ICCAD 1997 Tutorial
Design Technology for Building Wireless Systems Rajesh Gupta University of California, Irvine
[email protected] Mani Srivastava UCLA
[email protected] T Y• O F•
SI
C
E
A
R
E BE
•
ER
I
T
H
A
NIV
T
H
ORN
L IG H T
LE
E•U
LIF
A
•1868•
•T
Copyright 1997 Rajesh Gupta & Mani Srivastava
Phenomenal Growth in Wireless Voice & Data Services ●
35-60% annual growth in PCS users
●
By 2000, one in three phones will be mobile (42% in US)
●
Nordic countries: 10 mobile phones being added for every wireline phone
●
Japan: number of users doubled from 10M to 21M from March to october 1996
●
600M mobile phone users by 2001
●
$17B in PCS license auctions
●
300% growth in wireless data from 1995 to 1997
Big demand for portable computers: ●
2m ($290M) in 1988 to 74M ($54B) in 1998
●
20% of all computers sold are laptops 2
“Anytime Anywhere Anyform” Information Systems
PCS & Multimedia Messaging on the road
Fax & email on the beach
mani
UCLA
Wireless Sensors
Multimedia wireless LANs & PBXs in offices, schools, hospitals, homes
Networked sensors everywhere 3
Size & Battery Life are Critical in Wireless Devices ●
Battery technology is a key hurdle - no Moore’s Law here! Battery
Rechargeable?
Gravimetric Density (Wh/lb)
Volumetric Density (Wh/l)
alkaline-MnO2 (typical AA)
NO
65.8
347
NO
60
500
Li/MnO2
NO
105
550
zinc air
NO
140
1150
NiCd
YES
23
125
Li-Polymer
YES
65-90
300-415
Nominal Capacity (Watt-hours / lb)
silver oxide
40 NiMH
30 20
NiCd
10 0
65
70
75
80 Year
85
90
95 4
Where does the Battery Power go? Laptop
Cellular Phone
Laptop + Wireless Adapter
Personal Wireless Terminal
Microprocessor
1-4 W
1-4 W
Memory
1W
1W
Logic
2W
2W
Hard Disk
1W
1W
Display
2-6 W
2-6 W
0.185 W
0.6 / 1.8 W
0.6 / 1.8 W
2.5 W
2.5 W
?
0.085 W
Programmable DSP
0.5 W
RF Transceiver
2/4W
Commn. Processing Sound/Audio I/O
?
0.3 W
●
Typical laptop: 30% display, 30% CPU + memory, 30% rest
●
Wireless devices: increasing communication & multimedia processing
Low power VLSI are a key to wireless 5
Wireless Systems Design: Key Driving Forces
●
Increasing integration of communication & multimedia system components due to advances in semiconductor technology & circuits - RF CMOS circuits - MEMS structures RF components, display
●
Relentless digitization continues - high speed digital circuits & A/D converters IF and even RF processing in digital domain direct conversion techniques - complex communication algorithms favor digital implementation - increasing CPU MIPS make even a “software radio” possible
A wireless-system-on-a-chip is becoming possible
6
Building a Wireless System on a Chip RF & IF Transceiver Baseband Processing Custom ASIC Logic
Algorithm Acceleration Coprocessors
Wireless Network Protocol Processor (Microcontroller)
Application Processor
DSP Core RAM/ROM
RAM ROM DRAM
RAM/ROM DRAM
Network/Host/Peripheral Interface 7
Challenge to VLSI & CAD RF & IF Transceiver Baseband Processing Custom ASIC Logic
Algorithm Acceleration Coprocessors
Wireless Network Protocol Processor (Microcontroller)
DSP Core RAM/ROM
RAM ROM DRAM
Computer with Radios analog circuits that minimize special analog process steps maximize digital and minimize analog computation reusable communication & multimedia modules
energy efficient embedded software synthesis Application Processor
RAM/ROM DRAM
Network/Host/Peripheral Interface
low cost & low power protocol processor cores
8
Tutorial Goals
Present basics of wireless systems, and VLSI design issues, techniques, and tools for building integrated wireless systems
This tutorial will NOT describe: - detailed CAD algorithms for solving system design problems - theory of radio and communication systems design - detailed architecture of any wireless communication systems
9
Tutorial Outline ●
Introduction to Wireless Communication Systems
●
- system and medium characteristics - technological evolution in the design of wireless communication systems Wireless Systems Design
●
- digital communications: modulation, coding, multiple access - example designs VLSI Circuits for Wireless Systems
●
- micro-architecture for wireless systems-on-a-chip - direct-conversion for digital communications using VLSI Design technology for Wireless Systems
●
- design entry, validation, and analysis tools Pre-designed Core Blocks and IP Issues for Wireless
●
Future Outlook and Conclusions
10
Part 1: Introduction to Wireless Communication Systems
Wireless Spectrum Frequency in Hz 104
LF
106 MF
108
VHF HF
1010
1012
1014 IR
UHF
1016 UV
1018
1020 X-Ray
Light
1022
1024
Cosmic Rays
Radio
46 49
Cordless (CT-1)
824-849 869-894 902-928
Cellular (AMPS, IS-136, IS-95)
ISM
1850-1990
PCS
2400-2483
5.15 - 5.35 & 5.725 - 5.825 GHz
ISM
U-NII Frequency in MHz
12
Diversity of Applications in Wireless Communications
Low Voice Interactive Data Data Rate Video teleconferencing
Information Content (Mbps) 100.0 10.0
Wireless ATM Wireless LAN: IEEE 802.11
1.0 0.1
Mobile Wireless Multimedia
Cordless: DECT, PHS, PACS, WLL
Cellular: GSM, IS95, IS54, PDC, 0.01
Wireless Data: Mobitex, CDPD, pACT, GPS Office
Building Indoors
Stationary
Walking
Vehicular
Outdoors
Environment Multimega bits/sec throughput for robust, reliable multimedia networking over wide range of environments.
