[mythtv-users] E: nvidia new drivers hdtv/interlacing, any benefit for us non HDTV people?

Bruce Markey bjm at lvcm.com
Mon Jul 5 16:17:31 EDT 2004


Chris Vargas wrote:
> --- bjm wrote:
...
> OK, so the question then becomes: is it possible _via
> software_ to get nVidia tv-out to NOT do this scaling?
> Or is this a hardware limitation of nVidia's tv-out? 

AFAIK, there is not a switch to tell it to align X11 lines with
TV scanlines. However, it seems logical that some overscan value
would project the display onto 480 lines. There is no way to know
for a fact if it is truly aligned and no way to know if there is
still some blending or processing that may be going on.

Try setting "TVOverScan" to "0.979". In XF86Config-4 set the
display mode to "640x480". In mythfrontend Setup->Appearance->
Screen settings, uncheck "Use GUI size for TV playback". In
TV Settings Playback->Overscan set "Vertical over/underscan
percentage" to "0".

I would have thought that TVOverScan of 1.0 would be the maximum
of 480 lines. That doesn't appear to be the case, at least for
my fx5200. I tried 1.0 when TVOverScan was first available and
it is better than the default size but there was still some stair-
stepping and jaggedness so I gave up and went back to 800x600 with
overscan set to fit my screen.

About two weeks ago I spent some time restarting X over and over
with different TVOverScan values. I looked at the tops and
bottoms of circles and diagonal white lines for stair-stepping,
horizontal parallel lines to see if they varied in thickness
and lattice patterns for flicker. I found that it was better at
something less than 1.0 and after some divide and conquer narrowed
down that 0.979 was as smooth as I could find.

I believe that this is pretty close to output on 480 lines but
there is no way I can prove this and no way to verify it. Still,
the top of a circle is about as smooth as a direct TV signal and
a thrown baseball appears to be one round object in motion rather
than two ghosts flickering along.

I can't say if the TVOverScan value is different for different
chips or different cards and I'd guess it would be different
for PAL but this is what is as close to a one to one alignment
that I can find.

...
>>Placebo effect. 
> 
> 
> I agree, with respect to the new drivers (at least
> from what I'm hearing). 
> 
> I was actually referring to the more basic ability of
> someone to detect whether they're getting correct
> interlaced source playback via tv-out, regardless of
> which nVidia drivers are in use. It's very obvious to
> me when it's not correct--I'm surprised it's less
> obvious to others.

Ya, I know what you mean. I think there are a lot of factors
like the size of the screen and viewing distance for starters
along with the relative analog resolution, kell factor, etc.
In my living room, I have a 70 inch Mitsubishi at ~7 feet and
I need to be very picky about lots sources of problems beyond
just digital recordings. However, in my bedroom I have a 27 inch
glass tube at ~11 feet and I don't bother with de-interlace or
display size tricks because these problems are near or below
the limits of visual acuity and just aren't noticeable.

http://www.ntsc-tv.com/ntsc-index-04.htm

>>>Again, at least with the PVR-x50, the fields are 
>>>definitely there and maintained. 
>>
>>Um, capture of the input signal? 
> 
> 
> Yes sorry, capture of the input signal. And it's good
> to know that the other capture cards preserve field
> info when capturing at 480 lines (capturing at 240, I
> imagine, tosses out every other line).

Right, and there is no point to choosing any height other than
480 or 240 (576 or 288 PAL) because it would record some lines
and skip others. Not only is this uneven but it sometime just
behaves badly with with odd heights.

http://www.mythtv.org/docs/mythtv-HOWTO-21.html#ss21.3

--  bjm


More information about the mythtv-users mailing list