Re: saa7134 field order fix is not a fix.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]



On Tue, 2003-05-06 at 06:19, Tom Zoerner wrote:
> Richard Jones wrote:
> > CNN text crawler has classic field order jitter. Lips on speakers 
> > likewise. Fast movement also shows bad jitter. Scene changes appear
> > to flick back briefly to the previous scene during scene change.
> 
> Couldn't that have different causes too, like wrong frame processing
> sequence? (Since you say you're using code you modified yourself)
> 

Possibly but what I'm doing is pretty straight forward , I pass
the frame capture from the tv card and sending it straight to
the tv-out on the matrox g400. It is possible that the g400 has got
the field order flipped itself.

> If you do receive a network with teletext I think that would offer
> a much more reliable method for testing field sequence order. Get
> libzvbi from http://zapping.sourceforge.net/ and execute in the
> test directory: ./capture --dump-ttx; check the output if
> teletext pages continue across field borders, e.g.:
> 
>    WST 3 13 022 >.8 Schwiegersohn Daniel bietet ihr begeis-<
>    WST 3 14 319 >^/ tert eine Stelle in seiner Anlagebera- <
> 
> Here you see packet 13 in line 22/field A and packet 14 of the same
> teletext page (since the word "begeistert" is continued) in line
> 319/field B.  If the DMA bits were used wrongly you'd see the 2nd half
> of the previous frame's teletext page in field B.  Note that teletext
> pages are allowed to have missing packets and may not always cross
> fields, so you have to look at a larger number of pages to draw a
> conclusion.
> 

Ok with "if ((status & 0x10) == 0x10)" I get the following:
WST 1 12 323 >.8Muslim Lawyers says it's obtained power <
WST 1 13 007 >G8of attorney from the three men - Amrozi,<

and
WST 1 16 011 >.PSaturday night's home clash with        <
WST 1 17 319 >GPNational Rugby League pacesetter        <



and with  "if ((status & 0x10) == 0)"
WST 1 12 011 >.8Muslim Lawyers says it's obtained power <
WST 1 13 319 >G8of attorney from the three men - Amrozi,<

and
WST 1 12 323 >.8 1968, you must not modify,copy,        <
WST 1 13 007 >G8 reproduce,republish,frame,upload to a  <


Now I don't know much about VBI, how do you tell which field 
the line came from? On both DMA bit tests I get packets between
the two lines I've shown, I'm still not sure how to use the teletext
output to determine the field order as both cases seemed to
produce output in correct order when reading the dump (I'm not sure if
the line numbers make sense though), can you perhaps send (off list if
its too large) a capture showing both correct and incorrect field order.


> 
> > if ((status & 0x10) != 0x10)
>  > if ((status & 0x10) == 0x00))
>  > which isn't exactly equivalent.
> 
> Why not?  The result of (status & 0x10) can have two values: 0 or 0x10,
> since all other bits are masked out (set to zero).  Hence if the result
> is != 0x10 it must be == 0 so both checks are logically equivalent.
> 

Sorry stupid comment, proof I should think before typing.

> > (...) but given that you mentioned the change conflicted with vendor 
> > specified behaviour
> 
> Well, the same paragraph contains other inconsistencies, e.g. talking
> about DONE_RA1..RA4 while at other places it's RA0..RA3.  Also OE isn't
> defined in the same chapter, but 2 pages earlier for a different
> register...  unfortunately there's always some room for interpretation
> with specs like these.
> 

Do Phillips have the spec for download? I had a quick look on their web page
and was unable to find it.

Richard.





[Index of Archives]     [Linux DVB]     [Video Disk Recorder]     [Asterisk]     [Photo]     [DCCP]     [Netdev]     [Xorg]     [Util Linux NG]     [Xfree86]     [Free Photo Albums]     [Fedora Users]     [Fedora Women]     [ALSA Users]     [ALSA Devel]     [Linux USB]

Powered by Linux