●
13
Characteristics of Wireless Systems ●
Wireless - limited bandwidth, high latency - variable link quality (noise, disconnections, other users) - heterogeneous air interfaces - easier snooping necessitates encryption
●
Mobility - user and terminal location dynamically changes - speed of terminal mobility impacts wireless bandwidth - easier spoofing necessitate authentication
●
more signal processing
more protocol processing
Portability - limited battery capacity, computing, and storage - small dimensions
higher energy efficiency
14
Time Varying Wireless Environment
LOS
R S D
No LOS! D
●
Available wireless resource undergoes dramatic & rapid changes
●
- multipath reflection, doppler fading, frequency collisions Rapid signal fades & distortions as the receiver moves - e.g. noise-like Rayleigh Fading when multipath signals are summed 15
Sources
Simplified View of a Digital Radio Link antenna
Source Coder Multiplex Source Coder
Multiple Access
Channel Coder
Power Amplifier
Modulator
transmitted symbol stream
carrier fc “Limited b/w” “Highly variable b/w” “Random & Noisy” “Spurious disconnections”
RADIO CHANNEL
Destinations
received (corrupted) symbol stream antenna
Source Decoder Source Decoder
Demultiplex
Multiple Access
Channel Decoder
Demodulator & Equalizer
RF Filter
carrier fc 16
Propagation of Radio Waves ●
Line of Sight (LOS) - free space P r = ( P t G t G r λ 2 ) ⁄ ( ( 4π ) 2 d 2 L )
●
Reflection (with Transmittance and Absorption) - radio wave impinges on an object >> λ (30 cm @ 1 GHz) - surface of earth, walls, buildings, atmospheric layers - if perfect (lossless) dielectric object, then zero absorption - if perfect conductor, then 100% reflection - reflection a function of material, polarization, frequency, angle
●
Diffraction - radio path obstructed by an impenetrable surface with edges - secondary waves “bend” around the obstacle (Huygen’s principle) - explains how RF energy can travel even without LOS, a.k.a “shadowing”
●
Scattering (diffusion) - when medium has large number of objects < λ (30 cm @ 1 GHz) - similar principles as diffraction, energy reradiated in many directions - rough surfaces, small objects (e.g. foliage, lamp posts, street signs) 17
Log-normal Shadowing Path Loss Model ●
Assume average power (in dB) decreases proportional to log of distance d PL ( d ) = PL ( d 0 ) + 10n log ----- d 0
●
Path-loss exponent, n, depends on propagation environment Environment Free Space Urban area cellular radio Shadowed urban cellular radio In-building LOS Obstructed in building Obstructed in factories
n 2 2.7 - 3.5 3 to 5 1.6 to 1.8 4 to 6 2 to 3
●
Problem: “Environment clutter” may differ at two locations at same d
●
Measurements show that at a given d path loss has a normal distribution d PL ( d ) = PL ( d 0 ) + 10n log ----- + X σ d 0 - X σ is a zero-mean Gaussian r.v. (in dB) with standard deviation σ (in dB) - σ says how “good” the model is 18
Example Link Budget Calculation ●
Maximum separation distance vs. transmitted power (with fixed BW) Given: - cellular phone with 0.6W transmit power - unity gain antenna, 900 MHz carrier frequency - SNR must be at least 25 dB for proper reception - receiver BW is B = 30 KHz, and noise figure F = 10 dB What will be the maximum distance? Solution: N = -174 dBm + 10 log 30000 + 10 dB = -119 dBm For SNR > 25 dB, we must have Pr > (-119+25) = -94 dBm Pt = 0.6W = 27.78 dBm This allows path loss PL(d) = Pt - Pr < 122 dB λ = c/f = 1/3 m Assuming d0 = 1 km, PL(d0) = 91.5 dB For free space, n = 2, so that: 122 > 91.5 + 10*2*log(d/(1 km)) or, d < 33.5 km Similarly, for shadowed urban with n = 4, 122 > 91.5 + 10*2*log(d/(1 km)) or, d < 5.8 km
19
Small-Scale Fading
●
Fading manifests itself in three ways
●
1. time dispersion caused by different delays limits transmission rate - replicas of signals with different delays (reflection, diffraction etc.) 2. rapid changes in signal strength (up to 30-40 dB) over small ∆x20dB fade every 2.5s
●
Also, a function of frequency, and fade depth
●
Diversity techniques help - multiple antennas, multiple frequencies 21
Data Rate Limitation in Frequency Selective Fading
●
“Frequency selective fading” results in inter-symbol interference 0.1 maximum data rate without significant errors = -----------------------------delay spread - e.g. GSM has a bit period of 3.69 µs, or a rate of 270 kbps
●
Data rate can be improved by “equalization” - equalizer is a signal processing function (filter) cancels the inter-symbol interference usually implemented at baseband or IF in a receiver - must be adaptive since channel is unknown & time varying training, tracking, and re-training during data transmission
●
GSM example - with its equalizer, GSM can tolerate up to 15 µs of delay spread - otherwise, with 15 µs of delay spread, GSM would be limited to 7 kbps
22
Combating the Wireless Channel Problems ●
Increase transmitter power - counters flat fading, but costly and greatly reduces battery life
●
(Adaptive) Equalization - compensates for intersymbol interference
●
Antenna or space diversity for “multipath” - usually, two (or more) receiving antennas, separated by λ/2 - selection diversity vs. scanning diversity vs. combining diversity - “adaptive antenna arrays” or “smart antennas”
●
Forward error correction - transmit redundant data bits - “coding gain” provides “fading margin” - not very effective in slowly varying channels or long fades
●
Automatic Repeat Request (ARQ) protocols - retransmission protocol for blocks of data (e.g. packets) in error - stop-and-wait, go-back-N, selective-repeat etc. 23
A Digital Radio Link antenna
Source Coder Multiplex Source Coder
Multiple Access
Channel Coder
Power Amplifier
Modulator
transmitted symbol stream
carrier fc “Limited b/w” “Highly variable b/w” “Random & Noisy” “Spurious disconnections”
RADIO CHANNEL
Destinations
received (corrupted) symbol stream antenna
Source Decoder Source Decoder
Demultiplex
Multiple Access
Channel Decoder
Demodulator & Equalizer
RF Filter
carrier fc 24
Evolution of Mobile & RF Wireless Systems ●
First Generation: Analog - Voice - analog modulation - cellular phone (AMPS) with manual roaming - cordless phones - packet radio networks
●
Second Generation: Digital - Voice & Data - digital modulation - cellular & PCS phones with seamless roaming, integrated paging (IS-54, IS-95, IS-136, GSM etc.) - digital cordless, multi-zone cordless, wireless PBXs - wireless data LANs (802.11), MANs (Metricom), WANs (CDPD, ARDIS, RAM)
●
Third Generation: Digital - Multimedia - unified digital wireless access anytime, anywhere - voice, data, images, video, music, sensor etc.
