What is 2:2 Pulldown?

[Undead Sega]

Hi, i dont know if this is the right place, but i was wondering what exactly is 2:2 Pulldown?

i know its abit of a silly question and i read about it and stuff like in Wikipedia, and looked abit on the net, but i cannot get enough information, and plus i do not understand it fully.

also through afterdawn's glossary, it says "There are currently no consumer tools for adding 2:2 Pulldown to any type of video".

can anyone here kindly explain what it is exactly, i know one thing from it, and is it helps convert Film (24p) to PAL, but isnt it just like 2:2:2:2:2:2:2:2:2:2:2:3 Pulldown?

[Guest]

It should really be called "2:2 telecining". It just means that for a pair of frames you have 2 fields for the first and two for the second, i.e., no fields are pulled down and each source frame has its two fields. The rate conversion is achieved by speeding up the film by 4%.

[Undead Sega]

so, is it technically, PAL Speed-Up?

correct me if i am wrong. Also, either way, which is it? and can it be done?

[Guest]

so, is it technically, PAL Speed-Up? Isn't that the meaning of what I said?!

correct me if i am wrong. Also, either way, which is it? and can it be done? Which is what? Can what be undone?

There is no pulldown in "2:2 pulldown", so there's nothing to be undone.

[Undead Sega]

so there's nothing to be undone.

Oppps! sorry, i meant 'done' can it be 'done?' assume so with VirtualDub and BeHappy.

so its just a speed up, i think its a terrible name! '2:2 Pulldown' :( i was getting very happy, as i hate speed ups.

[Guest]

Oh, sorry, I misread your post.

If you can serve your video via Avisynth then speeding it up is as simple as using AssumeFPS(). I have never done audio speedups but I'm sure it is a common thing with available tools.

If you are wanting to convert film rate material to PAL, you can use real pulldown (2:2:2:2:2:2:2:2:2:2:2:3). Then you can leave the audio as is. If the target is MPEG2, use DGPulldown, otherwise do hard pulldown using an Avisynth script.

[Undead Sega]

its cool, and yeah, i already know how to do the pulldown, i find it alot better than PAL Speed-Up :D i wish studios could use this method, i suppose it is less noticable than 29.97 pulldown, but i also wish there was such thing as PAL 24p on DVD much like NTSC 23.976 :(

dont u agree?

[Guest]

i also wish there was such thing as PAL 24p on DVD much like NTSC 23.976 :(

dont u agree? I'm not aware of 23.976 being supported on NTSC DVD. Can you give me a reference for that?

[Undead Sega]

http://en.wikipedia.org/wiki/DVD-Video#Resolution_and_Frame_rate

i also own NTSC DVDs that are not 2:3 Pulldown either.

[2Bdecided]

i also own NTSC DVDs that are not 2:3 Pulldown either.No you don't. You may own NTSC DVDs with soft pulldown. You may own NTSC DVDs that are really 29.97fps. You don't own any DVDs that are actually ~24fps.

I can't believe you want to argue this point with neuron2 of all people!

Cheers,
David.

[Guest]

It's OK, I like arguments. I've lost arguments and I've won arguments. Winning is better. :)

My understanding of it is that for 23.976 you must have 3:2 pulldown flags to make a display rate of 29.97.

Do you have any reference that says you can make a 23.976 fps DVD without pulldown?

Better yet, please post a VOB fragment from the DVD that you claim is 23.976 without pulldown.

[reepa]

Would it be correct to say that an NTSC DVD may have 23.976 (en)coded pictures per second?

[Guest]

Would it be correct to say that an NTSC DVD may have 23.976 (en)coded pictures per second? Yes, of course. And it has 29.97 displayed pictures per second, achieved with 3:2 soft pulldown.

[Undead Sega]

according to Afterdawn: http://www.afterdawn.com/glossary/terms/22_pulldown.cfm

it says: 2:2 Pulldown is different than other pulldown patterns because it doesn't actually duplicate any frames. Rather it's a way to simply tell the video decoder to display the video as two interlaced fields rather than a single Progressive Frame.

nothing about Speed Up there either.

