dmunsil, Walking with Dinosaurs - film or video based?  

Can you give a specific chapter and title to look at? One of us will dump the flags and tell you how it is encoded. (of course we have to get a copy of the disc first.)

This documentary was done in Britain, so it's possible that they were using 25 frames per second, which converts well to the 50 fields per second of PAL. To convert this to an NTSC disc requires some odd pulldown cadences, which might cause odd issues with the JVC. I see this at times with other British TV series. In any event, if we can get our hands on a copy, we can answer the question pretty definitively.

When you put the JVC in Auto mode, do you get the steady blue light, or does it blink? If it's steady, that means the progressive_frame flag is set to 'true' on most or all of the frames. If it blinks, that's a sign that the flag is toggling off and on, like Titanic and Austin Powers.

When you watch it in film mode, do you get combing?

Don

The shows (episodes) are 35mm film based. All 35mm Film is 24 frames/sec. It doesn't matter where it is shown; USA, England, India, Africa, etc. When film is transfered to video is when the conjuring begins. 3/2 Pulldown, Skip Printing, etc is used to match whatever the frame rate for the specific video system is; PAL, NTSC, SECAM, etc.

Lee

Lee,

Film in theaters is always 24 frames per second, but for TV in this country, often 30 fps (really 29.97 fps) film cameras are used so it converts to video easier. Friends and some other sitcoms are shot this way. I believe that the same thing is done in Britain for TV shows.

Do some hunting around on the web and you can see discussions of the pros and cons of shooting 24fps vs. 29.97fps for TV.

For example, just poking around in Google, I find the following quote, from the Davinci Systems web site:

"SMPTE framing refers to the 3:2 pulldown relationship that exists when transferring 24 fps film (at about 23.98 fps) to NTSC video (at 29.97 fps). Note that when transferring 30 fps film in NTSC (at 29.97), and 25 fps film in PAL (at 25.00), one film frame exactly occupies one video frame"

And here's a quote from Soho Group, a London post-production house:

"All TV based productions shot at 25 fps are transferred and recorded in PAL at 25 fps whereby each film frame matches exactly to the frame rate of the VT recorder which also runs at 25 fps."

Best regards,
Don

[This message has been edited by dmunsil (edited 02-28-2001).]

Film is always 24 fps. TV cameras run at 29.97fps.
For PAL they speed up film 4% to fit 24fps into 25fps.
This makes film run faster and is really annoying
with music because it runs too fast.

I'll try again.

Film intended for movie theaters is always shot at 24 fps, everywhere in the world (excluding special effects like slow motion and fast motion).

Film intended for video is often (but not always) shot at a frame rate more compatible with video. This means 30 fps, or 29.97 fps, in the US, and 25 fps in Europe.

I'm not talking about video cameras here -- this is 35 mm film. "Friends," for example, is shot at 30 fps, on film, then transferred to video, then edited on video. Editing is much easier when there is no 3-2 pulldown to worry about.

Other shows that are striving for a film "look," like E.R., might shoot at 24 fps, edit at 24 fps (though probably using an Avid or something similar), then transfer the finished work to video, using 3-2 pulldown at that point. (I don't know for sure that E.R. is done this way, though I suspect it is.)

In Britain, most film is shot at 24 fps, but films that are specifically intended for TV, and TV shows shot on film, may be shot at 25 fps. Walking with Dinosaurs seems to me to be a perfect example of a show that would be convenient at 25 fps, but I don't know that they did it that way. It's just a possibility.

The 4% speedup you are talking about is when you take a theatrical film and transfer it to PAL, and it is to avoid this speedup that film intended for PAL video is shot at 25 fps.

Don

Don,
How can someone know how a show was filmed? Meaning at what frame rate? I'm curious as to how shows like the X-Files and Soprano's are filmed. Another thing I'm wondering, is how does this transfer over to DVD? Meaning, let's say a show is filmed at 30 fps, and transferred to NTSC DVD. Is this going to be transferred and flagged as 24 fps (film)? Or transferred and flagged as 30 fps (video)? I find this to be very, very interesting. I have Farscape on Region 2 PAL DVD. The show is shot in Australia. At 100 Hz (25 fps x4), camera pans and motion are quite fluid. I have viewed the exact same episodes on a Region 1 NTSC DVD at 72 Hz (24 fps x3). Motion didn't look smooth at all and appeared out of sync and jittery. I wondered how this could be, if it was originally filmed at 24 fps. But if it's filmed at 25 fps, that would explain a lot.

[This message has been edited by Larry Davis (edited 02-28-2001).]

sspears,
How do you "dump the flags"? Can the rest of us do this?