25
Tutorial Outline ●
Introduction to Wireless Communication Systems
●
- system and medium characteristics - technological evolution in the design of wireless communication systems Wireless Systems Design
●
- digital communications: modulation, coding, multiple access - example designs VLSI Circuits for Wireless Systems
●
- micro-architecture for wireless systems-on-a-chip - direct-conversion for digital communications using VLSI Design technology for Wireless Systems
●
- design entry, validation, and analysis tools Pre-designed Core Blocks and IP Issues for Wireless
●
Future Outlook and Conclusions
26
Part 2-A: Wireless Systems Design: Basics
Sources
Simplified View of a Digital Radio Link antenna
Source Coder Multiplex Source Coder
Multiple Access
Channel Coder
Power Amplifier
Modulator
transmitted symbol stream
carrier fc “Limited b/w” “Highly variable b/w” “Random & Noisy” “Spurious disconnections”
RADIO CHANNEL
Destinations
received (corrupted) symbol stream antenna
Source Decoder Source Decoder
Demultiplex
Multiple Access
Channel Decoder
Demodulator & Equalizer
RF Filter
carrier fc 28
Digital Modulation & Demodulation - A “User’s View” ●
Modulation: maps sequence of “digital symbols” (groups of n bits) to sequence of “analog symbols” (signal waveforms of length TS)
●
Demodulation: maps sequence of “corrupted analog symbols” to sequence “digital symbols” - e.g. maximum likelihood decision TS-long analog symbol
corrupted
n-bit digital symbol ...(0110) (0111) (0000)...
CHANNEL
MOD
DEMOD
best effort output ...(0110) (0111) (0000)...
noise, fading, etc.
S1 S2
Set S = {S1, S2,... SM} of M waveforms of length TS e.g. obtained by distinctively modifying the phase and/or frequency and/or amplitude of a carrier M=2 is “binary modulation” Otherwise, M-ary modulation n = floor(log2 M)
SM t=0
t=TS
29
Commonly Used Digital Modulation Techniques
Coherent
Non-Coherent
Phase-shift keying (PSK)
FSK
Frequency-shift keying (FSK)
ASK
Amplitude-shift keying (ASK)
Differential PSK (DPSK)
Continuous phase modulation (CPM)
CPM
Hybrids
Hybrids
●
Coherent or Synchronous Detection: process received signal with a local carrier of the same frequency and phase
●
Noncoherent or Envelope Detection: requires no reference wave
30
Selecting a Modulation Schemes
●
Provides low bit error rates (BER) at low signal-to-noise ratios (SNR)
●
Occupies minimal bandwidth
●
Performs well in multipath fading
●
Performs well in time varying channels (symbol timing jitter)
●
Low carrier-to-cochannel interference ratio
●
Low out of band radiation
●
Low cost and easy to implement
●
Constant or near-constant “envelope” - constant: only phase is modulated may use efficient non-linear amplifiers - non-constant: phase and amplitude modulated may need inefficient linear amplifiers
No perfect modulation scheme - a matter of trade-offs! 31
Metrics to Evaluate Modulation Schemes ●
Power Efficiency (or, Energy Efficiency) η P - ratio of signal energy per bit to noise power spectral density required required at the receiver for a certain BER (e.g. 10-5) ηP = Eb ⁄ N 0 - measures ability to give low BER at low signal power levels - impacts battery life!
●
Bandwidth Efficiency η B - ratio of throughput data rate to bandwidth occupied by modulated signal η B = R ⁄ B bps/Hz - measures ability to accommodate data within a given bandwidth
●
Often a trade-off between power and bandwidth efficiencies, e.g. - adding redundancy (FEC) reduces bandwidth efficiency, but reduces the received power required for a given BER - modulation schemes with higher values of M decrease B but increase E b for a given BER 32
Choice of a Modulation Scheme ●
At 0.001% BER and a fixed transmission bandwidth:
M
Power Penalty Factora
Bit-Rate Gain Factora
Energy Penalty Factora
2
1
1
1
4
2
2
1
8
4.7
3
1.56
16
10
4
2.5
32
20.7
5
4.1
64
42
6
7
a. Relative to BPSK (M=2)
●
BPSK and QPSK has the same energy efficiency but QPSK has two times more bandwidth efficiency (bit rate gain factor) than BPSK.
●
The drawback of using QPSK is in the poor achievable energy efficiency in practice => use GMSK to achieve a bandwidth efficiency of 1.25 with BT = 0.3. 33
A Geometric View of Modulation ●
Signal set S = { s 1(t), s 2(t), …, s M(t) } represents points in a vector space
●
Vector space defined by a set of N ≤ M orthonormal (i.e. orthogonal and with unit energy) basis signals { φ j(t) j = 1, 2, …, N } - N is the dimension of the vector space
●
Every s i(t) can be expressed as a linear combination of basis signals
●
Example: BPSK signals s 1(t) = s 2(t) =
2E b ⁄ T b cos ( 2π f c t ) 0 ≤ t ≤ T b and
2E b ⁄ T b cos ( 2π f c t + π ) can be represented as: φ 1(t) =
2 ⁄ T b cos ( 2π f c t )
s 1(t) =
E b φ 1(t)
s 2(t) = – E b φ 1(t)
34
The Constellation Space ●
Geometric representation of S is called the Constellation Diagram, e.g. for BPSK: Q
I – Eb
●
Eb
Bandwidth occupied by the modulation scheme decreases as the number of signal points / dimension increases - a densely packed modulation scheme is more bandwidth efficient - however, bandwidth increases with dimension N
●
Probability of bit error is a function of the distance between the closest points in the constellation diagram - a densely packed modulation scheme is less power efficient
35
Some Examples... ●
M-ary QAM Q d
I M=16
●
6 d 2 = --------------E s M–1
M-ary PSK Q
I d
M=4
π d = 2 E s sin ----M
36
Comparison of Several Modulation Methods
●
Ref.: Wireless Information Networks by Pahlavan & Levesque, 1995
37
Sources
Simplified View of a Digital Radio Link antenna
Source Coder Multiplex Source Coder
Multiple Access
Channel Coder
Power Amplifier
Modulator
transmitted symbol stream
carrier fc “Limited b/w” “Highly variable b/w” “Random & Noisy” “Spurious disconnections”
RADIO CHANNEL
Destinations
received (corrupted) symbol stream antenna
Source Decoder Source Decoder
Demultiplex
Multiple Access
Channel Decoder
Demodulator & Equalizer
RF Filter
carrier fc 38
Multiple Access ●
Fundamental problem
Shared Time-Frequency Subspace
Allocated Spectrum
Frequency
How to share the Time-Frequency space among multiple co-located transmitters?
