Download No Slide Title - Stanford University

Wireless Communications:
Trends and Challenges
ANDREA J. GOLDSMITH
Dept. of Electrical Engineering
Stanford University
http://ee.stanford.edu/~andrea
8C32810.1-Cimini-7/98
OUTLINE
• Introduction
• Radio Environment
• Physical Layer Issues
• Channel Access Issues
• Network Issues
• Standards and Future Systems
• Summary
8C32810.20-Cimini-7/98
WIRELESS DATA VISION
Region
TAXI
City
laptops, PDAs
Campus
In-Building
Seamless Multimedia Networks
with Mobility and Freedom from Tethers
[R. Katz, "Does Wireless Data Have a Future?", Plenary Talk, INFOCOM '96]
8C32810.35-Cimini-7/98
VOICE VERSUS DATA VERSUS VIDEO
Voice
Data
Video
< 100 ms
–
< 100 ms
< 1%
0
<1%
BER
10-2 - 10-3
< 10-5
< 10-7
Data Rate
8-32 kbps
1-100 Mbps
1-20 Mbps
Continuous
Bursty
Continuous
Delay
Packet Loss
Traffic
Wired Networks Trying to Integrate
(ATM, SONET, Multimedia Services)
8C32810.80-Cimini-7/98
WHAT IS THE FUTURE
OF WIRELESS DATA?
100
USA market
90
80
millions
70
cellular + PCS subs
Internet users
60
50
paging subs
40
30
dedicated wireless
data subs
20
10
annual laptop sales
0
1995
*Estimates as of 1996
7C29822.008-Cimini-9/97
laptop users
2000
THE ISSUE IS PERFORMANCE
"The mobile data market has been slow to take off,
but progress is being made. The most formidable
obstacle to user acceptance remains performance."
I. Brodsky, "Countdown to Mobile Blast Off",
Network World, February 19, 1996
Mobile
Multimedia
Terminal
Network
Adaptation
& Control
Wireless
Interface
Radio
Protocols
& Modem
Radio Link
Radio
Protocols
& Modem
Mobility
Control
Protocols
Network
Interface
Radio
Port
RADIO ACCESS SEGMENT
Signaling/
Routing
Mobility
Control
MOBILE NETWORK SEGMENT
BROADBAND WIRELESS
NETWORK
• Link Performance: Data Rate and Quality
• Network Performance: Access, Coverage, Reliability,
QoS, and Internetworking
8C32810.81-Cimini-7/98
GAP BETWEEN WIRED AND
WIRELESS NETWORK CAPABILITIES
LOCAL AREA PACKET SWITCHING
100 M
100,000
WIDE AREA CIRCUIT SWITCHING
ATM
ATM
100,000
Ethernet
10,000
FDDI
Ethernet
10,000
wired- wireless
bit-rate "gap"
100
wired- wireless
1000 User
Bit-Rate
1000 User
Bit-Rate
(kbps)
1st gen
WLAN
Polling
2nd
gen
WLAN
100
ISDN
(kbps)
28.8 modem
9.6 modem
2.4 modem
Packet
1
PCS
2.4 cellular
1
Radio
32 kbps
14.4
9.6 cellular digital
cellular
10
10
bit-rate "gap"
.1
.1
.01
.01
1970
1980
1990
YEAR
8C32810.79-Cimini-7/98
2000
1970
1980
1990
YEAR
2000
TECHNICAL CHALLENGES
• Low-Power/Low-Cost Implementations
• Scarce Radio Spectrum
• Radio Channel Characteristics
– Limits on Signal Coverage
– Limits on Data Rates
• Efficient Network Architectures and
Protocols
• Seamless Internetworking
• Authentication and Security
7C29822.010-Cimini-9/97
RADIO ENVIRONMENT
• Path Loss
• Shadow Fading
• Multipath
8C32810.83-Cimini-7/98
Limit the Bit Rate
and/or Coverage
PATH LOSS MODEL
• Different, often complicated, models are
used for different environments.
• A simple model for path loss, L, is
L=
Pr
Pt
=K
1
f da
2
where Pr is the local mean received signal
power, Pt is the transmitted power, d is the
transmitter-receiver distance, f is frequency,
and K is a transmission constant.
The path loss exponent a = 2 in free space;
2  a  4 in typical environments.
7C29822.011-Cimini-9/97
SHADOW FADING
• The received signal is shadowed by
obstructions such as hills and buildings.
• This results in variations in the local mean
received signal power,
Pr (dB) = Pr (dB) + Gs
where Gs ~ N(0, ss ), 4  ss  10 dB.
2
• Implications
– nonuniform coverage
– increases the required transmit power
R
8C32810.84-Cimini-7/98
P = Pr0
MULTIPATH
Received
Power
t
Delay Spread
h(t) =
Si aiejq d(t-ti)
i
• Constructive and Destructive Interference
of Arriving Rays
10
0 dB With Respect
-10 to RMS Value
-20
-30
0.5l
8C32810.85-Cimini-7/98
0
0.5
0
10
1
t, in seconds
20
x, in wavelength
1.5
30
Received
Power
DELAY SPREAD
TIME DOMAIN INTERPRETATION
Two-ray model
t = rms delay spread
2t
Delay
t small
Channel Input
1
0
T
1
T
Channel Output
0
T
2T
0
T
2T
2T
t large
T
t
• T small
• t large
negligible intersymbol interference
significant intersymbol interference,
T which causes an irreducible error floor
8C32810.88-Cimini-7/98
PHYSICAL LAYER ISSUES
• Link Performance Measures
• Modulation Tradeoffs
• Flat Fading Countermeasures
• Delay Spread Countermeasures
8C32810.91-Cimini-7/98
LINK PERFORMANCE MEASURES
PROBABILITY OF BIT ERROR
• The probability of bit error, Pb, in a radio
environment is a random variable.
– average Pb, Pb
– Pr [Pb > Pbtarget] D outage, Pout
=
• Typically only one of these measures is
useful, depending on the Doppler frequency
and the bit rate.
8C32810.92-Cimini-7/98
LINK PERFORMANCE MEASURES
EFFICIENCY
• Spectral Efficiency
– a measure of the data rate per unit
bandwidth for a given bit error
probability and transmitted power
• Power Efficiency
– a measure of the required received
power to achieve a given data rate
for a given bit error probability and
bandwidth
• Throughput/Delay
8C32810.72-Cimini-7/98
GOALS OF
MODULATION TECHNIQUES
• High Bit Rate
• High Spectral Efficiency
• High Power Efficiency
• Low-Cost/Low-Power Implementation
• Robustness to Impairments
8C32810.74-Cimini-7/98
DIGITAL MODULATION
• Any modulated signal can be represented as
s(t) = A(t) cos [wct + f(t)]
amplitude
phase or frequency
s(t) = A(t) cos f(t) cos wct
in-phase
- A(t) sin f(t) sin wct
quadrature
• Linear versus nonlinear modulation  impact
on spectral efficiency
• Constant envelope versus non-constant
envelope hardware implications with impact
on power efficiency
8C32810.94-Cimini-7/98
LINEAR MODULATION TECHNIQUES
s(t) = [ S an g (t-nT)]cos wct - [ S bn g (t-nT)] sin wct
n
n
I(t), in-phase
Q(t), quadrature
LINEAR MODULATIONS
Square
Constellations
M-ARY QUADRATURE
AMPLITUDE MOD.
(M-QAM)
M 4
M=4
(4-QAM = 4-PSK)
CONVENTIONAL
4-PSK
(QPSK)
8C32810.95-Cimini-7/98
M-ARY PHASE
SHIFT KEYING
(M-PSK)
Circular
Constellations
M 4
OFFSET
DIFFERENTIAL
4-PSK
4-PSK
(OQPSK) (DQPSK, p/4-DQPSK)
SIGNAL CONSTELLATIONS
M-PSK (Circular Constellations)
bn
4-PSK
16-PSK
an
M-QAM (Square Constellations)
bn
16-QAM
4-PSK
an
Tradeoffs
– Higher-order modulations (M large) are more spectrally
efficient but less power efficient.
– M-QAM is more spectrally efficient than M-PSK but
also more sensitive to system nonlinearities.
8C32810.96-Cimini-7/98
PULSE SHAPING
Rectangular pulses are spectrally inefficient
pulse shaping
intersymbol interference (ISI)
non-constant envelope
Nyquist pulses
8C32810.75-Cimini-7/98
RAISED COSINE PULSE SHAPING
g(t)
a = 0, 0.5
a=1
0
T
2T
3T
a = 0 a = 0.5
a=0
a = 0.5
4T
G(f)
a=1
-1
T
Relative Peak
Instantaneous Power (dB)
-
1
2T
0
1
2T
1
T
30
256 QAM
64 QAM
20
16 PSK
16 QAM
10
8 PSK
4 PSK
0
8C32810.97-Cimini-7/98
0.25 0.5 0.75 1.0
Cosine Rolloff Factor, a
f
t
DEMODULATION
• Coherent detection requires a coherent phase
reference.
– difficult to obtain in a rapidly fading
environment
– increases receiver complexity
• Differential detection uses the previous symbol
for the reference signal.
– eliminates need for coherent reference
– entails loss in power efficiency (up to 3 dB)
– Doppler causes irreducible error floor,
typically small for high bit rates
8C32810.133-Cimini-7/98
FREQUENCY SHIFT KEYING
• Continuous Phase FSK (CPFSK)
– digital data encoded in the frequency shift
– typically implemented with frequency
modulator to maintain continuous phase
t
s(t) = A cos [wct + 2 pkf  d(t) dt]
-
– nonlinear modulation but constant-envelope
• Minimum Shift Keying (MSK)
– minimum bandwidth, sidelobes large
– can be implemented using I-Q receiver
• Gaussian Minimum Shift Keying (GMSK)
– reduces sidelobes of MSK using a
premodulation filter
– used by RAM Mobile Data, CDPD,
and HIPERLAN
8C32810.98-Cimini-7/98
SPECTRAL CHARACTERISTICS
10
QPSK/DQPSK
GMSK
Power Spectral Density (dB)
0
-20
-40
-60
B3-dBTb = 0.16
0.25
-80
1.0
-100
-120
(MSK)
0
0.5
1.0
1.5
2.0
Normalized Frequency (f-fc)Tb
7C29822.013-Cimini-9/97
2.5
BIT ERROR PROBABILITY
AWGN CHANNEL
10-1
5
2
10-2
5
For Pb = 10-3
2
BPSK 6.5 dB
QPSK 6.5 dB
DBPSK ~8 dB
DQPSK ~9 dB
DBPSK
10-3
BPSK, QPSK
5
Pb
2
DQPSK
10-4
5
2
10-5
5
2
10-6
0
2
4
6
8
10
12
14
gb, SNR/bit, dB
• QPSK is more spectrally efficient than BPSK with the
same performance.
• M-PSK, for M>4, is more spectrally efficient but requires
more SNR per bit.
• There is ~3 dB power penalty for differential
detection.
