Next-Generation Wireless Communications Networks - From Chips to Architecture

workablejeansMobile - Wireless

Nov 21, 2013 (3 years and 4 months ago)

47 views

7 September 2010

IASTED
-
MS
-
2010

1

Next
-
Generation Wireless
Communications Networks
-

From Chips
to Architecture

Daniel Foty
1,2,3
, Bruce Smith
2
, Saurabh Sinha
3
, and Michael Schr
öter
4


1
Gilgamesh Associates LLC, Fletcher, Vermont, The United States of America

2
Sarissa Radio, Inc., Los Altos, California, The United States of America

3
Carl and Emily Fuchs Institute for Microelectronics, University of Pretoria, Pretoria, The
Republic of South Africa

4
The University of California / San Diego, La Jolla, California, The United States of
America


Presented at the IASTED
-
MS

Gaborone, Botswana


7 September 2010

7 September 2010

IASTED
-
MS
-
2010

2

Outline


Infrastructure, abundance, scarcity…


Wireless bandwidth demand


crisis
now


Old model


power
-
centric, bits/Hz (failing)


New model


bandwidth
-
centric, bits/
watt


Emerging mm
-
wave technology network ideas


Nuts and bolts (components, architectures, etc.)


Second
-
generation thoughts


Summary

7 September 2010

IASTED
-
MS
-
2010

3

Thought For Today

“Nothing is easier… than setting some arbitrary goal


preferably based on numbers


and [then going]
after it, in utter disregard of the costs or the
repercussions.”


--

Prof. Thomas Sowell, 2009

7 September 2010

IASTED
-
MS
-
2010

4

Beckmann’s Conjecture

“In a healthy society, engineering design gets
smarter and smarter; in an [unhealthy society], it
gets bigger and bigger.”


--

Prof. Petr Beckmann, “A History of
π,” 1971

7 September 2010

IASTED
-
MS
-
2010

5

1994 “Infrastructure Vision”

Source: Adapted from D. Foty
and E. Nowak,
IEEE Micro
,
June 1994


7 September 2010

IASTED
-
MS
-
2010

6

Abundance and Scarcity

“Every economic era is based on a key abundance
and a key scarcity.”


--

George Gilder,
Wired

magazine, May
1996(!!)

7 September 2010

IASTED
-
MS
-
2010

7

Abundance/Scarcity Matrix

(Gilder, 1996)

(Us, 2010)

Abundance

Scarcity

Power

Transistors

Bandwidth

-

MIPs

-

Bits

Time

Space

Fourteen

years on, we’re still stranded in the old paradigm…

And its penalties are worsening rapidly!!

7 September 2010

IASTED
-
MS
-
2010

8

The New Paradigm

(Gilder, 1996)

(Us, 2010)

Abundance

Scarcity

Bandwidth

Power

Time

Space

We need to make bandwidth much more abundant


on present course, dire shortage by… NOW!!

7 September 2010

IASTED
-
MS
-
2010

9

Bandwidth Demand

The Cisco Visual Networking Index (VNI) Forecast for 2007
-
2012 provides key findings on a variety of consumer and business Inter
net Protocol (IP)
networking trends that are driven by the increasing use of video and Web 2.0 social networking and collaboration applications
--
A
KA visual
networking.


Projections resulting from the project include:



* IP traffic will increase at a combined annual growth rate (CAGR) of 46 percent from 2007 to 2012, nearly doubling every tw
o years.


* The resulting annual bandwidth demand on the world's IP networks will be approximately 522 exabytes, or more than half a z
ettabyte.


* In 2012, Internet video traffic alone will be 400 times the traffic carried by the U.S. Internet backbone in 2000. Interne
t video jumped from 12 percent
of the global consumer Internet traffic in 2006 to 22 percent in 2007.


* Video on demand, IPTV, peer
-
to
-
peer (P2P) video, and Internet video are forecast to account for nearly 90 percent of all c
onsumer IP traffic in 2012.


* Global business IP traffic is forecast to grow strongly at a CAGR of 35 percent from 2007 to 2012. Increased broadband pen
etration in the small
-
business segment and the increased adoption of advanced video communications in the enterprise are major drivers for business

IP

traffic growth.


* Business IP traffic will grow fastest in the developing markets and Asia
-
Pacific. In volume, North America will continue t
o have the most business IP
traffic through 2012, followed by Asia
-
Pacific and Western Europe.


* Global IP traffic will reach 44 exabytes per month in 2012, compared to less than seven per month in 2007.


