[mythtv-users] What Resolution Are You Using?

Dan Hopper ku4nf at austin.rr.com
Tue Apr 22 05:47:00 UTC 2003


> Kind of but it depends on what you mean by de-interlacing.
> Both fields are written into a framebuffer with no
> deinterlace algorithms applied. There is code to fix

Oh, OK.  Good to know.

> Two problems. First, MythTV can be shown on a progressive
> scan monitor with refresh rates other than 60 and even if
> it is a TV, you might use 800x600 (or 1024x768) or have an
> overscan offset or record with a height other than 480, etc.
> All of these will cause the interlace fields to not line up.
> Commercial DVRs have the advantage of knowing that they will
> only be drawing a true NTSC raster.

True, but I expect most folks without expensive progressive TV
monitors will be doing the NTSC in -> NTSC out case, so it'd be nice
if this could be optimized to do a one-to-one kind of deal with
respect to interlaced scan lines and such.  Ideally, you'd just
overlay the menus on a signal that was otherwise unaltered with
respect to the input (well, ignoring compression artifacts).

> Second, AFAIK, there is no access to the vertical refresh
> timing of the graphics card so you can't know when to update
> the framebuffer to be sure it will be drawn on the next pass.
> If there is access to this information, I'd really like to
> fix the playback jitter due to framebuffer updates not
> accurately spaced with vertical refresh. This is what 'jitter
> reduction' tries to mitigate but it can't truly be fixed
> without knowing when the card is going to draw the next frame.

I don't know how to do it, but I do know that I've been able to get
my Timex Datalink watch working under Win2k on VMware under XF86 4.1
on Linux.  The Datalink screen download function relies on
synchronizing with the vertical refresh I believe in order to obtain
the proper timing to talk to the watch.  It throws the screen into a
60 Hz refresh mode and spaces lines to indicate bytes to download to
the watch.  I believe if it doesn't update the framebuffer properly
with respect to the vertical refresh the data will be corrupted.

The point being, this is done under XFree86 so somehow VMware is
getting the synchronization information back to the Windoze app
while XF86 is driving the screen.  It would seem _some_ sort of
mechanism exists to get the information.  It'd be interesting to
know how.

Dan


More information about the mythtv-users mailing list