32 bit float meets 24 bit ASIO sample size, what happens?

[Staccato]

If anyone knows this and cares to post it, thanks in advance!

 

[polygen]

32 bit float streams (as inside most DAWs) get dithered down to the audio driver's bit depth at the point the DAW passes the signal to the driver....like for example the output objects in the logic mixer.

 

[Staccato]

Thanks again, if dithering is disabled, then I'm looking at truncation to the driver depth!

 

[jcat]

Thanks again, if dithering is disabled, then I'm looking at truncation to the driver depth!

I don't know about every sequencer, but I know about SX.

SX will dither down to whatever the project is set to on the way out of the buses, and the mixer is always 32 float no matter what the project is set to.

So if you set your project to 16bit it will dither down to 16bit on the may out of the busses, and to 24bit if you set the project to 24. If you set the project to 32bit the signal won't be dithered on the way out. But you can always put a dithering plugin in slots 7 or 8 to do it for you in certain circumstances.

Hope that helps 8)



Cheers,

jcat

 

[Joey123]

Correct me if I'm wrong but I've thought that 32-bit float is just a scalable 24-bit; in other words, the 24-bit word length is \"floating\" in the greater bit-buffer of 32-bits' ability to hold greater allowance for variability.

But to my knowledge and again if I'm wrong please correct me, but I thought 32-bit float is jsut 24-bits in a nice padded, mathmatical blanket 8)

 

[Staccato]

That sounds reasonable, I was trying to sort out what host software does under certain circumstances, and now I realize, the host will hand off whatever it is told to, to the sound card (driver), whether the instructions come from the host, or the driver. Thank you for your reply.

 

[jcat]

Correct me if I'm wrong but I've thought that 32-bit float is just a scalable 24-bit; in other words, the 24-bit word length is "floating" in the greater bit-buffer of 32-bits' ability to hold greater allowance for variability.

But to my knowledge and again if I'm wrong please correct me, but I thought 32-bit float is jsut 24-bits in a nice padded, mathmatical blanket 8)

32bit does indeed have the same resolution as 24bit, but it still needs to be converted from 32bit to 24bit ASAIK..



Cheers,

jcat

 

[Sonic]

And when you have an 32bits file and want to convert to 24bits, do you have to use a dither?

 

[jcat]

Yes, a dithering plugin will make the necessary changes to the data.




Cheers,

jcat

 

[fishtank]

And when you have an 32bits file and want to convert to 24bits, do you have to use a dither?

NO!

You do not have to use dither - and although you can use a plug-in to add the *proper* amount of dither noise for 32 bit to 24 bit conversion, the difference will be inaudible as the quantization distortion is FAR below the limit of human hearing. It shouldn't hurt to use dither (when going from 32 to 24), but there is no good reason to do it and it is simply a waste of time.

You do need to use dither when going to 16 bit however, as the qantization distortion is quite audible at lower bit depths - but this should be at the last step. Dither noise is additive and if used multiple times on the same tracks, it could become too much of a good thing.

 

[Staccato]

And when you have an 32bits file and want to convert to 24bits, do you have to use a dither?

NO!

You do not have to use dither - and although you can use a plug-in to add the *proper* amount of dither noise for 32 bit to 24 bit conversion, the difference will be inaudible as the quantization distortion is FAR below the limit of human hearing. It shouldn't hurt to use dither (when going from 32 to 24), but there is no good reason to do it and it is simply a waste of time.

You do need to use dither when going to 16 bit however, as the qantization distortion is quite audible at lower bit depths - but this should be at the last step. Dither noise is additive and if used multiple times on the same tracks, it could become too much of a good thing.

In a way, this says there's no reason to have 32 in the first place?, since the LS 8 will be tossed, indeed, all my Sonar settings are 24.

 

[jcat]

The dither will get rid of the floating point, so it should realy be used. Obviousely 32bit gets no more headroom or resolution than 24bit, but it's damned handy in a digital mixer! =D>




Cheers,

jcat