AWR1243: implementing calibDcRangeSig algorithm

Hi,

Here is some more information on the "calibDcRangeSig" algorithm and how it is different from another algorithm support by the SDK "Clutter Removal"

calibDcRangeSig

Estimate once in a controlled environment where you know there are no true objects that you want to detect and there should be no motion also i.e only what you want to remove during normal operation (hence “calibration”). Remove (subtract) continuously post calibration. Number of range bins in the vicinity of zero (on “either side”) as specified by user are individually accumulated over time (and individually subtracted) i.e if range FFT output of a chirp k is x(1,k),x(2,k)..x(N,k), then during calibration say you want to estimate over K chirps and say you have first F range bins (F << N) and last L range bins (L << N) configured in calib DC (it is a subset of the full N) then estimates are s1,..sF,s(N-F+1),..sN where si = sum(x(i,1),..x(i,K))/K.  During regular (post calibration) operation, you perform a transformation on range FFT output of each chirp x1,x2,..xN to x1’,x2’..XF’, x(N-F+1)’..xN’ where xi’ = xi – si, before feeding to Doppler FFT.

Clutter Removal

Continuously remove all static clutter from the scene. So estimation and removal are done continuously. Two methods A) If Doppler FFT is done with rectangular window (basically no windowing), then zeroing out the Doppler FFT bin 0 does the job. B) If Doppler FFT is done with non-rectangular window, then zeroing out the Doppler bin 0 can take out some very slow moving objects along with the pure static clutter due to wider main lobe. So for this case, we estimate and subtract dc from all the samples before performing Doppler FFT i.e when doing Doppler FFT on x1,x2,..xN, you first do a transformation of the signal to x1’, x2’,..xN’ (before feeding to Doppler FFT) where xi’ = xi - sum(x1,..xN)/N.

Thank you

Cesar