and outputs would be required on the
MMIC. The loss associated with traditional bond wire
connections at 220 GHz would then be completely avoided.
In this work, which is done within the NanoComp project, a
Swedish/German Research Collaboration supported by FOI,
we aim to design a 220 GHz single-chip receiver MMIC with
integrated antenna. The MMIC will aside from the integrated
antenna (this paper) include a full mm-wave front-end, i.e. a
low noise amplifier (LNA), a mixer, and an X8 LO multiplier
chain [1]. All designs are manufactured in a 0.1 µm gate
length GaAs mHEMT MMIC process offered by the
Fraunhofer Institute for Applied Solid-State Physics (IAF) in
Germany.
The novel slot-square antenna will be used to feed a highresistivity Si substrate lens. The thickness of the thinned GaAs
substrate, 50 µm, roughly corresponds to 0.13λ. This rather
high thickness would be problematic for a planar antenna
because of the excitation of surface wave modes. However,
by using a substrate lens the excitation of surface waves is
prohibited and moreover, the field of view of the antenna
system is easily tailored to the application by choosing a
suitable diameter of the lens, see the conceptual drawing in
Fig. 1.
The 220 GHz slot-square feed antenna was designed with
requirements on the impedance bandwidth, a dual polarization
capability and compatibility with a microstrip (MS)
environment. The substrate lens under the MMIC focuses the
Fig. 1. Conceptual drawing of an MMIC
with feed antenna on top of a substrate
lens antenna.
Fig. 2. A 3D view of the novel
MS fed slot-square substrate lens
feed antenna.
REFERENCES
[1] S. E. Gunnarsson, N. Wadefalk, J. Svedin,
S. Cherednichenko, I. Angelov, H. Zirath, I. Kalfass, and
A. Leuther, “A 220 GHz Single-Chip Receiver MMIC
with Integrated Antenna”, submitted to IEEE Microwave
and Wireless Components Letter, Oct. 2007.
40
GHz Symposium 5-6 March 2008
Planar antennas for terahertz frequencies
Sergey Cherednichenko
Chalmers University of Technology, Dep.Microtechnology and Nanoscience, Fysikgrand
3, SE-41296, Gothenburg, Sweden
serguei@chalmers.se
During the last years, terahertz frequencies got much interest due to the progress
in the electronic and photonic devices for terahertz wave detection and generation. Many
applications require free space wave launching and re-collection. It brings up a question
of an effective antenna, suitable for integration with THz devices, and presenting required
properties, among of which high gain and polarization sensitivity are often named. A
high level of expertise has been reached in the field of radio astronomical receivers,
which can now be adopted for terrestrial applications.
Planar antennas are very efficient for integration not only with two terminal but,
as it was recently demonstrated with a subMM MMIC receiver, even with three terminal
devices. A narrowband response with a X-pol level below -20dB can be obtained with
either double-slot (see figure below) and double-dipole antennas. Clamped to a silicon
lens a high gain of about 20dB can be achieved. Due to a high symmetry of the beam
pattern this antennas can be efficiently (90%) coupled to a Gaussian telescope for a long
distance signal transmission. On contrary, logarithmic spiral (see the figure below) or
logo-periodic antennas have very broadband response, covering a few octaves.
We will present results of different antenna simulations using modern 3D
simulating software. Theoretical results will be compared with experimental once.
detector
elliptical silicon lens
1,2
0.6
1
1.6 1.8
Terahertz Spiral antenna
0.9
Normalised intensity
Signal (A.U.)
1
0,8
0,6
0,4
0,2
0
0
0,5
1
1,5
2
Frequency (THz)
2,5
3
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.3
0.8
1.3
1.8
2.3
2.8
3.3
Frequency (THz)
41
GHz Symposium 5-6 March 2008
Geosynchronous Earth Orbit Atmospheric Sounder
S. Andersson, J. Embretsén, A. Emrich, M. Ericson, M. Hjorth, J.Riesbeck, C. Tegnander
Omnisys Instruments AB, Contact: je@omnisys.se
ABSTRACT
Omnisys Instruments is currently developing a system which will demonstrate an innovative
instrument concept for future atmospheric microwave sounding. The instrument concept’s
primary objective is to achieve accurate observation at significant spatial resolution with
microwave and (sub) millimetre wave sensors from Geosynchronous Earth Orbit (GEO).
A 50 GHz receiver front-end
Atmospheric sounders measure the distribution of
radiation emitted by the atmosphere. From this
information vertical profiles of temperature and
humidity through the atmosphere may be
obtained. Current generation of sounders, which
are embarked on-board low Earth orbit (LEO)
satellites, provide meteorological data for weather
forecasting and global observations for climate
monitoring. Geostationary observations have the
key potential advantage for providing real-time
continuous coverage of a region, which is
essential for now casting.
Omnisys has led an ESA study to select candidate breakthrough concepts to meet the
requirements. In terms of performance, the specification is: 30 km resolution over the earth
disc (400x400 map), covering the frequency bands 53, 118, 166, 183, 340, 380 GHz, and an
update rate of 30 minutes. Due to the large aperture size required to achieve 30 km horizontal
resolution from GEO, its use has so far been restricted to the case of LEO satellites. This will
now be accomplished with a synthesised aperture using over 500 (sub) millimetre receivers
mounted on booms. The deployed booms stretch over an 8 meter diameter.
The overall objective of the GEO Atmospheric Sounder
activity is to develop and demonstrate enabling
instrument concepts to achieve accurate observation at
significant spatial resolution with microwave and (sub)
millimetre wave sensors from GEO. This will entail the
development and demonstration of the required
innovative technologies as part of a demonstrator.
The demonstrator will consist of 20 dually polarised
50 GHz receiver front-end units with multi-functional
single-chip receiver MMIC. These front-end units will
be mounted in a Y-shaped sparse array on a rotating,
900 mm in diameter, mechanical structure.
The instrument concept
demonstrator
The central processing core of the demonstrator performs IF conditioning, quantisation and
cross correlation of all combinations of the 40 (2x20) down converted signals. Adding to this
are the phased locked local oscillator with distribution, power conditioning and distribution
unit and electronics control unit.
42
GHz Symposium 5-6 March 2008
Back-end module demonstrator for radio-astronomy
applications
Juan Luis Cano, Beatriz Aja, Enrique Villa, Luisa de la Fuente, Eduardo Artal
(juanluis.cano@unican.es)
Departamento de Ingenieria de Comunicaciones. ETSIIT. Avenida de los Castros, s.n.
Universidad de Cantabria. Santander (Spain)
Summary
A broadband back-end module demonstrator for a Ka-band radiometer has been
designed and tested. It is a direct conversion receiver based on low noise amplification,
band pass filtering and Schottky diode detection. Simulations of the system are based on
individual subsystems measurements using a coplanar probe station and broadband
coplanar to microstrip transitions.
Introduction
Radio-astronomy radiometers for scientific applications, such as Cosmic Microwave
Background observations [1], are very sensitive and broadband receivers. Their frontend module is a very low noise amplifier operated at cryogenic temperature. A back-end
module at room temperature provides extra gain and direct detection. The branch
demonstrator is composed of two cascaded MMIC LNA, band pass filter and Schottky
diode detector. The receiver bandwidth is from 26 up to 36 GHz.
Results
The figure on the left shows a LNA assembled in a test fixture for its characterisation,
and its gain for different bias points. The plot on the right depicts the back-end output
response versus frequency for -53 dBm input power, which is the nominal power output
coming from the front-end.
30
|S21| (dB)
20
3.5 V ; 63 mA
10
3 V ; 62 mA
0
2.5 V ; 60 mA
-10
2 V ; 59 mA
-20
1.5 V ; 55 mA
-30
-40
15
20
25
30
35
40
45
50
freq, GHz
Conclusion
A broadband back-end module demonstrator has been assembled. Its behaviour has
been adequately predicted by a prior test of each individual subsystem.
References
[1] M. Bersanelli et al., ”Planck-LFI: instrument design and ground calibration strategy”, Proc.
of the European Microwave Association”, Vol. 1, Issue 3, September 2005, pp. 189-195.
This work has been supported by the Ministerio de Educación y Ciencia (Spain), grant references ESP2004-07067C03-02, AYA2007-68058-C03-03 and BES-2005-6730.
43
GHz Symposium 5-6 March 2008
ALMA Band 5 (163-211 GHz) Sideband Separating Mixer
1
B. Billade1, Igor Lapkin1, A. Pavolotsky1, R. Monje1, J. Kooi2, and V. Belitsky1.
Group of Advanced Receiver Development(GARD), Chalmers University of Technology,
Gothenburg, Sweden.
2
California Institute of Technology, Pasadena, USA
Email: bhushan.billade@chalmers.se
The ALMA radio telescope is a multi-antenna interferometric instrument under construction by
an international consortia consisting of European countries, USA, Canada, and Japan. ALMA
will use receivers in 10 different bands and this report presents work on ALMA Band 5, which
will be a duel polarization sideband separating heterodyne receiver for 163-211 GHz with 4 - 8
GHz IF. For each polarization, Band 5 receiver employs sideband rejection quadrature layout
(2SB) with SIS mixers. In such receiver, the RF signal appears with 90 degree phase difference
at the two mixers and LO appears in phase. In our design we use a 90-degree 3-dB waveguide
branch-line hybrid for RF and waveguide E-plane power divider for LO. In order to achieve
specified -15dB side band rejection the amplitude and phase imbalance of the quadrature
scheme should not be more than -2.5dB and 12 degrees respectively, which puts strict
constrain on the RF and IF hybrids. We have performed extensive simulations of all
components of the 2SB SIS mixer, including 3D EM modeling using Agilent EMDS and ADS
circuit simulator. Our simulation shows that the RF hybrid can be designed with 0.8 dB
amplitude and 3 degree phase balance at the best case, while the IF hybrid employing Lange
coupler attains 0.7 dB amplitude and 5 degree phase balance. The SIS mixer employs MMIClike approach with all mixer components integrated on the same crystal quartz substrate. The
waveguide-to-microstrip transition is done using an E-probe extending into the waveguide; the
RF choke at the end of the probe provides a virtual ground for the RF signal. The LO injection
is done using a microstrip line directional coupler with slots in the ground plane. The mixer
itself consists of two SIS junctions with area 3 square microns each, in twin junction
configuration, followed by a quarter wave transformer to couple it to the probe. This circuitry
provides optimum matching of the SIS junctions at RF frequencies though the problem with
such architecture is that there is no natural cold point to extract the IF. A high inductive line is
therefore used using additional layer of SiO2. A bias-T is integrated with the IF hybrid.
At the time of the conference we plan to present details of the mixer design and results of the
first experimental verification of the DSB mixer performance.
44
GHz Symposium 5-6 March 2008
High Power Photonic MW/THz Generation Using UTC-PD
Biddut Banik, Josip Vukusic, Hans Hjelmgren, Henrik Sunnerud, Andreas Wiberg and Jan Stake
Department of Microtechnology and Nanoscience, Chalmers University of Technology; SE41296 Göteborg, Sweden (e-mail: biddut.banik@chalmers.se).
Summary: The ongoing research work concentrates on extending the previously accomplished
UTC-PD fabrication and modelling techniques to 340 GHz and above. We have fabricated and
characterized UTC-PDs intended for high power MW/THz generation. Several integrated
antenna-detector circuits have been designed and characterised.
Introduction: Because of the inherent difficulty to generate power in the frequency range 0.l-10
THz, the term ‘THz-gap’ has been coined [1]. Among a number of MW/THz generation
techniques, the photomixer based sources hold high potential offering wide tunability and decent
amounts of output power. The photomixing technique relies on the nonlinear mixing of two
closely spaced laser wavelengths generating a beat oscillation at the difference frequency. In
recent years, there has been an increasing interest in the UTC-PD [2] for photomixing, photo
receivers, mm- and sub-mm-wave generation, fibre-optic communication systems, and wireless
communications. UTC-PDs have become very promising by demonstrating output powers of 20
mW at 100 GHz [2] and 25 µW at 0.9 THz [3].
Fig. 1. (a) The principle of photomixing (b) SEM of a fabricated UTC-PD (c) Integrated UTC-PD-antenna.
Results: In order to understand the device behaviour and its dependence on various factors, we
have developed an accurate device model [4] implementing hydrodynamic transport model. The
model has also enabled us to design and optimise the device for any specific application and
target frequency. With the aim to realise and evaluate the performance of the UTC-PD in terms of
high power MW/THz generation, several antenna and bias circuit integrated UTC-PDs have been
designed and fabricated. The characterisation and measurement results of those devices will be
presented.
Conclusion: We have designed, fabricated and characterised UTC-PDs and antenna integrated
UTC-PDs. Several other design methodologies, e.g. optimised optical coupling, thermal
management, waveguide integration etc. will be investigated in future.
References:
[1] P. H. Siegel, "Terahertz technology," IEEE Transactions on Microwave Theory and Techniques, vol. 50, pp.
910-28, 2002.
[2] H. Ito, T. Nagatsuma, A. Hirata, T. Minotani, A. Sasaki, Y. Hirota, and T. Ishibashi, "High-power photonic
millimetre wave generation at 100 GHz using matching-circuit-integrated uni-travelling-carrier photodiodes,"
Optoelectronics, IEE Proceedings, vol. 150, pp. 138-142, 2003.
[3] C. C. Renaud, M. Robertson, D. Rogers, R. Firth, P. J. Cannard, R. Moore, and A. J. Seeds, "A high responsivity,
broadband waveguide uni-travelling carrier photodiode," Proceedings of the SPIE, vol. 6194, pp. 61940C, 2006.
[4] S. M. M. Rahman, H. Hjelmgren, J. Vukusic, J. Stake, P. Andrekson, and H. Zirath, "Hydrodynamic simulations
of uni-traveling-carrier photodiodes," IEEE J. Quantum Electron., vol. 43, pp. 1088-1094, 2007.
45
GHz Symposium 5-6 March 2008
Towards a THz Sideband Separating Subharmonic Schottky Mixer
Peter Sobis (1,2), Jan Stake(1) and Anders Emrich(2)
(1)
Chalmers University of Technology
Department of Microtechnology and Nanoscience
SE-412 96, Göteborg, Sweden
Email: peter.sobis@chalmers.se
Email: jan.stake@chalmers.se
(2)
Omnisys Instruments AB
Gruvgatan 8, SE-421 30, Västra Frölunda, Sweden
Email: ps@omnisys.se
Email: ae@omnisys.se
GaAs Schottky mixers with state of the art planar submicron diodes are used for THzdetection up to 3 THz today [1]. GaAs Schottky diodes can operate in room
temperature which makes them good candidates for space applications and an
interesting low cost alternative to low noise cryogenic SIS and HEB technologies.
The main advantage of a sideband separation scheme besides that the lower and upper
sidebands are indeed separated, is that the IF bandwidth is increased by a factor of
two. Moreover, there is no need for image rejection filters on the RF input, which can
be bulky and increase the weight and cost of the overall receiver system. Sideband
separation mixers have been implemented at THz frequencies before [2], however up
to this point they have never been tried with Schottky diodes in a subharmonic mixer
configuration. We will present the current status of the development of a novel
sideband separating subharmonic reciever topology operating at 340 GHz, see Fig1.
The design of a subharmonic mixer and the LO and RF waveguide hybrids will be
presented followed by an account of measured results of the individual components.
Fig1. Schematic of the sideband separation mixer (left) and modular assembly (right).
[1] J.L. Hesler, T.W. Crowe, W.L. Bishop, R.M . Weikle, R.F. Bradley and Pan
Shing-Kuo, “The development of planar Schottky diode waveguide mixers at
submillimeter wavelengths”, IEEE MTT-S International Microwave Symposium
Digest, vol 2, pp. 953-6, 1997.
[2] C. Risacher, V. Vassilev, V. Belitsky, A. Pavolotsky, “Design of a 345 GHz
Sideband Separation SIS Mixer”, Proceedings of 3rd ESA Workshop on Millimeter
Wave Technology and Applications: Circuits, Systems and Measurement Techniques,
21-23 May 2003, Espoo, Finland
46
GHz Symposium 5-6 March 2008
HIFAS: High-performance Full-custom Autocorrelation Spectrometer ASIC
A. Emrich, S. Andersson, J. Dahlberg, L. Landén, M. Hjorth, Omnisys Instruments;
T. Kjellberg, Chalmers/MC2; N. Andersson, Acreo
Contact email: mh@omnisys.se
ABSTRACT
Autocorrelation spectrometers are often employed for radio astronomy and atmospheric
research applications, due to their high bandwidth and stability. The major advantage of the
autocorrelator over other types of spectrometers is that the real-time signal processing can be
implemented using relatively simple digital logic, which allows for compact implementations
with very high clock speeds and corresponding bandwidths (or low bandwidth, low power
implementations where that is preferred).
For the Odin research satellite, launched in 2001 and still in operation, Omnisys designed an
autocorrelation spectrometer. The core of the spectrometer consists of two ASICs, a sampler
and an autocorrelator. Each pair of such chips has 100 MHz of bandwidth and consumes 0.4
W of power. This performance-power ratio was outstanding at the time and is still impressive.
HIFAS, Omnisys fifth generation autocorrelation
spectrometer ASIC is being developed. It is a full-custom
design with over two million transistors, designed for
IBM’s 180 nm SiGe Bi-CMOS process. Unlike earlier
generations, it contains both the bipolar 3-level (“1.5-bit”)
A/D converter and the CMOS correlator on the same chip.
Thereby, the sensitive high-speed digital interface between
the two parts gets integrated on the chip.
The chip supports as input either a complex I/Q input signal
pair, measuring its spectrum from –fclk/2 to +fclk/2 or a
single baseband signal sampled on both clock edges,
measuring from 0 to fclk. This choice gives flexibility for
the system level design.
Fig. 1: Photo of the HIFAS
The first batch of the chip was produced in 2007. ASIC. ADC is in bottom right
Unfortunately it turned out to have a logic bug that makes it corner.
necessary to do a re-run. A second revision of the chip is
being designed at the time of this writing, and tape-out is
planned for early 2008.
Despite these initial problems with the chip, most of the chip’s
functions have been tested and shown to work. The analog parts
work in both of the two input modes with up to 8 GHz sample
clock, and most of the digital features are also working correctly.
Fig. 2: ASIC wire-bonded
on test board.
The goal is to reach a bandwidth of 8 GHz, a resolution of 1024
channels, and a power consumption of 3-5 W. When finished,
this chip will set a new world record in autocorrelator
performance, and open for new possibilities in radiometry on
both space and ground.
47
GHz Symposium 5-6 March 2008
Session II
1530-1730 Wednesday 5 March 2008
48
GHz Symposium 5-6 March 2008
Extremely Low-Noise Amplification with Cryogenic FET’s and
HFET’s: 1970-2006
(Where do we go from here?)
