High-Throughput Enhancements for 802.11: Features and Performance Month 2002 doc.: IEEE 802.11-02/xxxr0 November 2004 High-Throughput Enhancements for 802.11: Features and Performance John Ketchum, Sanjiv Nanda, Rod Walton, Steve Howard, Mark Wallace, Bjorn Bjerke, Irina Medvedev, Santosh Abraham, Arnaud Meylan, Shravan Surineni QUALCOMM, Incorporated 9 Damonmill Square, Suite 2A Concord, MA 01742 Phone: 781-276-0915 Fax: 781-276-0901 johnk@qualcomm.com John Ketchum, et al, Qualcomm John Doe, His Company

Goals Maximize Throughput, QoS, and Spectral Efficiency November 2004 Goals Maximize Throughput, QoS, and Spectral Efficiency Minimize complexity and assure backward compatibility Provide balance between TTM needs and 11n design longevity economics John Ketchum, et al, Qualcomm

Throughput Comparison November 2004 Throughput Comparison 0 dB: 120 m 10 dB: 60 m 20 dB: 30 m 30 dB: 15 m 37.5 dB: 10m SGI-52: 52 data subcarriers with short guard interval Results given with closed loop rate control, except STBC-OL SS-STBC can achieve 120Mbps at 30m (20dB) ES has > 6 dB advantage over other at 150 Mbps PHY throughput At 30 m (20 dB) ES has >50% PHY t’put advantage over others John Ketchum, et al, Qualcomm

Proposal Summary: PHY Builds on 802.11a waveform November 2004 Proposal Summary: PHY Builds on 802.11a waveform 20 MHz bandwidth with 802.11a/b/g spectral mask 802.11a modulation, coding, interleaving with expanded rate set Backward compatibility through legacy STF, LTF and SIG Supports a maximum of 4 wideband spatial streams Two forms of spatial processing Spatial Spreading (SS): modulation and coding per wideband spatial channel No calibration required SNR per wideband spatial stream known at Tx Eigenvector Steering (ES): via wideband spatial modes/SVD per subcarrier Tx and Rx steering Over the air calibration procedure required Rate adaptation enables sustained high rate operation PHY techniques proven in FPGA-based prototype John Ketchum, et al, Qualcomm

November 2004 Spatial Spreading Spatial spreading for 2 Tx and 4 Tx uses Hadamard matrix No multiplies required to execute Matrix-Vector multiply Flexible number of spatial streams 1 ≤ Ns ≤ Ntx All transmit antennas used, regardless of stream count John Ketchum, et al, Qualcomm

Spatial Spreading: Mandatory & Optional Features November 2004 Spatial Spreading: Mandatory & Optional Features Mandatory Hadamard matrix-vector multiply at transmitter Cyclic transmit diversity at transmitter Receiver must be capable of handling spatially spread signals (zero-forcing, MMSE, etc.) Support for rate feedback in PLCP/MAC header Optional Rate feedback functionality John Ketchum, et al, Qualcomm

November 2004 Eigenvector Steering Substantial throughput gains over baseline spatial spreading Full MIMO channel characterization required at Tx Tx steering using per-bin channel eigenvectors from SVD Rx steering renders multiple Tx streams orthogonal at receiver, allowing transmission of multiple independent spatial streams This approach maximizes both data rate and range Per-stream rate control and rate feedback required for robust high throughput operation John Ketchum, et al, Qualcomm

Support for Eigenvector Steering November 2004 Support for Eigenvector Steering Base standard mandatory features are required to support optional ES mode Independent rates per stream for up to four streams Modulation/coding/interleaving must support independent rates per stream Rate feedback Fields in PLCP header extension or MAC header MIMO training waveform design Must support steered reference Allows implicit channel state feedback in all PPDUs Tone interleaving (TGnSync) or Walsh cover (Qualcomm) Related elements such as signaling for mode control John Ketchum, et al, Qualcomm

November 2004 Eigenvector Steering Some features are mandatory for devices supporting optional ES mode Messaging and sounding waveforms to support over-the-air calibration Transmit steering and computation of Tx steering vectors Real-time matrix-vector multiply capability Determine steering vectors from unsteered training sequence Steered training sequence Other optional Eigenvector Steering features Bi-directional steering: Both STAs in a corresponding pair use Eigenvector steering Uni-directional steering: Only one STA in corresponding pair (e.g. AP) use Eigenvector Steering John Ketchum, et al, Qualcomm

