Bad A2DP Performance

[Steven **]

I don't get nearly as bad performance as the OP.  But it does make a
difference how close I hold the Freerunner to my headset.  With it
just a few inches away, I get a glitch in the sound every 10 - 20
seconds.  If I set it on the table in front of me about 2.5 - 3 feet
from the headset, I get constant glitches.  I also noticed what seem
like dropped packets or sync issues.  The song would just ahead by a
second or so every once in a while.  No clue what causes that.

Here are my versions:
bluez-utils - 3.23-r4 -
bluez-utils-alsa - 3.23-r2 -
bluez-utils-compat - 3.23-r2 -

I checked top while I was playing.  madplay was at about 18% usage and
aplay took another 12%.  If I just used aplay with a wav, it uses
about 20%.

Is gstreamer what openmoko-mediaplayer uses?  Cause that wouldn't even
play sound.  Not sure what that is doing (obviously my goal would be
for the mediaplayer to play to the BT headset).

I noticed in the ticket a comment about trying it at 16 KHz sample
rate.  So, I played the same song with that sample rate.  It performed
much better.  Of course, it sounds like crap...  But I only got a few
stutters/glitches.  It was fairly bearable.  I actually listened to
the whole song and I noticed the following output:
Playing WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Stereo
6746 frames decoded (0:02:56.2), +0.6 dB peak amplitude, 36 clipped samples

If I play the same clip at the same distance, but at the original 44.1
KHz, it's unbearably glitchy.  I get:
Playing WAVE 'stdin' : Signed 16 bit Little Endian, Rate 44100 Hz, Stereo
6746 frames decoded (0:02:56.2), +0.6 dB peak amplitude, 187 clipped samples

-Steven

On Tue, Aug 5, 2008 at 11:17 PM, Brad Midgley <bmidgley at gmail.com> wrote:
> Daniel
>
> On Tue, Aug 5, 2008 at 1:03 PM, Daniel Benoy <daniel at benoy.name> wrote:
>> I reported a problem to the openmoko bug tracker a little while ago and I haven't received any feedback yet.  I'm hoping I will get more feedback here.
>
> there are two sources of possible trouble for quality, cpu load and
> timing of bluetooth frames.
>
> The cpu is not likely the problem on neo. When driving it through the
> alsa driver, the neo1973 had really good a2dp performance even on
> older versions of bluez-utils, much better than the TI cpus (eg in the
> n800). Did you check the cpu load while playing? I don't have a
> freerunner, but the cpu is supposed to only be an improvement over the
> neo1973.
>
> btw, what version of bluez-utils is in the system?
>
> an aside, the gstreamer plugin has had a performance issue that burns
> a lot of cpu. I never did isolate it to either the encoder or
> transmission components. It isn't likely you're using the gstreamer
> route unless you specifically set out to do it that way.
>
> --
> Brad
>
> _______________________________________________
> Openmoko community mailing list
> community at lists.openmoko.org
> http://lists.openmoko.org/mailman/listinfo/community
>