Marian W. Pospieszalski
National Radio Astronomy Observatory
2551 Ivy Road, Ste. 219
Charlottesville, VA 22903
SUMMARY
Improvements in the noise temperature of field-effect transistors (FET’s) and, later,
heterostructure field-effect transistors (HFET’s) over the last several decades have been
quite dramatic. In 1970, a noise temperature of 120 K was reported at 1 GHz and
physical temperature of 77 K; in 2003, noise temperatures of 2, 8 and at 35 K were
reported at 4, 30 and 100 GHz, respectively, for physical temperatures of 14 to 20 K.
In the first part of the talk the developments in this field are briefly traced and an
attempt is made to identify important milestones. Examples of experimental results
obtained with different generations of FET’s (HFET’s) are compared with the model
predications. The current state of the art in cryogenic low noise InP HFET amplifiers is
presented and some gaps in our understanding of experimental results are emphasized.
Random gain fluctuations of these amplifiers important for applications in broadband
continuum radiometers for radio astronomy are also shortly discussed.
In the second part the question whether rapidly advancing technologies of microwave
heterostucture bipolar transistors (HBT’s) and CMOS can in the future offer alternatives
to the extremely low noise performance of InP HFET’s is addressed. For that purpose
noise models of unipolar and bipolar transistors are reviewed with emphasis on their
common noise properties. Simple close-form approximate expressions for the noise
parameters are given and comparison is made of calculated results with the available
experimental data.
49
GHz Symposium 5-6 March 2008
560 GHz ft, fmax operation of a refractory emitter metal InP DHBT
Adam M. Crook, Zach Griffith, Mark J. Rodwell
ECE Department, University of California
Santa Barbara, CA, USA
Erik Lind*
Solid State Physics, Lund University
Lund University, Sweden
Erik.Lind@ftf.lth.se
* work performed in part while at UCSB
Abstract—We present results of a hybrid
dry/wet-etched type I InGaA/InP DHBT using a
refractory emitter metal. Simultaneously high ft
and fmax of 560 GHz is obtained, with a
breakdown voltage BVceo of 3.4V.
INTRODUCTION
Scaling theory [1] of HBTs indicate that a 2:1
increase in bandwidth requires a 4:1 reduction in
emitter and collector widths – for THz operation this
requires emitter widths below 125nm. Traditional
lift-off techniques and wet etching techniques used
for triple-mesa HBTs are difficult to reliable scale
below 300 nm emitter widths. We have developed a
hybrid dry/wet etch technique that reliable scales to
emitter widths below 250nm. First results on a
22nm base thickness, 70 nm collector thickness with
~200 nm emitter width produced record
simultaneous ft and fmax of 560 GHz [2].
I.
Figure 1. Cross section view (52° tilt) of a DHBT
MEASURMENTS & CONCLUSIONS
The transistors were characterized from DC-67
GHz. The DC current gain was ~ 25. The
Breakdown voltages were BVceo ~ 3.4V, and BVcbo
~ 3.6 V, limited by band-to-band tunneling. For
devices with emitter widths of 200 nm, a
simultaneous extrapolated ft and fmax of 560 GHz
was obtained, which is the first report of a device
with both ft and fmax above 500 GHz. Peak ft was
600 GHz for a device with lower (430 GHz) fmax.
III.
FABRICATION
The epitaxial material was grown on 4” S.I. InP
wafers at commercial vendor IQE. The fabrication
starts with a blanket sputtered deposited Ti0.1W0.9
film, which is subsequently patterned using a SF6/Ar
dry etching. Using the emitter metal as mask, the
emitter is dry etched in a Cl2/N2 plasma, stopping
just short of the base. A InP wet etch is then used to
clear the In0.53Ga0.47As base. The transistors are
finished using self aligned base ohmics, forming a
triple-mesa transistor. A cross-section SEM image is
shown in Fig. 1. Emitter junctions with widths down
to 200 nm could controllable be fabricated, showing
a substantial improvement over fully wet etched
processes.
II.
ACKNOWLEDGMENT
This work was supported by the DARPA SWIFT program
and a grant from the Swedish Research Council.
REFERENCES
[1]
[2]
50
M. Rodwell et.al., IPRM 2007, pp. 9-13
E. Lind et.al., DRC 2007, Late News
GHz Symposium 5-6 March 2008
Low phase-noise balanced Colpitt InGaP-GaAs HBT VCOs with wide frequency
tuning range and small VCO-gain variation
Herbert Zirath
Ericsson AB, Microwave and High SpeedElectronics Research Centre, Mölndal, Sweden.
Microwave Electronics Laboratory, Göteborg, Sweden Chalmers University of Technology, Department of Microtechnology and
Nanoscience, GHzCentre Email: herbert.zirath@chalmers.se
Abstract—Low phase-noise InGaP-GaAs HBT VCOs, utilizing an
on-chip ‘wide tuning range varactor’, have been designed,
fabricated, and characterized. The primary design goals were low
phase noise, wide continuous frequency tuning and small VCOgain variation. Two types of varactor were compared, a square
varactor and a finger varactor. The square varactor achieves a
frequency tuning of 27%, and the finger varactor 21%. The
phase noise is typically -100dBc/Hz at 100kHz offset frequency. A
VCO gain (sensitivity) of 160 MHz/V is obtained with a variation
of less than 10%. The power consumption is controlled by the
core dc-current and is of the order 50-100mW.
RCC
VCC
CCC1
VVAR
VVAR
LTANK
LTANK
CVC
RB2
Finger varactor
CVC
CB
C1
C1
RB
RB
C2
RE
Cout
RE
Cout
RB1
I.
THE MMIC TECHNOLOGY
The MMIC-technology used in this study is a modified
commercially available InGaP-GaAs HBT MMIC-technology
with an emitter width of 3 μm. The base-collector doping
profile is specially designed and optimized for high
Cmax/Cmin ratio and a linear oscillation frequency versus
varactor voltage characteristic, i e a constant VCO gain.
II.
VOUT
Fig 1 Schematic of the VCO
DESIGN OF THE VCO
The schematic of the VCO is shown in Fig 1. The VCO
consist of two Colpitt oscillators which are coupled together
with the capacitor C2. The complimentary outputs are taken
from the emitters in order to minimize the loading of the VCO.
By controlling the dc-current of the oscillator core, the
amplitude in the tank can be controlled as well as the output
power. The current in the oscillator is controlled by the voltage
applied at RB2. This control voltage can be used for amplitude
level control of the VCO. In order to improve the Q-value of
the tank, the varactor geometry was optimized by using a
finger structure. A VCO with a square varactor was also
designed for comparison. The varactor is realized by using the
base-collector junction of the HBT. In this work, a specially
tailored doping profile in the base-collector junction is utilized.
The goal is to achieve a linear frequency-varactor voltage
characteristic as well as a large capacitance variation.
III.
VOUT
Fig 2 VCO with square varactor. 26% frequency tuning
A minimum phase noise of -99.5 and -101.5 dBc/Hz is
obtained with approximately 4 dB variation over the
frequency band for a fixed bias. The simulated minimum
phase noise is -102 and -104.5 dBc/Hz @100kHz offset
respectively i e 2.5/3 dB lower than the measurements. The
use of finger varactor resulted in a measured improvement of
phase noise with 2 dB, close to the simulated value of 2.5 dB.
The phase noise as a function of offset frequency was
measured utilizing an Agilent E5052A Signal Source
Analyzer.
This work shows that encouraging results on wideband
tuned Colpitt VCOs can be obtained if a special ‘tailored’
base-collector doping profile is utilized. The measured phase
noise is below most reported VCOs in the literature when the
oscillation frequency is normalized, and the variation in KVCO
is to the knowledge of the author the smallest reported in the
literature
MEASUREMENT RESULTS AND DISCUSSION
The output power and oscillator frequency versus the
varactor voltage at a collector supply of 8V and a core current
of 20 mA is plotted in Fig. 2. The oscillation frequency versus
varactor voltage is linear throughout the varactor control range
which is highly required in PLL-applications. The ‘square
varactor VCO’ can be tuned continuously from 6.8 to 8.9GHz.
51
GHz Symposium 5-6 March 2008
Feasibility of Filter-less RF Receiver Front-end
S.Ahmad, N.Ahsan, A.Blad, R.Ramzan, T.Sundström, H.Johansson, J.Dąbrowski, C.Svensson
Department of Electrical Engineering, Linköping University
This work is a feasibility study on a radio receiver design where RF filters are avoided or
considerably simplified. The problem originates from the idea of a software defined radio
(SDR), capable of receiving any radio standard in any frequency band. Following the
traditional radio paradigm towards multi-band SDR, we would end-up in a bunch of
highly selective band-preselect filters placed between antenna and the RF front-end. This
is an expensive, bulky and lossy solution. Similarly if we try to develop tunable passive
filters.
In a receiver with no band-preselect filter the wanted signal would be accompanied by
strong interferers (blockers). The idea is to receive the entire signal and convert it to the
digital domain where the interferers can be effectively removed by digital filtering. As
both signal and blocker are simultaneously present at the receiver input, automatic gain
control does not help. Instead, dynamic range of the front-end and the ADC, and their
linearity and noise floor must meet the full dynamic requirement.
For the receiver noise floor both the thermal and ADC quantization noise are vital. We
find the receiver dynamic range DR = Pbl /( FRx N ref ) = ( Pbl × SNRout ) /( S × γ ) where Pbl is the
blocker power maintained by the receiver, Nref is the reference thermal noise, SNRout is the
required SNR at the demodulator input, S is the receiver sensitivity and γ stands for the
despreading factor (if any). FRx reflects the contribution both by thermal and quantization
noise. The solution we suggest is a combination of a highly linear LNA/mixer (for large
blockers there may be very little room for LNA gain), followed by a low-pass ΣΔ-ADC
with very high sampling rate (of the order of the carrier frequency). This architecture
could be used as a homodyne or as a low IF superheterodyne with digital IF. By using a
very high sampling rate, we mitigate the need for antialias filters and we gain dynamic
range through the high oversampling ratio. We will give results on estimated
requirements on the components and parameters of such a solution.
To demonstrate the concept we have designed a zero-IF/low-IF RF front-end composed
of a highly linear wideband LNA and a quadrature mixer with a simple antialiasing
lowpass filter at its output, followed by a 1st order 4-bit ΔΣ ADC, which operates in
current mode. The front-end is intended for frequencies 1-6 GHz while the largest signal
bandwidth is assumed to be 20 MHz. In this way we aim to combine GSM900/1900,
UMTS FDD, DECT, and WLAN 802.11a,b standards in one hardware. The ADC
sampling frequency is 2.4GHz so the images at k×2.4GHz can be easily suppressed with
the 2nd order low pass filter. The circuit has been designed for a 90 nm CMOS process
and sent for fabrication at CMP.
Simulations indicate the following results for this demonstrator at 2.4/5.2 GHz:
Gain = 6dB/5dB, NF=6dB/7.5dB, S11= -15dB, and IIP3= -4.5dBm, Pbl ≤ -10dBm. The
NF including the ADC is by about 3dB larger. These data are not sufficient to meet the
required specifications, but show the feasibility of the concept.
52
GHz Symposium 5-6 March 2008
Small Size 2-10GHz Radar Receiver Si-RFIC
Håkan Berg€, Heiko Thiesies€, Marie Hertz€, Fredrik Norling€
€Microwave Technology, Saab Microwave Systems, Saab AB
SE-412 89 Göteborg, Sweden
Introduction
Using traditional RFIC design architecture a radar receiver IC is designed. Traditional
low frequency architectures allow for a dramatic size reduction compared to microwave design [1]. Furthermore, since there are no matching networks between the circuits on-chip the bandwidth is limited by transistor performance only. This gives a
larger bandwidth and the opportunity to cover the radar bands at S-, C-, and X-band
using the same receiver. The IF bandwidth is 200-2500 MHz. Due to the large bandwidth at both in- and output the IC is well suited for superheterodyne radar receivers
where it can be used for both frequency conversions.
The IC is manufactured by austriamicrosystems in their 0.35µm SiGe-BiCMOS process with an fT of 60 GHz and is packaged in a 4x4 mm QFN plastic package.
Results
Simulations have been performed at five different input frequency bands as shown below. The gain and IP3 of these simulations are presented with different gain settings.
The noise figure is simulated to be less than 13 dB at all frequencies at maximum
gain.
IF [MHz]
2000 ±250
2000 ±250
2000 ±250
2000 ±250
250 ±50
LO [GHz]
5.05
7.55
10.50
12.50
2.25
20
15
10
5
0
-5
-10
-15
-20
Application
S-band to IF
C-band to IF
X-band (low end) to IF
X-band (high end) to IF
IF to Baseband
Input IP3 [dBm]
Gain [dB]
RF [GHz]
3.05 ±0.25
5.55 ±0.25
8.50 ±0.25
10.5 ±0.25
2 ±0.05
1
2
3
4
5
6
7
8
9
10 11
26
24
22
20
18
16
14
12
10
1
Input frequency [GHz]
2
3
4
5
6
7
8
9
10 11
Input frequency [GHz]
Simulated gain and input referred IP3 for the maximum gain (triangles) and at 20 dB
attenuation (diamonds).
Conclusions
A broadband receiver IC has been developed. It is shown to be a potential replacement component for several single-band components and help building generic receivers. It has utilized RFIC design techniques rather than microwave ones; this
makes it possible to combine the small-size with broadband performance. The simulated performance is well compatible to the requirements of future radar systems.
[1] H. Berg, H. Thiesies, M. Hertz.and F. Norling, “A High Linearity Mixed Signal
Down Converter IC for C-band Radar Receivers” 2nd European Microwave Integrated Circuits Conference, pp 255-258, Okt. 2007.
53
GHz Symposium 5-6 March 2008
High frequency, current tunable spin torque oscillators: experimental characterization
S. Bonetti*, J. Garcia, J. Persson, and Johan Åkerman
Materials Physics, Department of Microelectronics and Applied Physics,
Royal Institute of Technology, Electrum 229, 16440 Stockholm-Kista, Sweden
*e-mail: bonetti@kth.se
We present characteristics of high frequency and current tunable spin torque oscillators (STOs),
and give a description of the experimental setup we implemented for their characterization. Our
setup allows an automated investigation of magneto-resistance and frequency response as a
function of varying applied magnetic field (both in magnitude and in-plane/out-of-plane angle) and
driving current. We could measure oscillators with Q-values (resonance frequency / bandwidth) up
to 2700 and high tunability range (10-26 GHz). Since we reached the upper frequency limit of the
spectrum analyzer (26.5 GHz) at far from critical values of current and field, we expect to observe
even higher tunability ranges (up to an estimated 65 GHz) with improvement of the setup.
State-of-the-art oscillators in wireless devices, such as bluetooth chips, use inductive coils in
order to generate microwave power. These coils consume large chip area and it has been estimated
that more than 75% of the silicon could be saved using a Spin Torque Oscillator (STO). The
working principle of this device lies in the interaction between a spin polarized dc current and the
magnetization of the various magnetic layers in the STO, resulting in an angular precession of the
magnetization itself. As the resistance of the device is proportional to the magnetization angle, the
device is an ac resistor, and the dc current causes an ac voltage output. Predictions suggest
oscillations ranging from 1 GHz up to 60 GHz with high Q-values (up to 18,000) [1].
In order to measure the signal generated by these devices, one must be able to vary the
applied magnetic field and current in an accurate way. Our setup can generate magnetic fields up to
1.7 T and since samples are mounted on a holder fixed on a turntable, the direction of the magnetic
field can be varied continuously in the direction in-/out-of- plane of the sample. A pulsed and
stepped current source and nanovoltmeter solution replaces the need of a more expensive lock-in
amplifier, still allowing a very accurate reading of the magneto-resistance. In order to connect to the
sample, non-magnetic ground/signal/ground high frequency (up to 40 GHz) probes are used. The
signal from the probe is amplified by a low noise, broadband (1-26 GHz), +45 dB amplifier before
connecting to a signal analyzer (20 Hz – 26.5 GHz frequency range). The dc current is applied to
the STO by means of a bias-T, connected in the vicinity of the amplifier in the transmission line. All
instruments are GPIB controlled and the measurements are fully automated.
A highly advantageous feature of these devices is that the oscillation frequency can be tuned
in a linear way with the applied dc current and field (typically 100 Mhz/mA and 20 GHz/T). We
observe oscillations up to 26.5 GHz only limited by the microwave amplifier and the signal
analyzer. In the literature, the highest oscillation frequency observed to date is 35 GHz [1], but
higher values are not theoretically precluded. From extrapolations of our results we find a
maximum operation frequency as high as 65 GHz. At present, the highest Q-value observed in our
STOs is 2700 at 12 GHz, but Q values above 18,000 [1] at about 34 GHz have been observed in
similar devices. Thanks to all these characteristics (linear current tunability, high Q, broadband
frequency range), STOs promise to be attractive devices for applications where the operational
frequency of the oscillator is required to vary on a wide range, on the order of tens of GHz.
We gratefully acknowledge financial support from The Swedish Foundation for Strategic
Research (SSF), The Swedish Research Council (VR), The Göran Gustafsson Foundation and The
Knut and Alice Wallenberg Foundation.
References: [1] W. H. Rippard et al, Phys. Rev. B 70, 100406(R) (2004).
54
GHz Symposium 5-6 March 2008
N-tupling the capacity of wireless
communications using electromagnetic
angular momentum
Bo Thidé∗
Swedish Institute of Space Physics
P. O. Box 537, SE-751 21 Uppsala, Sweden
E-mail: bt@irfu.se
Physically, electromagnetic radiation consists of two quantities: the well-known
electromagnetic linear momentum (or power/Poynting) flux and the less wellknown electromagnetic angular momentum flux, both of which are carried by
radio beams all the way out to the far zone and are detectable there.
Conventional radio communications has so far relied primarily on the transfer
of linear momentum/power/Poynting flux from a transmitting antenna to a receiving antenna. However, one often makes use of the spin part of the angular momentum (SAM), better known as wave polarisation. SAM spans two orthogonal
rotational states, corresponding to the (one and the smae) left-hand and right-hand
rotation of the electric field of all of the waves (photons) in the radio beam. Using
wave polarisation, the communication capacity in a given (oscillatory) frequency
bandwidth can therefore be effectively doubled.
However, as demonstrated in numerous experiments carried out during the
past 15 years it is also possible to make use of the orbital part of the electromagnetic angular momentum (OAM), a much subtler, differential polarisation quantity, related to but distinctly different from SAM. Since OAM spans l + 1 orthogonal rotational states, where l can be a large number, this effectively allows an
N-tupling of the communication capacity in a given frequency bandwidth, where
N = 2l + 1. A capacity increase of N > 100 has been experimentally demonstrated.
We have been able to show that it is possible to use essentially conventional
radio techniques to endow radio beams with OAM, thus allowing for a dramatic
increase in the capacity of wireless point-to-point communications. We report the
latest results in this area.
