802.11ac Transmit Beamforming
This example shows how the performance of an IEEE® 802.11ac™ link can be improved by beamforming the transmission when channel state information is available at the transmitter.
Transmit beamforming focuses energy towards a receiver to improve the SNR of a link. In this scheme the transmitter is called a beamformer and the receiver is called a beamformee. A steering matrix is used by the beamformer to direct the energy to the beamformee. The steering matrix is calculated using channel state information obtained through channel measurements. In IEEE 802.11ac [ 1 ] these measurements are obtained by sounding the channel between beamformer and beamformee. To sound the channel the beamformer sends an NDP (Null Data Packet) to the beamformee. The beamformee uses the channel information provided by sounding to calculate a feedback matrix. This matrix is fedback to the beamformer in a compressed format. The beamformer can then use the feedback matrix to create a steering matrix and beamform transmissions to the beamformee. The process of forming the steering matrix is shown in the diagram below.
In IEEE 802.11ac the single user beamformee capability is not mandatory. Therefore a multi-antenna transmitter may have to use a different scheme to transmit packets to a receiver which cannot act as a beamformee. One such scheme is spatial expansion. Spatial expansion allows a number of space-time streams to be transmitted on a greater number of transmit antennas. Using spatial expansion can provide a small transmit diversity gain in channels with flat fading when compared to directly mapping space-time streams to transmit antennas [ 2 ].
In this example a 4x2 MIMO configuration is considered between a transmitter and receiver, with two space-time streams used for a data packet transmission. First the scenario of a receiver which is not capable of being a beamformee is considered. A transmission is made using spatial expansion and the data symbols are recovered and the signal quality measured. To show the benefits of transmit beamforming the data packet is then transmitted over the same channel realization, but this time using transmit beamforming. The performance of the two schemes are then compared. These stages are shown in the diagram below.
A 4x2 MIMO configuration is used in this example with 2 space-time streams.
NumTxAnts = 4; % Number of transmit antennas NumSTS = 2; % Number of space-time streams NumRxAnts = 2; % Number of receive antennas
The format specific configuration of a VHT waveform is described using a VHT format configuration object. In this example the waveform is configured with a 20 MHz bandwidth and the MIMO configuration specified above.
cfgVHT = wlanVHTConfig; cfgVHT.ChannelBandwidth = 'CBW20'; cfgVHT.APEPLength = 4000; cfgVHT.NumTransmitAntennas = NumTxAnts; cfgVHT.NumSpaceTimeStreams = NumSTS; cfgVHT.MCS = 4; % 16-QAM, rate 3/4
In this example a TGac channel model is used with delay profile Model-B. The channel realization is controlled with a seed to allow repeatability.
tgacChannel = wlanTGacChannel; tgacChannel.DelayProfile = 'Model-B'; tgacChannel.ChannelBandwidth = cfgVHT.ChannelBandwidth; tgacChannel.SampleRate = wlanSampleRate(cfgVHT); tgacChannel.NumReceiveAntennas = NumRxAnts; tgacChannel.NumTransmitAntennas = NumTxAnts; tgacChannel.TransmitReceiveDistance = 100; % Meters tgacChannel.RandomStream = 'mt19937ar with seed'; tgacChannel.Seed = 70; % Seed to allow repeatability
Noise is added to the time domain waveform at the output of the channel with a power, noisePower.
noisePower = -37; % dBW
Setup other objects and variables for simulation.
