24p and 24pA pull-down flags

[Pete Bauer]

I'm trying to understand the pull-down flags that are used in miniDV and captured AVI files for 24p and 24pA video. I think this will go a long way toward helping to choose optimum settings and workflow for 24p work. With that information, those nagging questions about how clips are affected by various capture, editing, and especially export settings can then be reasoned through instead of guessed at. As an example, some people have expressed concern that under certain circumstances, cutting a single frame from an imported 24p clip could upset the cadence in an exported file; understanding the flags allows us to know if such concerns are warranted or not. And I've seen at least one instance of otherwise beautifully done 24p (2:3) footage ending up with interlace artifact in the final export because of a misunderstanding of the process.

There are a lot of good posts here and articles out on the internet that discuss the end result of those flags – the 2:3 or 2:3:3:2 cadence of 24p and 24pA – but try as I might, I can’t find enough information about the flags themselves to understand how these flags work. The nearest I’ve been able to find are these links:

http://mediasoftware.sonypictures.co...ctinfo/24p.pdf
(Early pages cover cadences; then page 7 discusses “Top Field First” and Repeat First Field” DVD flags)

http://catalog2.panasonic.com/webapp...del=AG-DVX100A
(click on the Operating manual link. Cadence is discussed on page 62 in “Shooting Progressive.”

Ok, four 24p frames are laid to 10 fields (5 frames) of video to comply with the DV spec of 29.97 by adding an extra frame with redundant fields, using one or the other of the cadences. I’ll call that extra frame and its fields “redundant” and the others “true.” At least some of these frames are flagged in the headers (“subcode”) that one assumes must identify the type of frame it is (redundant or true). That flag then presumably tells your editing software to either display a frame normally, or skip it, which is the more or less the reverse of what 24p DVDs do by flagging to repeat fields in the 24p (48i) file to give 60i playback on a TV. Then, of course, your NLE needs to be able to write the flags properly for rendered 24p files!

Without getting too terribly lost in the technical side, I'm hoping to get at least as far as being able to understand the following particulars:

1. Are the flags attached at the field level, or the frame level? (I'll assume full frames for the rest of the questions).

2. Does each of the five frames have its own flag? If not, is it just the redundant frame that is flagged?

3. What does a flag say about its associated frame?
-- Is it a specific identifier for the frame within a pull-down group (ie does it say, “I am Frame A or B or C or D or E”)
OR
-- Does it just generically identify the frame type, (“I am a full frame” or “I am a redundant frame.”)
OR
-- Is there just the "Top first field" or “Repeat first field” sort of flag scheme as DVDs use? (See p7 of the first link for more on that).

4. Assuming no more editing than cutting frames out of an otherwise unprocessed / unrendered timeline (which could possibly upset the cadence), do the original flags persist within a frame when exported back out of the common NLEs back to miniDV tape or to a DV AVI file?

Anybody have technical knowledge about this, or know of a good article that discusses the flags themselves, not just the cadence? The essence of it is, "How does the flagging scheme for 24p and 24pA work?"

With any luck, the answer will be a lot shorter than the question!

[Jan Crittenden Livingston]

<<<-- Originally posted by Pete Bauer : I'm trying to understand the pull-down flags that are used in miniDV and captured AVI files for 24p and 24pA video. I think this will go a long way toward helping to choose optimum settings and workflow for 24p work.

Peter, it is not the flags so much that will determine the optimum output but rather the pulldown and whether your system can extract the 24P from it. The easiest and cleanest is the 24PA or 2:3:3:2. So easy it can be done on the fly with many systems so that buy the time to footage is in the sytem , it is at 23.98. The 24P will require a rendering to remove the B form the BC frame, the D from the CD frame and the reversal of the two C fields as they are in inverse order.

24P
aa bb bc cd dd
12 12 12 12 12

See the two c field are numbered 2 and then 1. This has to be change to 12. Thisis a rendering process, and will require the footage to go into the system at 29.97 and then render it to a new file. There might be systems that can do this on the fly but they are not at the low end.

the 24PA

aa bb bc cc dd
12 12 12 12 12

Remove the bc frame and you have 23.98 or 24P.