∗ Also
at LOIS Space Centre, MSI, Växjö University, SE-351 95 Växjö, Sweden
1
55
GHz Symposium 5-6 March 2008
Session III
0830-1000 Thursday 6 March 2008
56
GHz Symposium 5-6 March 2008
The Next Wireless Wave is a millimeter Wave
Joy Laskar PhD
Schlumberger Chair in Microelectronics at Georgia Tech
Director Georgia Electronic Design Center
joy.laskar@ece.gatech.edu
The past few years has witnessed the emergence of CMOS based circuits
operating at millimeter wave frequencies. Integrated on a low cost organic
packaging, this technology has the promise for high volume fabrication, lowering
the cost and opening substantial markets for communications, sensing and
radar. As standardization efforts catalyze the interest and investment of industry
and agencies, one can be assured of ubiquitous millimeter-wave technology in
the consumer electronic market place in the fairly near future. In this talk we will
review our developments of radio technology with the highest speed (15Gbs),
robustness (better than 20dB link margin with bit-error rates better than 10^-9),
agility (beam steering and shadowing resistant) and lowest cost per bit. These
links are enabled by digital mmw radio technology which leverage new layer
fusion radio architectures.
57
GHz Symposium 5-6 March 2008
HF and Mixed Signal Challenges/Design for
Communication and Remote Sensing
(Invited talk)
Mehran Mokhtari
GigaHertz Symposium, 5-6 March 2008, Göteborg, Sweden
Abstract
Performance improvements in Communication- and remote sensing- systems and related
applications are strongly related to front-end capabilities in terms of Linearity, spectral
purity and noise, power consumption, size and weight, etc.
While, in most cases, the bulk of the complex signal processing is performed in the
digital domain, the significance of the analog front-ends, as well as mixed signal blocks
become apparent, in the overall performance and characteristics of the systems.
Digital signal processing utilizes the immense capabilities in constantly improving
CMOS technologies. Power consumption, device speed, and interconnect technology, as
well as the cost, in high volume production, in CMOS processes have been, amazingly
well, following the road maps set for DSP requirements, and appear to have more to offer
in the coming years. Mixed signal circuits, in the broader sense, in CMOS/BiCMOS
technologies, and circuit techniques, have satisfied the bulk of the immediate needs in
most high volume applications, and appear to have room for considerable improvements,
in the years to come.
However, some applications (mostly high-end and relatively low-volume), require
linearity and spectral purity beyond the capabilities of main stream circuit techniques and
technologies, especially in realizing analog and mixed signal functions that commonly set
the ultimate performance and capabilities of the systems.
The advancement of these systems, require technologies, circuit techniques, and overall
signal processing that rely on careful study of fundamental physical limitations set by
Solid-State Physics and Electromagnetics, for the boundary conditions of the desired
application, primarily in the micro-/millimeter- wave domain.
The presentation will recapture some of these fundamental limitations in the image of
application requirements. Some device technologies (Compound Semiconductor, Si), and
circuit approaches, will be covered and discussed.
58
GHz Symposium 5-6 March 2008
MMIC design at G-band (140-220 GHz) including a 220 GHz
Single-Chip Receiver MMIC with Integrated Antenna
Sten E. Gunnarsson1, Niklas Wadefalk1, Morteza Abbasi1, Camilla Kärnfelt1, Rumen Kozhuharov1,
Jan Svedin3, Bahar M. Motlagh1, Bertil Hansson1, Sergey Cherednichenko2, Iltcho Angelov1,
Dan Kuylenstierna1, Herbert Zirath1,4, Staffan Rudner3, Ingmar Kallfass5, and Arnulf Leuther5
1
Chalmers University of Technology, Microwave Electronics Laboratory, MC2, Göteborg, Sweden.
Chalmers University of Technology, Physical Electronics Laboratory, MC2, Göteborg, Sweden. 3Swedish Defense
Research Agency (FOI), Linköping, Sweden. 4Ericsson AB, MHSERC, Mölndal, Sweden,
5
Fraunhofer Institute for Applied Solid State Physics (IAF), Freiburg, Germany
Email: sten.gunnarsson@chalmers.se
2
So far, different antennas, LNAs, mixers of different
types, frequency multipliers, and power amplifiers have
been designed and characterized in the 200 to 220 GHz
band. Furthermore, different combinations of the former
have also been tested such as the 220 GHz single-chip
receiver MMIC with integrated antenna found in Fig 1, [1]
and a fundamentally pumped resistive mixer with
integrated slot-antenna. The receiver MMIC consists of a
novel slot-square substrate lens feed antenna, a three-stage
LNA and a sub-harmonically pumped resistive mixer. The
receiver MMIC is mounted on a 12 mm silicon substrate
lens which focuses the radiation from the calibration loads
to the on-chip antenna through an opening in the backside
metallization of the MMIC. The double sideband noise
figure of this quasioptical receiver is as low as 9.7 dB
(2470 K) at 220 GHz including the losses in the antenna
and in the lens. Other measured results include a
fundamentally pumped resistive mixer with less than 10
dB of loss between 200 and 220 GHz, [2].
I. ABSTRACT
Systems operating beyond 100 GHz have gained
increased academic and commercial interest over the
recent years. Traditional applications include radiometers
measuring the absorption lines of e.g. oxygen (118 GHz)
and water (183 GHz) for environmental studies. Emerging
applications has been proposed using the atmospheric
windows at 94, 140, and 220 GHz. These applications
include e.g. very high-speed (> 10 Gbps) wireless
communication, high-resolution passive and active
millimeter wave (mm-wave) imaging applications, and
systems for detection of concealed weapons, plastic
explosives, and biological weapons. In order to enhance
the usefulness of such systems, mobility is a key factor
and the systems should ideally be small and lightweight.
Thus, an MMIC based solution would be favourable over
a traditional discrete design due to its smaller size and
weight. Furthermore, an MMIC solution with its lower
price per unit and high manufacturing repeatability opens
up the possibility of multi-pixel array systems for
enhanced system performance.
G-band (140-220 GHz) MMICs have been designed and
measured in the MMIC group at Microwave Electronics
Laboratory (MEL), MC2, Chalmers University of
Technology. This work has been done within the
NanoComp project, a Swedish/German Research
Collaboration supported by the Swedish defense agency,
FOI, together with the Swedish Foundation for Strategic
Research (SSF) through the High Speed Electronics and
Photonics (HSEP) program. The goal is to design singlechip receiver MMICs operating at 220 GHz which
includes the full mm-wave front-end, i.e. on-chip antenna,
low noise amplifier (LNA), mixer, and an X8 LO
multiplier chain. All designs are manufactured in a 0.1 μm
gate length GaAs mHEMT MMIC process offered by the
Fraunhofer Institute for Applied Solid-State Physics (IAF)
in Germany, (www.iaf.fraunhofer.de).
Fig. 1. Chip photo of the 220 GHz single-chip receiver MMIC. The chip
measures 3.0×1.0 mm2.
REFERENCES
[1] S. E. Gunnarsson, N. Wadefalk, J. Svedin,
S. Cherednichenko, I. Angelov, H. Zirath, I. Kallfass, and
A. Leuther, ”A 220 GHz Single-Chip Receiver MMIC with
Integrated Antenna”, Submitted to IEEE Microwave and
Wireless Components Letter, October 2007.
[2] S. E. Gunnarsson, N. Wadefalk, I. Angelov, and H. Zirath,
”A 220 GHz (G-band) Microstrip MMIC Single-Ended
Resistive Mixer”, Accepted for publication in IEEE
Microwave and Wireless Components Letter.
59
GHz Symposium 5-6 March 2008
A Quad-Core 130-nm CMOS 57-64 GHz VCO
Vara Prasad Goluguri∗ , Johan Wernehag† , Henrik Sjöland† and Niklas Troedsson§
∗ Cambridge
Silicon Radio Sweden AB, Emdalav. 16, 223 69 Lund
of Electrical and Information Technology
Lund University, Box 118, 221 00 Lund Sweden
Email: Johan.Wernehag@eit.lth.se
§ UMC, 488 DeGuigne Drive, Sunnyvale, CA 94085, USA
† Department
current consumption is 3.1 mA from a 1.2V supply.
Abstract— In this paper a quad-core 130-nm CMOS oscillator is presented covering the license free 57-64 GHz
frequency band. Each core consists of an LC-oscillator,
the supply for the separate cores is switched on and off
by a combinatorial network using digital gates. The oscillator core consumes 3.1 mA from a 1.2 V supply and outputs an average power of -5.3 dBm. The phase noise is simulated to -110 dBc/Hz at 3 MHz offset.
−4
Output Power (dBm)
−4.5
I. I NTRODUCTION
A 7 GHz wide unlicensed band around 60 GHz is
available [1]. This spectrum allocation will permit
communication at several gigabits per second. Also
in Japan and Europe [2] frequency bands at 60 GHz
are opened for unlicensed WLAN communications
with a 5 GHz world wide overlap.
−5
11
10
01
00
−5.5
−6
−6.5
−7
−7.5
54 55 56 57 58 59 60 61 62 63 64 65 66 67
Frequency (GHz)
Fig. 1. Output power vs. frequency of the four oscillator cores. The
legend shows the two bit control word
II. S IMULATION R ESULTS
The circuit topology is the well known LCIII. C ONCLUSION
oscillator with a tail current source [3] sepaA quad-core 130-nm CMOS oscillator covrate for each core. The supply to only one core is acering
the 57-64 GHz frequency band is pretive at the same time. The supplies are controlled by
a digital network with two bit input control word. sented. It outputs -5 dBm and has -110 dBc/Hz
This digital network is built of NAND-gates and in- phase noise at 3 MHz offset frequency.
verters. NMOS varactors are used at each core
R EFERENCES
for fine tuning. Coarse frequency tuning is accom- [1] Federal Communications Commission, Amendment of Parts 2,
plished by changing core. The reason for using mul15 and 97 of the Commissions Rules to Permit Use of Radio
Frequencies Above 40 GHz for New Radio Applications, FCC
tiple cores is the limited performance of varac95-499, ET Docket No. 94-124, RM-8308, Dec. 1995.
tors at 60 GHz, preventing high phase noise per[2] E. R. C. (ERC), The European Table of Freformance and tuning range from being achievquency Allocations and Utilisations Covering the Frequency Range 9 kHz TO 275 GHz, ERC REable with a single core. The inductors are modPORT 25, http://www.ero.dk/documentation/docs/doc98/
eled using [4] and have a Q of about 30 at 60 GHz.
official/pdf/ ERCREP025.PDF, Copenhagen: 2004.
The cores output from -7 to -4 dBm, with an av- [3] A. Kral, F. Behbahani, and A. Abidi, “RF-CMOS Oscillator with
Switched Tuning,” Custom Integrated Circuits Conference, pp.
erage of -5 dBm output power over the fre555–557, May 1998.
quency range, Fig. 1. The simulated phase noise is
[4] N. Troedsson, http://www.indentro.com.
-110 dBc/Hz at 3 MHz offset frequency. The core
60
GHz Symposium 5-6 March 2008
Session IV
1030-1210 Thursday 6 March 2008
61
GHz Symposium 5-6 March 2008
GaN HEMT Development for Microwave Power Applications
- Current Status and Trends Masaaki Kuzuhara
Department of Electrical and Electronics Engineering, University of Fukui
3-9-1 Bunkyo, Fukui 910-8507, Japan
Email: kuzuhara@fuee.fukui-u.ac.jp
GaN is a direct semiconductor with a bandgap energy of 3.4eV. Nevertheless it has an
excellent high-field drift velocity of over 2x10 7 cm/s. making it suitable for high-voltage and
high-frequency applications. Furthermore, GaN has a capability of constructing various
heterojunctions for high-frequency operation using a modern epitaxial technology. Thus GaN and
related heterojunctions are more flexible to adjust material properties to their specific device
applications. For example, a properly designed AlGaN/GaN HEMT is easy to scale its gate length
down to less than 50nm without much sacrificing its breakdown behavior, indicating that
ultra-short channel GaN HEMTs are particularly suited for transmitter applications at
millimeter-wave frequencies and for post-CMOS n-channel FETs [1].
To achieve outstanding microwave performance, several improvements have been thus far
made in the basic device structure, including passivation, epi-layer design, metallization, and
field plates. Wu et al. employed a gate field-plate technology to achieve a record output power
density of 32.2W/mm operated at 4GHz with a drain bias (V dd ) of 120V using an AlGaN/GaN
HEMT fabricated on a SiC substrate [2]. Okamoto et al. reported a record one-chip microwave
output power of 230W (CW) at 2GHz with V dd =53V using a GaN HEMT with a gate width (W g ) of
48mm [3]. At C-band (5GHz), a record one-chip output power of 167W was reported with
W g =24mm operated at V dd =65V [4]. As a microwave amplifier, Mitani et al. reported a pulsed
output power of 1kW at S-band (3.2GHz) operated at V dd =80V using 4-dies of an internally
partial-matched GaN HEMT chip with W g =36mm fabricated on a SiC substrate [5]. At
millimeter-wave frequencies, Murase et al. achieved a record CW output power using a 0.2µm
recessed-gate GaN HEMT chip with a newly-designed source via-hole structure. The fabricated
one-chip FET amplifier with W g =6.3mm operated at V dd =25V delivered an output power of 20.7W
with a linear gain of 5.4dB at 26GHz [6].
At the moment, device-quality heterojunctions other than AlGaN/GaN are still not easy to
prepare because of the difficulty in epitaxial growth. With the development of high-quality bulk
nitride substrates, this problem would be considerably solved. One such promising example is the
use of InGaN with high In composition as a channel material. Simulation results for THz
operation using an AlInN/InN material system will be presented at the symposium.
References
[1] Y. Ohno and M. Kuzuhara, IEEE Trans. Electron Devices, vol.48, pp.517-523 (2001).
[2] Y.-F. Wu et al., IEEE Electron Device Lett., vol.25, pp.117-119, (2004).
[3] Y. Okamoto et al., IEEE Trans. Microwave Theory Tech., vol.52, pp.2536-2540 (2004).
[4] K. Yamanaka et al., IEICE Tech. Report, ED2007-210, pp.23-28 (2008) (in Japanese).
[5] E. Mitani et al., Proc. EuMIC Conf., pp.176-179 (2007).
[6] Y. Murase et al., IEICE Tech. Report, ED2007-212, pp.33-38 (2008) (in Japanese).
62
GHz Symposium 5-6 March 2008
Paving the road for integrated gallium nitride transceivers
Kristoffer Andersson, Mattias Thorsell, Niklas Billström*, Joakim Nilsson*, Jonas Holmkvist*, Ann-Marie
Andersson*, Mattias Südow, Martin Fagerlind, Per-Åke Nilsson, Anna Malmros, Hans Hjelmgren, Niklas
Rorsman
GigaHertz Centre, Microwave Electronics Laboratory, Department of Microtechnology and Nanoscience,
Chalmers University of Technology, Göteborg, SE-41296, Sweden
* Saab AB (incl. Saab Microwave Systems, Saab Avitronics and Saab Bofors Dynamics)
Gallium nitride is a very attractive material for manufacturing high-frequency microwave transistors. The
high saturated electron velocity, high breakdown field combined with the good thermal conductivity of
the silicon carbide substrate makes it almost ideal for high-power transistors. The output density of GaN
is at least 5 W/mm with reported hero numbers of 40 W/mm at GHz frequencies. Also at millimeter
wave frequencies there are reports of power densities of 2 W/mm. Besides the excellent power
properties of GaN recent results indicates noise performance close to what was achieved in GaAs a few
years ago. Typically Fmin values of 1 dB at X-band are reported. There are preliminary results showing
that GaN HEMTs are capable of withstanding significant input powers; up to 36 dBm. If these results
hold, it will be possible to eliminate the input protection circuitry and thus remedy the higher noise
figure of GaN. Thus GaN seems to be a very capable candidate for future integrated transceivers for
radar/communication and electronic warfare applications. This work reports on the first design steps
towards such integrated transceivers.
During the last year several integrated gallium nitride circuits have been fabricated at Chalmers. Below
are a SPDT-switch and a resistive feedback amplifier. The SPDT switch is a broadband switch (DC-18 GHz)
with an insertion loss less than 3 dB at the high-frequency end. The switch could handle at least 3W of
input power. The resistive feedback amplifier provides more than 8 dB of gain in the 1 – 6 GHz
frequency band with a noise figure of about 4 dB.
Acknowledgements
This research has been carried out in the GigaHertz Centre in a joint research project financed by
Swedish Governmental Agency of Innovation Systems (VINNOVA), Chalmers University of Technology
and Saab AB.
63
GHz Symposium 5-6 March 2008
Demonstrator of Class-S Power Amplifier
Andrzej Samulak + , Georg Fischer ∗ , and Robert Weigel +
+
Lehrstuhl für Technische Elektronik, Friedrich-Alexander-Universität Erlangen-Nürnberg,
91058 Germany, E-mails: {Samulak, Weigel}@lfte.de
∗
Alcatel-Lucent, Bell Labs Europe, 90411 Nürnberg, Germany,
E-mail: georgfischer@alcatel-lucent.com
Abstract— A realized demonstrator of a Class-S Power
Amplifier based on GaN transistors is presented. An input
signal is prepapred externally in simulative way (ADS,
Matlab) and is based on a 4th order Bandpass ∆-Σ
modulator definition. FPGA board is engaged as a signal
generator. The demonstrator is based on a Current Switching Class D (CSCD) topology. The presented solution offers
an output power of at least 37 dBm. Module efficiency is
20% (30% drain efficiency respectively) for full scale 3GPP
∆-Σ input signal. Center frequency is fixed at 225 MHz,
but analyses and maesurements requires wider bandwidth.
Fig. 1.
I. OVERVIEW
General idea of Class - S Amplifier.
amplifier are the switching transistors, which are driven
by the preamplifier. Devices work in antiphase, when
the first one is in ”ON” state (conducts current), the
second one is in ”OFF” state. In practice, active device
imperfections limit maximum bandwidth and sampling
rate. In addition, reactive parasitics (bond wires, Cds ,
Cgs ) and memory effects [3] contribute to the losses at
higher frequencies, which decreases total efficiency of
the SMPAs. The output signals from the switches are
combined and filtered in the last part of the whole circuit.
The transistor supply circuit is integrated with the output
part of the amplifier. Finally, losses in the output circuit
decide on the overall amplifier efficiency. The topology
of a Class S amplifier assumes an existing feedback
loop from SMPA output to ∆-Σ Modulator input, which
assists adopting modulation parameters to circuit nonidealities and output signal conditions. In addition this
can be seen as some sort of self linearization.
Linear classes of amplifiers offer high output power
and low distortions, but efficiency is still unsatisfactory.
Moreover, high power amplifiers dedicated for base
stations occupy more than a half the form factor of abasestation cabinet. These main limits and problems make
the idea of Switch Mode Power Amplifiers (SMPA)
more attractive. The presented Class-S amplifier is based
on Current Switching Class D (CSCD) configuration,
which despite supply problems, is more attractive with
RF demonstrator solutions [1], [2]. A Class-S amplifier consists of 4 main parts. The first part the ∆Σ Modulator converts the RF input signal to a delta
sigma modulated ”Digital RF” signal. This conversion
is necessary, because the transistors operate as switches
and a two-state input signal is required. The ∆-Σ Modulator is mainly dedicated to conversion narrow band
signals, but can be as well employed in 3GPP wideband
signal conversion. The main problem of bandpass ∆Σ Modulator is the operation on oversampled signals.