* By comparison, global IP traffic in 2002 was five exabytes which means that the volume of IP traffic in 2012 will be 100 t
imes as large.


* Monthly global IP traffic in December 2012 will be 11 exabytes higher than in December 2011, a single
-
year increase that w
ill exceed the amount by
which traffic has increased in the eight years since 2000.


* Mobile data traffic will roughly double each year from 2008 through 2012.


Note: A zettabyte is equal to: 1 trillion gigabytes; 1,000 exabytes; 250 billion DVDs, while an exabyte is equal to: 1 billio
n g
igabytes; 1,000 petabytes; 250
million DVDs.

7 September 2010

IASTED
-
MS
-
2010

10

Bandwidth Demand


AT&T Wireless


Disaster


iPhones/iPads


“Bandwidth Hogs”


NYC/SF


Can’t even make
phone calls

mid
-
day
and afternoon (I speak from experience)


Ending unlimited data plans


Re
-
introducing
“caps” beyond which usage becomes metered


iPhone4


Video calls?!?!?!?! (Yeah, right…)


Only allowed via WiFi network


Not over mobile network


Desperate need for more bandwidth

7 September 2010

IASTED
-
MS
-
2010

11

Network Crisis


Problem




Networks are voice
-
centric, not data
-
centric


Need to make a transition to genuinely data
-
centric networks


Track has been voice
-
centric networks plus band
-
aids


Now has reached a crisis point

7 September 2010

IASTED
-
MS
-
2010

12

The “Iron Triangle”



Shannon Information Theory:

C = W log (1 + SNR)



C”


Data rate


SNR = P/N*W, where P is the signal power and N is the noise power spectral density
(one
-
sided), so that N*W is the noise power


W is the bandwidth


(Thanks to Prof. Toby Berger, University of Virginia)


7 September 2010

IASTED
-
MS
-
2010

13

Old Model


Old model


High power,


Low bandwidth


Focus on bits/Hz


However, this clearly isn’t working out now


Empirical observation


Back with simple analysis

7 September 2010

IASTED
-
MS
-
2010

14

Much Thunder, Little Rain

“In any technology, reality must take precedence
over public relations, for nature cannot be fooled.”

--

Richard Feynman, 1986

“You can’t polish a turd.”



Old U.S. Army Proverb

PC Magazine,

2 October 2007

7 September 2010

IASTED
-
MS
-
2010

15

Northern Reaction

7 September 2010

IASTED
-
MS
-
2010

16

The Golden Calf


Bits/Hz


Can only increase w/
more power


Shannon: No limit on bits/Hz, but energy/bit…


“Spectral efficiency”


Not

best approach


All


High entropy methods

7 September 2010

IASTED
-
MS
-
2010

17

Bits Per Watt

BT (Cl
-
2)

BT (Cl
-
1)

802.11b

802.11g

WirelessHD

7 September 2010

IASTED
-
MS
-
2010

18

New Model


New model


Low power


High bandwidth


“Power
-
efficient bandwidth"


Power/bandwidth plane


Move from bandwidth
-
scarce/power
-
rich to power
-
scarce/bandwidth
-
rich


“Iron Triangle”


Harsh constraints


Need a different way forward

7 September 2010

IASTED
-
MS
-
2010

19

Need to Expand the Triangle


Only way to get higher data rate at good “cost”?


Implies
higher carrier frequencies


Net: mm
-
wave is the route to power
-
efficient bandwidth


Escape the tyranny of “bits per Hz”


bits per watt
!!



7 September 2010

IASTED
-
MS
-
2010

20

Propagation

7 September 2010

IASTED
-
MS
-
2010

21

Regulatory


mm
-
Wave


57
-

64 GHz "ISM"
-
like allocation


US/Japan/Europe vary slightly, but 5 GHz of overlap


Unlicensed


And “wide/absolute” but not “wide/relative”


US FCC


500 mW maximum EIRP


71


76 GHz


Licensed band for comms


77 GHz vehicular radar band (R.A.D.A.R for real!!)


81


86 GHz


Licensed band for comms


92


95 GHz


VERY

licensed band for comms


Big opportunity… But
terrible

regulatory environment

7 September 2010

IASTED
-
MS
-
2010

22

Regulatory Help Needed!!


