Post Go back to editing

ADRV9025 DPD is not working properly

Test environment: AD90255GNR signal,FDD working mode, all RX links closed, all TX links open, signal bandwidth 100MHz, central frequency point 3.65g.

Test result: DPD of TX1 and TX3 works normally, DPD of TX2 and TX4 does not.
Problem 1: The DPD calibration of link TX2 and link TX4 does not work, does not converge, and ACPR always remains uncalibrated ACPR=-33dB.

(1) TX2 link, the output power is 23dbm, and the READ DPD status is shown in the figure below: THE DPD feedback data shows that the DPD calibration coefficient is not updated due to the poor EVM.

After instrument test, the EVM of TX2 output is 3.1%, and that of ORX2 input is 3.36% without DPD.
TX4 is the same. This shows that the hardware is fine.

For help 1:Direct EVM, Indirect EVM, Indirect Error, Select Error is too large. How to reduce it? How can I make TX2 and TX4 work?

Problem 2: TX1 and TX3 can work normally, but the first DPD convergence time is not fixed, sometimes fast and sometimes slow, the fastest 15s, the slowest 10mins.

For help 2: What parameters can be set to shorten the first DPD convergence time?

Problem 3:When TX1 and TX3 DPD are working normally, ACPR deteriorates (from -51dB to -41dB) when the output power of TX1 or TX3 is changed. DPD convergence speed is slow, and it takes 1 minute to recover to -51dB.

For Help 3: Which parameters can be changed so that DPD can quickly converge?

Problem 4:In TX1 link DPD function normal calibration state, using "adi_adrv9025_DpdReset" and reset DPD, TX1 link output signal has no response, read TX1 DPD status value, dpDUpDatecount value is not recalculated.Why?

Parents
  • Which DPD model file are you using and how did you select that? Can you please share us the DPD tracking configuration parameters? Also, can you please share us the DPD statistics for the working condition on Tx1 or Tx3.I believe TX1 and Tx4 channels are same as Tx1 and TX3. DPD update time is about 4sec for all 4 channels. Whats the Tx-ORX mapping in your system? Can you increase the ORX gain by another 5dB and check the performance?

    For help 1:Direct EVM, Indirect EVM, Indirect Error, Select Error is too large. How to reduce it? How can I make TX2 and TX4 work?

    The EVM that we are measuring here is not the symbol EVM that we measure on the instrument. These parameters are related to DPD and please refer UG for more details.

    For help 2: What parameters can be set to shorten the first DPD convergence time?

    DPD update time is about 4secs and should not take what you mentioned. What is your Tx- ORX mapping?

    For Help 3: Which parameters can be changed so that DPD can quickly converge?

    Are you enabling the DPD at the target power and then decreasing the power or its the other way around? In any case, DPD convergence should happen

    I think the DPD model file you chose may be affecting all these phenomenon.

    Can you do a DPD model sweep by following the below link?

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

  • 1,Which DPD model file are you using and how did you select that? 
    Answer:model1.

    2, DPD tracking configuration parameters are as follows:

    #adi_adrv9025_DpdTrackingConfig parameters

    dpdIndirectRegularizationValue=20
    dpdDirectRegularizationValue=35
    dpdSamples=16384

    #dpdMThreshold=((10^MThreshold_dBFS/20)*FULL_SCALE_CODE)^2
    MThreshold_dBFS=-21
    FULL_SCALE_CODE=32768

    dpdPeakSearchWindowSize=655350
    dpdUpdateMode=1
    minAvgSignalLevel=512( equivalent of -36dBFs)

    #Default value
    dpdMu=50
    minAvgSignalLevelOrx=512(equivalent of -36dBFs)
    dpdFilterSel=0
    enableDirectLearning=0
    dpdIndirectRegularizationLowPowerValue=20

    3,DPD statistics for the working condition on Tx1 or Tx3:

    4,Whats the Tx-ORX mapping in your system? 

    Answer:

    5, Can you increase the ORX gain by another 5dB and check the performance?

    Answer:I tried to modify the ORX gain, but nothing changed.

    6,According to UG, indirectEVM and DirectEVM calculation formula is very simple, but requires time aligned and gain and phase equalized,Is it possible that the time is not aligned or the gain and phase are not equalize due to the failure or failure of a calibration, resulting in EVM value is  too big?

    7,Does it affect DPD if I turn off CLGC?

  • Can you confirm if you are using FDD mode?

    Can you confirm if the raw ACLR is similar as that of the working channels?

    Can you operate at 3dB back off on the non-working channels and check if the DPD is working fine? 

    Keep the ORX gain index in such a way that the peak ORX power is about -3dBFS

    CLGC tracking requires DPD tracking to be enabled but DPD does not have any dependency on CLGC.

    Can you do a model sweep using the below link:

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

    Model1 is of how many coefficients? Can you use 190 coeffs and enable direct learning?

  • We seem to have found the cause of the problem. We have readjust the PA. Although the EVM of PA is not changed, DPD can update parameters normally

    But I still have three questions to confirm:

    1. Now the test is conducted in FDD mode and DPD can work normally. If the switch is in TDD mode, is it possible to switch to TDD mode as long as TX and RX are enabled according to TDD time slot ratio?

    2, in the execution of DPD calibration, RadioCtrlCfgSet is configured into Pin mode, TX and RX enable is controlled by FPGA, ORX enable how to control?

    3. Is there any problem if the Direct EVM and Indirect EVM in Default Definition of Fault Condition are changed to 20% or higher?

  • Q1. Yes, DPD works in TDD mode as well

    Q2. ORX enable should also come from FPGA. Please refer page 113 of the UG 1727.

    Q3. No, by default we have set these values to 13% which can be changed. As long as the DPD performance is met, changing these fault condition values should be fine.

  • readjust the

    Hi 

    Do you mean you replace another PA  and solve the problem?

    Thanks

Reply Children
No Data