Any good reason to record in 24 bits ?

I doubt that recording and mixing in 24 bits will be
in any way better than in 16 bits.

Can any one prove (yes really prove) me wrong ?

/Goran Sweden

Don’t know about ‘prove’ it but this article makes a good enough case to convince me.

I happen to have 24 bit sound cards and version of NTrack so that’s what I use and the sound quality of recorded audio is good. If I was on a budget and had 16 bit gear I wouldn’t lose any sleep over it.

For a long time I remained unconvinced that 24 bit audio would ever be on demand in my modest set-up (I record my own efforts, as well as the group I happen to play in, on a non-profit basis). However, when it was time to renew my hardware set-up, I got a sound card with 24 bit, 48 kHz capability - plus the 24 bit version of n-Track, of course… - and did I note the difference!! :D

If the rest of your set-up - including your ears - are up to the increased demand placed on it by a 24 bit system, by any means, go make the change… - you might never want to go back to 16 bit again. As plumbum states, stay with your 16 bit system if it is adequate for your purposes.

Yes, you can actually hear the difference (no, that’s not a proof…).

regards, Nils

I use it mainly for convenience To get as good sound quality as I would get with 16bit at it’s max I get without having to aim to record at as high level possible.

I well may aim to -6db to -10db or even less so I don’t have to be afraid that the vocalsit screams just a bit louder and nearer than he did in the sound check, or drummer hitting harder or lead guitarist to turn his rig just a bit louder during the heat of the moment.

Quote (goran @ Mar. 15 2007,03:28)
Can any one prove (yes really prove) me wrong ?

No, I can't prove it, but it sure sounds better to me and makes my mixes easier.

according to digital logic - when a recording is made at 8 bits, then the resolution (the amount of steps between infinity and 0db per sample) is 256 - at 16 bits it is 65.536 steps and at 24 bits it is 16.772.216 steps - 32 bit depth would be 294.967.296 steps -

Dr J

First, a 24-bit card is likely to be better overall than your 16-bit card (which is probably an inexpensive built-in). So, that has nothing to do with 24 vs. 16 bits.

Second, I agree with Varakeef, with 24 bits it’s EASIER to make a good recording. To get the best results with 16 bits, you have to carefully cailbrate your analog system to your soundcard’s inputs’ “sweet spot”, and this involves some compromises: are you going to sacrifice S/N ratio or headroom? (That accounts for what Varakeef says, and I agree wholeheartedly.) With 16 bits, even an inexpensive analog chain with modern mixers has a wider dynamic range than the soundcard. With 24 bits, it’s the other way around. So, you focus more on optimizing the analog signal chain and set the soundcard input level so it won’t clip until the analog chain is well over its comfort range, and not worry about losing detail in quiet passages.

Third, I don’t think anyone can actually hear the difference between a 16-bit file and a 24-bit file in a double blind test. However, it’s pretty well acknowledged by so many serious pro engineers that there is a notable (though not astounding) difference between mixes made from 24-bit tracks versus 16-bit tracks. I have my suspicions about this but they’d be difficult to test scientifically. Also, you have to make very good mixes before this difference is significant.

So, at the level of home recordists, I feel that excellent results can be obtained using 16-bit cards. However, things are easier and quality is likely to be higher using 24-bit cards. Considering the expense most of us sink into gear, it’s silly not to add about $150 for a good soundcard. However, for a budget setup, I recommend spending money on monitors first.

Funny, that article asks the question “Should we record at 24 bits?” in a heading, but the text discusses whether we should record at 96kHz, which is an entirely different thing.

And it ignores a very significant point about that case: processing. In theory, if information over 44kHz really doesn’t matter to our ears and brains, there’s STILL a reason why 96 kHz can sound better, due to the quality of the software in plugins and EQ. You see, in theory, you can do any process on a 44KHz signal chain and get the same information-theoretic result as if it were a 96kHz signal chain with no information in it above 44kHz (and using processes that don’t add any information up there, which is important).

