page 1 robust ultra high frequency (uhf) satellite communications protocol for uuvs sbir topic #:...
TRANSCRIPT
Page 1
Robust Ultra High Frequency (UHF) Satellite Communications
Protocol for UUVs
SBIR Topic #: N02-019Contract #: N66604-02-C-4577
Deliverable Item 0001ACPresentation on Phase I Work
Wavix, Incorporated27 January 2003
Page 2
Phase-I Objectives
• Understand the maritime noise environment
• Characterize RF (UHF) communication impediments
• Identify mitigating (low-level) protocol techniques
• Recommend development paths to follow
Page 3
The Challenge
Page 4
Maritime Noise ConcernsOriginal List
• Signal fading from wave motions rocking UUV
• High seas coating antenna or changing local RF propagation characteristics
• Shadowing by waves• Surface wave reflections causing multi-path
interference• Changing atmospheric conditions in LOS• Nature of channel noise: high BER?
Page 5
Maritime Noise ConcernsNon-Obstacles I
• Antenna wash-over– Choose appropriate non-conductive housing– Water has small attenuation at UHF
• Ground-plane issues– Appropriate choice of antenna:
• Quadrifilar Helix has small sensitivity to ground plane
• Helical antenna is very sensitive to ground plane
Page 6
Maritime Noise ConcernsNon-Obstacles II
• Water aerosols– Water has little attenuation in UHF band– (Atmospheric water effects become
serious above ~2-4 GHz)
• UUV rocking– Motions are too small to disrupt signals– Motions are too slow for Doppler effects
Page 7
Characteristic Length Scales
• Wave Heights: 2 m• RF Wavelength: 1 m (@ 300
MHz)• Surface Roughness: 0.1 m
Page 8
Characteristic Time Scales• Wave times: ~5-10 s• SATCOM frames:
– 8.96 s, made of 1024 blocks [5 kHz waveform]– 1.3866 s, made of 0.052 ms time chips [25 kHz
waveform]• Geo propagation time: 500 ms• SATCOM blocks:
– 8.75 ms blocks in 5 kHz waveform; variable number/channel
– 0.052 ms “time chips” in 25 kHz waveform; channel format varies but order of 100 might be typical
• ATM packet size: ~2.5 ms – IP packets are 4 to 10 times larger, but variable
• Symbol size: ~0.05 ms (@ 19.6 bits/second)
Page 9
Maritime Noise ConcernsObstacles
• Small look angle to satellite
• Wave obscuring• Ocean-surface roughness• Nature of the RF channel
Page 10
Satellite Elevation vs. Latitude
Latitudes = Elevations
(for G = 0!)
20 = 82
30 = 71
40 = 56
50* = 40
60 = 25
70 = 12
(* US/Canadian border,
English Channel, etc.)
Page 11
Waves Obscuring LOS
• Fading by attenuation at small look angles; some multipath at large look angles
• Operations in northern latitudes: near-horizon viewing of SATCOM satellites
• Operations in sea-state four: wave heights of 1.25 to 2.5 meters (average = 1 PI)
• Depth of fades: ~7 dB @ 1 GHz• Wave period: 5 to 10 seconds • Time scale of fades: ~1-3 seconds
Page 12
Ocean Surface Roughness
• Fading effects from multi-path interference
• Length scales of ~0.1 meter mean less interaction with UHF signals
• Most serious at very small look angles
Page 13
Nature of the RF Channel
• The RF channel is principally a fading channel (Rayleigh channel, or channel with memory), and not a noisy channel (AWGN channel, or memoryless channel)
• BER has limited utility• Memory channels are much more
difficult to simulate (Markov chains)• Much less research exists for fading
channels
Page 14
Protocol Strategy
Look at low-level protocol strategies that will have the greatest utility in
mitigating the predominantly fading maritime-communications channel.
