AWR1642: CFAR calculation using range-doppler heatmap

Kei

Part Number: AWR1642

Hello,

Using the following settings, I got the following "detected object" data and "range-doppler heatmap" data.
Regarding CFAR threshold (=6000), I think it means about 17.5dB.
(6*6000/256/8 Q8 format, number of Rx antenna is 8)

But calculating CFAR by using values picked up from range-doppler heatmap, SN is under the threshold.
But it was definitely detected because it is included in detected object data.
Is there something wrong?

guiMonitor -1 1 0 0 0 1 0 <- detected object and range-doppler heatmap
cfarCfg -1 1 0 7 3 4 0 6000 <- doppler direction

x, y, z, range index, doppler index, peak val
11.9375, 8.5625, 0.0, 15, 34, 107

excerpt from range-doppler heatmap

	range Index→
doppler Index↓	. . .	12	13	14	15	16	17	18	. . .
	22	26817	28041	27701	27492	25226	25122	26045			liner	dB
	23	27769	28158	27834	27604	26251	25434	25826		average of noiseWin	26602.9	77.9
	24	27239	27403	26655	26626	26506	24928	26041		CUT	29587	86.7
	25	27677	27467	27661	27279	26434	25781	26423		SN		8.74
	26	27783	27674	28061	28303	27412	26551	26202
	27	26569	26700	27552	27764	27236	26379	25318
	28	26696	26738	26598	26996	27277	26212	25885
	29	26567	27192	26353	25587	25176	26791	27092
	30	26527	27132	26842	26621	25497	25450	24889
	31	26803	26547	25880	26672	26175	25749	26354
	32	26168	25778	25747	26914	26400	26165	25953
	33	26485	26632	26936	27771	27515	25656	26362
	34	27093	27246	28135	29587	29226	26761	25571
	35	27977	27180	27473	28504	28302	26138	26509
	36	27477	26386	25724	26156	26399	26599	27613
	37	27533	26490	26520	26268	26985	26664	26092
	38	26938	27580	26660	25741	26564	25448	25479
	39	24986	26609	26667	25958	25239	25291	25585
	40	27379	24655	26733	26753	27410	26075	25185
	41	27412	26494	26542	26480	26623	26167	26177
	42	26420	26404	25025	25511	26461	26890	25944
	43	25989	26427	26698	26050	25412	26703	26843
	44	25753	26660	27078	26772	26438	26593	26408
	45	26068	26963	27018	25384	24592	25458	25735
	46	27354	27440	27364	26765	25501	25038	25961
	. . .

over 3 years ago

0 Cesar over 3 years ago

TI__Guru*** 148804 points

Hi,

Please let us know if you are using sw released by TI or custom SW developed by your company.

If you are using sw released by TI , please let us know what version

thank you

Cesar

0 Kei over 3 years ago in reply to Cesar

Intellectual 750 points

Hi,

I am using pre-built binary xwr16xx_mmw_demo.bin located in C:\ti\mmwave_sdk_02_01_00_04\packages\ti\demo\xwr16xx\mmw.

SDK is 2.1.

In order to extract "detected object" and "range-doppler heatmap" on UART, I am using our custom tool written in python.

Regards,
Kei

0 Cesar over 3 years ago in reply to Kei

TI__Guru*** 148804 points

Please give me some time to review this with systems team

thank you
Cesar

+1 Cesar over 3 years ago in reply to Cesar

TI__Guru*** 148804 points

Hi,

The problem is in the setting of the cfarCfg command. The command used is

cfarCfg -1 1 0 7 3 4 0 6000 <- doppler direction

The length of the noise window is 7 (cfarCfg argument in yellow). This means 7 on the left side, plus 7 on the right side, which gives 14 noise elements, but this is divided by 16, i.e. 2^4 (cfarCfg argument in red). Typically the length of the noise window should be power of 2, since the division in average calculation can only be a power of 2. In this case we are right shifting the sum of 14 elements by 4 (dividing by 16).

So in this example, the SN is 18.48 (shown in yellow below), which is more than the threshold set to 17.5dB, and that’s why it is detected.

So, you just need to increase the noise length to 8.

cfarCfg -1 1 0 8 3 4 0 6000 <- doppler direction

Thank you

Cesar

0 Kei over 3 years ago in reply to Cesar

Intellectual 750 points

Hi,

Thank you for your answer. Now I understand.

Regards,
Kei

Sensors

Sensors forum

AWR1642: CFAR calculation using range-doppler heatmap