02-14-2019 10:25 AM - edited 02-16-2019 11:43 AM
We are finding glitches in captures from the MTS demo using the demonstration GUI. See attached images showing spikes and skips.
* Seemingly random spikes occur differently across all channels (3 shown).
* Somewhat periodic phase jumps occurs the same way across all channels.
I also should say that we once observed captures without any such glitches. Pleasantly surprised, we saved the preferences and configuration to file, power cycled the ZCU111 board, closed the GUI, reopened it, loaded the stored config and preferences, but the glitches were back.
1. Any ideas what causes these different artifacts?
2. Anyone else experience this (or not experience it)?
3. How can we reliably get a good/clean starting point with the MTS load for the ZCU111?
02-15-2019 07:26 AM
do you mean spurs ?
02-15-2019 07:49 AM
Two different artifacts are shown in the time domain. Their spectral responses are different.
02-15-2019 10:06 AM
02-17-2019 08:41 AM
02-18-2019 04:14 AM
Ive just seen this,
is it relevant ?
Your glitch looks way to drastic to be in the adc, more likely in the data capture side.
02-18-2019 05:31 AM
@drjohnsmith is right to point out this issue. In this case the problem with the DMA is still being debugged.
What size is your capture buffer in this example?
As an exercise can you reduce the size of the bufffer to 2097152 samples or below? Does the issue go away then?
02-18-2019 05:33 AM - edited 02-18-2019 05:34 AM
Thanks for finding that issue with large sample sizes. The affected captures in the case of this topic are only 8192 samples; this topic appears to be unrelated to that particular VDMA error issue.
02-18-2019 06:34 AM
OK so what is the metal log telling you?
03-04-2019 11:45 AM
We are having the same issues with spikes and skips as shown by the photos below. If anyone has any suggestions or ideas on why this is occuring it would be greatly appreciated.
03-04-2019 12:25 PM
03-11-2019 11:58 AM
Follow up to sehejahluwalia 's post:
We believe to have determined that the spikes have a high probability of occurring with periodicity of approximately 256 samples. Therefore, it is our assumption that the glitches are due to the memory buffers being written and read simultaneously. The way we had tested this was to set the frequency of the signal relatively low, 20 MHz, and samples to some low number; 1000 was used for our experiment. Phase skipping was uniformly observed under these conditions. Furthermore, upon decimation the signal had lower likelihood of glitching.
Question then becomes; is the FPGA reading from a register in memory being written to... If so, how can we resolve this issue?
06-19-2019 05:10 AM
we are having the same problem in our board. And due to these random phase jumps etc. there are a lot of spurs in the sampled signal.
Is there any solution that you could find? Can you please share them with us?
06-19-2019 05:33 AM
On review of the reference design, it looked like the buffers were overrun periodically, depending on the sample rate... there are various solutions, including simply reducing the sample rate. Also, but likely unrelated, we found that the rf_tool program on the PS will segfault or crash occasionally, and launching the program again did not resolve the issue (power cycle to the board or restart of the PS was required). There are numerous issues with the reference design, but it serves its purpose.
06-19-2019 05:52 AM
Thnx for your answer.by saying reducing sampling frequency, what is the reduced sampling frequency on your trials? Even we reduce from 4Gsps to 1Gsps we observe the same problems in our board. Can you please advise a sampling frequency with which you observe no problems.
09-12-2019 04:50 AM
In MTS mode we figured out that internal PLLs should always be bypassed. Otherwise no matter what you do, you cannot get rid of glitches etc. Hope that your problem is related to that.