16 VS 24 Bit. Performance Difference?

[a2zproductions]

I'm doing a 3 Track project, 2 Stereo, One
Mono. Is there a speed difference between
whether or not I set the Project up for 16
bits or 24 bits. I want maximum performance,
if setting the Project for 16 bits will allow
more plugins and stop the frequent DAE errors
I'll try it.

Also, what is the advantage of having a 24 bit project, when you just bounce it down
to 16 for a CD?

TIA,
Adrian

[loopmusic]

The definition & representation of the recorded sound is much better at 24 bit.

Fair enough, you are going to bounce down to a 16 bit CD; but it would be nice to have the extra quality available to you when it come to editing & using effects. Any errors (typically caused by the PC rounding out calculations) will be less noticable on the recording.

On the other hand - if you are going to record straight vocals / acoustic instruments and use absolutly no plugins; then a 16 bit recording won't matter in the slightest.

I tend to record synths etc. at 16 bit as the sound quality [to me] doesn't matter a damn as I am after a groove as opposed a particular sound. When doing vocals I try to use 24 bit as we use a lot of plugins.

Dan

[McGriffy]

I always record at 24 bit. When properly dithered, it will sound better at 16 bit than a straight 16 bit recording. This is true if you do any processing. And remember that so much as one fader at a level other than zero is processing.

If you want to hear the difference try this test. Record 15-20 seconds of a simple test tone at 16 bit. Now automate the volume so that it goes from about -40 to -96 and bounce it. Crank your monitors and listen to the very end of the resulting file. It should get really rough, grainy, whatever you call it. It's a form of distortion. Now repeat adding dither in the bounce and repeat at 24 bit. The volume change is computed in 32 bit float internally so even if your source is 16, there is stuff in those last 8 bits once you touch a slider.

This test (thanks to Bob Katz - www.digido.com) made it obvious to me. They say you can hear details at -110dB in a properly dithered 16 bit recording where the undithered dynamic range is only 96dB

[Toney Robinson]

I guess I thought the bit depth affected the dynamic range only, and the actual "quality" has more to do with sampling rate. You can read more about this stuff from a article by Lynn Fuston http://www.3daudioinc.com. To summarize: He was listening to classical music at high sampling rates, like 192khz and according to him there was definately an audible difference. But when the bit depth was reduced to 16 bit, he heard little or no difference.

BY THE WAY, Digi, if you are listening, I'd sure like to see you come out with a version of 001/PTLE that supported 88.2khz!

Originally posted by McGriffy:
I always record at 24 bit. When properly dithered, it will sound better at 16 bit than a straight 16 bit recording. This is true if you do any processing. And remember that so much as one fader at a level other than zero is processing.

If you want to hear the difference try this test. Record 15-20 seconds of a simple test tone at 16 bit. Now automate the volume so that it goes from about -40 to -96 and bounce it. Crank your monitors and listen to the very end of the resulting file. It should get really rough, grainy, whatever you call it. It's a form of distortion. Now repeat adding dither in the bounce and repeat at 24 bit. The volume change is computed in 32 bit float internally so even if your source is 16, there is stuff in those last 8 bits once you touch a slider.

This test (thanks to Bob Katz - www.digido.com) made it obvious to me. They say you can hear details at -110dB in a properly dithered 16 bit recording where the undithered dynamic range is only 96dB

[JMS40]

Hi Adrian-
Personally, I like to record and archive masters using the highest quality settings available to me at the time. Yes, we step down to release on the current CD format, but the industry standard won't always be so far behind us and, who knows, one day when they are rereleasing your greatest hits, you'll be happy it's there.
I'm curious about your system specs if you are having DAE errors on a 3 track project...

Originally posted by a2zproductions:
I'm doing a 3 Track project, 2 Stereo, One
Mono. Is there a speed difference between
whether or not I set the Project up for 16
bits or 24 bits. I want maximum performance,
if setting the Project for 16 bits will allow
more plugins and stop the frequent DAE errors
I'll try it.

