[mythtv-users] Caveats in upgrading from Ubuntu 18.04 + MythTV 29.1 to something newer?

Stephen Worthington stephen_agent at jsw.gen.nz
Tue Apr 25 16:27:45 UTC 2023


On Tue, 25 Apr 2023 10:24:28 -0400, you wrote:

>On 4/24/23 11:59 PM, Stephen Worthington wrote:
>> I find this interesting, as I am very sensitive to deinterlacing
>> problems and have not had this happen to me.  I have been using both
>> VDPAU and NVDEC at times on the 525 drivers and can see no difference
>> between them, except for the NVDEC drivers supporting more 4K files
>> than VDPAU, and NVDEC not supporting some older video files.  I do not
>> remember any difference in the deinterlacing as I moved through the
>> driver versions from 390 upwards.  I am using Nvidia GT1030 cards.
>
>Yeah, I can't rule out that it may be hardware-specific.  In my case 
>it's a GeForce GT 730.  I happened to have a horizontally-panning scene 
>on the TV at the time, and it was distractingly flickering.

Since a GT730 is obsolete (no longer sold), I doubt that the latest
drivers get much testing on it.  So it is entirely possible that they
are broken with respect to a GT730.  So sticking to an older driver
version that was around when they were still being sold is likely a
good idea, as the new things in the newer drivers tend to be only for
the new hardware anyway.  But then eventually you will run into the
problem that the older drivers will stop being supported and will then
stop working with the newer kernels.

Juddering during panning is pretty sure sign the deinterlacing is not
working properly, but you can get that when using a very crude
deinterlacing algorithm instead of the best ones.  Which could be a
configuration problem.  If you want to check the deinterlacing
actually being used for playback, start playing a problem recording
and then do M(enu) > Playback > Playback Data.  The information
displayed at the top of the picture includes a line right at the
bottom in the middle titled "Deint : " which tells you exactly what is
being used.  For VDPAU it should say "2 x VDPAU advanced" and for
NVDEC it should be "2 x NVDEC adaptive".

Here in New Zealand, it used to be common on the Choice and HGTV
channels to get recordings with judder during panning, due to the
setup being used for converting ATSC 29.95 Hz to DVB-T 25 Hz not being
correct.  They finally fixed that only after Discovery bought those
channels from the previous owners a couple of years ago.  When I first
met that, it took me quite a while to discover that no amount of
fiddling with my deinterlace settings was ever going to be able to fix
it.

>Given that the broken-for-reasons-I-didn't-get-around-to-figuring-out 
>6200ch channel changer was the more-pressing deal breaker, I'll admit 
>that I didn't spend a whole lot of time poking around with the 
>deinterlacing issues.  The fact that manually running 6200ch worked fine 
>from a shell, but not as expectd when invoked from MythTV sure was a 
>head scratcher for me.  Could be as simple as a race condition between 
>the channel changing being sent out over firewire and the HD-PVR 
>starting to record.  I did notice that the HD-PVR started recording 
>before 6200ch actually completes (tested via adding e.g. a 10-second 
>sleep to the channel changer).  That's about the point where the time I 
>had sunk into looking at it had exceeded the value of being on a recent 
>Ubuntu version.  :)

Race conditions are always nasty to deal with.  In this case, there
has been a recent bug fix in v33 that fixed the problem that
mythbackend was sometimes generating the pre-recording events at the
wrong times.  It is possible that the same bug was affecting when the
channel change command was being run.  If so, that might explain your
problem.  It would pay to ask the devs exactly when the channel change
command is supposed to be being run, and check that it is actually
happening then and not later.  If it is being run at the correct time,
you could consider running 6200ch from the pre-recording events (now
that they work properly), as there is a setting that can be used to
tell mythbackend how much before the recording start it should do the
first pre-recording event.

>But after reverting back to my backup that had the 390 Nvidia drivers, 
>it's back to smooth / non-flickering / no evidence of having been 
>interlaced.

Yes, that is vital to those of us who are bothered by it.  My mother
does not notice such things.


More information about the mythtv-users mailing list