[mythtv-users] What keeps causing "prebuffering pauses" ?!?!?!

Hamish Moffatt hamish at cloud.net.au
Sun Feb 1 02:17:27 EST 2004


On Fri, Jan 30, 2004 at 11:33:09AM -0500, Andrew Dodd wrote:
> Quoting Hamish Moffatt <hamish at cloud.net.au>:
> > Any resolution for this? I'm getting it sometimes on my DVB-T setup.
> > Switch to a particular channel and it occurs non-stop, switch to another
> > and it's fine, go back and it starts again. One channel is the most
> > affected.
> Possible causes:
> 
> XvMC and 16bpp are a Bad Combination (Sometimes X config utilities will drop you
> into 16bpp without you realizing it.  This happened to me.  I set my box up for
> 24 bpp default, at some point the default changed to 16, I have NO IDEA how.)

I'm running 24 bit. I'm using the NVidia binary driver with XvMC, but I
haven't enabled XvMC in Myth because I read that it prevents the
deinterlacing filter from being used. Also I have plenty of CPU (Athlon
XP 2600+) and have never seen myth use very much - only 20% even for
high definition decoding.

> Enable "extra audio buffering" - This helped for me.
> 
> Finally, define "more than adequate performance".  Raw read/write speeds are
> irrelevant for Live TV, as the HD has to keep seeking back and forth between
> locations on the drive.  I had a relatively recent 120GB 7200RPM Maxtor drive
> that was *insufficient* for watching Live TV.  (Although other factors were that
> it was in an external Firewire case and running a FAT32 filesystem.)  Raw
> read/write transfer rate was 20 MB/sec.  It couldn't even handle one LiveTV
> stream without occasional prebuffering pauses.  In situations of intense reading
> and writing at the same time (example:  Saving the raw video stream from a
> recording using avidemux), the transfer rate would drop to below 5 MB/sec due to
> the amount of seeking that was being done.
> 
> My new Seagate 200GB running ext3 internally works great.  hdparm read tests
> approach 60 MB/sec.

I only get the prebuffering pause problem when switching to particular
channels. Sometimes I can switch to those OK, and sometimes not. When
the problems occurs I can switch to another which will be fine but the
problem will occur again when I change to the problem channel. If I quit
live TV and start again, the problem channel can be fine.

I have noted an additional problem where 2 of my channels get errors in
the backend log every 25 seconds exactly. It reports 

2004-02-01 18:07:42 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:07:42 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:07:42 DVB#0 WARNING - Transport Stream Continuity Error.  PID = 769
2004-02-01 18:07:42 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:07:42 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:07:43 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:07:43 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:07:43 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:07:43 DVB#0 WARNING - Transport Stream Continuity Error.  PID = 769
2004-02-01 18:08:08 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:08:08 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:08:08 DVB#0 WARNING - Uncorrectable error in packet, dropped.
2004-02-01 18:08:08 DVB#0 WARNING - Transport Stream Continuity Error.  PID = 769
2004-02-01 18:08:08 DVB#0 WARNING - Uncorrectable error in packet, dropped.
etc

As you can see, 25 seconds apart. Two local channels cause that, though
two others seem to be OK. (Channels 2 & 7 in Melbourne are bad, 9 & 10
seem to be ok. SBS has other problems here currently.)

My setup:

Athlon XP 2600+, 512Mb, nForce2 chipset
GF4 440MX AGP card (not on-board)
Linux 2.6.2-rc3 (for latest DVB drivers).
ALSA drivers from the kernel.
Myth 0.14 compiled without ALSA support (using Debian packages)
and without XvMC support.
60Gb MAXTOR IDE, 5400 RPM, 2Mb cache getting 40 Mb/sec on hdparm -t.
ext3fs.

I'm particularly mystified by the 25 second issue. Is it just running
out of disk bandwidth? It's hard to tell, as the log doesn't indicate
anything. Myth is only using about 0.7% of CPU according to top, with
XFree86 at about 2.3%. System is 75% idle.

I intend to replace the disk with something bigger before "deployment"
so if that's the limiter I'll just do it now.


Hamish
-- 
Hamish Moffatt VK3SB <hamish at debian.org> <hamish at cloud.net.au>


More information about the mythtv-users mailing list