In practice, it’s not trivial to do what theory says we can do. Plugin writers can easily write plugins for things like chorus that work far better (sound dramatically better) on a 96 kHz signal chain than on a 44kHz signal (again, assuming there’s “nothing up there”). What they’re supposed to do is on the fly convert to 96kHz (or higher) and then back, but that takes lots of CPU power, so they often don’t.

As a result, 96kHz can sound significantly better than 44K, when using certain plugins.

That’s my hypothesis; I haven’t actually tested it. But I do know some plugins that sound like crap in exactly the cases I’d expect them to for this reason, and some day I’ll try them at 192 kHz – recording at 44 and upsampling – and see if I’m right. I’ve often heard experts say that there’s no point in upsampling, why use a bigger basket to carry the same eggs? But they’re ignoring potential flaws in the processing chain.

I’ve also read numerous articles about nyquist theory and recording and playback, that seem to prove that there can’t possibly be a difference if we really can’t hear or aren’t influenced by frequencies over 20 kHz. However, we’re not just recording and playing back, we’re processing, so those articles don’t address our situation.

I’ve also seen an excellent study that shows that folks who listen to gamelon music (in Bali) can indeed appreciate higher audio rates. However, (a) these results have not been duplicated, and (b) it uses hardware different from what we have available and doesn’t really apply to mixes for our stereos. Very few tweeters reproduce anything over 20kHz.

It is all about resolution. Now, I don’t necessarily buy the super high sample rates like 96 and 192khz… but I do buy the 24 deal. It does sound better. That extra headroom and resolution is a big dam n deal. Reveb tails are smoother, recording is easier due to the expanded head room, etc. Don’t buy the “but you mix down to 16 bit anyway.” Not the point. All this high bit depth stuff has to do with rounding. The more precision you have, the less rounding compromises you have to make and after several hundred million calculations throigh plugs ins, the mixing bus, etc, that can add up quickly. It is like selling a billion shares of stock and rounding off a half cent. A half of a cent times 1 billion is still a lot of cash.

I have a Delta 66 and a Delta 1010. I have compared 16 bit tracks with 24 bit tracks and I can’t really tell the difference. I record at 24 bits, however, and I think that 24 bits is better for a few reasons;

1) If you have a card that only does 16 bits then you have a cheap card. If you upgrade to a more expensive 24 bit card with better converters you will likely hear a difference even if you continue to record at 16 bits.

2) You dont’ have to run the levels as high and “bit hoard” when you record at 24 bits. I run my levels at at approximately -15 to -10 instead of trying to max out the meters. My mixes fall together better and I don’t have to worry about clipping. I’ve got plenty of headroom for unexpected transients or loud vocals.

3) -18 recording level on your DAW corresponds to 0 db on your preamp. By running at 24 bits and recording at lower levels you are running your preamps in their sweet spot instead of overdriving them. That will probably help the quality of your recordings more than anything else.

I have a question about this.

Presuming that your desired end result is a 16/44.1 wav file, say for a CD. You’re gonna do the whole thing on your own DAW, including your best effort at a mastering pass.

Say you record at 24/48. While mixing, you indeed get the advantages identified above. But then after you’ve mixed and done a home-mastering effort, you have to produce a 16/44.1 version of it. This requires a conversion/dither process.

In the situation most of us are in (i.e. we don’t send our mixes off to a pro mastering house) we are using our not-so-pro conversion tool, such as the one in N.

Would you be better off mixing/mastering at 16/44.1 and therefore not needing to convert to new bitdepth at the end, or have a better mix but a compromised final result anyway because of your final conversion requirement.

Perhaps I misunderstand the impact of those two conversions: i.e. from 24-16 and 48-44.1, but I always wondered.

I was using N at 24/48 for the last couple of years, but I recently got a new laptop which just can’t handle that, so I recorded my most recent project at 16/44.1. The results are better than I’ve got for a long time, so I am really wondering about this.

Ok, let me muddle this up a bit. I record in 24bit with standalone recorders. I then import the wave file as 16 bit and mix them in n-Track.