The second part of a Class-S PA is the preamplifier with
level shifter. The preamplifier provides ∆-Σ Modulated
signal in two antiphase bit trains, amplifies to the proper
level and shifts them to the appropriate gate voltage
swing. The preamplifier’s parameters depend on the
transistor technology used as switches.The main part of
R EFERENCES
[1] Johnson, T.; Stapleton, S.; ”Available Load Power in a RF Class
D Amplifier with a Sigma-Delta Modulator Driver” Radio and
Wireless Conference, Sept. 2004 Pages:439 - 442 .
[2] Kobayashi, H.; Hinrichs, J. M.; ”Current-mode class-D power
amplifiers for high-efficiency RF applications”, Microwave Theory and Techniques, IEEE Tran. on Vol. 49, Issue 12, Dec. 2001.
[3] S. C. Binari, P. B. Klein ”Trapping effects in GaN and SiC
Microwave FETs”, Proceedings of the IEEE, Vol. 90, June 2002.
64
GHz Symposium 5-6 March 2008
30/20 GHz Balanced Sub-harmonic MMIC Mixer
for Space Applications
Dennis Kleén dennis.kleen@space.se, Joakim Thelberg joakim.thelberg@space.se
Saab Space AB, Göteborg, Sweden
Summary: A sub-harmonic MMIC mixer with an integrated pre-amplifier has been developed. The RF
input frequency range is 27-31 GHz and the IF output frequency range is 18-21 GHz. Due to the subharmonic mixing the LO frequency is in the range of 4.3-5.8 GHz. The mixer shows a conversion gain
of 3 dB with a noise figure of 4.2 dB. The mixer is based on a zero bias diode implemented in an Emode technology. The circuits were fabricated in the OMMIC ED02AH process.
I.
Introduction
The telecommunication market in general is experiencing a rapidly increasing demand for
broadband connections and interactive services. These services may include HDTV,
interactive multimedia services, internet access etc. There is therefore a need for new
components and MMICs for these Ka-band systems.
II. Pre-Amplifier
The pre-amplifier was designed to cover 27-31 GHz, realised as a two-stage design, based on
a 6x25 um and a 6x50 um FET. The input stage utilizes an inductive series feed back in order
to achieve a good input match as well as low noise figure. The amplifier also contains a
voltage controlled 3 dB attenuator at the output. The overall maximum gain is 15 dB with a
noise figure of 3.2 dB and an output third order intercept point of >20 dBm.
III. Mixer Core
The balanced sub-harmonic mixer was designed for an RF input frequency of 27-31 GHz and
an IF output frequency of 18-21 GHz. The LO frequency range was 4.3-5.8 GHz. The mixer
core is based on eight 4x60 um zero bias diodes realised in an e-mode FET technology. For
sub-harmonic mixing the diodes are connected as anti-parallel pairs, with two diodes in each
pair for improved linearity. The LO is applied in anti-phase to each of the pair by a spiral
Marchand balun. The termination of the LO harmonics is crucial for high IP3 performance.
The RF signal is applied and the IF signal extracted via a distributed diplexer connected to the
common point of the anti-parallel diode pairs. Typical conversion loss for the mixer core is
10 dB. The 4LO-to-IF isolation was 75 dBc.
IV. Results
The multi-function chip with the pre-amplifier and
mixer core combined, showed a conversion gain of
+3 dB with a noise figure of 4.2 dB. The 4LO-to-IF
isolation was measured to 75 dBc. Chip size is
2.5x5 mm2, shown in figure 1.
Figure 1
MMIC Chip
V. Conclusion
A balanced sub-harmonic mixer with an integrated pre-amplifier has been described.
Demonstrated performance is good and makes it an interesting candidate for future broadband
space applications.
[1]
[2]
[3]
[4]
References
Andreas Ådahl, Herbert Zirath, “A Reliable Ka-Band Sub-harmonic Mixer for Satellite
Converters”, Compound Semiconductor IC Symposium 2006
Paul Blount & Charles Trantanella, “A high IP3, Subharmonically Pumped Mixer for LMDS
Applications”, IEEE GaAs IC Symposium 2000
C.A. Zelley, A.R. Barnes and R.W. Ashcroft, “A 60 GHz double balanced sub-harmonic mixer
MMIC”, Gallium Arsenide Applications Symposium. Gaas 2001
M van der Merwe, JB de Swardt, “The Design and Evaluation of a Harmonic Mixer Using an
Anti-Parallel Diode Pair”
65
GHz Symposium 5-6 March 2008
Water Vapour Radiometer for ALMA
A. Emrich, M. Wannerbratt, Omnisys Instruments
ABSTRACT
ALMA, the Atacama Large Millimeter/
submillimeter Array, will be a single research
instrument composed of 50 (or more) highprecision antennas, located on the Chajnantor
plain of the Chilean Andes in the District of
San Pedro de Atacama, 5000 m above sea
level. ALMA will operate at wavelengths of
0.3 to 9.6 millimetres, where the Earth’s
atmosphere above a high, dry site is largely
transparent, and will provide astronomers
unprecedented sensitivity and resolution. The
50 antennas of the 12 m Array will have
reconfigurable baselines ranging from 150 m
to 18 km forming a big interferometer.
The water vapour radiometer is a support equipment used to compensate for the differences in
phase due to different amount of water vapour in the atmosphere above each antenna. Other
interferometers have been built without this compensation, but the availability of the
instruments has then been very low.
The water vapour radiometer is measuring the energy on one of the spectral lines of water,
183.31GHz (1.6 mm). Since the measured signal is very weak, the measurement is calibrated
by alternately measure on well defined hot or cold loads, so called Dicke-Switch radiometer.
The overall accuracy of the WVR is 2K brightness temperature over the range from 50K to
370K To be able to reach this accuracy, the front end and parts of the back end is temperature
controlled to better than 0.1K.
Omnisys will build up to 58 radiometers, including spares and option of 5 extra. The expected
lifetime will be at least 15 years. This implies that high effort must be taken to design in a
way that makes the WVR cost effective by means of MAIT.
Subsystems of the WVR:
Quasi optics, one flat mirror and three
active mirrors, chopper wheel with mirrors
for hot and cold loads.
Front End, corrugated horn antenna, low
noise (NF=7.5 dB) integrated schottky
mixer and LNA (30 dB), local oscillator
system based on a DDS controlled 15 GHz
VCO and active multiplier chain.
Back End, four channel filter bank with
diode detectors and 50 dB of distributed
stable amplification.
Control and power, thermal control, motor
control, computing and communication.
66
GHz Symposium 5-6 March 2008
European Radio & Microwave Interest Group - EuRaMIG
An initiative from GigaHertz Centre
Status and coming activities
Peter Olanders, Ericsson AB, Steering Board Chairman GigaHertz Centre
peter.olanders@ericsson.com
Jan Grahn, Chalmers University of Technology, Director, GigaHertz Centre
jan.grahn@chalmers.se
The European industry depending upon RF/microwave technologies is scattered.
As a result, it has been difficult to gather and coordinate European microwave
stakeholders for creating impact on the formulation of European Framework
Programs. In order to meet this challenge, GigaHertz Centre took the initiative to
organise a meeting in Brussels on 10 January 2008 between representatives for the
European Commission, a broad range of European industrial stakeholders,
institutes and universities active in the RF/microwave field. The meeting resulted
in the formation of European Radio & Microwave Interest Group (EuRaMIG).
This is an industrial-academic-public network open to anyone who wants to
contribute and support the development of European RF/microwave technologies.
All viewgraphs from the meeting in Brussels including comments and current
status can be found on www.chalmers.se/ghz
EuRaMIG is setting up working groups among European stakeholders to describe
the future of European microwave industry and its position and impact on
European growth, and the necessary research priorities and academic education to
be carried out. The vision is a Europe taking the global lead in RF/microwave. The
output from the EuRaMIG working groups will be addressed to the European
Commission.
In GigaHertz Centre, Chalmers University of Technology and seven companies
(Ericsson, Infineon Technologies, Saab, NXP Semiconductors, Sivers IMA,
Omnisys Instruments, Comheat Microwave) conduct joint research and innovation
in microwave power (RF amplifiers and GaN circuits), system-on-chip mm-wave
solutions and low phase-noise circuit based oscillators. The research is partly
supported by the Swedish Governmental Agency for Innovation Systems
(VINNOVA) in the VINN Excellence program. Web page: www.chalmers.se/ghz
67
GHz Symposium 5-6 March 2008
Workshop
Antennas
1300-1430 Thursday 6 March 2008
68
GHz Symposium 5-6 March 2008
Integrated Antennas for RF MEMS Routers
#
A. Rydberg#, S. Cheng#, Paul Hallbjörner*, S. Ogden**, and K. Hjort**
Signals and Systems Group and **Microsystems Group, Department of Engineering Sciences,
Uppsala University. *SP Technical Research Institute of Sweden, Borås, Sweden
anders.rydberg@angstrom.uu.se
Summary:
Three antenna designs for flip-chip bonding with RF MEMS based switching routers at 20 GHz
will be presented. The prototypes are fabricated on 100 um thick Pyralux AP polyimide flexible
laminates [1]. Measured results show very good agreements with the full wave simulations.
Introduction:
A microwave router based on RF MEMS switching is a
novel application driven component for wireless
communications (e.g. in microwave links or sensor
networks). Integrating MEMS switches and transmission
lines on a single router chip will enable a cost-effective and
higher level of integration to be achieved.
A dipole, uniplanar Yagi antenna and four element Yagi
antenna array are investigated in this report. As shown in
Fig. 1, the dipole is placed in front of a ground plane which
represents the effects of the RF MEMS router on the
antenna performance and acts as a reflector. The uniplanar
Yagi antenna is proposed to achieve higher gain and better
front-to-back ratio than the dipole. Both the dipole and the
uniplanar Yagi antenna are designed for 50 Ω differential
feed. A 1:1 balun transformer on the same substrate as the
RF MEMS router is required in order to match each
antenna to a 50 Ω CPW of the router.
Fig. 1. Geometries of the four element
Yagi antenna array: Lsub=18 mm,
Lh=7 mm, L=5.9 mm, Ld=5.2 mm,
Lr=9 mm, Lf=0.73 mm, D=0.98 mm,
Dd=1.55 mm, Dr=1.25 mm, G=50
um, Wstrip=100 um, T=18 um and
h=100 um.
Results:
The antennas are well matched to the desired input impedance at the operation frequency. The
presented dipole and uniplanar Yagi antenna feature about 10 % 10 dB return loss bandwidth at
20 GHz. The maximum gain of 3.8, 4.8 and 6.8 dBi is achieved, respectively. Moreover, good
front-to-back ratio and appropriate beam coverage of all the antennas are observed.
Conclusion:
This report presents three antenna designs for flip-chip bonding with RF MEMS switching
routers at 20 GHz. The antennas are good candidates for electrically steerable antenna arrays.
References:
1. Shi Cheng, Erik Öjefors, Paul Hallbjörner, Sam Ogden, Joakim Margell, Klas Hjort, Anders Rydberg, "Body
Surface Backed Flexible Antennas for 17 GHz Wireless Body Area Networks Sensor Applications," published in the
10th European Conference on Wireless Technology, Oct. 2007.
69
GHz Symposium 5-6 March 2008
Microstrip patch antennas for wireless applications
Naima Amar Touhami, Beatriz Aja, Antonio Tazón, Eduardo Artal
(naima@dicom.unican.es)
Departamento de Ingenieria de Comunicaciones. ETSIIT. Avenida de los Castros, s.n.
Universidad de Cantabria. Santander (Spain)
Summary
A 3.5 GHz planar antenna system has been developed for wireless applications [1]. The
antenna structure is based on microstrip patch arrays [2]. Each patch is excited by a slot
antenna coupled to a microstrip line. The antenna system has two arrays, one for the
transmitter the other for the receiver, and meets the high isolation required between both.
Antenna manufacturing is based in the combination of different substrates for microstrip
lines and radiation elements.
Introduction
Planar antennas are well suited for wireless applications when space saving and
compact terminals is a must. For short and medium range applications at 3.5 GHz the
antenna gain requirements are about 14 dB. To achieve these gain values an array of
printed patch radiators was designed, simulated with an Electro-Magnetic software tool
and experimentally tested. Next Figure shows two pictures, first one with two antenna
prototypes (four or eight patches), the second one shows the bottom side with microstrip
power dividers and slot excitation (eight patches version).
Results
Different versions of such array antennas have been used to compare their performances
according to the dielectric substrates used and their mechanical arrangement. Typical
achieved values in a 3.5% bandwidth centred at 3.52 GHz are: input return losses better
than 12 dB, isolation between coaxial ports better than 60 dB and antenna gain around
14 dB.
Conclusion
Planar antennas based on microstrip patches are well suited for wireless terminals
working at 3.5 GHz. It is possible the operation of a complete transceiver with separate
antennas given the high isolation achieved between them.
References
[1] Satish K. Sharma and L. Shafai “Performance of a Microstrip Planar Array Antenna at
Millimeter Wave Frequencies Using a Series Parallel Feed Network” IEEE Antennas and
Propagation Society International Symposium, 2001, vol.3, pp. 594-597
[2] L. Cabria , J. A. García , A. Tazón , J. Vassal'lo , “Active reflectarray with beamsteering
capabilities”, Microwave and Optical Technology Letters, 2005, Vol. 48 , pp. 101 - 105
This work has been supported by the Ministerio de Industria, Comercio y Turismo (Spain), PROFIT grant reference
FIT-330220-2006-75
70
GHz Symposium 5-6 March 2008
Small Microstrip Fractal Antenna for RFID tag
P. Enoksson1, M. Rusu1,2, A. Curutiu1,3, H. Rahimi1, C. Rusu4
1
Chalmers University of Technology, Micro and Nanosystems group, Göteborg, Sweden
2
Faculty of Physics, Bucharest University, Bucharest-Magurele, Romania, mvrusu@yahoo.com
3
Bonn University, Bonn, Germany
4
Imego AB, Göteborg, Sweden
Feasibility study of using fractal-based multi-band antenna for RFID tag is presented. The
requirements are: one antenna layout for two bands (868MHz, 2.45GHz), small size (3x3cm), and
reading distance of at least 10cm and 30cm for 868MHz and 2.45GHz, respectively. The
simulations show that fractal antenna can be useful for RFID tags.
The RFID antenna requirements are at the limit of conventional antenna [1], so we
propose to use fractal-shape antenna. Fractals have long perimeter packed in a small area
and an intricate (tortuous) structure allowing for many possible bands of resonances which
are not necessarily harmonics [2]. This type of antenna has gained interest only in the last
decade [3-5].
A number of fractal curves have been simulated by ADS Momentum (Agilent
Technologies) and Minkowski is the most promising one (Figure 1).
(a)
(d)
(b)
(e)
(c)
(f)
Figure 1. The six fractal
shapes tested for antenna
layout: (a)Koch, (b)Sierpinski,
(c)‘tree’, (d)‘cuboid’, (e)‘logperiodic’, (f)Minkowski
Design a small antenna (physical length << wavelength of operation) satisfying
simultaneously all the requirements asks for a compromise among parameters such as
different resonance frequencies with reasonable high |S11| values, desired impedance and
the highest possible antenna efficiency. Statistical correlation between S11 and the
resonance frequency (Figure 2) for our Minkowski-based fractal antenna shows that the
desired requirements could be obtained.
Figure 2. Statistical
correlation between S11 and
resonance frequency for
simulated Minkowski-based
fractal antenna.
References:
1. Wheeler H., Small Antennas, IEEE Transactions on Antennas and Propagation, 23(4), 1975.
2. Cohen N., Fractal and Shaped Dipoles, Communications Quarterly, p.25-36, Spring 1996.
3. Puente C., "Fractal Antennas", Ph.D. Dissertation at the Dept. of Signal Theory and Communications,
Universitat Politècnica de Catalunya, June 1997.
4. M. Rusu, R. Baican, Adam Opel AG, Patent 01P09679, “Antenne mit einer fraktalen Struktur”, die auf
der Erfindungsmeldung 01M-4890 “Fractal Antenna for Automotive Applications” basiert, 18 Okt. 2001.
5. Rahimi H., Rusu M., Enoksson P, Sandström D., Rusu C., Small Patch Antenna Based on Fractal Design
for Wireless Sensors, MME07, 18th Workshop on Micromachining, Micromechanics, and Microsystems,
16-18 Sept. 2007, Portugal.
71
GHz Symposium 5-6 March 2008
This research has been carried out in the Chase VINN excellence Centre
Dual-band choke horn Eleven Feed
1
Adeel Yasin1, Jian Yang1, Tomas Östling2
Chalmers University of Technology ,2 Arkivator AB
E-mail: yasin@student.chalmers.se
Short Summary
The aim of this research is to propose a compact solution of combining Eleven antenna
and choke horn as a feed for dual band reflector antenna. A novel circular geometry of
Eleven antenna has been investigated for improving system performance. The combined
solution is a low-cost, low-profile wide-band solution for satellite terminals.
The simulated and measured performance of the prototype is presented in the paper.
Introduction
Multiband antennas are of interest for many applications, for example, in satellite
communications. Some of the benefits are simultaneous operation at several frequency
bands, independent tracking control and different polarization schemes. In this research,
frequency bands of interest are C and Ku bands for satellite communication applications.
Analysis studies have been done for best performance of combined feed parameters.
The novel circular geometry of Eleven feed (Fig.1) has better BOR1 efficiency as
compared to conventional linear Eleven feed configuration [1].Choke horn [3] design has
been modified to adjust circular Eleven feed for required operational bands(Fig.1). The
combined feed has been characterized in terms of aperture efficiency of reflector and
coupling efficiency due to mutual coupling between the two feeds. The mono-pulse
tracking capability has also been verified for multi-port model of circular Eleven feed.
Results
Frequency Bands: [5.5-6.5,10.5-14.5]GHz
Directivity ~10dBi
Peak XP sidelobe Level < -15dB
Aperture Efficiency > -2.5dB @ 50D
Coupling Efficiency > -0.5dB
Input Reflection co-efficient <-10dB
Conclusions
1. Dual-band and dual polarized.
Fig1.Dual Band choke horn Eleven feed
2. Wideband with capabilities for tri-bands as (C/Ku/Ka)
References
[1] R. Olsson, P.-S. Kildal and S. Weinreb, “The Eleven antenna: a compact low-profile
decade bandwidth dual polarized feed for reflector antennas,” IEEE Trans. Antennas
Propas., vol. 54, No. 2, Feb. 2006.
[2] P.-S. Kildal, R. Olsson and Jian Yang, “Development of three models of the Eleven
antenna: a new decade bandwidth high performance feed for reflectors,” Proceedings of
EuCAP 2006, Nice, November 2006.
[3] Z. Ying, P.-S. Kildal, and A. Kishk, “A broadband compact horn feed for prime-focus
reflectors”, Electronics Letters, vol. 31, 14, pp.1114‚1115, 6th July, 1995.
