1 hour ago
TL;DR: You can stretch/resample audio originally intended to run along with 24/1.001 fps (instead of "Captain Jack") to do so with 24.000 fps or also only reflag the sample rate to run slightly faster, theoretically preserving the quality.
Hi there,
the following is for sure rather academic and "nerdy" in terms of being most probably inaudible anyway. Technically, resampling audio is nothing bad per se, contrary to common audiophile misbelief, as most DACs for example will already to that for the sake of oversampling, jitter suppression (e.g. via ASRC) or matching some sweet spot of a crystal's/DAC's native sample rate (as Benchmark did prominently with their "DAC1"), etc.
However, from an archival's and psychological standpoint, it makes sense to mangle the data as little as possible.
That was the preamble so to say. Now recently, I again edited and synchronized a few Cinema DTS tracks to certain movie releases. Those usually run at 24/1.001 (~ 23.976) fps (thanks, NTSC legacy!) and are encoded at 48000 Hz. Cinema DTS however, being sampled at 44100 Hz, is intended to run along 24.000 fps as standard in the cinema and thus will be too fast by a factor of 1.001.
Most people here would thus stretch the Cinema DTS 44100 Hz audio to be a bit slower and match the 24/1.001 fps timing and maybe also resample the whole thing to become 48000 Hz to somewhat better match the usual DVD/BD/UHD BD specifications. Either step will be lossy by definition, but as mentioned in the introduction, not a real issue.
If one wants to avoid that and only keep the (usually best anyway) Cinema DTS track, one can quite easily go ahead and just change the corresponding nominal frame rate of the video track to be 24.000 fps instead of 24/1.001 fps by e.g. eac3to (the parameter "-changeto24.000" will do). Remux everything back and voila, one ends up with 24.000 fps video (which accordingly to eac3to is "unusual", but actually used as such officially for often European releases) and the proper untouched audio.
But, if one also wants to keep the usually provided audio tracks from the 24/1.001 fps source, one runs into the problem of having to alter the video playback frame rate depending on the used audio track (24.000 fps for the Cinema DTS track and 24/1.001 fps for "the others" (movie pun intended, Kidman is great)). Unfortunately, no usual player I'm aware of, supports that on the fly.
So I thought of another solution which turns out to work surprising well: instead of resampling or changing the speed of the 24/1.001 fps audio tracks, one can instead reflag the 48000 Hz of those to become 48048 Hz (e.g. with Audacity), covering for exactly the timing difference, as 48000 Hz for 24/1.001 fps equals 48048 Hz for 24 fps. It is a nice mathematical coincidence that the speedup works out to perfectly fit integer-wise as with 44100 Hz, it would already be fractorial.
48048 Hz are definitely non-standard and it could be very well that most players internally mangle the 48048 Hz to 48000 Hz anyway (then providing not even the theoretical advantage of having left the audio material untouched) but the data itself one will still have kept unaltered which is always nice for archiving. Sadly, my Denon's status information isn't exact enough as it only displays "48 kHz", so without further measurements, I don't know if it actually gets 48048 Hz from the Ugoos or resampled 48000 Hz. Maybe, I'll check that via S/PDIF some other time.
I haven't found a way to reflag AC3 or DTS files to run at 48048 Hz though and it could be very well possible that it's simply not allowed by any standard, although technically possible of course as it's only a matter of eventual processing and DAC timing and not the actual encoded data. So one would have to decode that stuff to PCM beforehand.
Feel free to let me know what you think of that, just some idea I came up with.
Hi there,
the following is for sure rather academic and "nerdy" in terms of being most probably inaudible anyway. Technically, resampling audio is nothing bad per se, contrary to common audiophile misbelief, as most DACs for example will already to that for the sake of oversampling, jitter suppression (e.g. via ASRC) or matching some sweet spot of a crystal's/DAC's native sample rate (as Benchmark did prominently with their "DAC1"), etc.
However, from an archival's and psychological standpoint, it makes sense to mangle the data as little as possible.
That was the preamble so to say. Now recently, I again edited and synchronized a few Cinema DTS tracks to certain movie releases. Those usually run at 24/1.001 (~ 23.976) fps (thanks, NTSC legacy!) and are encoded at 48000 Hz. Cinema DTS however, being sampled at 44100 Hz, is intended to run along 24.000 fps as standard in the cinema and thus will be too fast by a factor of 1.001.
Most people here would thus stretch the Cinema DTS 44100 Hz audio to be a bit slower and match the 24/1.001 fps timing and maybe also resample the whole thing to become 48000 Hz to somewhat better match the usual DVD/BD/UHD BD specifications. Either step will be lossy by definition, but as mentioned in the introduction, not a real issue.
If one wants to avoid that and only keep the (usually best anyway) Cinema DTS track, one can quite easily go ahead and just change the corresponding nominal frame rate of the video track to be 24.000 fps instead of 24/1.001 fps by e.g. eac3to (the parameter "-changeto24.000" will do). Remux everything back and voila, one ends up with 24.000 fps video (which accordingly to eac3to is "unusual", but actually used as such officially for often European releases) and the proper untouched audio.
But, if one also wants to keep the usually provided audio tracks from the 24/1.001 fps source, one runs into the problem of having to alter the video playback frame rate depending on the used audio track (24.000 fps for the Cinema DTS track and 24/1.001 fps for "the others" (movie pun intended, Kidman is great)). Unfortunately, no usual player I'm aware of, supports that on the fly.
So I thought of another solution which turns out to work surprising well: instead of resampling or changing the speed of the 24/1.001 fps audio tracks, one can instead reflag the 48000 Hz of those to become 48048 Hz (e.g. with Audacity), covering for exactly the timing difference, as 48000 Hz for 24/1.001 fps equals 48048 Hz for 24 fps. It is a nice mathematical coincidence that the speedup works out to perfectly fit integer-wise as with 44100 Hz, it would already be fractorial.
48048 Hz are definitely non-standard and it could be very well that most players internally mangle the 48048 Hz to 48000 Hz anyway (then providing not even the theoretical advantage of having left the audio material untouched) but the data itself one will still have kept unaltered which is always nice for archiving. Sadly, my Denon's status information isn't exact enough as it only displays "48 kHz", so without further measurements, I don't know if it actually gets 48048 Hz from the Ugoos or resampled 48000 Hz. Maybe, I'll check that via S/PDIF some other time.
I haven't found a way to reflag AC3 or DTS files to run at 48048 Hz though and it could be very well possible that it's simply not allowed by any standard, although technically possible of course as it's only a matter of eventual processing and DAC timing and not the actual encoded data. So one would have to decode that stuff to PCM beforehand.
Feel free to let me know what you think of that, just some idea I came up with.
