Post Go back to editing

Question about ACLR fluctuating up and down while running DPD on ADRV9029

The video below is the result of running DPD on EVB.


I have a problem with ACLR going up and down per about 1 sec.

Is there any way to improve this?

Below are the settings for the DPD test on ADRV9029 EVB.

Please advise on how to improve the above.

thank you.

Top Replies

Parents
  • If you are not using GaN PA, please change DPD update mode to 0 or 1. If you are using 190 coeffs model file, use damping factor between 20 and 40 and you may enable direct learning if required.In direct learning, regularisation value should be between 20 and 30.

    In the DPD statistics, The Tu power is close to the saturation (full scale) and the Tx power is beyond full scale. Test with BB input waveform of -12 or -13dBFS ( currently it is -9dBFS), there is no headroom for DPD gai. expansion in your case.

    You can increase the ORX gain index by another 10dB.(The peak ORX power can be upto -3dBFS)

  • Thanks for your reply

    Please let me know what could be the cause of ACLR repeating up and down once per second.
    (The power of the baseband signal does not change, the power coming through the ORx is also constant, but the ACLR moves up/down similar to the DPD actuator update cycle.)

    At the time of the above test, regularization was changed to 5, 10, 20, 30, but there was no significant difference.

    The PA used in this test is GaN PA.

    Please advise.

    Thank you.

  • This could be due to the DPD update period. Which model file are you using- 190 or 95 coeffs?

    If you are using 190 coeffs, use dampling factor between 20 and 40.

    The Tu power and Tx power are saturated, please back-off the digital input waveform by another 3dB and increase the ORX gain index by another 10dB.

    Let us know the observations once you test with the above changes.

  • I'm using it with a 95 coeffs.

    Are you talking about lowering the power of the Tx Source file (BB signal) to lower the TU Power you mentioned?
    Also, can I scale down the Tx BB signal to lower the TU Power? (The scale that appears when the "TONE" button is pressed in the Transmit tab of the TES GUI)

    Please answer my questions.

  • Yes, you can adjust the scaling from the 'Tones' window. 

    Back off by 3dB from the scaling that you are already using. So, if you are using 0dB, enter -3dB and test once. Check the DPD statistics if the change is impacted.

  • As you said to me, I tried changing the scale by -3dB to lower the Tu power and changing the Orx gain index, but there was no big difference from the video above.

    The figure below shows the results of DPD statistics during the test.

    Do you think the above phenomenon (ACLR moving up/down) could be a symptom of DPD?

    Thank you.

  • It could be due to DPD, can you monitor the spectrum without DPD and see if this variation is observed.

    Can you increase the ORX Gain index so the peak ORX power can be upto -3dBFS?

    Play with Peak search window size and Regularisation value parameters.

    Can you also check with a different waveform if the problem is due to a waveform?

  • According to the customer who asked this question, it comes from the end of the PA power .

    In addition, it is said that the above phenomenon occurs less when the CFR is given strong.

    I'd love to hear from you on the above customer comments.

  • Can you increase the ORX Gain index so the peak ORX power can be upto -3dBFS?

    Has this been tried out? The DPD performance will be good when the ORX Peak power is about -3dBFS.

    In addition, it is said that the above phenomenon occurs less when the CFR is given strong.

    When you say CFR is strong, you are trying to achieve less PAR, right? If so, the PA may be operating above the compression point when the CFR is not strong. Can you check at 1 to 3dB backoff when the CFR is not strong?

Reply
  • Can you increase the ORX Gain index so the peak ORX power can be upto -3dBFS?

    Has this been tried out? The DPD performance will be good when the ORX Peak power is about -3dBFS.

    In addition, it is said that the above phenomenon occurs less when the CFR is given strong.

    When you say CFR is strong, you are trying to achieve less PAR, right? If so, the PA may be operating above the compression point when the CFR is not strong. Can you check at 1 to 3dB backoff when the CFR is not strong?

Children
No Data