August 2004 doc.: IEEE /0873r0 August 2004

August 2004 doc.: IEEE /0873r0 August 2004 High-Throughput Enhancements for : Features and Performance of QUALCOMM’s Proposal 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 Phone: Fax: John Ketchum, et al, QUALCOMM John Ketchum, et al, QUALCOMM

Agenda Proposal guide and main points MAC Features System Performance
August 2004 doc.: IEEE /0873r0 August 2004 Agenda Proposal guide and main points MAC Features System Performance PHY Features Link Performance John Ketchum, et al, QUALCOMM John Ketchum, et al, QUALCOMM

Guide to Qualcomm’s Proposal
August 2004 Guide to Qualcomm’s Proposal The complete proposal submitted by QUALCOMM consists of the following four documents: High Throughput System Description and Operating Principles. Section 1 provides an overview of the proposed PHY and MAC enhancements Section 2 provides a detailed description and proposed text for the MAC and PLCP enhancements. Section 3 provides a detailed description and proposed text for the PHY enhancements. Appendix A provides the mathematical background and operating principles for MIMO applicable to the proposal. High Throughput Proposal Compliance Statement (this document.) Section 1 addresses compliance with the functional requirements of n. Section 2 addresses compliance with the PAR and Five Criteria of n. Section 3 addresses Comparison Criteria of n. Link Level and System Performance Results for High Throughput Enhancements. Section 1 describes the system simulation methodology Section 2 provides system performance results for the simulation scenarios defined in the n usage models document. Section 3 describes the PHY simulation methodology Section 4 provides link level simulation results for packet error rate and throughput. Section 5 defines the link abstraction used to capture the packet error model in system level simulations and also provides model verification results. Section 6 provides performance results for the modified preamble. High Throughput Enhancements Presentation – Features and Performance. Summary presentation of the proposal features and performance results. PHY Features MAC Features Link Performance System Performance John Ketchum, et al, QUALCOMM

Main Points 20 MHz operation Maximum PHY data rates:
August 2004 Main Points 20 MHz operation Maximum PHY data rates: 202 Mbps for stations with two antennas 404 Mbps for stations with four antennas Backward compatible modulation, coding and interleaving Highly reliable, high-performance operation with existing convolutional codes used in combination with Eigenvector Steering spatial multiplexing techniques Backward compatible preamble and PLCP with extended SIGNAL field. Adaptation of rates and spatial multiplexing mode through low overhead asynchronous feedback. Works with TXOPs obtained through EDCA, HCF or ACF. John Ketchum, et al, QUALCOMM

MAC – Outline Motivation MAC Enhancements: Common Features
August 2004 MAC – Outline Motivation MAC Enhancements: Common Features Scheduled Operation and Adaptive Coordination Function (ACF) QoS Capable IBSS Operation Summary John Ketchum, et al, QUALCOMM

MAC Design Objectives Objectives
August 2004 MAC Design Objectives Objectives Preserve the simplicity and robustness of distributed coordination Backward compatible Enhancements for high throughput, low latency operation Build on e, h feature set: TXOPs, Block Ack, Delayed Block Ack, Direct Link Protocol Dynamic Frequency Selection Transmit Power Control John Ketchum, et al, QUALCOMM

Critical Features for High Throughput Operation
August 2004 Critical Features for High Throughput Operation Critical Features for High Data Rates Adaptation of PHY rates and MIMO transmission mode Low overhead feedback Compatible with EDCA or HCCA Low latency To support PHY adaptation To satisfy end-to-end delay requirements of multimedia/interactive applications High MAC Efficiency, reduced contention overhead Frame aggregation, Compressed Block ACK Enhanced Polling Simplify QoS handling compared to e Exploit high data rates of n John Ketchum, et al, QUALCOMM

Different Operating Environments
August 2004 Different Operating Environments Application to different operating regimes Evolution of current deployments Solution: Simple enhancements: frame aggregation, closed loop rate control Low loads: EDCA High loads: HCCA Large enterprise networks Solution: Enhancements to HCCA for deployments with large numbers of STAs Optimized scheduled operation Implemented in Enterprise-class AP Flexible operation modes. See examples. Small networks with significant QoS traffic Solution: IBSS with distributed round-robin scheduling John Ketchum, et al, QUALCOMM

Flexible Frame Aggregation
August 2004 Flexible Frame Aggregation Eliminate MAC throughput bottleneck Throughput saturates at ~70 Mbps even with e features Permits aggregation of encrypted or unencrypted frames MAC headers in the aggregated frame can be compressed John Ketchum, et al, QUALCOMM

Eliminate Immediate ACK for MIMO Transmissions
August 2004 Eliminate Immediate ACK for MIMO Transmissions Receiver delay for demodulation and decoding of (coded) OFDM transmissions 802.11a SIFS is 16 us. 802.11g provides a 6 us OFDM signal extension MIMO OFDM transmissions impose even greater burden on the receiver Aggregated frames make matters worse Inefficient solution Larger SIFS or longer signal extension Efficient solution Eliminate Immediate ACK for MIMO OFDM transmissions Use e Block ACK and Delayed Block ACK mechanisms Reduced IFS for scheduled transmissions TXOP Bursting with zero IFS (AP transmissions) Consecutive scheduled STA TXOPs separated by GIFS (800 ns Guard IFS) TXOP Bursting with BIFS (STA transmissions) John Ketchum, et al, QUALCOMM

Backward Compatible PLCP Header
August 2004 Backward Compatible PLCP Header PPDU Type 0000 Compatible Preamble Changes described in PHY section Extended Backward Compatible SIGNAL Field Set Rate field in current SIGNAL1 field to one of eight unused values. Indicates presence of SIGNAL2. John Ketchum, et al, QUALCOMM

PPDU Size and LENGTH Fields
August 2004 PPDU Size and LENGTH Fields LENGTH Field in Legacy SIGNAL field is used by the Receiving STA to parse the received octet stream To determine location of FCS, length of PAD. For aggregated frame, need length per encapsulated MAC frame. Aggregation Header contains LENGTH field for each encapsulated MAC frame. PPDU Size Field included in Extended SIGNAL Indicates PPDU Size in number of standard or SGI OFDM symbols. SIGNAL at 6 Mbps can be decoded by all n STAs to determine medium time occupied by the PPDU. Solution: Replace LENGTH by PPDU Size in Extended SIGNAL. Include Aggregation Header whenever MIMO PPDU is used. John Ketchum, et al, QUALCOMM

