CCK-OFDM Summary Steve Halford Mark Webster Jim Zyren Paul Chiuchiolo May 2001 CCK-OFDM Summary Steve Halford Mark Webster Jim Zyren Paul Chiuchiolo Intersil Corporation S. Halford, et al Intersil Corporation

Why OFDM for High Rate? OFDM recognized as best solution for W-LAN May 2001 Why OFDM for High Rate? OFDM recognized as best solution for W-LAN Selected by 802.11a & ETSI for W-LAN at 5 GHz OFDM meets current & future needs: Highest rates and backward compatibility Meets consumer expectations set by 802.11a High throughput with ultra-short preamble New deployments & outdoor bridge applications Share baseband with 802.11a Dual band radios Multiple baseband vendors Best performance for complexity trade Multipath & Bluetooth S. Halford, et al Intersil Corporation

Overview of Intersil’s Proposal for 802.11g May 2001 Overview of Intersil’s Proposal for 802.11g S. Halford, et al Intersil Corporation

OFDM for High Rate Extension May 2001 OFDM for High Rate Extension Replace data portion of packet with OFDM modulation Data rates 6, 9, 12, 18, 24, 36, 48 or 54 Mbps using 20 MHz symbol rate Existing .11b radios will recognize preamble and header Length field will be correctly decoded CCA mechanisms maintained Use reserve bits in 802.11b header for OFDM parameters OFDM Proposal is compatible with 802.11b S. Halford, et al Intersil Corporation

OFDM-Specific Fields Add OFDM-only sync and SIFs pad May 2001 OFDM-Specific Fields Add OFDM-only sync and SIFs pad Reduces complexity of receiver & allows for flexible transmit filtering Short Sync allows time for clock rate change 4 useconds duration is half duration of 802.11a Allows time to switch rates Can also use to refine time & frequency estimates S. Halford, et al Intersil Corporation

May 2001 OFDM-Specific Fields Long Sync provides training data for channel estimation Provide 8 useconds of training data (same as 802.11a) Do not need to rate-change 802.11b channel estimate Can switch filters at transmitter for OFDM mode SIFs pad extends the SIFs time to match 802.11a 802.11g receivers will see a 16 usec SIFs during OFDM operation 802.11b receivers will still see a 10 usec SIFs during OFDM operation S. Halford, et al Intersil Corporation

Impact of OFDM-specific Fields May 2001 Impact of OFDM-specific Fields Simplify Radio design & add flexibility with added fields OFDM Sync: Transition time & channel estimation SIFs Pad: SIFs time compatibility between 802.11b & 802.11a What is the impact on throughput? Add 18 useconds of overhead Reduction in Throughput 100 byte: 310 kbps 1000 byte: 500 kbps 2346 byte: 704 kbps ** Short Preamble option at 24 Mbps, No ACK Throughput impact is negligible S. Halford, et al Intersil Corporation

Ultra-Short Preamble Option May 2001 Ultra-Short Preamble Option Modulation & preambles identical to 802.11a Reduces total preamble to 20 usecond New deployments Outdoor point-to-point links Gives a route to 802.11a in the 2.4 GHz band Compatible with mandatory 802.11g Single PHY solution to high rate S. Halford, et al Intersil Corporation

Radio Design Issues Baseband processor change only for 36 Mbps May 2001 Radio Design Issues Baseband processor change only for 36 Mbps Current RF supports all rates up to 36 Mbps OFDM preserves current channelization 3 channels spaced by 25 MHz (U.S. deployments) 48 & 54 Mbps supported with new RF Same baseband Higher density constellation has more stringent requirements on radio front end Requirements are well from 802.11a designs Design issues are well understood S. Halford, et al Intersil Corporation

Forward Compatibility of CCK-OFDM May 2001 Forward Compatibility of CCK-OFDM S. Halford, et al Intersil Corporation

Advantages of compatibility May 2001 Advantages of compatibility CCK-OFDM provides .11a & .11g compatibility Could also add HiperLAN2 compatibility Marketplace will see single waveform as high rate wireless LAN solution Introduction of a new waveform like PBCC will only fracture the marketplace Lower cost dual band radios Dual band 802.11a with PBCC requires one to build two complex basebands (OFDM & PBCC) Multiple vendors provide basebands & IP for .11g S. Halford, et al Intersil Corporation

Dual Band Radio Allow seamless transitions for laptop WLANs May 2001 Dual Band Radio Allow seamless transitions for laptop WLANs Single low-cost card could provide support for .11b, .11a, and .11g Auto-detect network or best connection type S. Halford, et al Intersil Corporation

Further Advantages of Compatibility May 2001 Further Advantages of Compatibility Consider the cost of compatibility Optional Mandatory TI Compromise Proposal .11a (OFDM-only) + .11b(CCK) + .11g (PBCC) .11b (CCK) + .11g(OFDM) Intersil Proposal inlcudes .11a(OFDM) S. Halford, et al Intersil Corporation

Adding CCK Baseband to OFDM May 2001 Adding CCK Baseband to OFDM Adding CCK support is much easier than alternative PBCC requires number of complex design efforts PBCC-22 requires 30x ops/bit over CCK (00/384r1) Must still support CCK CCK receivers based on rake receiver Implement with a channel matched filter & correlator S. Halford, et al Intersil Corporation

Complexity Comparison between May 2001 Complexity Comparison between OFDM and PBCC S. Halford, et al Intersil Corporation

PBCC:Reduced State Approach May 2001 PBCC:Reduced State Approach Suggested approach by PBCC proponents Still higher complexity than OFDM Minimum MF length will be 10 taps Retain only 64 states out of 216 states at each update 8 symbols form the channel state Each update, surviving state generate 4 new candidates 64 leads to 256 states Retain the most likely 64 (requires sort) Updates occur at symbol rate for this approach S. Halford, et al Intersil Corporation

Compare PBCC & OFDM Compare Complexity of WMF with FFT & FEQ May 2001 Compare PBCC & OFDM Compare Complexity of WMF with FFT & FEQ Compare Complexity of the two 64-state decoders S. Halford, et al Intersil Corporation

Compare WMF w/ FFT & FEQ PBCC: WMF Complexity is driven by length May 2001 Compare WMF w/ FFT & FEQ PBCC: WMF Complexity is driven by length Ignore the estimation problem Needs 10 taps to handle 5 multipath rays PBCC-22: Requires 880 x 106 real multiplies per second PBCC-33: Requires 1980 x 106 real multiplies per second Increase length to 15 to cover same delay spread OFDM: Consider the FEQ & FFT FEQ: 52 multiplies/symbol = 52 x 106 real multiplies per second FFT: Radix 4, 96 multiplies/symbol = 96 x 106 real mps Total: 148 x 106 real multiplies per second Remains fixed for all data rates S. Halford, et al Intersil Corporation

Compare 64-state Decoders May 2001 Compare 64-state Decoders Compare Trellis search approaches in terms of: branch metrics calculations path metric updates compare selects PBCC: 64 States  256 states  64-states for each symbol 256 branch metric calculations 256 path metric updates Select 64 best of 256 states -- Variety of approaches OFDM: 64 states  128 paths  64 states for each information bit 128 branch metric calculations 128 path metric updates Compare each pair & select best -- 64 compare-selects S. Halford, et al Intersil Corporation

Compare Complexity: Summary May 2001 Compare Complexity: Summary Waveform Real Multiplies (x 106) Branch Metric Path Metric PBCC-22 880 2816 PBCC-33 1980 4224 OFDM-24 148 3072 OFDM-36 4608 PBCC has much higher complexity due to matched filter Equivalent OFDM operation (FFT & FEQ) have fixed complexity OFDM & PBCC have nearly same complexity in trellis search (?) PM & BM difference is proportional to data rate difference S. Halford, et al Intersil Corporation

May 2001 Conclusions for 802.11g S. Halford, et al Intersil Corporation

Conclusions PBCC relies on coding & sophisticated receiver May 2001 Conclusions PBCC relies on coding & sophisticated receiver Non-standard code matched to 8-PSK signal Different code than used for optional PBCC-11 No interleaver to help spread burst errors Sensitive to burst errors like generated by Bluetooth Cover code benefit never demonstrated Why add unnecessary elements? Requires complex decoder design to get adequate performance 3 years in development with no product or public demo Single company provider ? Already lags OFDM systems for data rate S. Halford, et al Intersil Corporation

Conclusions OFDM is forward & backwards compatible May 2001 Conclusions OFDM is forward & backwards compatible Uses existing long & short preamble for compatibility 802.11a modulation in place of CCK OFDM-specific training data added for reduced complexity Dual band radios possible Offers 802.11a in the 2.4 GHz band through ultra-short preamble OFDM offers the highest rates of all proposals 36 Mpbs with current radio (baseband only change) 48 & 54 Mbps possible with new radio design Only proposal that meets consumer expectations S. Halford, et al Intersil Corporation

Conclusions OFDM is ideal for W-LAN environment May 2001 Conclusions OFDM is ideal for W-LAN environment Equalization split between transmitter & receiver for lower overall complexity Guard Interval -- absorbs multipath without complexity FFT & FEQ -- Low complexity & fixed for any rate Nearly MLSE without complexity of PBCC See Documents IEEE Submissions 01/153 & 01/060 Lower complexity error correction code 64 state code Single code used for all code rates via puncturing More robust to narrowband interference Simple to remove known interference S. Halford, et al Intersil Corporation

Conclusions OFDM now meets regulatory approval (5/10/01) May 2001 Conclusions OFDM now meets regulatory approval (5/10/01) OFDM is now in the 2.4 GHz band Higher rates than PBCC already being offered IEEE should embrace & ensure network compatibility OFDM (802.11a) development was collaborative Multiple companies contributed ideas Complexity & design is well known & proven Many companies will offer products S. Halford, et al Intersil Corporation