Berkeley Wireless Research Center Why Theorists and Implementers Should Work Together Bob Brodersen...

Berkeley Wireless Research Center

Why Theorists and Implementers Should Work

TogetherBob Bob

BrodersenBrodersenDept. of Dept. of

EECSEECSUniv. of Univ. of

Calif.Calif.Berkeley Berkeley

http://bwrc.eecs.berkeley.edu


In particular 17 GHz of New Unlicensed Bandwidth…

The UWB bands have some use restrictions, but FCC requirements will allow a wide variety of new applications

The 57-64 GHz band can transmit up to .5 Watt with little else constrained

Cognitive Radio allocations in the regulatory process (400-800MHz band to start with) – maybe also in 3-10 like UWB?

How can we use these new resources?

0 10 20

UWB/CR?UWB/CR

UWBMm

WaveBand

30 40 50 60 GHzComm Vehicular CommID


New radio technologies- UWB, 60 GHz and Cognitive

Carrier Frequency (GHz)

Pea

k D

ata

Rat

e (b

ps)

1 G

100 M

10 M

1 M

100 k

10 k

HDTV motion picture,Pt.-to-Pt. links

NTSC video;rapid file transfer

MPEG video;PC file transfer

Voice,Data

Cellular

3G

802.11b

60 GHzPt.-to-Pt.

60 GHzWLAN

Bluetooth

0.1 1 10 100

802.11a

ZigBee

UWB

UWB

CognitiveRadios


New Challenges

Interference channel is the one of interest» How do we model this channel» What is its capacity» How do we best use this channel

Non-sinusoidal radios» How to analyze and design with impulses

Microwave Radios» The path to Gbit/sec links» Requires optimal antenna systems

Cognitive Radios» How do we sense signals» How do we design radios with large in-band interferers


Lets start with UWB…

According to the FCC:

“Ultrawideband radio systems typically employ pulse modulation where extremely narrow (short) bursts of RF energy are modulated and emitted to convey information. … the emission bandwidths … often exceed one gigahertz. In some cases “impulse” transmitters are employed where the pulses do not modulate a carrier.”

-- Federal Communications Commission, ET Docket 98-153, First Report and Order, Feb. 2002


Two basically different signaling approaches

Sinusoidal, Narrowband

Frequency

Time

Time

Frequency

Impulse, Ultra-Wideband


First Major Application Area

High Speed, Inexpensive Short Range Communications (3.1-10.6 GHz)» FCC limit of -41dBm/Mhz at 10 feet severely

limits range – Power level roughly 1 mW– For short range communications this may be OK –

e.g. line of sight from 10 feet for connecting a camcorder to a set-top box, “wireless Firewire”

» Advantage is that it should be less expensive and lower power than a WLAN solution (since 802.11a > 100 Mbits/sec for short range)


Status of High Rate, Short Range UWB

Major standards battle in IEEE 802.15.3

Two competing approaches» Frequency hopping OFDM

– Exploits the wide bandwidth to provide higher rates with lower precision hardware (e.g. reduced A/D accuracy, linearity requirements)

– Uses a well understood technique (802.11a/g)

» Impulse radios – New approach, so somewhat unknown ultimate

performance and efficiency


OFDM or Impulse?

OFDM strategy makes sense from a theoretical standpoint (deals with multipath)

But what about the implementation ??


Lets compare to an 802.11a chip

What can we eliminate?

ADC/DACViterbi

Decoder

MAC Core

Time/FreqSynch

FFTDMA

PCI

AGCFSM


How about using Impulses?

Basically pulsed rate data transmission – sort of optical fiber without the fiber…

Key design problem, as in wireline transmission, is synchronization

0 2 4 6 8 10 12

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Time (nS)

Magnitude (

V)

“1” “0”Biphase signalling


Front-end is very simple

Mostly Digital Radio Architecture: - Wideband antenna - Wideband amplifier / matching network - RF bandpass filtering (low Q filter) - High bandwidth sample and track - High-speed and low resolution ADC - Sampling Clock generator - DSP

