Quoted from nosro:
I have sent you a link to the files.
I don't think this is an undersampled sound because it is one of the long songs. I have tried importing this WAV into many other sound scripts and always get the distortion.
The distortion seems to happen on many other sounds as well. It is not always audible, but looking at the exported WAV shows that there is definitely some alteration happening.
Yes, sound 0x454 is not undersamped.
To check that the Pinball Browser algorithm is working well, I used the "bootstrap" technique :
Export sound 0x454, then import it and export it again : The differences between the 2 exported files are barely noticeable :
To see some differences (and check that you're not comparing the file to itself !), you have to zoom a lot (amplitude) :
but obviously, it's not working like this with your file...
To understand what's going on with your file, display its spectrum; you'll see that you have frequencies near the Nyquist frequency (Sound at 24kHz => Nf=12kHz) :
(you can check that it's not the case with the original sound).
That's why your sound is distorted after Import/Export (see the Shannon sampling theorem and add some lossy compression in the middle to match the game format).
The solution to your problem is to apply a low-pass filter with Audacity. I did that on your test file with a 6kHz cut-off frequency (but you can try a higher frequency) :
Now, the signal is the same (more precisely "almost" the same, but it's barely noticeable).
I'll see if I can add such a filter in the PB embedded sound editor; it would be more convenient.