Folks,
I've got quite a bit of 24/96 material on my computer that is in 5.1 surround. Given that most of us accept it as true that 16/44.1 vs 24/96 makes no audiable difference, is there any reason not to dither these files down to 16/44.1 in 5.1 channels?
It would save me disc space, reduce my network activity and also reduce my CPUs cock cycles.
Thanks
I think the best is to try yourself, if you can hear any difference on your gear. If not, go for the conversion if you need the drive (and especially backup) space, I would not worry about network and CPU load in either case.
It would save me disc space
Maybe. Probably not much. Test some files with wildly different aural densities (think prog metal versus minimal electronica) and see if the average percentage is worth it.
Forgive me if I'm wrong, but I got the impression that you have a few dollars to spare, so instead of wondering about file quality versus size, just sidestep the issue and add a multi-terabyte hard drive. That's wht I did and it's the greatest decision I ever made. You can't buy peace of mind for less.
reduce my network activity
Barely.
reduce my CPUs cock cycles.
Barely.
Also Freudian slip?
Generally, for any desktop computer since '95, playing music is a lightweight operation.
If the source files are lossless you just need to convert to a lossy format like AAC, if the source is already lossy, you're probably better off leaving them the way they are.
Forgive me if I'm wrong, but I got the impression that you have a few dollars to spare, so instead of wondering about file quality versus size, just sidestep the issue and add a multi-terabyte hard drive. That's wht I did and it's the greatest decision I ever made. You can't buy peace of mind for less.
When thinking about disk space and just for peace of mind, remember to always think twice (at least, and physically separate)!
If it's lossy, don't touch it.
If it's uncompressed, use lossless compression.
If it's already lossless, you could save a surprising amount of space by remove the last four bits (i.e. making it 20-bit). It's hard to imagine any scenario (measurement, analysis, compression, never mind normal listening) where this would be detectable.
While 48kHz/20-bit, 48kHz 16-bit, and 44.1kHz 16-bit should all also be safe, I wouldn't do it, except for compatibility. I like having 24/96 content around to demonstrate that it sounds the same as 44.1kHz/16
With lossless compression, the difference in bitrate between 20/96 and 20/48 won't be as great as 2:1.
Cheers,
David.
With lossless compression, the difference in bitrate between 20/96 and 20/48 won't be as great as 2:1.
Yes, but wouldn't you expect the savings going from 24 bit to 16 bit to be more than a third?
With lossless compression, the difference in bitrate between 20/96 and 20/48 won't be as great as 2:1.
Yes, but wouldn't you expect the savings going from 24 bit to 16 bit to be more than a third?
Yes. Much more.
However, if someone releases a 24/96 5.1 mix with buckets of headroom, I wouldn't feel 100% confident saying that converting 24-bits to 16-bits would be inaudible in every possible circumstance. I'd struggle to find
any content that I could ABX, but it's not impossible that something somewhere is recorded/mixed so quietly that the difference can be heard. If anyone mixes 5.1 at the final Dolby reference level (dial norm -31dB), that's obscenely quiet by CD standards.
Cheers,
David.
..cock cycles.. Yeah, that's what i get for posting at 2:30 am. So far i don't have any bluray porngraphy, so dithering that to 16-bit is not an issue.
I did forget to mention that my files are in wavpack, if that makes any difference. I guess i'll just try it.
Anyone have any reccomendations for a good linux commandline tool to do this transparently? Mencoder? Sox?
Thanks for the responces.
With lossless compression, the difference in bitrate between 20/96 and 20/48 won't be as great as 2:1.
Yes, but wouldn't you expect the savings going from 24 bit to 16 bit to be more than a third?
Yes. Much more.
2Bdecided is right. I have experienced forty percent and more, if all the last 8 bits are used. Because then they are mainly "used" for noise, and noise is hard to compress.
Anyone have any reccomendations for a good linux commandline tool to do this transparently? Mencoder? Sox?
caudec (http://caudec.googlecode.com/) version 1.3.5 can do it easily (it uses sox). Run something like this:
$ caudec -o /path/to/new/dir -n 1 -b 16 -r 48000 -c wv /path/to/files/*.wv
Let me know if you try it and run into any problems. There's an AUR package if you run Arch Linux.
How about keeping the 24 bits but reencoding to wavpack lossy? You should still get a dramatic decrease in size, but you can always return to full 24 bit lossless whenever you want.
Edit: I mean, of course, that the file that you play will be much smaller, but the total size of the two files will be somewhat larger than the lossless file alone.
caudec (http://caudec.googlecode.com/) version 1.3.5 can do it easily (it uses sox). Run something like this:
$ caudec -o /path/to/new/dir -n 1 -b 16 -r 48000 -c wv /path/to/files/*.wv
Let me know if you try it and run into any problems. There's an AUR package if you run Arch Linux.
Thanks skamp. I'll give that a shot. If i can't get it to work then i'll just mess with it in sox directly.
@pdq: Something i've never quite understood: When you have a wavpack lossy file and a correction file, can you play back the equiv of a lossless version without re-encoding as lossless (currently using mpd as my player) If so, i might just do that to all my tracks, provided i can do so an keep all my custom tags.
Thanks
@pdq: Something i've never quite understood: When you have a wavpack lossy file and a correction file, can you play back the equiv of a lossless version without re-encoding as lossless (currently using mpd as my player) If so, i might just do that to all my tracks, provided i can do so an keep all my custom tags.
It depends on the player. On Windows, most players automatically look for and use the correction file for lossless playback (and even the DirectShow filter does it, which was quite a trick). On Linux it's a little more dicey because Gstreamer and FFmpeg do not use them, and a lot of players use one of those to play WavPack files. If a program uses libwavpack (like deadbeef) then it's easy to set it up so that it works. I'm not sure about mpd.
Of course, it would be easy to encode at a high enough bitrate that the lossy files will be completely transparent all by themselves (and still save a huge amount from your original files). I suggest something around 1024 kbps to start.
Thank you Bryant. I'll play with it and see what satisfies me.