[Guest]

It's nonsense. Video decoders do not display the video. The display process displays it based on the pulldown flags and the connected device. There is no pulldown and there are no flags set. If you want to telecine 24fps to PAL without pulldown then you need a speedup. There is no signalling at all for "2:2 pulldown". It's a nullity. It doesn't exist. If you think it does, cite the spec that describes the stream syntax for specifying "2:2 pulldown".

And where's the VOB you claimed you have? Did you give up on that but you're not gracious enough to admit that it doesn't exist?

[Undead Sega]

okay, maybe there isnt that kind of pulldown, but i havent given up on that claim to which i am trying to figure out how to cut a VOB file.

[Guest]

Use DGSplit. Set it for 20 MBytes. After it starts writing the second split file, kill it. Then upload the first one it made, which will be the start of the VOB.

http://neuron2.net/dgsplit/dgsplit10.zip

But before you go to all that trouble, do this:

Open the VOB in DGIndex.
Set Video/Field Operation to Honor Pulldown Flags. <--IMPORTANT -- DO NOT OMIT
Start a preview with F5.
Look at the Field Repeats box in the Info dialog.

If you see the count going up, then you have soft pulldown.

If you see just 0 there, and the frame rate box says 23.976 (and not 29.97), then please upload that VOB!

(You *have to be sure* you have Honor Pulldown Flags set when you do the test above. If you do Force Film, it does not count field repeats and reports the frame rate as 23.976, because setting Force Film causes the pulldown flags to be ignored.)

[Sagekilla]

Isn't 2:2 Pulldown also presenting progressive frames as interlaced to the display? Or is this a different concept.

[Guest]

What syntax element in the stream do you think is instructing the player to present progressive frames rather than interlaced ones for 2:2 pulldown?

If you say the progressive_frame flag, then how does that differ from a progressive encode WITHOUT "2:2 pulldown"? It doesn't, because "2:2 pulldown" is a nullity.

Do you think a player is going to see progessive_frame=true and present progressively to an interlaced display? Of course not. At best, if the display is capable of accepting interlaced or progressive, the player might use the one that corresponds to the stream encoding, but that has nothing to do with any imaginary signaling of "2:2 pulldown", which does not exist.

Waiting for you to cite the syntax elements that specify "2:2 pulldown". Maybe you could start by reading the MPEG2 specification. :)

I'll say it again: "2:2 pulldown" refers to film content that is telecined to PAL rate without using pulldown but rather by speeding up the audio and video. You have to see that the term is dodgy when they use "2:2 pulldown" to refer to a stream without pulldown! I mean, it's like admitting that a black cow is white; it's incoherent.

[2Bdecided]

It's just the original frames, as frames, no repeats, no messing about. Each frame may be split into two fields at some stage, or it may not be. It doesn't really matter either way. Similarly, any encoder flags saying "interlaced" or "progressive" are irrelevant - setting them incorrectly harms efficiency/quality, but this is so common that you shouldn't use them to determine what the content actually is. It tells you nothing about the frame rate. (30fps progressive wrapped in 60i is also "2:2 pulldown"!).

FWIW, and I don't know, but I always assumed that the name "2:2 pulldown" was an American invention - it's such a simple process in PAL land that no one ever bothered naming it - but in NTSC land with the 3:2 abomination, they had to invent a name to differentiate the so-simple-it-doesn't-need-a-name PAL process from their "normal" 3:2 pulldown.

Cheers,
David.

[Comatose]

The way I see it 2:2 pulldown is just ticking "Encode interlaced" when encoding progressive. I've seen Japanese TV stations do this when airing native 30 fps progressive material.

neuron2: You referred to 2:2:2:2:2:2:2:2:2:2:2:3 pulldown... how would you implement that in an Avisynth script?
I've been really bad results using ChangeFPS(50).SeparateFields().SelectEvery(4,0,3) in PAL... including some "color interlacing" showing up on a CRT and really bad IVTC (or deinterlacing?) on a LCD HDTV. The HDTV was doing much better on normal broadcast, so...

Also: How would I go on about figuring out a pattern and "reverse engineering" it?

Thanks

[Guest]

2:2 pulldown is just ticking "Encode interlaced" when encoding progressive.

I've seen Japanese TV stations do this when airing native 30 fps progressive material. It's very common to encode everything as interlaced; do you want to say that everything is therefore "2:2 pulldown"? As was pointed out, the interlaced/progressive encoding status doesn't control the display process. Stop trying to infuse sense into a non-sensical thing!

2BDecided is right. In the US we had this notation 3:2 for telecining. Then somebody thought "How do I use this fancy notation to describe telecining by speedup without pulldown?". They came up with the silly notation "2:2".
neuron2: You referred to 2:2:2:2:2:2:2:2:2:2:2:3 pulldown... how would you implement that in an Avisynth script? Use SeparateFields.SelectEvery.Weave. You can figure out the arguments to SelectEvery if you think about it. You'll need your cycle to cover one second and you'll need to double a field once in the middle and once at the end. That's for 24 -> 25 fps. If you have 23.976 input, well, that's more difficult. If your source is MPEG2, you can use DGPulldown and then serve the video, at which point you have hard telecining.
Also: How would I go on about figuring out a pattern and "reverse engineering" it? A pattern of what?

[FlimsyFeet]

My understanding of it is that for 23.976 you must have 3:2 pulldown flags to make a display rate of 29.97.
That's my understanding as well - however, if you make a DVD out of a 23.976 fps MPEG2 (ignoring any warnings your authoring package might throw up) then odds are it will play, even if it is out of spec. Also I wouldn't like to say with 100% certainty that there are no commercial DVDs like this, based on some of the weird standard conversions/telecining/field blending you can find on some Asian releases. :)

How would a PAL progressive scan DVD player cope with this 2:2:...:3 pulldown? For NTSC 3:2 pulldown, I understand the player reconstructs the progressive frames, then displays frame 1 for 0.05s, fraem 2 for 0.033s and so on. So with the PAL method the player shouild show each frame for 0.04 seconds then every 12th frame for 0.06. Would this not give a more noticeable judder than the NTSC alternative?

I don't mind speed up most of the time, but I can notice it if I have a correct speed version to compare against (or a soundtrack CD!)

[Didée]

neuron2: You referred to 2:2:2:2:2:2:2:2:2:2:2:3 pulldown... how would you implement that in an Avisynth script?
I've been really bad results using ChangeFPS(50).SeparateFields().SelectEvery(4,0,3) in PAL... including some "color interlacing" showing up on a CRT and really bad IVTC (or deinterlacing?) on a LCD HDTV. The HDTV was doing much better on normal broadcast, so...
Something else must've went wrong in the process. Perhaps ...

... the source wasn't truly progressive (or had to be IVTC'ed and something went wrong there) (both rather unlikely)

... you encoded as progressive instead of interlaced (possible)

... there was a YV12->RGB conversion with progressive sampling instead of interlaced sampling (very likely!)


In any case: If you have a progressive 24fps or 23.976fps source, and want to perform hard pulldown to 25i, then the sequence

ChangeFPS(50).SeparateFields()
SelectEvery(4,0,3).Weave()

is absolutely correct. Nothing more to say about that.

[Comatose]

The source was progressive (a 720p59.94 Hauppauge capture off PSN of anime - decimated down to 23.976fps), and I double checked that I did encode interlaced.

The DVD player I used for the CRT TV is a rather cheap (ok, very very very very cheap) player, and it's kinda crappy, so it might've messed up there, but I figured it didn't have to do much since it's hard telecined and at the right resolution.

The color interlacing part only happened on the CRT.
What I saw on the LCD were stuff that looked like bad field matching, for example, I had credits "jump" (bob?) a little after a scene change, as if they were matched with a field from the previous scene.
The LCD was also using a relatively expensive (high quality?) Panasonic player, so it was *probably* doing everything right.

Anyway, I'll wait for tonight and record some anime to see exactly what they do.

[Didée]

I was just referring to your comment that the Convert/Separate/Select/Weave method would have failed, which seems strange because the method impossibly can fail (unless some other error gets mixed in on the way).

You're very nebulous anyway ... the mentioned code was missing the Weave(), you're talking about how a 720p source plays on a DVD player after decimating, and when the discussion comes about how to correctly produce field pulldown (i.e. we speak interlacing), you're going to capture a 720p60 source to see what "they" do. Hard to make head or tails out of that.

Without having seen the exact script that you actually used, it's hard to tell what did go wrong. However, when one can see "color stripes" in the context of interlaced YV12, then it's almost sure that the YV12i has been treated progressively.


Example: source YV12 24p, hard pulldown to 25i in YV12 colorspace, with the code I posted above.

Using Doubleweave() to simulate an interlaced display device -
- right side: 25iYV12.ConvertToRGB()
- left side: 25iYV12.ConvertToRGB(interlaced=true)

http://img47.imageshack.us/img47/8123/interlacedyv12pitfallai6.th.png (http://img47.imageshack.us/my.php?image=interlacedyv12pitfallai6.png)http://img47.imageshack.us/images/thpix.gif (http://g.imageshack.us/thpix.php)

The one to the right looks exactly like the issue you mentioned.

[Comatose]

Sorry for being so unclear, I'm just kinda... not quite down to earth lately XD
Yeah, I was missing the Weave, but only here. Also, that 720p was downscaled, which is another "obvious" thing I forgot to state D:

And when I said I was going to capture from TV, I meant I'm going to capture SD 50i TV from the station that airs anime to see what kind of pulldown they use that makes it look good on both CRTs and LCD HDTVs... mainly to make sure it wasn't something I was doing wrong.

My script:
avcsource("blah.dga") # 1280x720p59.94
TDecimate(mode=1,cycleR=3) # 23.976
RemoveGrain(1)
spline36resize(720,576,0,4,-5,0)
ChangeFPS(50)
SeparateFields()
SelectEvery(4,0,3)
Weave()
and that went to QuEnc with interlaced encoding on.

Another important detail I forgot to mention is I only saw the color interlacing issue only 2 or 3 times for less than a second in a 1:30m video, and then it only appeared on very fast motion (there were high motion sequences that didn't have this issue). The horrible DVD player wouldn't let me pause, so I don't know exactly what it looks like.
I would've probably dismissed it as an issue with the cheap DVD player (this was on the CRT) but then I tested it on the better DVD player with the LCD HDTV and it looked like it went through the worst IVTC ever (although they are separate issues).

So what's bothering me is mainly the horrible IVTC or deinterlacing or whatever the HDTV is doing, which is why I suspect TV stations use the 2:2:...:3 telecine instead of this one.

A ton of thanks for your help! :D

[2Bdecided]

No, most TV stations use 2:2 and that's it. This implies a speed up. They like this - they can fit more commercials in!

Adding the extra two fields or one frame per second to maintain the original speed is less common. The one frame option is very visibly jumpy on smooth pans; the two fields option breaks some (many!) dumb deinterlacers.

Cheers,
David.

[Comatose]

Doesn't 2:2 imply it's just assumefps(25,sync_audio=true)?
Because it's definitely telecined D: Well, it's definitely something. It looks telecined on almost every frame with motion, but it doesn't look like what you'd see when watching interlaced material.

edit: I'm uploading a sample now. Done: http://www.mediafire.com/?nnnkedozmwe
edit2: It's definitely not sped up, Bleach was ending when I turned it on and I compared the ending to a YouTube video (fansub, so NTSC source) and it syncs perfectly/has the same pitch.

[2Bdecided]

It looks field blended to me.

If you want to find out what TV stations do when handling 24fps material properly, look at something that they were given as 24fps material (i.e. a real movie) - not something that was probably given to them on NTSC video, and then cheaply converted.

Cheers,
David.

[Comatose]

:thanks:

[Guest]

Now that the thread hijack appears to be over...

@Undead Sega

Are you going to upload that VOB sample?