Ambika - Voicecard data leak

Hi all, my first post here after completing my first MI build. First off I apologize if I am breaking any kind of forum etiquette, since I know the Ambika is no longer supported by MI. I’m posting here simply in the hopes of getting some advice from other Ambika owners/builders.

Ok - so I built an Ambika with 6 smr4 voicecards. 5 of the 6 work exactly as expected, and the 6th one works about 90% as expected with one slight difference from the others- there seems to be a bit of digital noise leaking into the audio signal. On the initial attack and then also on the release of a note there is a subtle and short (but clearly audible) burst/blip of digital noise. If the (vca) attack or release get too long it goes away, and if the attack or release is very short it’s hard to discern from the normal “fast envelope” sound. My best guess at this point is that it is some type of data signal from the 328 that is leaking into the audio path, but I can’t figure out where from. My first thought was that it was the VCA signal, but I later realized that the noise is affected by the filter cutoff (and the vca happens after the filter), so it must either be leaking in somewhere before the filtering, or is perhaps related to the filter cv input?
I have swapped both the 328 and mcp4822 with known working ones and it made no difference, so I assume that it must be something related to the voice card itself. I thought maybe the filtering on the PWM signals wasn’t working, but I have double checked the value of all those components and everything is correct. Can’t find any bridges or other soldering weirdness, the filter tracks (about) 1/v oct fine, and the scaling isn’t off from any of the other voice cards. The A/B jumpers are all set correctly, and everything else works as expected. At this point I’m a bit stumped, do any of you have any thoughts on where to look next?
Thanks!

Trace the signal path to see at which point of the signal chain it appears.

Just wanted to give an update on this odd problem. So following Pichenettes advice, I traced through the circuit, first with an oscilliscope (which turned out to not be sensitive or clean enough to discern the extra noise) and then just through speakers. Found out that it was for sure the VCA CV signal. I tested all the components around that section of the board (c10, R11, and the tl072) everything was good, so I tried replacing the LM13700 associated with the VCA, and that seemed to have cleaned things up mostly. Theres still a little bit of noise that leaks through, but it’s much more subtle now and is very hard to differentiate from the other voicecards. Anyway, happy to get on to playing with the Ambika now!