IWR6843AOP: Unnecessary Calibration

Hi,

The rangecompbias can vary widely across different devices. For best performance, I would calibrate each individually.

This function is often used to correct for the effects of an imperfect radome or housing, or to improve angle accuracy for the radar. Is that how you're using it? The received signal strength shouldn't vary strongly from device-to-device (maybe a maximum of 1-2 dB by my experience).

Best,

Nate

Hi Nate,

Thanks for your reply.

I want to clarify that the calibration was performed for each radar individually.

The table below pertains to a single radar calibration. During the calibration process using the demo visualizer, the demo continuously outputs compRangeBiasAndRxChanPhase values until you click "stop sensor." I took 10 measurements and calculated the mean and standard deviation for each channel separately. The results are summarized in the table below. These results are for a single radar, but similar results were obtained for the other three radars.

Here's my question: if I use the mean values as the calibration values, the corner received power does not change or gets lower, rendering the calibration ineffective. This was only checked on a corner in the boresight. Although the radar has housing and a radome, as I mentioned before, the calibration is ineffective on a corner in the boresight. Do you recommend performing the calibration at different azimuths, not just on the boresight?

Thanks,

Shlomi

Shlomi Shvartzman said:

Mean	-0.031	0.084	0.084	0.09	0.09	-0.154	-0.154	-0.037	-0.037	-0.09	-0.09	-0.048	-0.048	-0.1	-0.1	-0.08	-0.08	0.016	0.016	-0.128	-0.128	-0.023	-0.023	0.136	0.136
STD	0.083	0.276	0.287	0.46	0.204	0.236	0.347	0.483	0.472	0.211	0.282	0.242	0.293	0.139	0.284	0.357	0.234	0.282	0.33	0.159	0.455	0.292	0.301	0.387	0.276
STD/|Mean| (%)	267.7	328.6	341.7	511.1	226.7	153.2	225.3	1305.4	1275.7	234.4	313.3	504.2	610.4	139	284	446.3	292.5	1762.5	2062.5	124.2	355.5	1269.6	1308.7	284.6	202.9

Hi Nate,

I agree with you, but I'm still confused about the values I'm getting from the demo.

When I load the correct calibration configuration and place the radar in front of a corner, I can see an output of compRangeBiasAndRxChanPhase every second in the "Console Messages" window of the Demo Visualizer. Below are some examples of the outputs.

As I mentioned before, the STD value indicates that the measurements are very noisy. Given this, which values should I choose for the correct calibration?

Thanks,

Shlomi

Hi Nate,

If the measurements show high variation (with a high standard deviation relative to the mean), what does that say about the validity of the calibration?

Thanks,

Shlomi

Hi Shlomi,

Your point is fair that the single chirp measurement is noisy, but I think this doesn't account for the fact that a typical radar measurement in a single frame relies on many chirps aggregated together (reducing noise). What if you do 100 measurements and look at sample means from 16/32/64 chirps together to simulate a typical chirping profile? Does that reduce your variance?

Best,

nate

Hi Nate,

I believe the calibration values are printed every frame (not every chirp).

If that's the case, they are very noisy.

What do you think?

Thanks,

Shlomi

Hi Shlomi,

Yes it looks like you're right. Maybe you could try increasing the number of chirps per frame to achieve more averaging? We could also look at the relative phase values to see if the different vectors are not so far apart in 25-dimensional space. Try adding more chirps to the frame first and we'll see what we can do.

Best,

Nate