OP, I think your measurement setup inadvertantly includes a ground loop... more like two, in both directions (the D3020 v2 seems to be a Class I device). There is no way to get clean results with the equipment you have right now (though I guess the obvious difference in PE resistance between mains leads is good to know). At the very least, a balanced input would be very much recommended... typically this means a $150-ish audio interface. (*) A matching balanced attenuator for speaker to either input line or mic level (depending on which one performs better and what levels it'll handle... check Julian Krause's reviews) is recommended as well.
You are not the first to attempt such measurements either, check e.g.
this thread.
*) I mean, if you are
really, really cheap, a stereo unbalanced input can serve as a mono balanced input in a pinch. The only problem is that you have to invert one channel and add it to the other one, which is simple enough in an audio editor but may prove infeasible live... I don't think REW can analyze an offline recording like RMAA will do.
NAD d3020 v2 is Class D amplifier, NAD modified Hypex UcD102, with integrated DAC Cirrus Logic CS42528 (24bit ,192Hz is plenty for me), and PSU is modified Hypex
SMPS400A100. I wanted something with low power consumption.
1. Really reallly reaally cheap.
I have some technical background in IT and software, but i am not that good with electronics. I just wanted to make sure that i am not being sold snake-oil.
2. I planned to borrow USB DAC/ADC, isolated from the PC.
That way I could get much better measurements, lower noise floor, less spikes.
3. NAD d3020 v2 has Optical SPDIF input
That means I can feed the data to it over electrically isolated line, isolating any noise coming from PC.
That is however not true for the Line-In I used for measuring.
This is the software I use:
https://www.audiotester.de
Which means:
a) As an input for measurement is used LineIn in computer soundcard, and its clearly very noisy input.
b) As an output for measurement i used headphones jack on NAD d3020v2, not speaker output.
At this point i was accounting for really big margin of error, but the results are much better I expected.
c) I confirmed SNR and THD values provided by amp manufacturer to be true and I could compare the measurements with reviews on this forum.
d) I found volume sweetspot were THD goes much lower (-80dB as manufacturer says, i measured as low as -95dB THD)
e) I confirmed that audible coilwhine from GPU under stress is audible through speakers if used Analog AUX input, but its isolated if using Optical SPDIF.
f) I confirmed that Windows distorts sound if its above certain volume level (CAudioLimiter issue).
4. The software measures noise in audible spectrum, and some of it was picked by the wire for the loop
While I was trying to figure out the source i found busted power cord, which was causing issues in PC itself, but earlier today I confirmed that measured noise is not audible, which basically makes it a different kind of issue, not for this forum. I took it as a hint to replaced older power strip with newer power surge and I plan to replace some of the inadequate power cords.
So I managed to perform certain tweaks without much investment on testing equipment...