72
GHz Symposium 5-6 March 2008
OPTIMIZATION OF 200-800MHZ ELEVEN FEED FOR USE IN
REFLECTOR ANTENNAS OF GMRT
Y.B. Karandikar*, P.S. Kildal †
*PhD student,Chalmers Univ. of Technology, Sweden
†
Prof., Chalmers Univ. of Technology, Sweden.
yagesh@student.chalmers.se
per-simon.kildal@chalmers.se
This paper discusses in brief the electrical design of the Eleven Feed for the parabolic
reflector antennas of Giant Meterwave Radio Telescope (GMRT-India) optimized for 200
to 800 MHz bandwidth. Paper also gives the comparison between the simulated and
measured antenna input reflection coefficient and radiation patterns. The measured S11 of
the feed is less than -8dB and the computed total efficiency including the mismatch losses,
is better than -3dB over desired bandwidth for the reflector antennas of GMRT having
62.5° of half subtended angle.
The S11 of the current design shows 2-3dB improvement as compared to previous designs
of Eleven Feed [1][2][3]. To achieve this, efforts are taken to first understand conventional
log-periodic arrays of folded dipoles defined by set of optimization parameters like scaling
constants, spacing constants, folded dipole lengths, wire diameters etc. This understanding
is then further extended to match folded dipole log-periodic arrays made up of metal plate
of constant thickness to 200ohm in free space. Different matching techniques like shorting
stub on longest dipole, truncation of array by shortening the lengths of first and last folded
dipole, are applied to get simulated S11<-10dB. Finally the folded dipole log-periodic
arrays made up of plates designed in free space are used in Eleven Feed configuration and
designed is further optimized to get simulated S11<-10dB for Eleven Feed.
The difference between the measured and simulated S11 is due to mechanical tolerances in
the manufacturing process. And reason of low aperture efficiency is the deep dish of
GMRT having 62.5° of half subtended angle where the pattern of the Eleven Feed is more
suitable for half subtended angles of 50-55°. This means the GMRT reflector is somewhat
under illuminated.
Figure 1: Manufactured Eleven Feed
[1] R. Olsson., P. S. Kildal, S. Weinreb. “A novel low profile log-periodic ultra wideband feed for dual
reflector antenna of US-SKA”, IEEE Symposium Antennas & Propagation, pp. 3035-3038, (2004).
[2] R. Olsson., P. S. Kildal, S. Weinreb. “Measurements of a 1 to 13 GHz lab model of dual polarized low
profile log-periodic feed for US-SKA”, IEEE Symposium. Antennas & Propagation, pp. 700-703,
(2005).
[3] R. Olsson., P. S. Kildal, S. Weinreb. “The Eleven Antenna: A compact low profile decade bandwidth
dual polarized feed for reflector antennas”, IEEE Trans. Antennas & Propagation, volume 54, pp. 368375, (2006).
73
GHz Symposium 5-6 March 2008
METHOD FOR CIRCUIT BASED OPTIMIZATIONS
OF RADIATION CHARACTERISTICS OF MULTIPORT ANTENNAS
Kristian Karlsson(I)
(I)
Email: {kristian.karlsson, jan.carlsson}@sp.se
SP Technical Research Institute of Sweden
501 15 Borås, Sweden
Jan Carlsson(I) (II)
(II)
Chalmers University of Technology
412 96 Gothenburg, Sweden
Summary
This contribution presents a method for using data from full-wave EM
simulations in combination with a circuit simulator to simplify antenna integration and
antenna parameter evaluation.
Introduction Today’s compact multi functioning mobile terminals offers complex problems
to the antenna and system engineer. Several antennas have to be located within a limited
space and will of course couple to each other as well as to other units within the terminal
(EMC problems). The presented method [1,2] can be applied on multi-port antennas including
arbitrary circuits and feeding networks connected to their ports. Studies can be performed on
far field antenna parameters (e.g. efficiency, diversity gain, and correlation) as well as near
field properties (EMC, SAR). The contribution also describes how the method is used in
combination with a global optimization scheme.
Result
The method is implemented in a software named MPA (Multi Port Antenna
evaluator) which can compute antenna parameters such as: total radiated power, radiation
efficiency, total radiation efficiency, radiation pattern, near field, user defined transfer
functions (mutual coupling, isolation, impedance), correlation and apparent diversity gain.
The MPA software handles input data from several commercial full-wave EM simulators
(CST, EMDS, MicroStripes and IE3D) and has been used in several studies concerning
diversity antennas and EMC characteristics.
Conclusion A method for computation of radiation patterns and near fields using full-wave
EM simulator data in combination with a circuit simulator is described. The method will help
to reduce the number of full wave simulations needed during the integration process of e.g. a
multi port antenna in a terminal. It can also be used in combination with a global optimization
scheme to improve e.g. radiation efficiency, diversity gain or near field characteristics.
Acknowledgement This research has been carried out in the Chase VINN Excellence Centre
at Chalmers.
References
[1]
[2]
K. Karlsson, J. Carlsson, I. Belov, G. Nilsson and P.-S. Kildal, “Optimization of antenna
diversity gain by combining full-wave and circuit simulations”, Proceedings of EuCAP
2007, UK, 11-16 Nov. 2007, MoPA.36.
D.F. Kelley, “Embedded element patterns and mutual impedance matrices in the
terminated phased array environment”, Proceedings of IEEE AP-S International
Symposium, USA, 3-8 July 2005, vol. 3A, pp. 659-662.
74
GHz Symposium 5-6 March 2008
Circular Monopole and Dipole Antennas for
UWB Radio Utilizing a Flex-rigid Structure
Magnus Karlsson, and Shaofang Gong, Member, IEEE
Department of Science and Technology, Linköping University, SE-601 74 Norrköping, Sweden
Phone: +46 11 363491, E-mail: magka@itn.liu.se
T
he general focus during the era of UWB antenna
development has so far been on the antenna element but
not so much on how the antenna can be integrated and used in
a UWB system. In this paper the concept of utilizing a flexible
and rigid (flex-rigid) substrate is presented. Using this flexrigid concept the antenna is made on the flexible part of the
flex-rigid structure, and other circuits, e.g., a balun can be
integrated in the rigid part.
Figs. 1a and 1b show the substrate structure in which two
dual-layer NH9326 laminates are bonded together with a
polyimide-based flexible substrate. Fig. 1c shows a circular
dipole antenna integrated in the flex-rigid substrate. The balun
(in the rigid part) utilizes broadside-coupled microstrips.
Metal 1
Rigid
Flex
Rigid
Fig. 2a shows voltage standing wave ratio (VSWR)
simulation of a circular monopole, and a dipole antenna on the
flex-rigid substrate. It is seen that the designed circular
monopole antenna has a wide impedance bandwidth using the
proposed flex-rigid structure. It covers the entire UWB
frequency band 3.1-10.6 GHz at VSWR<2. Moreover, Fig. 2a
shows also VSWR simulation of the dipole antennas. It is seen
that the circular dipole antenna has a wide impedance
bandwidth using the suggested flex-rigid structure and the
balun is the component limiting the bandwidth. More
precisely, the circular dipole antenna implemented using the
flex-rigid substrate can cover the Mode 1 UWB frequencybandwidth (3.1-4.8) at VSWR<1.54 without a balun and
VSWR<1.68 with a balun. The circular dipole antenna
radiation simulation shown in Fig. 2b indicates that the
antenna has a typical dipole-antenna radiation pattern.
Metal 2: antenna
Monopole
Dipole
Dipole with balun
5
Metal 3: ground
4
VSWR
Metal 4
(a) Substrate cross-section.
Rigid
3
2
Flex
1
0
1
2
3
4
5
6
7
8
9 10 11 12
Frequency (GHz)
(a) VSWR simulation.
Rigid
(b) Bendable property.
φ=0º
0
0
Circular dipole
antenna
330
30
-5
-10
-15
300
60
-20
Polyimide
foil
-25
-30 270
90
-25
38 mm
-20
Rigid part
-15
240
120
-10
Single ended
feed-line
-5
0
210
150
180
(c) Circular dipole antenna.
(b) Normalized Eθ radiation pattern at 3.960 GHz, φ=0º.
Fig. 1. Antenna design: (a) detailed cross-section, (b) bendable property, and
(c) circular dipole antenna.
Fig. 2. Simulation of the monopole and dipole antennas: (a) VSWR
simulation, and (b) normalized radiation pattern.
75
GHz Symposium 5-6 March 2008
Design, Manufacture and Test of Eleven Feed for 1-13 GHz
Jian Yang, Ingmar Karlson, Xiaoming Chen and Per-Simon Kildal
Department of Signals and Systems
Chalmers University of Technology
S-412 96 Gothenburg, Sweden
E-mail: jian.yang@chalmers.se
Short Summary: A simple one-by-one parameter optimization scheme has been applied to
design the Eleven feed for 1-13GHz. The dipoles have been printed on a thin dielectric film
printed circuit board (PCB) technology is used to manufacture the Eleven feed (Fig.1). A new
method of combining several commercial software tools (WILP-D – CST – ADS) has also
been investigated in order to more efficiently compute the performance of the Eleven feed
with the center puck included.
Introduction: Wide frequency band systems are required for many applications, such as in
radio astronomy and ultra-wideband (UWB) technology. Chalmers University of Technology
has during the last years been developing a new ultra wideband antenna – the Eleven antenna.
It has very good features such as nearly constant beamwidth, 10-11 dBi directivity and
almost fixed phase centre location over the whole bandwidth, low profile and simple
geometry. The Eleven antenna is very suitable for use as a feed in reflector antennas.
Previously, we have reported three Eleven feeds designed for use in different radio
telescopes; Green Bank, GMRT and RATAN. The purpose of the present project is to report
the development of an Eleven feed for 1 – 13 GHz. One critical performance of a feed for a
reflector antenna used in radio telescope is impedance mismatch, because strong reflection at
the antenna input port will increase the system noise level. Previously, a measured reflection
coefficient of -8 dB up to 3 GHz was reported, whereas attempts to make feeds at higher
frequencies were unsuccessful so far due to problems in manufacturing the small dipoles at
the higher part of the frequency range. A new 1-13 GHz Eleven feed has been designed by
using a one-parameter at the time optimization technique, manufactured and tested. The
present paper reports the results of the testing and compares with simulations.
Results: The measured reflection coefficient of the Feed is below -10 dB from 1 – 8GHz
(agreed with calculation) and -5dB up to 13 GHz due to the effect of the center puck. A new
method of accounting for the center puck in the design has been developed, based on using
different software tools for the center puck and the dipole panels and combining the results.
With the new method we show that we are able to predict the effect of the center puck. The
effect of the crossing of the feed lines of the two polarizations at center puck has been
investigated and shown to be essential for a successful design.
Conclusion: The reported design and manufacture method have improved the performance
of the Eleven feed up to 8GHz. The effects of overlapping and center puck have been
investigated numerically, and new solutions based on this work are proposed and will
certainly cause further improvements and enable Eleven feeds at even higher frequencies.
References:
[1]
R. Olsson, P.-S. Kildal and S. Weinreb, “ The Eleven antenna: a compact low-profile
decade bandwidth dual polarized feed for reflector antennas,” IEEE Trans. Antennas Propas.,
vol. 54, No. 2, Feb. 2006.
P.-S. Kildal , R. Olsson and Jian Yang, “Development of three models of the Eleven antenna:
a new decade bandwidth high performance feed for reflectors,” Proceedings of EuCAP 2006,
Nice, November 2006.
1
76
GHz Symposium 5-6 March 2008
Workshop
RF Power Amplifiers (2)
1300-1430 Thursday 6 March 2008
77
GHz Symposium 5-6 March 2008
Output Power Density and Breakdown Voltage in Field-Plated Buried Gate
Microwave SiC MESFETs
P.Å. Nilsson, F. Allerstam, K. Andersson, M. Fagerlind, H. Hjelmgren, A. Malmros, M. Südow,
E. Ö. Sveinbjörnsson, H. Zirath, and N. Rorsman.
Microwave Electronics Laboratory, Microtechnology and Nanoscience, Chalmers University of
Technology, SE-412 96 Goteborg, Sweden. per-ake.nilsson@mc2.chalmers.se
Silicon Carbide MESFETs, with field plates, were simulated, fabricated, and characterized. Using
filed plates, it was possible to reach a breakdown voltage for a microwave power device of 170 V,
and an output power density at 3 GHz of 8W / mm.
MESFETs made of Silicon Carbide (SiC) have long been a good candidate for microwave power
applications, and are now commercially available. Their suitability for radar and communication
applications is due to the high thermal conductivity, the high breakdown field, and the high
saturation velocity of SiC.
The maximum output power of the devices depends on the drain voltage. The drain voltage can be
increased by increasing the breakdown voltage of the device. In this work, we study the effect of
field plates on the breakdown voltage of SiC MESFETs.
180
Lg
FP
Source
Gate
160
oxide
Drain
n-cap
B VDG
Breakdown Voltage (V)
n-channel
p-buffer
SiC substrate
140
B VDS
120
100
80
0
200
400
600
800
fie ld p la te le n g th (n m )
1000
Fig. 1:. Measured breakdown voltage vs. field plate length (left) and simulated electric field in the
device close to breakdown (right). The inset shows the device cross section.
The devices made for this study were field-plated buried gate MESFETs [1] made at Chalmers
University. The field plate overlap was varied from 50 nm to 800 nm. A clear increase of breakdown
voltage was seen for longer field plates (fig. 1.) This is in accordance with physical simulations. The
output power at 3 GHz was measured for the devices and values up to 8 W/mm were achieved.
We have shown that the use of field plates is a useful way to increase the breakdown voltage, and
thus the output power, of silicon carbide MESFETs.
This research has been carried out in the Microwave Wide Bandgap Technology project financed by
Swedish Governmental Agency of Innovation Systems (VINNOVA), Swedish Energy Agency
(STEM) Chalmers University of Technology, and Ericsson AB, Saab AB, Norstel AB, Infineon,
NXP, Furuno, and Norse Semiconductor Laboratories AB.
References
[1] K. Andersson, M. Sudow, P. A. Nilsson, E. Sveinbjornsson, H. Hjelmgren, J. Nilsson, J. Stahl, H.
Zirath, and N. Rorsman, IEEE Electron Dev. Lett. 27, 573, (2006).
78
GHz Symposium 5-6 March 2008
Silicon-on-SiC hybrid substrate with low RF-losses and
improved thermal performance
J. Olsson, Ö. Vallin, D. Martin, L. Vestling, U. Smith, and H. Norström*
Uppsala University, The Ångström Laboratory, Solid State Electronics, P. O. Box 534, SE-751 21 Uppsala, Sweden
*
permanent address: Infineon Technologies AB, SE-164 81 Kista, Sweden
Email: jorgen.olsson@angstrom.uu.se, Tel: +46(0)18 471 3035
up to 30 GHz, is significantly higher for
BaSiC compared to the reference substrates.
1. Abstract
A novel SOI hybrid substrate ("BaSiC")
consisting of silicon-on-silicon carbide is
presented. Compared to ordinary SOI-material,
MOS transistors on the BaSiC substrate show
no self-heating and RF-measurements up to 30
GHz show more than a factor of ten lower
substrate losses.
4. Conclusions
A recently presented Si-on-SiC hybrid
substrate, called BaSiC, has been investigated.
The high thermal conductivity of SiC results in
an effective thermal resistance about a factor
of two lower than the SOI reference. Due to
the semi-insulating properties of the SiC, the
BaSiC substrate has very low losses at high
frequency as compared to SOI-substrates.
2. Introduction
SOI-technology in general offers advantages at
high-frequencies compared to ordinary silicon
technology [1]. However, the low thermal
conductivity of the buried silicon dioxide
insulator may cause self-heating of devices
and reduced performance. A new Si-on-SiC
hybrid substrate for high performance SOI was
recently presented and it addresses this
problem [2]. The hybrid substrate uses
polysilicon as an intermediate layer, see Fig. 1,
thereby avoiding the thermally unfavorable
SiO2 [3]. Previous efforts to improve the SOI
thermal properties include replacing the buried
SiO2 insulator by diamond [4], AlN [5], or
Al2O3 [6].
References
[1] F. Gianesello et al. IEEE SOI Conf. p. 119, 2007
[2] J. Olsson et al. IEEE SOI Conf. p. 115, 2007
[3] S. Whipple et al. MRS Proc. v.911, p.383, 2006
[4] B. Edholm et al. IEEE SOI Conf., p. 30,1997
[5] D. M. Martin et al. EUROSOI, p. 67, 2007
[6] C. de Beaumont et al. EUROSOI, p. 123, 2006
Si
SiO2
Si
3. Results
Manufactured MOSFET devices on BaSiC
show well behaved subthreshold and Id-Vd
characteristics without any self-heating effects,
whereas the SOI reference devices show selfheating. The effective thermal resistance of the
different substrates was measured using
heating resistors. A factor of two lower
effective thermal resistance is observed for
BaSiC compared to SOI, which is in good
agreement with results for similar substrates
[3]. As to RF performance, Fig. 2 illustrates
the relative resistive substrate losses. The
equivalent parallel resistance, measured on an
open-pad structure for a wide frequency range
Si
α−Si
SiC
Fig.1: Bonding process of the BaSiC hybrid wafer.
3E 5
1/real(Y) [ohm)
1E 5
m1
f re q =3 0 .0 0 G H z
1 /re al(m Y )= 5 4 2 9 .5 6 8
B A S I C d ark
B A S I C lig ht
1E 4
m1
S O P S IC
1E 3
1E 2
1E7
SoPSiC:
Commercial Si-onSiO2-on-SiC for
GaN epi growth
1E8
SOI
1E9
1E10
3E 10
fre q , H z
Fig.2: NVA measurements of the substrate losses
79
GHz Symposium 5-6 March 2008
A review of validation criteria for behavioral power amplifier models
P. Landin and M. Isaksson
University of Gävle, Dept. of Electronics, SE-80176 Gävle, Sweden.
email: min@hig.se
Abstract
The radio frequency power amplifier is still one of the most interesting components in a
telecommunication system; its power consumption dominates the other parts in the system. The purpose of
the radio frequency power amplifier is to amplify the radio signal to a necessary power level for
transmission to the receiver. The transmission is performed through the air. It is important to handle the
power amplifier's conflicting behaviors of efficiency and linearity in this process. The design of linear and
efficient radio frequency power amplifiers in modern radio telecommunication systems has been described
in the literature as one of the most challenging design problems.
One of the most promising techniques to overcome the interference problem due to the nonlinear
behavior of the power amplifier is the technique of digital predistortion. In order to work, digital
predistortion requires knowledge of the nonlinear characteristics of the amplifier since it, in principle,
applies the inverse of the raw amplifier to the signal prior to amplification. Accurate nonlinear
characterization of the amplifiers is necessary for several of these techniques and for optimizing the
amplifier design. Behavioral models, also denoted as black-box models, have attracted interest as a means
for characterizing power amplifiers.
