On May 2, 2006, at 9:00 PM, Brad Midgley wrote:
> Apparently the bluetooth sco audio pcm i/o is wired to nssp.
> Can I configure nssp to deal properly with pcm and just present me
> a bytestream for our audio?
That was certainly the theory that I was operating under when I asked
gordon to wire it that way. The PBA31307 has a 4-wire physical
connection for its PCM audio. One line for PCM out, one for PCM in,
then one for "PCM clock" and one for "PCM frame". The docs made it
look at a glance like it was nicely matched to the 4-wire interface
that the NSSP provides, where the data which would be transfered over
the NSSP would simply be the PCM data. Given the way that the PXA
UARTs work, I figured that you'd basically just be able to either
read/write PCM data directly from/to some register address after
setting up the NSSP clock/frame details appropriately, or spend a bit
more time and even do DMA or something fancy.
Looking at the docs for the PBA, it looks like it ought to be pretty
straightforward to do this; and while I can't actually copy/paste the
relevant bits from the doc due to NDA issues, I should probably be
able to hack a quick sample driver together to be able to read/write
PCM audio in 16-bit linear PCM or 8-bit A-law or u-law log encoded
audio to/from the PBA. It looks like it should be *relatively*
straightforward to do that. Once the basic code is written, then it
should be pretty easy to continue/extend the work based on the
(heavily commented) source code that I'll stick in SVN.
I think I've got most of the major issues I've been cranking on off
my plate within the next day or two, which will free up some time to
take a look at this.
>> Can I configure nssp to deal properly with pcm and just present me with
>> a bytestream for our audio?
> I think I've got most of the major issues I've been cranking on off my
> plate within the next day or two, which will free up some time to take
> a look at this.
It wouldn't be as generally useful to leave out the cpu, but would it be
straightforward for me to have the cpu leave nssp and its pcm lines in
hi-z and physically jumper the pba31307 over to the ucb1400?