Comparison of IEEE 802.11g Proposals: PBCC, OFDM & MBCK Sean Coffey, Ph.D., Anuj Batra, Ph.D., Srikanth Gummadi, Matthew Shoemake, Ph.D., Ron Provencio and Chris Heegard, Ph.D. Home and Wireless Networking Texas Instruments 141 Stony Circle, Suite 130 Santa Rosa California 95401 (707) 521-3060, coffey@ti.com Sean Coffey, Texas Instruments

Outline Overview of comparison with CCK-OFDM Comparison with MBCK Details of PBCC vs. CCK-OFDM Conclusion Sean Coffey, Texas Instruments

Comparison with CCK-OFDM Sean Coffey, Texas Instruments

The fundamental PBCC performance edge The PBCC solution uses a more sophisticated code Advantage of PBCC solution of 3.75 dB over CCK-OFDM solution in received power (22Mbps versus 26.4Mbps) The CCK-OFDM solution requires approximately 130% more received power in the same environment. 3.35 dB, 22Mbps vs 24Mbps (115% more ) 2.95 dB over CCK-OFDM solution in received pwr per bit 100% more received power for the same performance after normalizing for rate differences. This doubles area coverage, battery-life, cell density or combination. Sean Coffey, Texas Instruments

PBCC solution in multipath The PBCC performance advantage carries over to all channel conditions slightly greater in multipath PBCC uses an optimized "cover code" that makes the code effectively time-varying and inherently more resistant to multipath This is the "P" in "PBCC" Sean Coffey, Texas Instruments

Achievable rates/throughput Top mandatory rates: 22 Mbps for PBCC 26.4 Mbps for CCK-OFDM Top optional rates: 33 Mbps for PBCC 59.4 Mbps for CCK-OFDM 10% higher than for 802.11a Sean Coffey, Texas Instruments

Achievable throughput (cont) Substantial extra overhead in CCK-OFDM reduces throughput: for MPEG packets (188 byte) PBCC 22 has higher throughput than CCK-OFDM 26.4 PBCC 33 has higher throughput than CCK-OFDM 52.8 PBCC 33 Mbps has a 2 dB signal-to-noise ratio advantage over CCK-OFDM 26.4 Mbps Sean Coffey, Texas Instruments

Time to market / solution feasibility TI has produced a chip implementing all mandatory parts of its proposal. PBCC proposed mandatory mode will work with existing 802.11b radios. The CCK-OFDM proposal requires design of a new radio Long delay in market availability of IEEE 802.11g products. Sean Coffey, Texas Instruments

Overall approach The fundamentals of the respective approaches: there is no fundamental performance difference between single-tone and multi-tone systems. TI is developing 802.11a-compliant hardware. Both single-tone and multi-tone make sense; it does not make sense to do both! as is required for backward compatibility in CCK-OFDM proposal. Sean Coffey, Texas Instruments

IP issues TI has offered a royalty-free license for all its inventions required to implement the mandatory portions of the standard if its solution is adopted as the standard There are serious third-party IP issues with CCK-OFDM proposal Sean Coffey, Texas Instruments

Comparison to MBCK Sean Coffey, Texas Instruments

Performance advantage Based on MBCK submission In Gaussian noise, Eb/N0: MBCK 22 Mbps: 8.0 dB, vs. PBCC: 5.5 dB PBCC advantage 2.5 dB Variable, 100 ns, MBCK 15.75 dB, vs. PBCC 11.75 dB PBCC advantage 4.0 dB Fixed, 25 ns: 7.2 dB vs PBCC 8.1 dB difference –0.9 dB Fixed, 100 ns: 16.2 dB vs PBCC 8.75 dB difference 7.45 dB Fixed, 250 ns: 17.5 dB vs. PBCC 9.75 dB difference 7.75 dB Sean Coffey, Texas Instruments

Discussion Essential components of MBCK proposal not currently revealed MBCK proposal beats CCK-OFDM in AWGN but is still well short of PBCC’s solution From MBCK submission, PBCC’s solution has advantage in most channel conditions Sean Coffey, Texas Instruments

Details Sean Coffey, Texas Instruments

Performance advantage PBCC-22 uses a novel 256-state, rate 2/3 binary convolutional code matched to 8-PSK CCK-OFDM adopts a standard 64-state, rate 1/2 convolutional code widely known to all coding practitioners since 1972 existing PBCC 11 Mbps option in 802.11b standard is a variation of this type of code A major component of overall 2.95 dB advantage comes from differences between the codes Sean Coffey, Texas Instruments

Performance curve /24 Sean Coffey, Texas Instruments

Performance curve (cont) /24 Sean Coffey, Texas Instruments

Performance curve (cont) Sean Coffey, Texas Instruments

Performance curve (cont) Sean Coffey, Texas Instruments

Performance gain (cont) CCK-OFDM solution spends 20% of transmission time on cyclic extension of every OFDM symbol non-information-carrying, performance cost CCK-OFDM solution has 4 pilot tones out of 52 tones overall spends 1/13 of energy on non-information-carrying signal Bottom line: PBCC advantage of 1.6 (coding) + 1.0 (cyclic extension) + 0.35 (pilot tones) = 2.95 dB Sean Coffey, Texas Instruments

Performance gain (cont) Required Eb/N0, experimental: Assume 1000 byte packets, PER = 0.01: CCK-OFDM Eb/N0 = 8.4 dB (document 00/392r1) PBCC Eb/N0 = 5.5 dB Documented difference: 2.9 dB Sean Coffey, Texas Instruments