Rate and Mode Adaptation
August 2004 Rate and Mode Adaptation PPDU Type 0000 Rate vector (DRV) included in Extended SIGNAL field. Rate and mode feedback (DRVF) included in FEEDBACK field MIMO OFDM Training symbols inserted as necessary Fast ramp up to exploit high PHY rates for bursty traffic Enormous throughput benefit with low overhead Robustness to interference, shadowing, channel and receiver impairments John Ketchum, et al, QUALCOMM

August 2004 Compressed Block Ack Compressed format 1: Do not indicate status of fragments. Shrink BlockAck Frame from 152 to 32 octets. Compressed format 2: Indicate status of fragments only if there are missing fragments Compressed format 3: Remove trailing zeroes from Bitmap. John Ketchum, et al, QUALCOMM

Scheduled Operation – SCHED Message
August 2004 Scheduled Operation – SCHED Message SCHED message and Scheduled Access Period (SCAP) are enhancements of HCCA CAP 802.11n AP acquires the medium after PIFS (as in the HCCA CAP) and transmits a SCHED message (instead of Poll). The SCHED message defines the schedule of transmissions for the SCAP. Default values of SCAP: 1.024ms, ms, 4 ms. SCHED is a Multiple Poll Message Lower overhead, more efficient Indicates Tx STA and Rx STA for TXOPs => Improved power saving John Ketchum, et al, QUALCOMM

Scheduled Operation – Protection and Recovery
August 2004 Scheduled Operation – Protection and Recovery Protection of SCAP High level procedures to avoid overlapping BSS: Mandatory DFS CTS-to-Self to clear out NAV for SCAP. For n STAs set NAV through Duration field in SCHED frame. Keep SCAP short (< 4 ms) to minimize impact of collisions with legacy STAs during SCAP No CCA required for transmissions during SCAP John Ketchum, et al, QUALCOMM

Scheduled Operation – Reduced IFS
August 2004 Scheduled Operation – Reduced IFS Reduced IFS Since, no CCA required for transmissions during SCAP PPDU Aggregation: IFS and preambles may be eliminated between consecutive Scheduled AP transmissions. Consecutive Scheduled TXOPs from STAs may be transmitted with GIFS (800 ns) Optionally, FRACH and Protected EDCA may be scheduled during a SCAP John Ketchum, et al, QUALCOMM

Scheduled Operation – Managed Peer-to-Peer
August 2004 Scheduled Operation – Managed Peer-to-Peer PPDU Type 0000 Managed Peer-to-Peer Operation is an enhancement of DLP In Scheduled STA-STA TXOPs PPDU Size in SIGNAL1 is replaced by Request AP promiscuously decodes Request field in STA-STA transmissions. STAs indicate SCHED Rate, QoS and requested length for subsequent TXOP. STAs do closed loop rate control AP does scheduling John Ketchum, et al, QUALCOMM

SCHED Frame Format and Fields
August 2004 SCHED Frame Format and Fields SCHED Frame Fields CTRL0, CTRL1, CTRL2, CTRL3 fields are separately coded and transmitted at 6, 12, 18, 24 Mbps, respectively Multiple Assignment Elements are included in each CTRLJ Each Assignment Element specifies: Tx STA (may be AP), Rx STA (may be AP), Start Time, TXOP Duration MAP field identifies start of FRACH and Protected EDCA within SCAP John Ketchum, et al, QUALCOMM

Summary of HCF Enhancements
August 2004 Summary of HCF Enhancements Advantages of SCHED over HCF Poll Reduced overhead: single message instead of multiple Polls, multiple IFS Efficient encoding of TXOP/RXOP assignments Improved Power Saving: After decoding the SCHED message, STAs not scheduled for Tx or Rx can sleep for the remaining SCAP Efficient feedback for ES operation: MIMO OFDM Training symbols attached to SCHED frame permit STAs to estimate the AP-STA channel and achievable rate. Improved QoS handling: Optimized low-latency operation for n STAs Managed peer-to-peer operation STAs do closed loop rate control. AP does scheduling Protected Contention Periods to complement scheduling FRACH Protected EDCA John Ketchum, et al, QUALCOMM

Operation of Adaptive Coordination Function (ACF)
August 2004 Operation of Adaptive Coordination Function (ACF) SCAP is an enhancement of the HCCA CAP Setting NAV The Duration field in the SIGNAL field of the SCHED frame sets the NAV for the SCAP at all n STAs. To set the NAV for the SCAP at legacy STAs, the AP may transmit a CTS-to-Self prior to the transmission of the SCHED frame. SCAP Timing 802.11n STAs respect the SCAP interval so that their transmissions terminate when the SCAP expires. The AP may schedule back-to-back SCAPs. John Ketchum, et al, QUALCOMM

ACF – Example Operating Mode
August 2004 ACF – Example Operating Mode Case: No CAP Legacy STAs, if present, can satisfy their QoS using EDCA Setting NAV The Duration field in the SIGNAL field of the SCHED PPDU sets the NAV for the SCAP at all n STAs. If only n STAs are present, there is no need for CTS-to-Self. Beacon announces CFP to protect most of the Beacon interval to avoid interference from arriving legacy STAs If medium is shared with legacy STAs, use CTS-to-Self at start of SCAP Interspersed SCAP and EDCA periods permit “fair” sharing of the medium 802.11n QoS Flows are served during SCAP 802.11n non-QoS flows use EDCA periods along with legacy STAs. John Ketchum, et al, QUALCOMM

ACF – Optimized Scheduled Operation
August 2004 ACF – Optimized Scheduled Operation Case: Limited resource required for legacy STAs. Legacy STAs with non-QoS flows that may be satisfied with only occasional allocations of EDCA periods (CP) Setting NAV Beacon sets NAV at legacy STA for CFP. The Duration field in the SIGNAL field of the SCHED frame sets the NAV for the SCAP at all n STAs. Protected EDCA periods for n STAs included in Scheduled Access Period John Ketchum, et al, QUALCOMM

