Confused about Framerates

[phatslug]

Ive got these two DivX Avi's I recently converted to MPEG2 for burning onto a DVD. Their frames rates are 23.976 (NTSC Film), and 25 (PAL). Now I did the conversion with TMPEG, and changed few options outside the DVD-NTSC preset, however, in my haste I left the Contents of Movie as 'Video Movie' from a previous DV encode, rather than 'Film Movie', like I thought it should be. Both MPEG2 files have a framerate of 29.970, and play--get this--perfectly on my DVD player.

How the hell?? No jerkiness or audio sync problems.

So out of curiosity I tried doing the encode again on a small section, this time with 'Film Movie'. TMPEG automatically selected Inverse Telecine for me. The result, was indeed a movie which looked like every second, several frames were dropped.

Im quite confused. I really thought I had all this framerate stuff under my belt, but I guess I don't. Whats happening here, what am I doing right?

[DJRumpy]

The only magic here is pulldown. It must have been selected, either by you, or the template wizard. Your 'film' movie is still 23.976, but pulldown makes it appear as 29.97 fps. It does this by flagging individual fields. These flagged fields are duplicated in a pattern sequence during playback to increase the framecount by 1 frame out of every 5. TMPgenc knows that a DVD requires a reported 29.97 fps, so it would apply pulldown for you if your using the ntsc dvd template, and your source was 23.976.

The settings your looking for are specifically the 'Frame rate', and 'Encode mode' settnigs on the VIDEO tab. These specify how the playback will be handled. If your video is 23.976 fps, and you set the Encode Mode to:
"3:2 pulldown when playback"
you should then be able to select for the Frame Rate setting:
"23.976 fps (internally 29.97 fps)"
The english is bad, and the framerate description is backwards (it should be "29.97 fps (internally 23.976 fps)", but it's the setting you need to properly encode a FILM video, or a PAL video that has been slowed to 23.976 fps.

So out of curiosity I tried doing the encode again on a small section, this time with 'Film Movie'. TMPEG automatically selected Inverse Telecine for me. The result, was indeed a movie which looked like every second, several frames were dropped.

By any chance, did you use the PAL avi to perform this test on?

[indolikaa]

DJ, I am very confused about this whole framerate thing, and I've gotten a little frustrated because I've read a lot on this site and still don't understand it.

I have some Star Wars laserdiscs that I've been playing with. Now, are these mastered on the disc itself at the NTSC Film rate or the NTSC Broadcast rate? I would have thought that the signal was 29.97 so that the TV could understand it, but now I'm just plain lost. I don't see how the Film rate would even be an issue if the media had to be mastered to play on our home LD players and televisions.

Help?

[pijetro]

Yes you are correct. If you are playing it on your TV (NTCS land), then it's most likely at 29.97fps. It's been telecined for viewers of television. What many people like to do, is to get the film back to 23.976fps, via IVTC, in order to get the movie back to its original source, which was film. The reason for this is several,
1) You don't need to deinterlace (usually)
2) There's about a 20% savings in filespace
3) To watch it in DVD (back onto your TV), running pulldown.exe over your file just tells your settop to play it as if is 29.97fps....

[indolikaa]

So if you capture the LD and run the inverse telecine, you have the original i<24 frames/second. But you have to run a special program before authoring the DVD so the set-top player can play the DVD.

OK. Then two questions.

1) How does the computer know which frames to remove? Without some form of SMPTE or other timecoding being transferred from the LD, I would think you'd get a very erratic end result.

2) What if my DVD player and TV are capable of progressive scan? Can they handle the i<24 frames/second video without any pulldown tricks?

[pijetro]

Very good questions.

Answers.

1) Your computer doesn't remove anything. Telecining is a very specific pattern (usually 3 progressive, 2 interlaced), and if you frameserve your video via Vdub or Avisynth, then the sequence of frames being removed are pretty easy to correct before going to the encoder
Basically, your encoder receives the frames that Avisynth and Vdub remove.
In Vdub, zoom in on a captured clip and see if you can see a pattern of interlaced and progressive by moving your keyframe one at a time. That's the fastest way to see if your source is telecined. I believe if you frameserve, it's under the video>framerate>check Inverse Telecine reconstruct from fields adaptive.

