AWR1843BOOST: Mirrored Range Doppler map

Hi,

Please apply an FFT shift on the 2D FFT data in the doppler dimension to get correct mapping to positive and negative velocities.

Regards,

Kaushik

Hi Kaushik,

Thank you for your reply. I apply fft shift after both the range and doppler fft. That is not the problem. The problem are the "mirrored" doppler signals on the range-doppler map.

See above on the plot on both ends of the vertical axes I see same reflections as in the center (it would be easier to it on the animated plot). It is some kind of wrap-around effect - do you know where it might stem from?

Many thanks,

jan

Hi Jan,

Can you please share your sample script, data and your config where you are observing this?

Regards,

Kaushik

Hello Kaushik,

Many thanks for your reply. In words, I read raw data from the AWR1843BOOST using the DCA1000 board. Then I reshape the data to obtain a radar data cube of size (32, 512, 12) - (chirps, samples, antennas). On the data cube I run a real FFT on the samples direction and then a complex FFT on the previous FFT output. I shift the final FFT. 

I suspect this "wrapped" range-doppler map might be due to a wrong interpretation of the raw data into the data cube. Can this be the case?

This is the code I use to interpret the raw data into the radar cube (note that I use AWR1843BOOST) chip -> https://github.com/moodoki/iwr_raw_rosnode/blob/master/mmWave/scripts/fft_viz.py#L61

I use the same script to generate the range-doppler map -> https://github.com/moodoki/iwr_raw_rosnode/blob/master/mmWave/scripts/fft_viz.py#L104

Many thanks for help!

jan

Hi Jan,

Thank you for sharing the same. Please give me two week's time to review this and ge back to you.

Regards,

Kaushik

hi Kaushik,

Any news on this issue?

Best,

jan

Hi Jan,

I would require an additional week's time to review this. Regret the delay caused.

Regards,

Kaushik

Hi,

Ok - thanks - let me know when you have some info.

Best,

jan

Hi Jan,

Looks like it is getting a bit delayed. Have you looked into it and made any progress?

Regards,

Kaushik

hi,

How do you mean delayed? In what sense? It is very weird because separately each these mirrored FFTs look fine and also the range dimension is fine.

Unfortunately I have not made any progress despite many trials - I'm at a dead end with this one. If you need some additional material for analysis just let me know.

Best,

jan

Hi Jan,

I had been out of office for a while and was not able to follow up in time. 

jaknm said:

I apply fft shift after both the range and doppler fft.

You are only supposed to apply FFT shift after doppler FFT and not after range FFT. Could you please check with this once?

Regards,

Kaushik

hi,

Kaushik Gowda said:

You are only supposed to apply FFT shift after doppler FFT and not after range FFT. Could you please check with this once?

Sorry, this was a misspell from my side - I do it only after the doppler FFT - doppler FFT happens after the range FFT so that is what I meant by saying I do it after both.

Do you have any hunch about what is going on there maybe? Hvae you seen something similar before? Can this have to do with the wrong raw data restructuring? (I use 3 TX and 4 RX).

Best,

jan

I see. No problem. Here are some more questions that can give us hints:

• Is your peak in the range bin matching the range of your test object?
• Are the mirrored peaks matching your velocity or integral multiple of your velocity?
• Can you briefly mention how you have chirped and how you have re-arranged your data?

Regards,

Kaushik

hi Kaushik,

Kaushik Gowda said:

Is your peak in the range bin matching the range of your test object?

Yes.

Kaushik Gowda said:

Are the mirrored peaks matching your velocity or integral multiple of your velocity?

Yes - the one around zero is matching the velocity and the mirrored ones are matching integral multiples.

Kaushik Gowda said:

Can you briefly mention how you have chirped and how you have re-arranged your data?

As for chirping - I configured the radar to send 3 times (one for each TX) 32 chirps and each chirp contains 512 samples.

As for rearranging the data - I use the scheme below from the online document - https://www.ti.com/lit/an/swra581b/swra581b.pdf. I use the data arrangement scheme as for complex data with xWR14xx (I am using complex data with AWR1843BOOST). After obtaining the samples I reshape them in N rows and 8 columns - that way each row has all 8 numbers representing the real parts (first 4) and imaginary parts (last 4) of the received receiver samples for each chirp. Then I add first 4 with the 1i times last four to get the actual complex number. Then I reconstruct the virtual 12 antennas data (3TX x 4RX) by assuming that the received data comes after one another from each TX - that is first three rows will contain first chirp samples from TX1, TX2, TX3.

Let me know if you any other info.

Best,

jan