[mythtv] Problems with Bob Playback

Kevin Cody Jr. kcodyjr at verizon.net
Sun Jan 15 23:01:28 UTC 2006

I'm still experimenting with this, so YMMV, but...

1.) If you're outputting to an interlaced mode, you shouldn't be doing 
-anything- to the signal: match the monitor resolution to the capture 
resolution, do not deinterlace by any means, and more importantly:

2.) Make sure your vertical refresh is synced to the source stream. 
You're in PAL, so you need a 50Hz refresh rate - no getting around it.

With this approach, you shouldn't see any motion or scaling artifacts at 
all. You will probably see the spaces between lines at 720x576, and in 
an interlaced mode you'll definitely see the odd line phosphors fade 
before the even lines are drawn, resulting in a twitchy picture. 
Televisions have much slower phosphors.

Alternatively, you can process the signal to improve visible image 
quality. YMMV, but try the following (only applies to CRT via VGA/DVI):

1.) Set your display to double the capture resolution, both directions.
2.) Use linear or kernel deinterlacing and progressive output.
3.) Set the vertical refresh to an even multiple of the source rate - 
for PAL, any multiple of 25 from 75Hz up will do. NTSC will want 120Hz 
if possible, 90Hz if not.

Allowing a noneven multiple refresh rate guarantees motion jitter. This 
is one place where NTSC is actually more useful than PAL, for there is 
no practical refresh rate that can be divided both by 25Hz (video) and 
24Hz (film). 120Hz, however, divides cleanly by both 24 and 30.

IMO it's better for NTSC users to pick 90Hz at 2x resolution over 120Hz 
at only 480 output lines; I can still see the spaces between lines even 
in progressive mode regardless of refresh rate. Older monitors, and more 
expensive recent monitors, might not have this problem.

Michael T. Dean wrote:

>Martin Ebourne wrote:
>>On Sat, 14 Jan 2006 21:12:05 -0500, Michael T. Dean wrote:
>>>Martin Ebourne wrote:
>>>>Anything other than timestretch 1.0 works great, just at 1.0 it is
>>>If using timestretch, bob deinterlacing is disabled, so that would explain
>>>why you don't see the problem with timestretch.  :)
>>Yes, that is certainly implicated. But then something else is still going
>>on because with bob on and timestretch other than 1.0 the picture is
>>clearly being deinterlaced somehow - with bob off completely (from the
>>settings page) scrolling text has a comb effect at any timestretch, but
>>with bob on and timestretch in action there is no comb.
>>Actually, looking at the picture some more I think I can see what's
>>happening. With bob on and timestretch active the vertical resolution
>>appears to be halved, so I guess it's doing one-field deinterlacing
>>instead. That would explain the lack of shake.
>Yes.  If you have enabled bob deinterlacing, you've told Myth that you 
>want deinterlacing, however bob only works at certain multiples of the 
>frame rate, so Myth falls back to one-field deint when you turn on 
>mythtv-dev mailing list
>mythtv-dev at mythtv.org

More information about the mythtv-dev mailing list