UWB attenna

BpLNA

ADC

DigitalBackend


Receive match filter

Basic approach is to create a match filter for the above received pulse shape

This collects all the energy associated with the waveform

PN0 PN1

Nripple<= 64 ns

Trep10ns ~ 100ns


Sampling Offset Effects

Unfortunately a small timing offset results in a very different waveform so the match filter output is very dependent on the timing

150 160 170 180 190 200

-1

0

1x 10

-4 0Ts

150 160 170 180 190 200

-1

0

1x 10

-4 0.125Ts

150 160 170 180 190 200

-1

0

1x 10

-4 0.25Ts

150 160 170 180 190 200

-1

0

1x 10

-4 0.375Ts

150 160 170 180 190 200

-1

0

1x 10

-4 0.5Ts

150 160 170 180 190 200

-1

0

1x 10

-4 0.625Ts

150 160 170 180 190 200

-1

0

1x 10

-4 0.75Ts

150 160 170 180 190 200

-1

0

1x 10

-4 0.875Ts


A solution to this… (Mike Chen)

Convert the single baseband pulse into an analytic signal (real and imaginary parts) via a Hilbert transformation.

The analogy is the use of the I and Q channel for sinusoidal systems

ADC

AnalyticMF

Hilbert

ShapeEst

Ave DetPulse

in (real)

Imag

Real


UWB impulse signal processing

Research into the signal processing for impulse detection is just beginning – so lots of opportunities

Analytic impulse signal processing also achieves

a timing resolution below the sampling period, what can this be used for?


Second Major Application Area

Low Data Rate, Short Range Communications with Locationing (< 960 MHz)» Round trip time for pulse provides range

information – multiple range estimates provides location

» Used for asset tracking – a sophisticated RFID tag that provides location

» Can be used to track people (children, firemen in buildings)

» Sensor networks


Location Determination Using UWB

UWB provides» Indoor measurements

» Relative location

» Insensitivity to multipath

» Material penetration (0-1 GHz band)

Time of flight

Transmit short discrete pulses instead of Transmit short discrete pulses instead of modulating code onto carrier signalmodulating code onto carrier signal– Pulses last ~1-2 nsPulses last ~1-2 ns

– Resolution of inches Resolution of inches


0 100 200 300 400 500 600 700 800 900-0.1

-0.05

0

0.05

0.1

0 100 200 300 400 500 600 700 800 900-10

-5

0

5

10

Signal processing for ranging

The problem is to determine the leading edge of the response

Simple averaging…

“Clean” algorithm (iterative best fit of time delayed waveforms)

0 10 20 30 40 50 60 70 80-0.05

0

0.05

0.1

0.15

0 10 20 30 40 50 60 70 800

0.005

0.01

0.015


Many new questions…

What are the limits on locationing accuracy and what are the dependences

What algorithms can be used to achieve these limits

How do we coordinate networks of devices and what advantages can we obtain


Next lets look at the 60 GHz band…

0 10 20

UWB/CRUWBCR

UWBMm

WaveBand

30 40 50 60 GHzComm Vehicular CommID

Microwave communications


Why is operation at 60 GHz interesting?

Lots of Bandwidth!!!» 7 GHz of unlicensed bandwidth in the U.S. and Japan » Europe CEPT “there is an urgent need to identify and

harmonize civil requirements in the frequency range 54–66GHz.”

57 dBm

40 dBm


Why isn’t 60 GHz in widespread use?

The technology to process signals at 60 GHz is expensive

Misconceptions about path loss and propagation at 60 GHz


VGS = 0.65 V

VDS = 1.2 V

IDS = 30 mA

W/L = 100x1u/0.13u

CMOS can do it - 130-nm CMOS has a gain of 7dB at 60 GHz


40-GHz and 60-GHz CMOS Amplifiers

Design and modeling can be incredibly accurate…. Power consumption: 36 mW (40 GHz), 54 mW (60 GHz)

18-dB Gain@ 40 GHz

11.5-dB Gain@ 60 GHz


A Leap Forward for CMOS

• CMOS offers two orders of magnitude cost reduction while providing higher integration and reliability

• Each new process generation moves it 20-40% higher

X Where we are nowwith 130 nm


Why isn’t 60 GHz in widespread use?