Need

for regulatory relief on 92
-

95GHz
band


Relax power constraint


Same power limits as 60 GHz


500 mW EIRP,
2
µ
W/cm
2

@ 3 m (FCC))


Not good for “pencil
beams”


Let people use it outdoors

7 September 2010

IASTED
-
MS
-
2010

23

Tiered mm
-
Wave Networking


Can divide into three “tiers”

1.
Short range


WP2P, low power, portable (UWB flop)

2.
Medium range (c. 2km)


E.g., building
-
to
-
building

3.
Long range (c. 10km)


Backhaul/bulkhaul


Comments on each

1.
Low cost / low power / on
-
board antenna (UWB space)

2.
Beefier (SiGe bipolar) PA / external antenna

3.
GaN PA / external antenna


Can greatly improve state
-
of
-
the
-
art both
technically
and

cost
-
wise


on all three

7 September 2010

IASTED
-
MS
-
2010

24

Three Network Tiers


Long
-
range (c. 10 km)


Backhaul/bulkhaul story is very promising


Particularly suitable for developing countries
(southern Africa)


Medium
-
range (c. 2 km)


E.g., building
-
to
-
building


70/80 GHz bands


1.2
°

“pencil beam”


90 GHz band would be marvelous


Regulatory relief needed here


Help!!

7 September 2010

IASTED
-
MS
-
2010

25

Extant mm
-
Wave Stuff


“BridgeWave gigabit point
-
to
-
point wireless links provide fiber
-
equivalent
connections between locations by transmitting data over highly secure 60 GHz &
80 GHz (E
-
Band) radio frequencies at gigabit and Fast Ethernet speeds with the
advantage of add/drop data ports, and optional wire
-
speed AES encryption built
-
in.”

7 September 2010

IASTED
-
MS
-
2010

26

Extant mm
-
Wave Stuff


“70/80” bands, full
-
duplex,1.25 Gbps bulk
-
haul wireless Ethernet


Range


c. 6 km


Produced by Gilland Electronics

7 September 2010

IASTED
-
MS
-
2010

27

Limiting Factors


Widespread use of 70/80 for
bulkhaul/backhaul limited by…


Cost


Power consumption


“Up to”


Problem if you live/operate in a rainy
place


Power amplifier technology


Output, cost (TWT & kV power supply, propagation loss
limitation (only overcome with power))


GaN not available yet

7 September 2010

IASTED
-
MS
-
2010

28

Short Range Networking


Short range


Mobile
-
device
-
centric


Requires low power consumption


Driven by…


“Media
-
rich content” (very data
-
intensive)


Enormous storage capacity in mobile devices


Obvious market need… but what happened?

7 September 2010

IASTED
-
MS
-
2010

29

“ETSI” 2004 Forecast

7 September 2010

IASTED
-
MS
-
2010

30

“ETSI” 2004 Forecast

7 September 2010

IASTED
-
MS
-
2010

31

“ETSI” 2004 Forecast

7 September 2010

IASTED
-
MS
-
2010

32

UWB


PR vs. Reality


ETSI 2004 forecast vs. reality

Year

Forecast (Millions)

2004

0.37

2005

15

2006

50

2007

165

2008

338

2009

543

Reality (Millions)

0

0

0

0

0

0



Where’s the beef??



Something clearly wrong

Matches Unit
Shipments of
Hoverponies
During Same
Time Period

7 September 2010

IASTED
-
MS
-
2010

33

Harsh Reality


“Embarrassed polar bear”


Obvious market need, technology couldn’t
reach it


Technological approaches fundamentally flawed


One failure after another


Requires new approaches

7 September 2010

IASTED
-
MS
-
2010

34

Link Issues
-

Reality

7 September 2010

IASTED
-
MS
-
2010

35

Devices


“Devices”


“Semiconductor technology”


Need to make sensible choices here


Avoid “ultimate CMOS” / “SoC” theology


Make good product engineering choices

7 September 2010

IASTED
-
MS
-
2010

36

IC Technology


Mm
-
wave has used III
-
Vs (HBT, HEMT)


Supply limited, very expensive


CMOS is mainstream and inexpensive…


But (despite hand
-
waving) not capable


Only “real” products on market c. 5 GHz


Been stuck there now for several years


Further
CMOS generational scaling doesn’t seem to be
increasing the useful RF
-
CMOS frequency value

7 September 2010

IASTED
-
MS
-
2010

37

SiGe Technology


SiGe BiCMOS can meet the goals


Available for some time


200 GHz/220 GHz


Some

very cost
-
effective


Now available


300 GHz


Delay due to lack of demand


not physics


(SiGe BiCMOS can mix bipolar at will)


Serious development underway for 500 GHz (!!)