Hence, the accuracy of the nonlinear behavior model is central and must be evaluated in some way so
that strong candidates for e.g. digital predistortion can be found among a variety of behavioral models. An
attractive way of comparing two models is to evaluate their performance on validation data. A validation
data set is one that has not been used to help construct any of the models that we would like to evaluate.
This procedure is often called cross-validation. The cross-validation technique is attractive because it
makes sense without any probabilistic arguments and assumptions about the true system. The quality of
the validation data is of great importance; the accuracy of the model can not be shown better than the
accuracy of the measurements. A challenge is therefore to create a measurements system that can measure
and deliver validation data with extra ordinary properties with respect to e.g. bandwidth and dynamic
range.
An obvious choice of a validation criteria is to calculate the total error of the model, i.e. the difference
between the measured output and the output of the model using validation data. Often it is calculated as
the normalized mean-square error (NMSE). A strong drawback with the NMSE is that it mainly measures
the in-band error while it is mainly the out-of-band error that is of interest.
In this paper we go through a number of validation criteria for radio frequency power amplifier
behavioral models, showing their pros and cons in a comparative way. Examples of measures that will be
considered is the NMSE, the adjacent channel power ratio, the power spectrum comparative method, the
adjacent channel error power ratio, the weighted error-to-signal power ratio, and the complementary
memory measures, the memory effect ratio and the memory effect modeling ratio.
[1]
[2]
[3]
[4]
S. C. Cripps, Advanced Techniques in RF Power Amplifier Design. Boston, MA: Artech House, 2002.
M. Isaksson, D. Wisell, and D. Rönnow, "A Comparative Analysis of Behavioral Models for RF Power Amplifiers,"
IEEE Trans. Microwave Theory Tech., vol. 54, pp. 348-359, Jan. 2006.
H. Ku and J. S. Kenney, "Behavioral Modeling of Nonlinear RF Power Amplifiers Considering Memory Effects," IEEE
Trans. Microwave Theory Tech., vol. 51, pp. 2495-2504, 2003.
D. Wisell, M. Isaksson, and N. Keskitalo, "A General Evaluation Criteria for Behavioral PA Models," in 69th ARFTG
Conf. Dig., Honolulu, HI, USA, 2007, pp. 251-255.
80
GHz Symposium 5-6 March 2008
CMOS for micro- and millimeter wave power
applications
Mattias Ferndahl1 , Hossein Nemati1 , Herbert Zirath1 .
1
Microwave Electronics Laboratory, Chalmers University of Technology. SE-41296 Göteborg, Sweden
S UMMARY
This abstract presents an empirical study of the large
signal behavior of CMOS devices from two different
nodes, 130 nm and sub 65 nm (Lphys ≤ 45nm) using
a load pull system at micro and millimeter wave frequencies to evaluate their use for millimeter wave power
amplifiers.
I. I NTRODUCTION
HE down scaling of CMOS technology has led to
very high transit frequencies and is becoming a
realistic alternative to III-V technologies for millimeter
wave applications with several successful design examples [1]–[3]. Not much has however been reported
regarding this issue. Above 15 GHz Vasylyev et al [4]
with 17.8 dBm output at 17 GHz represent highest power
and Yao et al [5] with 6.4 dBm at 60 GHz highest
power combined with high frequency. Previous work by
the authors [6] was limited to lower frequencies and
relatively small devices.
T
II. R ESULTS
Load pull measurements on different CMOS devices
with different gate lengths and widths have been carried
out to evaluate their power performance. At 23 GHz
output power levels of 22 dBm with 5 dB gain was
obtained for 1 mm devices using 130 nm CMOS. At
35 GHz the gain of these was found too low to be of
practical use. 40 nm gate length devices however showed
very promising performance with 12 dBm output power,
7 dB gain and a maximum power added efficiency (PAE)
of 33 %, for 192 μm devices. In the extensive work by
Scholvin et al [7], extending up to 20 GHz, they state a
maximum operating frequency of 20-25 GHz for the 65
nm node. The results presented here however shows that
this limit could be extended to 35 GHz, at least, for sub
65 nm nodes.
We also argue that the approach of using nonminimum dimensions in the transistor layout to increase
the breakdown voltage cannot be used higher up in
frequency. This due to that the fmax , ft of the nonminimum dimension transistors are too low.
III. C ONCLUSION
Through load pull measurements it is shown that
CMOS power amplifiers with acceptable gain and output
power can be made well up into the millimeter wave
region. It is also shown that, for these high frequencies,
the short gate length technologies surpass the longer gate
length technologies in terms of power added efficiency
while output power density stays fairly constant at 100
mW/mm. This is in contrast to lower frequencies, i.e.
below 5 GHz, there non-minimum gate length devices
with thicker gate oxide are used for power amplification.
R EFERENCES
[1] C.H. Doan, S. Emami, A.M. Niknejad, and R.W. Brodersen,
“Millimeter-Wave CMOS Design,” Solid-State Circuits, IEEE
Journal of, vol. 40, no. 1, pp. 144–155, 2005.
[2] M. Ferndahl, B. M. Motlagh, A. Masud, I. Angelov, H. O. Vickes,
and H. Zirath, “CMOS Devices and Circuits for Microwave and
Millimetre Wave Applications,” in European Gallium Arsenide
and Other Semiconductor Application Symposium, GAAS 2005,
Paris, 2005, p. 105.
[3] H. Shigematsu, T. Hirose, F. Brewer, and M. Rodwell,
“Millimeter-Wave CMOS Circuit Design,” Microwave Theory
and Techniques, IEEE Transactions on, vol. 53, no. 2, pp. 472–
477, 2005.
[4] A. V. Vasylyev, P. Weger, W. Bakalski, and W. Simbuerger, “17GHz 50-60 mW power amplifiers in 0.13μm standard CMOS,”
Microwave and Wireless Components Letters, IEEE, vol. 16, no.
1, pp. 37, 2006.
[5] T. Yao, M. Q. Gordon, K. K. W. Tang, K. H. K. Yau, M. T.
Yang, P. Schvan, and S. P. Voinigescu, “Algorithmic Design of
CMOS LNAs and PAs for 60-GHz Radio,” Solid-State Circuits,
IEEE Journal of, vol. 42, no. 5, pp. 1044–1057, 2007.
[6] M. Ferndahl, H. O. Vickes, H. Zirath, I. Angelov, F. Ingvarson,
and A. Litwin, “90-nm CMOS for microwave power applications,” IEEE Microwave and Wireless Components Letters, vol.
13, no. 12, pp. 523, 2003.
[7] Jorg Scholvin, David R. Greenberg, and Jesus A. del Alamo,
“Fundamental Power and Frequency Limits of Deeply-Scaled
CMOS for RF Power Applications,” in Electron Devices Meeting,
2006. IEDM ’06. International, 2006, pp. 1–4.
81
GHz Symposium 5-6 March 2008
Comparative analysis of the complexity/accuracy tradeoff for power amplifier behavior models
Ali Soltani Tehrani†, Haiying Cao, Thomas Eriksson†, Christian Fager
†Department of Signals and Systems
Department of Microtechnology and Nanoscience
Chalmers University of Technology, Göteborg, Sweden
asoltani@chalmers.se
Summary – A comparative study of state-of-art
nonlinear amplifier models are presented in this
document. The main focus is on the modeling
accuracy as a function of the computational
complexity.
and the ACEPR as
Introduction – Due to electrical consumption
requirements, power amplifiers are driven to high
efficiency regions, which have nonlinear
characteristics. Nonlinear dispersive effects also
exist, which are typically due to RF mismatching
and thermal effects. These impairments cause the
amplifier to distort the communication signal and
introduce spectral regrowth. Linearization of the
amplifiers is needed among many reasons in order
to avoid interfering with applications at neighboring
frequencies. Digital predistortion of the input data
to the amplifier can help us achieve this. In order to
utilize this technique, behavior models for power
amplifiers have to be established.
The models are compared in terms of NMSE and
ACEPR vs. the computational complexity. At this
stage, the number of parameters required to
construct the model is used as a measure of the
complexity. The Volterra was shown to reach low
NMSE values albeit at the cost of high complexity.
The memory polynomial model on the other hand
was shown to reach a limit in NMSE but at a fairly
low complexity. The models are arranged in the
figure below:
ACEPR =
∫ Meas( f ) − Model ( f )
2
df
out of band of the signal
∫ Meas( f )
2
df
inband of the signal
Model Definitions – In literature many behavior
models have been proposed to model the amplifier
characteristics. In this project some of the more
well-known models are characterized in terms of
their accuracy and computational complexity.
Some of these models include the Volterra model,
memory polynomial model, generalized memory
polynomial model and the Kautz Volterra model
[2]. The Volterra series is the most general model,
but due to high complexity is seldom used in
practice. The other models can be viewed as
reduced Volterra series models. The model
parameters are identified from data from a 3.5 GHz
WiMAX amplifier.
Conclusion – Depending on the amount of
complexity available for the transmitter, a model
can be chosen from the models characterized. The
behavior models are arranged according to their
performance in terms of complexity and accuracy.
Inverse modeling is also analyzed and characterized
for the different models.
Results – The modeling accuracy is defined by two
measures; the normalized mean square error
(NMSE) and the adjacent channel error power ratio
(ACEPR). The NMSE is defined as
⎛ var(Model − Meas ) ⎞
⎟⎟
NMSE = 10 log10 ⎜⎜
var(Meas )
⎝
⎠
[1] J. C. Pedro and S. A. Maas, "A comparative overview of
microwave and wireless power-amplifier behavioral
modeling approaches," Microwave Theory and
Techniques, IEEE Tran., vol. 53, pp. 1150-1163, 2005.
[2] M. Isaksson, D. Wisell, and D. Ronnow, "A comparative
analysis of behavioral models for RF power amplifiers,"
Microwave Theory and Techniques, IEEE Tran., vol. 54,
pp. 348-359, 2006.
82
GHz Symposium 5-6 March 2008
1
A Computational Load-Pull Investigation of Harmonic Loading effects on AM-PM conversion
1
O. Bengtsson, 2L. Vestling, and 2J. Olsson
1
University of Gävle, SE-801 76 Gävle, Sweden, phone: +46 26 648904, fax: +46 26 648828, e-mail:
bob@hig.se, 2Uppsala University, The Ångström Laboratory, Solid State Electronics, P.O. Box 534,
SE-751 21 Uppsala, Sweden
Summary— In this work computational harmonic load-pull have been used to study the effect of
harmonic loading on AM-PM conversion for an RF-Power LDMOS transistor. It is found that
especially the load impedance seen at the 2nd harmonic has a large impact (up to 2° or 15%
difference) on the phase distortion at P1dB in this investigation conducted at chip level.
I. INTRODUCTION
In computational load-pull transient simulations are
conducted for different load settings [1]-[2]. The effects of
harmonic loading can be investigated using a combination of
passive and active loads, Fig. 1, [3]. From the phase shift of
id versus vg at different power (voltage) levels the voltage
dependent phase shift (AM-PM conversion) can be
investigated by comparison to the ideal phase shift.
Fig. 1. Computational load-pull setup
II. RESULTS
The harmonic loads have been swept with the fundamental load in the primary optimum position for
maximum output power. Non optimum 2nd harmonic loading is shown to create up to 2° (15%)
increase in AM-PM conversion at P1dB, Figs . 2-5. 3rd harmonic has less impact.
Pout Class-AB
Pout Class-F
Pout Harm. Load.
Phase Class-AB
Phase Class-F
Phase Harm. Load.
Pout (dBm)
20
15
41
39
37
35
10
33
5
31
29
0
27
-5
25
-30
-25
-20
-15
-10
-5
0
Vin (dBV)
2.
3.
4.
5.
Fig. 2-5. Fundamental, 2nd and 3rd harmonic load-pull at -5 dBV input voltage (about P1dB) and for swept input voltage.
Output power contours are shown in grey with maximum at black (o). Absolute phase is shown in colors with maximum
at black (+). Contours show levels of 10° for f0, 0.5° for 2f0 and 0.25° for 3f0.
III. CONCLUSION
An AM-PM investigation using computational harmonic load-pull has been conducted for an RFPower LDMOS transistor. A large impact of the 2nd harmonic loading on phase distortion at chip
level is observed.
IV. REFERENCES
[1]
Loechelt GH, Blakey PA. A Computational Load-Pull System for Evaluating RF and Microwave Power
Amplifier Technologies. IEEE MTT-S Digest 2000;465-8
[2]
Jonsson R, Wahab Q, Rudner S, Svensson C. Computational load pull simulations os SiC microwave power
transistors. Solid-State Electronics 2003;47;1921-6.
[3]
Bengtsson O, Vestling , Olsson J. A Computational Load-Pull Method with Harmonic Loading for HighEfficiency Investigations. Submitted to Solid-State Electronics
83
GHz Symposium 5-6 March 2008
Deviation from Ideal Phase
(deg.)
43
25
Identification of Distortions in a RF Measurement System
Haiying Cao ∗ , Ali Soltani, Christian Fager ∗ , Thomas Eriksson
∗
Department of Microtechnology and Nanoscience, Department of Signals and Systems
Chalmers University of Technology, Gothenburg, Sweden
{haiying, asoltani, christian.fager, thomase}@chalmers.se
Summary: Distortions in a RF measurement system
can adversely impact the characteristics of the
measured signal. In this paper, signal processing
techniques are used to identify the sources for
distortions. Promising measurement results are
shown.
Introduction: The measurement system setup in Fig.
1 is used for characterization of power amplifiers.
Since high measurement accuracy is needed, it is
necessary to compensate for the system impairments.
Models for the measurement system have therefore
been evaluated. In Fig. 1 a computer is used to
generate a baseband signal, a vector signal generator
(VSG, Agilent E4438C) acts as the modulator and
transmitter, and a digital storage oscilloscope (DSO,
Agilent Infiniium 54854A) acts as the receiver. The
described system has inherent linear and nonlinear
distortions. Therefore, a Wiener filter is proposed to
model the linear distortion, and the well-known
Parallel Hammerstein (PH) model is used to model
the nonlinear distortion. Here, we use two
measurement setups to indentify the distortion. First,
the computer is used to download the baseband
signal to the VSG, and the RF signal is recorded
from one channel in the DSO. Second, the computer
is used to download the baseband signal to the VSG,
and the DSO records the same RF signal in two
separate channels.
the baseband and the measured signals. In Fig. 2,
SNRs versus the number of measurements is shown.
The SNR with the linear and nonlinear models is
higher compared to the SNR with just linear model
or with no model. We note that they get saturated
after a certain number of measurements since they
are limited by the inherent nonlinearities of the VSG
and DSO. In order to isolate the distortions from the
VSG and DSO, we feed the same signal using two
channels of the DSO to remove the nonlinearity
effect of the VSG, and we are able to see the
nonlinearity effect of the DSO. In this case, we
define the SNR between the two channels
as: 10 log10 (var(ch1 ) / var(ch1 − ch2 )) , and we get
higher SNR. This implies that the performance of the
measurement system is limited by the nonlinearity of
the VSG, and further research on modeling this
nonlinearity will be investigated.
Conclusion: A Wiener filter and PH model are used
to model the linear and nonlinear distortions of the
measurement system respectively, and the SNR of
the system is improved. Two channel measurements
are used to isolate the nonlinearity effects of the
VSG and DSO.
Trigger
VSG
Computer
ch 1
ESG
Baseband
signal
DSO
RF signal
1010MHz
MHz
reference
signal
reference
signal
ch 2
Fig. 1 Measurement system setup
Result: Pre-processing the signal is necessary before
the identification, which includes delay elimination,
normalization and statistical averaging. Since a
measurement noise floor exists in the devices,
statistical averaging is used to lower the noise
variance which enhances the dynamic range [1, 2].
Signal to noise ratio (SNR) is used to evaluate the
agreement between the measurement system and its
model. It is defined here as the difference between
Fig. 2 SNR versus number of measurements
Reference:
[1] Fager, C.; Andersson, K.; “Improvement of Oscilloscope
Based RF Measurements by Statistical Averaging Techniques”
Proc. of IEEE MTT-S, June 2006 pp. 1460 – 1463
[2] D. Wisell; “A baseband time domain measurement system
for dynamic characterization of power amplifiers with high
dynamic range over large bandwidths” Proc. of IMTC’ 03, May
2003, vol. 2, pp. 1177-1180.
84
GHz Symposium 5-6 March 2008
Further enhancement of Load pull simulation technique to study
non linear effects of LDMOS in TCAD
A. Kashif a, T. Johansson b, C. Svensson c, T. Arnborg d and Q. Wahab a,e
a
Department of Physics, Chemistry and Biology (IFM), Linköping University, SE-581 83 Linköping, Sweden
b
Infineon Technologies Nordic AB, SE-164 81 Kista, Sweden
c
Department of Electrical Engineering(ISY), Linköping University, SE-581 83 Linköping, Sweden
d
Ericsson AB, SE-221 83 Lund, Sweden
e
Swedish Defense Research Agency (FOI), SE-581 11 Linköping, Sweden
E-mail: ahska@ifm.liu.se
High linearity and efficiency in power amplifiers (PAs) are always demanding. In PAs, the
RF transistor is major component and should have high performance. The accurate large
signal transistor’s model is important to attain high performance in terms of RF output power,
gain, efficiency and linearity at desired bandwidth [1]. Computational load pull simulation
technique is found a suitable solution to study the transistor’s performance in real circuit
context [2]. The technique has been further extended to study the intermodulation distortion (IMD) of
the intrinsic transistor exclusively [3].
The LDMOS transistor is cost effective technology and used in communication system due to
reasonable linearity at medium power density [4]. We studied LDMOS transistor’s structure
provided by Infineion Technologies. The structure was optimized by observing the effects of
interface charges at the RESURF region. The optimized structure showed lower on-resistance,
which in turn increases the drain current. This enhances the RF output power as well as
frequency of operation. This optimized transistor delivered 1.3 W/mm RF output power upto
4 GHz in class AB operation.
The optimum performance in PAs can be achieved by understanding the non-linear behaviour
at device level under large signal operation. We studied intermodulation distortion by two
tone test in TCAD. The carrier frequency (f1) was selected at 1 GHz with 200 MHz tone
spacing to reduce the computational resources. A complete harmonic current and voltage
signals were evaluated with 5 cycles of 1 GHz signal in time domain. At input, two ac voltage
signal sources (Carrier (f1) and 2nd tone (f2)) were applied in series while at the output these ac
voltage signals sources were used with 180° phase difference with respect to the input signal.
The time domain current and voltage signals were transformed into frequency domain by Fast
Fourier Transformation (FFT) using Matlab. The IMD3 were observed at 2f2 -f1 and 2f1-f2. In
this technique the higher order harmonics are shorted. To estimate the power of IMD3, we
calculated the power by current flowing into the real part of impedance of the carrier
frequency. By optimization of LDMOS structure, the value at the power level 10 dB back-off
from P1dB enhances from -17 to -36 dBc.