2) True, but most authouring softwares won't recognize anything without the 2:3 pulldown flag, and if you share with friends, they won't be friends any longer if the vids don't play on their non-progressive settops..
BTW, the filesize of a pulldowned .M2V, is the exact same size as the filmed

[Yirkin]

I have read that animation is usually recorded at 23 fps. When capturing it from TV, is it best to capture at 23 fps or is that only done when the source is already 23 (I assume it is converted to 29 when broadcast)? I see a lot of mpg4 avis that are encoded at 23 fps. Do they capture at that frame rate or does someone convert it down to 23 when they are reencoding/compressing after the capture? Are 23 fps encodings primarily done to save on file size?

Thanks, interesting topic!

Howard

[dcsos]

Only in the case of internet fare will you find them cutting down on file size by using what is essentially the lowest frame rate that doesn't stutter

Otherwise your seeing a from film not video format file which could be at the FILM rate of 23.976

SO yes that has to go to 29.97 to AIR in the usa, but not if you get it on a dvd or such

[phatslug]

So correct me if Im wrong--

3:2 Pulldown is from 23.976 fps, non-interlaced -> 29.976 fps, interlaced
Inverse Telecine is the reverse of that operation?

So when I buy a DVD, someone has taken the original film content and done a 3:2 Pulldown to it? Is telecine the same as 3:2 Pulldown?

So what exactly does 29fps(23fps internally) do? And what about pulldown.exe?

Theres an excellent article about framerates and interlacing, originally posted by KetchSumAir, which helped alot -
http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html

Now Im wondering about FILM -> PAL conversions. I mean the FILM -> NTSC is allright because 24fps & 60 fields ps are a relatively nice ratio, 4:10. But 24fps & 50 fields ps, well, thats seems more complicated.

[KetchSumAir]

I'm not at all an expert on this, but I'll try my best to answer the questions:

3:2 Pulldown is from 23.976 fps, non-interlaced -> 29.976 fps, interlaced
Inverse Telecine is the reverse of that operation?

Actually, the 3:2 Pulldown is just a flag that gets set in the file. The DVD player takes that and converts the 23.976fps to 29.976fps on the fly, unless it's driving a progressive display source (such as an HDTV). Telecining is basically taking 24frames ps, changing it to 48 fields ps, then adding the required 12 fields ps via the 3:2 method. And, yes, inverse telecining is taking 60 fields ps that were ORIGINALLY in a 23.974fps format and telecined, and reverting them to the original 23.976 fps format. (Can be confusing - hopefully, I explained well...)

So when I buy a DVD, someone has taken the original film content and done a 3:2 Pulldown to it? Is telecine the same as 3:2 Pulldown?

Yes - they have added the 3:2 pulldown flag. Partially described above, telecining is really just following the directions of the 3:2 pulldown flag.

So what exactly does 29fps(23fps internally) do? And what about pulldown.exe?

Don't have a clue here (could guess... no, I'll let someone else answer this)...

Theres an excellent article about framerates and interlacing, originally posted by KetchSumAir, which helped alot -
http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html

Actually, I can't take credit for the original post. As far as I know, it was malducci who originally posted this in yet another post.

Now Im wondering about FILM -> PAL conversions. I mean the FILM -> NTSC is allright because 24fps & 60 fields ps are a relatively nice ratio, 4:10. But 24fps & 50 fields ps, well, thats seems more complicated.

Again, I'm not an expert here, but I believe it's quite simple too:
In PAL's case, for every 12 frames, one interlaced field is added - thus providing an equivalent 50 fields ps ((12 frames * 2 = 24 fields + 1 extra field)*2) The bigger problem is they have to slow the audio down just a bit, which effects the tone. So, of course, there is tonal adjustment needed as well.

[DJRumpy]

You're over complicating this by doubling the framerate. FILM tyo PAL conversions are just the reverse of PAL to FILM. The film is sped up by 1 frame per second, from 23.976 to 25fps. No frames are duplicated, or deleted.

