[mythtv-users] pcHDTV with PVR-350 for output

[Dan Lanciani]

Daniel Thor Kristjansson <danielk at mrl.nyu.edu> wrote:

|On Sun, 24 Oct 2004, Dan Lanciani wrote:
|]Daniel Thor Kristjansson <danielk at mrl.nyu.edu> wrote:
|]If "very high" equates to about twice what we have now, it shouldn't be
|]long before this is not a problem.  In any case, it would be nice to have
|]at least *some* independence...
|
|Very high would be 10 to 20 times the CPUs of today if you wanted to 
|encode 1080i. Outputting 640x480 MPEG2 while decoding 1080i should be 
|doable in a few years.

I'll settle for decoding 480i/p and outputting 640x480 MPEG2. :)

|]|We can get fast-play working with HDTV, but it's not yet high priority.
|]|It's mostly a ffmpeg w/MythTV issue.
|]Ah, I didn't realize that this was not supposed to work.  Does any kind
|]of fast forward functionality work?  Rewind?
|
|Skip works. And fast forward works on very powerful processors.

In this context is my 3.2GHz P4 very powerful or are we still talking
about the future?

|]|(For instance, I can get my plain old TV to sync to a 
|]|800x600 signal from the FX5200.)
|]I'm not sure what this means.  Are you actually generating 600 (300 
|]interlaced) scan lines?  If so that's not real NTSC anymore.  Maybe you 
|
|Yes, it works with my specific TV. It should actually work with most TVs 
|with composite input, but I've never published my modeline because it 
|can destroy some older TV's.

I think you will find that the ability to sync to a signal that far from
the standard is not common, unless maybe your TV happens to be some sort
of combined multi-sync monitor.  Are you certain that the card is putting
out a 600/300 line signal rather than down converting?

|]can get a specific TV so sync to it, but it won't work as a general 
|]video source.  And this is my main gripe with the TV-out support on 
|]every video card I've ever tried.  You can tweak this and that to get 
|]it to almost look right on a particular TV, but the signal is so far 
|]out of spec that it's useless for anything else.
|
|Hmmm, well safe modeline adjustment could be added to MythTV. Right now 
|MythTV setup is admitedly a pain. There are some standalone programs out 
|there to generate modelines. One used to actually come with both RedHat 
|and Mandrake, but has been dumped because most computer monitors these 
|days tell X11 how to drive the monitor. But TV's don't and setting them 
|up is very much like setting up XFree86 was like 10 years ago.

But again, I really don't want to set the output up for a particular TV.
I want to generate NTSC standard video that I can feed into my distribution
system and watch on any TV (or even tape).  We have many existence proofs
that it is possible to generate an NTSC signal that is generally correct
without being TV-specific (e.g., my DTC-100 HDTV box, many DSS receivers,
DVD players, etc.)  Just the fact that we are talking about setting up for
a TV suggests that something is seriously lacking in TV-out support...

|]I haven't even noticed the ringing yet.  It is completely overshadowed by
|]the horrible motion artifacts.  Even as someone moves their hand across the
|]screen you can see lots of black scan line "checks" around the motion.  I
|]know it isn't a problem with the source stream because my ancient DTC-100
|]renders a perfect NTSC representation of the same program.  I don't know if
|]it is an MPEG decoding issue or an NTSC encoding problem, but it (IMHO) makes
|]the output useless.  (See my previous message about the FX5200.)
|
|Are you using XvMC?

Beats me.  I'm just using the defaults.

|Try XV if you are. This sounds like a bug. Does 
|it look good in mplayer?

Haven't tried it.

|Does it look good on your monitor?

I've been staring at this for a while now and I think I can see similar
artifacts on the monitor.  However, they are much less noticeable and I
may just be fooling myself that they are there at all.  I would not have
noticed them were I not comparing to the TV version.  On the other hand,
the monitor is a lot smaller than the TV.

The last time I saw artifacts similar to this I discovered that the signal
had its chroma and luma components offset by a frame or a field (it wasn't
clear which).  This seemed to upset the color decoder, especially when a
bright object moved "through" the chroma information of another part of
the scene.  In the current case the effect shows up well with dark moving
objects, so while it may also be a matter of an NTSC error confusing the
color decoder, it probably isn't the same error.


				Dan Lanciani
				ddl at danlan.*com