The TCAD load pull simulation technique is now able to predict the large signal parameters
such as RF output power, efficiency, gain, impedances and non-linear effects of the transistor
in real context for the designing of power amplifiers.
References
[1] Frederick H. Raab, Peter Asbeck, Steve Cripps, Peter B. Kenington, Zoya B. Popovic et al, “Power Amplifiers and Transmitters for RF
and Microwave”, IEEE Transactions on Microwave Theory and Techniques, vol. 50, no. 3, March 2002: pp. 814-26.
[2] R. Jonsson, Q. Wahab, S. Rudner, and C. Svensson, “Computational load pull simulations of SiC microwave power transistors,” in Solid
State Electronics, vol. 47, pp. 1921-1926, 2003.
[3] A. Kashif, C. Svensson, T. Johansson, S.Azam, T. Arnborg and Q. Wahab, “A non-Linear TCAD large signal method to enhance the
linearity of transistor” in International Semiconductor Device Research Symposium (ISDRS’07), 12-14 December 2007.
[4] A. Wood, W. Brakensick,.C. Dragon, and W. Burger, “120 Watt, 2 GHz, Si LDMOS RF power transistor for PCS base station
applications” in IEEE MTT-S ,vol. 2, 7-12 June 1998: pp. 707-710
85
GHz Symposium 5-6 March 2008
GaN device and MMIC development at Chalmers University of Technology
M. Fagerlind, M. Südow, K. Andersson, M. Thorsell, A. Malmros, P.Å. Nilsson, H. Zirath, and
N. Rorsman
Microwave Electronics Laboratory, Microtechnology and Nanoscience, Chalmers University of
Technology, SE-41296 Göteborg, Sweden. martin.fagerlind@chalmers.se
Introduction
At Chalmers we are working with the development of GaN based HEMT (High Electron Mobility
Transistor) devices and MMICs (Monolithic Microwave Integrated Circuits). GaN and its III-nitride
alloys are interesting in microwave power applications, where properties like high breakdown field
and high saturation velocity are required.
Device
In the device area we are working towards an optimization of device characteristics like
maximization of transconductance and power handling capabilities i.e. maximization of breakdown
fields and source to drain current densities. This is done in-house by process and device layout
optimization, as well as physical simulations. Feedback to external suppliers of GaN epitaxial
wafers is also part of the optimization process.
Typical DC device performance is, drain saturation current (at zero gate voltage) around 1 A/mm.
Maximum transconductance in the range of 250 mS/mm. Typical small signal measurements for
similar devices are extracted to a maximum of fT=40 GHz and a maximum of fmax=70 GHz.
MMIC
There is a strong consensus about the benefits of power amplifiers in GaN. These devices are
however, also well fit for receiver electronics. Characterization of receiver blocks for robustness
and low noise, with the ultimate goal of realizing transceiver modules monolithically integrated on
one chip is therefore one of our major areas of research.
As an example of this effort, a high linearity X-band mixer based on a single 4x100µm GaN
HEMT is presented, Fig1a. The mixer has an RF bandwidth of 7-16GHz, an IF bandwidth of 2GHz
and exhibits a conversion loss of <8dB across the entire RF band, Fig1b. This mixer has very good
intermodulation properties seen in an IIP3 of 30dBm at 10.2GHz.
Fig1a. Photograph of X-band mixer MMIC
Fig1b. Conversion loss vs. IF and RF.
Acknowledgement
This research has been carried out in the Microwave Wide Bandgap Technology project financed
by Swedish Governmental Agency of Innovation Systems (VINNOVA), Swedish Energy Agency
(STEM) Chalmers University of Technology, and Ericsson AB, Saab AB, Norstel AB, Infineon,
NXP, Furuno, and Norse Semiconductor Laboratories AB.
86
GHz Symposium 5-6 March 2008
Workshop
Measurement - Modeling
1300-1430 Thursday 6 March 2008
87
GHz Symposium 5-6 March 2008
Model-Based Pre-distortion for Signal Generators
Carolina Luque and Niclas Björsell
University of Gävle, ITB/Electronics, SE-801 76 Gävle, Sweden
e-mail: caalue@hig.se, Phone: +46739823405
Currently, devices such as power amplifiers (PA) and analog to digital converters (ADC)
targeted to achieve the demanding WiMax and 3G wireless communications applications
must have performances high enough to support the current wireless standards and ensure an
appropriate response of the whole system.
To fulfill the requirements on spurious free dynamic range (SFDR) and signal to noise and
distortion ratio (SINAD) in ADCs [1] and to have exceptional linearity and efficiency at high
output power for PAs [2], high measuring performance of these high-quality components are
required. One must ensure that the test setup has superior performance compared to the device
under test (DUT) [3].
Even state-of-the-art signal generators (SG) can have problem to generate spectrally pure
signals. Due to imperfections in the SG such as nonlinearities, the generated signal will
contain unwanted intermodulation products (IMD). This work approaches the problem by
modeling the SG and using model-based digital pre-distortion (DPD), to reduce the 3rd–order
IMD products. Modeling and DPD are applied for a three tones input signal; initially for a
polynomial memory-less model, and in the near future to a grey-box with memory that also
takes the internal architecture of the SG into account.
The results from the memory-less model shows that the accuracy of the model and the
effectiveness of the DPD to some extent depend on the amplitude of the input signal. We have
shown that DPD, for certainly range of frequencies and using polynomials of 9th order, can
reduce the 3rd order IMD product up to 15dB.
The extension of the memory-less model to also include dynamic behavior, may improve the
performance in the DPD, and decrease the distortion even more due to the memory effects in
the system.
References
[1]
D. Brandenburg, "High-Speed ADC Specs for Wireless Applications" Department of
technology, Signal Conditioning Group, Fairchild Semiconductor Article, 2007.
[2]
P.M. Asbeck, L.E. Larson, I.G. Galton, “Synergistic design of DSP and power
amplifiers for wireless communications " IEEE Transactions on Microwave Theory
and Techniques, Volume 49, Issue 11, Nov 2001.
[3]
N. Björsell, O. Andersen, P. Händel, "High dynamic range test-bed for characterization
of analog-to-digital converters up to 500 MSPS," 14th IMEKO Symposium on New
Technologies in Measurement and Instrumentation & 10th Workshop on ADC
modeling and testing, September 12-15, 2005, Gdynia, Poland , pp. 601-604.
.
88
GHz Symposium 5-6 March 2008
A COMPARISON OF ANTENNA DIVERSITY
CHARACTERIZATION METHODS USING
REVERBERATION CHAMBERS AND DRIVE
TESTS
D. Nyberg*, M. Franzén‡ and P.-S. Kildal†
*Chalmers University of Technology, Sweden, daniel.nyberg@chalmers.se
‡Bluetest AB, Sweden, magnus.franzen@bluetest.se
†Chalmers University of Technology, Sweden, per-simon.kildal@chalmers.se
The aim of this paper is to compare two ways of characterizing diversity antennas; by
using conventional drive tests and using the high performance reverberation chamber
available from Bluetest AB. This work is collaboration between Chalmers, Bluetest
AB and Rayspan Corporation and it is done within the Chase research center at
Chalmers.
Three different antennas are used throughout the comparison and they are of two
types. The first antenna is a circular array consisting of 6 conventional monopole
elements. Those are mounted on a circular metallic ground plane of 140 mm radius.
Furthermore 3 different sets of holes have been drilled in the ground plane allowing
three different configurations of the array. The spacing between the elements in the
different configurations is 0.24λ 0.14λ, 0.06λ at 900 MHz. This antenna is a new
version of a similar antenna that has been used extensively as an example antenna at
Chalmers when the measurement technique in the reverberation chamber was
developed [1]. The previous antenna was measured in the standard Bluetest chamber,
whereas the present measurements are done in the new high performance chamber.
The purpose of the present paper is to compare performance obtained in the standard
and high performance Bluetest reverberation chambers. We will also include results
of the planned drive test measurements of the antenna. We will not expect very good
agreement between the drive tests and the measurements in the reverberation
chamber. In that sense, the reverberation chamber is superior, since all chambers
creates the same statistical isotropic environment which also is repeatable, where as
for drive test different values will be obtained for different test locations and even
different individual measurements, thus making comparisons between different
antennas very difficult. We also hope to report the results of similar measurements of
antennas manufactured and designed by Rayspan Corporation. One of them is a 3element array working at 2.5 GHz. The other is a 4-port dual-band antenna at 2.5 &
5.2 GHz.
References:
[1] K. Rosengren & P.-S. Kildal, “Radiation efficiency, correlation, diversity gain
and capacity of a six-monopole antenna array for a MIMO system: theory,
simulation and measurement in reverberation chamber”, Proceedings IEE
Microwaves, Optics and Antennas, pp.7-16. vol. 152, no. 1, Feb 2005. See also
Erratum published in Proceedings IEE, Microw. Antennas Propag., Vol. 153,
No. 4, August 2006
89
GHz Symposium 5-6 March 2008
Measuring Relative Receiver Sensitivity of Wireless Terminals in One Minute in
a Reverberation Chamber
M. Andersson*, C. Orlenius, M. Franzén
*Bluetest AB, Götaverksgatan 1, SE-417 55 Gothenburg, Sweden,
mats.andersson@bluetest.se
Summary: The traditional way of measuring receiver sensitivity of wireless
terminals is the Total Isotropic Sensitivity (TIS) measured in an anechoic
chamber. This paper describes an alternative, up to 60 times faster, method to
optimize receiver sensitivity during the design process.
Introduction:The downlink speed of mobile broadband services is directly affected
by the receiver sensitivity of the terminal. The traditional way of evaluating receiver
sensitivity during design to optimize performance is to measure the Total Isotropic
Sensitivity (TIS) in an anechoic environment [1]. An alternative method is to measure
the Average Fading Sensitivity (AFS) in a reverberation chamber with continuous
Rayleigh fading [2].
Results: The AFS value can be found from fitting a line to a number of one minute
relative receiver sensitivity values where the average bit error rate corresponding to a
given output power from a base station simulator is measured, see fig. 1.
Conclusions: For many measurements, especially during antenna design, it is not the
absolute receiver sensitivity of the terminal which is of most importance but the
ability to quickly optimize the design, i.e. to be able to say if a new antenna or
antenna configuration is better than another. By using the same average transmit
power and to see if the average bit error rate has changed up or down it is possible in
only one minutes time to see if the relative receiver sensitivity has improved or not by
a design change. This offers a tremendous saving in time during development to find
the optimum receiver sensitivity configuration.
References:
[1]
“Test Plan for Mobile Station Over The Air Performance” CTIA Certification,
rev. 2.2, November 2006.
[2]
C. Orlenius, P-S. Kildal, and G. Poilasne, “Measurements of total isotropic
sensitivity and average fading sensitivity of CDMA phones in reverberation
chamber”, IEEE AP-S International Symposium, Washington DC, July 2005.
90
GHz Symposium 5-6 March 2008
Modeling of SiGe HBTs Operation in Extreme Temperature Environment
1
Modeling of SiGe HBT Operation in Extreme
Temperature Environment
P.Sakalas1,2, M.Ramonas2,4, M.Schroter1,5, A.Kittlaus3, H.Geissler3, C.Jungemann4, A.Shimukovitch2
1
Chair for Electron Devices and Integrated Circuits, Dresden University of Technology, 01062 Dresden, Germany,
2
Fluctuation Phenomena Laboratory, Semiconductoe Physics Institute, 01108 Vilnius, Lithuania,
3
SUSS MicroTec Test Systems, 01561 Sacka, Germany,
4
EIT4, Bundeswehr University 85577 Neubiberg, Germany
5University of California San Diego, ECE, La Jolla, CA 92093-0407, USA.
Abstract - DC, RF and noise characteristics of SiGe HBTs, featuring peak fT of 80 GHz, were measured and modeled in different ambient temperatures. Forward Gummel, output characteristics and S-parameters were measured in the range T0=4473K. Very good agreement of simulated DC data with experimental enabled analysis of temperature dependent RF performance and noise sources of investigated SiGe HBTs.
this effect is taken into account neither in [12] nor the compact model.
Cut-off frequency fT (cf. Fig.2) increases with a drop in T0 due to
reduced electron scattering and bandgap narrowing, while it degrades
with heating. Measured noise parameters are in Fig.3 compared to
GALENE as well as to HICUM. The respective HICUM modeling
and noise source analysis will be presented at the conference.
140
I. INTRODUCTION
120
Automotive radar and cryogenic amplifier [1][2] satellite communication [3] applications require transistors capable of operating in
extreme environments. Recently, it was shown that SiGe HBT BiCMOS technology can be successfully used for cryogenic environment
applications [4][5][6][7][8], including extremely low ambient temperatures (T0=4K) [9][10]. Though resulting in degradation of DC and
RF performance, SiGe HBTs - opposite to conventional Si BJTs - can
operate at high ambient temperatures, reaching T0=573K [11]. Due to
the demonstrated reasonable performance of SiGe HBTs over a wide
T0 range, compact modeling becomes increasingly important for
designing circuits operating at extreme conditions. This work presents
a comparison of TCAD simulation, compact modeling (using
HICUM) with measured data for temperature dependent DC, RF and
noise characteristics.
fT [GHz]
fT [GHz]
100
0.8
0.9
JJcC[mA/µm^2]
[mA/µm2]
0.001
0.01
JC [mA/µm2]
0.1
1
10
NFmin [dB]
III. REFERENCES
[1] D.Gupta et al., IEEE Trans.on Applied Superconductivity, V. 13, pp. 477483, 2003.
[2] M.R.Murti et al., IEEE TED, V. 48, No. 12, pp. 2579-2587, 2000.
[3] E.R.Soares et al., IEEE TED, V. 48, No. 7, pp. 1190-1198, 2000.
[4] J.D.Cressler: Silicon Heterostructure Handbook, Boca Raton,
Taylor&Francis, 2005.
[5] J.D.Cressler, Proc. of the IEEE, V. 93, Issue 9, pp. 1159-1582, 2005.
[6] B.Banerjee et al., IEEE BCTM, pp.171-173, 2003.
[7] S.Pruvost et al., IEEE EDL, V.26, No.2, pp.105-108, 2000.
[8] N.Zerounian et al., Electronic Letters, V.36, No.12, pp.1076-1077, 2000.
[9] Geissler et al., Proc. ARTFG Microwave Meas. Conf., pp.137-140, 2006.
[10] P.Chevalier et al., IEEE BCTM, pp. 26-29, 2007.
[11] T.C.Wei-Min et al., IEEE TED, V. 51, No. 11, pp. 1825-1832, 2005.
[12] B.Neinhus et al., VLSI Design 8, pp.387-391, 1998.
[13] M.Schroter, IEICE Transactions on Electronics, V. E88-C, No.6, pp.10981113, 2005.
1E-3
0.7
0.0001
Fig. 3. Measured NFmin (symbols) and HD simulation (lines) vs. JC.
1E-2
0.6
1e-005
JC [mA/µm2]
1E-1
1E-6
0.5
1e-006
Fig. 2. Measured fT versus JC for a wide temperature range
1
Galene
HICUM
40
Jc [mA/µm^2]
1E1
1E-5
60
0
1e-007
Noise parameters were measured with an automated tuner system
from Maury in the 1-26 GHz frequency range at lattice temperatures
of T0=293-423K. The temperatures of the chuck, probes and shield
were measured and controlled. Very good agreement of forward Gummel characteristics was obtained with the hydrodynamic (HD) device
simulator GALENE [12] and with the compact model HICUM [13]
(cf. Fig. 1). A fair agreement was obtained for the base current as well.
Since at T0<78K base current changes its origin from diffusion to tunneling [4] modeling was performed for a limited T0 range since
T=25C, 50C, 100C, 150C
80
20
II. EXPERIMENTAL AND RESULTS
1E-4
4K
75K
493K
373K
300K
1.0
VBE
Vbe[V]
[V]
Fig. 1. Measured (symbols) and HD (blue lines with crosses) and
HICUM (red solid lines) JC versus VBE.
91
GHz Symposium 5-6 March 2008
On-wafer network analyser uncertainty estimation
J. Stenarson∗ , K. Andersson † , C. Fager
∗
†
and K. Yhland
∗
SP Technical Research Institute of Sweden, email: jorgen.stenarson@sp.se
† Chalmers University of Technology
Residual directivity
Summary—This paper presents a simple method for
the estimation of the most important uncertainty components/contributions for on-wafer Vector Network Analyser
(VNA) measurements. The method is based on the estimation of residual errors, of a calibrated VNA, using transmission lines. Results are presented using two different
models for the calibration standards.
0.03
0.02
0.01
Residual match
0.00
I. I NTRODUCTION
In VNA calibration, the residual errors are traditionally estimated by terminated one-port transmission line measurements using the one-port ripple
technique [1]. I.e. the residual errors are related
to the characteristic impedance of the transmission
lines. This method is inconvenient for on-wafer
work because it requires many terminated transmission lines.
This paper applies the two-port ripple method [2,
3] to on-wafer uncertainty estimation. The residual
errors are obtained from transmission line and thru
measurements using a SOLT/SOLR calibrated network analyser. The method compares the system
impedance defined by the VNA calibration standards to that of the reference transmission lines
giving the residual errors directivity, tracking and
match.
0.10
0.05
Residual tracking
0.00
Table based model
Manufacturer model
0.2
0.1
0.0
0
10
20
30
Frequency [GHz]
40
50
Fig. 1. Residual errors for calibrations using the table based model
(blue) or the manufacturer model (red) of the same standards.
short and open. However, for the manufacturer
model, the residual match and tracking show a
substantial increase which indicates that the open
and/or short standards were not accurately modeled.
II. R ESULTS
The method was applied to 150 µm GroundSignal-Ground (GSG) coplanar wafer probes and
a standard alumina calibration substrate containing
SOLT/SOLR standards. For comparison, two sets
of data were used for the SOLT/SOLR kit, a manufacturer model and a table based model. The table
based model was measured with a TRL calibration.
The resulting residual errors are shown in Fig. 1.
Since the residual directivity is approximately equal
between using the table based model and the manufacturer model we can conclude that the load is
of good quality. This is because the manufacturer’s
model of the load is a perfect match.
The residual match when using the table based
model is similar to the residual directivity, which
together with the excellent residual tracking means
that the table based model accurately predicts the
III. C ONCLUSION
We have shown how to use the two-port ripple
method for estimating the residual errors of VNA
measurements in the context of on-wafer measurements. The results show that the residual errors can
be improved by modelling the calibration kit.
R EFERENCES
[1] EA, “Guidelines on the evaluation of vector network analysers
(VNA),” Tech. Rep. EA-10/12 (rev.00), European co-operation
for Accreditation, 2000.
[2] J. Stenarson and K. Yhland, “Residual error models for the
SOLT and SOLR VNA calibration algorithms,” in 69th ARFTG
Conference, (Honolulu Hawaii), 2007.
[3] J. Stenarson and K. Yhland, “A new assessment method for the
residual errors in SOLT and SOLR calibrated VNAs,” in 69th
ARFTG Conference, (Honolulu Hawaii), 2007.
