While it is good that MiniDSP works to minimize this noise, keep in mind that this is all digital and the errors are rounding from 32-bit effective number of bits (194.4 SINAD) to a 24-bit ENOB (146.2 SINAD) and sometimes the rounding driving that down to a 20-bit ENOB. Here in the 'analog world' a 20+ bit ENOB is perfectly excellent. It is only relative to the 32-bit perfect digital signal that this looks like a huge problem.
It would be nice to be 20-bit as it's the theoretical results, but measuring the analog performance that wasn't the case at all.
Loading it with couple of filters (nothing heavy, just couple of L-R24dB and couple of corrections down low) performance was degrading rapidly as measuring lower.
While the hi-section was maxing my silly measuring rig (around -105 THD+N) the low section had a straight rising line going lower, resulting in mid 70's at the lower registers.
A subsonic, as many would like to do for example, would rise the whole way noise floor near 35dB! I have a +10dB room mode at ~30Hz which I sometimes knock down (sometimes I don't, feels like a guilty pleasure late night with low levels for example) , hitting a correction there would create havoc.
The strangest of all? Even engaging one single filter was giving the effect, as if it was changing working state, that's the way to describe it.
And it's a shame, cause flex looked perfectly nice.
Looking at 1kHz THD+N looks nice and all. But my way of measuring these days is starting at 20Hz. If it does this right then I go higher, if not I feel it's a lost cause.
Edit: the funny part of the story is that I returned it as broken, that's what I thought it was, as a sample.
If
@mdsimon2 didn't up the game the way he did, and if I wasn't searching the minidsp forums, diyaudio, etc after that I would still think it was broken.