Another sample rate/bit depth question...

[IIRs]

I think the best idea is to export the final file with as high bitrate

Yes

and samplerate as possible,

No: stick to the sample rate you mixed at.

then use a good dither

Yes

and a good resamler (eg r8brain pro)

better still not to re-sample at all.

[C00kie]

The error isn't neccessarily in the least significant bit.

Sorry, I don't buy it. Please tell me, where do you expect the quantization error to be other than in the least significant bit? Because that seems to be what you have been explaining.

For anybody thats not interesting in splitting hairs, here is some solid advice. For peace of mind it is safe to record, process & save everything in the highest possible bit depth supported, and stick to the sampling rate you intend to publish on (e.g. 44.1kHz)

[audiobot202]

If I mix down from the app at 24/44.1 and then open in Audition at 32 bit floating point for mastering...I then apply Ozone which has inbuilt dither. I choose to dither to 16 bit and hit the render button. The track still seems to be a 24 bit one, but now with the added dither noise. So, should I simply master with Ozone with inbuilt dither turned off and then save from Audition to a 16 bit file...will Audition automatically introduce dither noise to compensate?

[IIRs]

If you export at 24 bit you should really dither to 24 bit, even if its going to be mastered and re-dithered to 16 bit in the end.

Better to export at 32-bit if you can, as you lose no data at all and therefore need no dither at this stage.

I choose to dither to 16 bit and hit the render button. The track still seems to be a 24 bit one, but now with the added dither noise

Thats fine, you just need to bit-truncate the dithered audio, making sure that no changes take place at all between adding the dither noise and throwing away the bottom few bits.

[birrbits]

Cookie, if say a voltage level is interpreted as 00100000 instead of 00011111 is this quantization error isolated to the least significant bit? No, because quantization error isn't bit error! It is the error in representing an actual voltage level with a quantized voltage level. The only reason I express the final result in bits is that it shows the degree to which the problem can occur....I hope this makes sense

[Warmonger]

OMG! 1111111111111111 is 16 bits different from 000000000000000!

[birrbits]

OMG! 1111111111111111 is 16 bits different from 000000000000000!

bah no, not what i mean, its just easy to confuse the issue by only thinking of the lsb being affected.... division/multiplication is shifting last time i looked....

Ok C00kie, you win, my lil calculations did NOT included dividing by 8. So what does all this mean?? Is birrman a moron? I hope not, because the possible bit error from quantization noise with 8 16 bit tracks is still eliminated with 8 24 bit tracks.

[Warmonger]

I agree that you should record at the highest possible bitrate to reduce quantization error (which in turn reduces accumulated error). I also think that this example of 8 tracks simply being mixed misses the point.

For any project, you will probably be using some kind of effects on the audio. Any processing you do will increase the amount of error, and recording with 24 bit waves reduces the impact of this processing on the accumulated error.

In the end, I'd be surprised if anyone could tell the difference in a mix in a well-designed scientific test.

[MrMenace]

There had been an interesting conversation or two regarding this exact topic on the Cakewalk boards. One of the groups of threads there had to do with bringing in multiple audio files recorded at different rates. The resulting master take due to the dithering process and other processes (track compression and such) ended up with less sound than expected (The description of the music was that it was thin.)

Where I am going with this was that during these threads the conclusion most came up with was that if you have a quality package for dithering the 24 bit file down to 16 bit, it is a good idea to use 24 bit for recording.

As stated here, it provides more bits per sample so that when you use compression or other effects and finally dither down to 16 bits, you are more likely to have most of your 16 bits per sample be usable data and not zeros introduced from the various plugins and dithering.


Dennis

[flex42]

I keep reading here at KVR that some VSTis sound better if you run at 96kHz. Is that true?

[IIRs]

In the end, I'd be surprised if anyone could tell the difference in a mix in a well-designed scientific test.

