Transmit processing: a viable scheme for MIMO-OFDM in 802.11n André Bourdoux Bart Van Poucke Liesbet Van der Perre IMEC, Wireless Research bourdoux@imec.be

Need for 4G High-Speed WLANs 5 m 1 Mbps 10 Mbps 100 Mbps 1 Gbps Maximum Data rate 100 kbps 50 m 500 m Range 1G WLAN 802.11 1-2 Mbps 3G WLAN 802.11a/g 6-54 Mbps 4G WLAN 802.11n > 100 Mbps 2G WLAN 802.11b 5.5-11 Mbps IEEE802.11n calls for 100 Mbps at the MAC SAP in a 20 MHz band MIMO is seen as a ley technology to meet this requirement MIMO techniques can result in rate increase and/or range increase Higher data rates Larger range More users

MIMO-OFDM boosts performances in frequency selective environments Higher capacity SDM, SDMA MIMO offers: Higher robustness Diversity (MRC, STBC) Frequency Magnitude The MIMO advantages: - Multiplexing (either SDM or SDMA) allows dramatic increase in capacity. Multiple antennas are needed at both ends of the link. Theoretical capacity improvement is approximately equal to min(NT,NR), hence linear in the number of antennas. - Diversity (Combining diversity such as Maximum Ratio Combining or transmit diversity such as Space-time Block Coding) allows to improve the link budget by combining signals travelling over independently fading paths. Multiple antennas are needed at one side of the link only. A single stream is transmitted. There is still some capacity gain as the SNR improves, but the gain is only a logarithmic in the number of antennas.

A smart MIMO system adapts to scene and actual user needs SDMA multiplies cell capacity SDM brings higher throughput in DL/UL MIMO-TX ! MRC brings robustness STBC brings robustness MIMO transmission can be used in an intelligent (adaptive) manner: different user scenarios exist: single user, multi-user, file transfer, multimedia, upstream, downstream, etc.. Different environment: corporate, home, malls, … Different user termanals classes and needs: single antennna, multi-antenna, power- or not power-limited, legacy terminals, etc… MIMO-TX can play a significant role for SDMA downlink and TX-MRC for legacy terminals SDM downlink for multi-antenna terminals

A wide variety of MIMO schemes are available “MIMO-TX” schemes, focus of this presentation SDM s1 s2 TX-SDM SDMA TX-SDMA MRC TX-MRC STBC stbc Downlink H SDMA s1 s2 RX-SDMA SDM RX-SDM MRC RX-MRC STBC Uplink MIMO-TX schemes are not the only MIMO mode. A variety of MIMO modes is possible and should be included in 11n. The rest of this presentation focuses on specific topics for MIMO-TX NB: RX-SDMA is not recommended because it requires a complicated pre-synchronization of the transmitters so that the OFDM symbols from different transmitters arrive with good time alignment at the AP antennas. This is especially acute in environments with large delay spreads.

MIMO-TX requires “TX-CSI” … H S1 MIMO TX SDM(A) MIMO with TX pre-processing SN h S1 H MIMO TX MRC … MIMO with TX pre-processing Simpler receiver (no MIMO processing needed, SISO equalizer needed) Transmitter processing: low complexity (Zero-forcing, MMSE, MRC) Channel H needed at TX ( “TX-CSI” ) Option 1: by feedback Option 2: from channel estimation in reverse link, reciprocity needed only works for TDD set-ups Reciprocity: Propagation channel: OK if delay kept short Front-ends: not OK, calibration needed to measure transfer function of the front-ends MIMO-TX simplifies the receiver processing. The SISO equalizer is still needed to compensate for scaling and rotation of the constellation at the receiver. For SDM, suitable schemes are zero-forcing or MMSE. This is just a linear processing (matrix multiplication). Non-linear schemes, dual of the successsive interference cancellation for MIMO-RX schemes, also exist but require cooperation of the receiver. Hence, they are not suitable for SDMA with legacy terminals The rest of the slide is self-explanetory or detailed in the next slides

Example of Channel estimation and TX processing DL 2 s2 AP s1 UT1 UT2 DL 1 UL1 AP UT1 UL2 AP UT2 Spatial dimension Channel estimation, AGC couples H Time Spatial pre-filter Example of Channel estimation procedure for MIMO-TX processing: The intended receivers are the UTs in this example Short uplink bursts from the UTs are used to estimate the channel at the AP side. IF the UTs are single antenna UTs (SDMA) then each UT transmits a burst in turn. If the UT is multi-antenna (SDM), a single burst is transmitted. The AGC coefficients used during the channel estimation must be known (4 coefficients for SDMA, 2 coefficients for SDM). The estimated channel and the front-end calibration values (including AGC) are then used to compute the TX pre-filter, MMSE in this example Parallel SDM or SDMA transmission can then take place. H with FE. C.

Front-end non-reciprocity can be solved by calibration The channel includes: the propagation channel ( H ) front-end circuits (filters, etc..) linear and non linear ( Dxx,yy ) Base Station Receiver Front-end From modulator To demodulator Terminal Transmitter Front-end Terminal Receiver Front-end Base Station Transmitter Front-end H DTX,AP DTX,MT DRX,AP DRX,MT Propagation Channel Downlink: Uplink: Beause the estimated channel includes the RF, IF and baseband circuits that have an impact on the frequency response, the uplink and downlink channel are not reciprocal, even in a TDD set-up. uplink and downlink channel matrices are just the transpose of each other for perfectly reciprocal channels. It can be shown that the TX processing is only affected by the non-reciprocity at the side where MIMO-TX is applied, usually at the AP side. To solve the problem, the front-end frequecy responses must be known at this side. Hence, front-end calibration is required at this side only. HDL  (HUL)T

Non-perfect reciprocity  MUI Downlink model (for Inversion) : Not diagonal anymore  Multi-user interference

The Base Station Transceiver alone is responsible for the MUI Only the base station creates MUI (DTX,BS & DRX,BS) The terminal effects (DTX,MT & DRX,MT) only create scalar complex multiplication, can be equalized H includes all common (reciprocal) terms, including antenna coupling Common LO is mandatory Valid for all pre-filtering technique, including MIMO-TX, beamforming, SVD

OFDM-MIMO Demo Set-up 2 Terminals with 1 antenna each Access Point with 2 antennas

Our advise for 802.11n MIMO-TX and MIMO-RX schemes are both interesting for 802.11n MIMO-TX needs channel knowledge at TX side Estimation in reverse link has lower latency Delay between reverse link estimation and MIMO-TX transmission must be minimized must be supported by MAC Protocol MIMO-TX has been demonstrated Real-time (VHDL, 5GHz band) Wireless, 2x2 antennas MIMO-OFDM-SDM (108 Mbps) and MIMO-OFDM-MRC (8 dB SNR improvement)