Hi AD experts,

I have some troubles for having an effective gain with DPD with the waveform LTE provided.

I use the evb ADRV9002 with a driver and the PA sky66391-12. According to the datasheet, the PA is able to achieve 35dB raw ACLR at 28dBm and 50dB ACLR with DPD.

I measured a raw aclr at 37dB for a 28dBm output

I notice any improvment with the ACLR at 28dBm but I have some improvment at 25dBm as the raw ACLR goes from 38dB to 46dB with DPD.

The hardware set up :

the ORX is set on the RX1B with 60dB attenuation from the output PA

ADVR9002 TX1 is activated

TES set up :

waveform = CFR_sample_rate_15p36M_bw_10M.csv

Device : FDD/LTE/15.36MSPS

Board : external PA loopback after PA = ON

Path Delay = 36500 ( I measured with the API the delay : 18500 ; the DPD was unstable and with 36500 it is better, but there are still some fluctuations)

DPD configured by default)

Could you help us?

[edited by: nicolas_22 at 4:27 PM (GMT -5) on 17 Feb 2021]
  • 0
    •  Analog Employees 
    on Feb 18, 2021 5:31 PM

    Hello Nicolas,

    Just responding to let you know that I am investigating this issue offline and will return with a solution when it's ready.

    Best Regards,


  • Hello Oisin

    Thanks for your help. Here are more details about theses tests :

    - I measured the RF path delay with the PA = 10,5ns. With the 18,5ns measured by the API, it is coherent with the 30ns written in the TES GUI

    - I measured the 60dB coupling path at -28dBm so with a 7dB PAR the peak power should be around -21dBm closed to the -18dBm required at the input of the EVB

    - I try different couplings from  70dB to 50dB with no success (70dB being the worst case)

    - I try with the TETRA waveform i have some improvment with DPD but not so good as the one showned in the user guide

    - The device system configuration is : (see picture below)



  • 0
    •  Analog Employees 
    on Feb 22, 2021 11:42 AM in reply to OWatkins

    Hello Nicolas,

    I've brought your issues to one of our applications engineers who's very experienced with measuring DPD performance. He had the following questions and comments:

    1) He would like to know how you've measured the external path delay. Internally we use an IronPython script that runs Adrv9001.cals.ExternalPathDelay_Calibrate(...). Your comments on "DPD being unstable at Path Delay = 18500" confused him a bit.

    2) Do you have an LPF placed at the Tx output? This is necessary to filter out the LO harmonics. Refer to figure 182 in the User Guide for more details.

    3) Do you know the P1dB of the PA you're using? Our engineer is wondering if you're pushing the PA too hard, stating "ADRV9002 DPD requires the Tx peak signal doesn’t surpass P1db".

    Let us know your responses to these questions and we'll keep working to get you an answer.

    Best Regards,


  • Hello Oisin,

    You will find my reponse below :

    1/ I measured the delay with the API that you provide in Analog Devices\ADRV9001 Transceiver Evaluation Software\IronPython according to user guide Measuring the External Path Delay p190. It gives me 18500. 

    For the measurment I remove the driver, PA, filter  and coupler. So with a VNA, I measured the delay for thoses devices which give me 10,6ns.

    If i write 18500 in TES GUI, the DPD is unstable, with 29000 it is fine

    2/ I have a 1805-1880 ceramic band pass filter between the driver and the PA. I will try tomorrow to move the filter before the driver

    3/ I don't know the P1dB, but this PA sky66391-12 is able to have 50dB ACLR with DPD from 35dB raw ACLR at 28dBm. TYou achieved this goal with the ADRV9375:



  • Hi,

    I moved the filter before the driver. The LO Harmonics are below 65dBc. The results with DPD are the same.