It's the same as the argument over cheap snares versus boutique snares.
Sorry but this is apples and oranges. Hit a snare a thousand times and measure it and you'll have a thousand different sets of data. Convert an original WAV to MP3 and back a thousand times and you'll have a thousand identical sets of data.We're talking about a measurable difference between an original and a compressed/decompressed set of 0 and 1. Since there's no original snare drum to compare to this example unfortunately doesn't fly.
Can you see a difference between 16/44 and 96/24 ? It looks impossible for humans..
There's a difference between 44.1kHz and 48kHz depending on your DAC. Because of lower quality some of them start to roll-off the heights way too early. Nyquist theorem prohibits frequencies above half the sample rate, so all DAC come with this roll-off, some do it better, some do it worse. But in general there's no need to go any higher as 44.1 for any kind of musical consumption. 48kHz just raises the Nyquist frequency into a definitely inaudible range so the roll-off definitely has no impact onto the signal covering the audible range.
What Chris is talking about is production. It's highly important to have a multiple of samples while producing audio, since lots of audio processors might add frequencies above Nyquist (e.g. any kind of non-linear amplification like distortion and exciters, bit crushing, (brickwall) limiting, even some harsh compressors), being processed further down the stream, introducing unwanted artifacts. Although todays professional grade DSP routines over-sample internally if needed.
But this is another esoteric discussion, broadly argued about in all kinds of home recording forums all over the world. Pros don't care, they just throw money at the highest grade shtuff (or better: what they
perceive as the highest grade stuff).
there I hear a difference between FLAC 0 or FLAC 7 against MP3 320CBR.
Could you point your finger on the lossless compressed version? Although there is tiny differences (we're talking about 0.001-0.01dB every few samples) the majority of semi- to professionals claiming they could is (scientifically proven) wrong here, saying the MP3 is of higher quality if tested blindly.
Ask someone you trust to set up at least three (better 5) different tracks for a real blind test, mixing in some WAV to WAV tests, just for fooling your confirmation bias. In my opinion it wouldn't work outside of a laboratory (or any other controlled, unfamiliar environment) since I e.g. could just throw various mp3 decoders onto the files and compare just the diff of the data. So I could tell which is which within seconds (okay, maybe minutes) sitting in front of a Linux machine - without listening.