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: Is it bad to convert 32 bit to [24 bit] flac? (Read 9239 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Is it bad to convert 32 bit to [24 bit] flac?

I've ran into two problems today for the first time in a row. I think it's because I've recently installed the SoX resampler component into foobar, because I read that it was better.

First of all I was trying to rip an SACD I have to .flac. I have it as an .iso image, and got the proper component to read it in foobar, thinking that I would just convert it to 24-88 .flac and be done with it. Turns out for the first time, I got a "Trancode warning [sic]", telling me that it will not be lossless. I never knew that DSD and 32 bit PCM were not supported by .flac. I mention the 32 bit PCM because right after that I put the .iso aside and tried converting these different WavPack files (32 bit) to 24/96 .flac for better storage. This one perplexed me because I've done this before and never got a transcode warning. It must have been because of the SoX resampler component. Learn something new every day, I guess.

So my main concern is this: Would converting DSD or 32 bit audio to 24 bit 96000 hz FLAC be any better or worse than if I were converting from a format that 'fits' the FLAC codec?

Let's say for example I had the exact same album, only ripped to 24/192 WAV. Would the resulting 24/96 .flac files be any different than if I were to convert from 32/384000 WavPack to 24/96 FLAC? Same exact album, ripped differently.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #1
So my main concern is this: Would converting DSD or 32 bit audio to 24 bit 96000 hz FLAC be any better or worse than if I were converting from a format that 'fits' the FLAC codec?
That depends on what you're trying to accomplish.

Do you want to losslessly compress the original SACD audio? If so, you can't use FLAC because FLAC only supports PCM, and converting DSD to PCM is not lossless. WavPack supports lossless DSD compression (but I'm pretty sure it's not lossless if you use foobar2000 to convert it).

Do you want to save space by resampling the audio to a lesser format? If so, I'd recommend 16-bit, 44.1 or 48 kHz. Anything more than that is wasting space, since you won't be able to hear the additional bits or the higher frequency.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #2
That depends on what you're trying to accomplish.

Well I'm wondering if, for example, converting to 16-44 flac would be the same as the non-SACD version of said album. Let's say you have two CD's, one is an SACD and another is a regular pressed CD, same edition because they do that sometimes. Is converting the SACD to 16-44 redbook going to be the same thing as ripping the regular CD? And if so, my further question would be would ripping the SACD to 24-88.2 FLAC in foobar be any higher fidelity than the 16-44 redbook CD?

Do you want to losslessly compress the original SACD audio?

Not necessarily, but I'm also curious to know how I would do that. I would be interested in getting .dsf files out of it instead of a single file, .iso disc image. I tried converting to wavpack with foobar but like you said, I still got the message about DSD passthrough. I also downloaded a program called ISO2DSD and it won't recognize the file for some reason, I can't get it to convert my .iso file to multiple .dsf files.

I know it's hard to explain, but I'm really just trying to figure out if any additional kind of transcoding is going on here. I know that 16/44 is basically transparent, but I've gotten into the habit of storing 24 bit files because space is cheap anyway and they're relatively popular with artists these days.

I'll try to give another example for the sake of brevity. Let's say, for example, hdtracks.com is selling 24/88.2 FLAC files for an album. Let's say I also own the physical SACD release. If I bought those files from hdtracks, would they be ANY different than the files that would come out of converting my .iso image file to 24-88.2 FLAC using SoX resampler on foobar?

And additionally, I would have the same question about converting from a 32 bit DSD master.

Do you want to save space by resampling the audio to a lesser format?

I personally think DSD is overkill, at an absolute maximum I store 24/96 files, so yes I don't mind losing some of that data, I'm just worried about 'dithering', I think it's called.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #3
Do you want to save space by resampling the audio to a lesser format? If so, I'd recommend 16-bit, 44.1 or 48 kHz. Anything more than that is wasting space, since you won't be able to hear the additional bits or the higher frequency.
even this can be considered wasting space, because constant bit depth is not optimal if the sound has varied loudness levels (almost always true, except perhaps some music genres called "noise"...). I mean, if 16 bits are enough to achieve good enough signal to noise ratio in quieter parts of the recording, then they are obviously more than enough for louder parts.

the good thing about constant bit depth is that it's simple&standard and there's no loss from conversion from one format to other, and no loss from splitting and joining files at any points.

to do better than this, I think you'd want at least a floating-point format (with 16 bits or less; and that's currently super exotic) or some kind of lossy compression.