RRBSS – QoS capable IBSS operation
August 2004 RRBSS – QoS capable IBSS operation Provide QoS capability without AP May also be used by low-end AP Applicable to usage scenarios with CE devices with high throughput, high QoS needs Exploit the large PHY data rates of MIMO OFDM to simplify scheduling and QoS management. Designed for up to 15 STAs Distributed admission control Self identification of QoS flows Distributed Round-Robin Scheduling Short Beacon Period for low latency John Ketchum, et al, QUALCOMM

August 2004 RRBSS – Long Token PPDU Long Token PPDU PPDU Type 1010 Round Robin order for current Beacon period included in Complete RR List Up to 15 RRIDs RR Seq indicates changes in RR List Long Token must be transmitted by each STA if RR Seq changes Connectivity Vector indicates RRIDs that the STA can hear Permits clustering of contiguous STAs on RR List RR Bandwidth Management field permits distributed sharing Simple standardized rules John Ketchum, et al, QUALCOMM

RRBSS – Short Token PPDU
August 2004 RRBSS – Short Token PPDU Short Token PPDU PPDU Type 1000 STA must transmit Long or Short Token PPDU if no data to send. Explicit Token Passing using Long or Short Token Implicit Acknowledgment by Next STA Compact RR List contains RRID of STA RRID of Next STA for Token Passing RRID of Last STA on RR List John Ketchum, et al, QUALCOMM

RRBSS – Robust Operation
August 2004 RRBSS – Robust Operation Explicit token passing. Implicit acknowledgment of token passing by Next STA on RR List. Otherwise STA must pass token to the following STA. RR List rotates at each Beacon period No single STA is “designated master” Last STA in Beacon period n, becomes First STA in Beacon period n+1. Transmits Beacon and Long Token in Beacon period n+1. First STA seizes medium at PIFS. If medium is idle at DIFS, previous First STA transmits Beacon and Long Token. Last STA is dropped from RR List. Changes in RR List indicated through RR Seq Each STA must transmit Long Token when RR Seq increments John Ketchum, et al, QUALCOMM

Simplified QoS Handling
August 2004 Simplified QoS Handling In all operating regimes Exploit high data rates to simplify QoS handling Simple admission control Based on simplified TSPEC: Mean data rate, Delay bound Mean data rate Mapped to symbols per second for resource allocation Delay bound Mapped to scheduling period, and ARQ operation Low latency operation is critical To operate with small buffers. This is critical at high data rates. To meet low delay guarantees in all operating regimes EDCA/HCCA with lightly loaded system RRBSS (with or without AP) Scheduled operation for heavily loaded system John Ketchum, et al, QUALCOMM

Summary of MAC Enhancements
August 2004 Summary of MAC Enhancements Detailed design of MAC enhancements for MIMO OFDM Completely Backward Compatible Enhancements required for high throughput, low latency operation Features applicable to different operating regimes List of proposed features Frame Aggregation. Aggregation Header. Eliminate Immediate ACK for MIMO transmissions Extended SIGNAL field and PPDU Type Rate and MIMO Mode Adaptation Compressed Block Ack SCHED Message, SCAP and Scheduled TXOPs Reduced IFS between scheduled transmissions Flexible Operating Modes with ACF RRBSS – QoS capable IBSS Operation. Token PPDUs. John Ketchum, et al, QUALCOMM

System Simulation Methodology
August 2004 doc.: IEEE /0873r0 August 2004 System Simulation Methodology The simulator is based on ns2 Includes physical layer features TGn Channel Models PHY Abstraction determines frame loss events MAC features EDCA Adaptive Coordination Function (ACF): SCHED and SCAP Frame Aggregation ARQ with Block Ack Closed Loop Rate Control (DRVF and DRV) MIMO Modes (ES and SS) Transport File Transfer mapped to TCP QoS Flows mapped to UDP John Ketchum, et al, QUALCOMM John Ketchum, et al, QUALCOMM

Simulation Conditions – Fixed
August 2004 Simulation Conditions – Fixed The following parameters are fixed for all system simulation results. Bandwidth: 20 MHz. Frame Aggregation Fragmentation Threshold: 100 kB Delayed Block Ack Adaptive Rate Control Adaptive Mode Control between ES and SS AC CW min CW max AIFS 127 1023 2 BlockAck/VoIP 1 4 Video HDTV 8 Other QoS 3 10 Best effort John Ketchum, et al, QUALCOMM

Simulation Conditions – Varied
August 2004 Simulation Conditions – Varied The following parameters are varied. Results are provided for different combinations of these parameters. Bands: 2.4 GHz 5.25 GHz MIMO: 2x2: All STAs with 2 antennas 4x4: All STAs with 4 antennas Mixed: Scenario 1: the AP and the HDTV/SDTV displays are assumed to have 4 antennas; all other STAs have 2 antennas. Scenario 6: AP and all STAs, except VoIP terminals have 4 antennas; VoIP terminals have 2 antennas. OFDM symbols Standard: 0.8 μs Guard Interval, 48 data subcarriers SGI-EXP: 0.4 μs Shortened Guard Interval, 52 data subcarriers Access Mechanisms ACF (SCHED) HCF (Poll) EDCA with additional AC for Block Ack John Ketchum, et al, QUALCOMM

Additional Scenarios Scenario 1 HT is an extension of Scenario 1:
August 2004 Additional Scenarios Scenario 1 HT is an extension of Scenario 1: Additonal FTP flow of up to 130 Mbps at 15.6 m from the AP for 2×2. Scenario 1 EXT is an extension of Scenario 1: Maximum delay requirement for all video/audio streaming flows is decreased from 100/200 ms to 50 ms. Two HDTV flows are moved from 5 m from the AP, to 25 m from the AP. Scenario 6 EXT is an extension of Scenario 6: One FTP flow of 2 Mbps at 31.1 m from the AP is increased up to 80 Mbps for 4x4. John Ketchum, et al, QUALCOMM

Summary of Total Throughput Results
August 2004 Summary of Total Throughput Results John Ketchum, et al, QUALCOMM

