AWR2243BOOST: Phase discontinuity issue

Hi Daniel,

Are you seeing this phase shift in the Tx power output spectrum? Can you provide more details on your observation along with the respective chirp configs?

Regards,

Kaushik

This is a field result under single target scenario for 8 virtual antennas. The phase discontinuity from Tx1 to Tx3 makes angle processing invalid in the next step. 

Hi Kaushik,

Any update on this issue?

Thanks!

Hi Daniel,

I have not come across this before. Before diving deeper, I would like to just confirm the following top-level info if possible:

1. Just to rule out SW/config issues, do you see the same error if you repeat the test with a different EVM sample?
2. Have you applied any phase shift (BPM or per chirp) for either Tx1 or Tx3 chirps? (Most likely not in your case, just to check)
3. Is it possible to share your chirp configuration?
4. Have you performed any Tx phase shift/mismatch monitoring using mmWavelink in your test?
5. What is the exact phase shift error stemming from all Tx3 based samples? What is the median and deviation of error across runs?
6. Are you using mmWave studio for your test or some kind of custom SW on chip?

Regards,

Kaushik

Hi Kaushik, please find my responses below:

1. Just to rule out SW/config issues, do you see the same error if you repeat the test with a different EVM sample?
   Answer: Yes, issue is easily reproducible.
2. Have you applied any phase shift (BPM or per chirp) for either Tx1 or Tx3 chirps? (Most likely not in your case, just to check)
   Answer: No.
3. Is it possible to share your chirp configuration?
   -- Chirp 0
   if (ar1.ChirpConfig(0, 0, 0, 0, 0, 0, 0, 1, 0, 0) == 0) then ...
   -- Chirp 1
   if (ar1.ChirpConfig(1, 1, 0, 0, 0, 0, 0, 0, 0, 1) == 0) then ...
   -- Chirp 2
   if (ar1.ChirpConfig(2, 2, 0, 0, 0, 0, 0, 0, 1, 0) == 0) then ...
4. Have you performed any Tx phase shift/mismatch monitoring using mmWavelink in your test?
   Answer: No.
5. What is the exact phase shift error stemming from all Tx3 based samples? What is the median and deviation of error across runs?
   Answer: even goes above 3 radians.
6. Are you using mmWave studio for your test or some kind of custom SW on chip?
   Answer: mmWave

Hi Daniel,

Thank you for the inputs. Let me review this and get back to you by this EOW.

Regards,

Kaushik

Hi Daniel,

After looking into this and discussing with the team there could be one of two possible reasons behind this. Can you let me know the details:

• Have you performed zero angle calibration?
• Another possible cause is that there could be antenna related angle mismatch. What is the angle at which you have placed your test object? How does the mismatch error change when you move your object in an arc from -60 to +60 degrees? 

Regards,

Kaushik

Thanks Kaushik,

Zero calibration is performed but does not have a major impact for the following reasons.

Per our observation, the mismatch error:
1- is angle dependent.
2- changes over time.
3- is periodic vs time.

We have plotted the error for 1000 frames for a stationary single target in the following plot. By moving to a different angle, a different error pattern will be observed.

Hi Daniel,

For a given EVM, angle dependent mismatch will be constant across chirps/frames. Your observation is strange indeed. I would like you to perform some experiments:

• How does this trend change with respect to change in frame periodicity and chirp definition?
• Does this pattern hold true across EVMs? (You previously mentioned that you could see this across EVMs)

Regards,

Kaushik

Hi Kaushik,

We have tried various config settings while the error seems to be persistent. We are not sure what parameter is influencing on the error.

As for other EVMs, we will investigate it thoroughly and update.

Thanks,

Daniel

Hi Daniel,

I am also curious as to how you have obtained these phases. Can you provide a few details on your phase extraction method in your test?

Regards,

Kaushik

Sure.
The phase is extracted from the radar cube after Range and Doppler FFTs are calculated. There are a total of 12 virtual antennas but the first 8 were taken for plotting: Tx1 -> Tx3 -> Tx2

We are not expecting a major phase drop or jump in the phase from antenna 1 and 8 for a single target (corner reflector), and an almost linear phase curve is expected.

We are testing in stationary scenarios. But phase looks to be dynamically changing over time!

Hi Daniel,

It is supposed to be linear as you mentioned. I will add a few more pointers below:

• I hope the phase is taken for same exact range-doppler bin across antennae.
• Do you also see this phase change just after 1D FFT?
• Does anything about the mismatch curve change if you change the type of reflector that you are using as a test object?

Please let me know the results once you have more info from your testing.

Regards,

Kaushik

• I hope the phase is taken for same exact range-doppler bin across antennae: Correct, same bin.
• Do you also see this phase change just after 1D FFT? Yes, the error seems to be matching with that after Doppler FFT.
• Does anything about the mismatch curve change if you change the type of reflector that you are using as a test object? We can try this, but we wanted to rule out elevation signal components from the data.

Hi Daniel,

Thanks for the info. Let me know once you have more details following your tests.

Regards,

Kaushik

Hi Kaushik,

Here are the updates:
- We used a different AWR2243 evm on a different setup, the phase mismatch error is showing on this evm too. The error pattern looks different though.
- We replaced the corner reflectors with some other objects, the error was still observed.

Could this be due to the lua script? Do you have a validated script for 3Tx, 4Rx which we can try?

Thanks,

Daniel

Hi Kaushik,

You mentioned that angle dependent mismatch will be constant across Tx's? For instance, how much phase mismatch is expected between one chirp on Tx1 and another on Tx3?
Daniel.

Hi Daniel,

Please find my responses as follows:

1. Could this be due to the lua script?
   You can try a data capture with just two chirps (Tx1 and Tx3) manually instead of the lua script though I highly doubt that the lua script would be causing phase related issues unless there is some config that would mess with it. 
2. Do you have a validated script for 3Tx, 4Rx which we can try?
   We do not have one as of now.
3. You mentioned that angle dependent mismatch will be constant across Tx's? For instance, how much phase mismatch is expected between one chirp on Tx1 and another on Tx3?
   By constant, I meant constant for a particular EVM therefore it can be resolved with calibration. I do not know the exact limits within which this mismatch can occur (I can get more info regarding this), but it is supposedly not to be as high as 2 radians as you are observing. The margin of error is much lower.

Regards,

Kaushik

Thanks Kaushik.
We switched to DFP this time and calibrated the Rx gains at theta=0. We collected 60 frames and the plots before angle processing (after doppler FFT) are shown below. We used a corner reflector for this particular test.

With DFP,  we don't see a variable/periodic phase mismatch between Tx1 and Tx3 as we were seeing on mmwave studio. However, there is still a fixed large phase offset between Tx1 and Tx3. The error seems to be about 1 radian now.

Have you seen this before?

Hi Daniel,

I don't understand what you mean by we switched to DFP. Did you change your DFP version?

You say you calibrated Rx gains. Have you calibrated existing phase mismatches across channels?

Since the change is not varying, it is something that can be offset and rectified with calibration. 

Kaushik Gowda said:

I can get more info regarding this

Regarding this, the typical mismatch ranges are about -10 to +10 degrees.

Regards,

Kaushik

Hi Kaushik,

We were using mmwave studio with lua script where seeing large variable/periodic phase mistmatch between Tx1 and Tx3.
We don't see this issue when using DFP (instead of mmwave studio). I hope this clarifies the difference.

Can you please share the procedure for calibration?

Thanks,
Daniel

Hi Kaushik.

We are still seeing phase shift from Tx1 to Tx3. The mismatch is almost constant from frame to frame (see plot above) but changes from run to run (after rebooting the radar).
A fixed calibration at theta = 0 doesn't seem to work as the phase mismatch changes from run to run. Do you think calibration can offset it? How? 

Thanks,

Daniel

HI Kaushik,

Any idea what's going on here?
We are blocked due to this issue!

Thanks for your consideration.

Daniel

Hi Kaushik,

We are still blocked and waiting for a solution. Any help would be appreciated.

Daniel

HI Daniel,

Kaushik is OoO on vacation for another week, please expect next reply beyond 30Nov.

Meanwhile, could you perform the factory calibration for boot-time and phase shift for your setup?

https://www.ti.com/lit/an/spracf4c/spracf4c.pdf

https://www.ti.com/lit/an/spracv2/spracv2.pdf

.

Regards,

Jitendra