OK, that is what I was thinking as well.
Short update: I did the test again and you were right! I missed to align the tracks properly so in the end the test with ONLY THE CABLES to conect I/O was a little better than the cables plus the compressor in bypass (around 0.2 dB)
But still I’m not near what the AI suggested! (except if I take two signals - i.e. the one with only cables and the one with cables patchbay and compressor in bypass - that where AD converted and compare these two: then they almost dissapear!)
@Johnny_Moneto I think I cannot follow (again 
)
Is it a mistake that you one time write 0.1 dB and the next time 1.0 dB? If not than I don’t understand your test and the level-matching…
Also I can not follow your conclusion. What do you mean by:
I made sure that all levels in my test are “the same”. So that shouldn’t be the problem…
No, this was not a mistake. The level difference after adjusting one of the two audio files brought them within 0.1dB of each other, so almost identical. Before they were roughly 1dB apart.
The point is that I reduced the level difference from ~1..0dB to ~0.1dB but the null tests were 6dB apart. Small change on the input, big change on the output.
Furthermore a difference of ~0.1dB between the two source signals resulted in -28dBFS. Small differences would result in a level of -150dBFS. So this is really loud.
Keep in mind that I used a clean triangle waveform. If you use a “normal”, ie. more complex signal small differences in level will result in even bigger discrepancies in the null test result.
Again, a null test is really only for comparing two digital signals, not a digital to an analogue signal.
Ok, I think I get it now - thanks for clarifying!!
Could that mean that mean that there is nothing wrong with my setup, just a lot more going on that I expected?
I knew that and like I said above: I didn’t expect the signals to null. I just didn’t think the effect would be that strong. So my whole Idea was just to try how much noise floor there would be left, or how much the coloring of the unit was in bypass compared to the original signal that I was sending through.
So the big difference might be coming from a small difference in level or alignment?
That could also mean that the one-sidedness of my drumloop could then appear because of some very hard panned instruments of the set…??!!
The differences will also come from the signal being different due to DA and AD conversion and the impact of the cables and connectors. What I tried to show is that it doesn’t take a lot of difference for the null test to completely fail.
Again: Don’t use it in the analogue domain. It is the wrong analysis tool.
Good idea. If you use hard panned signals within the signal, I guess it could be.