Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Going lossless with 32bit & CoolEditPro 2.0 (Read 5844 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Going lossless with 32bit & CoolEditPro 2.0

Hi, All

I have a library of 32bit audio files and editing them via CEP 2.0. Lossless coding gives me avg 70% reduction in size - pretty impressive, given the fact that i need all the library(20Gb in wavs) on the harddrive.
Used FLAC (sp. CEP FLAC filter) successfully until some problems arose(here). As I know FLAC format does support 32bit, while current encoder does not. Seems like CEP filter managed to overcome this restriction, maybe unintentionally , but not 100% stable & bugfree

So, are there any lossless codecs that I can use for 32bit? (GUI or CEP filter is a plus) 
 
P.S.Rendering audio back and forth between 16bit(for packing) & 32bit(for editing) is hardly an option 

Going lossless with 32bit & CoolEditPro 2.0

Reply #1 –
My FLAC filter saves 32 bit float data in 24 bit linear format. Reference FLAC library does not support anything higher and float isn't supported at all. This is not lossless but quality loss is minimal. Only lossless format that supports 32 bit float format that Cool Edit/Audition uses is latest alpha version of OptimFROG.

Going lossless with 32bit & CoolEditPro 2.0

Reply #2 –
Yeah, checked it once more - 32bit becomes 24 while FLACing. Effectively, what I'm doing now is resampling to 24 for saving, returning to 32 while in CEP. Well, max -140dB distortion is fine  , could be much worse with 16.
However, sometimes for no apparent reason the filter just refuses to do the job, and that's the reason for my question. Thanks for OptimFrog idea. Still have to wait for the filter, though, I guess

Going lossless with 32bit & CoolEditPro 2.0

Reply #3 –
Minor corrections: when you shrink bit depth, it is called truncating or quantizing, not resampling.

Also, the loss of bits is not as great as it looks. The 32 bits of the float format are divided into 1 sign bit, 8 exponent bits and 24 mantissa bits (including one hidden mantissa bit). The 24 bit integer is effectively 1 sign bit and 23 mantissa bits. So you lose the exponent, which is not important for storage.

Going lossless with 32bit & CoolEditPro 2.0

Reply #4 –
Quote
So you lose the exponent, which is not important for storage.

But 6dB down from full scale, the exponent will ensure that you have a leading 0, and 24 bits of data (down to the 25th linear bit). 12dB down from full scale, the exponent will ensure that you have two leading 0s, and 24 bits of data (down to the 26th linear bit). and so on...

So, assuming you processed the file in some way to fill those extra bits, you do lose accuracy that was present in the 32-bit version when converting to 24-bits.

That's probably irrelevant. What could be relevant is if you expect to maintain the 32-bit data, and so you don't watch your levels when processing. CEP/audition 32-bit data maintains values over digital full scale (several hundred dB above digital full scale) without clipping. Converting to 24-bit integers will hard clip anything above digital full scale.

That could be a nasty shock if you intend to reduce the level later, and find there's nothing above digital full scale, and it's all now clipped.

This is something I often use 32-bit data for. So long as it's clear that it's being converted to 24-bit data, it's not a problem. But it would be wrong for the programmer to assume that a silent conversion is OK, because it could cause very serious problems in the above instance.

Cheers,
David.

Going lossless with 32bit & CoolEditPro 2.0

Reply #5 –
The extra resolution and dynamic range are extremely useful for processing, but not important for storage, as I said. Data beyond 20 bits is inaudible, it's just there to fend errors off. If you make sure that you don't clip when truncating (by making sure 24 bit full scale > highest peak in float original), you won't experience any discernible loss.

Going lossless with 32bit & CoolEditPro 2.0

Reply #6 –
Quote
So, are there any lossless codecs that I can use for 32bit? (GUI or CEP filter is a plus)

Another lossless compressor that can handle 32-bit files is TTA. There's no CoolEdit filter available, but there is a front-end.

Edit: typo

Going lossless with 32bit & CoolEditPro 2.0

Reply #7 –
Wow. I had not heard about TTA.

But, it seems to me it compresses 32 bit integers, and the guy here needs to compress 32 bit floats. Using it on floats straight will probably degrade compression.

I think lossless compression of float audio is a very difficult proposition because of round-off errors, especially during subtraction.

Going lossless with 32bit & CoolEditPro 2.0

Reply #8 –
I've had good luck with Monkey's Audio and Cool Edit Pro for 24 bit files. Look for the Cool Edit Ape Filter. Silly name, good codec.

Going lossless with 32bit & CoolEditPro 2.0

Reply #9 –
Quote
My FLAC filter saves 32 bit float data in 24 bit linear format. Reference FLAC library does not support anything higher and float isn't supported at all. This is not lossless but quality loss is minimal. Only lossless format that supports 32 bit float format that Cool Edit/Audition uses is latest alpha version of OptimFROG.

I got quite a surprise while reading this. I remember a very big point made by the FLAC project - that FLAC is 100% lossless, and that never will it include lossy modes of any kind. Isn't this going against the primary goal? I know it's not lossy in the sense of MP3/MPC/OGG etc, but it's still not 100% lossless as they claim.

Going lossless with 32bit & CoolEditPro 2.0

Reply #10 –
Quote
I got quite a surprise while reading this. I remember a very big point made by the FLAC project - that FLAC is 100% lossless, and that never will it include lossy modes of any kind. Isn't this going against the primary goal? I know it's not lossy in the sense of MP3/MPC/OGG etc, but it's still not 100% lossless as they claim.

They are using FLAC in a way that it wasn't designed for, so I don't think this counts as the developers going back on their word.
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Going lossless with 32bit & CoolEditPro 2.0

Reply #11 –
Quote
They are using FLAC in a way that it wasn't designed for, so I don't think this counts as the developers going back on their word.

Well ok, that makes sense.  However, I'd prefer if it gave a warning or a flat out error saying it's not supported and "some loss may occur". As was mentioned above, doing this silently can mislead people into thinking it's still a lossless file...

Going lossless with 32bit & CoolEditPro 2.0

Reply #12 –
Quote
However, I'd prefer if it gave a warning or a flat out error saying it's not supported and "some loss may occur". As was mentioned above, doing this silently can mislead people into thinking it's still a lossless file...

I agree that confusion is possible there. The WavPack plugin for Cool Edit has a dialog that clearly shows that float data is being converted to something less (16, 20, or 24-bits), but if you don't open the dialog you will never see it.

BTW, the WavPack plugin is (I believe) the only lossless Cool Edit plugin that stores the non-audio information (like cues).

Going lossless with 32bit & CoolEditPro 2.0

Reply #13 –
Quote
However, I'd prefer if it gave a warning or a flat out error saying it's not supported and "some loss may occur". As was mentioned above, doing this silently can mislead people into thinking it's still a lossless file...

Well, I should say that CEP definitely displays dialog window with smth like "this might be a lossy operation, be careful" when saving into FLAC via filter. Switched it off long ago, so I can't say if this is said for every initial format of the wave, or just those >24.

Going lossless with 32bit & CoolEditPro 2.0

Reply #14 –
If you're getting 70% compression I suspect that 32 bit float is overkill. In practice I'd say it's overkill anyway. I can't imagine what sort of sound your recording that needs that much percision. Full symphonies benefit from 24 bit but I've never heard any sort of benefit from 32 . . .

Going lossless with 32bit & CoolEditPro 2.0

Reply #15 –
Quote
32 bit float is overkill

Well, CEP has three options - 8,16,32bit. I constantly use filters, noise reduction and other complex stuff. I'm not sure if CEP can do right dithering, so I prefer the last one