(but then when a new better lossy compression method is developed eventually, there will be no upgrade path if you don't have originals.)

so... I guess there's just no ultimate answer, to each their own.
a fan of AutoEq + Meier Crossfeed

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #4
Well I'm wondering if, for example, converting to 16-44 flac would be the same as the non-SACD version of said album. Let's say you have two CD's, one is an SACD and another is a regular pressed CD, same edition because they do that sometimes. Is converting the SACD to 16-44 redbook going to be the same thing as ripping the regular CD? And if so, my further question would be would ripping the SACD to 24-88.2 FLAC in foobar be any higher fidelity than the 16-44 redbook CD?
...
I'll try to give another example for the sake of brevity. Let's say, for example, hdtracks.com is selling 24/88.2 FLAC files for an album. Let's say I also own the physical SACD release. If I bought those files from hdtracks, would they be ANY different than the files that would come out of converting my .iso image file to 24-88.2 FLAC using SoX resampler on foobar?
they will almost certainly be different, because
1. often these are different tracks to begin with (different mastering)
2. there are a lot of variables in resampling&bit depth reduction, even if differences are inaudible. there's almost no chance that you can guess the exact settings which they used to convert.
but that's not something to worry about IMO.

I personally think DSD is overkill, at an absolute maximum I store 24/96 files, so yes I don't mind losing some of that data, I'm just worried about 'dithering', I think it's called.
Which kind of dithering do you mean here exactly?
a fan of AutoEq + Meier Crossfeed

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #5
Well I'm wondering if, for example, converting to 16-44 flac would be the same as the non-SACD version of said album. Let's say you have two CD's, one is an SACD and another is a regular pressed CD, same edition because they do that sometimes. Is converting the SACD to 16-44 redbook going to be the same thing as ripping the regular CD?

Remember different CD pressings could have different authoring, and without knowing exactly how each pressing was authored, you'd just have to assume. For example the may or may not opt to change the volume or loudness compression.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #6
Ahh, what am I trying to ask here...

Which kind of dithering do you mean here exactly?

I don't know, I'm not really well versed in audio production. I just mean like, is it the same reason that audiophiles don't like MQA? It's hard for me to tell because my converted files look fine in spek (spectogram program). It's not like spotting a transcode where there's a clear cutoff.

For (another) example, I know that if you were to convert a 1411 kbps PCM > 320 kbps MP3 > 192 kbps MP3 then that would actually be WORSE than just doing 1411 kbps PCM > 192 kbps MP3 in the first place, i.e. you can only transcode to a lossy source once. I'm not 100% sure on why this is, but I think it's because you lower the "resolution" of the spectral waves each time. I think that if you were to zoom in on the spectrals of the former and the latter then the former would look more pixelated, right? I don't know how to spot that on my own, I'm not sure of a program that would do that or if it's even a viable option. Maybe Adobe Audition will?

Maybe you guys are right, the mastering and settings could be different so the SACD or 32 bit wavpack files wouldn't be the same had they been mixed with the intent of distributing in PCM. I will revise my question again; I want to know if in relation to the example in my previous post, would the 24-88.2 khz FLAC sourced from the SACD be any lower FIDELITY than the FLAC that was released on hdtracks? Yes, mastering may be different. I apologize for not being clear enough, I'm mainly concerned with the fidelity here.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #7
DSD to PCM can never be "bit perfect".

But as long as you stay at "CD quality" or better, almost anything you do (including multiple conversions) should be transparent unless something goes terribly wrong.   Not "basically transparent", actually transparent.

Quote
I'm just worried about 'dithering',
Theoretically, you should dither whenever downsampling.  But at 16-bits or better you can't hear dither or the effects of under any normal listening conditions.    At 8-bits you can hear quantization noise and the dither noise is supposed to be less objectionable.   (I have heard 8-bit quantization noise but I've never tried 8-bit dither.)

So my advice is to do whatever conversion(s) you have to do to get the format you want, and try not to worry about it!

Quote
Is it bad to convert 32 bit to [24 bit] flac?
32-bit floating-point can go over 0dB and if it does it will clip when you convert to FLAC.  

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #8
Theoretically, you should dither whenever downsampling.  But at 16-bits or better you can't hear dither or the effects of under any normal listening conditions.    At 8-bits you can hear quantization noise and the dither noise is supposed to be less objectionable.   (I have heard 8-bit quantization noise but I've never tried 8-bit dither.)

So my advice is to do whatever conversion(s) you have to do to get the format you want, and try not to worry about it!
Ah, well that is reassuring. Now I'm curious as to why you should dither when downsampling; I didn't know that either. But as far as transparency goes, that's what's important to me, as long as it's not "barely transparent" and well over the realms of human hearing range like you said.
32-bit floating-point can go over 0dB and if it does it will clip when you convert to FLAC.  
This, on the other hand is more worrisome to me. I did not know that either but now I have a better understanding. I would assume that most people wouldn't mix their masters so that they exceed 0db, but then again maybe their whole reason for using 32 bit was so that they could. Maybe I'll have to keep those files as they are. I could have sworn ive seen 32-192 bit flac before. It must have been wavpack and I'm misremembering. What would be the best way to store 32 bit floating point files in lossless and with the most compression? Doesn't seem like there's that many options.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #9
I don't really care much for 24bit audio unless I intend to work with something in an audio editor at some point (at which point, why not just keep the 32bit integrer/float?), but, jumping on the issue as stated on the topic's title, would dithering be irrelevant in an scenario of downsampling down to 24 for listening? Is the quantization noise expected to be audible at all in this case? I'm guessing no, but I have to ask.
Surely clipping is a different matter entirely here, I'm not asking about that. Just the noise/dither issue.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #10
would the 24-88.2 khz FLAC sourced from the SACD be any lower FIDELITY than the FLAC that was released on hdtracks?

Assuming the source is 24/88 PCM audio, the answer is yes imho.
SACD=DSD64
Inherent to DSD is a tremendous dynamic range, a tremendous freqency range and this comes at the expense of a tremendous amount of quantisation noise. In case of DSD64 right after 20 kHz.
Assuming there is musical live in this recording above 20 kHz, the PCM can reproduces it, the DSD will swamp it in the quantisation noise.



If you are converting DSD to PCM don't forget to filter out this noise.
TheWellTemperedComputer.com

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #11
If you are converting DSD to PCM don't forget to filter out this noise.
Oh man, well I'm out of luck, at least for a good while then. I'm at the mercy of what foobar decides to do, I can't really do that myself at this point. You think if I check "use dithering" it will filter out that noise? Or does dithering add noise? I guess this will be good motivation for me to finally stop procrastinating and learn more about music production. I'm just upset that any SACD rips and 32 bit floating point vinyl rips I've accumulated in the past are compromised. So many times I've had to 'redo' my collection, whether it be from mp3>flac or flac>flac with .logs and .cue.. It's really time consuming, but it's kind of worth it to me because I have decent speakers.

I'm glad foobar decided to tell me that the conversion wouldn't be lossless, I think as a result of foobar updating and not the SoX component like I said earlier in this thread. I would have never known had that warning not come up. I'm going to keep my 24/88 FLAC's for now, but the 32 bit-sourced ones I'll have to listen to and see if there's any clipping. I will be keeping backups of all the original files for now, until I decide if there's a better way to store/compress.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #12
Quote
32 bit floating point vinyl rips I've accumulated in the past are compromised.

Are they?
A common practice is to record in 32 or even 64 bit.
This of course has noting to do with dynamic range as 16 bits gives us 96 dB and 24 bit 144 dB dynamic range (vinyl is aprox 50 dB) and as our gear can probably resolve 20/21 bits max at playback time, 32 bit recording looks silly.
But most of the time there is a post-production stage. You mix or  in case of vinyl, you remove pops and clicks. Whatever you do, DSP is calculating and you want to keep the quantization error down. You can’t do that with 16 or 24 bit integers.
So using 32 bit float is a good practice.

If you are going to publish, most of the time this will be Redbook as this is by far the most popular format.
Redbook (CD) is 2 channel, 16 bit , 44.1 kHz.
If you truncate you must dither (indeed adding noise) to decorrelate the quantization error from the audio.
Observe that this is at -96 dBFS so although soft,  it still might be audible.
Anyway, everybody does.
If you export to 24 bit, the LSB is at – 144 dBFS.
I don’t think there is a playback chain able to resolve -144. Maybe a whisper quit DAC can but no poweramp is quiet enough to reproduce this.
It is probably nonsense to dither but it won't do harm either.

