Project status

1
Project status
BWRC Winter Retreat 2005
Bob Brodersen Dept. of EECS Univ. of
Calif. Berkeley
2
Basic Research Theme Designing Radios to
Improve Spectrum Utilization

• Exploit new frequencies 60 GHz
• Avoid interfering with primary users
• UWB use low transmit power
• Cognitive dynamically find spectra that isnt
  being used
• Infrastructure to support this research
• BEE2
• Multi-antenna, multi-radio front end

3
Basic Research Theme Designing Radios to
Improve Spectrum Utilization

• Exploit new frequencies 60 GHz
• Avoid interfering with primary users
• UWB use low transmit power
• Cognitive dynamically find spectra that isnt
  being used
• Infrastructure to support this research
• BEE2
• Multi-antenna, multi-radio front end

4
60 GHz CMOS

• Ali Niknejad will report on this effort tomorrow
• Key results
• Best Paper award at 2004 ISSCC (Technology
  directions)
• Measurements to 85 GHz shows models still hold
• Noise measurements
• Phase shifter chip in fabrication

5
Performance to 85 GHz
11-dB Gain _at_ 60 GHz

• We have developed a design methodology that gives
  repeatable results for microwave CMOS design

6
60-GHz Amplifier Noise Figure (Talk by Chinh
Doan tomorrow on noise modeling)
7
Transceiver Integration (Sohrab Emami)

• First pass earlier this year, 7 dB conversion
  gain
• Next generation in design

8
RF Phase Shifter Architecture (Sayf Alalusi -
talk this afternoon)

• 1 data stream, RF phase shifters only, digitally
  controlled
• Achieves high antenna gain in an arbitrary
  direction
• Chip in final stage of fabrication
• T-line package characterization underway

r(t)
s(t)
a0
a0
a1
a1
S
a2
a2
9
Basic Research Theme Designing Radios to
Improve Spectrum Utilization

• Exploit new frequencies 60 GHz
• Avoid interfering with primary users
• Cognitive dynamically find spectra that isnt
  being used
• UWB use low transmit power
• Infrastructure to support this research
• BEE2
• Multi-antenna, multi-radio front end

10
Cognitive Radios (Talk this afternoon- Danijela
Cabric)

• Paradigm shift from a passive receiver to a radio
  that can sense and learn, then find and adapt to
  the environment.
• FCC is releasing spectra in the 400-800 Mhz band
  for this work
• We had a workshop in October on Cognitive Radios
  (40 people attended)
• White paper completed on protocol design (CORVUS)
• Focus on testbed for real system simulation

11
Active Interference Cancellation (Jing Yang)

• Concept of Interference Cancellation System
• Predicting the future interference
• Taking full advantage of AGC A/D
• to amplify and quantize the signal
• Assumption
• Loop fast enough
• Interference is much larger than the signal we
  are trying to send
• INR is high for all channels
• Open Problems
• Most effective adaptive algorithm
• Signal Protection (Notching) and constraints for
  Transmitter
• Multiuser Detection way -- Frequency
  domain filtering
• Subtraction done at RF or Baseband
• Distortion and power issues

12
Basic Research Theme Designing Radios to
Improve Spectrum Utilization

• Exploit new frequencies 60 GHz
• Avoid interfering with primary users
• Cognitive dynamically find spectra that isnt
  being used
• UWB use low transmit power
• Infrastructure to support this research
• BEE2
• Multi-antenna, multi-radio front end

13
Berkeley Impulse Transceiver (Ian ODonnell)
Targeting Sensor Network Application

• Specifications
• 100kbps over 10m with 10-3 BER
• 1mW total (TXRX) power consumption
• 0-1GHz bandwidth

ADC
GAIN
CLK
DIGITAL
TX

• All-CMOS integrated UWB transceiver
• for comm. and ranging/locationing
• Aggressive low-power design
• 10x lower power than comparable performing
  narrowband radios
• Provide flexible platform for further research

14
UWB Transceiver Prototype
Based on Digital Sampling/Acquisition
Oscilloscopes
TIA
GAIN
ADC
Analog Front-End
ADC
Digital Interface
Control Logic and Interface
BIAS
OSC
ADC
Pulser (by Stanley Wang)
CLKGEN
DLL
PULSE
15
UWB Transceiver (0-1 GHz band) in fabrication
16
3-10GHz UWB CMOS LNA (Stanley Wang)

