Years ago, I had the pleasure of working with some serious digital to analog converters. I could experiment with the conversion, varying the dither mode, as well as some very cool options in signal processing.
The noise floor makes a significant difference. As GB mentioned, the mastering makes a huge difference as well.
If you’re hearing the noise floor on 16 bit audio, you’re probably not doing the comparison right. It sounds to me that you took poorly leveled audio, cranked up the volume, and compared the noise floors?
If so, yeah that’ll make noise shaping sound really good But thats because its a bad way to test.
fuze_owner-GB wrote:
I almost always am working with pre-1950 material; then run it through usually numerous noise removal processes to get a pleasing sound, post-restoration. I work in the 24 bit domain because the filtering algorithms work best with this format.
I think you’re misunderstanding whats happening. Audio that old probably is about 50-60dB. 16 bit audio is ~96dB. You’re no where near being limited by quantization error. Nothing you hear in it has anything at all to do with quantization error. Nothing being discussed here applies to such old analog recordings.
Additionally, if you’re doing noise removal, you’re not working with 24 bit audio. You’re working with floating point audio. At the end you may export to a 24 bit PCM or FLAC file, but its definitely not being processed at 24 bit and the bit depth you export to has nothing to do with the filters used.
fuze_owner-GB wrote:
Consistently I can hear artifacts when sampling down to red book standard audio format if no dithering is used. So, maybe it’s a case where my sound restoration software is playing a role.
Yeah sounds like you have lousy software or are using it incorrectly. It would help if you could explain in more detail what it is you hear. You may also want to read up a little more on the exact tools you’re using. It sounds like theres more going on.
fuze_owner-GB wrote:
If I had a pristine original in 24 bit; maybe dithering might not be necessary. But due to the nature of my work, I almost get no opportunity to work with pristine files.
No its the opposite. If you had 24 bit recordings, dithering would do something useful. But if your recordings are almost all analog noise, adding 0.001% more isn’t helping. Theres already a lot there. Dithering helps with very, very high SNR recordings where the noise from the actual quantization error dominates. For everything else its not helping since you never hear the quantization error.
Like I said, I’m not an engineer. I may not understand correctly what is happening to my files; all I know is that, in the end, I am getting outstanding results with the finished product. And at the end of the day, that is what matters.
I’ll defer any technical expertise to you on this subject. Sorry for muddying the water; I was just trying to help.
The noise floor is dependent upon many factors. More resolution in the form of sampling gives you the dynamic range to keep your signal well above the background.
With the Soundstream digital recorder, I had plenty of “space”. I liked the dbx 700, as it didn’t use pulse code modulation (PCM), but single-bit sigma delta modulation and a sampling rate above 600kHz! The noise floor of the console upstream of the two-track 700 was all you could hear.
To keep my LP recordings nice and quiet, I use the dbx 4bx after the preamp, so my final digital on the Sansa is quite nice. I use the dbx to push the audible noise threshold to a pleasant level. Compared against the “volume wars” editions of the classics, I have greater dynamic range using LP than the “modern” CD, and no clipping or nasty compression artifacts.
If you’re hearing the noise floor on 16 bit audio, you’re probably not doing the comparison right. It sounds to me that you took poorly leveled audio, cranked up the volume, and compared the noise floors?
For my personal real world calibration, “Monochrome” by Kodo starts off peaking at -77 dB and crescendos over about 4 minutes to full scale. For the beginning part, if volume is set to be comfortable at full scale, the level is comparable to the mechanical noise of my HD based DAP a couple of feet from my head (at first I thought it was some weird artifact from Vorbis). A fridge or PC vent fan running in the next room would wipe it out. It would be buried in the noise if playing from an LP. Even that quiet little tick tick tick on the drum is still 18 dB above the 16 bit noise floor.
Moral of the story is its tough to hear a 16 bit noise floor with real world noise around.
Im one of those guys who has ripped all my 1500+ cd collection in FLAC 24 bit for my proud hi-fi system (using the cds as well :))
You can’t rip a CD to 24 bit FLAC. A CD is only 16 bit, so a FLAC of it is 16 bit too.
After reading this thread I ripped a CD to 16 + 24bit FLAC and the clip would only play the 16 bit files. I dont want to convert all to 16 bit .
So my question is: Does anyone know if there will be a firmware upgrade allowing the clip to play FLAC 24 bit?
I wouldn’t count on it. And anyway, it sounds like theres just a bug in the decoder if you’ve enabled 24 bit input, so you could losslessly convert those fake ‘24 bit’ FLACs you have to 16 bit FLACs without loosing quality. It’ll take a while, but at least you can start it running and come back when its finished. Probably a good idea even if sandisk does fix the bug.
Garrett wrote:
But I would like NOT to convert all my files to 16 bit, so that me and my wife can enjoy our music anywhere, without needing to convert or have
2 different formats. Especially when it comes to those SACD’s I own :(.
