Urukul/AD9912 - Output frequency lower than requested, noisy signal

aletofful

I am working with 2 racks of Sinara hardware with a total of 3 Urukul AD9912 boards. Two of the Urukuls used to work fine in rack #1. However, after running them in rack #2, the output signal from the DDS suddenly dropped to 67 % of the frequency requested by ARTIQ, and the signal appears very noisy (see picture attached). This behaviour was then reproduced after moving the boards back to rack #1. The third Urukul board which was run for the first time in rack #2 shows the same behaviour.

It seems like rack #2 might be the culprit and is breaking the Urukuls, however I am not sure what exactly could be causing this problem. Note that I powered both racks separately with the same exact power supply, since the two Kaslis are identical and have the same requirements.

rjo

If your configuration demands a reference clock: are you providing that reference clock? At which power?

aletofful

rjo It is clocked at 100 MHz at 9.1 dbm. The same reference clock was used for both Kaslis.

charlesbaynham

@aletofful Correct me if I'm wrong, but I believe your reference clock is distributed from Kasli to the Urukuls (for both racks). @rjo, is it possible that this distribution has somehow failed and the Urukuls are expecting direct clocks now?

aletofful

charlesbaynham Yes, we provide an external clock reference to Kasli, and that then distributes it to the Urukuls. We do not directly clock the individual Urukuls.

@rjo Another thing that came to mind is that when I observed this problem, I was running on rack #1 the most up-to-date version of ARTIQ that matched the up-to-date binaries, while on rack #2 I ran the same new version of ARTIQ but with some old binaries. Could this have maybe caused the Urukuls to break when running them in rack #2?

rjo

AFAIK there has been a lot of manual disassembly and reassembly of that crate. So I don't know whether the MMCX refclock connections are in place. But it doesn't seem to be the clock you want to use. It's certainly not 100 MHz but 125 MHz. The device DB that I received would expect a 100 MHz reference clock at each Urukul front panel SMA clock input. 9.1 dBm sounds a bit high. 6 dBm would be better.

aletofful

rjo The MMCX were always reconnected after the reassemblies. I was not aware that 100 MHz was the wrong frequency to provide Kasli as this had always been set up this way on rack #1 when the Urukuls used to output the correct frequency. Your comments have been very helpful - I will now test if this fixes the issue.

rjo

100 MHz is the right frequency for your configuration for both Urukul and Kasli. But the internal MMCX connection carries 125 MHz and in your device db you are not using it.

aletofful

I can now confirm that after clocking the Urukuls directly via SMA with 100 MHz at ~6.5 dBm, the output signal is now clean and at the right frequency. Thank you for pinpointing the problem!