8C32810.99-Cimini-7/98
BIT ERROR PROBABILITY
FADING CHANNEL
1
5
2
10-1
5
2
10-2
5
Pb
DBPSK
2
BPSK
10-3
5
AWGN
2
10-4
5
2
10-5
0
5
10
15
20
25
30
35
gb, SNR/bit, dB
• Pb is inversely proportion to the average SNR per bit.
• Transmission in a fading environment requires about
18 dB more power for Pb = 10-3.
8C32810.100-Cimini-7/98
BIT ERROR PROBABILITY
EFFECTS OF DOPPLER SPREAD
• Doppler causes an irreducible error floor when differential
detection is used decorrelation of reference signal.
100
QPSK
DQPSK
10 -1
Rayleigh Fading
10 -2
-3
Pb 10
fDT=0.003
No Fading
10 -4
0.002
0.001
10 -5
0
10 -6
0
10
20
30
40
50
60
gb, SNR/bit, dB
• The irreducible Pb depends on the data rate and the Doppler.
For fD = 80 Hz,
data rate
T
Pbfloor
10 kbps
10-4s
3x10-4
100 kbps
10-5s
3x10-6
1 Mbps
10-6s
3x10-8
The implication is that Doppler is not an issue for high-speed
wireless data.
[M. D. Yacoub, Foundations of Mobile Radio Engineering , CRC Press, 1993]
8C32810.101-Cimini-7/98
BIT ERROR PROBABILITY
EFFECTS OF DELAY SPREAD
• ISI causes an irreducible error floor.
10-1
Irreducible Pb
Coherent Detection
+ BPSK
QPSK
OQPSK Modulation
x MSK
x
10-2
x
x
+
+
x
+
10-3
+
x
+
10-4
10-2
•
10-1
100
rms delay spread t
=
symbol period
T
The rms delay spread imposes a limit on the maximum bit rate
in a multipath environment.
For example, for QPSK,
t
Maximum Bit Rate
Mobile (rural)
25 msec
8 kbps
Mobile (city)
2.5 msec
80 kbps
Microcells
500 nsec
400 kbps
Large Building
100 nsec
2 Mbps
[J. C.-I. Chuang, "The Effects of Time Delay Spread on Portable Radio
Communications Channels with Digital Modulation," IEEE JSAC, June 1987]
8C32810.102-Cimini-7/98
SUMMARY OF
MODULATION ISSUES
• Tradeoffs
– linear versus nonlinear modulation
– constant envelope versus non-constant
envelope
– coherent versus differential detection
– power efficiency versus spectral efficiency
• Limitations
– flat fading
– doppler
– delay spread
8C32810.103-Cimini-7/98
HOW DO WE OVERCOME THE
LIMITATIONS IMPOSED BY THE
RADIO CHANNEL?
• Flat Fading Countermeasures
– Fade Margin
– Diversity
– Coding and Interleaving
– Adaptive Techniques
• Delay Spread Countermeasures
– Equalization
– Multicarrier
– Spread Spectrum
– Antenna Solutions
8C32810.104-Cimini-7/98
DIVERSITY
• Independent signal paths have a low probability
Received Signal Power
(dBm)
of experiencing deep fades simultaneously.
0
-20
-40
-60
-80
-100
0
4
8
12
16
d
The chance that two deep fades
occur simultaneously is rare.
• The basic concept is to send the same
information over independently fading radio
• Independent fading paths can be achieved by
separating the signal in time, frequency, space,
polarization, etc.
8C32810.105-Cimini-7/98
DIVERSITY COMBINING TECHNIQUES
• • •
a1
a2
a3
aM
Combiner
Output
• Selection Combining: picks the branch with the
highest SNR.
• Equal-Gain Combining: all branches are coherently
combined with equal weights.
• Maximal-Ratio Combining: all branches are coherently
combined with weights which depend on the branch
SNR.
8C32810.106-Cimini-7/98
DIVERSITY PERFORMANCE
• There is dramatic improvement even with two-branch
selection combining.
– 10 dB reduction in required SNR for 1% outage 
less transmitted power or higher bit rates or larger
coverage area
Pb
10-1
5
Pout
99.99
99.9
99.5
98.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
Maximal
Ratio
Combining
2
10-2
5
M=2
20.0
2
10-3
10.0
M=1
5
5.0
2
2.0
10-4
5
1.0
Maximal
Ratio
Equal
Gain
0.5
M=2
Selection
0.2
0.1
2
10-5
5
0.05
M=4
2
0.02
10-6
5
10
15
20 25 30
gb, SNR/bit, dB
35
40
0.01
-40
-30
-20
10log
(
-10
1
margin
0
)
• The output SNR with Maximal-Ratio Combining improves
linearly with the number of diversity branches, M  the
complexity becomes prohibitive.
7C29822.014-Cimini-9/97
10
CHANNEL CODING
• Channel coding reduces Pb by introducing redundancy
in the transmitted bit stream.
• Block and convolutional codes acheive this improvement
at the expense of increased signal bandwidth or a lower
data rate.
• Bit error probability–AWGN channel
10-2
5
BPSK
2
10-3
5
For Pb = 10-6
2
Uncoded
Hamming
BCH
Conv.
10-4
5
Pb
Uncoded
2
Hamming
(7,4,1)
10-5
10.5 dB
10.0 dB
6.5 dB
5.0 dB
BCH
(127,64,10)
5
Conv.
1/2 rate
(k=7)
2
10-6
5
2
10-7
0
2
4
6
8
10
12
14
gb, SNR/bit, dB
• Fading causes burst errors. If the fading is slow enough
relative to the symbol rate, coding will not be effective.
7C29822.015-Cimini-9/97
CODING PERFORMANCE
FADING CHANNEL
• Pb performance for the IS-136 rate-1/2 convolutional
code on a simulated mobile radio channel (harddecision decoding).
1
Uncoded
50 km/hr
Coded
1 km/hr
Coded
8 km/hr
Coded
50 km/hr
Coded
100 km/hr
10-1
Pb 10-2
10-3
10-4
8
10
12
14
16
gb, SNR/bit, dB
18
20
• Negligible coding gain if fading is slow compared
to bit rate  interleaving
[V. Iyengar and J. Michaelides, "Performance Evaluations of RLPs (Radio Link Protocols)
for TDMA Data Services," ITIA Contribution TR45.3.2.5/93.03.30.10, Chicago, March 30, 1993]
8C32810.17-Cimini-7/98
CODING PERFORMANCE
FADING CHANNEL
• Pb performance for the IS-136 rate-1/2 convolutional
code on a simulated mobile radio channel (soft
decision decoding).
1
Uncoded
50 km/hr
Coded
1 km/hr
Coded
8 km/hr
Coded
50 km/hr
Coded
100 km/hr
-1
10
10-2
Pb
10-3
10-4
10 -5
8
10
12 14 16
gb, SNR/bit, dB
18
20
[V. Iyengar and J. Michaelides, "Performance Evaluations of RLPs (Radio Link Protocols)
for TDMA Data Services," ITIA Contribution TR45.3.2.5/93.03.30.10, Chicago, March 30, 1993]
8C32810.18-Cimini-7/98
CODING AND INTERLEAVING
• The basic principle is to spread the burst errors
over many code words.
1 codeword
read
into
interleaver
by rows
1
2
3
4
5
6
7
8
9
10
11
12
read
out
by
columns
1,5,9,2,6,10,3,7,
11,4,8,12
channel
reads out
by rows
1, 2 ,3,4,5, 6 ,7,8,
9, 10 ,11,12
1
2
3
4
5
6
7
8
9
10
11
12
1,5,9, 2 , 6 ,10 ,3,7,11,
4,8,12
reads in
by rows
• The required interleaver size can be large if the
relative fading rate is slow, as is usually the case
for high-speed data. For example, fD = 10 Hz,
bit rate = 10 Mb/s, error burst = 330,000 bits.
delay and complexity
8C32819.16-Cimini-7/98
ADVANCED CODING TECHNIQUES
• Trellis Codes
– reduce Pb without bandwidth expansion
through joint design of the channel code
and signal constellation
– can be designed with “built-in” time diversity
• Turbo Codes
– exhibit enormous coding gains
– interleaving inherent to code design
– very complex with large delays
– not well-understood for fading channels
8C32810.19-Cimini-7/98
CODING PERFORMANCE TCM
8PSK TCM
-1
10
10
-3
10
-4
Uncoded
4 PSK
Ungerboeck
Code
R=2/3, M=4
Pb
10
-2
MSB
b3,
R=2/3
10
-5
-6
10 10
b1,
R=2/3
b2,
R=2/3
12
14
16
18
Es/N0 (dB)
8C32810.69-Cimini-7/98
LSB
20
22
ADAPTIVE TECHNIQUES
• Adaptive Modulation
• Automatic Repeat Request
8C32810.21-Cimini-7/98
ADAPTIVE MODULATION
TRANSMITTER
Adaptive
Modulation
and Coding
RECEIVER
Power
Control
Channel
noise
Demodulation
and Decoding
+
Channel
Estimate
Delay
FEEDBACK CHANNEL
• Power and/or data rate adapted at transmitter to
channel conditions
• Potential for large increase in spectral efficiency
• Can be combined with adaptive compression
– requires reliable feedback channel and accurate
channel estimation
– increases transmitter and receiver complexity
8C32810.22-Cimini-7/98
AUTOMATIC REPEAT REQUEST (ARQ)
• Method of "self-adapting" the data rate to
the channel conditions
• Used in combination with error-detecting code
• Variations of ARQ used in Mobitex and CDPD
• Types: Stop-and-Wait, Go-Back-N, SelectiveRepeat
– power and spectrally inefficient
– impacts higher layer protocols
– necessary for meeting stringent Pb
requirements or data
7C29822.023-Cimini-9/97
DELAY SPREAD COUNTERMEASURES
• Signal Processing
– at the receiver, to alleviate the problems
caused by delay spread (equalization)
– at the transmitter, to make the signal less
sensitive to delay spread (multicarrier,
spread spectrum)
• Antenna Solutions
– change the environment to reduce, or
eliminate, the delay spread (distributed
antenna system, small cells, directive
antennas)
7C29822.024-Cimini-9/97
EQUALIZER TYPES AND STRUCTURES
The goal of equalization is to cancel the ISI
or, equivalently, to flatten the frequency response.
Equalizer
Types
Linear
Nonlinear
DFE
ML Symbol
Detector
MLSE
Structures
Transversal
Lattice
Transversal
Lattice
Transversal
Channel
Estimator
[J. G. Proakis, "Adaptive Equalization for TDMA Digital Mobile Radio,"
IEEE Trans. on Veh. Tech. , May 1991]
8C32810.107-Cimini-7/98
LINEAR EQUALIZER
• A linear equalizer effectively inverts the channel.
n(t)
Channel
Hc(f)
Equalizer
1
Heq(f)
Hc(f)
• The linear equalizer is usually implemented as a
tapped delay line.
• On a channel with deep spectral nulls, this equalizer
enhances the noise.
poor performance on frequency-selective
fading channels
8C32810.108-Cimini-7/98
DECISION FEEDBACK EQUALIZER
DFE
n(t)
x(t)
Hc(f)
Forward
Filter
^
x(t)
+
Feedback
Filter
• The DFE determines the ISI from the previously detected
symbols and subtracts it from the incoming symbols.
• This equalizer does not suffer from noise enhancement
because it estimates the channel rather than inverting it.
 The DFE has better performance than the linear
equalizer in a frequency-selective fading channel.
• The DFE is subject to error propagation if decisions are
made incorrectly.
• Decisions are made on coded symbols.  no coding gain
7C29822.025-Cimini-9/97
MAXIMUM LIKELIHOOD
SEQUENCE ESTIMATION
• MLSE has theoretically optimum performance.
• It requires knowledge of the channel parameters and
the noise distribution.
• The implementation complexity grows exponentially
with the length of the channel impulse response 
not practical for high bit rates.
8C32810.109-Cimini-7/98
EQUALIZER ISSUES FOR
HIGH-SPEED WIRELESS DATA
• The number of required equalizer taps, N, is proportional
to the delay spread.
• The equalizer taps must be adapted at the highest
Doppler rate.
– The length and periodicity of the training sequence
impacts the spectral efficiency.
– There is a tradeoff between speed of convergence
and complexity.
Algorithms
(for DFE)
Number of
Multiply
Operations
Least Mean
Square (LMS)
2N + 1
Kalman
Recursive Least
Squares (RLS)
2.5N2 + 4.5N
Square Root
Fast Kalman
7C29822.026-Cimini-9/97
Convergence
Advantages
Disadvantages
Low computational
complexity
Slow convergence,
depends on
channel
~N
Fast convergence,
good tracking ability
High
computational
complexity
1.5N2 + 6.5N
~N
Better stability
than Kalman
High
computational
complexity
20N + 5
~N
Fast convergence
and good tracking
Could be
unstable
~10-100N
EQUALIZER PERFORMANCE
BPSK
1
10 Mbps
5
10 -1
10 -2
1
10
Pb 10 -3
10 -4
10 -5
.1
5
.1
1
no equalizer
DFE
10 -6
25
30
35
40
45
50
SNR (dB)
BPSK
1
16 Mbps
8
4,16
10 -1
1
8
.1
.1,4
1
Pout 10 -2
10 -3
no equalizer
DFE
10 -4
1
10-4
10-8
Target Pb
10-12
• Pahlavan has shown that, for 30-meter cells (t = 50 ns), 20 Mb/s
can be achieved using a DFE with 3 forward taps and 3 feedback taps.
[K. Pahlavan, S. J. Howard, and T. A. Sexton, "Decision Feedback Equalization
of the Indoor Radio Channel," IEEE Trans. on Commun., January 1993]
8C32810.110-Cimini-7/98
MULTICARRIER MODULATION
• The transmission bandwidth is divided into many
narrow subchannels which are transmitted in
parallel.
• Ideally, each subchannel is narrow enough so
that the fading it experiences is flat  no ISI.
Transmitter
R/N b/s
R/N b/s
R/N b/s
QAM
filter
QAM
filter
QAM
filter
d0(t)
f0
d1(t)
RF
D(t)
f1
d N-1(t)
fN-1
Bandlimited
signals
f0
Receiver
f2
f1
filter
f0
RF
filter
f1
f0
QAM
f1
filter
fN-1
8C32810.111-Cimini-7/98
QAM
QAM
fN-1
OFDM RECEIVER STRUCTURE
Receiver

f0
RF

d(0)
d(1)
f1

parallel
to
serial
converter
QAM
d(N-1)
f N-1
• Subchannel Separation
1
NT
^
– integrate over NT, then d(m) = d(m)
– choose fn = f0 + nDf, with Df =
• Efficient FFT Implementation
• A guard interval can virtually eliminate ISI
(or, interblock interference)  lower spectral
or power efficiency.
8C32810.113-Cimini-7/98
WHAT TO DO WITH
BAD SUBCHANNELS?
• Coding Across Subchannels  works best
with large delay spread
• Frequency Equalization  requires accurate
channel estimation
• Adaptive Loading  requires reliable
feedback channel and accurate channel
estimation
8C32810.114-Cimini-7/98
MULTICARRIER MODULATION
ISSUES FOR HIGH-SPEED
WIRELESS DATA
• Minimal training is required.
• Time-varying fading, frequency offset, and timing
mismatch impair the orthogonality of the
subchannels.
• Large peak-to-average power ratio is a serious
problem when transmitting through a nonlinearity.
– possible solutions: nonlinear coding,
clipping and filtering
8C32810.115-Cimini-7/98
CURRENT AND PROPOSED
APPLICATIONS OF OFDM
• Asymmetric Digital Subscriber Line
• Digital Audio Broadcasting
• Wireless LAN
• Digital Terrestrial Television
• High Speed Cellular
8C32810.116-Cimini-7/98
SPREAD SPECTRUM
• Spread spectrum increases the transmit signal
bandwidth to reduce the effects of flat fading,
ISI and interference.
• SS is used in all wireless LAN products in the ISM
band
– required for operation with reasonable power
levels
– minimal performance impact on other systems
– IEEE 802.11 standard
• There are two SS methods: direct sequence and
frequency hopping.
– Direct sequence multiplies the data sequence
by a faster chip sequence.
– Frequency hopping varies the carrier
frequency by the same chip sequence.
8C32810.117-Cimini-7/98
DIRECT SEQUENCE
SPREAD SPECTRUM
Interference
Data
(T)
Modulator
Carrier
Recovery
Channel
Spreading
(PN) Code
Tc << T
Data
(T)
Demod
Spreading
(PN) Code
Synch
Transmitter
Receiver
Narrowband
Filter
Narrowband
Interference
Data Signal
with Spreading
Modulated
Data
8C32810.117-Cimini-7/98
Original
Data Signal
ISI
Receiver
Input
Other
SS Users
Other
SS Users
Demodulator
Filtering
ISI
RAKE RECEIVER
sc(t)
Received
Signal
sc(t-Tc)
Coherent
Combiner
Data
Output
Demodulator
sc(t-2Tc)
•
•
•
sc(t-TM)
• When the chip time is much less than the rms delay spread,
each branch has independent fading  equivalent to
diversity combining.
• When the chip time is greater than the rms delay spread,
the paths cannot be resolved  no diversity gain.
8C32810.119-Cimini-7/98
PERFORMANCE OF RAKE RECEIVER
FADING CHANNEL
0.5
DPSK
10-1
Rayleigh
10-2
Pb
10-3
RAKE
AWGN
10-4
10-5
8C32810.27-Cimini-7/98
0
5
10
gb, SNR/bit, dB
15
SPREAD SPECTRUM ISSUES
FOR HIGH-SPEED WIRELESS DATA
• Hardware Complexity
– synchronization
– high processing speeds for high
bit rates
– RAKE receiver
• High Required Bandwidth to Accommodate
Spreading
Spread spectrum is difficult at
high bit rates and not really
needed.
8C32810.120-Cimini-7/98
ANTENNA SOLUTIONS
Goal: Reduce (or eliminate) delay spread
• Distributed Antenna System
• Very Small Cells  antenna in every room
• Sectorization
• Directive Antennas/Beam Steering
Omnidirectional
120
90
60
150
180
0
330
210
7C29822.028-Cimini-9/97
120
30
240
270
Sectorized
300
90
Directive
60
150
120
30
180
0
330
210
240
270
300
90
60
150
30
180
0
330
210
240
270
300
DISTRIBUTED ANTENNA SYSTEM
Probability Abscissa Exceeded
1
Distributed
Monopoles
.5
0
Central
Monopole
0
10
20
30
RMS Delay Spread (ns)
40
50
[A. A. M. Saleh, A. J. Rustako, Jr., and R. S. Roman, "Distributed Antennas
for Indoor Radio Communications," IEEE Trans. on Commun., December 1987]
7C29822.029-Cimini-9/97
EXAMPLES OF
PERFORMANCE IMPROVEMENTS
• High-Speed Narrowbeam Antenna Experiment
[P. F. Driessen "Gigabit/s Indoor Wireless Systems with
Directional Antennas," IEEE Trans. on Comm., August 1996]
– directional antennas (15° beamwidth) at both ends of
LOS link
– no equalization
– 622 Mbps BPSK transmission without errors
• Sectored Antennas [G. Yang and K. Pahlavan, "Comparative
Performance Evaluation of Sector Antenna and DFE Systems
in Indoor Radio Channels," Proc. of ICC '92]
– 6 sectors at base and mobile
– best combination chosen
– for Pout = 0.01, 5 Mbps with omni, 25 Mbps with
sectored antenna
Omnidirectional Antennas
1
.001
30 Mbps
.01
10 Mbps
5 Mbps
3 Mbps
2 Mbps
.01
.0001
.02
30 Mbps
20 Mbps
.1
Pout
Six Sector Antennas
Pout
20 Mbps
.005
.002
10 Mbps
.001
1 Mbps
.0005
10 10-1 10-2 10 10 10 10 10 10 10 10 10 10 10
100 10-1 -2 10-3 10-4 10-5 10-6 10-7 10-8 10-9 10-10 10-11 10-12 10-13
10
Target Pb
Target Pb
0
8C32810.121-Cimini-7/98
-3
-4
-5
-6
-7
-8
-9
-10
-11
-12
-13
SUMMARY OF COUNTERMEASURES
• Diversity
• Coding and Interleaving
• Adaptive Techniques
• Equalization
• Multicarrier
• Spread Spectrum
• Antenna Solutions
These techniques can be combined.
8C32810.123-Cimini-7/98
COMBINED EQUALIZATION AND
SECTORED ANTENNAS
1
Pt = 100 mW
Rb = 20 Mbps
Omni
.1
Pout
Omni+DFE
.01
Sector
Sec+DFE
.001
.0001
20
40
60
Square room length (meter)
1
30mx30m
Omni+DFE
.1
Pout
Omni
Sector
.01
.001
.0001
Sec+DFE
0
10
20
30
Rb (Mbps)
40
50
[G. Yang and K. Pahlavan, "Comparative Performance Evaluation of Sector
Antenna and DFE Systems in Indoor Radio Channels," Proc. of ICC '92]
8C32810.122-Cimini-7/98
CHANNEL ACCESS ISSUES
• Multiple Access
• Random Access
• Frequency Reuse
8C32810.125-Cimini-7/98
MULTIPLE ACCESS TECHNIQUES
• Frequency Division (FDMA)
• Time Division (TDMA)
• Code Division (CDMA)
• Hybrid Approaches
7C29822.033-Cimini-9/97
FDMA
The total system bandwidth is divided into
channels which are allocated to the different
users.
Code Space
Time
Frequency
7C29822.030-Cimini-9/97
TDMA
Time is divided into slots which are allocated
to the different users.
Code Space
Time
Frequency
7C29822.031-Cimini-9/97
CDMA
Time and bandwidth are used simultaneously by
different users, modulated by orthogonal or semiorthogonal codes (e.g. spread spectrum).
Code Space
Time
Frequency
7C29822.032-Cimini-9/97
IMPLICATIONS FOR HIGH-SPEED
WIRELESS DATA
• Perform well with continuous stream traffic but
inefficient for bursty traffic
• Complexity
Frequency Division < Time Division < Code Division
• Multiple Data Rates
– multiple frequency bands
– multiple timeslots
– multiple codes
7C29822.034-Cimini-9/97
RANDOM ACCESS TECHNIQUES
• ALOHA
• Carrier-Sense Techniques
• Reservation Protocols
• Implication for High-Speed
Wireless Data
7C29822.038-Cimini-9/97
ALOHA
• Data is packetized.
• Retransmission is required when packets collide.
• Pure ALOHA
– send packet whenever data is available
– a collision occurs for any partial overlap of
packets
• Slotted ALOHA
S (Throughput per
Packet Time)
– send packets during predefined timeslots
– avoids partial overlap of packets
.40
Slotted Aloha
.30
.20
Pure Aloha
.10
0
0.5
1.0
1.5
2.0
3.0
G (Attempts per Packet TIme)
• Comments
– inefficient for heavily loaded systems
– capture effect improves efficiency
– combining SS with ALOHA reduces collisions
8C32810.39-Cimini-7/98
CARRIER-SENSE TECHNIQUES
• Channel is sensed before transmission to determine
if it is occupied.
• More efficient than ALOHA  fewer retransmissions
• Carrier sensing is often combined with collision
detection in wired networks (e.g., Ethernet).
not possible in a radio environment
Busy Tone
Wired Network
Wireless Network
• Collision avoidance is used in current wireless LANs.
(WaveLAN, IEEE802.11, Spectral Etiquette)
8C32810.40-Cimini-7/98
RESERVATION PROTOCOLS
• Demand–Based Assignment
– a common reservation channel is used to
assign bandwidth on demand
– reservation channel requires extra bandwidth
– very efficient if overhead traffic is a small
percentage of the message traffic
• Packet Reservation Multiple Access (PRMA)
– similar to reservation ALOHA
– uses a slotted channel structure
– all unreserved slots are open for contention
– a successful transmission in an unreserved
slot effectively reserves that slot for future
transmissions
7C29822.041-Cimini-9/97
EXAMPLES
• ARDIS
– slotted CSMA
• RAM Mobile Data
– slotted CSMA
• CDPD
– DSMA/CD - Digital Sense Multiple Access
– collisions detected at receiver and
transmitted back
• WaveLAN
– CSMA/CA
8C32810.126-Cimini-7/98
IMPLICATIONS FOR HIGH SPEED
WIRELESS DATA
• Retransmissions are power and spectrally
inefficient.
• ALOHA cannot satisfy high-speed data
throughput requirements.
• Reservation protocols are also ineffective
for short messaging.
• Delay constraints impose throughput
limitations.
7C29822.042-Cimini-9/97
FREQUENCY REUSE
BASE
STATION
• Frequencies (or time slots or codes) are reused at
spatially-separated locations.
• Introduces interference  system capacity is
interference-limited.
• Mainly designed for circuit-switched communications
• Base stations perform centralized control functions.
(call setup, handoff, routing, etc.)
8C32810.43-Cimini-7/98
DESIGN CONSIDERATIONS
• Reuse Distance (D)
– distance between cells using the same
frequency, time slot, or code
– smaller reuse distance packs more users
into a given area, but also increases their
co-channel interference
• Cell Radius
– decreasing the cell size increases system
capacity, but complicates the network
functions of handoff and routing
8C32810.44-Cimini-7/98
CHANNEL ASSIGNMENT
• Fixed Channel Assignment (FCA)
– each cell is assigned a fixed number
of channels
– channels used for both handoff and
new calls
• Reservation Channels with FCA
– each cell reserves some channels for
hand off calls
• Channel Borrowing
– a cell may borrow free channels from
neighboring cells
• Dynamic Channel Assignment
7C29822.67-Cimini-9/97
METHODS TO IMPROVE
SPECTRUM UTILIZATION
• Interference Averaging (CDMA)
• Interference Reduction
(power adaption, sectorization)
• Interference Cancellation
(smart antennas, multiuser detection)
• Interference Avoidance
(dynamic resource allocation)
8C32810.45-Cimini-7/98
Ad-Hoc Networks
•
•
•
•
•
Each node generates independent data.
Source-destination pairs are chosen at random.
Routing can be multihop.
Topology is dynamic
Generally a fully connected network with
different link SNRs
• Can allocate resources dynamically (rate, power,
BW, routes,…)
NETWORK ISSUES
• Network Architectures
• Mobility Management
• Network Reliability
• Internetworking
• Security
8C32810.53-Cimini-7/98
NETWORK ARCHITECTURES
• Hierarchical/Tree
• Star
• Ad-Hoc
• Implications for High-Speed Wireless Data
– single hop versus multiple hops
– static versus dynamic topology
– single points of failure
7C29822.054-Cimini-9/97
NETWORK CONTROL
• Centralized
– RAM Mobile Data
– CDPD
– Altair
• Distributed/Peer-to-Peer
– WaveLAN
• Implications for High-Speed Wireless Data
– less channel estimation required with
centralized control  increases efficiency
of packet transmission
– centralized control provides more efficient
resource management with setup-time overhead
– an extensive infrastructure is not required for
distributed control
7C29822.055-Cimini-9/97
MOBILITY MANAGEMENT
• Location Management
– identification and authentication
– home and visitor location data bases (cellular)
– discovery and registration (Mobile IP)
• Routing
– fixed data bases (connection-oriented)
– Mobile IP (connectionless)
– tree (virtual connection)
overhead and delay impact throughput
suboptimal (triangle) routing  delay
inefficiency and higher congestion
• Handoff
– transmissions may be delayed or dropped 
impacts higher layer protocols
– multi-homing  inefficient use of resources
8C32810.131-Cimini-7/98
NETWORK RELIABILITY
• End-to-End connection is composed of many
wireless/wired hops.
– widely varying data rates
– high BERs on some/all hops
– large, varying latencies
– user mobility causes hop characteristics
to vary
• Problem with reliability protocols like TCP.
– wireless losses mistaken for congestion
– bulk losses cause timeouts
– large round-trip time variances and
asymmetric channels
8C32810.58-Cimini-7/98
APPROACHES TO
NETWORK RELIABILITY
• Local (link-layer) solutions
– Forward error correction does not work well in fading
– ARQ introduces large latency
• End-to-end solutions
– Difficult to distinguish if packet loss due to congestion
or link quality
– Difficult to design for changing hop characteristics
End-to-end performance guarantees are difficult
to make
• Potential solutions
– Hierarchical/layered coding of voice/video/images
– Different Quality-of-Service classes
– Application awareness
– Local solution with end-to-end awareness
Requires interaction between all layers
8C32810.58a-Cimini-7/98
QUALITY OF SERVICE (QoS)
• Traffic dependent performance metrics required for
type of data transmitted
– bandwidth
– latency
– likelihood of packet (message) loss
• Categories
– guaranteed
– predictive
– best effort
• Implications for high speed wireless data
– QoS performance generally based on switched,
fiber-optic, wired networks
– wireless links have high Pb and high latency due
to link layer retransmission and unpredictable
link bandwidths
– QoS guarantees and predictions are difficult to
make for wireless networks  it is not clear
that the best effort is good enough for most
applications
7C29822.059-Cimini-9/97
INTERNETWORKING
• TCP/IP
– Compatible with existing wired networks
– Works well over large range of wired subnet
performance
– TCP has problems operating over wireless links
• Wireless ATM
– ATM is emerging standard for multimedia
transmission over wired networks
– ATM protocol based on links with 10-10 BER
and Mbps/Gbps data rates
• high overhead in packet structure
• QOS guarantees
– Not clear that ATM protocol can be modified
for wireless links
7C29822.68-Cimini-9/97
STANDARDS AND FUTURE SYSTEMS
• Bluetooth
• Wireless LANs
• High-Speed Digital Cellular (3G)
• 4G Cellular
• Wireless "Cable"
– Multichannel Multipoint Distribution
Service (2.2 GHz)
– Local Multipoint Distribution
Service (28 GHz)
• Satellite Networks
- Iridium, Globalstar, Others
• HomeRF
8C32810.61-Cimini-7/98
BLUETOOTH
• Cable replacement RF technology
• Short range (10 meters)
• 2.4 GHz band
• 1 Data (700 Kbps) and 3 Voice channels
• Supported by over 200 telecommunications
and computer companies
8C32810.61-Cimini-7/98
802.11 Wireless LANs
• 802.11b: standard for 2.4 GHz ISM band
• Frequency hopped spread spectrum
• 1.6 Mbps data rates, 500 foot range
• Star or peer-to-peer architecture
• 802.11a extends rates to 10-70 Mbps
• Extensions trying to add QoS
8C32810.63b-Cimini-7/98
HIPERLAN
• Types 1-4 for different user types
- Frequency bands: 5.15-5.3 GHz, 17.117.3 GHz
• Type 1
- 5.15-5.3 GHz band
- 23 Mbps, 20 MHz Channels
- 150 foot range (local access only)
- Protocol support similar to 802.11
- Peer to peer architecture
- ALOHA channel access
• Types 2-3
- Wireless ATM
- Local access and wide area services
- Standard under development
- Two components: access and
mobility support
8C32810.63a-Cimini-7/98
HIGH-SPEED DIGITAL CELLULAR
• North American Digital Cellular
– CDMA (IS-95) enhancements
– TDMA (IS-136) enhancements
– IS-136+
 32-64 kbps
– IS-136HS  384 kbps
• GSM
– General Packet Radio System (GPRS)
– Enhanced Data Rates for GSM Evolution
(EDGE)
8C32810.62-Cimini-7/98
EDGE
• Evolution of GSM / GPRS
• ETSI standardization as GSM evolution
chosen for data services for IS136HS
• Higher-level modulation (adaptive)
• 200 kHz carrier spacing
• Up to 384 kbps in 200 kHz
8C32810.137ppt-Cimini-7/98
WIDEBAND CDMA (3G)
• The W-CDMA concept:
– 4.096 Mcps Direct Sequence CDMA
– Variable spreading and multicode operation
– Coherent in both up-and downlink
= Codes with different spreading,
giving 8-500 kbps
P
...
.
f
4.4-5
MHz
t
High rate
multicode user
Variable rate users
8C32810.138ppt-Cimini-7/98
10 ms frame
W-CDMA
KEY TECHNICAL FEATURES
• High bit-rate services require wideband
• Flexibility for different services
• Optimized for packet data transfer
• Capacity and coverage gain from frequency
diversity
• Built in support for
– adaptive antenna arrays
– multi-user detection
– hierarchical cell structures
– transmitter diversity
• Low infrastructure cost (many users/
transceiver)
• BS synchronization not required
8C32810.139ppt-Cimini-7/98
SUMMARY
• The desire for mobility coupled with the demand
for Internet and multimedia services indicate a
bright future for wireless data.
• Current products and services have unsatisfactory
performance for high-speed wireless data applications.
• The inherent limitations of the radio channel can be
significantly reduced using signal processing and
architectural techniques, at the expense of cost
and complexity.
• The network-level design must take into account
the physical layer limitations of the wireless channel,
as well as the impact of user mobility.
8C32810.65-Cimini-7/98