IWR6843: Difference in SNR Values between SDK 3.5.0.4 and SDK 3.3.0.3

Hi,

As you said, I tried modifying the profile_2d.cfg of SDK 3.5.0.4. In that, I changed the RX Gain Parameter of profileCfg from 158 to 30 and tried running the Demo Visualizer with the edited profile_2d.cfg. SNR Observed was 89dB.

Can you suggest any other ideas ?

Regards,

N.Shankar

Hi, Shankar:

If different RX setting does not change the SNR, then I will need to consult with our software expert on this.  I will get back to you soon.

Best,

Zigang

HI, Shankar:

Our software expert suggest to 

1) Try on SDK 03.04.00.03, if you see the same problem,  it is very likely that the calibration is causing the difference. 

2) On the new SDK version, try to disable the calibration by changing the calibration and periodicity to false in mss_main.c, and then recompile the code. 

Let me know how it goes.

Best,

Zigang

Hi,

1. I tried with SDK 03.04.00.03, SNR captured using Demo Visualizer was 88.5dB

2. I disabled the calibration in SDK 03.05.00.04 and compiled the code and flashed the binary, SNR captured using Demo Visualizer was 89dB.

So, Change in Calibration is not causing any difference.

Can you suggest a solution?

Regards,

N.Shankar

Hi,

As you said, I checked the Signal Profile and Noise Profile in the mmWave Demo Visualizer 3.5.0.

1. With SDK 3.5.0.4(Disabled Calibration) and its profile_2d.cfg, we observe SNR value as 89dB(Range Profile) and Noise Profile as 35dB.

2. Can you tell where to look about the parameters related to FFT? Is this causing the difference in SNR Values?

Regards,

N.Shankar.

HI, Shankar:

For 1) you may want to do the same with older version of SDK and compare side by side.   It will be great if you can send the signal/noise profile figure with peak marked.  

For 2), the FFT size may affect the FFT scaling, 

• At SDK 3.3.0.3 Out-of-Box, it is hardcoded at C:\ti\mmwave_sdk_03_03_00_03\packages\ti\datapath\dpu\rangeproc\src\rangeprochwa.c

hwaParamCfg[paramsetIdx].dest.dstScale = 0;

hwaParamCfg[paramsetIdx].accelModeArgs.fftMode.butterflyScaling = 0x3;

• At SDK 3.4.0.3 and above, it is programmed at the same file

hwaParamCfg[paramsetIdx].dest.dstScale = pDPParams->fftOutputDivShift;hwaParamCfg[paramsetIdx].accelModeArgs.fftMode.fftEn = 1;

hwaParamCfg[paramsetIdx].accelModeArgs.fftMode.butterflyScaling =
(1 << pDPParams->numLastButterflyStagesToScale) - 1U;

Best,

Zigang

Hi,

Thanks for the reply.

I have used the same setup as mentioned above. I have used mmWave Demo Visualizer 3.5.0 for capturing the Signal/Noise Profile.

I have attached the images of the signal/noise profile of SDK 3.3, SDK 3.4 and SDK 3.5 in the below pdf. Setup is same.

Signal_Noise_Values.pdf

Regards,

Shankar.N

HI, Shankar:

I will have to consult with some SDK expert and trying it myself before set back to you.

Best,

Zigang

Hi, Shankar:

Sorry for the delay, I did some raw data capture and does observe the lower SNR in the SDK 3.5.0.4 comparing to the SDK version 3.3.0.3.  The test at home with a small CR located at 0.9m looks more obvious than a large CR located at 5m in the Anechoic chamber.  I also observed the loop back gain level is different between the two versions.  

Can you try a target with different distance?   

I reported this problem to the design team but have not received any feedback from them yet.

Best,

Zigang