Telecine and it's reverse, IVTC are a bit more complicated though. When performing IVTC, you are not deleting any content, you are simply removing those duplicated fields from the video. If you had 4 frames of video; Each frame is broken into two field (top and bottom), denoted by the T and B below. The telecine process simply creates a new FRAME by copying fields from existing frames (frames 1 and 3), to create a new frame like so:

4 original frames from a FILM video (23.976) split into top and bottom fields looks like this:

1T 2T 3T 4T
1B 2B 3B 4B

The telecine process duplicates the top field from frame 1, and adds a copy of frame #2's bottom field. This generates a new frame (#2) (1T/2B). The same process is used to create a new Frame #3. The top field from frame #2 is combined with a copy of the bottom field from Frame #3 (2T/3B).

1T 1T 2T 3T 4T
1B 2B 3B 3B 4B

So you started with 4 frames, and now you have 5. Two frames had their top and bottom fields duplicated to produce a 3rd new frame. IVTC simply removes these duplciate fields, restoring the frame count to 4 frames. No information is lost, since the new frames were made up of copies of existing fields. If you notice from the second example, frames #1, #4, and #5 are all unmodified (progressive). This is why telecined video has a pattern of 3 progressive, and 2 interlaced frames. To perform IVTC on this, you would simply remove the top field of frame #2, and the bottom field of frame #3.

There are only three different types of video to remember as far as television goes.

PAL (25 fps - Progressive or Interlaced)
FILM (23.976 fps - Progressive)
NTSC (29.97 fps - Interlaced / Telecined )

If your NTSC video is telecined, then it was originally created at 23.976 fps, and can be restored to FILM. If it is 29.97 fps (Interlaced), then it cannot be converted to FILM without chopping frames and losing content as a result, since no dupilcated fields exist in the video. You should NOT capture NTSC at 23.976, because you're dropping frames to capture at a lower framerate. ALWAYS capture at the rate of output (29.97 for ntsc), (you can capture at double the framerate if you want to capture every FIELD seperately, but that's outside the scope of this conversation). You need to, because if you capture at a lower framerate, you will not capture all fields (inlcluding the duplicated field) from the video.

So what exactly does 29fps(23fps internally) do? And what about pulldown.exe?

Both of these put pulldown flags on the MPEG. "23fps (29fps internally)" is a setting in TMPGenc to do this. PULLDOWN.EXE is an external program that does the same thing. The flags for the above example, would be placed on each frame that contains duplicated fields (#2, and #3) to indicate these are fields to be duplicated, and what field to duplicate. This way, the MPEG doesn't actaully need to have the extra FRAME actually encoded in the mpeg, but it can produce it during playback as if it actually was encoded. This saves space, and bitrate.

[phatslug]

Well shit... thanks DJRumpy & KetchSumAir, I understand it!
Thanks alot. Im extremely grateful.

Great info in those last coupla posts. That should be a guide or something

I got the theory, however now TMPEG is confusing me a little. Whats the difference between setting these three?

Encode Mode to 3:2 Pulldown when playback
Frame rate to 23.976(29.97 internally)
Checking the 3:2 Pulldown under the Advanced tab

Thanks again....

[DJRumpy]

The first two should be used together to produce a FILM encoded mpeg. The first adds the pulldown flags to the encoded mpeg. The second specifies the framerate on the video.

The last one supposed to do the same thing as the first two working together. I've never see it work though. Using the first two will produce a proper MPEG. The filter in the Advanced section seems to do nothing, although they may have fixed it in later versions. Remember I don't actually use TMPGenc

[adam]

The 3:2 pulldown filter on the advanced tab is for performing hard telecines to 29.97fps. If you have a NTSCfilm source, and enable this filter, and set output to 29.97fps, then TMPGenc will perform the 3:2 telecine and duplicate the fields to physically bring the framerate up to 29.97fps, as opposed to flagging the stream and letting the DVD player do the telecine. This filter isn't all that useful.

[DJRumpy]

It must have been the framerate setting I used then. I set my output to 23.976, assuming it added flags. Instead, it simply gave me what I asked for (23.976 fps).

What a pointless filter. I can't think of any reason, other than to show someone what NOT to do when converting to NTSC.