% Recovery configuration cfgRec = wlanRecoveryConfig('PilotPhaseTracking','None'); % Indices for extracting fields ind = wlanFieldIndices(cfgVHT); % AWGN channel to add noise with a specified noise power. The random % process controlling noise generation is seeded to allow repeatability. awgnChannel = comm.AWGNChannel; awgnChannel.RandomStream = 'mt19937ar with seed'; awgnChannel.Seed = 5; awgnChannel.NoiseMethod = 'Variance'; awgnChannel.Variance = 10^(noisePower/10); % Calculate the expected noise variance after OFDM demodulation noiseVar = vhtBeamformingNoiseVariance(noisePower,cfgVHT); % Number of spatial streams Nss = NumSTS/(cfgVHT.STBC+1); % Get the number of occupied subcarriers in VHT fields [data,pilots] = helperSubcarrierIndices(cfgVHT,'VHT'); Nsd = numel(data); % Number of data subcarriers Nsp = numel(pilots); % Number of pilot subcarriers Nst = Nsd+Nsp; % Total number of occupied subcarriers % Generate a random PSDU which will be transmitted rng(0); % Set random state for repeatability psdu = randi([0 1],cfgVHT.PSDULength*8,1);
Transmission with Spatial Expansion
First a transmission is made using spatial expansion. This type of transmission may be made by a multi-antenna transmitter to a receiver which is not capable of being a beamformee. The SpatialMapping property of the format configuration object allows different spatial mapping schemes to be selected. In this example the example spatial expansion matrix provided in Section 220.127.116.11.1.2 of [ 3 ] is used. Therefore a 'Custom' spatial mapping is configured. The custom spatial mapping matrix is used by assigning the SpatialMappingMatrix of the format configuration object. This matrix describes the mapping of each subcarrier for each space-time stream to all transmit antennas. Therefore the size of the spatial mapping matrix used is Nst-by-Nsts-by-Nt. Nst is the number of occupied subcarriers, Nsts is the number of space-time streams, and Nt is the number of transmit antennas. The spatial mapping matrix duplicates some of the space-time streams to form the desired number of transmit streams.
% Configure a spatial expansion transmission vhtSE = cfgVHT; vhtSE.SpatialMapping = 'Custom'; % Use custom spatial expansion matrix vhtSE.SpatialMappingMatrix = helperSpatialExpansionMatrix(vhtSE); % Generate waveform tx = wlanWaveformGenerator(psdu,vhtSE); % Pass waveform through a fading channel and add noise. Trailing zeros % are added to allow for channel filter delay. rx = tgacChannel([tx; zeros(15,NumTxAnts)]); % Allow same channel realization to be used subsequently reset(tgacChannel); rx = awgnChannel(rx); % Allow same noise realization to be used subsequently reset(awgnChannel); % Estimate symbol timing tOff = wlanSymbolTimingEstimate(rx(ind.LSTF(1):ind.LSIG(2),:),vhtSE.ChannelBandwidth); % Channel estimation vhtltf = rx(tOff+(ind.VHTLTF(1):ind.VHTLTF(2)),:); vhtltfDemod = wlanVHTLTFDemodulate(vhtltf,vhtSE); chanEstSE = wlanVHTLTFChannelEstimate(vhtltfDemod,vhtSE);
The received data field is demodulated and equalized to recover OFDM symbols for each spatial stream.
% Demodulate and equalize the data vhtdata = rx(tOff+(ind.VHTData(1):ind.VHTData(2)),:); [~,~,symSE] = wlanVHTDataRecover(vhtdata,chanEstSE,noiseVar,vhtSE,cfgRec);
The constellation of each spatial stream is plotted below.
vhtBeamformingPlotConstellation(symSE, ... 'Spatial Expansion Transmission Equalized Symbols');
The variance in the constellation is approximately the same for each spatial stream as the SNRs are approximately the same. This is because the average power in the channel is on average approximately the same per space-time stream:
disp(['Mean received channel power per space-time stream ' ... 'with spatial expansion: ']) for i = 1:NumSTS fprintf(' Space-time stream %d: %2.2f W\n',i, ... sum(mean(chanEstSE(:,i,:).*conj(chanEstSE(:,i,:)),1),3)) end
Mean received channel power per space-time stream with spatial expansion: Space-time stream 1: 0.73 W Space-time stream 2: 0.50 W
Transmission with Beamforming
When the receiver is capable of being a beamformee, a beamformed transmission can create a higher SNR compared to spatial expansion. We will now show the advantage of having channel state information available to create and use a steering matrix. To calculate a beamforming steering matrix, an NDP is passed through the channel. 'Direct' spatial mapping is used for the NDP transmission and the number of space-time streams is configured to match the number of transmit antennas. This allows the VHT-LTF to be used to sound channels between each of the transmit antennas and receive antennas. The calculated beamforming matrix is then used to beamform a transmission through the channel. The same channel realization is used for sounding and data transmission and there is no feedback compression between beamformee and beamformer, therefore the beamforming can be regarded as perfect in this example.