Page 15
Illustration courtesy Catherine Werst
Page 16
BER vs. S/N
0 1 2 3 4 5 6 7 8 9 10 11 12 13 1410
-6
10-5
10-4
10-3
10-2
10-1
100
BE
R
QPSK(BPSK)
Non-coherent FSK
Eb/N
o (dB)
Page 17
Tradeoffs Discussed
• UUV System Configuration
• Satellite Access Methods• Physical Layer:
Modulation• Physical Layer: Coding• Link-Layer Directions &
Recommendations
Page 18
UUV System Configuration
• Antenna– Type of antenna– Housing for antenna– Antenna on mast– Antenna diversity
• Power– A system issue that
affects RF performance
Page 19
Satellite Access Methods
• Single-User Channels• Frequency Division: FDMA• Code Division: CDMA
– Spread-spectrum approach would be challenging in satellite environment because of dynamic signal balancing
Time Division: TDMA [& GSM] – Provides adequate multi-use capabilities– Mature technology– Compatible with SATCOM, although different
specifications may be desirable
Page 20
System-Level Schematic
Bit-StreamTo Packet
ChannelEncoding
Mod.& Xmit
Packet toBit-Stream
ChannelDecoding
Rec. & De-Mod.
Air Link
PhysicalLayer
LinkLayer
Page 21
Bits rate vs. Symbol rate
Symbol Transmissions (baud)
Bi-state:1 bit/symbol
4-state:2 bit/symbol
Page 22
Physical Layer: Modulation
I
Q
I
Q
01 00
11
01
10
I
Q
I
BPSK QPSK
“M-ary” PSK
Trellis-coded
Page 23
Modulation Tradeoff
The result from information theory limits
information rate/bandwidth(i.e., baud rate), but not bits/second,
which can be increased with a compensating increase in transmit
power.Higher-order modulation techniques
are useful for bandwidth-limited applications, but we recommend
QPSK .
Page 24
Channel Coding &Forward Error Correction
• Channel coding describes the process by which a logical bit stream gets turned into a modulated signal suitable for transmission of the desired information.
• Forward Error Correction (FEC) describes coding techniques that encode the bit stream so that errors in the received bit stream can be corrected.
• Coding & FEC all take place in the lowest layer of the protocol stack, transforming a bit stream into its most desirable form for modulating the carrier for transmission.
• The benefit of an FEC technique is described as its coding gain.
• FEC is not the same as error detection (e.g., CRC bits).• Not all channel coding is FEC (e.g., NRZI).
Page 25
FEC Techniques & Tradeoffs• Block Codes, which operate on blocks of
symbols, are generally better on block errors– Hamming Codes– BCH codes– Reed-Solomon Coding
• Convolutional Codes, which operate on the stream of bits, are generally better on bit errors– Convolutional Codes – Viterbi Decoding – Turbo Codes (Turbo anything is very hot)– Turbo Product Codes
• Coding gains are generally 2 – 3 dB
Page 26
Reed-Solomon Coding I
• RS coding operates on a block of symbols, not modifying it but adding additional bits to the stream that can correct errors in the block.
• RS(n,k) – n encoded symbols; k message symbols– t = (n – k)/2 symbols can be corrected
• The degree of error correction depends on the number of added bits (2t).
• Adding bits increases bandwidth overhead.
Page 27
Reed-Solomon Coding II• In principal: Modern theoretical treatments of Reed-
Solomon coding use Galois group theory [GF(28)]!• In brief: combinations of correcting bits can
efficiently identify erroneous combinations of information bits
• In practice: The algorithms are widely available as firmware.
• NASA specifies RS(255,239) or RS(255, 223) for deep-space missions.
We recommend RS(255,239) for its generally good performance and easy availability.
Page 28
Block Code Performance IG
rap
h b
y K
. Aza
det
IEEE 8
02
.3 H
igh
-Sp
eed
Stu
dy
Gro
up
Ple
nary
meetin
g, M
on
treal
July
19
99
Page 29
Block Code Performance IIG
rap
h b
y K
. Aza
det
IEEE 8
02
.3 H
igh
-Sp
eed
Stu
dy
Gro
up
Ple
nary
meetin
g, M
on
treal
July
19
99
Page 30
Convolutional Encoding Concept
1 2 3 4 5 … n
Constraint length (n); rate (1/2); puncturing.