Message Edited by Garrett on 12-31-2009 01:28 AM
You may be interested to know that its not actually possible to rip SACDs with a PC, and even if you could, since they do not contain PCM audio, you would not be able to losslessly listen to them using FLAC (as FLAC compresses PCM audio). If you have lossless “SACD” rips, you’ve probably been duped, as such a thing is not possible with FLAC.
You can’t rip a CD to 24 bit FLAC. A CD is only 16 bit, so a FLAC of it is 16 bit too.
This had me a little confused. I know that CD’s are in 16 bit, however I own a 24 bit HQ soundcard (Auzentech) and therefore used the 24 bit setting, when ripping with fb2k (EAC to slow for me).
To all of my knowledge this would give me a lot better FP values when converting and that’s the reason that I enabled it, before I ripped my collection.
(This would also mean i would lose quality when converting back to 16 bit)
My Squeeze system plays the FLAC files without any problems and so does fb2k, Songbird, MediaMonkey on regular pc’s.
Maybe the Sansa Clip only supports an older version of FLAC, as I believe there have been some issued with 24 bit in previous versions of FLAC or their decoder has bug as you suggest?
This leaves me with only 2 options I can think of:
Wait for a firmware update.
Find another portable player. (They really should have written FLAC supports 16 bit only in their documentation)
Please enlighten me, if I am misunderstanding something here.
I know that CD’s are in 16 bit, however I own a 24 bit HQ soundcard (Auzentech) and therefore used the 24 bit setting, when ripping with fb2k (EAC to slow for me).
To all of my knowledge this would give me a lot better FP values when converting and that’s the reason that I enabled it, before I ripped my collection.
FLAC is a lossless codec. There are no FP values at all. It simply stores exactly what is on the CD. In this case, that would be 16 bit integers (not floats). If setting it to 24 bit mode actually changed the audio in some way, then FLAC wouldn’t be lossless.
Please enlighten me, if I am misunderstanding something here.
You are. You don’t lose quality by recompressing a lossless file losselessly. Thats why its called lossless. There is no loss. Just go back and fix your files and you won’t have to worry about this.
You can use Goldwave v5.55 to batch convert all your 24bit 96KHz flac files into 16bit 41Khz Flac files that will play on the Sandisk Fuze and even on a two year old Sandisk Clip that has been updated with the latest Clip software.
Mp3tag V2.45a, which is freely available, also shows that Goldwave will convert all your tags too. That means that all of your embedded tag info, including album covers etc., gets ported over to the new file.
I know that CD’s are in 16 bit, however I own a 24 bit HQ soundcard (Auzentech) and therefore used the 24 bit setting, when ripping with fb2k (EAC to slow for me).
To all of my knowledge this would give me a lot better FP values when converting and that’s the reason that I enabled it, before I ripped my collection.
FLAC is a lossless codec. There are no FP values at all. It simply stores exactly what is on the CD. In this case, that would be 16 bit integers (not floats). If setting it to 24 bit mode actually changed the audio in some way, then FLAC wouldn’t be lossless.
Since he mentioned using a 24 bit setting on the soundcard, it sounds like rather than a digitial rip he played the audio in through his sound card and digitized that, much as you would ripping an LP. In that situation, the extra 8 bits gives you some wiggle room in setting the level.
If he did do a digital rip and is converting to 24 bit flac just because of a 24 bit output device, then you’re right.
I know that CD’s are in 16 bit, however I own a 24 bit HQ soundcard (Auzentech) and therefore used the 24 bit setting, when ripping with fb2k (EAC to slow for me).
To all of my knowledge this would give me a lot better FP values when converting and that’s the reason that I enabled it, before I ripped my collection.
FLAC is a lossless codec. There are no FP values at all. It simply stores exactly what is on the CD. In this case, that would be 16 bit integers (not floats). If setting it to 24 bit mode actually changed the audio in some way, then FLAC wouldn’t be lossless.
Since he mentioned using a 24 bit setting on the soundcard, it sounds like rather than a digitial rip he played the audio in through his sound card and digitized that, much as you would ripping an LP. In that situation, the extra 8 bits gives you some wiggle room in setting the level.
If he did do a digital rip and is converting to 24 bit flac just because of a 24 bit output device, then you’re right.
Message Edited by donp on 01-04-2010 07:45 AM
He did the second using foobar as a ripper. He mentioned his sound card settings because he wasn’t clear on how ripping CDs worked.
That’s correct, I was not clear how the ripping with foobar did work. But now I have converted it all to 16 bit FLAC and cannot see or hear any difference (took about 24 hours).
Also I took the liberty of posting in the Hydrogenaudio forum also just to be sure about converting from 24 bit to 16 bit FLAC, if anyone is interested.
But thats because its a bad way to test.
.