The technology to process signals at 60 GHz is expensive

Misconceptions about path loss and propagation at 60 GHz


Typical path loss (Friis) formula is a function of antenna gain Gr and Gt:

But maximum antenna gain increases with frequency for the same antenna area, A

Path loss of line-of-sight transmission – is that a major problem?

22

4 r

GG

P

P tr

t

r

2

4

A

G


Using the same effective area then…

There is an improvement by the frequency squared It is better to be at higher frequencies!

22

1

r

AA

P

P tr

t

r


Material penetration Not that much worse for most materials

» Compensated for by larger receive antenna (> 5 cm2) » Exploit wider available bandwidths as need through stronger

coding (trade data rate for range)

Frequency (GHz)3 5 8 10 20 30 50 80 100 200

35

30

25

20

15

10

5

0

Concrete Block Painted 2X6 Board

Clay Brick

3/4" Plywood

3/4" Pine Board

Wet Paper Towel

GlassDrywall

Asphalt Shingle

To

tal O

ne W

ay A

tten

uati

on

(d

B)

Kevlar SheetPolyethylenePaper Towel (Dry)Fiberglass Insul.

(from Bob Scholtz)


The future will need GBit/sec wireless links

What technology will get us there? Lets compare 3 systems links

» UWB – (OFDM)» 802.11n – (MIMO)» 60 GHz


802.11n (TgnSync an Wwise proposals)

Configuration Rate ½, 16-QAM

Rate ¾, 16-QAM

Rate 2/3, 64-QAM

Rate ¾, 64-QAM

Rate 5/6, 64-QAM

1 Tx, 40 MHz 54 81 108 121.5 135

2 Tx, 40 MHz 108 162 216 243 270

3 Tx, 40 MHz 162 243 324 364.5 405

4 Tx, 40 MHz 216 364 432 486 540

TgnSync

Wwise


Comparison Lets take an antenna that has an effective

area of 5 cm2

Frequency AntennaGain

UWB (7 band mode) 5.092 GHz 2.4 dB

802.11n (highest range)

5 GHz 2.4 dB

60 GHz 60 GHz 24 dB

We can have 24 dB of gain on both antennas – or even more on the receive side

We can have 24 dB of gain on both antennas – or even more on the receive side


Power limitations

In all three cases near limit of FCC regulation (much higher in Japan)

An advantage of from 17 to 43 dB

How about bandwidth?

Power Ant. Gain Tx power

UWB -6.6 dBm 2.4 dB -4.2 dBm

802.11n 20 dBm 2.4 dB 22.4 dBm

60 GHz 15 dBm 24 dB 39 dBm


Bandwidth efficiency requirement for a Gbit/sec

The efficiencies required to meet the Gbit/sec goals are unrealistic for any approach but at 60 GHz

Bandwidth Efficiency needed

Actual Maximum

Design Goal

UWB 480 MHz 2 Bits/Hz 1 Bit/Hz (480 Mbits/sec)

802.11n 40 MHz 25 Bits/Hz 17 Bits/Hz (680 Mbits/sec)

60 GHz 2 GHz .5 Bits/Hz .5 Bit/Hz (1 Gbits/sec)


We now need adaptive beamforming algorithms

Very closely related to MIMO algorithms – similar problem, need to spatially localize transmission

Need to rapidly adapt to varying conditions

)(ty

ArrayProcessing

)(tx

ArrayProcessing

1st path, 1 = 1

2nd path, 2 = 0.6


Multiple antenna beamforming hardware is straight-forward

Wavelength is 5mm, so in a few square inches a large antenna array can be implemented

The challenge is how to determine the coefficients of this beamformer to maximize range while minimizing interference

a1

b1

a0

b0

a2

b2

PA

PA

PASingle Channel

Transceiver


The open questions… How best to implement a flexible, adaptive

antenna system What is the best way to use 7 GHz of

bandwidth to implement a high datarate link?» Extremely inefficient modulation but at a very

high rate? (say 2 GHz of bandwidth for 1 Gigabit/sec) – requires analog processing

» Or use an efficient modulation, so lower bandwidth. e.g. OFDM – but needs digital processing and a fast A/D


Last topic – Cognitive Radios

According to the FCC:

“We recognize the importance of new cognitive radio

technologies, which are likely to become more

prevalent over the next few years and which hold

tremendous promise in helping to facilitate more

effective and efficient access to spectrum”

- Federal Communications Commission,

ET Docket No. 03-108, Dec 30th 2003


The spectrum shortage….