+
+
Xmit a
Xmit b
Page 31
Convolutional Encoding Characteristics
• Operational choices:– Constraint length (length of shift register)– Generating polynomials (useful constraint lengths
are generally known to have optimal choices)– Symbol rate (determined by number of adders)– Puncturing (to increase the symbol rate at the cost of
decoding difficulty)• Our recommendation, again, is based on
simplicity, easy availability and, in this case, SATCOM compatibility.
• We recommend: Constraint length 7 Rate to be determined, but ½ and ¾ are common
Page 32
Viterbi Decoding(with Soft-Decoding Information)
•In principal: Maximum-Likelihood Estimation•In practice: Algorithms available in firmware Soft-decoding uses signal strength to assist in decision making, for a gain of ~2 dB.
Rec. bitstream
etc.
11
10
01
00
Page 33
Concatenation Concept
If one coding scheme is good,wouldn’t two be better?
In fact, Reed-Solomon and Convolutional Encoding are
complementary and commonly used together.
Page 34
The Importance of Interleaving
• Interleaving effectively transforms block errors into bit errors.
• Interleaving is neither error-correcting or error-detecting; it is an error avoidance technique.
• There is no coding gain associated with interleaving
• The tradeoff in choosing the size of the interleaver is between time scale of bit dispersal and tolerable delay times in transmission.
x1
x2
x3
x4
x1’ …
r1 …
r1’
r2’
r3’
r4’
Page 35
Recommended Concatenation
Reed-Solomon Outer Code – RS(255, 239)
Large-Order Interleaving– Scale to be determined by communication
constraints, but scaled to be effective against fading from waves
Convolutional Inner Code– Constraint length 7– Rate ½ to ¾– Viterbi decoder with soft-decision
• Coding gain ~5 dB => • BER reduction from ~10-3 to 10-9
Page 36
Physical Layer Diagram
(Link-Layer Packets)
Reed-Solomon Encoding, RS(255,238)
Large-Scale Bit Interleaving
Convolutional EncodingL=7, R=1/2 or 3/4
NRZI Encoding
QPSK modulation
Transmit
Receive
(Link-Layer Packets)
Reed-Solomon Decoding, RS(255,238)
Large-Scale Bit De-Interleaving
Viterbi Decodingwith Soft-Decision
NRZI Decoding
Quadrature Demodulation
Antenna Diversity
Page 37
Link Layer Recommendations
• Minimize Automatic Retransmit Request (ARQ)– Reserve it for higher-level protocols
• Minimize handshaking• Recognize transmission delay times• Build a delay-tolerant network
Page 38
Metaframing
• A concept we introduced in the proposal, but didn’t develop in Phase I
• Not SATCOM compatible, but similar in framework.
• Requires:– Good receiver timing and clock
synchronization– Reconsideration of guard times– Data backcapture to achieve gains
Page 39
Hierarchy of Recommendations
• Easy– Implement within UHF SATCOM
TDMA/DAMA capability• Medium
– Protocols beyond SATCOM capabilities– Possibly implemented over dedicated
SATCOM channels– May require new radios– May affect UUV system design
• Hard– Not compatible with existing SATCOM
specifications
Page 40
Easy Recommendations
• Use Quadrifilar Helix Antenna• Use existing TDMA/DAMA satellite
access • Use existing QPSK Modulation• Use existing convolutional inner
code– length 7, rate ½ or ¾
• Add additional interleaving and RS(255,239) outer code if possible
Page 41
Medium Recommendations
• Implement antenna diversity• Implement in custom-designed radio:
– Full concatenated channel coding:• Reed-Solomon outer code: RS(255, 238)• Interleaving (depth to be determined)• Convolutional inner code: length 7, rate ½ or ¾
– Viterbi decoder with soft-decoding
• Define delay-tolerant link-layer through transport-layer protocols.
• Possibly redesign TDMA specifications (over dedicated SATCOM channel)
Page 42
Hard Recommendation
Build a new, non-geostationary satellite system that will give
significantly better coverage over the oceans and ease the RF
communication channel’s susceptibility to fading because of high seas and small look angles.