Another Approach Space-Time Block Coding with Spatial Spreading November 2004 Another Approach Space-Time Block Coding with Spatial Spreading Additional Tx diversity benefit of STBC with flexibility of SS Number of STBC streams decoupled from number of Tx antennas Can adjust power allocations in unequal diversity cases Possible compromise approach – best of both worlds John Ketchum, et al, Qualcomm

Throughput Comparison November 2004 Throughput Comparison 0 dB: 120 m 10 dB: 60 m 20 dB: 30 m 30 dB: 15 m 37.5 dB: 10m SGI-52: 52 data subcarriers with short guard interval Results given with closed loop rate control, except STBC-OL SS-STBC can achieve 120Mbps at 30m (20dB) ES has > 6 dB advantage over other at 150 Mbps PHY throughput At 30 m (20 dB) ES has >50% PHY t’put advantage over others John Ketchum, et al, Qualcomm

802.11n PLCP Preamble/Header November 2004 802.11n PLCP Preamble/Header Legacy portion 100% backward compatible HT portion supports up to four wideband spatial channels using Spatial Spreading (SS) or Eigenvector Steering (ES) PLCP header extension carries scrambler init and rate feedback John Ketchum, et al, Qualcomm

Preamble Legacy Portion November 2004 Preamble Legacy Portion Legacy portion of preamble transmitted using cyclic transmit diversity (no spatial multiplexing or eigenvector steering) Legacy SIGNAL field used to indicate HT Rate field set to unused value indicates HT Size/Request field indicates HT PPDU length. John Ketchum, et al, Qualcomm

HT Portion HT-SIG conveys rates, MIMO training type and length November 2004 HT Portion HT-SIG conveys rates, MIMO training type and length MIMO training can be either steered training or direct training Uses Walsh functions to establish orthogonality among eigenmodes or Tx antennas Uses unique training sequence on each mode or Tx antenna to ensure equal levels at Rx Used by Rx STA to calculate Rx steering Used by Rx STA to calculate Tx steering when using ES John Ketchum, et al, Qualcomm

Legacy and MIMO Training for 2, 3, and 4 Tx November 2004 Legacy and MIMO Training for 2, 3, and 4 Tx STS: 802.11a STS LTS: 802.11a LTS L_SIG: 802.11a SIGNAL HT-SIG: Extended SIGNAL MTSn: MIMO training symbol for Tx antenna n CDx: x ns cyclic delay Shows unsteered MIMO training John Ketchum, et al, Qualcomm

Modulation/Coding/Interleaving November 2004 Modulation/Coding/Interleaving Proposal specifies parallel coding/decoding to support multiple rates in parallel Legacy BCC with extended rates/puncturing patterns to provide expanded MCS set Tail per stream per PPDU– requires parallel decoding for best performance Alternative is tail per stream per OFDM symbol Small increase in overhead, allows single-decoder architecture John Ketchum, et al, Qualcomm

Alternative Modulation/Coding/Interleaving November 2004 Alternative Modulation/Coding/Interleaving Simplified single-decoder architecture Parse/demux must be coordinated with puncture patterns John Ketchum, et al, Qualcomm

Advanced Coding No advanced coding included in proposal November 2004 Advanced Coding No advanced coding included in proposal Advanced coding must support independent rates per stream for eigenvector steering. Single coder/decoder architectures are more feasible with advanced coding such as Turbo codes. John Ketchum, et al, Qualcomm

Summary of MAC Objectives November 2004 Summary of MAC Objectives Enhanced efficiency built on 802.11e Ensure high QoS and high throughput Support MIMO operation with limited overhead Limit introduction of new features Minimize burden on transmit and receive processing John Ketchum, et al, Qualcomm

MAC Throughput vs Range November 2004 MAC Throughput vs Range Throughput above the MAC of 100 Mbps is achieved at: 5.25 GHz : 2x2 – 29 m, 4x4 – 47 m 2.4 GHz: 2x2 - 40 m, 4x4 -75 m. Highest throughput of all proposals. John Ketchum, et al, Qualcomm

