NTSCTelevision

Shooting Film at 24 fps for NTSC Television

extract from Digital Filmaking, 2nd Edition, Chapter 10

Television shows shot on film are shot at 24 fps and the 2:3 will exist for the EDL and negative conform. Since 24 fps does not equal 30 fps, one of the lists will always be in a match-backed state.

The choice is clear: either the editor edits on 1/30th of second boundaries to create a 30 frame EDL, which is frame-for-frame accurate or the editor edits on 1/24th of a frame boundary and creates a negative cut list that is not frame-for-frame accurate.

When the opposite list needs to be created, a match-back situation ensues. Since it is impossible to create 1:1 lists for both EDL and the film cut list, either list will be in a match-back state. We can refer to these lists as primary and secondary target lists. If the viewing audience first in the queue sees a video version, the primary concern is a video conform during online editing. The primary list is a timecode EDL that reflects the frame boundaries of the source videotapes. What the editor saw during the editing stage, is what will be seen after the online. For films, where the primary list is the negative conform, editing at 24 fps allows the editor to see every film frame, and where the editor cuts is how the resulting conform will appear.

MATCHING BACK

When a sequence is edited at 30 fps, the match-back process happens in order to translate the relationship between the timecode and the key numbers. The direction is from a 30-frame timecode EDL to a 24 frame key number cut list. There are several software applications available on the market that can provide this functionality (FilmLogic, Slingshot , Avid FilmScribe etc. ). Typically, there exists a database that contains all the sync-point relationships between the timecode and key number. All data is referenced to the timecode and to the reel ID of the timecode. The database can be logged in entire camera rolls or by separate takes. The minimum information required to create a match-back cut list

is:

Timecode = Key number = Pulldown phase

For example:

01:00:00:00 KJ 35 1234-6578+00 A

The database will most often contain extra information to better facilitate conform. Additional information such as camera roll, lab roll, scene/take can be integrated into the database.

Some programs will also track an extra field of timecode so that an audio conform of the original sound elements can be made. These extra fields of information allow the negative cutter to find camera rolls much faster and pull the required film elements. Some programs will also create pull lists that show how to pull the film in the order of lab or camera roll, or heads out or tails out depending on the wind of the camera roll.

When we have a situation where every video frame matches back to a film frame, we can call this a "natural relationship" between video and film, where one video frame matches back to one film frame. It is also a very naive way to think, that for each frame of video you will get a frame of film. The match-back program uses very simple calculations to find a key number. This can be done via a 4/5ths equation. For example, suppose our sync points are:

01:00:00:00 = AA-0000+00 = A

Let's say that we are looking for the corresponding key number for timecode number 01:01:10:00. The offset from the sync point would be calculated: 01:01:10:00 minus 01:00:00:00 = 01:10:00. We then take 01:10:00 and turn it into a total frame count: 1800 frames (30fps/sec - 60 sec/min) + 300 frames + 0 frames = 2100 video frames.

Since 5 video frames equal 4 film frames, we divide 2100 by 5 and multiply by 4 = 1680 film frames.

If we are working in 35mm film, there are 16 frames per foot.

We divide 1680 by 16 to get the footage offset: 1680/16 = 105 feet.

Since there is no remainder it matches exactly to AA-0000+00 + 105 giving a number of AA-0105+00.

Frame Calculations

The match-back becomes a bit more complicated when frames are added to the equation. Let's take the same video duration and add three frames to it. Instead of 01:10:00, let's use 01:10:03.

01:10:03 = 1800 + 300 + 3 = 2103 video frames divided by 5 = 420.6 video frames multiplied by 4 = 1682.4 film frames divided by 16 frames (per foot) = 105.15 feet.

We next multiply the remainder .15 by 16 to provide us with a frame count, which equals 2.4 film frames. During the match-back equation, the pull- down phase must also be considered. In this case, the sync point was an A frame at timecode frame :00. This means that for this range of transfer all time- codes ending in :00 and :05 will be A frames. The 2.4 frames left over will match to the C frame of the pull- down phase. Timecodes :01 and :02 would matchback to the same film frame thus adding only one frame to the key number offset. There still remains 0.4 of a frame. This 0.4 needs to be carried over to the next edit event and entered into the equation at the end of the next cut. We cannot simply discard the 0.4 of a frame, and it must be factored into the next edit event; it must be accounted for or a cumulative error will occur.

