Should tapes/vinyl be recorded to 24bit wav?

Topic: Should tapes/vinyl be recorded to 24bit wav? (Read 4488 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

Should tapes/vinyl be recorded to 24bit wav?

2003-05-02 12:59:19

Hi,

I'm probably going to encode some tapes & vinyl to MPC soon but in the process I will probably be doing quite a bit of noise reduction, scratch removal etc.

As each operation will reduce the accuracy of the WAV file and possibly cause audible noise is it better for me to record to a 24bit WAV? Is there any advantage to using a higher sample rate as well?

If so what would then be the best method to convert the WAV to 16/44?

Cheers,
-dave

Should tapes/vinyl be recorded to 24bit wav?

Reply #1 – 2003-05-02 13:04:33

The guys at the cool edit forums reccommend recording 24/96 if you can, especially if youa re going to do any clean-up processing.

The final output is then dithered down. I use cool edit for this.

There is a small gui for ssrc as well that will do the dithering for you.

Should tapes/vinyl be recorded to 24bit wav?

Reply #2 – 2003-05-02 13:19:13

Quote

The guys at the cool edit forums reccommend recording 24/96 if you can, especially if youa re going to do any clean-up processing.

The final output is then dithered down. I use cool edit for this.

There is a small gui for ssrc as well that will do the dithering for you.

> There is a small gui for ssrc as well that will do the dithering for you.

Am I right in thinking that SSRC gives the best quality?

Whats the file size for a 24/96 wav compared to 16/44 wav?

Cheers,
-dave

Should tapes/vinyl be recorded to 24bit wav?

Reply #3 – 2003-05-02 13:32:42

Quote

Whats the file size for a 24/96 wav compared to 16/44 wav?

24/96 is about 3.3 times larger than 16/44.

Should tapes/vinyl be recorded to 24bit wav?

Reply #4 – 2003-05-02 13:52:53

The theory:

1. record at the highest sample rate and bitdepth that your soundcard will allow.
2. Process at this sample rate, 32-bit (or 24-bit) processing.
3. Convert to either 44.1kHz (if you think you might burn them onto audio CD) or 48kHz (if you'll only ever listen to them on a recent PC)
4. Encode to mpc.

IIRC You can encode from 24-bit wavs, so no need to dither down to 16-bits.

The practice:

1. With a decent sound card, record at 44.1kHz 16-bit. With a poor soundcard, record at 48kHz 16-bit. Make sure the recording levels are reasonable. If it never gets within 6dB of the top, it's too quiet. If it clips during the music, it's too loud.
2. Perform as much restoration as you want at 16-bit resolution, confident in the knowledge that there's no possibility of the dither noise ever reaching anything like background noise level of the miserable specimen of an LP that you've foolishly decided to rescue
3. Convert it to 44.1kHz if it's not already.
4. Save it at 16-bits, and encode it from this file.

The advantage of the first method is that you'll get that warm fuzzy feeling^tm, knowing you've done the best you can. The advantage of the second method is that it's ten times faster - and LP restoration is slow enough as it is, so this is a good thing^tm.

I've used both methods. You've got to use some common sense: no amount of 32-bit processing is going to help with
1. poor source
2. poor de-noising software
3. poor de-clicking software.
Junk is still junk with 32-bit accuracy. And a bad de-clicker is still a bad de-clicker even if the output has 32-bits of precision.

I would seriously suggest you take 30 seconds of the best condition and widest dynamic range LP you have, and try both methods (identical processing both times). Time both methods. Compare the resulting audio (after you've mpc'd it) in a blind test. Then draw your own conclusions.

Cheers,
David.

Should tapes/vinyl be recorded to 24bit wav?

Reply #5 – 2003-05-02 14:31:25

> The theory:

> 1. record at the highest sample rate and bitdepth that your soundcard will allow.
> 2. Process at this sample rate, 32-bit (or 24-bit) processing.
> 3. Convert to either 44.1kHz (if you think you might burn them onto audio CD) or 48kHz > (if you'll only ever listen to them on a recent PC)

I assume the only reason you are recommending 48khz is to avoid the resampling present in AC97 and if my soundcard does not resample a 44.1khz source then I may as well stick to 44.1khz?

> 4. Encode to mpc.

If I encode a 48khz wav to MPC will the resulting file have a sample rate of 44.1khz or 48khz?

> IIRC You can encode from 24-bit wavs, so no need to dither down to 16-bits.

I think LAME can do this, I don't know about MPC. The thing is though, it's probably going to dither down to 16bit first right? Will it do as good a job as e.g SSRC? If on the other hand I'll end up with a 24bit MPC file (is that even possible?) will this cause me any problems?

