Gets stranger... doing the click file test in each app using a stereo click file, using exactly the same procedure as with the mono file, gives a correct result in both Reaper and Audition. So it's the music file that's giving the difference.
Aha. The music file is 16 bit, and I'm doing a 32 bit float mixdown. The click test file is 32 bit float. Perhaps Reaper is doing something different internally with that 16 bit file compared to the way Audition is treating it, although Audition's result (based on examining one sample in the mixdown) is correct. Next test, use a 32 bit float music file for the tests.
DAW Audio Quality does vary... and Live 7 sneak peak!
-
- KVRist
- 291 posts since 16 Apr, 2006
OK, got it pinned - if I use a 32 bit float file as the source file for the test, Reaper and Audition come up with an identical result (perfect null).
However, I would question why Audition will take the 16 bit file, sum it to itself into a 32 bit float file, and come up with a demonstrably correct result (the original sample values are doubled). Reaper produces a slightly incorrect result in that scenario (the sample values are not quite doubled).
If anyone can see a flaw in my tests, do tell me!
Maybe I should repeat with an undithered all-16-bit test....
However, I would question why Audition will take the 16 bit file, sum it to itself into a 32 bit float file, and come up with a demonstrably correct result (the original sample values are doubled). Reaper produces a slightly incorrect result in that scenario (the sample values are not quite doubled).
If anyone can see a flaw in my tests, do tell me!
Maybe I should repeat with an undithered all-16-bit test....
-
- KVRist
- 291 posts since 16 Apr, 2006
Yup, 16 bit sum of 16 bit stereo file in both apps does give a perfect inversion, so there's no problem with Reaper in this scenario. And my test procedure is validated IMHO.
So, the question remains concerning Reaper's handling of 16 bit files in a 32 bit mix. Reaper, unlike Audition, is capable of using a variety of file types within a session, and perhaps it's a consequence of some sort of resampling going on during mixdown of a 16 bit file to 32 bits that is causing the problem??? Probably Justin could quickly explain what happens but he's holidaying - and it can very easily await his return!
So, the question remains concerning Reaper's handling of 16 bit files in a 32 bit mix. Reaper, unlike Audition, is capable of using a variety of file types within a session, and perhaps it's a consequence of some sort of resampling going on during mixdown of a 16 bit file to 32 bits that is causing the problem??? Probably Justin could quickly explain what happens but he's holidaying - and it can very easily await his return!
-
- KVRist
- 291 posts since 16 Apr, 2006
And a further question might be....
If I play that 16 bit file on two tracks in Audition, and the same in Reaper, what is being passed to the soundcard? Is it the correct 16 bit to 16 bit sum in Reaper, or is it the seemingly incorrect 16 bit to 32 bit sum, and could that account for an audible difference if you have really golden ears?
If I play that 16 bit file on two tracks in Audition, and the same in Reaper, what is being passed to the soundcard? Is it the correct 16 bit to 16 bit sum in Reaper, or is it the seemingly incorrect 16 bit to 32 bit sum, and could that account for an audible difference if you have really golden ears?
-
- KVRist
- 357 posts since 16 Jun, 2006 from Stockholm - Sweden
Just thinking aloud here:
A 16 bit wav goes from -32767 to 32768. So the positive side has slightly more resolution!
(0 is 0 in both float and int)
When converting a float file to 16 bit integer you could do two things:
(this is based on a float file that doesn't exceed -1.0 or 1.0)
A:
float value * 32767 - results in a file that doesn't clip cos it goes from -32767 to +32767. It never goes to 32768 which is the maximum positive value.
B:
positive float value * 32768
negative float value * 32767
I think the last one is the correct one, but maybe some conversion in some softwares use the simpler A solution?
It would certainly not be audible, but probably measurable.
Same is true for 24 bit files, but the error in using soultion A is extremely small then.
A 16 bit wav goes from -32767 to 32768. So the positive side has slightly more resolution!
(0 is 0 in both float and int)
When converting a float file to 16 bit integer you could do two things:
(this is based on a float file that doesn't exceed -1.0 or 1.0)
A:
float value * 32767 - results in a file that doesn't clip cos it goes from -32767 to +32767. It never goes to 32768 which is the maximum positive value.
B:
positive float value * 32768
negative float value * 32767
I think the last one is the correct one, but maybe some conversion in some softwares use the simpler A solution?
It would certainly not be audible, but probably measurable.
Same is true for 24 bit files, but the error in using soultion A is extremely small then.
www.xlnaudio.com - Creators of "Addictive Drums"
Demo version available for download!
Check out the new Jazz Paks too
Demo version available for download!
Check out the new Jazz Paks too
-
- KVRAF
- Topic Starter
- 7489 posts since 6 Jul, 2004
Lots of interesting stuff there Art - thanks for doing those tests and sharing your results.
(by the way, I'm assuming you kinow there is an option in Audition to choose whether all files are converted to 32-bit on import...? Anyway, it's nice to get your confirmation that it does this accurately)
(by the way, I'm assuming you kinow there is an option in Audition to choose whether all files are converted to 32-bit on import...? Anyway, it's nice to get your confirmation that it does this accurately
)