Also, what is the advantage of having a 24 bit project, when you just bounce it down
to 16 for a CD?

TIA,
Adrian

[McGriffy]

Thanks for the link to that article on 3DAudio Toney. I am fascinated by this subject even when (like this case) I have no ability to test it out for myself. I would never argue against high sample rates and am in no position to argue the relative merits of bits vs. rate. One thing I wonder about is what form of dithering they were using. A discussion at this level can't ignore the subject.

I use the best I have available which is 24 bit. Actually I guess that's not quite true because I always record at 44.1 when heading for CD's because I think the conversion will be worse than the original limitation. (Any thoughts on that one?).

What I know is that I can now hear the difference in 16 to 24 bit. There are some old CD's that I now listen to and I'm pretty sure that they are all 16 bit productions. A little harsh, but mostly fatiguing to listen to long term.

And remember, it's not the noise floor. The pre's and converters in the 001 have a noise floor about the same as most 16 bit systems. It's the distortion caused by simple truncations.

DMcG

[a2zproductions]

Thanks for all the helpful info...
I believe I'm going to buy a G4.

My Win PIII 500 is not getting the
job done with mutiple plugins.

Regards,
Adrian

[Toney Robinson]

The side effect of inadiquate bit rate (dynamic range) is quantization noise which sounds like gross distortion (square waves). It is irrelevant to most pop music because everything is so squashed dynamically. Where you typically hear it is in classical stuff where there are very quiet passages, or as a single instrument is just decaying down to total silence. Good dithering can eliminate it. By the way, I too always use 24 bit for tracking, but converting sampling rates is a much more complicated. It's all related to math. If you can convert a sample rate in a manner that is divisible by a whole number, like 88.2khz / 2 = 44.1, you've got it made. 48khz to 44.1 is mathimatically and sonically bad news.

Originally posted by McGriffy:
Thanks for the link to that article on 3DAudio Toney. I am fascinated by this subject even when (like this case) I have no ability to test it out for myself. I would never argue against high sample rates and am in no position to argue the relative merits of bits vs. rate. One thing I wonder about is what form of dithering they were using. A discussion at this level can't ignore the subject.

I use the best I have available which is 24 bit. Actually I guess that's not quite true because I always record at 44.1 when heading for CD's because I think the conversion will be worse than the original limitation. (Any thoughts on that one?).

What I know is that I can now hear the difference in 16 to 24 bit. There are some old CD's that I now listen to and I'm pretty sure that they are all 16 bit productions. A little harsh, but mostly fatiguing to listen to long term.

And remember, it's not the noise floor. The pre's and converters in the 001 have a noise floor about the same as most 16 bit systems. It's the distortion caused by simple truncations.

DMcG

[Toney Robinson]

P.S. There is one more aspect a lot of people tend to forget about. In order to insure that no frequency above the Nyquist rate (22.05khz for 44.1khz sampling rate) gets "digitized", a "brick wall" low pass filter must be employed. It must be done in the analog realm and even the most sophisticated modern filters introduce artifacts into the signal, the most noticable being phase shift. If you sample at 88.2khz, the Nyquist rate is 44.1, well above human hearing. You can employ a much gentler sloped filter beginning at around 22khz and thus dramatically reduce the undesireable artifacts.

[McGriffy]

Another aspect of high sample rate is stereo imaging. I'm working from memory here so please forgive any slight factual errors, but it seems that most humans can distinguish times as small as 15ms. I believe the test was a pulse in each ear seperated by some small delay. Some people could reliably hear 5ms. Now consider that one sample at 44.1 is 23ms. Not hard to see where this could greatly affect that part of stereo imaging that is based on phase. Dithering can alleviate this to some extent but it's a pretty good argument for high sample rates as far as I'm concerned. This fits well with the observation in the 3daudio link above that the stereo imaging was all over the place at low sample rates.

As to brick wall filters - oversampling converters can help a lot, but it's always an issue. This is one of the reasons people like SACD so much. It's basically a 1 bit 2.88 MHz system.

DMcG