> The practice:

> 1. With a decent sound card, record at 44.1kHz 16-bit. With a poor soundcard, record
> at 48kHz 16-bit. Make sure the recording levels are reasonable. If it never gets within
> 6dB of the top, it's too quiet. If it clips during the music, it's too loud.
> 2. Perform as much restoration as you want at 16-bit resolution, confident in the
> knowledge that there's no possibility of the dither noise ever reaching anything like
> background noise level of the miserable specimen of an LP that you've foolishly decided > to rescue

Actually I'm not rescuing LP's. I'm encoding records that I bought recently (and so they are in good condition) so that I can listen to the tracks on my PC. The background noise level shouldn't be particularly high, not compared to 'a miserable specimen of an LP' anyway

> 3. Convert it to 44.1kHz if it's not already.
> 4. Save it at 16-bits, and encode it from this file.

> The advantage of the first method is that you'll get that warm fuzzy feeling^tm
>, knowing you've done the best you can. The advantage of the second method is > that it's ten times faster - and LP restoration is slow enough as it is, so this is a good
> thing^tm.

Why exactly is it 10 times faster, what process takes time. Dithering from 24 to 16 bit?

> I've used both methods. You've got to use some common sense: no amount of 32-bit
> processing is going to help with
> 1. poor source

In the case of my vinyl, they are 1 track a side 12" singles mainly, so the quality is significantly higher than an old LP, also because they are in good condition.

> 2. poor de-noising software
> 3. poor de-clicking software.

I was thinking of using Sonic Foundry software, unless anybody knows of a better solution?

> Junk is still junk with 32-bit accuracy. And a bad de-clicker is still a bad de-clicker even if > the output has 32-bits of precision.

Agreed but my question was based on the assumption that there are no weak links in the chain.

> I would seriously suggest you take 30 seconds of the best condition and widest
> dynamic range LP you have, and try both methods (identical processing both times).
> Time both methods. Compare the resulting audio (after you've mpc'd it) in a blind test. > Then draw your own conclusions.

As much as I agree with ABX'ing I'm not sure it tells the whole story. What I mean is, just because I can't ABX 2 files now, doesn't mean that at some point in future I won't be able to in future when I may be able to notice artifacts more easily either because my hearing has become better trained or I have better equipment.

Anybody else have an opinion on this?

Cheers,
-dave

Should tapes/vinyl be recorded to 24bit wav?

Reply #6 – 2003-05-02 14:57:22

Quote

I assume the only reason you are recommending 48khz is to avoid the resampling present in AC97 and if my soundcard does not resample a 44.1khz source then I may as well stick to 44.1khz?

Probably. The kind of people who like 24-bit audio also tend to like 48kHz sampling better than 44.1kHz sampling. I'd stick with 44.1kHz for convenience.

Quote

> 4. Encode to mpc.

If I encode a 48khz wav to MPC will the resulting file have a sample rate of 44.1khz or 48khz?

48kHz.

Quote

> IIRC You can encode from 24-bit wavs, so no need to dither down to 16-bits.

I think LAME can do this, I don't know about MPC. The thing is though, it's probably going to dither down to 16bit first right? Will it do as good a job as e.g SSRC? If on the other hand I'll end up with a 24bit MPC file (is that even possible?) will this cause me any problems?

If it accepts 24-bit input, it will not dither it to 16-bits. Lossy codecs do not have a "native resolution" like this. The resulting mpc or mp3 will be a standard file - it's neither 16-bit, nor 24-bits. Lossy codecs don't work like this. It doesn't have a conventional bitdepth until it's decoded - and then it's the choice of the decoder - it has nothing to do with the source bitdepth.

In simple terms, lossy codecs maintain the sample rate (unless the bitrate is so low that it would be better to drop it), but lose the bitdepth.

Quote

> The practice:

> 1. With a decent sound card, record at 44.1kHz 16-bit. With a poor soundcard, record
> at 48kHz 16-bit. Make sure the recording levels are reasonable. If it never gets within
> 6dB of the top, it's too quiet. If it clips during the music, it's too loud.
> 2. Perform as much restoration as you want at 16-bit resolution, confident in the
> knowledge that there's no possibility of the dither noise ever reaching anything like
> background noise level of the miserable specimen of an LP that you've foolishly decided > to rescue

Actually I'm not rescuing LP's. I'm encoding records that I bought recently (and so they are in good condition) so that I can listen to the tracks on my PC. The background noise level shouldn't be particularly high, not compared to 'a miserable specimen of an LP' anyway

I've just tried in Cool Edit: 10 operations will add about 10dB of noise (using a 16-bit file, starting with correctly dithered silence).

Let's say you start with -90dB. You'll end up at -80dB. You would be lucky to find an LP with this little noise to start with. If you start with -80dB, you'll end up with about -79dB (that's not a typo), and if you start with -70dB, you'll end up with -70dB, because all the additional noise is two orders of magnitude smaller than the noise you started off with, and so will have no effect.

