I’ll try to find some time to do the CPU comparisons ASAP.
To answer your question, no the volume of the noise does not change based on speed, but the characteristic does. Since it’s bleed from the RAND waveform, slowing down the LFO also slows that waveform.
The noise goes away entirely when I clock the SH1oh1 to external midi and don’t play an external clock. This seems to deactivate the RAND, causing it not to bleed into the post filter path.
If we can’t find any solution in the future, perhaps just a “mute/kill” option for the RAND would be a reasonable work around??
I’ll get back to you on the comparison.