Observations on Total Throughput
August 2004 Observations on Total Throughput ACF provides highest total throughput compared to HCF and EDCA. ACF satisfies all QoS flows for all Sceanrios when SGI-EXP symbols are used. Only in the case standard symbols are used (giving reduced throughput) at 5.25 GHz (giving reduced range), the PLR requirement of gaming flows is not satisfied. No increase in throughput for EDCA with 4x4 compared to 2x2. Scenario 1 EXT imposes stringent delay (less than 50 ms for streaming) and range (HDTV flows at 25 m) requirements on QoS flows. When 2x2 is used, one or two QoS flows are not satisfied. In the Mixed case, by equipping the AP and the HDTV and SDTV displays with 4 antennas, all QoS flows except the gaming flow are satisfied with an almost 50% increase in total throughput compared to 2x2. In Scenario 4, throughput achieved is over 100 Mbps with 2x2 and almost 200 Mbps with 4x4. Scenario 6 EXT Mixed case (mixture of 4-antenna and 2-antenna STAs) gives higher TCP throughput than the 4x4 case. This is because there is more time available for TCP flows due to the reduced training sequence overhead for VoIP STAs with 2 antennas compared to VoIP STAs with 4 antennas. Sceanrio 6 EXT has 30 VoIP flows. John Ketchum, et al, QUALCOMM

Summary of MAC Efficiency Results
August 2004 Summary of MAC Efficiency Results As defined, MAC Efficiency is meaningful only when the offered load for a scenario exceeds the carried load and there is always backlogged traffic at some flow. In the above table, the MAC Efficiency numbers are shown in red for the cases where the medium is forced idle due to no backlog. These numbers are not meaningful. John Ketchum, et al, QUALCOMM

Observations on MAC Efficiency
August 2004 Observations on MAC Efficiency For 2x2, the MAC Efficiency for ACF is between For 2x2, the MAC Efficiency for HCF and EDCA is around 0.5. For 4x4, the MAC Efficiency for HCF and EDCA reduces to 0.4 and 0.2, respectively. ACF manages to sustain a MAC Efficiency around 0.6, even with 4x4. John Ketchum, et al, QUALCOMM

Summary of QoS Flows Satisfied
August 2004 Summary of QoS Flows Satisfied John Ketchum, et al, QUALCOMM

Observations on QoS Flows
August 2004 Observations on QoS Flows Scenario 1 EXT imposes stringent delay (less than 50 ms for streaming) and range (HDTV flows at 25 m) requirements on QoS flows. When 2x2 is used, one or two QoS flows are not satisfied. In the Mixed case, by equipping the AP and the HDTV and SDTV displays with 4 antennas, all QoS flows except the gaming flow are satisfied with an almost 50% increase in total throughput compared to 2x2. More QoS flows are satisfied with HCF than with EDCA. However, ACF is required to address stringent QoS requirements. QoS for uplink EDCA VoIP flows is not satisfied. All QoS Flows are satisfied for Scenario 4. John Ketchum, et al, QUALCOMM

Summary of non-QoS Flow Throughput
August 2004 Summary of non-QoS Flow Throughput Scenario 6 EXT Mixed case (mixture of 4-antenna and 2-antenna STAs) gives higher TCP throughput than the 4x4 case. This is because there is more time available for TCP flows due to the reduced training sequence overhead for VoIP STAs with 2 antennas compared to VoIP STAs with 4 antennas. Sceanrio 6 EXT has 30 VoIP flows. John Ketchum, et al, QUALCOMM

Throughput versus Range for Channel Model B
August 2004 Throughput versus Range for Channel Model B John Ketchum, et al, QUALCOMM

Throughput versus Range for Channel Model D
August 2004 Throughput versus Range for Channel Model D John Ketchum, et al, QUALCOMM

Observations on Throughput versus Range
August 2004 Observations on Throughput versus Range The plots for Channel Model B and Channel Model D are roughly similar. Throughput above the MAC of 100 Mbps is achieved at: 29 m for 2x2, 5.25 GHz 40 m for 2x2, 2.4 GHz 47 m for 4x4, 5.25 GHz 75 m for 4x4, 2.4 GHz John Ketchum, et al, QUALCOMM

Qualcomm 802.11n PHY Design Fully backward compatible with 802.11a/b/g
August 2004 Qualcomm n PHY Design Fully backward compatible with a/b/g 20 MHz bandwidth with a/b/g spectral mask OFDM based on a waveform with additional expanded OFDM symbol and shortened guard interval Modulation, coding, interleaving based on a Expanded rate set Scalable MIMO architecture Supports a maximum of 4 wideband spatial streams Two forms of spatial processing Eigenvector Steering (ES): via wideband spatial modes/SVD per subcarrier Tx and Rx steering Over the air calibration procedure required Spatial Spreading (SS): modulation and coding per wideband spatial channel No calibration required SNR per wideband spatial stream known at Tx Sustained high rate operation possible via rate adaptation John Ketchum, et al, QUALCOMM

August 2004 Observation Detailed, up-to-date feedback on channel state is fundamental to achieving high throughput in a TDD MIMO WLAN The challenge is to achieve this reliably with low overhead We believe that the design described here achieves this goal John Ketchum, et al, QUALCOMM

OFDM Waveform Baseline OFDM structure identical to 802.11a/g
August 2004 OFDM Waveform Baseline OFDM structure identical to a/g 312.5 kHz subcarrier spacing/20 MHz carrier spacing Same subcarrier structure 48 subcarriers for data, 4 subcarriers for pilot “DC” subcarrier empty, 11 subcarriers for guard band 3.2 µs symbol, 800 ns guard interval 40% Physical-layer overhead (1 - (48*312.5 kHz/20 MHz)*(3.2 µs/4 µs)) Expanded OFDM symbol and shortened cyclic prefix introduced Same subcarrier spacing 4 additional data subcarriers—52 total same four pilot subcarriers Physical-layer overhead < 28% More vulnerable to time dispersion and ACI John Ketchum, et al, QUALCOMM

August 2004 Modulation and Coding Use existing constraint length 7, rate ½ convolutional code and punctured rates. Retain PSK/QAM modulation from Additional rates adopted to provide increased spectral efficiency 256 QAM modulation gives increased rates and spectral efficiency Code rates range from ½ bit per modulation symbol to 7 bits per modulation symbol. Up to four wideband spatial channels supported with separate coding/interleaving for each spatial channel. Enhanced interleaving over single OFDM symbol for MIMO OFDM Based on a/g interleaver Simple backward compatible mode John Ketchum, et al, QUALCOMM

Code Rates and Modulation
August 2004 Code Rates and Modulation Bits/subcarrier Bit/s/spatial chan1 Bit/s/spatial chan2 Code Rate Modulation 0.50 6 Mbit/s 7.2 Mbit/s r=1/2 BPSK 0.75 9 10.8 r=3/4 1.00 12 14.4 QPSK 1.50 18 21.7 2.00 24 28.9 16 QAM 2.50 30 36.1 r=5/8 3.00 36 43.3 3.50 42 50.6 r=7/12 64QAM 4.00 48 57.8 r=2/3 4.50 54 65 5.00 60 72.2 r=5/6 256 QAM 6.00 72 86.7 7.00 84 101.1 r=7/8 Notes: 1) short OFDM symbols; 2) expanded OFDM symbols with short guard interval John Ketchum, et al, QUALCOMM

August 2004 Spatial Processing Two forms of Spatial Processing for data transmission Eigenvector Steering (ES): Tx attempts to steer optimally to intended Rx Spatial Spreading (SS): Tx does not attempt to steer optimally to specific Rx ES operating modes take advantage of channel reciprocity inherent in TDD systems Full MIMO channel characterization required at Tx Calibration procedure required 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 data rate and range John Ketchum, et al, QUALCOMM

Spatial Channels and Spatial Streams
August 2004 Spatial Channels and Spatial Streams ES and SS approaches result in synthesis of spatial channels, or wideband spatial channels. Also referred to as eigenmodes, or wideband eigenmodes On MIMO channel between a transmitting STA with NTx antennas and a receiving STA with NRx antennas, maximum of wideband spatial channels available. Each resulting spatial channel may carry a payload, referred to as a spatial stream. Number of spatial streams, NS, may not be greater than the Nm John Ketchum, et al, QUALCOMM

Spatial Processing Tools
August 2004 Spatial Processing Tools Basic spatial processing techniques used in different combinations to maximize throughput, range, and reliability under a wide range of conditions Cyclic transmit diversity per Tx antenna Orthogonal cover across symbols and spatial channels Spatial spreading with simple orthogonal matrices Eigenvector steering to synthesize wideband eigenmodes Eigenvector Steering simplifies processing burden of AP support of many STAs Spatial Spreading allows STAs without full channel characterization to achieve high throughput without Tx steering ~80% of the throughput of calibrated modes with simple Rx processing Increased SNR variance across subcarriers in a spatial channel reduces effectiveness of legacy convolutional codes John Ketchum, et al, QUALCOMM

Over-the-Air Calibration
August 2004 Over-the-Air Calibration ES approach requires over-the-air calibration procedure Compensates for amplitude and phase differences in Tx and Rx chains Calibration required infrequently– typically on association only Simple exchange of calibration symbols and measurement information requires little overhead and background processing Total of ~1000 bytes of calibration data exchanged for 2x2 link ~2800 bytes for 4x4 link John Ketchum, et al, QUALCOMM

Feedback for ES and SS Modes
August 2004 Feedback for ES and SS Modes Rate adaptation Receiving STA determines preferred rates on each of up to four wideband spatial channels One rate per wideband spatial channel – NO adaptive bit loading Sends one 4-bit value per spatial channel, differentially encoded, (13 bits total) to inform corresponding STA/AP of rate selections Corresponding STA/AP uses this info to select Tx rates Piggy-backed on out-going PPDUs SS Mode can use single rate across all spatial channels Channel state information For ES operation, Tx must have full channel state information This is obtained through exchange of transmitted training sequences that are part of PLCP header Very low overhead. Distributed computation of steering vectors (SVD calculation) STAs do SVD, send resulting training sequence to AP For SS operation, unsteered training sequences transmitted in PLCP header to support channel estimation at receiver John Ketchum, et al, QUALCOMM

Feedback Operates with Asynchronous MAC Transmissions
August 2004 Feedback Operates with Asynchronous MAC Transmissions TXOPs obtained through EDCA, HCCA, or enhanced HCCA Transmitting STA sends steered or unsteered training sequences in each TXOP If operating in ES mode, receiving STA uses received training sequences to calculate transmit and receive steering vectors If operating in SS mode, receiving STA uses received training sequences to determine Rx processing Receiving STA estimates rates per wideband spatial stream and includes in feedback Transmitting STA determines age of Tx steering vectors and falls back to SS mode if vectors are too old Transmitting STA determines age of rate feedback and backs off Tx rates if feedback is too old John Ketchum, et al, QUALCOMM

Supported Antenna Configurations
August 2004 Supported Antenna Configurations Scalable deployments 2x2 → 4x4 Support for up 4 wideband spatial channels Typical max antenna configuration is four antennas per STA Can support more than four antennas on a STA, but without adding spatial channels May provide increase range or throughput through diversity/steering gains STAs may have any number of antennas STAs in network may have fewer antennas Maximum spatial channels available on a link between two nodes is limited by STA with fewest antennas John Ketchum, et al, QUALCOMM

Wideband Eigenmodes and OFDM
August 2004 Wideband Eigenmodes and OFDM OFDM chosen so that subcarrier spacing << coherence bandwidth Find ranked eigenmodes/eigenvalues in each OFDM subcarrier: Ensemble of eigenmodes of a given rank across OFDM symbol comprise a wideband eigenmode Highest ranked wideband eigenmodes exhibit very little frequency selectivity Smallest ranked wideband eigenmode exhibits frequency selectivity of underlying channel John Ketchum, et al, QUALCOMM

Wideband Eigenmodes TGn Channel B
August 2004 Wideband Eigenmodes TGn Channel B Power is relative to average total receive power at a single antenna John Ketchum, et al, QUALCOMM

Wideband Eigenmodes TGn Channel E
August 2004 Wideband Eigenmodes TGn Channel E Power is relative to average total receive power at a single antenna John Ketchum, et al, QUALCOMM

Spatial Spreading: Partial CSI Spatial Multiplexing
August 2004 Spatial Spreading: Partial CSI Spatial Multiplexing Transmitter is partially informed No explicit knowledge of channel or channel eigenvectors at Tx Tx has only data rate per wideband spatial channel Primary objectives Transmit full power regardless of the number of streams Tx’d Requirement for robust CSMA operation Maximize diversity per transmitted data stream Minimize outage probability  maximize throughput Backwards compatible operation Basic Concept Spatial spreading of data with simple unitary matrices Cyclic diversity transmission per Tx antenna to introduce additional diversity John Ketchum, et al, QUALCOMM

