Are high sample rates making your music sound worse?

Quote: (partial) "There has been an industry trend towards sampling rates well beyond the basic requirements: such as 96 kHz and even 192 kHz[7] This is in contrast with laboratory experiments, which have failed to show that ultrasonic frequencies are audible to human observers. "
Comment:    Whoa!  None of this is done to reproduce ultrasonic frequencies. In fact the analog filters remove them as best they can. It is done so that the analog filters can be lower-order and more phase-correct, and/or to allow them to BETTER remove the ultrasonic frequencies (which exist, whether you like it or not in a stepped eave out of a DAC) more completely. Heck, read the app notes from any major DAC provider like TI/Burr Brown and see the noise after their reconstruction filters. not so good you will find. But with proper up sampling, we move the noise higher making it easier to filter.
It is these partial truths that continue the crazy arguments between engineers trying to make better sound and audiophiles that are being a) confused and b) sold stuff that does not sound as good as it otherwise might.
Rant off.

Folks, I don't have time to go into all of it now, but there are LOTS of really serious errors in the statements being made here. e.g. "thinning out the music", e.g.: "you cant hear above 20 kHz so why 192?"  These make really critical mis-assumptions about what is going on in up sampling.
First, go learn about up-sampling and interpolation filters. Then learn about reconstruction filters and their issues. Then think about how much better you make things if you first interpolate and then feed it to the reconstruction filter.  Lots of analog issues get much easier.
It need not change the original data one bit (both literally and figuratively)
Bottom line: this is all about making the job of the end analog filters easier and less likely to produce artifacts. And done right it works terrifically well.  Done wrong, all bets are off. Don't do it wrong :-)

Wow, this is getting out of hand in breadth and how fast it progresses. Someone commented on what i said (quoted me) and then went on to speak of Chord DACs having the best filtering they knew of. It **appeared** that they were using it to counter my argument that up/over sampling helps with the filtering.
Let me be clear:  Chord over/up samples.  Their data stream runs at 104 mHz/mbps - far, far, far above the raw output of redbook ( around 1.5 mbps).
Their sound is not all due to filtering either.  They pay special attention to timing and jitter (as one must, since the reconstruction filter integrates over the sample value and the time).
They further use a different method to get the analog pulses (flip-flops, let's not go down that art-hole) that has some claimed advantages.
But it in NO WAY counters what I wrote.
But with the loose writing here it difficult to even understand what point people are trying to make.
G

" know the Chord upsamples, so does the Benchmark, and so do many others, I’m specifically referring to upsampling before the DAC. DACs do this as it helps with jitter, doing this before the DAC doesn’t help."
No, they do it to interpolate and reduce the challenge for the reconstruction filter.  Again, don't believe me, read a Burr-Brown application note if you prefer.
There was also a point about up sampling being no longer bit perfect.  If you understand interpolation, which is a distinction without a difference. If you understand interpolation, you know that it creates new data points which a) raise the frequency of the noise to be filtered and b) invent a new intermediate level - beyond 16 bits -- which in effect raises the resolution.  It is absolutely true that no new data exists and it is absolutely true that in a world of perfect, phase-coherent, 200 dB/octave filters, it would be un-necessary. But we are not in that world. So, at 88.2 kHz we get effectively 17 bits.  at 176.whatever we get 18 bits.
I acknowledge that at 96 we do nto have a simple multiple, BUT -- HUGE BUT -- all the original data points remain adn can be reconstructed so in practice it is bit perfect.  Its like saying "i used to have $100, this crook gave me $2 and now I don't have $100 anymore".  Uh, true, you have $102.  You are free to throw $2 on the ground - and the DAC is free to toss any bits on the ground; but it will use them to make its later job easier.
And yes, there are differences pre-DAC and in-DAC but they are ALL BEFORE the actual conversion - either via PDM (sigma-delta) or PAM (ladder).