Skip to main content
Topic: 32bit WAV to 32bit FLAC, what are my options? (Read 30478 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

32bit WAV to 32bit FLAC, what are my options?

Reply #25
From the FLAC documentation:
"FLAC supports from 4 to 32 bits per sample. Currently the reference encoder and decoders only support up to 24 bits per sample."

32bit WAV to 32bit FLAC, what are my options?

Reply #26
Wottha, please note that foobar2000 switched to 32-bit float audio pipeline with version 0.9 and has remained that way since. This means it can't handle 32-bit integer format losslessly itself.

32bit WAV to 32bit FLAC, what are my options?

Reply #27
Thank you everyone.  So if someone, for whatever reason, wanted to offer "32 bit" files for download, the only "practical" option would be 32 bit float (not fixed) WAV or AIFF?  Because those could be handled correctly by Foobar2000 and other players?  So not FLAC (or ALAC?) if they wanted to retain "32 bit"?  And 32 bit fixed is not a good option in any case because of the players?
I'm not suggesting this is a good idea or will make the least difference, but it's a request not unheard of.

32bit WAV to 32bit FLAC, what are my options?

Reply #28
So not FLAC (or ALAC?) if they wanted to retain "32 bit"?

As mentioned earlier in this thread, the only option is WavPack or WAV.

Quote
I'm not suggesting this is a good idea or will make the least difference, but it's a request not unheard of.

It is a very, very bad idea. Any bit beyond 21 can only be noise, thats pretty much dictated by the laws of physics. More reasonable, any bit beyond 18 is probably only containing noise in a final mix/master, and most music these days does not even contain 16 bits of real information. 24 bit as a format is pretty dumb (20 bit is enough, even if you want to be absolutely sure), 32-bit int is downright stupid. 32-bit float is stupid as well, as a final format. Its only purpose is to conserve non-normalized data which might be the result of some mixing step. So, it should only ever be used as an intermediate format.
Music: sounds arranged such that they construct feelings.


32bit WAV to 32bit FLAC, what are my options?

Reply #30
So if someone, for whatever reason, wanted to offer "32 bit" files for download, the only "practical" option would be 32 bit float (not fixed) WAV or AIFF?

There are no practical options. BTW, 32-bit float is roughly equivalent to 25-bit fixed.

Because those could be handled correctly by Foobar2000 and other players?

Why do you think that discarding 7 bits from 32-bit fixed input is not a "correct handling"?

32bit WAV to 32bit FLAC, what are my options?

Reply #31
I can imagine someone needs storing e.g. computer-generated 32bit samples for further processing/research etc. But a player is by principle followed by a DA conversion. No DA conversion can utilize more than 24 bits (actually much less, as has already been discussed).

32bit WAV to 32bit FLAC, what are my options?

Reply #32
Also, about 32-bit files...


32bit WAV to 32bit FLAC, what are my options?

Reply #33
Because those could be handled correctly by Foobar2000 and other players?

Why do you think that discarding 7 bits from 32-bit fixed input is not a "correct handling"?

Sorry, bad choice of wording?  Should have dropped the word "correctly"?  Everything I say is followed by a question mark because I don't know. 
I found this discussion where the "correct answer" doesn't appear to be correct if the whole thread is read.
https://forums.adobe.com/message/5662484#5662484
Since Pro Tools added 32 bit float files, more and more of these files are passing between studios.  Somebody else gets wind of this and thinks it'll be a good idea to sell it to purists.  So I think your cartoon might come to pass.

32bit WAV to 32bit FLAC, what are my options?

Reply #34
Since Pro Tools added 32 bit float files, more and more of these files are passing between studios.  Somebody else gets wind of this and thinks it'll be a good idea to sell it to purists.  So I think your cartoon might come to pass.
I guess most of these people still treat audio files like magnetic tape or similar. With tapes you'd want to use the absolute best quality, because of degradation issue and wear, so that even after some time you still have enough fidelity for a well sounding record. When they apply this reasoning to digital audio, they will choose whatever format has the biggest number. From my impression, what is rarely understood by those engineers, is that once your signal is in the digital domain (for example as PCM) there will be no degradation upon copying and storing files. Bits cannot rot. So even if you want to do heavy processing and leave some headroom, the aforementioned 20 or 24 bits are ample, because you only need that headroom for processing, not for storage. The same logic applies in the frequency domain.
It's only audiophile if it's inconvenient.

32bit WAV to 32bit FLAC, what are my options?

Reply #35
So not FLAC (or ALAC?) if they wanted to retain "32 bit"?

As mentioned earlier in this thread, the only option is WavPack or WAV.

Well, I suppose there is OptimFROG float and I thought another older program could also handle 32-bit float. I recommend WavPack because there are DAW programs that recognize it natively (i.e. Reaper) or with a plug-in (Audition).
"Something bothering you, Mister Spock?"

32bit WAV to 32bit FLAC, what are my options?

Reply #36
This is what I show to my class regarding 32 bit float:

http://www.youtube.com/watch?v=Qt-EJhDDHUI

It's not a delivery format as far as I'm concerned, just processing.

32bit WAV to 32bit FLAC, what are my options?

Reply #37
This is what I show to my class regarding 32 bit float:
http://www.youtube.com/watch?v=Qt-EJhDDHUI
It's not a delivery format as far as I'm concerned, just processing.

