You may recall this better than I, but there was a link in the past that instead of showing what little difference there was between a lossy file and a lossless file, they showed what was actually removed. You could listen to an audio file that played you the audio that was actually removed. There was plenty
it depends on actual encoding algo and its parameters.
generally they trim the higher frequencies because that's the most space efficient technique -- its a lot of information that is not perceivable by most people and with most equipment.
so if you feed a flat signal into the algo and measure the output, you'll typically see roll off around ~15khz area (really top end treble area).
now the thing is (1) many people have difficulties with hearing in that range AND (2) most consumer products have trouble reproducing sounds faithfully at that end.
stack on top of that, (3) most people are listening on the go, in public environments, etc, and any theoretical difference in the sound is very very difficult to
actually discern by a listener.
remember, these algos have been around for 3 decades. engineers have gotten really good at tuning the encoders.
when all of #1 (good ears), #2 (good equipment), and #3 (good environment) are aligned, some people can genuinely tell a difference.
but there are many people who
think they can tell a difference, but they (unknowingly) already failed at #1 or #2.
and for most people, even with #1 and #2 satisfied, they don't have #3, so the point is entirely moot.
