Tell me if I screwed up here, but the so called “intersample peak” problem is a non issue, caused by people with just a little knowledge of sample data systems but not enough of it.
Start with any band limited analog signal with fmax < N.
Sample that signal at a rate >2N, creating a digital signal.
Nyquist guarantees that you can get back the EXACT original analog signal. Including the AMPLITUDE which is often larger than the sample values.
Very few (perhaps not any) of the digital values will occur at a peak of the analog signal.
Consequently many (perhaps most) digital samples will be lower than the recreated analog signal when you plot them together on the same graph.
The so called “intersample peaks” problem confuses the two domains.
When you convert the digital signal back to analog you get exactly the original analog signal. There is no peak problem with that signal due to intersample peaks or anything other than poor D/A circuit implementation. Or using the playback on a cheap consumer device that does not have the ability to handle the peakiest peaks.
Unless you manipulated the digital domain to make it clip, the digital domain will be an exact representation of the original analog signal and can be converted back to analog with no problems.
If the original analog signal had saturated that is not an intersample peak problem either. That was just poor use of the analog equipment. And the digital signal will exactly recreate that saturated analog signal. Again there is no intersample peak problem.
Some people point to Gibbs phenomenon, which only occurs at a single point in the mathematical description of all this where there was a discontinuity in the original analog signal. In practice this will not be a problem either due to intersample peaks or anything else as analog signals can never have a jump discontinuity in real life, and the band limiting guarantees that.
Now, if you do the A/D with one set of equipment, and then do the D/A with another, I suppose that it could be possible that the second unit may not have the voltage range to fully play the proper signal in the analog domain. That is not an intersample peak problem either. If you just turn the gain down on the converter you would get a faithful frequency and sound from it , although smaller in amplitude than the original , which could be undone by a power amp following thus fully recreating the original signal in all aspects.
OK. Bash away or tell me that I am right.