Over time all of the remaining 0.4 frames will accumulate over the course of an entire show. A drift of sync will be about one frame for every 30 seconds.

Since we matched-back to the C frame, we are now 1/2 frame ahead in duration. These cumulative "roundings" are added and adjusted on a per-cut basis. If the end of the next cut left a ½ frame addition to the duration, then the total frame cum. at this point would be +1 frame. When the error margin reaches +/-1 frame, an adjustment can be made. Depending on the error the adjustment will be made to make the cumulative error return to zero.

Following is an example of a cutlist that indicates the conform for each event and the amount of offset at each event from the original edit.

Project: 
civil wars Assemble List for edl file master: Seq First Edge Number Last Length 
Total Conform 001 83011599-9640+13- 9651+04 10+08 10+08 0.0 Source: 04;00;03;06 
04;00;10;06 Record: 01:00:00:00 01:00:07:00 002 83011598-8034+09- 8043+10 9+02 
19+10- 1.0 Source: 04;02;31;22 04;02;37;25 Record: 01:00:07:00 01:00:13:03 003 
83011598-8224+06- 8240+08 16+03 35+13 0.5 Source: 04;04;04;18 04;04;15;11 Record: 
01:00:13:03 01:00:23:26 004 83011598-8538+03- 8553+07 15+05 51+02 1.0 Source: 
04;04;58;04 04;05;08;12 Record: 01:00:23:26 01:00:34:02

Worst Case Examples for Duration Adjustment

Now let's look at some worst case examples where we can better illustrate the reason for duration adjustment. There are two cases that may exist: one, when the duration has gone longer and the adjustment is made by removing frames, and, two, when the duration is short and frames are added to make the adjustment.

Going Long

Shown in figure 10-2 is an example of how a duration has gone longer and a frame removal adjustment must be made. As we know:

Video = 30 frames = 1 second
Film = 30 frames = 1.25 seconds

Here, the editor has put together a sequence of single-frame edits. All the video frames happen to match back to A film frames. Remember that the A frame is the only frame where one frame is equal to two video fields without a timecode change.

The editor's sequence is 30 frames long. The editor expects the resulting sequence to play for one second, which it should since 30 frames in NTSC video is equal to one second. These 30 video frames match back exactly to 30 unique film frames. Since all frames used were "A" frames, there is no compensation involved and these 30 video frames are now conformed back to 30 individual film frames.

However there is now a discrepancy in duration. If we played these two sequences, one video and one film, at their native play rates, the video would be one second long, and the film would be 1.25 seconds long (24 frames + 6 frames). This is a case where the film has gone long. We say gone long when the playback of the conformed sequence longer than the expected duration as determined by the duration of the video sequence. To solve this discrepancy, the software application would have to remove six frames from the tail of the sequence to provide a one second duration for the film medium.

Going Short

Another example with the already established relationship that we know:

Video = 60 frames = 2 seconds
Film = 30 frames = 1.25 seconds

This is a case where the duration of the film is shorter than the expected video duration. Here, the editor has put together a two second video sequence, consisting of 60 video frames. The editor created this sequence by editing together short, two-frame edits. To show how difficult the match-back process can become, we have chosen an example where the editor has used only B frames, meaning that both video frames match-back to the same film frame.

Therefore, what will happen is that the original sequence of 60 video frames will not match-back to 60 film frames but will, instead, match-back to only 30 film frames. However the result is that what is a two-second video duration is only 1.25 seconds long in film duration. The film has come up short. Here, a total of 18 frames would have to he added in order to create a film duration of 48 frames, or two seconds.

Obviously, these are worst case examples, but what should be clear is how the importance of pulldown identification and tracking is essential to a frame accurate negative cut list. In most cases, events will compensate in overall duration, requiring only the single frame adjustment to remain in sync between the two mediums.

GENERAL MATCH-BACK

When a "normal" production sequence is run through match-back software, tracking of video duration and film duration uses this method of natural mapping. By also keeping track of the pulldown frames, certain events will have a frame added to or subtracted from the tail of the cut. It is most often preferred to add or remove frames from the tail rather than at the head of a cut. By doing this, any point in the EDL will have equal duration to its cut negative counterpoint, and knowing this pulldown relationship allows the match-back systems to find the correct key number. This is why matchback can never be "frame accurate" but only exact to +/- 1 frame at any single point in the sequence. This type of match-back only occurs in NTSC where there is a 2:3 pulldown relationship between film and video.