Time 39
Basestation versus Peer-to-Peer Models
Basestation (infrastructure - centralized)
Peer-to-Peer (ad hoc network - fully-connected vs. multihop) 40
Approaches to Wireless Multiple Access Sharing of Time-Frequency Space
Slotted-time vs. Non-slotted time
Demand-based Assignment
Static (Fixed) Assignment e.g. Time-division & Frequency-division
Contention-based
“Connection Oriented”
Conflict-free Random Access e.g. ALOHA, PRMA Carrier-sensing
Scheduled Access
e.g. Token-passing & Polling
e.g. DQRUMA
“Packet Oriented”
Controlled Random Access 41
Frequency Division Multiple Access (FDMA) Assign different frequency bands to individual users or circuits - frequency band (“channel”) assigned on demand to users who request service - no sharing of the frequency bands: idle if not used - usually available spectrum divided into number of “narrowband” channels symbol time >> average delay spread, little or no equalization required - continuous transmission implies no framing or synchronization bits needed - tight RF filtering to minimize adjacent band interference - costly bandpass filters at basestation to eliminate spurious radiation - usually combined with FDD for duplexing f2 f1
f 2’ f1’
f2’ f1’
Frequency
●
f2 f1
Time 42
Time Division Multiple Access (TDMA) ●
Multiple users share frequency band via cyclically repeating “time slots” - “channel” == particular time slot reoccurring every frame of N slots - transmission for any user is non-continuous: buffer-and-burst digital data & modulation needed, lower battery consumption - adaptive equalization is usually needed due to high symbol rate - larger overhead - synchronization bits for each data burst, guard bits guard bits for variations in propagation delay and in delay spread - usually combined with either TDD or FDD for duplexing TDMA/TDD: half the slots in a frame used for uplink, half downlink TDMA/FDD: identical frames, with skew (why?), on two frequencies Sync
Data
Guard
slot 2
Frequency
slot 1
slot 6
slot 5
frame i-1
1 2 56 frame i
frame i+1
Time 43
Some TDMA Systems
Bit rate
IS-54
DECT
PHS
270.8 kbps
48.6 kbps
1.152 Mbps
384 kbps
Carrier spacing (b/w)
200 kHz
30 kHz
1.728 MHz
300 kHz
Time slot duration
0.577 ms
6.7 ms
0.417 ms
0.625 ms
Slots/frame
8 (or 16)
3 (or 6)
12
4
FDD
FDD
TDD
TDD
73% adaptive equalizer training overhead
80% adaptive equalizer training overhead
67% system control overhead
71%
Modulation
GMSK
π/4 DQPSK
GMSK
π/4 DQPSK
Adaptive equalizer
required
required
none
none
FDD or TDD? % payload in time slot
●
GSM
GSM handles time dispersion widths up to 18-20 µs... i.e. 5 bits of ISI - transmission bandwidth >> channel coherence bandwidth
●
IS-54 handles time dispersion up to 40 µs... i.e. 2 symbols might interfere - less complex equalizer needed than GSM|
●
Need equalization indoors at rates > 2 Mbps (DECT is only 1.152 Mbps) 44
Hybrid FDMA/TDMA ●
“Pure” TDMA with single frequency band is undesirable - require tight timing tolerances
●
Most TDMA systems actually employ hybrid FDMA/TDMA - multiple carriers with multiple channels per carrier - channel == (frequency band, time slot) tuple - may do “frequency hopping” on a frame-by-frame basis to combat multipath interference (Time Division Frequency Hopping: TDFH) increases system capacity
(f5, t1) t1 t2 t3 t4
(f1, t1) Frequency
(f3, t4)
f6 f5 f4 f3 f2 f1
(f2, t3)
frame i-1
frame i
frame i+1
45
Code Division Multiple Access (CDMA) ●
Multiplexing in the Code Space - multiple transmitters occupy the same frequency-time space - transmissions encoded with codes with very low cross-correlation - receiver retrieves a specific transmission with its corresponding code CDMA may be combined with TDMA or FDMA
Frequency
Code
●
c1 c5 c3 c2
46
Spread Spectrum Signalling ●
Spread Spectrum is the most common CDMA encoding technique
●
- originally developed for military communication systems - “spread” the signal over a much larger bandwidth than the minimum - signal appears pseudo-random with noise like properties - uniform small energy (W/Hz) over a large bandwidth hides the signal ⇒ Note: use of spread-spectrum does not imply use of CDMA Spreading is done using a unique code
●
Receiver does the “despreading” by using a time-synchronized duplicate of the spreading code
●
Inefficient for a single user, but multiple users can share band
●
Inherent interference rejection capabilities (e.g. narrowband interferers)
●
Resistant to multipath effects - delayed versions appear as uncorrelated noise - can even exploit multipath signals by combining them
●
Processing Gain: Gp = Bspread / Bsignal - indicates improvement in signal-to-interference ratio due to spreading 47
What is Spread Spectrum Communication? spectral density
interference
Ai
spread signal Aspread fspread
TRANSMIT
spectral density
unspread signal fdata
RECEIVE
Spreading Code running at f spread .
