Question about DPD error on my board using ADRV9029

It is most likely the PA peak power must be compressed. With the signal level and its PAR that you are transmitting, was the peak power of the PA is in line?

Can you please use CFR/ Pre-clipped waveform and check for DPD performance?

Can you please check at 1,2,3 dB back off power and let us know the DPD statistics and the observations?

I hope the path delay cal was run and there were no setup changes after the path delay cal was completed?

https://wiki.analog.com/resources/eval/user-guides/adrv9029/dpd_error_troubleshooting

When conducting this test, the CFR was disabled, and the test was performed. The PAR of my BB signal was approximately 11.8dB, matching the DPD statistics results.

Below are my additional questions:

1. Would it be beneficial for DPD to conduct the test by adjusting the PAR to around 8.5dB using CFR? (Especially for STABILITY_ERROR)
2. Does AM_AM_OUTLIERS_ERROR relate to path delay calibration?
3. Regarding your mention of '1, 2, 3 dB back off power,' is it correct to adjust the ORX gain or Tx Attenuation in a situation where there is no PA, as in the above scenario, to lower the DPD statistics by 1-3dB based on ORx Power?"

After enabling CFR and applying PAR at a level of approximately 8dB, the DPD statistics are as follows.
(The PAR of the baseband signal is 11.8dB, but it changed to approximately 8dB after applying CFR.)

[RFIC_GET_TSSI] 
Tx channel:1, mask:1 
meanTu  : -12.517063 dB 
peakTu  : -4.630517 dB 
meanTx  : -12.517063 dB 
peakTx  : -4.630517 dB 
meanOrx : -14.046754 dB 
peakOrx : -5.825554 dB 
DirectEvm : 0.263926 
IndirectEvm : 0.263926 
SelectError : 0.243279 
IndirectError: 0.238437 

dpd error_code : 0x341d 
dpd percent complete : 0/100 
dpd iter count : 277 
dpd update count : 0 
root@O-RU:/home/user/rf_test/RF_cmd#  

[RFIC_CFR_STATUS]  
Tx channel:1, mask:1 
cfrErrorCode 0000 
cfrEngine1PeaksDetected 964826008 
cfrEngine1PeaksSkipped 565907621 
cfrEngine2PeaksDetected 834302965 
cfrEngine2PeaksSkipped 483178584 
cfrEngine3PeaksDetected 722550727 
cfrEngine3PeaksSkipped 403320389 
cfrNumSamplesClipped 1965486624

Stability errors are still occurring.

Please advise on the above issue.

After enabling CFR, it looks like the AM-AM outlier errors are gone, my understanding is correct? However, i could still see large no.of peaks being skipped even at CFR Engine 3. Did you enable Hard clipper as well? Do you see proper output spectrum without spurs in the Spectrum analyser after enabling CFR? Assuming that the CFR is working fine, now we have only stability error.

PaulJung said:

Would it be beneficial for DPD to conduct the test by adjusting the PAR to around 8.5dB using CFR? (Especially for STABILITY_ERROR)

Yes, use CFR so the peaks would be compressed.The peak power/ PAR should match with the final stage PA peak power.

PaulJung said:

Does AM_AM_OUTLIERS_ERROR relate to path delay calibration?

Could be.

PaulJung said:

Regarding your mention of '1, 2, 3 dB back off power,' is it correct to adjust the ORX gain or Tx Attenuation in a situation where there is no PA, as in the above scenario, to lower the DPD statistics by 1-3dB based on ORx Power?"

You mean you do not have PA in your setup? I'm requesting you to test at 1 dB, 2 dB and 3 dB back-off power from the PA target power by reducing the digital input signal level.

If you are testing without PA, you should not see these errors. Please let me know if you have any additional gain in the Tx lineup and you are just testing TX to ORX feedback for DPD.

I have also experienced a situation where, in a configuration without AMP, on the ADRV9029 EVB, during Tx to ORx feedback, the update count remained at 0 with only the iteration count increasing, and no error code was observed.

Currently, I am DPD tests using the Tx-ORx feedback configuration without AMP.

Would it be meaningful to conduct tests to achieve a No error state with DPD without AMP and then conduct tests after obtaining AMP?

Or would it be more meaningful to conduct DPD tests with AMP?

I'm curious about your opinion.

It is better to test with Amplifier in the path as without which the DPD might do the inverse correction.