Am I getting the advantage of recording in 24 bit (not just headroom but audio quality) and losing it when I convert and mix in 16 bit?

I don’t think you are, mojo.

Re: the original question - I’m curious what “prove” means in this context. I can hear a rather significant difference between 16/44 and 24/44, and esp. 24/96. Isn’t that subjective experience proof in the relevant sense in this context? Next question, what does “better” mean? Here it can only mean that one subjectively prefers it. So, if one likes the sound one gets at 16 with one’s equipment, then that’s what one should use.

Thanks for the response TomS, but I’m not sure I understand your answer.

Let me rephrase my question: By converting and mixing in 16 bit, am I losing the sound quaility I gained when recording in 24 bit?

Don’t mean to hijack the post, just curious.

Depends on how the import is done - you might just be truncating the last 8 bits and hence losing that info and raising the sound floor, or you might be dithering to 16, in which case you get some but not all of the benefits of 24 bits.

if instead of ears you had a LINE IN socket, then your brain would have a fixed 20 bit resolution - 16 bits is not enough to obtain the best hearing resolution, 24 bits is to high, but as it is far easier for designers to add 8 bit resolution than it is to add 4 bits you have 24 bit soundcards - the actual difference is only 4 bits thats why you cannot (unless you have GOLDEN ears) detect any difference - its there OK but it takes more than the average PC (even with a 24 bit soundcard and pro monitors) can reproduce -

look here for the sort of system that can reproduce audio as it should be heard (no i cant aford one)

by the time your audio has been smashed down to iPod level it does matter if it started out as 16/20/or 1meg bit its rubbished and most CD players have useless audio stages -

if you really want to obtain a good quality CD you have to take a different route one that preserves the extra quality (although slight) that 24 bit soundcards reproduce - and that is to (record and) playback at 24 bit 48000khz sample rate to an professional outboard CD recorder, the audio then is recorded at 16 bits not down-sampled to 16 bits - does it make a difference ? OH YES it does - and as a pro CD recorder costs less than some plugins maybe you should look into getting one -

Dr J

I have some thoughts about this “headroom thing”.

IF you record in 16 bits you may strive for a signal peeking at maybe -6 to -12 db.
-12dB is 2bits (26=12) so there is 14bits left (16-2=14) and 14bits is 146db = 84dB.

Now look at your “recording Vu meter” in nTrack. It just shows 60 dB.
I can’t get mics - noise(room+mics) below 60dB (can You ?).

So there is no need for more than 16bits when it comes to noise.

Now you say you have more headroom with 24bits. So you must mean you will record at something like signal peeking at maybe -12 to -24 dB.

But then the analog-signal is quite low and I could suspect the soundcards AD-converter
not to be really precise and that will mean distortion. Why should we take that risk.

Also the analog amplifiers on the soundcard will give more (and the nasty form) distortion
at low signal levels. As long as the amplifiers are not class-A.

My conclusion is that there is no need for more than 16bits
and 24bits could lead to a lower audio quality if you record at a very low signal level.

(I have a “ESI Maya 44” 24bits soundcard but just the 16bits-nTrack_4.)

/Goran Sweden

I still say the technical discussions are beside the point - if you get the sound you want to get at 16 bits, then 16 bits is “better” than 24. If the sound you want to get is at 24/192, then that is “better.” “Better” is always relative to your purposes. Someday I’d like to own a big tape machine. But I never will.

There was that cool article in tape op recently about the fellow who recorded an album on everything from edison cylinders to wire to tape. Each has an aesthetic.

That said, if lower noise floor and more detail is what you mean by “better”, then with good converters and good speakers (not even great) you can hear the difference.

Plus - won’t we someday soon be listening as a regular thing to formats better than 16 bits? It’s like 40+ years ago when folks were doing mostly mono, but a few people recognized that stereo mixes would be valuable in the future, so they’d do two mixes, one mono one stereo. Plan ahead. If you’ve got it, use it.