|
Well, actually, 32-bit is the standard. Almost every sequencer uses 32-bit floating point internally.
So you must be referring to one of two things:
a) Recording;
b) Final mixdowns.
Let me address (a) first. The redbook audio standard is 16 bits, 44 kHz, and this was based on a very extensive study to determine what it would take to make recordings with effectively zero audible noise - and this includes a pretty good "safety margin". The noise floor with 16/44 is far lower than what any recording studio could ever hope to achieve in terms of ambient noise.
The standard doesn't really take into account the possibility of mixing 20 or 30 of these recordings together. If you take all of your synth and other tracks in a production, and bounce them all to 16-bit audio, it is definitely possible that you'll get some noise. That is why most producers actually do render intermediate audio as 32-bit float, which is the same as what the sequencer itself does.
So high-end production equipment mostly switched to 24-bit recording. Most ADCs and DACs are 24-bit. In theory, a 24-bit recording allows producers to mix 30, 40, 50 of these recorded tracks together and still have the pristine output quality of a typical 16-bit one-track recording of a band. In practice, again, the conversion noise is generally dwarfed by ambient noise, but I guess every little bit of noise reduction helps.
32-bit converters exist but they are pretty rare. And here's the kicker: a 32-bit DAC is going to be 32-bit fixed point. When you convert that to 32-bit floating point in a sequencer, you are actually going to start losing fidelity. If you've ever seen a service bill from some rinky-dink web company stating that you owe $32.000000007 in charges, that is an example of what happens when you try to convert from fixed to floating point and then back again. So even if you could get your hands on a 32-bit ADC, it wouldn't end up much better than 24-bit. Sequencers would have to switch to doubles (64-bit float) in order to take advantage of this.
So on the recording end, to summarize: the current standard is 24-bit, not 16-bit, so using 16 bits as a reference point for comparison is flawed. But even then, I'd be surprised if you could really hear the difference unless you're mixing a huge number of tracks all at 16 bits.
Now onto (b) the final mixdown:
One of the things you need to understand is that DACs are fixed-point. You can't actually take a floating-point signal and convert it directly to analog. Even if you found a DAC that claimed to do it - and I've never heard of such a thing - it would still have to convert to fixed-point internally first, which would require many more bits of precision than the original floating-point signal to do accurately.
I'm willing to bet that you do hear a difference when you do your final mixdown in 32 bit float, but not for the reason you think. Your sound card/Audio Interface, in all likelihood, has a 24-bit DAC on it (or maybe even 16-bit if it's cheap). If your sound driver even allows you to play 32-bit float audio directly, then it's getting down-converted somewhere in the chain, either by the driver itself or by the AI internally. And it's probably using some cheap, fast dithering algorithm, or maybe even doing straight truncation.
So what you're actually hearing when you try to work with 32-bit float is not higher fidelity, but lower fidelity. It's a combination of conversion loss and aliasing distortion. This is why we have dithering algorithms like the UV22-HR - complicated algorithms that obviously can't prevent artifacts, but create the artifacts in just such a way that they're virtually impossible to hear.
Even doing your mixdown in 24-bit fixed point is a bad idea (unless your destination is a mastering studio). It may sound better on your system, but you're going to get the same sort of crappy dithering/truncation if you encode as MP3 (which is 16 bits) or mix it onto a CD or vinyl that gets played on some much lower-fidelity consumer or club system.
Hope that answers your question. For mixdowns, stay away from 32 bits. Stick to 24 if it's headed to a studio, or 16 if your target is physical media. For recordings, they are likely happening at 24 bits at the hardware level, but are already getting converted into 32-bit float when you record from within the sequencer.
___________________
My party schedule:
2009-02-21 - DJ Attention @ I'm So Popular
2009-06-18 - DJ Annoying @ People Need To Know Where I'll Be
2012-11-32 - DJ Insufferable ɸ Or At Least the Stalkers I Complain About
2048-06-66 - Spastic & Whocares ¶ Although I'm Actually Flattered
9999-45-81 - Tweaker Gimp ☼ I Probably Won't Even Go To This But I Have To Make Sure I Fill Up All The Available Space Here
|