Quote

> 3. Convert it to 44.1kHz if it's not already.
> 4. Save it at 16-bits, and encode it from this file.

> The advantage of the first method is that you'll get that warm fuzzy feeling^tm
>, knowing you've done the best you can. The advantage of the second method is > that it's ten times faster - and LP restoration is slow enough as it is, so this is a good
> thing^tm.

Why exactly is it 10 times faster, what process takes time. Dithering from 24 to 16 bit?

You're processing over 3 times as much data in every operation. In my experience, the time doesn't scale linearly with the amount of data, but YMMV.

Quote

> I've used both methods. You've got to use some common sense: no amount of 32-bit
> processing is going to help with
> 1. poor source

In the case of my vinyl, they are 1 track a side 12" singles mainly, so the quality is significantly higher than an old LP, also because they are in good condition.

> 2. poor de-noising software
> 3. poor de-clicking software.

I was thinking of using Sonic Foundry software, unless anybody knows of a better solution?

That's the best solution I've found, if you mean the SF NR-2 declicker. TBH, if the records are that good, I wouldn't bother. You can easily do much more harm than good.

Quote

> I would seriously suggest you take 30 seconds of the best condition and widest
> dynamic range LP you have, and try both methods (identical processing both times).
> Time both methods. Compare the resulting audio (after you've mpc'd it) in a blind test. > Then draw your own conclusions.

As much as I agree with ABX'ing I'm not sure it tells the whole story. What I mean is, just because I can't ABX 2 files now, doesn't mean that at some point in future I won't be able to in future when I may be able to notice artifacts more easily either because my hearing has become better trained or I have better equipment.

Well, don't declick then, if you're worried you might impose damage that you might regret later. The artefacts introduced by the best software declicker are usually stratospherically larger than those introduced by 16-bit processing.

Quote

Anybody else have an opinion on this?

Why do I suspect that you don't trust my advice here?

Cheers,
David.

Should tapes/vinyl be recorded to 24bit wav?

Reply #7 – 2003-05-02 15:55:35

I'm curious why after you're going through all the work of getting your tapes & vinyl into digital that you want to encode into mpc? It is an excellent lossy codec but for the work involved (sampling, cleaning, mastering, resampling, reencoding...) that it would be more worth it to save it in a lossless format (flac, ape, etc) since at least it will avoid one level of deteration.

You might have hard drive limits or reasons to go to mpc but this is my $0.01 ($0.01 - 50% brain not functioning)

Should tapes/vinyl be recorded to 24bit wav?

Reply #8 – 2003-05-02 19:20:06

> Probably. The kind of people who like 24-bit audio also tend to like 48kHz sampling better than 44.1kHz
> sampling. I'd stick with 44.1kHz for convenience.

I always wondered if the people who claim DAT sounds superior to CD were crazy. Unfortunately I've never had the opportunity to listen to DAT to find out myself.

[snip]

> If it accepts 24-bit input, it will not dither it to 16-bits. Lossy codecs do not have a "native resolution" like this.
> The resulting mpc or mp3 will be a standard file - it's neither 16-bit, nor 24-bits. Lossy codecs don't work like
> this. It doesn't have a conventional bitdepth until it's decoded - and then it's the choice of the decoder - it has
> nothing to do with the source bitdepth.

I read another topic here earlier that more or less said the same. Thanks for the explanation.

ever reaching anything like
> background noise level of the miserable specimen of an LP that you've foolishly decided > to rescue

> I've just tried in Cool Edit: 10 operations will add about 10dB of noise (using a 16-bit file, starting with correctly > dithered silence).

> Let's say you start with -90dB. You'll end up at -80dB. You would be lucky to find an LP with this little noise to
> start with. If you start with -80dB, you'll end up with about -79dB (that's not a typo), and if you start with -70dB, > you'll end up with -70dB, because all the additional noise is two orders of magnitude smaller than the noise you > started off with, and so will have no effect.

So what you're telling me really is that 24/96 is overkill in any case?

[snip]

> That's the best solution I've found, if you mean the SF NR-2 declicker. TBH, if the records are that good, I
> wouldn't bother. You can easily do much more harm than good.

I would probably only be de-clicking small sections of the WAV and only when the click is plainly audible. Due to the way my records are used (DJ'ing) some damage here and there is unavoidable. Overall though, the records sound very clean.

> Well, don't declick then, if you're worried you might impose damage that you might regret later. The artefacts
> introduced by the best software declicker are usually stratospherically larger than those introduced by 16-bit
> processing.

I think the only way for me to figure out what to do is to take a 16/44 WAV, do numerous operations on it (e.g Normalize), then reverse those operations to get back to the 'original' file then ABX to see if I can hear the distortions introduced. Does that sound like a good way to do it?

> Why do I suspect that you don't trust my advice here?

Because I don't trust anyones advice? Actually it's not really a matter of trust, I'll go with whatever can be proven I guess.