7 September 2010

IASTED
-
MS
-
2010

38

SiGe vs. CMOS
-

Cost


Cost


CMOS
less

cost
-
effective than good
SiGe BiCMOS (when including “frequency”)

Process Generation

Per
-
Wafer Cost (Vol.)

0.6

m (no RF features)

$597

0.5

m (no RF features)

$639

0.35

m (no RF features)

$891

0.25

m (no RF features)

$1317

0.18

m (no RF features)

$1639

0.15

m (no RF features)

$1901

0.13

m (no RF features)

$2965

0.18

m
SiGe

(w/ RF)

$2800

Source: Fabless
Semiconductor
Association (FSA), Wafer
Fabrication Pricing
Report, Q2
-
2004

7 September 2010

IASTED
-
MS
-
2010

39

Partition


Two ICs

Chip 1: RF SiGe HBT, IF CMOS; Chip 2


Baseband analog & digital (all CMOS)

7 September 2010

IASTED
-
MS
-
2010

40

Harnessing the SiGe HBT


Next
-
generation methods, scaling techniques

(Credit: Michael Schr
öter)

7 September 2010

IASTED
-
MS
-
2010

41

RF
-
CMOS “Non
-
Scaling”

Threshold Voltage

Power Supply Voltage

Now… Cheating!!

0

1

2

3

4

5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Power Supply Voltage (V)

Drawn
L
min

of the Process Generation (
μ
m)

?!?!

'vddvsl.dat'

'vtvsl.dat'

7 September 2010

IASTED
-
MS
-
2010

42

CMOS Theology vs. RF
-
CMOS


Low power


RF
-
CMOS = AC path to ground


“Proof”


802.11g routers


Linksys (all
-
CMOS chipset) vs. Buffalo (bipolar PA)


Buffalo consumes much less power, runs cooler

7 September 2010

IASTED
-
MS
-
2010

43

New Approach to CMOS

0.1

1

0.001

0.01

0.1

1

10

100

1000

Normalized
g
m
/
I
D

Inversion Coefficient (IC)

Weak Inversion

Asymptote

Square Law

Asymptote

SC

Limit

Modern

LC

'gmid2.dat'

'asympt1.dat'

'asympt2.dat'

'iceq1.dat'

'gmxic20.dat'

'gmxic018.dat'

7 September 2010

IASTED
-
MS
-
2010

44

Characterizing g
m
/I
d



0.18
μm

0.01
0.1
1
1e-05
0.0001
0.001
0.01
0.1
1
10
100
1000
Normalized gm/Id
Inversion Coefficient (IC)
'gmsic20.dat'
'gmsic018.dat'
'gmsic027.dat'
'gmsic036.dat'
'gmsic072.dat'
Simple unified description across
entire spectrum of charge response

(L = 20.0
μm, 0.72 μm, 0.36 μm, 0.24 μm, 0.18 μm)

7 September 2010

IASTED
-
MS
-
2010

45

Critical


Loss of High
-
End

0.01
0.1
1
0.0001
0.001
0.01
0.1
1
10
100
1000
Normalized gm/Id
Inversion Coefficient (IC)
'gmic05.dat'
'gmic3008.dat'
'gmic013b.dat'


Loss of strong inversion



Not getting more bandwidth
with shrinkage!!

0.5

m

0.13

m

0.18

m

7 September 2010

IASTED
-
MS
-
2010

46

GaN Power Amplifiers


This is the piece that is needed to make cost
-
effective 70/80 bulkhaul/backhaul a reality


Conundrum


High
-
power PAs available


but for < 6 GHz


Cellular base stations


70/80/90 HEMT
-
based PAs available


but for
77 GHz automotive radar (350 mW output)


Need to catalyze development of higher
-
power PAs for 70/80 backhaul/bulkhaul

7 September 2010

IASTED
-
MS
-
2010

47

Chips, Packaging, Systems


Long
-
standing RF
-
IC problem


Stability


Not enough chips “work”