Well this isn't a scientific test obviously, but heres a clip of my old band from about 7 years ago, which was recorded and mixed entirely with my Roland VS880. (I got one of the first ones in the shops! )

Elfin - Lizard Jam (Intro)

Although I feel I got the recording and the mix about as good as could be expected (everything but the vocals was recorded live to the 4 available analogue inputs! I actually recorded a stereo mix of drums on two tracks, and a stereo mix of guitars on the other two, so most of the mixing had to be done by automating the channel EQ's with Cubase running on an old Atari. ) and we even got a Sound on Sound top tape award, but if you listen carefully to the guitar intro, you can hear a little bit of "modem noise" type distortion coinciding with each chord, and the mix as a whole has a brittle 2-dimensional quality from the lack of low end detail.

The same 16-bit audio recordings would have sounded noticably better were they exported from the VS in some way and mixed with a modern 32-bit app, but the results would be better still if the original recordings had a higher bit-depth.

[IIRs]

I keep reading here at KVR that some VSTis sound better if you run at 96kHz. Is that true?

Probably, but that's a different issue.. any vst that produces aliasing (whether its a synth or a distortion effect) will have that problem reduced by the use of a higher sample-rate (which is why many quality plugs use "oversampling")

It is also possible that filter algo's will sound different at the same frequency setting, because behind the scenes the filter calculates in terms of "fractions of nyquist".. but again this becomes irrelevant if the plug-in uses oversampling.

[flex42]

Ok, thanks!

[JumpingJackFlash]

Ok, it seems I asked a controversial question here!

A couple more questions first;
If render my track to video and put on DVD-R, is the DVD limited to 44.1/16 in the same way as an audio CD is?

From what (little) I understand, the only possible disadvantage to recording at the highest sample rate and bit depth as possible is due to bad dithering? - Correct?? (Asside from high CPU usage)

Does high bit depth and sample rate actually make the track sound better? - Or does it just give more room (and are these connected?)

Does dithering only apply to bit depth, or does it work with sample rate too?

If, as I suspect, dithering has nothing to do with sample rate, then surely there is no possible disadvantage to going with the highest sample rate possible? - Assuming the track is for distribution by a means other than audio CD. Correct??

Now, my setup is as follows; I'm using several VSTs (both instruments and effects), my tracks do not have lots of dynamic varience, nor am I doing anything wacky with pitch. The only dithering thing I have is the default thing that comes with Logic 5.5.
The final track is being disrtibuted in two ways, firstly by audio CD, secondly as a piece of underscore to a movie, rendered to DVD in Adobe Premiere.

What would be the best sample rate and bit depth to use for both of the above mediums?

Sorry, long post, but the responces are interesting.

[IIRs]

Does high bit depth and sample rate actually make the track sound better? - Or does it just give more room (and are these connected?)

High bit-depth repsents low level signals more accurately. This is almost always audibly better, and it allows more headroom to be left without signals dropping into the grainy low-level region, which means there is less need for potentially damaging compression on the way in.

High sample-rates are more controversial: all they do is allow higher frequencies to be represented. As the typical human being is lucky if their hearing still goes up to 20KHz, many people argue that CD rates are high enough already.

Even if you think those higher frequncies contribute meaningfully to the sound, whether it is useful to keep them in during the mixing process when they will only be thrown away to go on CD, is another question again!

The final track is being disrtibuted in two ways, firstly by audio CD, secondly as a piece of underscore to a movie, rendered to DVD in Adobe Premiere.

What would be the best sample rate and bit depth to use for both of the above mediums?

I'm not sure about DVD movie soundtracks.. anyone know what sample-rate they use?

Your best bet for both mediums is to record at 24-bit.

For the CD release, either stick to 44.1KHz throughout, or use an exact multiple (ie: 88.2KHz) as sample-rate conversion is much easier (and better sounding) when you just have to throw away every other sample (otherwise it has to "interpolate" to guess what would have been happening between the values it knows)