Well, being the geek that I also am, I came home and ran some similar tests. I'll only explain the one I found most interesting, which I think throws a bit of a kink in Damon's findings. Here's what I did.
1. took a 24/96 source file (source.wav)
2. dithered source.wav using Wavelab uv22hr and "save as new file" called it (dith_test1.wav)
3. closed everything, then re-opened source.wav
4. dithered source.wav using Wavelab uv22hr and "save as new file" called it (dith_test2.wav)
5. opened dith_test1.wav and dith_test2.wav in Wavelab and used Analysis>Audio File Comparer to compare
RESULT: The two files are identical.
NEXT>>
6. opened dith_test1.wav in Wavelab and resampled to 44.1 and "save as new file" called it (dith_resamp_test1.wav)
7. opened dith_test2.wav in Wavelab and resampled to 44.1 and "save as new file" called it (dith_resamp_test2.wav)
8. opened dith_resamp_test1.wav and dith_resamp_test2.wav in Wavelab and used Analysis>Audio File Comparer to compare
RESULT: The two files are
NOT IDENTICALFrom what I can tell, Wavelab does not identically resample the same file. Every time you run it against a source file, save as a new file, then compare the outputs, you'll find differences. The fact that this is the case makes it impossible to do a comparison where resampling is involved.
Moving on to a batch dither comparison on it's own.
1. took track1.wav and track2.wav (24/96 tracks) and combined them in Wavelab and called it combine.wav
2. dithered combine.wav in Wavelab using uv22hr and "save as new file" and called it combine_dith.wav
3. used Wavelab batch process and uv22hr to dither track1.wav and track2.wav and "save as new file" and called them track1_dith.wav and track2_dith.wav
4. used Wavelab to combine track1_dith.wav and track2_dith.wav and saved the new file calling it separate_dith.wav
5. opened combine_dith.wav and separate_dith.wav in Wavelab and used Analysis>Audio File Comparer to compare
RESULT: The two files are
NOT IDENTICALHmm. Seems that applying the uv22hr dither algorithm to one big file vs. two small files does yeild different results when the two files are combined. I had the Wavelab compare process create a delta file with markers. Visually, it appears to be a flat line. Listening to it yeilds dead silence. Viewing it on the FFT meter shows noise at aprox -140db across the frequency range.
What does this tell us? Somethign to debate, but my inital thoughts are simply that the complicated math problem that is dither doesn't yeild identical results when the source files are different...which they fundamentally are (in a bits and bytes sort of way) in this situation. To me, the more I think about it, that seems to make some logical sense. Does it really matter or mean anything? I could argue not, b/c there really is no basis for comparison. Whether you do it in batch or as one file, you are processing the file and altering it resulting in a file that IS different from the 24/96 source. True, they are different relative to each other, but both are different relative to the original source.
Moving on to the issue w/ the .wav not lining up when combined. Personally I don't see it...mine appear to join fine when viewed at the smallest zoom possible in Wavelab. Am I missing something here?
I very well may not know what I'm looking for.
And finally, the SBEs. I didn't have to test it, b/c I know it to be true. 24/96 files dithered and resampled result in files w/ SBE's. HOWEVER, I believe the issue resides in Wavelab itself, NOT the fact that this processing is done in batch. In my experience, every 16/44.1 file saved in Wavelab has a SBE b/c by default Wavelab doesn't save the file on a CD defined sector boundry. CDWav does, which is why we all use it to cut tracks. As a result, the comparison we're talking about relative to SBEs is really a comparison of how Wavelab and CDWav save 16/44.1 files. True, they are different, and Wavelab does screw it up. I got an interesting PM about how to address this issue...
---------
markers.....make sure you do CD track start, CD Track splice markers(for the individual tracks) and cd track end...make sure you check "quantize to CD Frame" THEN do an autosplit.no more SBEs!
Teddy
---------
Doesn't seem possible to work Teddy's process into this since the files are already split. The only way to really address the SBEs is w/ shntool fix. It's essentially the same thing though, since it just adjusts the split relative to the next file from the first SBE to the last and pads the final file.
So there you have it. Feel free to shoot holes in my theory/testing, but it appears to me this is 6 one way, 1/2 a dozen the other unless there is more to the wavs not lining up when re-joined. Which I agree in theory certainly seems possible if the dither/resample process alters the end of the first wav and the beginning of the second wav differently. Any difference would seem to be way down at -140db and not audible. But again, I didn't see it, and would love to see a pic of an example if Damon or someone else can grab one.
Finally, and I'll shut up now. After all of this, I would agree that the process Nick outlined makes the most sense and seems the "safest" and that is what I'll be doing from now on although I find this all very interesting from a purely "academic" perspective.
Damon, guess we can be dork buddies.
mitch