spectral density Adata
frequency
frequency
Wide Band Anti-jam -> high capacity CDMA Combats multipath -> diversity LPI -> Privacy LPD -> low power density
Adata
despread signal spread interference
Ai,received frequency
fdata
f spread PG = --------------------f bit
48
CDMA Using Direct Sequence (DS) Spread Spectrum ●
Spread the narrowband data by multiplying with a wideband pseudorandom code sequence - bits sampled, or “chipped”, at a higher frequency (e.g. 1.228 Mcps in IS-95) - signal energy is “spread” over a wider frequency (e.g. 1.25MHz in IS-95) - code sequences have little cross-correlation (orthogonal) - code sequences have little correlation with shifted versions of self
●
Received signal multiplied by synchronized replica of the code sequence
●
Energy of each “chip” is accumulated over a full data bit time
transmitted signal
=
X
01101011
01101011
PN Sequence (code)
Recovered signal
Intended receiver
X =
X 10110010
Chip
Noise - can be low pass filtered
Other receivers
digital data
49
CDMA Using Frequency Hopping Spread Spectrum ●
Transmission frequency is periodically changed - available spectrum divided into bands with central frequencies as carriers - sequence of data bursts with time-varying pseudo-random carrier frequencies - time duration between hops is the hop duration or hopping period Th - bandwidth of a frequency band in the hopset is the instantaneous b/w B - bandwidth of spectrum over which hopping occurs is total hopping b/w Wss - processing gain is Wss/B
●
Fast frequency hopping: more than one hop during each transmitted symbol
●
Slow frequency hop: one or more symbols transmitted in a hop
Frequency
channel #2
channel #1 f6 f5 f4 f3 f2 f1
50
Contention-based Multiple Access ●
Many transmitters access a channel with no or minimal coordination
●
Transmission in bursts of data
●
Collisions may happen: need ACK or NACK with retransmission - delays induced - lower spectral efficiency
●
Three categories: random access, scheduled access, hybrid access
Transmitter # 1
Transmitter # 2
Packet B
Packet C
Packet A
One Packet Time (τ)
Time
Vulnerable Period (2τ)
51
Contention-based Multiple Access in Wireless Systems?
●
Ethernet uses contention-based medium access...
●
Following attributes make contention-based multiple access interesting with wireless: - “carrier sensing” is much costlier in wireless 20-30 µs - can’t listen while transmitting therefore cannot detect collisions - what matters is the collision at a receiver ... but the transmitter can’t sense the channel at the receiver! - effects of spatial distribution of wireless nodes hidden terminal problem exposed terminal problem near-far problem (capture effect)
52
IEEE 802.11 MAC ●
Support for multiple PHYs: ISM band DSSS and FHSS, IR @ 1 and 2 Mbps
●
Efficient medium sharing without overlap restrictions
●
- multiple networks in same area and channel space - Distributed Coordination Function: using CSMA /CA - based on carrier sense mechanism called Clear Channel Assessment (CCA) Robust against interferers (e.g. co-channel interference)
●
- CSMA/CA+ACK for unicast frames with MAC level retransmission Protection against Hidden Terminal problem: Virtual Carrier Sense
●
- via parameterized use of RTS/CTS frames with duration information Provision for Time Bounded Services via Point Coordination Function
●
Configurations: ad hoc & distribution system connecting access points
●
Mobile-controlled hand-offs with registration at new basestation ad hoc network
distribution system
infrastructure network 53
IEEE 802.11 MAC (contd.) ●
CSMA/CA: direct access if medium free for > DIFS, else defer & back-off DIFS
DIFS
source
PIFS DATA
other
contention window
SIFS
DATA select slot & decrement back-off as long as idle
defer access
●
CSMA/CA + ACK: receiver sends ACK immediately if CRC okay - if no ACK, retransmit frame after a random back-off DIFS
source
contention window
DATA
receiver
SIFS ACK
other
DIFS defer access
●
DATA select slot & decrement back-off as long as idle
RTS/CTS with duration: distribute medium reservation information - also used in the defer decision 54
Cellular Systems MSC Pre-Cellular
●
PSTN
Post-Cellular
Replace single high power transmitter covering the entire service area with lots of low power transmitters (basestations) each covering a fraction of the service area (cell) - mobiles in sufficiently distant basestations may be assigned identical channel (frequency, time slot, & code) - system capacity may be increased without adding more spectrum
●
Major conceptual breakthrough in spectral congestion & user capacity - required relatively minor technological changes frequency reuse & co-channel interference channel allocation hand-offs 55
Space Division Multiple Access (SDMA) ●
Control radiated energy for each user in space - spot beam antennas (sectorized antennas) - different areas served by different antenna beams may use same frequency (CDMA, TDMA) or different frequencies (FDMA) - in future, adaptive antennas
56
Part 2-B: Wireless Systems Design: Standards, Design Issues, and Examples
The Un-wired World Wireless Communications
Amateur
Automotive - IVHS - GPS
Industrial
Analog
Digital - DECT - CT-2 - PHP - USCT - ISM
Business
Military/Aero Long-Haul
Monitoring - AMR - Control Cordless
- CT-0 - CT-1 - CT-300
Consumer
Analog - AMPS - ETACS - NMT450 - NMT900 - NMT-0 - Comvik - JTACS
Cellular
Digital - GSM - IS-54 - IS-95 - IS-136 - RCR-27
Paging WPABX - POSCAG - DECT - CT-2 - ERMES - PHP - SSB - USCT - ISM
WLAN PMR/SMR Mobile Data - 802.11 - ARDIS - DECT - Mobitex Conv - HIPerLAN - Omnitracs - ISM - Cellular/CDPD ESMR - Metricom - MIRS - TETRA PCN/PCS
- DCS1800 - PHP - LEO
- FPLMTS - UMTS - RACE 58
Evolution of PCS Technologies, Systems, and Services Macro-cellular Cellular
Satellites?
Micro-cellular Messaging
Paging
?
