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: FLAC 32 bit float (Read 3719 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

FLAC 32 bit float

Hi,

I am using the latest version of foobar2000 v2.0 and Encoder Pack 2023-07-05. I am converting wav 96khz 32 bit float, to FLAC, and I choose auto as bit depth. The output is FLAC 32 bit.

It plays on foobar, but it doesn't play on anything else such as VLC or media players etc.

Is this because FLAC in 32 bit is too new, and nothing supports it apart from foobar?

Thanks!


Re: FLAC 32 bit float

Reply #1
[1] Latest version of flac encoder supports up to 32-bit integer (aka fixed point), not float. Read below:
https://xiph.org/flac/2022/09/09/flac-1-4-0-released.html

[2] Older hardware and software may not support 32-bit integer flac files.
[3] If 32-bit float lossless compression is required, use WavPack.


Re: FLAC 32 bit float

Reply #3
This will allow you doing some interactive math using a real programming language instead of simply reading.
https://hydrogenaud.io/index.php/topic,112361.msg1012996.html#msg1012996

[edit]
For hardware devices claimed to support 32-bit float recording from analog input, have a look on this thread as well:
https://hydrogenaud.io/index.php/topic,124898.0.html

Re: FLAC 32 bit float

Reply #4
Quote
I heard somewhere that the audio quality for 24 vs 32 is same?,
32-bits gives you more resolution and more dynamic range.   But there are limits to your hearing and electronics.     16-bits is generally better than human hearing so you can down-sample from high-resolution to "CD quality" and you won't hear a difference in a blind ABX test.  i.e. The "sound quality" isn't better.   That's assuming the floating-point file doesn't go over 0dB.    If it does, the integer data will be clipped (distorted) unless you lower the volume before converting.

I don't think there are any 24-bit ADCs or DACs that have more than about 20-bits of accuracy.   They are limited by noise and/or non-linearity (on the analog side).

Quote
the extra 8 bits is to allow increase of gain when editing without distortion??
True.  Integer formats can't go over 0dBFS (zero decibels full-scale).   0dBFS is defined as the maximum you can "count to" with a given number of bits.  The numbers in a 24 bit file are bigger than those in a 16-bit file but it's not "louder" because everything is automatically scaled up or down to match your DAC at playback time.   The increased range comes at the quiet-end.

32-bit floating point has a range of more than -700 to +700dB so for all practical purposes it has no upper or lower limits.       This is ONLY in the digital domain....   There are limits in the physical (electrical and acoustic) world.  In floating point a numerical value of 1.0 is 0dBFS.

DAWs and audio editors normally use floating point internally.   It makes DSP (digital signal processing )"easier" and if you do something that pushes the peaks over 0dB (boosting the bass or mixing, etc.) the audio isn't damaged (yet) and you can simply lower the level before rendering the output file.

Although 32-bit audio can go over 0dB, it will clip your DAC when played at "full digital volume".    ADCs, DACs, and audio CDs are all integer based and limited to 0dB.*  (MP3 can  go over 0dB but I'm not sure what the limit is.)    You shouldn't publish or release a track that goes over 0dB.




* Zoom has some 32-bit floating point recorders & ADCs and there may be some floating-point DACs.   But they still have voltage & noise limits and you can't count-on the playback system having this capability...    0dB should still be considered the digital maximum. 

Re: FLAC 32 bit float

Reply #5
Thank you for that, I will check out wavpack also...

I heard somewhere that the audio quality for 24 vs 32 is same?, the extra 8 bits is to allow increase of gain when editing without distortion??

https://www.makeuseof.com/16-bit-vs-24-bit-audio/

https://www.wired.com/story/32-bit-float-audio-explained/
as I know 32-bit audio can be processed with a large margin before distortion occurs. However, if you are just listening to music, 24 bit will be enough.

Re: FLAC 32 bit float

Reply #6
I was comparing 24-bit to 32-bit float.
24-bit seems fine but 32-bit float seems to have less bass.

I'm probably the only one that hears that.
I left my golden ears back in my 20s.
EZ CD Audio Converter

Re: FLAC 32 bit float

Reply #7
Windows audio engine uses 32-bit float data format in shared mode, so apps not using ASIO or WASAPI exclusive mode always use 32-bit float when the audio data is somewhere between the output of the app (e.g. web browser) and the audio hardware.

Also, don't forget about CAudioLimiter as well.
https://hydrogenaud.io/index.php/topic,104051.0.html

...and this:
https://archimago.blogspot.com/2019/06/guest-post-why-we-should-use-software.html

Windows started to use 32-bit float audio engine since Vista, it is nothing new at all.

Re: FLAC 32 bit float

Reply #8
I was comparing 24-bit to 32-bit float.
24-bit seems fine but 32-bit float seems to have less bass.

Listening test log, please. Terms of service item #8.

For your information, the 32-bit float format can losslessly contain 24-bit integer signals.

Re: FLAC 32 bit float

Reply #9
X
24bit wav


X
32bit float wav

The 32bit has more headroom.
EZ CD Audio Converter

Re: FLAC 32 bit float

Reply #10
The 32bit has more headroom.
The format is, but it has nothing to do with the stats in the two screenshots because everything between the two files remains the same, apart from file format.

Re: FLAC 32 bit float

Reply #11
[attach type=thumb]27508[/attach]
24bit wav


[attach type=thumb]27510[/attach]
32bit float wav

The 32bit has more headroom.

1: Those would not be acceptable per TOS#8, even if they reported different numbers.
2: Which they don't.
3: And if they were to, it would be because you botched up the conversion.

User error, and you still don't know where you went wrong.