About bit depth, mixing, processing and audio quality.

[Gabriel Plalame]

Is the bit dept of a sample library permit any influences about the process that a sampler engine have to assume to mix several samples in polyphonic mode ?


Here a discussion about the pertinence of a 32bit sample library on Nortern Sound Source :

http://www.northernsounds.com/ubb/NonCGI/ultimatebb.php?ubb=get_topic;f=3;t=007011;p=4

What’s really the benefit of incrementing bit depth ?


Any opinion to share here ?



Regards

[Nick Batzdorf]

More bits means more low-level detail. It's a tradeoff between getting mileage out of computers and sound quality.

The difference between 16-bit samples (as the VSL has in its current release) and 24-bit samples dithered down to 16 bits after mixing is going to be subtle but audible; the difference between 24- and 32-bit samples dithered to 16 bits is likely to be subatomic.

Giga 3 is supposed to have a lossless compression scheme that lets you use 24-bit samples without any performance penalty over 16-bit ones. i don't understand how that works, but it probably has something to do with the fact that all sound is made up of sine waves; knowing that, you can predict where it's going without storing all the data.

Dietz?

[Dietz]

@Nick Batzdorf said:

More bits means more low-level detail. It's a tradeoff between getting mileage out of computers and sound quality.

The difference between 16-bit samples (as the VSL has in its current release) and 24-bit samples dithered down to 16 bits after mixing is going to be subtle but audible; the difference between 24- and 32-bit samples dithered to 16 bits is likely to be subatomic.
[...]
Dietz?

Uh - I couldn't put it better :-]

/Dietz - Vienna Symphonic Library

[Gabriel Plalame]

I understand that point, but the problem is less the dithering step after mixing, than the mix step itself.

If a 32bit depth gives nothing more, so why Steinberg create a 32flt Asio driver for recording, and so processing and mixing ?

[Peter Roos]

@Crystal said:

I understand that point, but the problem is less the dithering step after mixing, than the mix step itself.

If a 32bit depth gives nothing more, so why Steinberg create a 32flt Asio driver for recording, and so processing and mixing ?

32 bit floating point gives an enormous headroom, you can hardly clip anything within the audio path. Also if you apply a lot of calculations in a row, 24 bit fixed point will lose accuracy due to roundings during the calculations. 32 bit FP does not have this problem.

Peter

[Gabriel Plalame]

That’s probably why GS mix in 32bit float too...

Because... GS is a mixer too.

This is a bit OT, I know, but,

(Not really, in fact)

(Anyway this is my thread !)

(Hum.)

...this is more a GS post, but I was wondering : are Gsif drivers delivering a 32bit flt signal to Asio 32bit flt ?

Probably yes.

Or it depends of the sound card Gsif drivers...

I own a Creamware card, and can’t ask them anything.

There are practically ‘out’, now.


But if Gsif drivers deliver a 32bit signal into Asio 32flt, the GS mixing process is safe, and the result in the sequencer mix optimal...

[cm]

crystal, as far my understanding of the architecture lasts, the 32-bit process is what happens within the software/application. if it comes to output of sound through your sondcard it depends on the bit-depth the drivers provide. you will see in settings of gigastudio a dropdown to choose between 16/24/32 bit (if provided!). therefore GSIF doesn't deliver anything to ASIO - both deliver to/from your ins/outs on your soundcard
christian