High-tier PCS
Phone point Cordless
PABX
Wide Area Data
WLANs
Low-tier PCS
Cordless
?
Micro-cells
?
Macro-cells
?
WLANs
Grand Unification?
WLANs PRESENT
PAST
FUTURE 59
AMPS System (First Generation Analog) ●
Two 25 MHz bands: 824-849 MHz upstream, 869-894 MHz downstream
●
Divided into 30 MHz frequency bands - pair needed for a duplex channel
●
FDD+FDMA: 834 duplex channels
●
7-way frequency reuse (18 dB min. signal-to-co-channel interference)
●
Two types of channels: control and voice channels
●
Network controlled handoff - MSC becomes a bottleneck
●
Capacity constraints - 40-50 connections per cell
●
No on-air privacy, fraud a major problem Proprietary
SS7
BS
AMPS Common Air interface
MSC (MTSO)
OMC mobility management
BS
MSC (MTSO)
MS BS
PSTN
BS
HLR
BS BS
VLR
AUC
databases
60
GSM System (Second Generation Digital) ●
Two 25 MHz bands: 890-915 MHz upstream, 935-960 MHz downstream
●
Divided into 200 KHz frequency bands - 125 in each direction
●
FDD+TDMA+FH: 8 slots/4.615 ms frame, 270.833333 kbps raw, 22.8 kbps/user
●
Frequency hopping to combat multipath problems
●
Two types of logical channels: traffic channels and control channels
●
Mobile assisted handoff - BSC reduce the load on MSC
●
Features: subscriber identity module and on-air privacy
●
Services: telephone, data or bearer, short messaging
GSM Radio Air interface
BTS
MSC (MTSO)
OMC
BSC BTS
MSC (MTSO)
MS BTS BTS BTS BTS
PSTN
BSC HLR Abis Interface
A Interface
VLR
AUC
databases
SS7
61
Cellular Data Packet Network (CDPD) ●
Packet data network overlay on AMPS - same 30 KHz channels
●
Data packets are sent over unused voice channels
●
Channel hopping ensures non-interference with voice
●
Raw data rate is 19.2 kbps Reed-Solomon coded - real rate much less
●
Broadcast downlink, Data Sense Multiple Access (DSMA) MAC on uplink
●
Variety of connection-less, connection-oriented, and multipoint services
●
Reliable and unreliable classes - handling over radio link
●
In particular, IP (Internet Protocol) datagram connectivity
●
Mobile controlled handoff, registration at basestation to reduce paging
●
“Home MD-IS” tunnels incoming traffic to current MD-IS MD-BS MD-BS
M-ES
MD-IS mobility management
IS
IS
MD-IS
Data n/w (internet) MD-BS
M-ES
connection-less router
MD-BS F-ES 62
Designing Mobile Wireless Multimedia Systems
PSTN BASE STATION
WIRED NODE WIRELESS NODE
http://www. N
• • • • • •
PHONE
http://www. N
antenna RF + A/D digital transmitter/receiver channel codec source codec network protocols
modem ethernet transceiver
ETHERNET
63
Generic Mobile & Wireless System Architecture
Application & Services
Partitioning Source Coding & DSP Context Adaptation
OS & Middleware
Disconnection Mgmnt. Power Management QoS Management
Network
Rerouting Impact on TCP Location Tracking
Data Link
Multiple Access Link Error Control Channel Allocation
Radio, IR
Modulation Schemes Channel Coding RF/Optical Circuits
64
Radio Design Challenges ●
High speed digital processing
●
High performance in Eb/N0
●
Low complexity
●
Energy efficient (mW/MSps or nJ/OP)
Algorithm
RF Front-end Architecture
Fixed Point
Digital Modem IC Architecture
Partition 65
Partition between Analog and Digital Processing
●
Analog RF Signal Processing
Analog IF Transceiver
Baseband Analog-to-Digital Converter
Digital Baseband Signal Processing
Analog RF Signal Processing
IF Analog-to-Digital Converter
Digital IF Transceiver
Digital Baseband Signal Processing
Advantages allows for adaptability with little component replacements achieves Eb/N0 performance close to optimum (coherent BPSK) parameterizable to provide ease of redesign and upgrade
●
Challenges digital circuits operate at IF signal rate rather than baseband rate digital implementation can be more complex to minimize loss in Eb/N0 66
A Direct-Sequence Spread-Spectrum Radio Modem CODE PROCESS SELECT GAIN
Carrier Detect
TX Data POWER CONTROL
PN GENERATOR
PN Acquisition LOOP
Spread Data TX
VGA
LPF
AMP
BPF FREQ CNTRL
CLOCK RECOVERY LOOP
CARRIER RECOVERY LOOP
A/D
FREQUENCY SYNTHESIZER
AGC
LPF
LNA
6 Decision
Ack.: C. Chien & R. Jain, UCLA To SIR Est. Recv. Data
●
Low complexity, high speed, adaptable, and energy efficient transceiver in a single-chip
67
Transceiver Chip Design Issues
●
Challenge: Implement a complete coherent receiver on a single chip
●
Circuit Design Issues finite wordlength parameterizability critical path optimization complexity reduction
●
System Design Issues maintain stability in three feedback loops.
68
Costas Loop Filter Optimization INPUT Ec/N0= -17 dB 20
25
−20
10 dB
−40
20
N2
Eb/N0 (dB)
30
0
−60
0 dB
9 dB 15
−80 40 10
30
N2
-10 dB
20 10 0
0
20
15
10
5
25
30
5
35
N1 5
10
15
20
25
30
N1
Coefficient as powers of two shifts: C1 = 2 C2 = 2
–N 1
C1
–N 2 C2
D
Optimization Criteria: min ( max ( N 1, N 2 ) ), E b ⁄ N 0 ≥ 10 ± 0.5
Ack.: C. Chien & R. Jain, UCLA 69
40
0 dB 20 -11 dB 10
0 5 10 15 IF Input Quantization Size (Bits)
Complexity increase in receiver Sample rate through the multiplier 50.8 MHz sample rate requirement
200
100
100
50
0 4
8 12 16 IF Input Quantization Size (Bits)
Minimize IF quantization size reduce complexity and power dissipation at required throughput.
