Newbie question: FLAC 24 bit with fb2k, have I made a serious mistake?, Conversion of 24 bit FLAC to 16 bit Options

[Master_Garrett]

Hi all


First off, im not an experienced jetbrain when it comes to audio formats, but nonetheless about a ½ year ago, after a lot of
research, I decided to rip my entire CD collection to FLAC in 24 bit with fb2k, to ensure the best quality possible for my music I could get without
binding me to Windows/Leopard/etc. (I started out by using EAC, but it was to slow a process for my patience)

I have an Auzentech audio card with 24 bit unit, therefore i followed what several guides recommended me to do and used the best resolution my soundcard would support in the fb2k
setup under Output Bit Depth, which was 24 bit.

This works superb with my SqueezeBoxes and my wonderfull DynAudio speaker setup and i have been so happy with the quality over my old mp3/wma's .

Problem:
Yesterday i bought a portable player for my FLAC's, a Sansa Clip+, so that I could take with me when running.
The Sansa Clip+ claimed to support FLAC and is capable of playing FLAC's ripped to 16 bit, but not FLAC's ripped to 24 bit.
I posted in the Sansa forum and got an answer telling me to convert all my 24 bit FLAC's to 16 bit, instead of waiting for them to upgrade the firmware. Also they
claimed that this would NOT affect my FLAC's quality.

Question:
Can I convert my 24bit ripped FLAC's to 16bit without losing ANY quality of my FLAC files?
I know that CD's only are 16 bit, but I thought that setting the Output Bit Depth to 24 bit, matching my soundcard, would give me some degree of better resolution when running the compression algorithms
and encoding/decoding.
I have ripped +1500 CD's to 24 bit FLAC, so it's a big job starting the conversion or restart the ripping/tagging.


Please if anyone could enlighten me I would very much appreciate it, as im scared of messing up my entire collection when converting to 16 bit and having to start all over with the rips :S.
If that is what you guys would recommend?


Kind regards
Christian

[funkyblue]

Why did you rip your CD's in 24-bit without asking if that was a good idea or not?

[Remedial Sound]

Very short answers:

1. Standard (redbook) cd audio is 16 bits, so in retrospect you should have ripped to 16-bit flac.

2. That said, converting 24-bit to 16-bit flac is a virtually lossless process. I'm not certain if it's mathematically lossless, but for all intents and purposes you needn't worry about "messing up" your collection.

[varoot]

Can I convert my 24bit ripped FLAC's to 16bit without losing ANY quality of my FLAC files?

Yes, in your case your FLAC file should contain only 16-bit data padding to 24-bit.

I know that CD's only are 16 bit, but I thought that setting the Output Bit Depth to 24 bit, matching my soundcard, would give me some degree of better resolution when running the compression algorithms
and encoding/decoding.

AFAIK, setting Output Bit Depth should only affects decoding lossy files. Even so, I think its effect would be very subtle.

[DigitalMan]

No loss in quality as stated above. Note that you can just run a batch conversion in Foobar, for example, from 24 to 16 bit depth and it will preserve tags, filename and even folder structure - you do not have to re-rip / re-tag anything.

[saratoga]

2. That said, converting 24-bit to 16-bit flac is a virtually lossless process. I'm not certain if it's mathematically lossless, but for all intents and purposes you needn't worry about "messing up" your collection.

Its lossless. All he did was take his audio, stick 8 bits of zero on front of sample, and then flac it. If he decodes to PCM, removes the zeros, he'll get his original CD audio back.

[_m²_]

One more question, the are 44.1 kHz, right?
Because otherwise there is no way to keep them lossless (and unless they are x*44.1kHz where x is integer, they are lossy already).

[Master_Garrett]

Why did you rip your CD's in 24-bit without asking if that was a good idea or not?

Well I figured that since 24 bit audio is on its way, it couldn't harm to have it all in 24 bit and I did really think that fb2k was taking advantage of
my 24 bit soundcard when ripping the cd (filtering,compressing etc). If it had done that there would be a difference in the FP values, since the round
methods would not have to shorten a value to 16 bit.
But alas I was to optimistic, lacked the knowledge and didn't even consider asking the question on the forum . Can only blame myself.

Regards
Christian

[funkyblue]

OK Cool. Apologies for my abrupt post! I was just curious Only way to learn is to make a mistake. That how I've learnt

Cheers

[Master_Garrett]

No loss in quality as stated above. Note that you can just run a batch conversion in Foobar, for example, from 24 to 16 bit depth and it will preserve tags, filename and even folder structure - you do not have to re-rip / re-tag anything.

Thank you guys ( Remedial Sound, varoot, DigitalMan and Mike Giacomelli)

Your answers are the reassurance I was seeking with my post .
That means all i'll have to do is to select convert and set the bit depth to 16 bit instead of 24, right?
Or do I need to decode to PCM first, before converting to FLAC 16 bit, as Mike Giacomelli suggests?

Kind regards
Christian

[Master_Garrett]

One more question, the are 44.1 kHz, right?
Because otherwise there is no way to keep them lossless (and unless they are x*44.1kHz where x is integer, they are lossy already).

Thankfully they are all in 44.1 kHz. That much I do know and understand , all though you had me worried for a minute .

[probedb]

Or do I need to decode to PCM first, before converting to FLAC 16 bit, as Mike Giacomelli suggests?

Nope, no need. Mike was simply meaning that decoding to PCM and removing the 8 bits of padding you'd have the same data as on your CD

[funkyblue]

Just out of curiosity would the new files be bit-identical or not?

[_m²_]

Just out of curiosity would the new files be bit-identical or not?

They should be bit-identical with what's on the CDs, except for metadata and gaps between tracks.

This post has been edited by _m²_: Jan 2 2010, 12:57

[lvqcl]

Just out of curiosity would the new files be bit-identical or not?

If 8.0 and 8.00 are bit-identical, then yes.

[Porcus]

Well, actually: if you did rip with a HDCD decoder, then the output is up to 20 bits (plus padding), right?

What if the rip is performed with a de-emphasis filter? Probably more than 16 bits also then?

BTW, is there any utility which takes some X bit file and returns "this file is nowhere more than x bits effectively"?

[saratoga]

BTW, is there any utility which takes some X bit file and returns "this file is nowhere more than x bits effectively"?

lossywav can more or less do that.

[saratoga]

That means all i'll have to do is to select convert and set the bit depth to 16 bit instead of 24, right?
Or do I need to decode to PCM first, before converting to FLAC 16 bit, as Mike Giacomelli suggests?

The two are actually the same. If you use foobar's converter, select 16 bit output, and then flac as the codec (without specifying a bit depth), foobar should convert to PCM, drop the 8 fake bits, and then encode to 16 bit flac.

Try it on one file first and see if it works on your clip. If you're really worried, you could also rerip one file from CD and compare to the converted flac (foobar has a tool for checking if files are identical)

[Porcus]

BTW, is there any utility which takes some X bit file and returns "this file is nowhere more than x bits effectively"?

lossywav can more or less do that.

It truncates to a file no more than x bits, but can it analyse and find that a file contains no more than x from the beginning?

(A file may have all zeroes in both the most and least significant bits too, so it is not only a question of padding, but it was the padding the issue did concern for these purposes.)

[saratoga]

BTW, is there any utility which takes some X bit file and returns "this file is nowhere more than x bits effectively"?

lossywav can more or less do that.

It truncates to a file no more than x bits, but can it analyse and find that a file contains no more than x from the beginning?

Yes of course. It would be useless without that. Theres already a million programs that can convert audio between bit depths (e.g. foobar).

(A file may have all zeroes in both the most and least significant bits too, so it is not only a question of padding, but it was the padding the issue did concern for these purposes.)

Thats extremely rare given that most audio these days is peak normalized, but if you're concerned, you can quite easily renormalize your file (or just use replaygain) before using lossywav. Though maybe lossywav has a normalization switch (i haven't checked).

[2Bdecided]

AFAIK, lossyWAV isn't designed to check for trailing zeros - though such a thing is trivial to do.

It's FLAC which sees trailing zeros (from lossyWAV processing, or from a 16>24-bit "conversion") and dumps them en-block in the encoder, re-adding them in the decoder. All the codecs which lossyWAY "works" with do something similar...

http://wiki.hydrogenaudio.org/index.php?ti...c_compatibility

Cheers,
David.

[Nick.C]

One way to check for the distribution of trailing zeroes in lossyWAV is to use the --sampledist and --blockdist parameters which are calculated for the audio data "before" and "after" lossyWAV processing. The former looks for the lowest set bit per sample and the latter looks for the lowest set bit per codec-block (default=512 samples).

[edit] This can be done from the command line for console output or using the --writetolog parameter to store in the logfile. [/edit]

This post has been edited by Nick.C: Jan 4 2010, 19:23

[Master_Garrett]

So I did the conversion to 16 bit a yesterday using fb2k, took a while though , but everything is working smoothly now.
It took just around 24 hours all together, with converting, backup, dbl checking everything etc.

The Sansa Clip now plays my FLAC's and I can neither hear or see any difference when inspecting my files (file size a couple of kb
smaller in general (makes sense) but that's about it).

I'm going todo some more experimenting and studying based on your answers here so I'll be better
prepared the next time i'll be ripping/converting . So I might popup a question again every once in a while.

Thanks for all your help guys, really appreciate it.

FLAC rocks!


Kind regards
Christian

[2Bdecided]

@Nick C - Sorry Nick, I'd missed that - that's useful!

Cheers,
David.

[Kohlrabi]

Master_Garrett: If you're paranoid about the quality you could rip your CDs again to just plain WAV or 16 bit FLAC and bitcompare the tracks to your new files using foobar. Also, I think there are quite a lot people here interested in proper blind test results regarding 16bit and 24bit