I am not aware of any public domain tools that do this. If you have a question about a certain disc, I can look at it and tell you. This is of course providing I have access to the disc.

sspears,
In that case, nevermind.

Sometimes you can find out from the flags on the disc. Unfortunately, finding out what the flags are requires a tool that can tell you. There are commercial tools that do this, but they tend to be really insanely expensive. Stacey and I have such a tool, but we can't release it.

Another way is to do some detective work with a software DVD player. In general, most players tend to show you one MPEG picture per freeze frame. So if you step through a chunk of the DVD a frame at a time (using the T key on PowerDVD; not sure about WinDVD), and watch for the seconds indicator to tick over, then start counting frames until it ticks over again, you should find out how many separate images there are per second. There might be accuracy issues with the counter, so it might be worthwhile to do several seconds and average.

However, on many discs, the software won't do the pulldown using flags -- it will just be dumped on to the disc as 60 field per second video. You'll see 30 frames per second, but several of the still frames will be combed. In this case, you have to look at the number of combed frames, and figure out what kind of pulldown was used to create the 60 field per second master. (Make sure to use "force weave" when doing this test, to ensure the combed frames stand out.)

If you see 2 combs in a row every 5 frames, that's 3-2 pulldown, and the source was 24 fps.

For 25 fps, the pulldown is going to be 2-2-3-2-3, so you'd get 2 combs in a row every 6 frames, or possibly 3 combs in a row every 6 frames, depending on where they start the sequence (they could do 3-2-2-3-2, for example).

For 30 fps, there is no pulldown, so you'll just get 30 clean frames.

For video shot with a camera, you'll get 30 frames, all with very small combs.

Certainly they're going to try to preserve all the frames, but there's no guarantee how they're going to do it. Keep in mind that there is no standard for flagging "film" on a DVD. The 3-2 pulldown sequence is just a convention. There's no "original frame rate" field on a DVD, and no flag that says "this was from film." At least, there's no flag that is required to be used.

For a 30 fps film, if you're lucky they'll transfer it as 30 MPEG pictures, all with the "progressive_frame" flag set to TRUE. But in practice, I imagine they'd leave the flag set FALSE, because most of the encoders aren't that smart. They see 30 fps; they assume video. So on an HTPC, the vertical filtering algorithm kicks in, and you lose vertical resolution.

For a 25 fps film, they might recreate the pulldown (2-2-3-2-3) with flags, keeping the progressive_frame flag TRUE. But more likely, the master tape they got from Europe gets transferred to an NTSC master, generating the pulldown at that time, and by the time it goes on the DVD, it's just flagged as 30 fps video. Again, the video deinterlacing algorithm then kicks in, and you lose vertical resolution.

What is interesting, and I never have thought about this before, is what does a good deinterlacing chip, like the Silicon Image, do with a 2-2-3-2-3 pulldown sequence? Does it recognize it as 25 fps film converted to 60 field video? If so, that's damn good. My guess is that they do, based on conversations I've had with the engineers. They have a library of weird pulldown signatures, and they try very hard to handle them all. I'll ask Silicon Image specifically about this particular situation.

I think your theory makes a lot of sense. It could be, however, that 100 Hz is just so high that anything would look smooth. It would be worthwhile trying something that isn't easily divisible by 100, and seeing if it looks smooth as well. I read a paper by a researcher who said that if refresh rates got over 75 Hz, then pulldown ceased to matter, because it was over the flicker-fusion frequency or something. But that may not be applicable to this.

Best regards,
Don

Thanks for the great reply, Don. I just bought a DLP that can do 100 Hz. Unfortunately, it's being repaired. But when I get it back, I'm looking forward to doing some testing.

I want to jump in here with a question for Bob C or anyone else. Does WWD look better on the progressive DVD player or a typical interlaced player? I recently had another odd experience with the British TV series DVD of "The Prisoner" that clearly looked better on my interlaced DVD vs. my HTPC. On the HTPC the playback was very choppy as if there were missed frames. It looked like bad motion artifacts.

I just got my HTPC and I will take a look at WWD and see if its comparable to "The Prisoner"

John Moschella

I do have this disc. Can you point me to an episode in paticular? I will take a look on a couple of different processors/displays to see how it looks.

OK. If you just put in vol 1 and play the Arrival episode chapter 1. After the credits are finished #6 runs around the Village a little bit then talks to a lady in an outdoor cafe. He then walks to a phone. This sequence, where he walks to a phone, I thought was particularly bad. As he walks by a stone wall it looks very choppy, what I would consider motion artifacts. I also notice hesitations throughout Chapter 1 up to this point. We're only talking about 5-10 minutes from the very begining. These kinds of effects are not nearly as noticable on my television via interlaced DVD player as it was on the HTPC.

Much interest in your impressions.

John Moschella