AVS Forum banner
Status
Not open for further replies.
1 - 11 of 11 Posts

·
Premium Member
Joined
·
8,100 Posts
Discussion Starter · #1 ·
OK,


Just when I thought I had all the theories covered, I've just gotten a bit more confused.


3:2 pulldown is obvious (why it's there, how it's done & how it's eliminated).


However 2:2 pulldown got me thinking...


Since the film is 24fps and PAL is 50hz (25fps), they speed up the film by 4% and each frame contains two fields from the same movie frame.


Wouldn't 2:2 pulldown be simply contructing a frame by joining those two fields together and simply disabling video mode? Doing this on non-2:2 pulldown material would, of course, produce the zipper effect, right?


Would a scaler with PAL support, but no 2:2 pulldown support simply recognize this as video source and try to eliminate a zipper effect that, in fact, wouldn't be there?!
 

·
Registered
Joined
·
145 Posts
Yes. It would do calculation and loose vertical resolution with no need to do it.
 

·
Registered
Joined
·
1,687 Posts
Ofer


Not quite.


2:2 detection is just as important as it is for 3:2.


OK


Interlaced input @ 50Hz



Odd1

Even1

Odd2

Even2

Odd3

Even3


Correct Progressive output @ 50Hz


Odd1Even1

Odd1Even1

Odd2Even2

Odd2Even2

Odd3Even3

Odd3Even3


Output with weave @ 50Hz


Odd1Even1

Odd2Even1

Odd2Even2

Odd3Even2

Odd3Even3

Odd4Even3


So in this case half the frames are correct and half have artefacts.


If you throw in video processing you get the same problem you get good frames half the time and the other half you get reduced resolution frames.


The difference is pretty noticable.


John
 

·
Premium Member
Joined
·
8,100 Posts
Discussion Starter · #5 ·
OK,


Obviously, there would be awful artifacts if what John's example is what non 2:2 pulldown scalers would do.


However, isn't matching the odd & even dead easy? Is it just a question of determining what comes first - odd or even?


Isn't the sequence ALWAYS:


E1, O1, E2, O2, E3, O3....


So, all the scaler has to do is get it right the first time & it's OK.


What I mean is, is it possible for some material to have it as

E1, O1, E2, O2, E3, O3.... and some other material can have it backwards (so that the odd & even bits on the same frame are from different film frames)?


Why would such material be created? Is there really any such material?


If some material has this, than obviously you need 2:2 pulldown - to find the right sequence of frames that match the pattern (since even can be the first field sometimes and odd some other times). Matching them together causes an annoying zipper effect that enhances scan lines (I've seen this on a PC that was running odd & even in the wrong order).


However, isn't it always the case that material is produced properly (or mostly the case). i.e., that odd fields and even fields from the same frame (BTW, does Even come before odd, or the other way around?) end up together in 2:2 pulldown material?


If the odd & even can switch like that, how would you even be able to detect it?
 

·
Registered
Joined
·
1,687 Posts
Ofer


Each of my various devices seems to have a natural order for film mode.


DVD on my setup is even first

Video is odd first

DVB is odd first


The above is for well mastered PAL films.


There are cases of bad edit creeping in normally when something is shot on film you can get a stray odd/even pair creeping in which causes very visible combing if not picked up.


I have a couple of tapes of programs where this happens.


The other touture test is stuff transfered from the US.


With this you get some sequences on both of the polairites with pretty random switching causes by the convertion from 59.94 with 3:2 pulldown with maybe bad edits to 50Hz.


ER used to be shown here in that way and that would be a tourture test for a 2:2 scaler.


John
 

·
Registered
Joined
·
82 Posts
Sadly there is a variation with 2:2 as to which field comes first (i.e. Field Dominance)


Basically Telecine machines have a switch on them to determine their field dominance - it's usually set during installation and then forgotten - unless someone starts playing about with it!


The problem is that on certain material the elements can be transferred on different telecine machines and hence there can be variations. You can also get bad edits where you get a pair of non-related fields together which is obviously bad as well.


So for a scaler in the 'PAL' world 2:2 detection is as important as 3:2 detection - but generally the detection is a little easier (but only a little).


Hope this helps,


Richard Ansell

Snell & Wilcox
 

·
Registered
Joined
·
9,884 Posts
Quote:
So what I undestand from all this is that 2:2 is primarily a detection scheme for field dominance
From my own coding experiance, field dominance is the easier part of it. If you just compare fields any even field will better match either the preceding or next odd field.


The harder part is automatically deciding whether you should be doing pulldown matching at all since it's possible you have a video source where they are not supposed to match but be separated at equal intervals in time.


To me this seems to be futher confused because either video cameras or something in the post production process is not quite symetrical. It appears that a lot of video material has sort of a statistical 2:2 pulldown patten to it, even with NTSC video in the US. That is, there appears to be a dominance effect where pairs of fields are not actually from the same point in time but not equally distributed either. I've been thinking about it as sort of half-a$$ pulldown. ;)


- Tom
 
1 - 11 of 11 Posts
Status
Not open for further replies.
Top