Floating point and integer

[goula]

Hi everybody, hi Har-Bal team;

I'm used to import 32 bit float files coming from Samplitude into Har-Bal.
After I make corrections with Har-Bal, and record them, I get 32 bit fixed point (integer) audio files as results.

So I have to import these 32 bit integer files into 32 bit float files in Samplitude, to finalize the job (dithering...and so on).

Being not a DSP programmer, I'm afraid these 32 bit float>>>32 bit fixed point>>>32 bit float again conversions could be neither totally conservatives nor totally "transparent"?... Am I right to be afraid of that?

Is there a way to get Har-Bal to stay in 32 bit floating calculations or whatsoever the original audio file format is?


Thanks a lot!

[HarBal]

This bug should be addressed in the next release of HarBal. I have not fixed date for that yet though.

You are right to be concerned about the possible affects of this conversion but unless you are working on sounds recorded at an increadibly low level it is highly unlikely that you'll notice any audible degradation. If you've recorded from analog sources (that is anything not synthesized or anything via analog inputs) then the noise from the signal sources are likely to be at least 10-20dB above that added by the conversion dither.

Regards,


Paavo.