Thank you Moni.  That's a great video.
As far as delivery formats though, aren't they all post-processed, by upsampling in all D to A converters?  Can't some "32 bit" DAC chips accept 32 bit float without truncation?  Not to mention all the MP3 conversions done from delivery formats.  Apple makes a point of converting to 32 bit float caf before encoding AAC for iTunes.

32bit WAV to 32bit FLAC, what are my options?

Reply #38
As far as delivery formats though, aren't they all post-processed, by upsampling in all D to A converters?  Can't some "32 bit" DAC chips accept 32 bit float without truncation?


The best you can get from a DAC these days is about 20-21 bits, so 32 bit DAC chips don't actually exist. 

Not to mention all the MP3 conversions done from delivery formats.  Apple makes a point of converting to 32 bit float caf before encoding AAC for iTunes.


All processing these days is done in floating point.  That doesn't have anything to do with the delivery format though.  The precision of the operating used determines the accumulation of rounding error.

32bit WAV to 32bit FLAC, what are my options?

Reply #39
The best you can get from a DAC these days is about 20-21 bits, so 32 bit DAC chips don't actually exist.

This is what has me confused.  The DACs that call themselves 32 bit.  Can they not accept 32 bit float input?

All processing these days is done in floating point.  That doesn't have anything to do with the delivery format though.  The precision of the operating used determines the accumulation of rounding error.

If all 24 and 16 bit files are going to be post processed 32 bit float by SRC in DAC, why the back and forth, 32 in DAW to 24 from DAW to 32 float SRC in DAC.  Is it better to deliver 24?  I guess the back and forth is miniscule, and I'll drop it, but why is the back and forth so considered necessary?  And preferable?

32bit WAV to 32bit FLAC, what are my options?

Reply #40
The best you can get from a DAC these days is about 20-21 bits, so 32 bit DAC chips don't actually exist.

This is what has me confused.  The DACs that call themselves 32 bit.  Can they not accept 32 bit float input?

Some of them accept 32 bit input - that doesn't mean that they have a resolution that is more than 20 bits.

Quote
If all 24 and 16 bit files are going to be post processed 32 bit float by SRC in DAC, why the back and forth, 32 in DAW to 24 from DAW to 32 float SRC in DAC.  Is it better to deliver 24?  I guess the back and forth is miniscule, and I'll drop it, but why is the back and forth so considered necessary?  And preferable?

Less network traffic, less i/o traffic, less disc traffic, less CPU load, and less wasted disk space.

32bit WAV to 32bit FLAC, what are my options?

Reply #41
are going to be post processed 32 bit float by SRC in DAC


Floating point calculations in DAC? I doubt it.

32bit WAV to 32bit FLAC, what are my options?

Reply #42
Less network traffic, less i/o traffic, less disc traffic, less CPU load, and less wasted disk space.

Yes, sorry I meant to put in "besides file size".  All very good reasons, but I meant in the case of a purist end user who doesn't care about those things and has gear that can handle it.  Certainly not for everyone.  Sorry.

32bit WAV to 32bit FLAC, what are my options?

Reply #43
The input format a DAC uses doesn't matter so long as it is greater than its effective number of bits.

32bit WAV to 32bit FLAC, what are my options?

Reply #44
I doubt regular dacs (i.e. not embedded computers) accept float32. I have never heard of any. Accepting int32 is for convenience purposes - no convertsion from the commonly used int32 to the audio-world-only int24 format.

I doubt regular dacs implement their filters using floating-point calculations. Extending the calculation to int48 is common though, technically simple solution.

32bit WAV to 32bit FLAC, what are my options?

Reply #45
Use DACs generally use integer filters. Floating point is rarely used by hardware devices.


Re: 32bit WAV to 32bit FLAC, what are my options?

Reply #47
This is what I show to my class regarding 32 bit float
I should say that this is clearly not about bit depth but using floating point samples instead of integer samples. Floating point can repesent numbers above 1.0 which is the clipping threshold during playback. The data is there, but it's clipped when sent to the DAC which expects integer samples, so 0.0 to 1.0 is converted to the integer sample range.
It would work even with 16bit floating point audio. (however I don't know how 16bit floating point compares to 16bit integer audio in quality terms, but 24bit floating point should be still high quality enough for even studio stuff i guess)
WavPack -b4x4hc
Opus --cvbr --bitrate 256 --framesize 5

Re: 32bit WAV to 32bit FLAC, what are my options?

Reply #48
16 bit (half precision) floats have (effectively) 11 bits of precision, so quality would be quite poor.  There is no single standard definition for 24 bit floats, but they commonly have around 17 bits of precision.

 

Re: 32bit WAV to 32bit FLAC, what are my options?

Reply #49
From the FLAC documentation:
"FLAC supports from 4 to 32 bits per sample. Currently the reference encoder and decoders only support up to 24 bits per sample."

I don't want to discuss if you like 32 bits or why I need 32 bits so please stay on topic... Not all people use audio just for listening music.

Now. Flac can support 32 bits so... not the official one, but what encoder do I need? Or the developers supported it to then not create a encoder capable of 32 bits? (which seems funny to me...)

 
SimplePortal 1.0.0 RC1 © 2008-2019