AWR6843ISK: AWR6843ISK

Mani D

Prodigy 90 points

Part Number: AWR6843ISK

Tool/software:

Hi ,

the mentioned formula is used to implement capon (ie mVDR))beam forming .

I have seen in matlab that MVDR beamforming is implemented using below formula

mvdr = 1./real(sum(sv'.*(Cx\sv).',2));

where Sv- streeing vector

Cx- covariance matrix inverse

i have attached sample matlab example code and its result for your reference.

%%%%%%MVDR (Capon) spatial spectrum estimator for ULA

fs = 8000;
t = (0:1/fs:1).';
x1 = cos(2*pi*t*300);
x2 = cos(2*pi*t*400);
array = phased.ULA('NumElements',10,'ElementSpacing',1);
array.Element.FrequencyRange = [100e6 300e6];
fc = 150.0e6;
x = collectPlaneWave(array,[x1 x2],[10 20;60 -5]',fc);
noise = 0.1*(randn(size(x)) + 1i*randn(size(x)));

estimator = phased.MVDREstimator('SensorArray',array,...
'OperatingFrequency',fc,'DOAOutputPort',true,'NumSignals',2);

[y,doas] = estimator(x + noise);
doas = broadside2az(sort(doas),[20 -5])

plotSpectrum(estimator)

%%%

My question:

1.Can you confirm what is the difference between capon beamforming used in your case and implemented in matlab using "mvdr = 1./real(sum(sv'.*(Cx\sv).',2)); "?

2.does both implement the same thing or different? can you elaborate about it?

regards,

mani

12 days ago

0 Clinton Rodgers 12 days ago

TI__Genius 12955 points

Hi,

Thanks for posting. Please allow us a day or so to follow up.

Thanks gain,

Clinton

0 Kristien Everett 9 days ago in reply to Clinton Rodgers

TI__Expert 6745 points

Hey Mani,

Thanks for reaching out regarding your question on MVDR. First of all, I'm unfamiliar on the complete implementation of MATLAB's MVDREstimator function besides the equation you've outlined. However, the main difference between our MVDR implementation and the traditional MVDR implementation is that we apply a diagonal loading factor, alpha (sometimes specified as gamma), to the covariance matrix to help improve robustness of the beamformer.

As alpha approaches infinity, the beamformer's behavior starts to resemble the standard beamformer/digital beamformer (DBF). Conversely, as alpha approaches 0, the beamformer's behavior starts to resemble the traditional MVDR beamformer. Therefore, we can use this alpha parameter to mix between the two beamformer behavior's which we've found 0.03 to be optimal for most of our applications.

It doesn't appear a diagonal loading factor is used in the MATLAB implementation, so I would wager that the MATLAB MVDR is a traditional MVDR.

Regards,

Kristien

0 Mani D 9 days ago in reply to Kristien Everett

Prodigy 90 points

Dear kristien,

thanks for your quick reply.

i have captured ADC data using AWR6843ISK+DCA 1000 card for the following Chirp design parameter and trying to implement capon beam forming as per the equation used by you:

system parameter:
Maximum range - 2.7 m
Range resolution - 0.053 m
Maximum velocity - 0.406 m/s
velocity resolution - 0.015 m/s

chirp design 1:
profileCfg 0 60 943 11 41 0 0 97 1 64 2200 0 0 30
channelCfg 15 7 0
adcbufCfg -1 0 1 1 1
chirpCfg 0 0 0 0 0 0 0 1
chirpCfg 1 1 0 0 0 0 0 2
chirpCfg 2 2 0 0 0 0 0 4
frameCfg 0 2 54 0 200 1 0

antenna pattern attached for your reference:

I have used the same capon beamforming equation and calculating Range_Azimuth spectrum .(step by step calculation screenshots attached below for your reference):