% Configure a sounding packet vhtSound = cfgVHT; vhtSound.APEPLength = 0; % NDP so no data vhtSound.NumSpaceTimeStreams = NumTxAnts; vhtSound.SpatialMapping = 'Direct'; % Each TxAnt carries a STS % Generate sounding waveform soundingPSDU = ; tx = wlanWaveformGenerator(soundingPSDU,vhtSound); % Pass sounding waveform through the channel and add noise. Trailing zeros % are added to allow for channel filter delay. rx = tgacChannel([tx; zeros(15,NumTxAnts)]); % Allow same channel realization to be used subsequently reset(tgacChannel); rx = awgnChannel(rx); % Allow same noise realization to be used subsequently reset(awgnChannel); % Estimate symbol timing tOff = wlanSymbolTimingEstimate(rx(ind.LSTF(1):ind.LSIG(2),:),vhtSound.ChannelBandwidth);
Channel estimation is performed using the sounding packet to estimate the actual channel response between each transmit and receive antenna.
% Channel estimation vhtLLTFInd = wlanFieldIndices(vhtSound,'VHT-LTF'); vhtltf = rx(tOff+(vhtLLTFInd(1):vhtLLTFInd(2)),:); vhtltfDemod = wlanVHTLTFDemodulate(vhtltf,vhtSound); chanEstSound = wlanVHTLTFChannelEstimate(vhtltfDemod,vhtSound);
The channel estimated using wlanVHTLTFChannelEstimate includes cyclic shifts applied at the transmitter to each space-time stream. To calculate a beamforming steering matrix the cyclic shifts applied at the transmitter are removed from the channel estimate.
% Remove impact of cyclic shift from channel estimate chanEstSound = vhtBeamformingRemoveCSD(chanEstSound, ... vhtSound.ChannelBandwidth,vhtSound.NumSpaceTimeStreams);
In this example the beamforming steering matrix is calculated using singular value decomposition (SVD) [ 3 ]. The SVD of the channel matrix results in two unitary matrices, U and V, and a diagonal matrix of singular values S. The first NumSTS columns of V per subcarrier are used as the beamforming steering matrix. The SVD is computed using the function svd.
chanEstPerm = permute(chanEstSound,[3 2 1]); % permute to Nr-by-Nt-by-Nst V = zeros(Nst,NumTxAnts,NumRxAnts); for i = 1:Nst [U,S,V(i,:,:)] = svd(chanEstPerm(:,:,i),'econ'); end steeringMatrix = V(:,:,1:NumSTS); % Nst-by-Nt-by-Nsts
The beamforming steering matrix calculated above is applied as a custom spatial mapping matrix and is used to send data through the same channel.
% Configure a transmission with beamforming vhtBF = cfgVHT; vhtBF.SpatialMapping = 'Custom'; % Permute steering matrix to Nst-by-Nsts-by-Nt vhtBF.SpatialMappingMatrix = permute(steeringMatrix,[1 3 2]); % Generate beamformed data transmission tx = wlanWaveformGenerator(psdu,vhtBF); % Pass through the channel and add noise. Trailing zeros % are added to allow for channel filter delay. rx = tgacChannel([tx; zeros(15,NumTxAnts)]); rx = awgnChannel(rx); % Estimate symbol timing tOff = wlanSymbolTimingEstimate(rx(ind.LSTF(1):ind.LSIG(2),:),vhtBF.ChannelBandwidth); % Channel estimation vhtltf = rx(tOff+(ind.VHTLTF(1):ind.VHTLTF(2)),:); vhtltfDemod = wlanVHTLTFDemodulate(vhtltf,vhtBF); chanEstBF = wlanVHTLTFChannelEstimate(vhtltfDemod,vhtBF);
The received data field is demodulated and equalized to recover OFDM symbols for each spatial stream.