Chips that “work” as chips don’t in package/system


Everything

interacts with
everything else


Extensive/regular test required


Very

expensive


Use expensive (6x) ceramic packaging


Innovation


Electrical tuning


Won’t say more for now here…


Work in progress…

7 September 2010

IASTED
-
MS
-
2010

48

7 September 2010

IASTED
-
MS
-
2010

49

Propagation/Modulation/Antenna

7 September 2010

IASTED
-
MS
-
2010

50

POINT to POINT CHANNEL PERFORMANCE @
86.0
GHz.
Antenna aperture
0.0209
steradians (12.5664 in a sphere) conical beam of
reg max 1.20
1.197361
degrees
Antenna gain
43.00
dbi (reg minimum)
Target distance
10 000
meters
32917
feet
6.2
miles
10
Km
Transmitter input power
41.74
dBm
14 928
mw.
Transmitter & coupling losses
6.00
dB
EIRP
35.74
dBm
3749.73
mw.
reg maximum 150mw/100 MHz=
3750
mw
Antenna output
78.74
dBm
Fade loss
3.0
dB
Other losses / Interference
6.0
dB
Coding gain
4
dB
Frequency
86.0
GHz.
wavelength=
0.349
cm.
Rain loss
10.00
dB
Moderate Rain or fog
Free space loss
151.1
dB
10*Log(4*PI*d/lambda)^2'
Received isotropic power
-87.39
dBm
Receiver antenna gain
43.00
dB
Edge of coverage loss
2.00
dB
Received signal power
-46.39
dBm
Sky temperature
36.2
dB - K
4124
degrees K
6931
degrees F
Sky Temp Direct Sun (worst case)
Antenna temperature
25.7
dB - K
373
degrees K
180
degrees F
System noise temperature
36.5
dB - K
4497
degrees K
Boltzmann's constant
-199
dBm/K-Hz
Receiver figure of merit
1.18
Noise spectral density, No
-162
dBm/Hz
Spectral efficiency b/Hz
2.00
b/Hz.
(reg limit 0.125 min)
Received Pr/No
115.68
dB - Hz
Data rate in channel
97
dB-bits/s
5 000 000 000
Channel bit rate per sec.
Received, Eb/No
19
dB
Safety margin
3
dB
Covers edge of beam diminishment
Required Sp/Np
15
dB for BER=10^-8
REMAINING BUDGET
0.69
db
7 September 2010

IASTED
-
MS
-
2010

51

Dynamic Channel Building


A.k.a. (these days)


“Cognitive radio”


Knows the “facts on the ground”


Uses policy
-
based decisions to allocate
spectrum…


… and manages the error rate


(Another work
-
in
-
progress…)


7 September 2010

IASTED
-
MS
-
2010

52

System Design


EVERYTHING is system design now!!


Interactions too strong to neglect


Can’t build sub
-
parts in isolation and then toss
them together


This is particularly true in radio/wireless


This is particularly, particularly true at mm
-
wave
frequencies


Requires complete and comprehensive approach
to engineering


end
-
to
-
end


7 September 2010

IASTED
-
MS
-
2010

53

Mm
-
Wave Radio System

7 September 2010

IASTED
-
MS
-
2010

54

More Complex Challenge


“Using watts as a replacement for bandwidth, you get radio and
television stations all over the air with high
-
powered signals. You get
cellular phone systems with one base station every 30 miles…”


--

George Gilder, 1996


Translation: “
Densify the network grid!!”

7 September 2010

IASTED
-
MS
-
2010

55

Mesh Network Basis?

7 September 2010

IASTED
-
MS
-
2010

56

77 GHz


“Vehicular”


Allocation at 77 GHz for “vehicular radar”


Main interest


Automotive CAS radar


Possible extension to autonomous vehicles

7 September 2010

IASTED
-
MS
-
2010

57

Conclusions


Strong identified need for more bandwidth


Crisis is here


and will only get worse


E.g., AT&T


NYC and SF


Various technologies have flopped though


Failures clearly due to structural shortcomings


Mm
-
wave offers the best way forward


Natural bandwidth


Abundant bandwidth without
burning input power to “create” it


Break design
-
iteration mess with better methods

7 September 2010

IASTED
-
MS
-
2010

58

Conclusions


Goals:



Unify “UWB” targets with mm
-
wave (“Tier 1”)


Develop for medium/long range wireless
infrastructure needs (“Tier 2” and “Tier 3”)


Need cost/performance upgrade


Various “enablers” bring this within reach


Must catalyze GaN PA development


Regulatory relief on 92


95 GHz needed!!


Move from power
-
rich/bandwidth scarce to
bandwidth
-
rich/power
-
scarce

7 September 2010

IASTED
-
MS
-
2010

59

CRITICAL CONCLUSION!!


There is clear real
-
world demand for Gbps
wireless networking with much better
performance and cost numbers


Failure to get there is due to
fundamental
shortcomings in engineering thought


Main consequence


All “methods”
create bandwidth by burning power


Need
fundamentally new strategies

7 September 2010

IASTED
-
MS
-
2010

60



Thank You!!


Falem Nderit!!