Cheers,
-dave

Should tapes/vinyl be recorded to 24bit wav?

Reply #9 – 2003-05-02 20:09:46

> I'm curious why after you're going through all the work of getting your tapes & vinyl into digital that you want to > encode into mpc? It is an excellent lossy codec but for the work involved (sampling, cleaning, mastering,
> resampling, reencoding...) that it would be more worth it to save it in a lossless format (flac, ape, etc) since at
> least it will avoid one level of deteration.

You have a point there but I already have lots of MP3 (6000 files) and MPC (1000 files) I'd rather not add another format into the mix, it all gets a bit annoying and confusing with various plugins each with different settings and incompatibilities etc.

> You might have hard drive limits or reasons to go to mpc but this is my $0.01 ($0.01 - 50% brain not functioning)

Well hard drive space is one reason. If I store all of my tapes & vinyl lossless it will end up taking quite a bit of space which is already running a bit low for my liking. Of course I chose MPC because it sounds great while taking relatively little space.

-dave

Should tapes/vinyl be recorded to 24bit wav?

Reply #10 – 2003-05-02 20:58:14

96 khz is probably overkill.
24 bits is reasonable (50% more space during mastering).

Should tapes/vinyl be recorded to 24bit wav?

Reply #11 – 2003-05-03 18:47:58

So a good compromise would be to record 24bit but keep a 44khz sample rate?

I still want to know the best way to dither from 24 -> 16 though.

Anyone?

Cheers,
-dave

Should tapes/vinyl be recorded to 24bit wav?

Reply #12 – 2003-05-04 13:04:03

I think there is a 99% chance that even recording at 16 bit mode will do, because vinyl noise will act as dither.

Anyway, there's no harm in recording at 24 bit and dither to 16 bit at the end. For dithering, just use FB2K diskwriter, set to 16 bit with dither, and "triangular dither" and "triangular" noiseshaping.

Should tapes/vinyl be recorded to 24bit wav?

Reply #13 – 2003-05-04 15:36:25

Is that an agreed HA standard for the "best" way to dither?

Cheers,
-dave

Should tapes/vinyl be recorded to 24bit wav?

Reply #14 – 2003-05-04 21:18:28

Quote

Is that an agreed HA standard for the "best" way to dither?

Cheers,
-dave

Well, I would disagree, but I suspect KikeG had good reasons to suggest this. I'd like to know why.

Should tapes/vinyl be recorded to 24bit wav?

Reply #15 – 2003-05-05 05:53:05

>>>'So a good compromise would be to record 24bit but keep a 44khz sample rate?'<<<

yes, but you can safely apply an anti-aliasing filter when you reduce the sampling rate from, say, 48 to 44.1 kHz

>>>'I still want to know the best way to dither from 24 -> 16 though.'<<<

hmmm... i'm not too sure what you're asking here. you mean what's the preferred dither from 24 to 16 bits?

well, if this is your question, most music CDs today that are mastered by a studio or mastering house are dithered to 16 bits with pow-r, waves IDR, or apogee VU22. pow-r seems to be the favourite of many engineers nowadays (in fact, some record labels insist on it), but it is not available in software form, only hardware. waves IDR and apogee, however, are both available in software form

most of these dither algorithms (except for apogee i think, which does not do noise shaping) have the ability to use different dither noise and noise shaping curves. the general settings are:

dither type: 1 or 2 (1 generally being the best all-around dither noise designed to eliminate quantization distortion, 2 being more quiet type of dither noise but quantization distortion may be more audible). the choice is up to you

noise shaping: moderate, normal, high (aka ultra) (or simply 1, 2, 3 in some systems) - the higher the number/selection, the more noise is moved to higher frequencies of the audio spectrum (> 16 kHz), and away from the ear's most sensitive area. the only problem with that is you can end up with quite a bit of noise in higher frequencies, so using "high/ultra" noise shaping is not always the ideal solution. with most high-end dither algorithms, "normal" noise shaping should do just fine

of course, this is all very subjective stuff. you just have to do listening tests and see what suits you best

ps- i don't know if what i wrote is the HA standard, but it's certainly my standard for high quality results

Should tapes/vinyl be recorded to 24bit wav?

Reply #16 – 2003-05-05 16:38:07

Quote

Well, I would disagree, but I suspect KikeG had good reasons to suggest this. I'd like to know why.

I suggested triangular noise shaping without entering into more details because I think it's the best compromise. ATH shaped dither profiles squeeze more 16-bit dynamic range, but put a lot of noise at high frequencies, and the dithering is more "fragile", so that they are less suited for further post-processing. Triangular shape doesn't suffer as much from these problams, it also doesn't squeeze as much dynamic range of 16-bit audio, but and I think that in practice it's good enough, more if we are talking about vinyl recordings.

Notice