MAC Elements Summary Mandatory Enhancements to 802.11e Aggregation November 2004 MAC Elements Summary Mandatory Enhancements to 802.11e Aggregation Frame Aggregation to a single RA. PPDU Aggregation: Reduced or zero IFS Adaptive Coordination Function (ACF) Multi-poll enhancement to HCCA Low latency Data rate feedback from Rx to Tx Enhanced rate adaptation Very low overhead John Ketchum, et al, Qualcomm

Aggregation Significant performance gains at higher date rates: November 2004 Aggregation Significant performance gains at higher date rates: 25-60% greater throughput for PHY rates of 50-100 Mbps Key Attributes: Frame Aggregation to a single RA. PPDU Aggregation: Reduced or zero IFS John Ketchum, et al, Qualcomm

Adaptive Coordination Function November 2004 Adaptive Coordination Function SCAP (Scheduled Access Period) initiated by SCHED message Acts as consolidated multi-STA poll Indicate TA, RA, start offset and duration of TXOP. Permits effective PPDU Aggregation Eliminate Immediate ACK for Block Ack frames MIMO training in SCHED message functions as broadcast sounding waveform for channel estimation and SVD calculation John Ketchum, et al, Qualcomm

Adaptive Coordination Function November 2004 Adaptive Coordination Function Benefits ACF offers 50% to 100% throughput gain over EDCA & HCCA depending on traffic model ACF meets and exceeds QoS requirements with greater efficiency Scenario Throughput (Mbps) Scenario 1 – 5.25 GHz SGI-52 HCF 55.73 Scenario 1 – 5.25 GHz SGI-52 ACF 105.92   Scenario 6 – 5.25 GHz HCF 44.72 Scenario 6 – 5.25 GHz ACF 69.42 John Ketchum, et al, Qualcomm

November 2004 Data Rate Feedback John Ketchum, et al, Qualcomm

November 2004 Data Rate Feedback 16-bit field in PLCP header extension specifies up to four preferred rates Tx PHY rate is maximized after single ACK received Accurate PHY rate tracking for time varying channels Substantial throughput gains Scenario 1: 50% Scenario 6: 38% John Ketchum, et al, Qualcomm

Goals Maximize Throughput, QoS, and Spectral Efficiency Month 2002 doc.: IEEE 802.11-02/xxxr0 November 2004 Goals Maximize Throughput, QoS, and Spectral Efficiency Eigenvector Steering (ES) and rate feedback provide the highest throughput and QoS performance. ES should be an Optional Feature that can provide significant longevity to the 11n standard. Provision for optional ES in 802.11n requires a few mandatory and some specified optional features. John Ketchum, et al, Qualcomm John Doe, His Company

Goals Minimize complexity and assure backward compatibility Month 2002 doc.: IEEE 802.11-02/xxxr0 November 2004 Goals Minimize complexity and assure backward compatibility Builds on 802.11a waveform 20 MHz bandwidth with 802.11a/b/g spectral mask 802.11a modulation, coding, interleaving with expanded rate set Backward compatibility through legacy STF, LTF and SIG Provision for optional ES in 802.11n requires a few mandatory and some specified optional features. John Ketchum, et al, Qualcomm John Doe, His Company

November 2004 Goals Provide balance between TTM needs and 11n design longevity economics Both Spatial Spreading and Spatial Spreading with Space Time Block Coding are good mandatory alternatives that meet TTM objectives ES should be an Optional Feature that can provide significant longevity to the 11n standard. John Ketchum, et al, Qualcomm

Summary Qualcomm proposal builds on existing 802.11a,g,e design November 2004 Summary Qualcomm proposal builds on existing 802.11a,g,e design 802.11n can enable new markets & applications: Multimedia distribution in the home Enhanced enterprise applications (e.g. VoD, Video Conf.) These applications require: High throughput  SS/ES, ACF, rate feedback High QoS  SS/ES, ACF, rate feedback Maximized range  ES Maximum spectral efficiency  ES SS/ES + rate feedback + ACF meet the requirements associated with these new markets & applications: Highest network capacity: greater than 100 mbps above the MAC inside 30 m (20 MHz, 2x2, 5 GHz) Reliable coverage QoS: Less than 50 ms latency with “ZERO packet loss” John Ketchum, et al, Qualcomm