Steering for Spatial Spreading
August 2004 Steering for Spatial Spreading Tx data vector in OFDM subcarrier k, s(k), preconditioned with orthogonal “spreading” matrix, W For Nt = 2 or 4  W is Hadamard matrix (real Walsh functions) For Nt = 3  W is Fourier matrix (complex-valued Fourier functions) Transmit vector is Random steering introduces full Tx diversity per stream Each stream is transmitted out all Nt antennas Full Tx power is used, regardless of the number of streams, NS, transmitted If NS < Nm, spreading matrix is reduced to NS columns John Ketchum, et al, QUALCOMM

Cyclic Diversity Transmission
August 2004 Cyclic Diversity Transmission Each Tx antenna introduces a different cyclic delay Creates linear phase shift across OFDM subcarriers per antenna Each spatial stream is subjected to frequency selective fading across the subcarriers maximizes spatial diversity per spatial stream No phase discontinuities introduced from subcarrier to subcarrier minimizes degradation in legacy STAs channel estimation John Ketchum, et al, QUALCOMM

Spatial Spreading + Cyclic Delay
August 2004 Spatial Spreading + Cyclic Delay Random steering matrix, W, is transformed by phase shift matrix I is the cyclic shift increment Transmit vector is Equivalent to cyclic shift of Ws(k) in time domain John Ketchum, et al, QUALCOMM

Tx Functional Block Diagram
August 2004 Tx Functional Block Diagram John Ketchum, et al, QUALCOMM

TDD Reciprocal Channel
August 2004 TDD Reciprocal Channel In a TDD MIMO system, the over-the-air portion of the channel is reciprocal The up-link channel, , (entity A to entity B) is the transpose of the down-link channel, , ( is the OFDM subcarrier index): Due to gain differences in Tx and Rx chains at both ends of the link, the baseband-to-baseband channel is not reciprocal. The observed channel is weighted by two diagonal matrices with the complex gains of the transmit and receive chains: These gain differences can be removed with a simple over-the-air calibration procedure that learns the gain matrices Result is a very stable calibrated reciprocal channel John Ketchum, et al, QUALCOMM

Calibration Procedure
August 2004 Calibration Procedure Find diagonal calibration matrices that can be applied to transmit vectors to compensate for amplitude/phase variations between Rx and Tx chains in device Calibration required once per session; i.e., upon association Procedure as follows: Entity A (typically a STA) observes MIMO pilot from entity B (typically an AP) Entity A forms an estimate of channel, Entity A transmits MIMO pilot, which entity B observes Entity B forms channel estimate, Entity B transmits to A Requires transmitting bit values 624 B for 2x2; B for 4x4 John Ketchum, et al, QUALCOMM

Calibration Procedure
August 2004 Calibration Procedure Entity A now has both Entity A now solves for diagonal calibration matrices such that Entity A sends to entity B, then both ends of link have calibration matrix Requires transmitting bit values 624 B for 4 antennas; 312 B for 2 antennas Calibration matrices are incorporated into Tx steering vectors. John Ketchum, et al, QUALCOMM

Wideband Eigenmodes on TDD Reciprocal Channel
August 2004 Wideband Eigenmodes on TDD Reciprocal Channel Uplink channel SVD: Tx steering matrix: Rx matched filter: Downlink SVD: : downlink Tx steering matrix Transmit steering vectors at one end of the link can be computed directly from the receive matched filter at the same end of the link Normalize and conjugate Since Tx steering vectors can be obtained directly from Rx matched filter, eigenvectors only need to be computed at one end of the link John Ketchum, et al, QUALCOMM

Channel estimation for Wideband Eigenmodes
August 2004 Channel estimation for Wideband Eigenmodes Two kinds of training sequences: MIMO Training Sequence: orthogonal pilot is transmitted on each antenna allowing the receiver to directly form an estimate of the channel matrix, H(k), in each OFDM subcarrier. Steered MIMO Training Sequence: orthogonal pilot is transmitted on each eigenmode, allowing the receiver to directly form an estimate of the received matched filter, M(k) in each OFDM subcarrier. Some definitions: p(k): pseudo-random sequence across OFDM subcarriers (unique word) w(n); n[0, Ntx-1]: vector of length Ntx orthogonal sequences (n is index over time) w(n), 0≤n≤Ntx-1 are columns of Hadamard matrix for Ntx= 2,4; or Fourier matrix for Ntx= 3 John Ketchum, et al, QUALCOMM

MIMO Training Sequences
August 2004 MIMO Training Sequences Common MIMO Training Sequence: MIMO Training Sequence sent by AP as part of a control message containing scheduling information Contains 0,2,3,4 MIMO OFDM training symbols Number of training symbols equals number of Tx antennas For single-antenna AP (if allowed), long PLCP preamble serves as training sequence, no MIMO training sequence required Occurs immediately after PLCP Preamble and SIGNAL field Dedicated MIMO Training Sequence: STA sends steered MIMO Training Sequence as part of header of every PPDU Number of steered MIMO OFDM training symbols = number of Tx antennas For single-antenna station (if allowed), long training sequence serves as dedicated training sequence. John Ketchum, et al, QUALCOMM

Channel estimation: MIMO Training Sequence
August 2004 Channel estimation: MIMO Training Sequence Transmitted waveform is Ntx vector OFDM symbols w/orthogonal cover, transformed by cyclic shift matrix: Received waveform is Calculate the channel estimate by correlating with orthogonal sequence: John Ketchum, et al, QUALCOMM

Channel estimation: Steered Training Sequence
August 2004 Channel estimation: Steered Training Sequence Steered MIMO OFDM Training Sequence Transmitted waveform is NTx vector OFDM symbols on eigenmodes w/orthogonal cover and cyclic shift: Received waveform is Calculate the channel estimate by correlating with orthogonal cover and integrating: Then the estimated spatial matched filter is John Ketchum, et al, QUALCOMM