All frequency bands up to 60 GHz (and beyond) have FCC allocations for multiple users

The allocation from 3-6 GHz is typical - seems very crowded….

3 4 5 6 GHz


The reality…

Even though the spectra is allocated it is almost unused Cognitive radios would allow unlicensed users to share the

spectrum with primary users The TV band is interesting, but higher frequencies are

even more attractive

0 1 2 3 4 5 6 GHz

The TV band


What is a Cognitive Radio? Cognitive radio requirements

» co-exists with legacy wireless systems» uses their spectrum resources » does not interfere with them

Cognitive radio properties» RF technology that "listens" to huge swaths of

spectrum » Knowledge of primary users’ spectrum usage as a

function of location and time» Rules of sharing the available resources (time,

frequency, space)» Embedded database to determine optimal

transmission (bandwidth, latency, QoS) based on primary users’ behavior


Cognitive Radio Functions

D/APA

LNA A/D

IFFT

FFT

ADAPTIVELOADING

INTERFERENCEMEAS/CANCEL

MAE/POWER CTRL

CHANNELSEL/EST

TIME, FREQ,SPACE SEL

LEARN ENVIRONMENT

QoS vs.RATE

FEEDBACKTO CRs

Sensing Radio• Wideband Antenna, PA

and LNA • High speed A/D & D/A,

moderate resolution• Simultaneous Tx & Rx• Scalable for MIMO

Physical Layer• OFDM transmission

• Spectrum monitoring

• Dynamic frequency selection, modulation, power control

• Analog impairments compensation

MAC Layer • Optimize transmission

parameters

• Adapt rates through feedback

• Negotiate or opportunistically use resources

RF/Analog Front-end Digital Baseband MAC Layer


Spectrum Sensing

Key challenge: detecting weak Primary user signals

Analog Processing

Digital Processing NetworkingA/D

Approach:

Spectrum Sensing is a cross-layer functionality


Wideband Sensing Front-end

0 1 2 3 4 5 6 GHz

widebandantenna

A/D

RF Filter

LNA

Huge dynamic range

High speed A/D converter

Wideband Receiver

Nyquist sampling -> Multi-GHz A/D

Large dynamic range signal

Limitation: number of bits in A/D

E.g. 70dB dynamic range needs 12 bits

A/D figure of merit fs*2n

Need dynamic range reduction

Possible solutions

Tunable notch filters

Active cancellation

Spatial filtering using multiple antennas

AGC


Sensing Radio Function

Subdivide the spectrum into sub-channels (say 1 MHz)

Detect primary user occupancy in each location/direction

Continually monitor for appearance of primary user

Provide information to MAC layer

0 0.5 1 1.5 2 2.5

x 109

-90

-85

-80

-75

-70

-65

-60

-55

-50

-45

-40

Frequency (Hz)S

ign

al

Str

en

gth

(d

B)

TV bands

Cell

PCS

Spectrum usage in (0, 2.5) GHz


Spectrum Allocation and Access

What access scheme can assign

ANY sub-channel ANY CR user If we restrict one user per sub-channel:

» Orthogonal Frequency Division Multiple Access (OFDMA)

More general solution:» One user to multiple non-contiguous subchannels (how?)

CR1 CR2 CR3

CR

4

Spectrum Allocation

Spectrum poolf1 fN

PU present

PU absent

Interference


Summary

UWB – Need a new approach which yields high data rates with low complexity or new algorithms for positioning

60 GHz – Requires sophisticated adaptive antenna systems and modulation techniques which have reduced analog complexity

Cognitive radios – Requires sensing and highly adaptive transmission


Conclusion

Lots of New Opportunities That Require Close Cooperation

between Theorists and Implementers!!!

Berkeley Wireless Research Center Why Theorists and Implementers Should Work Together Bob Brodersen...

Documents

Transcript of Berkeley Wireless Research Center Why Theorists and Implementers Should Work Together Bob Brodersen...