Performance difference CCK-OFDM 26.4 Mbps vs. PBCC 22 Mbps Difference in received power must be 3.75 dB PBCC’s 22 Mbps has 130% more coverage, or battery life, or cell density than CCK-OFDM 26.4 Mbps (3.35dB, 115% more @ 24 Mbps) CCK-OFDM 26.4 Mbps vs. PBCC’s 33 Mbps Difference in received power must be 2.0 dB CCK-OFDM requires more! (1.6 dB @ 24 Mbps) PBCC’s 33 Mbps mode has 60% more coverage, or battery life, or cell density, than CCK-OFDM 26.4 Mbps (45% more @ 24 Mbps) Sean Coffey, Texas Instruments

Multipath performance: Variable, 100 ns PBCC 22 Mbps Eb/N0 = 11.75 dB CCK-OFDM 26.4 Mbps Eb/N0 = 15 dB Difference 3.25 dB Fixed, 25 ns 8.1 dB vs. 11 dB, difference 2.9 dB Fixed, 100 ns 8.75 dB vs. 12 dB, difference 3.25 dB Fixed, 250 ns 9.75 dB vs. 13.2 dB, difference 3.55 dB Sean Coffey, Texas Instruments

Achievable rates/throughputs CCK-OFDM requires extra OFDM preamble: 10.9 msec extra SIFS time postamble: 6 msec extra end-symbol pad time: up to 3.64 msec Average extra overhead is 18.7 msec per packet in addition to 96 msec 11b short preamble Sean Coffey, Texas Instruments

Breakeven points Below certain packet lengths, PBCC has higher throughput: PBCC 22 vs. CCK-OFDM 26.4, no acks: 306 bytes PBCC 33 vs. CCK-OFDM 39.6, no acks: 429 bytes PBCC 33 vs. CCK-OFDM 52.8, no acks: 189 bytes PBCC 33 vs CCK-OFDM 59.4, no acks: 159 bytes Compare MPEG packet length of 188 bytes! Sean Coffey, Texas Instruments

Breakeven points, contd. PBCC 22 vs. CCK-OFDM 26.4, with acks: 466 bytes PBCC 22 vs. CCK-OFDM 39.6, with acks: 176 bytes PBCC 33 vs. CCK-OFDM 39.6, with acks: 699 bytes PBCC 33 vs. CCK-OFDM 52.8, with acks: 309 bytes PBCC 33 vs CCK-OFDM 59.4, with acks: 264 bytes Sean Coffey, Texas Instruments

Complexity requirements: Decoder PBCC decode a 256-state (1024-edge) BCC with an 11/16.5 MHz clock CCK-OFDM decode a 64-state (128-edge) BCC with a 60 MHz clock if top optional mode implemented Similar complexity and cost TI’s PBCC 22 Mbps decoder has been implemented its practicality is proven Sean Coffey, Texas Instruments

Complexity requirements: Receiver PBCC a sophisticated single-tone receiver to handle multipath robustly Same receiver as all IEEE 802.11b modes CCK-OFDM a large FFT based processor at both the transmitter and receiver Similar complexity with TI’s single tone receiver Plus, a single-tone receiver for backward compatibility Undue extra cost, for what benefit? TI’s 22 Mbps receiver has been implemented its practicality is proven. Sean Coffey, Texas Instruments

Required radio PBCC uses 8-PSK, with 802.11b’s 11 MHz clock no problems of dynamic range or peak-to-average ratio. CCK-OFDM proposal involves 16- and 64-QAM much larger peak-to-average ratio Poses problems with linearity and dynamic range... … and is dramatically different as an interferer CCK-OFDM proposal requires a new radio and baseband a long delay in the debut of standardized products recall HomeRF 10 Mbps products expected in 2001 Sean Coffey, Texas Instruments

Temporal Character: Peak to Average Power Barker-2 PAR: 2.1 = 3.2 dB CCK-11 Sean Coffey, Texas Instruments

PAR (cont) PBCC-24: OFDM-26.4: PAR: 2.1 = 3.2 dB 11.5 = 10.6 dB Sean Coffey, Texas Instruments

Overall approach Both single-tone and multi-tone systems achieve “capacity” on the channels seen by wireless LANs. “Which is better: many bits on one tone or few on many?” Well studied in information theory, answer is well known: highest achievable performance in each system is exactly the same This is neither obvious nor easy to show, but has been known since at least the 1960's. Sean Coffey, Texas Instruments

IP issues Wi-LAN and others claim broad patents covering use of OFDM in wireless LANs Are attempting to enforce these patents via litigation Are not participants in IEEE 802.11g TI has offered a royalty-free license, without time limit, for all its inventions required to implement the mandatory portions of the standard, if its solution is adopted as the standard Sean Coffey, Texas Instruments

Wi-LAN's Zaghloul: "Wi-LAN is ready for the worst," he says, "and antcipates the best. We anticipate [Cisco] will pay without much of a fight, but if they don't, we're prepared to go all the way [in a legal fight to secure license payments.]” http://www.isp-planet.com/technology/cisco_wi-lan.htm Sean Coffey, Texas Instruments

Sean Coffey, Texas Instruments

Conclusions PBCC proposed mandatory 22 Mbps mode Has large performance advantage (3.75 dB over 26.4 Mbps) Works in same environment as IEEE 802.11b Put PBCC-22 Mbps next to 11 Mbps CCK: if the CCK mode works, the PBCC-22 mode also works Has been implemented, is ready for market now Spectrally and temporally identical to existing 802.11b Works with existing 802.11b radios No new interference type Royalty-free license offered PBCC proposed optional 33 Mbps mode requires 2dB less SNR than the mandatory 26.4 Mbps OFDM Sean Coffey, Texas Instruments