N N
min ( N ),
E b ⁄ N 0 ≥ 10 ± 0.5
0
DDFS
❥
-17 dB
Complexity Increase (%)
Output Eb/N0 (dB)
30
0
150
300 10 dB
Multiplier Sample Rate (MHz)
IF Wordlength Optimization
N
Ack.: C. Chien & R. Jain, UCLA 70
PN-Acquisition: Complexity/Performance Trade-off ●
PN acquisition: correlation between the incoming bits and the P/N sequence of the desired transmitter Serial Acquisition Energy Slope Detection
Received PN
Clock
❥
Timing
Generation
800 Gates
PN-Code Generator
Match Filter Acquisition Timing N-Tap Matched Filter
Received PN
Energy
Clock
Detection
Generation
Generator PN-Code
❥
Nc * Nif * 12 Gates + 800 Nc = #chips/bit Nif = IF Quantization
❥
10 000 Gates with Nc = 127 and Nif = 6 71
A Single-Chip 1.2 Micron CMOS DSSS Radio Modem DIGITAL BASEBAND TRANSMITTER DATA INPUT
DIFFERENTIAL ENCODER
SPREAD DATA
GOLD CODE GENERATOR (PNGEN)
DIGITAL IF RECEIVER 50.8 MHz
12.7 MHz INTEGRATE DUMP I1
(100-800) kHz INTEGRATE DUMP I2
DIFFERENTIAL DECODER
DATA OUT
100 kHz -12.7 MHz DDFS
LOOP FILTER
PHASE DETECTOR
IF SAMPLING CLK
PN TRACK CONTROL
INTEGRATE DUMP Q2
COSTAS LOOP
EARLY PN
LATE PN
INTEGRATE DUMP Q1
PN-ACQUISITION LOOP
Performance INTEGRATE DUMP I1
+
CHIP DELAY
LOOP FILTER
INTEGRATE DUMP Q1
NCO
CLOCK RECOVERY 50.8 MHz
❥
Low Complexity -- 51 K Transistors
❥
High Power Efficiency -- 21.7 mJ/MSample
❥
Maximum Chip Rate -- 12.7 Mchips/sec
❥
Scalable Performance -- Data Rates and Processing Gain: 100, 200, 400, 800 kbps at 12, 15, 18, 21 dB, respectively
-
IF SIGNAL
12.7 MHz
406.4 MHz
Ack.: C. Chien & R. Jain, UCLA
72
Integration of Radio into a System Custom Frame Grabber
Camera
Video Codec FPGA
DT Frame Grabber
CPU
Proxim RangeLAN2
Keyboard Single-chip DSSS Modem IC Memory and Mass Storage
Adaptive Direct Sequence Spread Spectrum Radio RF Front-end DSSS IF modem, Packet Interface, Adaptation Interface, Analog-Digital Conversion
Ack.: C. Chien & R. Jain, UCLA 73
Example 1: UCLA’s Wireless Multimedia Node
Video Capture 16-bit YUV
Control
VGA
Video Codec
Compressed Data Interface
Host Interface
Network Interface Chip Serial Data Interface Packet Buffer
Controller
12-bit RGB
PC-104 Bus
Video Buffer
Frame Buffer
Host Interface Modem
Host CPU
Wireless Channel
74
Example 2: Bell Labs’ SWAN Wireless ATM System BASESTATION CPU
Mobility Management
Mobile Notebook
Drivers for Adapter Cards
BusInterface Interface Bus
Peripheral Interface Interface
Interface Peripheral Peripheral
Host Interface
BACKBONE ATM ADAPTER CARD
MAC PHY
CPU CPU
CPU
XCVRInterface Interface XCVR XCVR Interface
mani
Host Interface
FAWN Flexible Adapter For Wireless Networking
CPU
Peripheral Interface
ETHER WA RE SYSTEM S/ W
Connection Switching
Lucent
Personal Multimedia Terminal
XCVR Interface
FHSSRFRFXCVR XCVR FHSS
FHSS RF XCVR
FHSS RF XCVR
To Antenna
BASESTATION
Personal Communicator
MOBILE END-POINTS
ATM SWITCH 75
FAWN Reconfigurable Wireless Adapter
to host processor
ARM CPU
PCMCIA
Peripheral Interface
PCMCIA Interface
RF Modem ADC
Dual Port RAM
Modem Controller
SRAM UART Control PAL
Dimensions
10.8 cm (W) x 1.9 cm (H) x 11.4 cm (D)
Power Consumption of FAWN
2.0 W
Power Consumption of radio transceiver
0.6 W (receive) 1.8 W (transmit)
Firmware resources
20 MIPS, 4 MByte
Reconfigurable hardware resources
10000 Gates equivalent
76
Example 3: Personal Mobile Terminal
microphone LCD display
Soft keys
Personal Terminal 6808
PRESS TO SCAN
↓
Scanner switch
●
SCANNER
↓
Bar code scanner
Simple hardware - peripheral card + FAWN adapter
●
Multimedia interface - audio, graphics, soft keys, bar code
●
Dumb end-point for “network-hosted mobile services”
77
Example 4: Berkeley’s Infopad Project ●
Infopad: low power wireless multimedia terminal
●
- no local general purpose processing (“dumb terminal” model) - speech and pen controlled user interface - audio, video, and text/graphics streams to the terminal Infonet: network infrastructure for Infopads
●
- based on cell, pad, and type servers Medley Gateway: transport & coding of video, audio, & graphics to Infopad
●
http://infopad.eecs.berkeley.edu/
78
Infopad Terminal Architecture 250 Kbps
1 Mbps
Proxim Uplink Radio
RX/TX Interface
Plessey Downlink Radio
ARM Subsystem
Low Power Infopad Bus
LCD IF
PEN IF
AUDIO IF
VIDEO IF
ucb
Color
Subsystem
mW
Radios
1490
ARM
877 - 2475
Custom H/W
137 - 297
B&W LCD
550 - 3800
Color LCD
3900
Pen Digitizer
150
Codec
50
Voltage Converters
2411
Crystals
75
Test H/W
629
Total
9.9W - 15W
Infopad
●
References:
1. http://infopad.eecs.berkeley.edu/research/terminal 2. [Narayanaswamy96] Narayanaswamy et. al., “Application and Network Support for Infopad,” in IEEE Personal Communications, April ‘96 79
Example 5: Xerox PARCTAB ●
Extremely portable mobile unit
●
- 7.8x10.5x2.4 cm3, 215 gm, 6.2x4.5 cm2 & 128x64x1 touch screen, 3 buttons - IR communication at 19.2 kbaud with CSMA MAC, PWM modulation - 12 MHz Signetics 87C524/528 CPU, 128K memory Basestation transceiver (on ceiling of a room nanocell)
