Other Parts Discussed in Thread: IWR1642
Hi,
I've been experiencing similar issues as described in the previous posts with people count demo for IWR1642. Judging from the data,I get 56 beams, 28 on each side, each a bit less than 2 degrees apart covering 110 deg FOV in total. However, there is a blind spot that covers about a sector of 5 beams missing from the middle. This almost 9 deg blind spot causes target detection algorithms to create new targets in the middle of the area and losing track of the target that entered the blind spot.
From what I have gathered from the previous posts, the CFAR algorithm is the culprit, which makes sense, as there is no neighboring cells for the edges which to compare the cell under the test. I have been digging into the source code for the people counting demo for 68xx board as it should work as per the forum posts. However, I am a little bit lost here, as the all significant difference in the CFAR as far as I can see, are RADARDEMO_detectionCFAR_priv.c function RADARDEMO_detectionCFAR_CAAll() save for a one additional flag and a type change in the .h file.
However, compiling the people counting demo for 16xx using the contents from the RADARDEMO_detectionCFAR_priv.c for the 68xx did not affect the situation. Is there something else I should change and has anyone succeeded in patching the demo so that the blind spot is not there?
Best regards,
Marko Leppänen
