Dithering, bouncing... whatever you want to call it... need some help...

[AudioPhile777]

Ok... so, a buddy of mine is just getting into recording and he posed an interesting discussion...

I record at 24/48... he records at 16/48... I bounce to 24/48 to pseudo-master my stuff... he bounces to the same bit/sample rate to pseudo-master...

He claims that is gives you more headroom going from 16 to 24 to master...

And, in a way that makes sense... but is it worth it if you're right back out of the 24 into 16 to render it to cd?

if someone could explain the mathematical pros and cons... not just preference, or "I hear this or that difference", but an explanation of the numbers behind it... I would appreciate it...

also, if this has been explained... I wouldn't know how to begin searching for it.

[colonel kurtz]

there's absolutely no benefit in converting a 16-bit file to 24-bit. all it does is add 8 bits of nothing.

[AudioPhile777]

Originally Posted by colonel kurtz

there's absolutely no benefit in converting a 16-bit file to 24-bit. all it does is add 8 bits of nothing.

Thats what I thought too... but in a simple test, the audio bounced to 16 sounded slightly smaller and more compressed than the same audio, same 2bus plugs on the bounce, bounced to 24...

I had my back turned and told him to play a random clip and then the other... so I didn't have a clue which one was which... but it did sound like the 24 clip had more depth to it...

[Tym_ex]

I think that's because even when you record your stuff @ 16 bits, your DAW and plugins use a minimum of 32 bits for all calculations to have more headroom. So in the end it does use the extra 8 bits for all plugin calculations and the summing process. Obviously, the difference will be much smaller then using 24 bit recording in the first place. Correct me if I'm wrong though.

[AudioPhile777]

That was my friend's argument... the computer runs at 32 bit... the audio is 16bit... going up to 24 or even 32 bit to master then bounce back down to 24....

either way... is there any point to doing that since you will end up back at 16 bit?

That is my question.

[::XeS::]

There is a reason because Andy and Fredman record their stuff at 24/44 and bounce to 16/44

[AudioPhile777]

Originally Posted by ::XeS::

There is a reason because Andy and Fredman record their stuff at 24/44 and bounce to 16/44

You didn't understand what I said.

I'm asking if there is any point to record at 16/48, bounce to 24/48 to master, then bounce to 16/44 for disc.

I've already assumed that there is no benefit or point to doing that.

[::XeS::]

Yes I understood...I was only in funny phase
Anyway... No, there are no benefits.

[2012]

There is NO benefit.

You're just adding 0's.

If you upsampled (i.e. 44.1 to 96k), there would be a difference, but not very significant, and the conversions will probably degrade the audio more than you would gain by upsampling.

Some people upsample the entire mix from 44.1 to 96 because some plugins sound better at higher sample rates.

P.S. Tell your friend to quit recording in 16-bit. It's 2007!

[QV]

If you convert before summing (all tracks) you're adding zero's which give you more headroom when you're actually summing, so all eq's and any digital changes have more headroom before they get rounded off and lost or approximated. Since these effects are cumulative there would be a difference in the final stereo track.
I agree it would make no diff if your're converting stereo file and just play it back with no manipulation.

[Tym_ex]

I don't think that's true, because your DAW will process the audio @ 32bit or higher anyways. So if you wouldn't convert all tracks to 24 bits but bounce your 16 bit project @ 24 bits the extra headroom will still be used... I think...

[BassTard]

Going for the mathematical theory I would recommend to check with a bitmeter.
Record something using 16 bit and convert to 24 - play back in a track with no volume-changes and no plugs. Watch the output using a bitmeter... usually it should still indicate that you're working with 16 bit - just until you change something.

So for me there's no point in recording 16bit - I think every modern DAW uses a higher bit depth for mixing (as has been said before) - so you will have more headroom nevertheless.
If you want to make sure, convert all tracks to 32-float, that may spare CPU as well. Mix down to 32 and decrease depth afterwards using a good dither or Voxengo's R8brain.

I believe to whitness less transparency and discernibility in 16-bit recordings - you definitly loose overtones that may proove important in a mix.

[QV]

Originally Posted by Tym_ex

I don't think that's true, because your DAW will process the audio @ 32bit or higher anyways. So if you wouldn't convert all tracks to 24 bits but bounce your 16 bit project @ 24 bits the extra headroom will still be used... I think...

if this were true, then wouldn't the word length be irrelevant? By the same logic - If I understand your point correctly - if one recorded in 8 bit, it would still sound as good as 24 bit.
The daw may be at 32 but the audio wordlength will be at 16... 24 bit gives you extra headroom which will make a difference in summing.

[Tym_ex]

Originally Posted by QV

if this were true, then wouldn't the word length be irrelevant? By the same logic - If I understand your point correctly - if one recorded in 8 bit, it would still sound as good as 24 bit.
The daw may be at 32 but the audio wordlength will be at 16... 24 bit gives you extra headroom which will make a difference in summing.

Hmm no wrong conclusion. I'm not saying there's no point in using 24 bit recording, there is! I'm just stating that there IS a difference in bouncing a mixed 16 bit project @ 16 bit or 24 because of the higher internal bitrate your DAW uses.

A 16 bit recording will have 16 bits of information, it will always have those 16 bits and not more, your DAW can't magically guess the other 8 bits when converting to 24 bit. But, when you have an entire project of 16 bit files your DAW uses 32 bits or more to sum those tracks together so there will be a difference. That's all the topic starter was asking...

Anyways, there's obviously no good reason to record @ 16 bit in the year 2007, but that wasn't the question.

[Jevil]

I can't notice any difference between 16/24/32. Should I go to check my ears? Last time I went to the doctor he told me my ears were great.

And talking about Khz, which is better, 44.1 or 48? Does it make any difference for a human ear? I was told that 48 is better for CPU performance.

[fatalforce]

If you record in 24 bit you need to add dither. Or else when you bounce, your truncating the bit rate (aka not good). When you could record at 16 bit, I prefer 24 bit because the quantizing error is less when you go from analog to digital.

[Metaltastic]

Originally Posted by Jevil

I can't notice any difference between 16/24/32. Should I go to check my ears? Last time I went to the doctor he told me my ears were great.

And talking about Khz, which is better, 44.1 or 48? Does it make any difference for a human ear? I was told that 48 is better for CPU performance.

Well the higher the sample rate, the higher the frequency that can be recorded. It's called the "Nyquist Theorem," and it denotes that the highest possible freq. that can be captured is half of the sample rate (so 22.05 kHz at 44.1 kHz SR, 24 kHz at 48 kHZ SR, etc.). When a freq. higher than that cutoff point (the "Nyquist") is sampled, it creates aliasing, a false sound at a lower freq. - but this isn't an issue, since every converter has a Low-Pass filter at the Nyquist.

As you can probably imagine, because we can't hear above 20kHz, a higher sample rate really doesn't make a difference. A higher bit depth, however, does, because it greatly increases the potential dynamic range (and thus headroom).

Raising either of them doesn't tax your CPU any more, but it does gobble up more HD space.

[fatalforce]

Originally Posted by Metaltastic

\s you can probably imagine, because we can't hear above 20kHz, a higher sample rate really doesn't make a difference. A higher bit depth, however, does, because it greatly increases the potential dynamic range (and thus headroom).

This is true but in theory, an analog wave is continuous. When you convert that wave into digital it takes samples depending on the sample rate. The lower the sample rate the less your wave will truthfully turn out. Wouldn't you want your waveform to be more accurate by taking more samples in the A to D conversion?