Just to clarify something here in your post that is perhaps not so obvious. You are talking about noise floor modulation but you did not talk about HOW this modulation occurs and why it would be bad for sound.
OK, let's call it NFM for short. If you look at a plot of IMD you will see the spurious IM tones created because of the non-linearity of the device. Multiply the number of tones to 500 or whatever you consider best represents a music signal & the number of IMD tones become very numerous & overlapping to the extent that they form a broadband distortion which is modulating with the signal. It's modulating in complex ways with changes along the axes of pattern, spectral makeup & amplitude. But that's just one simple explanation & there are many other ways that the noise floor can be modulated - I use liFePO4 battery power because of its low noise & high stability & anywhere I have tried it (In the reply chain), it has improved the sound Vs stock PSes. This is power direct from the battery output - even using a low noise voltage regulator such as TPS7A4700 or LT3042 degrades the sound. My conclusion: active devices are introducing some level of instability in the ground noise.
It's difficult to isolate just NFM & say how it sounds alone but in my experience & my reading of others experiments where I believe ground noise stability is being improved, the predominant audible difference is that there is a perceived increase in dynamics, in solidity of the soundstage, in realism of the illusion created by the playback. Look at RogerD's thread on heavy gauge ground wires between chassis of his playback devices - to my way of thinking what is happening here is that the ground impedance differences between devices are being eliminated/minimised & noise on the signal ground (reference ground) is now stable (He's effectively providing a lower impedance path for ground currents which may have been on his interconnect shields or ground wires) His reports of the sound is exactly the same as what I & others describe & I believe the underlying mechanism is the same - a more stable signal ground/reference.
Why this should be perceived in this way requires a deep dive into the workings of auditory perception but if we consider just NFM then auditory perception continually categorises the incoming soundstream into foreground & background layers - the background is relatively ignored with just a watching brief from auditory perception. So, for instance, we generally can ignore (hear through) the hiss of analogue playback, of room ambiance, etc. We continue to ignore this background sound unless something changes in it (the watching brief picks this up) & our focus is automatically shifted to the change in the background sound (this is a survival mechanism) & is called mismatch negativity in perceptual research.
It's not difficult to see what's happening in NFM somehow affecting our perception but it isn't as simple as above, in that it seems to affect the way we perceive the foreground sounds 7 we don't perceive our attention being drawn to the background sound layer. But I don't subscribe to the idea that all we perceive must be consciously registered - there is a lot happening to our perception subconsciously which colours our final perception.
If I am understanding you correctly (please correct me if I am wrong), you are talking about the phenomenon that was described by Crowhurst where he posits that because of negative feedback a signal will create a myriad of distortion products that will be indistingushable from noise EXCEPT for the fact that it is correlated with the signal and therefore not true noise.
The problem with this correlation with the signal is how our ear/brain processes such a "noise" is not the same as true random noise...like tape hiss. With true random noise it has been shown that you can hear below this noise floor to signals that are correlated with the music...kind of a neural "lock-in amplifier" that allows us to hear those signals because they are not noise. Of course there are limits to this. With a music correlated and modulated "noise" floor you have two issues, 1) the "noise" floor is changing constantly, rising and falling with the music amplitude and complexity and 2) Now that it is not random and uncorrelated with the signal you can no longer hear below this "noise" floor.
This impacts how we perceive low level signals and hearing soft sounds in the presence of much louder sounds...or as my late friend Allen Wright used to call it "Downward Dynamic Range". Soft sounds will now get masked by a correlated mush caused by a myriad of distortion products. This will also impact decay of notes and ambient sound cues.
Yes, you got it but I don't think it is just due to negative feedback - NFM can be created from many different sources - I mentioned PSes. I also believe that noise on the data wires of USB also intrudes & causes NFM. I've experimented with the output stage of certain DAC chips which have their current bias exposed through a pin & using a CCS instead of the recommended resistor noticeably improves the sound
When you think about it, it's not just soft sounds that are affected, t's also affecting the start of all sounds (they all start from soft) - so the precise timing of the start of sounds is being slightly changed by NFM. The solidity of the soundstage is one of the characteristics noted when NFM is reduced/eliminated.
I believe an understanding of NFM is the next breakthrough to be made in improving audio playback but it will require a measurement approach which can isolate it or maybe a way of simulating it & studying the various characteristics & their effect on perception.
This is what was behind my post about electronics distortions being more noticeable to our auditory perception than speaker distortion even though the gross THD figures would suggest otherwise - we are perceiving the distortions in different ways.
Look at horn speakers - usually very efficient but regarded as highly distorting, yet when not cranked too loud they can sound exquisite. I'm sure you can pitch in about Aries Cerat, horns, morricab?
Thanks morricab