Hello list, I need some help.
When running GRC 3.6.1 I am getting a problem with occasional choppiness
in
my audio output. Details:
- The choppiness sounds like only part of the buffer is filling, similar
to
the sound you get when a filmstrip’s playback gets unstable and you get
that ‘jitter’ - The issue usually does not immediately occur however occurs a few
minutes
into use. - Sometimes computer activity such as un-minimizing a window will
trigger
the problem. Sometimes the same will fix it. - Input is Funcube Dongle Pro+ 192khz on alsa hw:3
- Occurs regardless of using FCDPP+ Block or the Audio Input block.
- Output is pulse, originally set to 192khz (pulse resamples it),
brought
it down to native 48khz and used Rational Resampler block to bring the
192
to 48. Did not resolve, however the issue occurred less often. (still
intolerable) - Turned on Realtime Scheduling and then adapted the JACK instructions
for
enabling system realtime scheduling permissions (
http://jackaudio.org/linux_rt_config). Afterward confirmed the python
process to have a nice of -30 which is realtime. Issue did not resolve
but
again, it happened less often. (still intolerable) - Occurs regardless of graph complexity. Simple in->out arrangement will
reproduce issue. - Sometimes you can hear it get worse as time goes by, and sometimes it
will just as randomly fix itself, only to occur again later. - I am using pulse as my audio output.
- When the issue occurs, the console output in GRC fills rapidly with
“aUaU” - When the issue occurs, looking into pavucontrol playback tab shows
“ALSA
plug-in [python 2.7]” entry flickering rapidly, as if it is resetting
its
connection over and over in an unreasonable manner. - When the issue occurs, the realtime python process on core 2 shows no
change in core usage, however the non-realtime gnuradio-companion
process,
also on core 2, spikes from its usual 0.9% usage to 93%. The only way I
can
see a high-priority process being pushed out by a low-priority process
is
if the high-priority process is waiting on the low-priority process for
some operation to complete (this is a DSP no-no). - When the issue occurs, pulseaudio’s verbose log fills rapidly.
attached
is a snippet of what repeatedly fills the log. - There is chatter about similar underrun issues with mozilla and pulse.
However the fixes appear to be occurring in mozilla only so there wasn’t
much help from that.
(779392 - Improve the ALSA PulseAudio bug workaround used to fix bug 761274) - When attempting to specify ‘pulse’ as signal input in GRC and then
specify FCDPP as its input in pavucontrol, GRC gives silence and a
stream
of “aU” - pulseaudio 4.0
- Xubuntu 13.10 AMD64.
There was a similar issue posted earlier however it was different in
nature
in the sense that the underruns occurred immediately instead of on
occasion, implying a consistent sample rate mismatch.
https://lists.gnu.org/archive/html/discuss-gnuradio/2013-08/msg00516.html
It is reasonable to rule out sample-rate mismatch in my case as the
issue
occurs only on occasion and sometimes fixes on its own, a mismatch would
be
expected to immediately fail. It would be mathematically impossible for
a
mismatched sample rate to cause such cascading underruns after minutes
of
proper functioning except for clock drift, which would be expected to be
solved with a single buffer reset (a slight “blip”). This is not what is
happening. It is thousands of consecutive blips after a while of proper
functioning.
Thanks for reading. Log dump attached.
