Mailing List Archive: MythTV: Mythtvnz

0.25 video frame jitter.

Index | Next | Previous | View Threaded

mythicalbeast at slingshot

Apr 4, 2012, 3:22 AM

Post #1 of 7 (537 views)
Permalink

0.25 video frame jitter.

Does anyone using 0.25 (build couple weeks old) have any issues with
very high frame jitter.
The frame jitter is reported in stdout logs & in the OSD playback
status.

VDPAU & GT240 core2duo..
The frame jitter occurs with h264 i50 material & only at x1 playback
speed. Time stretch playback other than x1 was okay.
Selecting different audio tracks/channels made no difference.
Non vdpau playback was not possible with 0.25 (10fps) so could not
compare.
VLC 1.0.6 can play same material with yadifx1 on 50% CPU..

I note that for the first time 0.25 can play the Samsung 40Mbps HD demo
(30p) with no problems (no jitter).
It also plays hi bitrate BDs (24p) okay.

Thanks,
Brett.

_______________________________________________
mythtvnz mailing list
mythtvnz [at] lists
http://lists.ourshack.com/mailman/listinfo/mythtvnz
Archives http://www.gossamer-threads.com/lists/mythtv/mythtvnz/

nick.rout at gmail

Apr 4, 2012, 12:42 PM

Post #2 of 7 (513 views)
Permalink

Re: 0.25 video frame jitter. [In reply to]

On Wed, Apr 4, 2012 at 10:22 PM, Brett <mythicalbeast [at] slingshot> wrote:
> Does anyone using 0.25 (build couple weeks old) have any issues with
> very high frame jitter.
> The frame jitter is reported in stdout logs & in the OSD playback
> status.

Might pay to actually post the error messages.

Would definitely pay to update, 2 weeks is old at the moment, 0.25 is
close to release and there are a lot of bugs being worked on.

>
> VDPAU & GT240 core2duo..
> The frame jitter occurs with h264 i50 material & only at x1 playback
> speed. Time stretch playback other than x1 was okay.
> Selecting different audio tracks/channels made no difference.
> Non vdpau playback was not possible with 0.25 (10fps) so could not
> compare.
> VLC 1.0.6 can play same material with yadifx1 on 50% CPU..
>
> I note that for the first time 0.25 can play the Samsung 40Mbps HD demo
> (30p) with no problems (no jitter).
> It also plays hi bitrate BDs (24p) okay.
>
> Thanks,
> Brett.
>
>
> _______________________________________________
> mythtvnz mailing list
> mythtvnz [at] lists
> http://lists.ourshack.com/mailman/listinfo/mythtvnz
> Archives http://www.gossamer-threads.com/lists/mythtv/mythtvnz/

_______________________________________________
mythtvnz mailing list
mythtvnz [at] lists
http://lists.ourshack.com/mailman/listinfo/mythtvnz
Archives http://www.gossamer-threads.com/lists/mythtv/mythtvnz/

stevehodge at gmail

Apr 4, 2012, 5:47 PM

Post #3 of 7 (509 views)
Permalink

Re: 0.25 video frame jitter. [In reply to]

On Wed, Apr 4, 2012 at 22:22, Brett <mythicalbeast [at] slingshot> wrote:

> Does anyone using 0.25 (build couple weeks old) have any issues with
> very high frame jitter.
> The frame jitter is reported in stdout logs & in the OSD playback
> status.
>
> VDPAU & GT240 core2duo..
> The frame jitter occurs with h264 i50 material & only at x1 playback
> speed. Time stretch playback other than x1 was okay.
>

It sounds like something to do with deinterlacing - which deinterlacer have
you got selected?

Cheers,
Steve

mythicalbeast at slingshot

Apr 4, 2012, 11:11 PM

Post #4 of 7 (510 views)
Permalink

Re: 0.25 video frame jitter. [In reply to]

On Thu, 2012-04-05 at 12:47 +1200, Steve Hodge wrote:
> On Wed, Apr 4, 2012 at 22:22, Brett <mythicalbeast [at] slingshot>
> wrote:
> Does anyone using 0.25 (build couple weeks old) have any
> issues with
> very high frame jitter.
> The frame jitter is reported in stdout logs & in the OSD
> playback
> status.
>
> VDPAU & GT240 core2duo..
> The frame jitter occurs with h264 i50 material & only at x1
> playback
> speed. Time stretch playback other than x1 was okay.
>
> It sounds like something to do with deinterlacing - which deinterlacer
> have you got selected?
>
> Cheers,
> Steve
>
Normally I use VA temporal-spatial x2 VDPAU with hqscaling & sharpen.

I tried all the de-interlacers including none, with no improvement.
The de-interlacer still functions in time stretch playback.
Time stretch playback (not x1) is very watch-able..

There are no errors in playback logs as such..just very high reported
standard deviation & low PQ ..

I have been following the mailing lists for other reports.