Use of Training Sequences in AP-centric System
August 2004 Use of Training Sequences in AP-centric System AP transmits MIMO training sequence at the beginning of each SCAP This is in addition to legacy training sequences STAs receive MIMO training sequence and compute channel estimate STA computes transmit steering vectors via eigen-analysis or SVD on an as-needed basis, but not more frequently than every 2 ms Up-to-date channel estimate and SVD always available at STA When an STA transmits a PPDU, steered MIMO training sequence (aka dedicated Pilot) is included in the preamble AP estimates Rx steering from steered MIMO training sequence AP also derives Tx steering vectors from received steered MIMO training sequence SVD calculation at AP not necessary No need for AP to perform SVD for all associated STAs When AP transmits a PPDU, includes steered MIMO training sequence If AP does not have recent steered MIMO training sequence from STA, reverts to non-eigensteered mode. John Ketchum, et al, QUALCOMM

Use of Training Sequences in Peer-to-Peer Mode
August 2004 Use of Training Sequences in Peer-to-Peer Mode One end of link plays role of AP Sends MIMO training sequence and possibly steered MIMO training sequence Other end plays role of STA Sends steered MIMO training sequence only Training sequences are included as part of PLCP headers Low duty-cycle exchanges revert to non-eigensteered mode John Ketchum, et al, QUALCOMM

August 2004 Rate Adaptation Rate selection performed based on post-detection SNR per stream Post-detection SNR per stream is unique per subcarrier Ensemble of SNRs per stream across subcarriers used to drive rate selection Independent coding for each of up to NTx wideband spatial modes Code rate (modulation + coding) assigned based on observed SNRs, etc., across wideband spatial mode Single rate across all wideband spatial channels in SS mode. Rate decisions communicated via short rate control words (13 bits—differential encoding of rates for each for up to four modes) Transmitted rates indicated via Data Rate Vector (DRV) in SIGNAL field Receiver makes rate selection based on observation of received signal, and communicates result to transmitter at other end of link via DRVF (DRV feedback) in feedback field in data segment John Ketchum, et al, QUALCOMM

PLCP Preamble Standard 802.11a preamble with enhancements
August 2004 PLCP Preamble Standard a preamble with enhancements Last short preamble symbol is inverted to provide improved timing reference Cyclic delay is applied across Tx antennas Cyclic delay applied to entire 8 µs short preamble Cyclic delay applied to entire 8 µs long preamble Delay increment Tcd=50 ns John Ketchum, et al, QUALCOMM

Summary MIMO PHY design builds on existing 802.11a,g PHY design
August 2004 Summary MIMO PHY design builds on existing a,g PHY design Two operating modes provide highly robust operation under a wide range of conditions Eigenvector Steering provides best rate/range performance Spatial Spreading Adaptive rate control through low-overhead rate feedback supports sustained high throughput operation Low-overhead training sequence exchange supports high-capacity Eigenvector Steered operation for best rate/range performance Spatial Spreading operation provides robust high throughput operation when Tx does not have sufficiently accurate channel state information John Ketchum, et al, QUALCOMM

Simulation of Spatial Multiplexing Using Tx & Rx Eigensteering
August 2004 Simulation of Spatial Multiplexing Using Tx & Rx Eigensteering Common MIMO Training Sequence broadcast by AP once every SCAP (Scheduled Access Period) (…,t0,t3,…). Forward link (FL) channel coefficients estimated by STA receiver FL Dedicated MIMO Training Sequence (steered) transmitted by AP at t1=0.5 ms, immediately followed by FL data PPDU Reverse link (RL) Dedicated MIMO Training Sequence transmitted by STA at t2=1.5 ms, immediately followed by RL data PPDU Transmit and receive steering vectors derived from most recent channel estimates Closed-loop rate adaptation: FL and RL data rates determined based on receive SNRs observed in previous frames John Ketchum, et al, QUALCOMM

Simulation Sequence – Eigenvector steering
August 2004 Simulation Sequence – Eigenvector steering Channel estimation and rate control occurs over three SCAPs (2.048 ms each) First SCAP: STA receives Common MIMO Training Sequence Computes channel estimate Transmits Dedicated MIMO Training Sequence (steered) AP receives Dedicated MIMO Training Sequence Computes estimate of receive & transmit steering vectors Second SCAP: AP sends Dedicated MIMO Training Sequence (steered) STA makes FL rate selection based on received Dedicated MIMO Training Sequence STA sends Dedicated MIMO Training Sequence (steered) with DL rate selection AP makes RL rate selection based on received dedicated MIMO Training Sequence Third SCAP: AP sends data PPDU based on rate choice in previous SCAP Includes RL rate selection STA sends data PPDU SCAP = Scheduled Access Period John Ketchum, et al, QUALCOMM

Simulation Sequence – Spatial Spreading
August 2004 Simulation Sequence – Spatial Spreading First SCAP: STA receives Common MIMO Training Sequence from AP and computes channel estimate at t0. Based on the received training sequence, the STA makes FL rate selection. STA sends unsteered Dedicated MIMO Training Sequence on RL at t2 with FL rate selection. Based on the received training sequence, the AP makes RL rate selection. Second SCAP: STA receives Common MIMO Training Sequence from AP and computes receive vectors based on channel estimate at (2.048 ms + t0). AP sends data PPDU + RL rate selection on FL at (2.048 ms + t1) based on rate choice made in first SCAP. The STA’s receive vectors are based on the previously received training sequence. STA sends unsteered Dedicated MIMO Training Sequence followed by data PPDU on RL at (2.048 ms + t2 ) with data rates selected in first SCAP. The AP’s receive vectors are based on the received training sequence. John Ketchum, et al, QUALCOMM

Simulation Parameters
August 2004 Simulation Parameters 2x2, 4x2, and 4x4 system configurations IEEE n channel models B, D and E IEEE n impairment models: Time-domain channel simulator with 5x oversampling rate (Ts=10 ns) Rapp nonlinear power amplifier model (IM1): Total Tx power = 17 dBm; Psat = 25 dBm 2x2 backoff = 11 dB per PA; 4x4 backoff = 14 dB per PA Carrier frequency offset : PPM (IM2) Sampling clock frequency offset: PPM (IM2) Phase noise at both transmitter and receiver (IM4) 100 channel realizations generated for each SNR point In each channel realization the Doppler process evolves over three SCAPs to allow simulation of channel estimation, closed-loop rate adaptation and FL/RL data transmission in fading conditions Stopping criterion: 10 packet errors or 400 packets transmitted per channel realization Targeted packet error rate performance: mean PER <= 1% John Ketchum, et al, QUALCOMM

