[gst-devel] audio/raw vs float

Owen Fraser-Green owen at discobabe.net
Mon Mar 19 15:05:19 CET 2001


Zaheer wrote:

> Why don't we rename audio/raw to audio/int and also have audio/float and
> audio/fixed (and whatever else).

I think this sounds like a logical approach. As I wrote a couple of weeks 
ago I don't like the idea of audio/raw being int-centric and the 
name "audio/raw" implies that it is the most fundamental audio format so it 
seems silly for that format to come in so many flavours.

> > On Mon, 19 Mar 2001, Steve Baker wrote:
> > 
> > > So, what if we create a new mime type audio/float with these 
> > > format - an enum including standard 32 bit and 64 bit formats.  New 
> > > can be included in future releases.
> > > lower-bound - usually -1.0 (sometimes 0.0)
> > > upper-bound - usually 1.0
> > > rate - The sample rate of the data, in samples per second.
> > > channels - The number of channels of audio data.
> > >

I can understand the advantage in allowing the sample depth to be variable 
in audio/int because it affords the simplest implementation when the depth 
required by the audio hardware matches the audio source. For example, when 
a 16-bit MP3 is played is played through a 16-bit soundcard then everything 
can happen in 16-bit without any conversion. However, when "fictitous" 
formats such as floats and fixeds are concerned is there any advantage in 
cateing for different flovours of these, i.e. why not standardize on -1.0 
to 1.0 32-bit? Is there (or will there ever be) any application which would 
need 64-bit?

Also, is there any need for float and fixed formats to operate on anything 
more than single channels? If so then can't they just use seperate pads for 
the seperate channels? I guess I feel that the float/fixed format would be, 
much more than int, an abstraction of an analogue audio stream and, 
therefore, each audio channel should be treated as a seperate stream.


     Owen  Fraser-Green             "Hard work never killed anyone,
     owen at discobabe.net              but why give it a chance?"

More information about the gstreamer-devel mailing list