>With that information, those nagging questions about how clips are affected by various capture, editing, and especially export settings can then be reasoned through instead of guessed at.

All capture of the footage in SD is at 29.97. The reason is that there is no SD machine that records in 24P. Only HD. The timeline or sequence has settings and those at set for 23.98, with the 2:3:3:2 remove. This is done before capture. Frome there the NLE knows what to do.

Tf you have a system that cannot remove the pulldown, then you should shoot in 24P, and edit in 29.97. Be happy.


>As an example, some people have expressed concern that under certain circumstances, cutting a single frame from an imported 24p clip could upset the cadence in an exported file;

This is true, and you should never make an edit near the end of any of the composite frames or inbetween them. This only applies to sraight 60i editing of 24P footage. Imagine if you cut inbetween the bc frame and the cd frame. Yuck!

>understanding the flags allows us to know if such concerns are warranted or not. And I've seen at least one instance of otherwise beautifully done 24p (2:3) footage ending up with interlace artifact in the final export because of a misunderstanding of the process.

This is mostly because they edited in 60i and did something incorrect. The flags won't help you here. An understanding of what you are editing will. I have seen huge errors like you describe, when someone sees the 23.98 potential in FCP and uses that for capture. Oops, now it is just editing against a 23.98 time line without regard for the extraction. This setting is for HD. I have seen same error by using the wrong extraction. I actually have one of these errors on our demo disk, but it was a student project and the pictures were lovely so I used it any how.

>There are a lot of good posts here and articles out on the internet that discuss the end result of those flags – the 2:3 or 2:3:3:2 cadence of 24p and 24pA – but try as I might, I can’t find enough information about the flags themselves to understand how these flags work. The nearest I’ve been able to find are these links:

It is because the flags are only relevant to the edit syste, you can't see them, but you cann see the pulldown.


Ok, four 24p frames are laid to 10 fields (5 frames) of video to comply with the DV spec of 29.97 by adding an extra frame with redundant fields, using one or the other of the cadences. I’ll call that extra frame and its fields “redundant” and the others “true.”

There are no redundent frames, only in the DVCPROHD are there redundant frames. There are cobination frames like those described above.

1. Are the flags attached at the field level, or the frame level? (I'll assume full frames for the rest of the questions).

Frame level.

2. Does each of the five frames have its own flag? If not, is it just the redundant frame that is flagged?

Only the A frame is flagged.

3. What does a flag say about its associated frame?
-- Is it a specific identifier for the frame within a pull-down group (ie does it say, “I am Frame A or B or C or D or E”)

I am an A frame.

4. Assuming no more editing than cutting frames out of an otherwise unprocessed / unrendered timeline (which could possibly upset the cadence), do the original flags persist within a frame when exported back out of the common NLEs back to miniDV tape or to a DV AVI file?

In general there are no flags in the 24P footage, only in the 24PA footage. Any system that can render out the 24P from a 29.97 has its own method for determining the A frame. In Cinetools, unless they have made it smarter, you have to manually tell it which is the A frame. In the Andreniline, it knows the A frames are always :00. So in 24P the flag isn't necessary, but in the compressed transfer the flag serves a purpose, and tells the NLE where the A frame is and thus it knows where the BC frame is and dumps it.


Anybody have technical knowledge about this, or know of a good article that discusses the flags themselves, not just the cadence? The essence of it is, "How does the flagging scheme for 24p and 24pA work?"

Hope this helps, if not, please ask again.

Best,

Jan

[Pete Bauer]

Jan,

Thanks very much for taking the time to answer in such detail. It'll take me a while to digest all of this and do little tests to fully understand it. For now, the one thing that I wasn't quite clear on:

"In general there are no flags in the 24P footage, only in the 24PA footage."

Do both 24p and 24pA have an "I am an A Frame" flag, or did you mean that 24p has no such flags at all? I'm guessing both do have the flag, since p62 of the DVX100a manual comments that both modes will be shot in 5-frame increments?

Time allowing, I'm going to set up some end-to-end torture tests over the next week or so. That may generate a small follow-up question or two, but again, thank you very much!