AWR1642: CFAR detector problem

Hi

Here is the response from algorithm team

You could do it but it would weaken the intent of log CFAR [what is the pro/con of log CFAR is a separate question that I have answered in another forum post], it may look more like a hybrid of linear and log CFAR. One way to think of this is in the case of one tx, one rx, when you sum the noise samples, you do sum(log(.)) as per log CFAR definition where the sum is along the range or doppler bins. Now when you add more rx-es [virtual or not], then you want to continue this kind of sum i.e sum(log(.)) across the rx-es also else you are doing part linear and part log. I can respond on the forum if you want me to.

Yet another way to think of this is if you had done an angle FFT and then done a 2D CFAR say range-angle or even 3D CFAR [range-angle-doppler], then you would sum energies in the 2D or 3D domain in the same way i.e sum(log2(.)), here we don’t do angle FFT so in effect we are summing the energies at all angles [it is like a thick arc at a given range bin, thickness is equal to the CFAR noise window] and so that sum also needs to be as sum(log2())