% Demodulate and equalize the data vhtdata = rx(tOff+(ind.VHTData(1):ind.VHTData(2)),:); [~,~,symBF] = wlanVHTDataRecover(vhtdata,chanEstBF,noiseVar,vhtBF,cfgRec);
The equalized constellation for each spatial stream is plotted below. Note that the higher order spatial stream has a larger variance. This is due to the ordered singular values of the channels used in SVD beamforming.
vhtBeamformingPlotConstellation(symBF, ... 'Beamformed Transmission Equalized Symbols');
This ordering is also visible in the average power of the received space-time streams. The power of the received first space-time stream is larger than the second space-time stream. This is because the received signal strength is a function of the singular values of the channel which SVD orders in a decreasing fashion.
disp(['Mean received channel power per space-time stream ' ... 'with SVD transmit beamforming: ']) for i = 1:NumSTS fprintf(' Space-time stream %d: %2.2f W\n',i, ... sum(mean(chanEstBF(:,i,:).*conj(chanEstBF(:,i,:)),1),3)) end
Mean received channel power per space-time stream with SVD transmit beamforming: Space-time stream 1: 2.08 W Space-time stream 2: 0.45 W
Comparison and Conclusion
The figure below plots the equalized constellation from the spatial expansion and beamformed transmissions for all spatial streams. Note the improved constellation using SVD-based transmit beamforming.
figure plot(symSE(:),'Marker','.','LineStyle','none','Color',[0.929 0.694 0.125]) hold on plot(symBF(:),'Marker','.','LineStyle','none','Color',[0.494 0.1840 0.556]) title('Equalized Symbol Comparison') str = sprintf('%dx%d',NumTxAnts,NumRxAnts); legend([str ' Spatial Expansion'],[str ' Transmit Beamforming']) xlabel('Real') ylabel('Imag') xlim([-2 2]) ylim([-2 2])
The improvement can also be measured through the RMS and maximum error vector magnitude (EVM). EVM is a measure of demodulated signal quality.
EVM = comm.EVM; EVM.AveragingDimensions = [1 2]; % Average over all subcarriers and symbols EVM.MaximumEVMOutputPort = true; EVM.ReferenceSignalSource = 'Estimated from reference constellation'; EVM.ReferenceConstellation = helperReferenceSymbols(cfgVHT); [rmsEVMSE,maxEVMSE] = EVM(symSE); % EVM using spatial expansion [rmsEVMBF,maxEVMBF] = EVM(symBF); % EVM using beamforming for i = 1:Nss fprintf(['Spatial stream %d EVM:\n' ... ' Spatial expansion: %2.1f%% RMS, %2.1f%% max\n' ... ' Transmit beamforming: %2.1f%% RMS, %2.1f%% max\n'], ... i,rmsEVMSE(i),maxEVMSE(i),rmsEVMBF(i),maxEVMBF(i)); end
Spatial stream 1 EVM: Spatial expansion: 9.2% RMS, 44.8% max Transmit beamforming: 2.0% RMS, 8.6% max Spatial stream 2 EVM: Spatial expansion: 9.2% RMS, 52.3% max Transmit beamforming: 4.1% RMS, 12.7% max
This example demonstrates that if a receiver is capable of being a beamformee, the SNR can potentially be improved when a transmission is beamformed compared to a spatial expansion transmission. The increase in received power when using beamforming can lead to more reliable demodulation or potentially even a higher order modulation and coding scheme to be used for the transmission.
In a realistic operational simulation the performance of beamforming would be degraded due to the delay between channel state information calculation and feedback by the beamformee and feedback quantization. For more information see [ 2 ].
This example uses the following helper functions:
- IEEE Std 802.11ac™-2013 IEEE Standard for Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications - Amendment 4: Enhancements for Very High Throughput for Operation in Bands below 6 GHz.
- Perahia, Eldad, and Robert Stacey. Next Generation Wireless LANS: 802.11n and 802.11ac. Cambridge University Press, 2013.
- IEEE Std 802.11™-2012 IEEE Standard for Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications.