You brought up the dCs paper by Mike Story so I will explain what is wrong with the white paper titled "A Suggested Explanation For (Some Of) The Audible Differences Between High Sample Rate And Conventional Sample Rate Audio Material" found on the dCs website.
I would first recommend that you read my paper "Sampling Theory" that explains the immediate relationship between bandwidth and impulse response width. It is entirely incorrect to suggest than an impulse width and bandwidth are separate issues.
I have been bothered by that "Suggested Explanation" for quite some time. When doing analysis, appropriate tools must be used for each job.
For example:
If I borrow a million dollars I have to pay back at least a million dollars. It would be great if I could choose a different method to pay such as to take the log of a million which is $6. The lender will scream "bloody murder" because I used the "wrong tool." I used a logarithmic scale because it is advantageous to me.
Well, it is OK to talk about audio volume in log scale. The ear is nearly logarithmic, when it comes to relating POWER and volume, so the log scale is a good tool to use. But it is wrong to use a logarithmic scale of a TIME DOMAIN WAVEFORM because that does NOT correspond to the ear perception of a waveform.
We do not take the log of the waves that make music. Doing that would create unbelievable distortions of both harmonics and volume perception.
The impulse response of a 96KHz bandwidth is half the width of 48KHz. The paper by Mike Story from dCs does take a log of a time domain function. This is very wrong, and leads to wrong conclusions. If you arbitrarily put me in front of a warped mirror to show me that I am very thin or fat, the picture is distorted. This presentation takes liberties beyond engineering and science, and the use of a log scale with a totaly arbitaray -30dB referance distorts the picture totally. It states that:
"the 44.1 and 48KS/s filters spread the audible energy over 1msec or more. The 96KHz filter is much better, keeping the vast bulk of energy within 100uSec. The 192KS/s filter can be very good indeed, keeping the energy within 50uSec."
This statement may make one believe that the width ratio going from 48 KHz to 96 KHz is 10 to 1 instead of 2 to 1.
Also, any worthwile EE knows that there is NO POINT in offering 96KHz bandwidth when the mic and speakers are 20KHz or so. Any EE knows, or should know that the time domain behaviour and the frequency domain are one of the same, just a different way to present and view the same thing. It is in my paper, which is a good paper, not full of holes.
My guess is that they will continue to display that "paper" in their web. The honorable thing to do is to publicaly appolgise for diservice they have done. This very faulty paper was one of the rationals for doing 192KHz conversion. I learned to not hold my breath where marketing plays a role
BR
Dan Lavry