PHY Simulation Results
August 2004 PHY Simulation Results What we simulated Standard OFDM symbols Eigenvector Steering Spatial Spreading Expanded OFDM symbols (52 data tones/400ns guard interval: SGI-52) PER vs SNR for Fourier channel 1×1, 2×2, 3×3, and 4×4 (CC59) All above cases PHY throughput and PER vs SNR; CDFs of throughput and PER Standard OFDM symbols, ES & SS 2×2, 4×4, and 4×2 Channels B, D, and E SGI-52 OFDM symbols, ES & SS Channel B John Ketchum, et al, QUALCOMM

PHY Simulation Results (2)
August 2004 PHY Simulation Results (2) Average PER vs SNR Standard OFDM symbols, ES & SS 2×2, 4×4, and 4×2 Channels B, D, and E John Ketchum, et al, QUALCOMM

Highlights of PHY Simulation Results
August 2004 Highlights of PHY Simulation Results Highest PHY throughputs achieved in Eigenvector Steering mode Eigenvector steering is the very effective in combination with convolutional codes 256-QAM contributes substantially to throughput in ES mode. ES overcomes effects of phase noise in these cases Convolutional codes not as effective in Spatial Spreading mode High SNR variance across subcarriers within an OFDM symbol on an SS spatial channel degrades the performance of convolutional codes This is particularly pronounced on channel B and on link with 4 Tx and 2 Rx. Reducing number of streams (NS < min(NTx,NRx)) reduces variance and improves overall performance. Rate adaptation has clearly demonstrated benefits Many cases where a given fixed rate has poor performance, but using rate adaptation, higher overall throughput is achieved with lower PER Part of rate adaptation is controlling the number of streams used John Ketchum, et al, QUALCOMM

Highlights of PHY Simulation Results
August 2004 Highlights of PHY Simulation Results Use of shortened guard interval and extra data subcarriers contributes to increased throughput Increased vulnerability to delay spread and ACI. Improved receiver design should help with this Optional mode can be turned off in the presence of too much delay spread Many environments where high rates will be required, such as residential media distribution, have naturally low delay spread. John Ketchum, et al, QUALCOMM

CC59 Performance: ES & Standard OFDM
August 2004 CC59 Performance: ES & Standard OFDM John Ketchum, et al, QUALCOMM

CC59 Performance: SS & Standard OFDM
August 2004 CC59 Performance: SS & Standard OFDM John Ketchum, et al, QUALCOMM

CC59 Performance: ES & SGI-52 OFDM
August 2004 CC59 Performance: ES & SGI-52 OFDM John Ketchum, et al, QUALCOMM

PHY Throughput and PER Ch. B, 2×2: Eigenvector Steering
August 2004 PHY Throughput and PER Ch. B, 2×2: Eigenvector Steering John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. B, 2×2 : Eigenvector Steering
August 2004 Average PER w/fixed rates Ch. B, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Throughput and PER Statistics Ch. B, 2×2 : Eigenvector Steering
August 2004 Throughput and PER Statistics Ch. B, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

PHY Throughput and PER Ch. B, 2×2: Spatial Spreading
August 2004 PHY Throughput and PER Ch. B, 2×2: Spatial Spreading John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. B, 2×2: Spatial Spreading
August 2004 Average PER w/fixed rates Ch. B, 2×2: Spatial Spreading John Ketchum, et al, QUALCOMM

Throughput and PER Statistics Ch. B, 2×2: Spatial Spreading
August 2004 Throughput and PER Statistics Ch. B, 2×2: Spatial Spreading John Ketchum, et al, QUALCOMM

PHY Throughput and PER Ch. B, 2×2: Eigenvector Steering, SGI-52
August 2004 PHY Throughput and PER Ch. B, 2×2: Eigenvector Steering, SGI-52 John Ketchum, et al, QUALCOMM

Throughput and PER Statistics Ch. B, 2×2: Eigenvector Steering, SGI-52
August 2004 Throughput and PER Statistics Ch. B, 2×2: Eigenvector Steering, SGI-52 John Ketchum, et al, QUALCOMM

PHY Throughput and PER Ch. B, 4×4 : Eigenvector Steering
August 2004 PHY Throughput and PER Ch. B, 4×4 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. B, 4×4 : Eigenvector Steering
August 2004 Average PER w/fixed rates Ch. B, 4×4 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. B, 4×4: Spatial Spreading
August 2004 Average PER w/fixed rates Ch. B, 4×4: Spatial Spreading John Ketchum, et al, QUALCOMM

PHY Throughput and PER Ch. B, 4×2 : Eigenvector Steering
August 2004 PHY Throughput and PER Ch. B, 4×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

PHY Throughput and PER Ch. D, 2×2 : Eigenvector Steering
August 2004 PHY Throughput and PER Ch. D, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. D, 2×2 : Eigenvector Steering
August 2004 Average PER w/fixed rates Ch. D, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Throughput and PER Statistics Ch. D, 2×2 : Eigenvector Steering
August 2004 Throughput and PER Statistics Ch. D, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. D, 4×4 : Eigenvector Steering
August 2004 Average PER w/fixed rates Ch. D, 4×4 : Eigenvector Steering John Ketchum, et al, QUALCOMM

PHY Throughput and PER Ch. E, 2×2 : Eigenvector Steering
August 2004 PHY Throughput and PER Ch. E, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. E, 2×2 : Eigenvector Steering
August 2004 Average PER w/fixed rates Ch. E, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Throughput and PER Statistics Ch. E, 2×2 : Eigenvector Steering
August 2004 Throughput and PER Statistics Ch. E, 2×2 : Eigenvector Steering John Ketchum, et al, QUALCOMM

Average PER w/fixed rates Ch. E, 4×4 : Eigenvector Steering
August 2004 Average PER w/fixed rates Ch. E, 4×4 : Eigenvector Steering John Ketchum, et al, QUALCOMM

August 2004 doc.: IEEE /0873r0 August 2004

Similar presentations

Presentation on theme: "August 2004 doc.: IEEE /0873r0 August 2004"— Presentation transcript:

Similar presentations

About project

Feedback

Log in

Auth with social network:

August 2004 doc.: IEEE /0873r0 August 2004

Similar presentations

Presentation on theme: "August 2004 doc.: IEEE /0873r0 August 2004"— Presentation transcript:

Similar presentations

About project

Feedback