Robust Ultra High Frequency UHF Satellite Communications Protocol
Robust Ultra High Frequency (UHF) Satellite Communications Protocol for UUVs SBIR Topic #: N 02 -019 Contract #: N 66604 -02 -C-4577 Deliverable Item 0001 AC Presentation on Phase I Work Wavix, Incorporated 27 January 2003 1
Phase-I Objectives • Understand the maritime noise environment • Characterize RF (UHF) communication impediments • Identify mitigating (low-level) protocol techniques • Recommend development paths to follow 2
The Challenge 3
Maritime Noise Concerns Original 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? 4
Maritime Noise Concerns Non-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 5
Maritime Noise Concerns Non-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 6
Characteristic Length Scales • Wave Heights: 2 m • RF Wavelength: 1 m (@ 300 MHz) • Surface Roughness: 0. 1 m 7
Characteristic Time Scales • Wave times: ~5 -10 s • SATCOM frames: – 8. 96 s, made of 1024 blocks [5 k. Hz waveform] – 1. 3866 s, made of 0. 052 ms time chips [25 k. Hz waveform] • Geo propagation time: 500 ms • SATCOM blocks: – 8. 75 ms blocks in 5 k. Hz waveform; variable number/channel – 0. 052 ms “time chips” in 25 k. Hz 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) 8
Maritime Noise Concerns Obstacles • • Small look angle to satellite Wave obscuring Ocean-surface roughness Nature of the RF channel 9
Satellite Elevation vs. Latitudes = Elevations (for G = 0!) 20 = 82 30 = 71 40 = 56 50* = 40 60 = 25 70 = 12 71 72 (* US/Canadian border, English Channel, etc. ) 10
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 d. B @ 1 GHz • Wave period: 5 to 10 seconds • Time scale of fades: ~1 -3 seconds 11
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 12
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 13
Protocol Strategy Look at low-level protocol strategies that will have the greatest utility in mitigating the predominantly fading maritimecommunications channel. 14
Illustration courtesy Catherine Werst 15
BER vs. S/N 10 0 QPSK(BPSK) 10 BER 10 10 10 -1 Non-coherent FSK -2 -3 -4 -5 -6 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 E b /N o (d. B) 16
Tradeoffs Discussed • • • UUV System Configuration Satellite Access Methods Physical Layer: Modulation Physical Layer: Coding Link-Layer Directions & Recommendations 17
UUV System Configuration • Antenna – – Type of antenna Housing for antenna Antenna on mast Antenna diversity • Power – A system issue that affects RF performance 18
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 19
System-Level Schematic Bit-Stream To Packet Link Layer Packet to Bit-Stream Channel Encoding Physical Layer Channel Decoding Mod. & Xmit Air Link Rec. & De-Mod. 20
Bits rate vs. Symbol rate Bi-state: 1 bit/symbol 4 -state: 2 bit/symbol Symbol Transmissions (baud) 21
Physical Layer: Modulation Q BPSK Q 10 I 1 “M-ary” PSK 0 QPSK 11 00 I 01 Q I Trelliscoded I 22
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 . 23
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). 24
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 d. B 25
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 (2 t). • Adding bits increases bandwidth overhead. 26
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. 27
Block Code Performance I Graph by K. Azadet IEEE 802. 3 High-Speed Study Group Plenary meeting, Montreal July 1999 28
Block Code Performance II Graph by K. Azadet IEEE 802. 3 High-Speed Study Group Plenary meeting, Montreal July 1999 29
Convolutional Encoding Concept + 1 2 3 4 Xmit a 5 + … n Xmit b Constraint length (n); rate (1/2); puncturing. 30
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 31
Viterbi Decoding (with Soft-Decoding Information) 11 10 etc. 01 00 Rec. bitstream • 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 d. B. 32
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. 33
The Importance of Interleaving r 1 … • x 1 r 1’ x 2 r 2’ x 3 r 3’ x 4 r 4’ x 1’ … • • • 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. 34
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 d. B => • BER reduction from ~10 -3 to 10 -9 35
Physical Layer Diagram Transmit Receive (Link-Layer Packets) Reed-Solomon Encoding, RS(255, 238) Reed-Solomon Decoding, RS(255, 238) Large-Scale Bit Interleaving Large-Scale Bit De-Interleaving Convolutional Encoding L=7, R=1/2 or 3/4 Viterbi Decoding with Soft-Decision NRZI Encoding NRZI Decoding QPSK modulation Quadrature Demodulation Antenna Diversity 36
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 37
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 38
Hierarchy of Recommendations • Easy – • Medium – – • Implement within UHF SATCOM TDMA/DAMA capability 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 39
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 40
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 transportlayer protocols. • Possibly redesign TDMA specifications (over dedicated SATCOM channel) 41
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. 42
- Slides: 42