The quantization noise of DSD has nothing to do with dither.
It is simply a bunch of powerfull high frequencies.
If the software you use don’t allow you to choose a filter, just limit the sample rate.
Convert to 44 kHz will remove anything above 44/2 = 22 kHz. likewise 48/2=24, etc.
TheWellTemperedComputer.com

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #13
Roseval
I used to convert everything to 16-44, but now I save certain web downloads and especially vinyl rips in 24/96 because
a) on a surface level it quickly ensures me that the files are definitely not just sourced from a CD and
b) If I ever decide to go into some sort of DAW and edit them it would be good for mastering.

I also read on here somewhere that 24 bits allows more volume points (dynamic range? i don't know the term).
I do have a Behringer UMC 404HD 192k and a pair of Klipsch R-15M's so DSD would be pointless, but theoretically it should be able to emit frequencies up to 192 kHz, I think the klipsch speakers only go up to 96 kHz though (still well past transparency).

If you truncate you must dither (indeed adding noise) to decorrelate the quantization error from the audio.
Yeah, I think you can kind of see the noise left over from the SACD source (I didn't use dithering here because I didn't know what it meant):

And here is the 24/96 FLAC sourced from the 32/384200hz WavPack vinyl rip (no dithering):

Surprisingly this one looks better. But maybe there are differences that can't be seen. I still have to listen to it to hear/see if there's any clipping.

So I'm still ambivalent about this. Maybe my files are fine. Should I go back and do it again, but add dithering this time? I think I should.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #14
And here is the 24/96 FLAC sourced from the 32/384200hz WavPack vinyl rip (no dithering):
Ahh, my bad. sourced from 32 bit 384000 hz WavPack.

Edit: And it looks like I used more improper terminology. The vinyl rip mentioned above isn't floating point. Foobar calls it "fixed point". I just tried converting a 32/192 "floating point" wavpack file and this time it didn't give me a warning so I can only assume that the 24/96 flac output was indeed lossless.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #15


This chart can only be done by a marketing office. Download a song in DSD, FLAC or whatever you like from 2L and try to reproduce this result. Pure science fiction!

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #16
@celona You can't get this kind of graphic with a "song". This is about SNR and frequency resolution.

One cannot negate the fact that DSD has an increasing noise shaped dither which is inherent to the format.
Do you have a silent DSD record?
Do you have a DSD decoder that does not lowpass filter the output?
Do you have a frequency analyzer that goes up to 1.4Mhz (or at least to 192KHZ)?


Since I don't think that DSD has any kind of "Dolby-NR" system integrated, the lowpass filtering lowers the peak for higher frequencies.

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #17
This chart can only be done by a marketing office. Download a song in DSD, FLAC or whatever you like from 2L and try to reproduce this result. Pure science fiction!
TheWellTemperedComputer.com

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #18
Code: [Select]
Download a song in DSD, FLAC or whatever you like from 2L and try to reproduce this result.

I say, do!
Download a DSD64 (FLAC makes no sense as it is PCM) and convert it to FLAC WITHOUT using any filter.
You get something like this


What you probably missed is that this quantisation noise is not only substantial but also above the upper threshold of our hearing.
Hence  DSD64 sound good as you won't hear the noise.

BTW: the marketing department is Malcolm Hawksford, Essex University, UK
If you google a bit on DSD versus PCM you will find tons of articles all showing the same artifacts generated by DSD.

Talking about marketing: the CD was a tremendous success.
It generated substantial revenues for its patent holders Sony and Philips.
What about if you can repeat this?
Of course you cannot patent the same thing twice so there must be something new and also better.
The new was DSD and "better" was the marketing part.
As DSD wasn't an improvement compared with PCM, the SACD failed.
But up to today, some still believe in this marketing talk.
TheWellTemperedComputer.com

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #19
@Roseval The decreasing frequency content above 60Khz makes me think that the decoder on MusicScope has a gentle lowpass filter.

I didn't know that software. Went to the site and seems they are closing.
There's a newer version than the one that you have ( 2.0.7 although the link says 2.0.5). Requires a license and has a 30-seconds-limit demo mode so I could try it.

Showed a similar behavior but now the filter is more pronounced and is at -144dBFS already at 84Khz (I tested 2L-125_stereo-2822k-1b_04.dsf )

I agree with the rest of your comments.

 

Re: Is it bad to convert 32 bit to [24 bit] flac?

Reply #20
Unfortunately the company indeed cheased operation: https://hydrogenaud.io/index.php?topic=118744.0
Didn't realize it was a filter in MusicScope.
Thanks for taking the trouble to test.



TheWellTemperedComputer.com