Adding a pulldown??

[swiperstopswiping]

I am confused. I have been reading forums discussing different pulldown's such as the 2:3:3:2 and just don't get it. Is a pulldown something that one must manually do within a NLE?? Is something actually being pulled down while the camera is filming? I've searched the forums, but can't seem to find anything that is actually explaining what a pulldown is/does.

 

[mikkowilson]

"pulldown" is a way to record one framerate within another. It's normally done by the camera, though sometimes in post too.


In the case of the DVX, when it is shooting at 24 frames per seccond, it has to somehow record those onto tape at 30 frames per seccond - beacuse that's what the DV standard requires.

There are many ways to do this.
First would be to record 24 pictures of every seccond, and then leave the last 6 frames blank. - But that would look very bad with a black flash every seccond.

So the next idea would be to record 24 frames, and then repeat the last frame another 6 times. This gets rid of the black - we always have a picture - but it would still leave a huge pause in every seccond of video.

Well next step would be to figure out how to spread out those repeated frames.

We could record 12 frames, then repeat 3, then another 12 and repeat 3. This would fill up the 30, but still be jumpy.

Ok, so how about we repeat every 4th frame to smoothen it up?
Like this: 1,2,3,4,4,5,6,7,8,8,9,10,11,12,12,13,14,15,16,16,17,18,19,20,20,21,22,23,24,24 = a total of 30 frames.
Now we are getting close! We have 24 shot frames spread out and recorded over 30 frames of video. This is much better, but still has a slight stutter every 4th frame. If video was progressive, this is the best we could do! (interlace haters, take note, without interlaced NTSC video, 24P would be much harder to do well.)

However the 30 frames of NTSC video, are actually made from 60 "fields" - each frame is a double exposure basically. With every other line of pixels in one field, and every other line in the 2nd. So these even smaller increments allow us to spread the 24 frames over an even larger area (60, not 30), but we have to be carefull to make sure we get both halves of each frame recorded, at least once!
The result is a system that spreads out 4 shot frames (from 24fps) over the space of 5 frames, or rather; 10 fields, of recorded video.
The 2:3:3:2 notation reflects exactly this.
It shows that the first frame of shot video takes up 2 recorded fields (or one full frame) the 2nd shot frame takes up 3 fields (so one field is repeated) when recorded, and so on. The next frame again is recorded over 3 fields (another repeat), and the last frame over another 2 fields. When played back from tape, some fields are repeated, but they don't stutter nearly as badly as repeated frames would. Ta-Da! 24P playing (fairly) smoothely at 60 fields per seccond.

However it's really hard to edit video that's encoded in this pulldown system. So when you import the video into an NLE, it normally performs what's called a "reverse pulldown" which extracts the original 24 frames from the 10 fields of video, by dropping out all the spare ones. Then the editor actually runs at 24 frames per seccond (all nice and smooth without any jitters.)

If you use and external video monitor with your editor, often the NLE will do a quick 2:3:3:2 pulldown again so the NTSC (60 fields per seccond) monitor can display the picture properly. Your NLE will also do it for you if you want to record your final 24fps video back out to an NTSC tape. - In fact unless you are exporting to computer based media (Windwos Media, Quicktime, etc) or a DVD (which all support 24p [and then leave the pulldown to the DVD player), you will normally end up dooing another pulldown on the final video every time.


The bad news about pulldown is that if you are recording to a hard drive (when capturing, or using a Firestore type device), it will take up unecesary space. But the good news is that the conversion itself won't effet the video quality, as all it's ever dooing is repeating data, or removing repeated data. Nothing is ever lost.


- Mikko

 

[swiperstopswiping]

Wow, thank you for that in depth response. So no extra work in a NLE such as premiere 2.0 is required to perform a "reverse pulldown?"

 

[mikkowilson]

Working mostly in PAL land, I don't have any practical experience with working with 24P in editing (as we have 25p - dead easy in the 50 fields (25 frames) that is PAL video). Check the help files or ask in the Premiere section of the forum for deatails on what you need to do. I can't remember if it's automatic or not when you are working in a 24P project.

- Mikko