Originally Posted by Chu Gai
Absolutely just as there are golden noses, taste buds, and other senses. In this test that you're doing (I'm on an iPad and the wife's laptop is verbotten, so I haven't done the test but plan to), is the point you're making or suggesting that the audible differences you're picking up on are related to non-linear distortions? If so, which in your opinion is the more egregious?
I think you are being too specific with respect to the use of the term non-linear in my post. So allow me to explain.
Our audio systems can suffer from two types of distortions: linear and non-linear. Linear distortions do not create additional frequencies whereas non-linear ones can. An example of the former is what the room does to your speaker's low frequency response. You will get peaks and valleys but they are all of linear type. While the response has changed, there is nothing new in the audio.
We tend to have better sensitivity to linear distortions because we are more familiar with them. You would for example know what "too much bass" or too much highs sound like.
Non-linear distortions is everything else. This type of distortion can be any and all things. For example when you compress files, the transients create what is called "pre-echo" where a version of the sound is heard prior to it actually playing! This is data dependent distortion and highly non-linear. Yet, people listen to and enjoy compressed music despite high levels of distortion there.
When you reduce the bit depth, assuming you have added dither (noise), you are hiding the low level details. What would that sound to you? You probably couldn't describe it even though this seems like an intuitive concept. Due to masking often these distortions are hidden. If you don't know what they sound like, you can't find them in the track.
Jitter is another type of non-linear distortion where two sidebands of distortion are created. I often ask people what jitter sounds like and of course they can't answer.
So going back full circle, the point I was making is that non-linear distortions are tricky in that to most people they are not audible. Because of that people have a strong tendency to declare them inaudible. But as you have seen, just like audio compression, training helps significantly in identifying these distortions. Training comes in two forms which is understanding the nature of the system, the distortion and hearing its audible form in many variations.
As to your question, I am not ready to declare the source of degradations. That would require more testing to narrow it down. As a minimum we would want to separate the effect of sample rate from bit depth.