• Common-gate input stage provides wideband
  matching
• Matching network also provides current gain due
  to the impedance mismatch
• For low-voltage operation, output loading must be
  DC short, but on-chip RF chokes (e.g. Lload) have
  a low self-resonant frequency, which leads to
  high-frequency gain roll-off

? Internal gain peaking to boost up the gain at
high frequency
17
UWB LNA (3-6 GHz) Layout being submitted Thursday
Area 1.5mm x 1.1mm

18
Multiple Antenna Algorithm Implementation
Tradeoffs (Dejan Markovic)

• Goal investigate energy/delay-efficient
  implementation of multi-dimensional algorithms
• Implemented an SVD-based multiple-antenna channel
  decoupling algorithm on Simulink and BEE.
• Driver for Energy/Delay architecture optimization
  by using time-multiplexing and interleaving.
• A driver for large scale Simulink design

19
The algorithm.
20
Optimized implementation
12,8
10,8
14,9
12,11
12,8
12,9
12,8
8,5
12,9
10,8
21
Component 1/sqrt() Block (Debugging interaction
with Intel foundry)
(10k FPGA slices equivalent to 1mm2 in 130nm)
22
Basic Research Theme Designing Radios to
Improve Spectrum Utilization

• Exploit new frequencies 60 GHz
• Avoid interfering with primary users
• Cognitive dynamically find spectra that isnt
  being used
• UWB use low transmit power
• Infrastructure to support this research
• BEE2
• Multi-antenna, multi-radio front end

23
BEE2 Lives! (Chen Chang will talk this afternoon
and demo in poster session)
14X17 inch 22 layer PCB

• 500 Gigaops/sec 16 multiply and adds
• 20 Gigabytes of memory
• 160 Gigabits/sec full duplex off board bandwidth

24
A major issue as we scale up our designs to BEE2
size

• Verifying the functionality of a design is
  typically done via simulation
• Effectively software emulation of a large,
  real-time hardware system
• Orders of magnitude slower, with huge memory
  requirements

25
Debugging designs on the Hardware

• Inspecting the running hardware system itself can
  be a monumental effort
• Requires the critical data to be exposed off-chip
• Requires an adequate number of test pins to probe
• Requires complex test equipment
• Requires manual interpretation of signals

Photos F. Burghardt
26
FPGA Debugging Support (Kevin Camera)

• User tags signals of interest with debugging
  testpoints
• Defines a variable name
• Defines other parameters of interest for data
  observation

27
Debugging Runtime

• User can monitor variables and control process
  execution from remote client
• Inserted logic interfaces to PowerPC core on FPGA
• Embedded software implements thin protocol for
  communicating with client

28
Analog Front-ends for BEE2

• Radio Telescopes (ADC only)
• SETI 300 MHz bands
• ATA32 400MHz bands
• CARMA 500MHz bands
• VLBI 1024MHz bands
• TV-band Cognitive radio (ADC/DAC)
• 0.51GHz bandwidth
• Ultra-Wide Band radio (ADC/Pulse driver)
• 1GHz bandwidth
• Direct digital synthesis of cable TV signals
• 800MHz bandwidth

29
Basic Research Theme Designing Radios to
Improve Spectrum Utilization

• Exploit new frequencies 60 GHz
• Avoid interfering with primary users
• Cognitive dynamically find spectra that isnt
  being used
• UWB use low transmit power
• Infrastructure to support this research
• BEE2
• Multi-antenna, multi-radio front end

30
Multi-antenna Testbed

• BEE (Berkeley Emulation Engine)
• 600 GOP/s FPGA array previously developed at the
  BWRC
• Used for digital baseband signal processing,
  analysis, and combining
• 32 Radio Front-ends
• D/A 14 bit and A/D 12-bit resolution
• Up to 64 MS/s sampling rate
• Common LVDS clock synchronous to BEE clock
• 2.4 GHz unlicensed band
• 20 MHz channel bandwidth
• Common local oscillator synthesizer
• Linear Power Amplifiers with output power up to
  10 dBm
• RF Front-end ?? BEE interface
• 1GHz fiber optical connection per front-end,
  range 1/3 mile

Baseband
RF analog
31
Testbed System Components (Bill Baringer demo
in poster session)
Power, LO and CLK generation and distribution
BEE / BEE2
Fiber connection
RF Modem
32
Summary

• Moving to next stage in 60 GHz project
  transceiver design
• Cognitive radio system design completed
• UWB transceiver front-end in fabrication, LNA and
  A/D in layout
• Multiple antenna, multiple radio analog test-bed
  operational
• BEE2 Modules fabricated and operational
• Multi-antenna Testbed near completion