● ●
- IR with variable data rate: 9.6K, 19.2K, 38.4K; CSMA MAC - 38.4K serial link up to 30m with 10 unit daisy chain capability - performs coding/decoding, buffering, link level protocol checks - connected to LAN via serial port of nearby workstations Remote host based applications, proxy agents (per tab), and gateways (datagram service to tab) http://www.ubiq.com/parctab
Tab
Basestation 80
Design Trade-offs in Wireless Nodes
Terminal Complexity
Laptops Notebooks
ge
ra Sto
Palmtops PDAs
n
tio uta
Terminals
mp
Co
Communication Needs & Infrastructure Dependence ●
Computation-communication trade-off affects: - terminal cost - service cost 81
Design Issues
●
Adaptive process gain improves throughput
●
Multipath fading requires equalization
●
Bit rate limited by equalizer complexity
Throughput can be improved by physical layer processing
82
Adaptive Process Gain Improves Throughput 100 Desired
PG = 12 dB
Throughput (kbps)
80
60 PG = 15 dB
40
Achieved
20
PG = 21 dB
0 −15
−10
−5
0
5
Signal-to-Interference Ratio (dB) 83
RF Processing: Power Dissipation Top
Bottom Transmit
Control
Freq. Synth.
AGC
Receive
Power Reg.
Total RF Power = 5.75W Total IF Power = 6.118 W Total Radio Power = 11.87W
84
IF/Baseband Processing: Power Dissipation Top
Bottom
DSSS Analog IF Control
DSSS
Packet Interface Power Regulation
Total IF Power = 6.12W Total RF Power = 5.75 W Total Radio Power = 11.87W
Note: Power budget figures includes power dissipation from regulation inefficiencies.
85
Multipath Fading Requires Equalization
2
0
t
t0 t2 t1
t3
τ • τ > Its / 10 ⇒ ISI causes degradation in BER and will require equalization • τ is a function of transmit power and cluttering in the environment
Mobile
} Wireless
Linear feedback equalizer
5
3
Channel
Transversal equalizer
Linear feedback equalizer
2
Probability of error
0
Transversal equalizer
10-1
1
10-2 5 Linear feedback equalizer
2 10-3
31 taps in transversal equalizer 16 feedforward and 15 feedback taps in linear feedback equalizer
Transversal No interference equalizer
5 1 2 γ = ------- ∑ f k No k
2 10-4 0
5
10
15
20
25
30
35
SNR, db (10 log γ)
Dense Foliage
Urban Clutter
86
Bit Rate Limited by Equalizer Complexity ●
Improved performance using MLSE over DFE/FFE
Probability of error
1
MlSE simulation
10-1
Destination-feedback equalizer
10-2
Correct bits fed back No interference
10-3
Detected bits fed back MlSE bounds
10-4 0
5
10
15
20
25
SNR, dB (10 log γ)
- short training sequence O(100) vs. O(1000) bits ●
But, MLSE has high complexity and processing requirements - complexity ∼ O (4 τ Rs M τRs) - e.g. M=2, τ = 3ms, Rs = 2 Mbaud = 2 Mbps then, complexity ~ 1600 operations ~ 30k gates processing ~ 1600 * 2MHz = 3.2 GOPS
87
Physical Layer Processing to Improve Throughput preamble
DATA
header
throughput = max(throughput)
⇒
Tdata Tpreamble + Theader + Tdata min(Tpreamble), min(Theader)
Theader is protocol dependent
capture-time accumulates in multihop networks
• TCP/IP header • ATM header • MAC/link layer header
Tpreamble is physical layer dependent • time to acquire / capture packet • settling time of LO frequency
Aggressive signal processing can reduce this! 88
Understanding Energy Efficiency P = α C V2 f “Event-Driven” Latency is Important (Burst throughput)
“Continuous” Only Throughput is Important
Reduce V Increase h/w and algorithmic concurrency e.g., Speech Coding Video Compression
Make f low or 0 Shutdown when inactive
Reduce αC Energy efficient s/w System partitioning Efficient Circuits & Layouts
e.g., X Display Server Disk I/O Communication
89
7.0
7.0
5.0
5.0
Speedup
Normalized Delay
Voltage-Parallelism Trade-Off for Low Power
3.0 1.0 1.0
1.5
2.0
2.5
3.0
Ideal Speedup
3.0 1.0
1
2
Supply Voltage, V ●
3
4
5
6
7
8
Parallelism, N
Increased parallelism & reduced voltage can increase energy efficiency - more processors or functional units or pipelining - compiler techniques are the key
●
Architectural bottlenecks: - degradation of speed-up - capacitance overhead due to increased communication 90
Energy Efficiency is not just an Architecture Issue! ●
Radios consume a significant fraction of node power Lucent’s WaveLAN: 23 dBm 915MHz radio network interface transmit = 3W receive = 1.48W sleep = 0.18W GEC Plessey DE6003: 20 dBm, 2.4GHz radio transceiver transmit = 1.8W receive = 0.6W sleep = 0.05W Newton PDA active = 1.2W sleep = 0.164W Magic Link PDA active = 0.7W sleep = 0.3W
Radios need to be actively managed for low power via energy efficient wireless link protocols. 91
Low Power Design for Wireless • Display
Hardware has been addressed • Low power CMOS, • Displays, • Hard drives, etc.
HDD •
Low power protocols remain
µProc
DSPs
Link Layer Protocols MAC Layer Protocols
Radio Modem
92