Thanks,
Brett

_______________________________________________
mythtvnz mailing list
mythtvnz [at] lists
http://lists.ourshack.com/mailman/listinfo/mythtvnz
Archives http://www.gossamer-threads.com/lists/mythtv/mythtvnz/

mythicalbeast at slingshot

Apr 7, 2012, 3:19 AM

Post #5 of 7 (508 views)
Permalink

Re: 0.25 video frame jitter. [In reply to]

Maybe this was obvious..
The frame jitter I was experiencing was reduced/eliminated by increasing
the alsa buffer /proc/asound/card0/pcm1p/sub0/prealloc.
I have to do this manually.

Using default value of 64:
This value does not cause jitter problems with all recordings.
High data rate recordings, with AC3 only, play well by IEC958 pass-thru'
e.g. Samsung HD demo h264 30p 40Mbps AC3 448Kbps plays perfectly as
pass-thru' but jitter increases/deteriorates when using decoded PCM
output.
All my high data rate recordings are progressive (sadly).

Interlaced 50Hz material has double frame jitter of interlaced 60Hz.
(using avsforums 'cheese-slices' test tracks AC3 audio)
Both play with low jitter when using higher alsa buffer size.

There are no errors/warnings in playback logging.

Summary:
- interlaced recordings with only AC3 audio (& iec958 pass thru') have
high jitter with small alsa buffer.
- progressive recordings with only AC3 audio (& using pass-thru') have
low jitter with small alsa buffer.
- i60 plays with half frame jitter of i50 with small alsa buffer.
- increased alsa buffer eliminates the jitter problem.
- same alsa buffer size in 0.24+fixes does not demonstrate jitter.

Note:
0.24+fixes can not play the Samsung HD demo.
Thanks..

_______________________________________________
mythtvnz mailing list
mythtvnz [at] lists
http://lists.ourshack.com/mailman/listinfo/mythtvnz
Archives http://www.gossamer-threads.com/lists/mythtv/mythtvnz/

jyavenard at gmail

Apr 7, 2012, 10:02 PM

Post #6 of 7 (485 views)
Permalink

Re: 0.25 video frame jitter. [In reply to]

On 7 April 2012 20:19, Brett <mythicalbeast [at] slingshot> wrote:
> Using default value of 64:
> This value does not cause jitter problems with all recordings.
> High data rate recordings, with AC3 only, play well by IEC958 pass-thru'
> e.g. Samsung HD demo h264 30p 40Mbps AC3 448Kbps plays perfectly as
> pass-thru' but jitter increases/deteriorates when using decoded PCM
> output.

64kB is only good enough for stereo output (that includes AC3 and DTS
which fundamentally are 44.1-48kHz stereo PCM streams)

More than that, in particular LPCM multichannels stream and you will
get buffer underruns.

Ubuntu sets the default alsa buffer to 4MB.
When using TrueHD or DTS-HD MA, which are 16 bits, 8 channels, 192kHz
PCM. You'll need close to 4MB.
So setting 4MB alsa buffer is actually a value that makes sense and
that will work across the spectrum of content you could play

_______________________________________________
mythtvnz mailing list
mythtvnz [at] lists
http://lists.ourshack.com/mailman/listinfo/mythtvnz
Archives http://www.gossamer-threads.com/lists/mythtv/mythtvnz/

mythicalbeast at slingshot

Apr 8, 2012, 4:25 PM

Post #7 of 7 (476 views)
Permalink

Re: 0.25 video frame jitter. [In reply to]

>
> 64kB is only good enough for stereo output (that includes AC3 and DTS
> which fundamentally are 44.1-48kHz stereo PCM streams)
>
> More than that, in particular LPCM multichannels stream and you will
> get buffer underruns.
>
> Ubuntu sets the default alsa buffer to 4MB.
> When using TrueHD or DTS-HD MA, which are 16 bits, 8 channels, 192kHz
> PCM. You'll need close to 4MB.
> So setting 4MB alsa buffer is actually a value that makes sense and
> that will work across the spectrum of content you could play
>
My lucid/alsa 1.0.24.2 (user libs) defaults to 64.
Mythtv complains it requires 128..
But I only use 48KHz AC3 & PCM with Mythtv.

I (like many) ignored the alsa buffer errors reported because they
seemed to have no consequence with 0.24.

The real point was that 0.25 is badly affected by alsa buffer problem.
This issue was good motivation to fix (hack) the alsa /proc/asound
permissions/ownership.

What is the proper fix/soln to alsa /proc/asound ?:
- alsa code mods
- udev rule
- module options (possible in the past)
My solution was none of these, just chown/chmod in /etc/rc.local.

The frame jitter standard deviation with 0.24 is in the range 17-40.
The best jitter stddev with 0.25 is in range 80 - 400, these values do
not cause any visible impact (PQ).
These measurements are on the same machine & recordings & 0.24+fixes
backend.

Thanks,
Brett

_______________________________________________
mythtvnz mailing list
mythtvnz [at] lists
http://lists.ourshack.com/mailman/listinfo/mythtvnz
Archives http://www.gossamer-threads.com/lists/mythtv/mythtvnz/

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads