Re: gsoc2013 octave audio system

[Mike Miller]

On Sun, Apr 21, 2013 at 3:21 PM, Vytautas Jancauskas wrote:
> Well, for one there is some disrepancy, I think, between what matlab
> calls an audio device and what portaudio calls an audio device. I
> don't have access to matlab right now but I'm pretty certain they mean
> physical devices while portaudio enumerates things like "pulse",
> "jack", "default", "system default", etc. as devices.

IMHO this kind of difference is ok. In either system, the user will
have to interpret what those devices are and choose an identifier to
use when selecting a non-default audio device. If we provide
identifiers for all physical audio devices as well as different
library interfaces to those devices, that's more choice which is a
good thing.

> "Bits per sample. Specify only when signal Y is represented by
> floating-point values. Valid values depend on the audio hardware
> installed: 8, 16, or 24."
>
> This bit of matlab documentation cannot possibly be right can it? It
> certainly should be "Specify only when signal Y is represented by
> integer values"?

I'm not certain, but I guess they are speaking about the number of
bits the DAC can support on the physical sound card. So if the input
signal is an integer type, the number of bits are explicitly bound to
the type (int8, int16, etc). If the signal is a floating point type,
the user would have to tell the library how many bits to quantize to
when writing to the DAC. Same thing in reverse when reading from the
ADC.

For playing to the sound card, our implementation could do the
quantization in software before sending it to portaudio, right? For
recording it may not be as useful, since the getaudiodata function
defines how values are to be converted when being read into a matrix.

-- 
mike