How to threat values after ConvertAudioToFloat

[dimzon]

What is "Float Audio"? What is acceptable range for it?

[IanB]

Audio data stored a set of IEEE 32 bit floating point values instead of 16 bit signed integers.

In Avisynth the conversion from 16 bit to float isF32=I16/32768.0and from float to 16 bit isI16=Saturate_int16(F32*32768.0)The value 1.0 is normally consider 0dbm.

The range is the normal range of an IEEE floating point number, i.e. approx 1.176e-38 to 3.4e+38 a dynamic range of some 1500db (note the range of 32bit int is aprox 186db, 24bit int is approx 138db, 16bit is 90db and 8bit int is 42db)

[sh0dan]

Exactly! To put it simple:

* Range is scaled to -1.0 to 1.0
* Values outside the range are perfectly valid, but will be clipped when converting to 8/16/24/32 bit ints for output.