92
GHz Symposium 5-6 March 2008
mm-wave device testing using wideband coplanar transitions
Enrique Villa, Beatriz Aja, Luisa de la Fuente, Eduardo Artal (villae@unican.es)
Departamento de Ingenieria de Comunicaciones. ETSIIT. Avenida de los Castros, s.n.
Universidad de Cantabria. Santander (Spain)
Summary
A wideband coplanar to microstrip transition without via holes has been used for
passive device testing at mm-wave frequencies (20 to 40 GHz). Testing circuits are built
on 0.254 mm thick Alumina substrate. Test method, with a coplanar probe station, has
been checked and validated with passive devices.
Introduction
At mm-wave frequencies coaxial test fixtures do not provide accurate characterisation
of passive devices assembled in microstrip lines. A more repeatable method is to
perform tests with coplanar probes through broadband coplanar to microstrip transitions.
A coplanar transition has been designed to avoid metallized via holes in the dielectric
substrate, using virtual grounds at the end of special shaped radial stubs [1], [2]. For
calibration purposes a TRL kit has been included in the test set. Next Figure shows
pictures of the TRL kit, a broadband pass filter at 30 GHz and two transitions to test
beam lead diodes.
Results
Coplanar to microstrip transition design has been checked through electromagnetic
simulations and S parameter tests. Band pass filter characterisation has been validated
by comparison with tests based on a commercial coplanar to microstrip transition
having via holes. Beam lead Schottky diodes and thin film resistors were tested and
characterised from 20 to 40 GHz. Results showed a good repeatability and consistency.
Conclusion
A test method for S-parameters measurements of mm-wave passive devices based on
broadband coplanar transitions has been developed. Transitions have a good
performance from 20 to 40 GHz, and can be easily implemented on Alumina substrate
without via holes.
References
[1] K. Schmidt von Behren et al.,”77 GHz Si-Schottky Diode Harmonic Mixer”,
Proceedings of 32nd European Microwave Conference, October 2002, pp. 1-4.
[2] G. Zheng et al., “Wideband Coplanar Waveguide RF Probe Pad to Microstrip
Transitions Without Via Holes”, Microwave and Wireless Component Letters, IEEE,
pp. 544-546, Vol. 12, Issue 13, December 2003.
This work has been supported by the Ministerio de Educación y Ciencia (Spain), grant references ESP2004-07067C03-02 and AYA2007-68058-C03-03
93
GHz Symposium 5-6 March 2008
High Efficiency using Optimized SOI-Substrates
Lars Vestling∗ , Olof Bengtsson† , Klas-Håkan Eklund∗ and Jörgen Olsson∗
∗ Uppsala
University, The Ångström Laboratory, Solid State Electronics,
Box 534, SE-751 21 Uppsala, Sweden. E-mail: lars.vestling@angstrom.uu.se
† University of Gävle, SE-801 76 Gävle, Sweden.
Abstract— The effect of the substrate resistivity on
the efficiency for high-frequency SOI-LDMOS transistors is studied using computational load-pull simulations. It is shown that very low resistivity and
high resistivity SOI-substrates both result in high
efficiency. It is also shown that a normally doped,
medium resistivity, substrate results in significantly
lower efficiency.
I. I NTRODUCTION
It is well known that high-resistivity (HR) SOIsubstrates can improve the performance of highfrequency devices [1]. On the other hand, it has
been shown that very low resistivity (LR) SOIsubstrates may reduce substrate losses for RFpower devices [2].
In this paper the efficiency of an SOI-LDMOS
transistor on three different substrate resistivities
is studied. This is done using computational
load-pull [3] in class AB at 1 GHz.
II. R ESULTS AND C ONCLUSION
The substrate losses may be represented by
the off-state small-signal output resistance,
ROU T =1/Re(Y22 ). Figure 1 shows the simulated
output resistance for the three substrates, indicating that the HR and LR substrates are expected
109
ROUT [Ωmm]
107
6
10
105
104
3
10
102
101 6
10
7
10
8
9
10
10
Frequency [Hz]
10
10
[1]
[2]
[3]
[4]
F. Gianesello, et al., IEEE SOI Conf., pp. 119–120, 2007.
J. Ankarcrona, et al., EuroSOI2006, pp. 69–70, 2006.
G.H. Loechelt, et al., IEEE MTT-S Dig., pp. 465–468, 2000.
J. Scholvin et al., IEDM Tech. Dig., pp. 363–366, 2003.
Efficiency [%], POUT [dBm]
10
R EFERENCES
70
HR, 1 kΩcm
MR, 1 Ωcm
LR, 10 mΩcm
8
to have lower losses than the MR substrate at
1 GHz.
Figure 2 shows the results from the computational load-pull simulations. It is observed
that all three devices have almost the same
output power characteristics. However, there is
a significant difference in the efficiency, where
the MR substrate results in 10% lower efficiency
than the LR and HR substrates. This is mainly
explained by the low output resistance for the
MR substrate, as was shown in Fig. 1.
The results show that devices on LR substrates would perform as good as devices on
HR substrates from an efficiency point of view.
However, when designing the transistors it is
more difficult to do this using an HR substrate
due to problems with depletion/accumulation in
the region under the buried oxide [4].
The conclusion is that a LR substrate is the
best choice for RF-power devices on SOI.
HR
60 MR
LR
50
40
30
10
Fig. 1.
The small-signal output resistance for the different
substrates (VD =50V, VG =0V, tSU B =150µm).
94
POUT
20
10
0
-25
11
Efficiency
-20
-15
-10
-5
VIN [dBV]
0
5
10
Fig. 2. The efficiency and output power from computational
load-pull simulations (class AB, VDD =28V, VGS =1.3V, 1 GHz).
GHz Symposium 5-6 March 2008
Spin Torque Oscillator Simulations and Circuit Designs
Yan Zhou*, Sandeep Srinivasan, Johan Persson, and Johan Åkerman
Materials Physics, Department of Microelectronics and Applied Physics,
Royal Institute of Technology, Electrum 229, 16440 Kista, Sweden
*E-mail: zhouyan@kth.se
With the increasing integration and compactness of radio devices, the demand for further
miniaturization of high-quality, highly tunable oscillators (over a large frequency range)
continues to grow. To address this demand, we focus on the development of a novel,
nanometer-sized, magneto-electronic device – the Spin Torque Oscillator (STO). The STO is
compatible with the back-end flow of a standard Si process, and suitable for integration into
high-frequency CMOS, SiGe, or ultrahigh-frequency InP, InGaP, and InGaAs. Because of its
nano-scale dimensions (~50-100nm x 80-300nm), integration can be done without taking up
large chip area. The small size, wide tuning range, low power consumption and compatibility
with CMOS process make the STO an ideal candidate to replace traditional oscillator designs.
The Giant Magnetoresistance [1,2] based STO (Fig. 1.a) consists of one Cu layer
sandwiched by two ferromagnetic layers. The electrons flowing through one of the layers get
spin-polarized [3,4] and exert a torque on the localized magnetization of the other, creating a
current-tunable oscillator (1-40 GHz) with a high quality factor (Q up to 18000) [5]. In this
work we present numerical simulations of the STO in Matlab, Fortran, and Verilog-A. The
magnetodynamics is found from the Landau-Lifshitz-Gilbert- Slonczewski (LLGS) equation,
dmˆ
dmˆ
ηhI
= −γmˆ × Hˆ eff + αmˆ ×
+γ
mˆ × (mˆ × Mˆ ), (1)
dt
dt
2μ 0 M s eV
where γ is the gyromagnetic ratio, α is the damping parameter, η is the polarization ratio, μ0 is
the magnetic vacuum permeability, Ms is the saturation magnetization, and V the volume of
the free layer. Ĥeff is the effective magnetic field, which includes an applied magnetic field
Ĥapp, the uniaxial magnetic anisotropy field Ĥk, and the demagnetization field Ĥd.
(a)
30
(b)
by Cadence
by Matlab
z
x
20
y
Static
In-plane
state
Out-of-plane
15
z
frequency (GHz)
25
10
x
5
2
4
y
6
8
Current (mA)
10
12
Fig. 1: (a) Schematic structure of the spintronics nano-oscillator (SNO) (b) Cadence and Matlab simulations of the STO precession
frequency vs. DC current drive. The insets show the precessional trajectory of the STO for two different oscillation modes: in-plane and outof-plane precession (inset). There is no perceptible difference in the end result.
Fig. 1(b) shows how the STO precession frequency fSTO changes with external drive signal Idc.
When running identical simulations in Cadence and in Matlab, we find virtually no difference
in the end result. Examples of STO based circuits simulated in Cadence will also be presented.
We gratefully acknowledge financial support from The Swedish Foundation for
Strategic Research (SSF), The Swedish Research Council (VR), The Göran Gustafsson
Foundation and The Knut and Alice Wallenberg Foundation.
1.
2.
3.
4.
5.
M. N. Baibich, J. M. Broto, A. Fert, F. N. Vandau, F. Petroff, P. Eitenne, G. Creuzet, A. Friederich, J.
Chazelas, Phys. Rev. Lett. Vol. 61, pp. 2472 (1988).
G. Binasch, P. Grünberg, F. Saurenbach, and W. Zinn, Phys. Rev. B Vol. 39, pp. 4828 (1989).
J. Slonczewski, J. Magn. Magn. Mater. Vol. 159, L1 (1996).
L. Berger, Phys. Rev. B, vol. 54, pp. 9353 (1996).
W. H. Rippard, M. R. Pufall, S. Kaka, S. E. Russek, T. J. Silva, Phys. Rev. Lett. Vol. 92, pp. 027201
(2004).
95
GHz Symposium 5-6 March 2008
Workshop
The GHz Entrepreneur
1300-1430 Thursday 6 March 2008
The workshop is about doing business from innovations and/or IP(R)
in RF/Microwave. Personal reflections are given by representatives
from three small enterprises, one global company and one venture
company. The workshop is concluded by a discussion.
96
GHz Symposium 5-6 March 2008
Mikael Reimers, CEO
Foodradar Systems AB
www.foodradar.com
97
GHz Symposium 5-6 March 2008
Tomas Ornstein, CEO
Ranatec Instrument AB
www.ranatec.se
98
GHz Symposium 5-6 March 2008
Johan Lassing, CEO
Qamcom Technology AB
www.qamcom.se
99
GHz Symposium 5-6 March 2008
Peter Olanders, Technology Strategist
Ericsson AB
www.ericsson.com
100
GHz Symposium 5-6 March 2008
Bengt Gustafsson, CEO
Microwave Technologies AB
101
GHz Symposium 5-6 March 2008
Session V
1500-1600 Thursday 6 March 2008
102
GHz Symposium 5-6 March 2008
Industrial Aspects of 100 Gbit/s Optical Communication
Bengt-Erik Olsson
Ericsson AB
Ericsson Research
431 84 Mölndal
bengt-erik.olsson@ericsson.com
10 Gigabit Ethernet (GbE) [1] is today the interface of choice for high-end servers and routers
and the standard specifies implementations for up to 40 km on a dedicated optical fiber. For
longer distance communication, various non-standard implementations of 10 GbE exist even
though the Ethernet traffic often is encapsulated in another communication format, e.g.
Optical Transport Network (OTN) [2] in order to facilitate traffic monitoring and quality
assurance as well as coexistence with other traffic. There is now a need to aggregate traffic
from multiple 10 GbE sources and therefore there are requests for even higher-speed
communication channels and thus a next generation Ethernet is considered. Since Ethernet
traditionally has taken speed increases in steps of x10, a natural next step is to consider 100
Gb/s as the next speed for Ethernet traffic. However, even in an optical fiber there are
physical limitations to what bit-rate a single wavelength channel easily can accommodate and
a serial bit rate of 100 Gb/s might be very difficult to transmit for more than a few hundred
meters due to chromatic dispersion in the fiber. Recently, the IEEE [3] has launched a
standardization effort for 100 Gb/s Ethernet for short and moderate transmission distances
over dedicated optical fibers, where multiple low bit-rate optical wavelength channels in the
fiber are considered. Proposed solutions include 10 channels carrying 10 Gbit/s each or 4
channels carrying 25 Gbit/s each. These solutions target primarily short distances e.g. up to 10
or 40 km on a dedicated fiber and often new fibers can be added at these distances to increase
the total capacity. For longer distance communication, adding new fibers are usually not
possible and these systems are already utilizing dense wavelength division multiplexing
(DWDM) to increase the total fiber capacity. A DWDM system utilizes standardized
wavelength windows for multichannel communication allowing e.g. 160 optical channels
through the fiber. The use of a standardized frequency grid for optical communication
imposes new constrains on the optical signal at ultra high bit-rates, e.g. 100 Gbit/s, since the
bandwidth of the optical channel will severely limit the allowed bandwidth of the transmitted
signal. Traditionally the available bandwidth in an optical channel has been much greater than
the signal bandwidth, but transmitting 100 Gbit/s data using simple on-off keying will require
more bandwidth than available on most DWDM systems. Typically commercial DWDM
systems operate at either 100 GHz or 50 GHz optical channel spacing and the 3 dB bandwidth
available for the data channel is often limited to 55 GHz on a 100 GHz grid and of course
even lower on a 50 Ghz grid. Therefore more bandwidth efficient signaling formats must be
considered for 100 Gbit/s communication over DWDM channels. In this presentation the
issues of standardizing 100 Gbit/s Ethernet will be discussed as well as possible
implementations of bandwidth efficient 100 Gbit/s communication for use on DWDM
channels using novel optical modulation formats.
[1] http://grouper.ieee.org/groups/802/3/ae
[2] ITU-T G.709, "Interfaces for the Optical Transport Network", March 2003.
[3] http://grouper.ieee.org/groups/802/3/hssg/index.html
103
GHz Symposium 5-6 March 2008
All-Optical Waveform Sampling with Terahertz Capacity
Mathias Westlund, Peter A. Andrekson and Henrik Sunnerud
Chalmers University of Technology, MC2, Photonics Laboratory
Authors are also with PicoSolve Inc. (www.picosolve.com)
We present a fiber based all-optical sampling (AOS) capable of visualizing optical signals
beyond 500 Gb/s. AOS provides bandwidth and precision for next generation optical
communication.
The main driver behind the development of AOS techniques has been the exceptionally good
temporal resolution these techniques provide (<< 1 ps). Since the electronic means of
measuring optical waveforms are limited in bandwidth by the use of photodiodes and
electronic sampling gates to <70 GHz, the attraction to optical sampling systems providing in
principle THz bandwidth is natural.
At the GigaHertz symposium we would like to focus on the following aspects of AOS:
o Principle of operation and performance of a fiber based AOS
o How software algorithms can simplify hardware design and improve performance
o Advantages compared to conventional electronic solutions
o Alternative implementations of AOS
o Experimental implementation and results
The AOS (see Fig. 1) that has been developed at Chalmers [1-3] utilize an optical pulse
source that generates short intense sampling pulses together with a nonlinear phenomenon in
the optical fiber to create an optical gate which is open only when the sampling pulse is
present. The nonlinear phenomenon is called four-wave mixing and is an extremely fast
process which enables sub-picosecond gating. At the output of the gate samples of the signal
appear at a new wavelength, well separated from the sampling pulses and signal, and can be
extracted using an optical filter. At this point the samples can be detected using low
bandwidth electronics and fed into a computer for time-base processing and visualization.
Fig. 1: Principle of optical sampling
Figure 2 shows an example of the optical sampling capacity by visualizing a 640 Gb/s data
signal as a wide open eye-diagram. This AOS (which was provided by PicoSolve) has a
temporal resolution of 1 ps, 40 nm optical bandwidth (full C-band), 2 mW signal sensitivity
and is polarization independent, which defines the current state-of-the-art.
Fig 2. Optically sampled eye-diagram of a 640 Gb/s data signal.
1.
2.
3.
M. Westlund et al., “High performance optical-fiber-nonlinearity-based optical waveform monitoring,” JLT.,23, pp. 2012-2022, 2005
M. Westlund et al. “Software-synchronized all-optical sampling for fiber communication systems,” JLT,23, pp. 1088-1099, 2005.
M. Westlund et al., “Simple scheme for polarization-independent all-optical sampling,” PTL, 16, pp. 2108-2110, 2004.
104
GHz Symposium 5-6 March 2008
High Speed 1.3 µm VCSELs for FTTH and RoF
P. Westbergh, E. Söderberg, J.S. Gustavsson, P. Modh and A. Larsson
Photonics Laboratory, Department of Microtechnology and Nanoscience – MC2
Chalmers University of Technology, SE-412 96 Göteborg, Sweden
(contact e-mail: petter.westbergh@chalmers.se)
Z.Z. Zhang, J. Berggren and M. Hammar
Department of Microelectronics and Applied Physics
Royal Institute of Technology, SE-164 40 Kista, Sweden
With the cost/performance ratio being a critical factor for access networks such as fiber-to-thehome (FTTH) and radio-over-fiber (RoF) systems, vertical cavity surface emitting lasers (VCSELs)
emitting at wavelengths compatible with single mode fibers are of interest. The VCSEL offers a
combination of cost-efficient batch-level fabrication and testing/screening together with good beam
quality, good high frequency modulation response at low currents and good power efficiency.
We have developed GaAs-based single mode VCSELs emitting at 1.3 µm using highly strained
InGaAs quantum wells and a large detuning between the gain peak and the cavity resonance. Oxide
confinement is used for current and optical confinement and a surface relief technique is used for
selecting the fundamental mode and suppressing oxide modes (Fig.1).
Under large signal digital modulation, clear open eyes and error free transmission over 9 km of
standard single mode fiber were demonstrated at OC-48 (2.488 Gbit/s) and 10 GbE (10.31 Gbit/s)
bit rates up to 85°C (Fig.2), which proves the applicability for FTTH links.
Under large signal RF modulation, a spurious-free dynamic range (SFDR) of 100 and 95 dB⋅Hz2/3
was obtained at 2 and 5 GHz, respectively (Fig.3), which is in the range of those required for RoF
links in several systems for mobile communication and wireless access.
E. Söderberg, J.S. Gustavsson, P. Modh, A. Larsson, Z.Z. Zhang, J. Berggren and M. Hammar, “High temperature
dynamics, high speed modulation and transmission experiments using 1.3 µm InGaAs single mode VCSELs”, IEEE J.
Lightwave Techn. 25, 2791, 2007.
surface relief
top contact
oxide aperture
BCB
BCB
Fig. 1 Left: VCSEL design with surface relief for single mode emission. Right: Cross-sectional view showing the oxide
aperture used for transverse current and optical confinement and a BCB layer used to reduce the capacitance.
OC-48
10 GbE
25°C
85°C
Fig. 2 Eye diagrams recorded under OC-48 and 10 GbE
modulation at 25 and 85°C.
Fig. 3 SFDR at 2 GHz as a function of bias current for
VCSELs with two different surface relief diameters.
105
GHz Symposium 5-6 March 2008
106
GHz Symposium 5-6 March 2008
107
GHz Symposium 5-6 March 2008
108
GHz Symposium 5-6 March 2008
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising