Download - Andrea Goldsmith
-
7/31/2019 Andrea Goldsmith
1/99
Wireless Communications:Trends and Challenges
ANDREA J. GOLDSMITH
Dept. of Electrical EngineeringStanford Universityhttp://ee.stanford.edu/~andrea
810.1-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
2/99
-
7/31/2019 Andrea Goldsmith
3/99810.35-Cimini-7/98
[R. Katz, "Does Wireless Data Have a Future?", Plenary Talk, INFOCOM '96]
Seamless Multimedia Networks
with Mobility and Freedom from Tethers
WIRELESS DATA VISION
TAXI
Region
Campus
City
In-Building
laptops, PDAs
-
7/31/2019 Andrea Goldsmith
4/99
VOICE VERSUS DATA VERSUS VIDEO
810.80-Cimini-7/98
Wired Networks Trying to Integrate
(ATM, SONET, Multimedia Services)
Voice Data Video
Delay < 100 ms < 100 ms
Packet Loss < 1% 0
-
7/31/2019 Andrea Goldsmith
5/99
WHAT IS THE FUTUREOF WIRELESS DATA?
USA market
1995 20000
10
20
30
40
50
60
70
80
90
100
Internet users
paging subs
laptop users
cellular + PCS subs
annual laptop sales
dedicated wirelessdata subs
822.008-Cimini-9/97
*Estimates as of 1996
millions
-
7/31/2019 Andrea Goldsmith
6/99
THE ISSUE IS PERFORMANCE
810.81-Cimini-7/98
"The mobile data market has been slow to take off,but progress is being made. The most formidableobstacle to user acceptance remains performance."
I. Brodsky, "Countdown to Mobile Blast Off",
Network World, February 19, 1996
Link Performance: Data Rate and Quality
Network Performance: Access, Coverage, Reliability,QoS, and Internetworking
RadioPort
NetworkInterface
Signaling/RoutingMobilityControl
MobilityControl
Protocols
Radio Link
MobileMultimedia
Terminal
WirelessInterface
RadioProtocols& Modem
RadioProtocols& Modem
NetworkAdaptation& Control
RADIO ACCESS SEGMENT MOBILE NETWORK SEGMENT
BROADBAND WIRELESSNETWORK
-
7/31/2019 Andrea Goldsmith
7/99
GAP BETWEEN WIRED AND
WIRELESS NETWORK CAPABILITIES
10.79-Cimini-7/98
WIDE AREA CIRCUIT SWITCHING
User
Bit-Rate
(kbps)
14.4digitalcellular
28.8 modem
ISDN
ATM
9.6 modem
2.4 modem2.4 cellular
32 kb
PCS
9.6 cellular
wired- wireles
bit-rate "gap"
1970 20019901980YEAR
LOCAL AREA PACKET SWITCHING
User
Bit-Rate
(kbps)
Ethernet
FDDI
ATM100 M
Ethernet
Polling
Packet
Radio
1st gen
WLAN
2nd
gen
WLAN
wired- wirelessbit-rate "gap"
1970 200019901980.01
.1
1
10
100
1000
10,000
00,000
YEAR
.01
.1
1
10
100
1000
10,000
100,000
-
7/31/2019 Andrea Goldsmith
8/99
-
7/31/2019 Andrea Goldsmith
9/99
RADIO ENVIRONMENT
810.83-Cimini-7/98
Path Loss Shadow Fading Multipath
Limit the Bit Rateand/or Coverage
-
7/31/2019 Andrea Goldsmith
10/99
where Pr is the local mean received signal
PATH LOSS MODEL
Different, often complicated, models areused for different environments.
A simple model for path loss, L, is
The path loss exponent = 2 in free space;2 4 in typical environments.
power, Pt is the transmitted power, d is thetransmitter-receiver distance, f is frequency,
and K is a transmission constant.
822.011-Cimini-9/97
Pr 1
Pt f2d
= KL =
-
7/31/2019 Andrea Goldsmith
11/99
SHADOW FADING
810.84-Cimini-7/98
The received signal is shadowed byobstructions such as hills and buildings.
This results in variations in the local meanreceived signal power,
Implications
nonuniform coverage increases the required transmit power
Pr (dB) = Pr (dB) + Gs
where Gs ~ N(0, s ), 4 s 10 dB.2
R P = Pr0
-
7/31/2019 Andrea Goldsmith
12/99
MULTIPATH
Constructive and Destructive Interferenceof Arriving Rays
Received
Power
Delay Spreadt
dB With Respectto RMS Value
0 0.5
0.5
1.5
-30-20
-10
10
0
1t, in seconds
0 10 3020x, in wavelength
810.85-Cimini-7/98
h(t) = aiej id(t-ti)Si
-
7/31/2019 Andrea Goldsmith
13/99
DELAY SPREADTIME DOMAIN INTERPRETATION
large
T
smallT
0
11
T 2T
Channel Input
Channel Output
0 T 2T
0 T 2T
810.88-Cimini-7/98
Two-ray model= rms delay spread
2
Delay
Received
Power
T small negligible intersymbol interference
large significant intersymbol interference,which causes an irreducible error floorT
-
7/31/2019 Andrea Goldsmith
14/99
PHYSICAL LAYER ISSUES
Link Performance Measures
Modulation Tradeoffs
Flat Fading Countermeasures
Delay Spread Countermeasures
810.91-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
15/99
LINK PERFORMANCE MEASURESPROBABILITY OF BIT ERROR
810.92-Cimini-7/98
The probability of bit error, Pb, in a radioenvironment is a random variable.
Typically only one of these measures isuseful, depending on the Doppler frequency
and the bit rate.
average Pb, Pb
Pr [Pb > Pbtarget] outage, Pout=
-
7/31/2019 Andrea Goldsmith
16/99
LINK PERFORMANCE MEASURESEFFICIENCY
810.72-Cimini-7/98
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 andbandwidth
Throughput/Delay
-
7/31/2019 Andrea Goldsmith
17/99
GOALS OFMODULATION TECHNIQUES
810.74-Cimini-7/98
High Bit Rate
High Spectral Efficiency
High Power Efficiency
Low-Cost/Low-Power Implementation
Robustness to Impairments
-
7/31/2019 Andrea Goldsmith
18/99
DIGITAL MODULATION
810.94-Cimini-7/98
Any modulated signal can be represented as
Linear versus nonlinear modulation impact
on spectral efficiency
Constant envelope versus non-constantenvelope hardware implications with impact
on power efficiency
s(t) = A(t) cos [ ct + (t)]
s(t) = A(t) cos (t) cos ct
- A(t) sin (t) sin ct
amplitude
in-phase
quadrature
phase or frequency
-
7/31/2019 Andrea Goldsmith
19/99
LINEAR MODULATION TECHNIQUES
810.95-Cimini-7/98
s(t) = [ an g (t-nT)]cos ct - [ bn g (t-nT)] sin ct
I(t), in-phase Q(t), quadrature
LINEAR MODULATIONS
CONVENTIONAL
4-PSK
(QPSK)
OFFSET
4-PSK
(OQPSK)
DIFFERENTIAL
4-PSK
(DQPSK, /4-DQPSK)
M-ARY QUADRATURE
AMPLITUDE MOD.
(M-QAM)
M-ARY PHASE
SHIFT KEYING
(M-PSK)
M 4 M 4M=4(4-QAM = 4-PSK)
Square
onstellations
Circular
Constellation
n n
-
7/31/2019 Andrea Goldsmith
20/99
SIGNAL CONSTELLATIONS
Tradeoffs
Higher-order modulations (M large) are more spectrallyefficient but less power efficient.
M-QAM is more spectrally efficient than M-PSK butalso more sensitive to system nonlinearities.
M-QAM (Square Constellations)
16-QAM
4-PSK
an
bn
M-PSK (Circular Constellations)
16-PSK
an
bn4-PSK
810.96-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
21/99
PULSE SHAPING
810.75-Cimini-7/98
Rectangular pulses are spectrally inefficient
pulse shaping
intersymbol interference (ISI)
non-constant envelope
Nyquist pulses
-
7/31/2019 Andrea Goldsmith
22/99
RAISED COSINE PULSE SHAPING
G(f)
12T
0
= 1
= 0
= 0.5
f12T
-1T
- 1T
810.97-Cimini-7/98
30
256 QAM
64 QAM
16 PSK16 QAM
8 PSK
4 PSK
20
10
0 0.25 0.750.5 1.0
Cosine Rolloff Factor,
RelativeP
eak
InstantaneousP
ower(dB)
g(t)
0 4T3T
= 1
= 0 = 0.5
= 0, 0.5
tT 2T
-
7/31/2019 Andrea Goldsmith
23/99
DEMODULATION
for the reference signal.
reference.
Coherent detection requires a coherent phase
difficult to obtain in a rapidly fading
environment
increases receiver complexity
Differential detection uses the previous symbol
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
810.133-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
24/99
FREQUENCY SHIFT KEYING
810.98-Cimini-7/98
Continuous Phase FSK (CPFSK)digital data encoded in the frequency shift
typically implemented with frequency
modulator to maintain continuous phase
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
s(t) = A cos [ ct + 2 kf
d( ) d ]
t
-
7/31/2019 Andrea Goldsmith
25/99
SPECTRAL CHARACTERISTICS
822.013-Cimini-9/97
QPSK/DQPSKGMSK
(MSK)
1.0
1.0 1.5 2.0 2.50.50
-120
-100
-80
-60
-40
-20
0
10
0.25
B3-dBTb = 0.16
Normalized Frequency (f-fc)Tb
PowerSpectralDensity(dB)
-
7/31/2019 Andrea Goldsmith
26/99
BIT ERROR PROBABILITYAWGN CHANNEL
QPSK is more spectrally efficient than BPSK with thesame performance.
M-PSK, for M>4, is more spectrally efficient but requiresmore SNR per bit.
There is ~3 dB power penalty for differentialdetection.
810.99-Cimini-7/98
Pb
BPSK, QPSK
DBPSK
DQPSK
010-6
2
5
2
5
2
5
2
5
2
5
10-1
10-3
10-4
10-5
10-2
2 4 6 8 10 12 14
For Pb = 10-3
BPSK 6.5 dB
QPSK 6.5 dB
DBPSK ~8 dB
DQPSK ~9 dB
b, SNR/bit, dB
-
7/31/2019 Andrea Goldsmith
27/99
BIT ERROR PROBABILITYFADING CHANNEL
Pb is inversely proportion to the average SNR per bit.
Transmission in a fading environment requires about18 dB more power for Pb = 10
-3.
0.100-Cimini-7/98
DBPSK
AWGN
BPSK
010-5
2
5
2
5
2
5
2
5
2
5
1
10-2
10-3
10-4
10-1
5 10 15 20 25 30 35
Pb
b, SNR/bit, dB
-
7/31/2019 Andrea Goldsmith
28/99
BIT ERROR PROBABILITYEFFECTS OF DOPPLER SPREAD
Doppler causes an irreducible error floor when differentialdetection is used decorrelation of reference signal.
The implication is that Doppler is not an issue for high-speed
wireless data.
data rate T Pbfloor
10 kbps 10-4s 3x10-4
100 kbps 10-5s 3x10-6
1 Mbps 10-6s 3x10-8
The irreducible Pb depends on the data rate and the Doppler.For fD = 80 Hz,
0
0 60504030201010-6
10-5
10-4
10-3
10-2
10-1
10
DQPSK
Rayleigh Fading
No Fading
fDT=0.003
0.002
0.001
0
QPSK
Pb
b, SNR/bit, dB
[M. D. Yacoub, Foundations of Mobile Radio Engineering, CRC Press, 1993]
10.101-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
29/99
BIT ERROR PROBABILITYEFFECTS OF DELAY SPREAD
810.102-Cimini-7/98
The rms delay spread imposes a limit on the maximum bit ratein a multipath environment.For example, for QPSK,
ISI causes an irreducible error floor.
Maximum Bit RateMobile (rural) 25 sec 8 kbpsMobile (city) 2.5 sec 80 kbpsMicrocells 500 nsec 400 kbpsLarge Building 100 nsec 2 Mbps
[J. C.-I. Chuang, "The Effects of Time Delay Spread on Portable RadioCommunications Channels with Digital Modulation," IEEE JSAC, June 1987]
+
x
+
+
+
+
+
x
x
x
x
x
10-210-4
10-3
10-2
10-1
10-1 100
BPSKQPSKOQPSKMSK
Modulation
Coherent Detection
IrreducibleP
b
T=
rms delay spreadsymbol period
-
7/31/2019 Andrea Goldsmith
30/99
SUMMARY OFMODULATION 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
10.103-Cimini-7/98
delay spread
-
7/31/2019 Andrea Goldsmith
31/99
HOW DO WE OVERCOME THELIMITATIONS 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
10.104-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
32/99
DIVERSITY
Independent signal paths have a low probabilityof experiencing deep fades simultaneously.
The basic concept is to send the sameinformation over independently fading radio
Independent fading paths can be achieved byseparating the signal in time, frequency, space,polarization, etc.
10.105-Cimini-7/98
The chance that two deep fadesoccur simultaneously is rare.
0
0
-20
-40
-60
-80
-1004 8 12 16 d
ReceivedSignalPower
(dBm)
-
7/31/2019 Andrea Goldsmith
33/99
DIVERSITY COMBINING TECHNIQUES
Selection Combining: picks the branch with thehighest SNR.
Equal-Gain Combining: all branches are coherentlycombined with equal weights.
Maximal-Ratio Combining: all branches are coherentlycombined with weights which depend on the branchSNR.
810.106-Cimini-7/98
Combiner
Output
1 2 3 M
-
7/31/2019 Andrea Goldsmith
34/99
DIVERSITY PERFORMANCE
The output SNR with Maximal-Ratio Combining improveslinearly with the number of diversity branches, M thecomplexity becomes prohibitive.
There is dramatic improvement even with two-branchselection combining.
10 dB reduction in required SNR for 1% outage less transmitted power or higher bit rates or largercoverage area
5
2
10-6
10-5
5
2
10-4
5
2
10-3
5
2
10-2
5
2
10-1
5
10 15 20 25 30 35 40
Pb
b, SNR/bit, dB
M = 4
M = 2
M = 1
Maximal
RatioCombining
( )1margin
Pout
-400.01
0.02
0.05
0.1
0.2
0.5
1.0
2.0
5.0
10.0
20.030.040.050.060.0
99.99
99.999.598.0
90.080.070.0
-30 -20 -10 0 10
MaximalRatio
EqualGain
Selection
10log
M = 2
22.014-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
35/99
CHANNEL CODING
Channel coding reduces Pb by introducing redundancy
in the transmitted bit stream.
Block and convolutional codes acheive this improvementat the expense of increased signal bandwidth or a lowerdata rate.
Bit error probabilityAWGN channel
822.015-Cimini-9/97
Fading causes burst errors. If the fading is slow enoughrelative to the symbol rate, coding will not be effective.
For Pb = 10-6
Uncoded 10.5 dB
Hamming 10.0 dBBCH 6.5 dB
Conv. 5.0 dBPb
010-7
10-2
10-4
10-5
10-6
10-3
2 4 6 8 10 12 14
5
2
5
2
5
2
5
2
5
2
Uncoded
Hamming(7,4,1)
BCH(127,64,10)
Conv.1/2 rate
(k=7)
BPSK
b, SNR/bit, dB
-
7/31/2019 Andrea Goldsmith
36/99
CODING PERFORMANCEFADING CHANNEL
2810.17-Cimini-7/98
Pb performance for the IS-136 rate-1/2 convolutionalcode on a simulated mobile radio channel (hard-
decision decoding).
Negligible coding gain if fading is slow comparedto bit rate interleaving
Iyengar and J. Michaelides, "Performance Evaluations of RLPs (Radio Link Protocols)
r TDMA Data Services," ITIA Contribution TR45.3.2.5/93.03.30.10, Chicago, March 30, 199
1
10-1
10-2
10-3
10-4
Pb
8 10 12 14 16 18 20
Uncoded50 km/hr
Coded1 km/hr
Coded8 km/hr
Coded50 km/hr
Coded100 km/hr
b, SNR/bit, dB
-
7/31/2019 Andrea Goldsmith
37/99
CODING PERFORMANCEFADING CHANNEL
Pb performance for the IS-136 rate-1/2 convolutionalcode on a simulated mobile radio channel (softdecision decoding).
810.18-Cimini-7/98
Iyengar and J. Michaelides, "Performance Evaluations of RLPs (Radio Link Protocols)
r TDMA Data Services," ITIA Contribution TR45.3.2.5/93.03.30.10, Chicago, March 30, 199
1
Uncoded50 km/hr
Coded1 km/hr
Coded8 km/hr
Coded50 km/hr
Coded100 km/hr
10-1
10-2
10
-3
10-4
10-5
Pb
8 10 12 14 16 18 20
b, SNR/bit, dB
-
7/31/2019 Andrea Goldsmith
38/99
-
7/31/2019 Andrea Goldsmith
39/99
Trellis Codes reduce Pb without bandwidth expansion
through joint design of the channel codeand 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
ADVANCED CODING TECHNIQUES
810.19-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
40/99
CODING PERFORMANCE TCM
b1,R=2/3
10-1
10-2
10-3
10-4
10-5
10-6
10 12 14 16 18 20 22
Es/N0 (dB)
Pb
Uncoded4 PSK
UngerboeckCode
R=2/3, M=4
LSB
b2,R=2/3
b3,
R=2/3
MSB
8PSK TCM
810.69-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
41/99
Adaptive Modulation
Automatic Repeat Request
ADAPTIVE TECHNIQUES
810.21-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
42/99
Power and/or data rate adapted at transmitter tochannel conditions
Potential for large increase in spectral efficiency
Can be combined with adaptive compression
requires reliable feedback channel and accuratechannel estimation
increases transmitter and receiver complexity
810.22-Cimini-7/98
ADAPTIVE MODULATION
AdaptiveModulationand Coding
PowerControl
Demodulationand Decoding
ChannelEstimate+
Delay
TRANSMITTER RECEIVER
FEEDBACK CHANNEL
noise
Channel
-
7/31/2019 Andrea Goldsmith
43/99
Method of "self-adapting" the data rate tothe channel conditions
Used in combination with error-detecting code
Variations of ARQ used in Mobitex and CDPD
Types: Stop-and-Wait, Go-Back-N, Selective-
Repeat
22.023-Cimini-9/97
power and spectrally inefficient
impacts higher layer protocols
necessary for meeting stringent Pb
requirements or data
AUTOMATIC REPEAT REQUEST (ARQ)
-
7/31/2019 Andrea Goldsmith
44/99
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 (distributedantenna system, small cells, directive
antennas)
22.024-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
45/99
EQUALIZER TYPES AND STRUCTURES
The goal of equalization is to cancel the ISI
or, equivalently, to flatten the frequency response.
[J. G. Proakis, "Adaptive Equalization for TDMA Digital Mobile Radio,"IEEE Trans. on Veh. Tech. , May 1991]
Equalizer
Nonlinear
ML SymbolDetector
DFELinear MLSE
TransversalTransversal
ChannelEstimator
Types
Structures
LatticeTransversal Lattice
10.107-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
46/99
LINEAR EQUALIZER
Equalizer
Heq(f)1
Hc(f)
Channel
Hc(f)
n(t)
10.108-Cimini-7/98
A linear equalizer effectively inverts the channel.
The linear equalizer is usually implemented as atapped delay line.
On a channel with deep spectral nulls, this equalizerenhances the noise.
poor performance on frequency-selective
fading channels
-
7/31/2019 Andrea Goldsmith
47/99
DECISION FEEDBACK EQUALIZER
The DFE determines the ISI from the previously detectedsymbols and subtracts it from the incoming symbols.
This equalizer does not suffer from noise enhancementbecause 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 aremade incorrectly.
Decisions are made on coded symbols. no coding gain
822.025-Cimini-9/97
Hc(f)Forward
Filter
n(t)
x(t)
DFE
Feedback
Filter
+
-
x(t)^
-
7/31/2019 Andrea Goldsmith
48/99
MAXIMUM LIKELIHOODSEQUENCE ESTIMATION
10.109-Cimini-7/98
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.
-
7/31/2019 Andrea Goldsmith
49/99
EQUALIZER ISSUES FORHIGH-SPEED WIRELESS DATA
The number of required equalizer taps, N, is proportionalto the delay spread.
The equalizer taps must be adapted at the highestDoppler rate.
The length and periodicity of the training sequence
impacts the spectral efficiency.
There is a tradeoff between speed of convergence
and complexity.
Number ofAlgorithms Multiply Convergence Advantages Disadvantages
(for DFE) Operations
Least Mean 2N + 1 ~10-100N Low computational Slow convergence,Square (LMS) complexity depends on
channel
Kalman 2.5N2
+ 4.5N ~N Fast convergence, HighRecursive Least good tracking ability computational
Squares (RLS) complexity
Square Root 1.5N2
+ 6.5N ~N Better stability Highthan Kalman computational
complexity
Fast Kalman 20N + 5 ~N Fast convergence Could beand good tracking unstable
22.026-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
50/99
EQUALIZER PERFORMANCE
Pahlavan has shown that, for 30-meter cells ( = 50 ns), 20 Mb/scan be achieved using a DFE with 3 forward taps and 3 feedback taps.
[K. Pahlavan, S. J. Howard, and T. A. Sexton, "Decision Feedback Equalizationof the Indoor Radio Channel," IEEE Trans. on Commun., January 1993]
810.110-Cimini-7/98
BPSK
25
10-6
10-5
10-4
10-3
10-2
10-1
1
30 35 40 45 50
no equalizerDFE
10 Mbps
5
110
.1
.15
1
SNR (dB)
Pb
Target Pb
BPSK
1
10-4
10-3
10-1210-810-4
10-2
10-1
1
no equalizerDFE
16 Mbps
4,16
8
1
1
8
.1
.1,4
Pout
-
7/31/2019 Andrea Goldsmith
51/99
MULTICARRIER MODULATION
The transmission bandwidth is divided into manynarrow subchannels which are transmitted inparallel.
Ideally, each subchannel is narrow enough sothat the fading it experiences is flat no ISI.
10.111-Cimini-7/98
RF
Transmitter
R/N b/s
D(t)
f0
f1
fN-1
dN-1
(t)
QAM filter
R/N b/sQAM filter
R/N b/s QAM filter
(t)d0
(t)d1
Bandlimitedsignals
f0 f1 f2
f0
fN-1
f1
N-1
Receiver
RF
filter
QAM
QAM
QAM
filterf1
f0
filterfN-1
-
7/31/2019 Andrea Goldsmith
52/99
OFDM RECEIVER STRUCTURE
Subchannel Separation choose fn = f0 + n f, with f =
integrate over NT, then d(m) = d(m)
1
NT
^
Efficient FFT Implementation
A guard interval can virtually eliminate ISI(or, interblock interference) lower spectral
or power efficiency.
10.113-Cimini-7/98
parallelto
serialconverter
QAM
f0
f1
fN-1
d(0)Receiver
d(1)
d(N-1)
RF
-
7/31/2019 Andrea Goldsmith
53/99
WHAT TO DO WITHBAD SUBCHANNELS?
Coding Across Subchannels works bestwith large delay spread
Frequency Equalization requires accuratechannel estimation
Adaptive Loading requires reliablefeedback channel and accurate channel
estimation
10.114-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
54/99
MULTICARRIER MODULATIONISSUES 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
10.115-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
55/99
CURRENT AND PROPOSEDAPPLICATIONS OF OFDM
Asymmetric Digital Subscriber Line
Digital Audio Broadcasting
Wireless LAN
Digital Terrestrial Television
High Speed Cellular
10.116-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
56/99
SPREAD SPECTRUM
10.117-Cimini-7/98
Spread spectrum increases the transmit signalbandwidth 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 andfrequency hopping.
Direct sequence multiplies the data sequence
by a faster chip sequence.
Frequency hopping varies the carrier
frequency by the same chip sequence.
-
7/31/2019 Andrea Goldsmith
57/99
DIRECT SEQUENCESPREAD SPECTRUM
Modulator
Transmitter
Channel
Spreading(PN) Code
Tc
-
7/31/2019 Andrea Goldsmith
58/99
RAKE RECEIVER
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.
10.119-Cimini-7/98
CoherentCombiner Demodulator
DataOutputReceivedSignal
sc(t)
sc(t-Tc)
sc(t-2Tc)
sc(t-TM)
-
7/31/2019 Andrea Goldsmith
59/99
PERFORMANCE OF RAKE RECEIVERFADING CHANNEL
810.27-Cimini-7/98
0.5
10-1
10-2
10
-3
10-4
10-5
Pb
0 5 10 15
b, SNR/bit, dB
Rayleigh
DPSK
AWGN
RAKE
-
7/31/2019 Andrea Goldsmith
60/99
SPREAD SPECTRUM ISSUESFOR HIGH-SPEED WIRELESS DATA
810.120-Cimini-7/98
Hardware Complexity synchronization high processing speeds for high
bit rates RAKE receiver
High Required Bandwidth to AccommodateSpreading
Spread spectrum is difficult athigh bit rates and not reallyneeded.
-
7/31/2019 Andrea Goldsmith
61/99
Goal: Reduce (or eliminate) delay spread
Distributed Antenna System
Very Small Cells antenna in every room
Sectorization
Directive Antennas/Beam Steering
Omnidirectional Sectorized Directive
90
270
180
150
120
30
300240
210 330
60
0
90
270
180
150
120
30
300240
210 330
60
0
90
270
180
150
120
30
300240
210 330
60
0
ANTENNA SOLUTIONS
22.028-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
62/99
DISTRIBUTED ANTENNA SYSTEM
00
10
.5
1
20 30 40 50
RMS Delay Spread (ns)
Distributed
Monopoles
CentralMonopole
Pr
obabilityAbscissa
Exceeded
[A. A. M. Saleh, A. J. Rustako, Jr., and R. S. Roman, "Distributed Antennas
for Indoor Radio Communications," IEEE Trans. on Commun., December 1987]
22.029-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
63/99
EXAMPLES OFPERFORMANCE IMPROVEMENTS
810.121-Cimini-7/98
High-Speed Narrowbeam Antenna Experiment[P. F. Driessen "Gigabit/s Indoor Wireless Systems withDirectional Antennas," IEEE Trans. on Comm., August 1996]
directional antennas (15 beamwidth) at both ends ofLOS link
no equalization 622 Mbps BPSK transmission without errors
Sectored Antennas [G. Yang and K. Pahlavan, "ComparativePerformance 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
Six Sector Antennas
Target Pb
Pout
30 Mbps
20 Mbps
10 Mbps
10010-1 10-210-3 10-410-510-610-710-810-9 10-10-1210-1110-10
.0005
.002
.001
.005.01
.02
.0001
.001
.01
.1
1
Omnidirectional Antennas
out
30 Mbps20 Mbps
10 Mbps
5 Mbps3 Mbps2 Mbps
1 Mbps
Target Pb
10010-1 10-210-3 10-410-510-610-710-810-9 10-1310-1210-1110-10
-
7/31/2019 Andrea Goldsmith
64/99
SUMMARY OF COUNTERMEASURES
Diversity
Coding and Interleaving
Adaptive Techniques
Equalization
Multicarrier
Spread Spectrum
Antenna Solutions
These techniques can be combined.
10.123-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
65/99
COMBINED EQUALIZATION ANDSECTORED ANTENNAS
[G. Yang and K. Pahlavan, "Comparative Performance Evaluation of Sector
1
.1
20 40 60
.01
.001
.0001
Square room length (meter)
Pout
Omni
Omni+DFE
Sector
Sec+DFE
Pt = 100 mWRb = 20 Mbps
1
.1
20100 30 40 50
.01
.001
.0001
Rb (Mbps)
Pout
Omni
Omni+DFE
Sector
Sec+DFE
30mx30m
Antenna and DFE Systems in Indoor Radio Channels," Proc. of ICC '92]
10.122-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
66/99
CHANNEL ACCESS ISSUES
Multiple Access
Random Access
Frequency Reuse
810.125-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
67/99
MULTIPLE ACCESS TECHNIQUES
Frequency Division (FDMA)
Time Division (TDMA)
Code Division (CDMA)
Hybrid Approaches
22.033-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
68/99
FDMA
The total system bandwidth is divided intochannels which are allocated to the differentusers.
22.030-Cimini-9/97
Code Space
Time
Frequency
-
7/31/2019 Andrea Goldsmith
69/99
TDMA
Time is divided into slots which are allocatedto the different users.
22.031-Cimini-9/97
Code Space
Time
Frequency
-
7/31/2019 Andrea Goldsmith
70/99
CDMA
Time and bandwidth are used simultaneously bydifferent users, modulated by orthogonal or semi-orthogonal codes (e.g. spread spectrum).
22.032-Cimini-9/97
Code Space
Time
Frequency
-
7/31/2019 Andrea Goldsmith
71/99
IMPLICATIONS FOR HIGH-SPEEDWIRELESS DATA
Perform well with continuous stream traffic butinefficient for bursty traffic
ComplexityFrequency Division < Time Division < Code Division
Multiple Data Rates multiple frequency bands
multiple timeslots
multiple codes
22.034-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
72/99
ALOHA
Carrier-Sense Techniques
Reservation Protocols
Implication for High-SpeedWireless Data
RANDOM ACCESS TECHNIQUES
822.038-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
73/99
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 send packets during predefined timeslots avoids partial overlap of packets
Comments inefficient for heavily loaded systems capture effect improves efficiency combining SS with ALOHA reduces collisions
810.39-Cimini-7/98
.40
.30
.20
.10
0 0.5 1.0 1.5 2.0 3.0
G (Attempts per Packet TIme)
S(Throughputper
PacketTime
)
Slotted Aloha
Pure Aloha
-
7/31/2019 Andrea Goldsmith
74/99
CARRIER-SENSE TECHNIQUES
Channel is sensed before transmission to determineif it is occupied.
More efficient than ALOHA fewer retransmissions
Carrier sensing is often combined with collisiondetection in wired networks (e.g., Ethernet).not possible in a radio environment
Collision avoidance is used in current wireless LANs.(WaveLAN, IEEE802.11, Spectral Etiquette)
810.40-Cimini-7/98
Wired Network
Busy Tone
Wireless Network
-
7/31/2019 Andrea Goldsmith
75/99
DemandBased 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 futuretransmissions
RESERVATION PROTOCOLS
822.041-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
76/99
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
810.126-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
77/99
Retransmissions are power and spectrallyinefficient.
ALOHA cannot satisfy high-speed data
throughput requirements.
Reservation protocols are also ineffectivefor short messaging.
Delay constraints impose throughputlimitations.
IMPLICATIONS FOR HIGH SPEEDWIRELESS DATA
822.042-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
78/99
-
7/31/2019 Andrea Goldsmith
79/99
Reuse Distance (D)distance between cells using the same
frequency, time slot, or code
smaller reuse distance packs more usersinto a given area, but also increases their
co-channel interference
Cell Radiusdecreasing the cell size increases system
capacity, but complicates the networkfunctions of handoff and routing
810.44-Cimini-7/98
DESIGN CONSIDERATIONS
-
7/31/2019 Andrea Goldsmith
80/99
CHANNEL ASSIGNMENT
Fixed Channel Assignment (FCA) each cell is assigned a fixed number
of channels
channels used for both handoff andnew calls
Reservation Channels with FCA each cell reserves some channels for
hand off calls
Channel Borrowing a cell may borrow free channels fromneighboring cells
Dynamic Channel Assignment
822.67-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
81/99
Interference Averaging (CDMA)
Interference Reduction
(power adaption, sectorization)
Interference Cancellation(smart antennas, multiuser detection)
Interference Avoidance(dynamic resource allocation)
METHODS TO IMPROVESPECTRUM UTILIZATION
810.45-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
82/99
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,)
-
7/31/2019 Andrea Goldsmith
83/99
NETWORK ISSUES
Network Architectures
Mobility Management
Network Reliability
Internetworking
Security
810.53-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
84/99
NETWORK ARCHITECTURES
22.054-Cimini-9/97
Implications for High-Speed Wireless Data single hop versus multiple hops
static versus dynamic topology
single points of failure
Hierarchical/Tree
Star
Ad-Hoc
-
7/31/2019 Andrea Goldsmith
85/99
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
22.055-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
86/99
MOBILITY MANAGEMENT
810.131-Cimini-7/98
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)
Handoff
transmissions may be delayed or dropped impacts higher layer protocols
multi-homing inefficient use of resources
overhead and delay impact throughput
suboptimal (triangle) routing delay
inefficiency and higher congestion
-
7/31/2019 Andrea Goldsmith
87/99
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
810.58-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
88/99
APPROACHES TONETWORK RELIABILITY
10.58a-Cimini-7/98
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 difficultto 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
-
7/31/2019 Andrea Goldsmith
89/99
QUALITY OF SERVICE (QoS)
Traffic dependent performance metrics required fortype 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
822.059-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
90/99
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
822.68-Cimini-9/97
-
7/31/2019 Andrea Goldsmith
91/99
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
810.61-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
92/99
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
810.61-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
93/99
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
810.63b-Cimini-7/98
802.11 Wireless LANs
-
7/31/2019 Andrea Goldsmith
94/99
HIPERLAN
Types 1-4 for different user types- Frequency bands: 5.15-5.3 GHz, 17.1-17.3 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 andmobility support
810.63a-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
95/99
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)
810.62-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
96/99
-
7/31/2019 Andrea Goldsmith
97/99
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
f
t
10 ms frame
4.4-5MHz
High ratemulticode user
Variable rate users
...
.P
810.138ppt-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
98/99
W-CDMAKEY TECHNICAL FEATURES
High bit-rate services require wideband
Flexibility for different services
Optimized for packet data transfer
Capacity and coverage gain from frequencydiversity
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
810.139ppt-Cimini-7/98
-
7/31/2019 Andrea Goldsmith
99/99
The desire for mobility coupled with the demandfor Internet and multimedia services indicate abright future for wireless data.
Current products and services have unsatisfactoryperformance for high-speed wireless data applications.
The inherent limitations of the radio channel can besignificantly reduced using signal processing andarchitectural techniques, at the expense of costand complexity.
The network-level design must take into accountthe physical layer limitations of the wireless channel,as well as the impact of user mobility.
SUMMARY