EVM Varies with DAC NCO Frequency - Root Cause?

Hi PICOSATS 

Sorry to hear that you're running into EVM issue over the frequency. Thanks, forwarding the EVM vs IF frequency, its very helpful. I don't think this is an issue with DAC's self-calibration.

Unfortunately, I don't have the hardware and software options to measure EVM for your use case. I believe 2 tone IMD3 response should be close representative of wideband data

Looking at AD9082 (AD9081 should be similar performance) datasheet Figure 19 IMD3 at FDAC= 12GSPS case with 1x8x interpolation which slightly different from your case but close enough. I see some variation in IMD3 around 2.8G to 3.5GHz when the back-off is 0dBFS for the 2 tone.

Can you increase your back off to see if performance changes?

I picked a use case with 1x8x interpolation from ACE quick configuration. Please see details of my setup at the end of the post. Using 2 tones I tried to measure IMD3, I didn't see any deterioration at 2.8G when compared to 3.4G. It's better than ~ -80dBc performance. I could have zoomed in and taken better measurements, but I think for first order I didn't see anything unusual. If you could try this on your side that would be great.

I didn't see any images or spurs that seem to be of concern at these frequencies either.

I do see the DAC frequency bandwidth rolloff between 3G to 4G for FDAC = 12GSPS case,not sure if this is causing issues, something to look into.

If you could do similar measurements with a highly linear vector signal generator and get a baseline of your EVM measurement equipment that would be good data to compare with.

Sometimes equipment band or internal path changes at certain frequencies, that might change the EVM measurement optimization. Something for you to explore

My setup details

Hello Sam, thank you for your prompt reply.

- I've tested back-off (if by that you intend reducing the amplitude of the digitally modulated signal) by reducing the samples amplitude via an FPGA gain block but I had no luck: the effect is present regardless of the samples gain value.

Now I'm hunting the registers. There must be (in my opinion)  some sort of incorrect settings.

I've tested all the possible values for register 0x01c9 to check if some different NCOs were combining their outputs but again I had no luck

Now I'm trying to disable register 0x0140 and 0x0143 which control DAC's bit shuffling & rotation

Hi PICOSATS , 

Yes, I meant reducing samples amplitude

I would recommend following the software API procedure that ADI provides as reference to bring up the device and setup the datapath.www.analog.com/.../ad9081.html Software Request Form. You will see it in on your myanalog account -> Resources -> Software Downloads

What is the software and hardware that you using for your application?

If you have the following hardware, you can try to evaluate using ACE GUI or C-API. Once it works satisfactorily you can compare the registers (although I still recommend APi's), if need be.

https://www.analog.com/en/resources/evaluation-hardware-and-software/evaluation-boards-kits/eval-ad9081.html

https://www.analog.com/en/resources/evaluation-hardware-and-software/evaluation-boards-kits/ads9-v2ebz.html

Try a 2 Tone test to see how the DAC output spectrum looks.

Hello, turns out what I was observing was a DAC line mismatched at 50Ohm and that's simply a reflection! Sorry for the inconvience, at first I really thought it was the NCO's fault. In truth, I was "sweeping" the line over frequency, doing a sort of impedance characterization....

Hi PICOSATS , I am glad you are able to resolve your issue.