LD#12: Film/video 3-2 pulldown and "white flags" Last Revised: 01 Apr 93 re: > I've got a [CLD-] 3070 [Pioneer laserdisc player] and quite like it. > The digital effects are very clean and in some cases even better than > CAV effects. I've noticed that some CAV disks don't have clean > picture frames (they are merged with another frame somehow, causing a > wobbly picture. can anyone explain why this happens? I've only > noticed it on a few CAV disks). What you are seeing is called "field motion", and on film->video transfers is [usually, but not always] a result of someone's ignorance or laziness. Here is an updated version of an article I wrote about this in 1989... Background: In the case of NTSC, the screen is refreshed at 59.94 Hz, or approximately 60 Hz. |<-----------------1/30 Sec.----------------->| |<-------------------FRAME------------------->| Apparent FRAME |<--------FIELD------->|<-------FIELD-------->| ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### ##################### Time = N Time = N + 1/60 sec. The 60 Hz FIELD rate actually generates a 30 Hz FRAME rate. Because of persistence of vision in the human eye, the first field has not completely faded before the next one arrives. Thus we see a complete frame, usually, most of the time, sort of... Most motion picture film is shot at a simple 24 frames per second, although made-for-TV material is sometimes shot at 30. Trivia: when projected on a motion picture screen, each (24 fps) frame is flashed twice, by a two-bladed shutter, resulting in a 48-per-sec image rate, reducing flicker. NTSC laser video discs always rotate to produce a 30 Hz frame / 60 Hz field rate. Although it would be simple to build an LD player that slowed to 24 (or 18 or 16) video frames/sec., many TV sets wouldn't sync properly at the lower rates, the phosphors would fade, increasing flicker, and mixed source material on the same disc side would still require a 30 Hz rate. On material transferred from 24 frames/sec film to 30 frame (60 field)/sec video, the ideal practice is that each frame is mapped onto 2.5 video fields. However, this requires performing the "pull down" halfway through a field, during a horizontal retrace, and thus is only practical in the digital domain, on the latest telecine scanners, such as the Snell&Wilcox "DEFT". The balance of this article assumes we aren't this advanced, and to my knowledge, no LDs were made this way in any case. In the analog domain, or on "real-time" digital telecine scanners, the optimal practice is to map a film frame onto three video fields, pull down the next frame, map it onto two video fields, then back to 3 field; hence the name "3/2 pulldown". Older low-budget practice is to repeat every fourth film frame for one video frame. See postscript #3. As confirmed by the 3M Video Disc pre-mastering guide, 24/60 telecine mapping looks like this (all numbers refer to original film frames): |<------------------------1/4 second----------------------->| +=========+=========+=========+=========+=========+=========+==== Film frame | 1 | 2 | 3 | 4 | 5 | 6 | +=========+=========+=========+=========+=========+=========+==== Transfer / \ \ / \ / \ \ / \ / \ \ / \ / .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---. Video field | 1 | 1 | 1 | 2 | 2 | 3 | 3 | 3 | 4 | 4 | 5 | 5 | 5 | 6 | 6 | 7 +-------+-------+-------+-------+-------+-------+-------+---+---+ Video frame | 1+1 | 1+2 | 2+3 | 3+3 | 4+4 | 5+5 | 5+6 | 6+7 +-------+-------+-------+-------+-------+-------+-------+---+---+ Field motion| No | Yes | Yes | No | No | No | Yes | Yes +-------+-------+-------+-------+-------+-------+-------+---+---+ LD frame # | 1 : 1 : 1 | 2 : 2 | 3 : 3 : 3 | 4 : 4 | 5 : 5 : 5 | 6 : 6 | 7 +-------+-------+-------+-------+-------+-------+-------+---+---+ White flag |W | | |W | |W | | |W | |W | | |W | |W .-------.-------. .-------.-------. .-------.-------. Laser still |<--1-->|<--2-->| |<--3-->|<--4-->| |<--5-->|<--6-->| (no field `-------'-------' `-------^-------' `-------^-------' motion) Unless steps are taken, two of every five video frames will contain field motion (non-matching fields from different film frames, each being alternately re-displayed at 30 times/sec). The Warner CAV set of "Empire of the Sun" (cat# 11844) is mis-mastered this way. The only way to individually view film frames 2, 6...etc. without field motion is to enable the player's digital (field) store (if present, and a control provided). Correct LD mastering places an "action point" signal called a "white flag" ("W" above) at the start of each sequence of RELATED fields. As it is placed several scan lines prior to the start of a field, it can also be considered to be just AFTER the end of a sequence of related fields, and I suspect it is actually used in this latter manner. In CAV still-frame mode, it tells the laser servo when to perform a one-track (two-field) backstep to re-display the current frame. The signal is called the "white flag" because it consists of a single horizontal video scan line with the active line period set to maximum white level. It is placed in the vertical interval in either line 11 of an odd field or line 274 of an even field. Lines containing visual picture information start at 21 or 284 (22 or 285 if closed-captioning is present). Another way that the logical film frames is marked is that the "picture number", a 24-bit bi-phase coded signal present in vertical interval lines 17&18 or 280&281. A player theoretically also knows where the film frame changes because the frame number changes. The LD spec (IEC-857 for NTSC) suggests that the white flag is used for still framing, but I have seen some indications that some players may be using picture number instead. The white flag and/or picture number tell the LD player where the start/end of the logical frame is for still-frame purposes only. During normal "PLAY", the LD player will free-run and show all the fields in the "Video field" line above. (If the white flags are left out on a CAV disc, the LD player will never display a still frame at all; it will just keep rolling. Consumer video media with this severe defect are exceedingly rare.) What is evidently happening on discs like "Empire of the Sun" is that the white flags are carelessly placed at the start of every second video field, and the picture numbers are changing every second field, instead of at the video boundaries of the original film frames, starting with the appropriate video field (field dominance). Based on a scan of my LD database, Warners didn't have a lot of experience with CAV when EoTs was transferred, so they were probably not even aware of this issue. CAV discs with incorrect white flags are relatively common. Even Disney (who have made many CAV discs) got them wrong on the first pressings of "The Little Mermaid". Another way to mess up the white flags is to get the field dominance wrong and get the flags mispositioned, off by one field or frame. If this is done, then various mis-behaviours result (see postscripts). Sometimes it is impossible or impractical to avoid field motion on film-source material. Even some Criterion Collection discs have brief interludes of field motion because short inserts were made from two video sources. Frame rate: each video field (of the 2 or 3) in each logical frame has the same frame number encoded in its vertical interval. On a CAV disc correctly mastered from 24fps film, the video field rate is 60, for a video frame rate of 30, but the LD frame counter will only change at 24/sec. The disc is NOT slowed down to 24 rotations/sec. CLV: Still "frame" on CLV discs does not rely on white flags. Also, CLV "still" on consumer LD players is still-field (not frame). A field is grabbed semi-randomly from the current track, and the lines are doubled (not interpolated) into a video frame. CLV "step" on most consumer players steps forward one CLV track (not by one or two video fields, unfortunately). Reportedly some CLV discs have white flags and some don't (and those that do may be assumed to include many improperly marked for 3/2 pulldown). The industrial Pioneer LD-V8000 has CLV still-FRAME mode and honors CLV white flags. Pioneer could have introduced a true consumer still-frame CLV capability, but I suspect it would have reverted to still-field if the disc did not have white flags. Frankly, due to the number of mis-mastered discs, I would rather they provided us with more user control of presumed pull-down, starting field and number of fields for digital stills. > My reaction: I'd rather have a frame freeze that occasionally screws up > than be stuck with a field freeze all the time. If you stop on a bad frame, > you can always move backwards or forwards one or two frames to clear it up. You don't need this compromise. It is possible to make a CAV disc that can properly freeze every original 24 fps source frame. Unfortunately, some producers don't bother, particularly when the side contains multiple selections, or material from both film and video sources. The edit sheet that needs to be filled out can be very complicated and time consuming. Silent sources: Silent and 8mm films may have been shot at 16 fps, 18 fps or something close. It is possible to use a periodic pull-down (16:4/4/4/3 and 18:4/3/3) on some rates, but it is not clear if video transfer and LD premastering equipment currently support this. Video sources: If the material was originally from live NTSC video source, the image was constantly changing during imaging. EVERY original frame (field pair) will have field motion, regardless of where the white flags are set. Only digital field store will provide a stable still image. A transfer of 60 fps film material, like "ShowScan" footage, would have the same problem as live video. In both cases, however, the higher image rate increases apparent resolution, reduces stroboscopic effects and makes the presentation more alive. I would be happy to surrender stable CAV still frame in order to get the benefits of 60 frames/sec mapped onto 60 fields. PAL discs: All PAL discs spin at 1500 rpm (25 Hz). Carefully mastered PAL discs could repeat every 12th (24fps) film frame for an extra video field, but the more usual practice is a simple 2/2 pulldown, which speeds up the film by 4%, and maps 1 film frame onto 2 fields. NTSC discs from PAL video source gets very messy indeed. Regards, PO Box 248 Bob Niland Enterprise mailto:name@isp-name.domain Kansas which, due to spam, is: 67441-0248 USA rjn AT access DASH one DOT com Unless otherwise specifically stated expressing personal opinions and NOT speaking for any employer, client or Internet Service Provider. ____________________________________________________________________________ Postscripts: A rogues gallery of other white flag follies 1. Forward field shift: White flags are placed at 3-2 field intervals, but are one field too late. Results in EVERY LD frame having field motion. Since discs made from live video sources also have field motion in every frame, this defect can be difficult to isolate. Tips: frame counter advances 24/second, or you know that the source element was film shot at less than 60 fps. |<------------------------1/4 second----------------------->| +=========+=========+=========+=========+=========+=========+==== Film frame | 1 | 2 | 3 | 4 | 5 | 6 | +=========+=========+=========+=========+=========+=========+==== Transfer / \ \ / \ / \ \ / \ / \ \ / \ / .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---. Video field | 1 | 1 | 1 | 2 | 2 | 3 | 3 | 3 | 4 | 4 | 5 | 5 | 5 | 6 | 6 | 7 +-------+-------+-------+-------+-------+-------+-------+---+---+ Video frame | 1+1 | 1+2 | 2+3 | 3+3 | 4+4 | 5+5 | 5+6 | 6+7 +-------+-------+-------+-------+-------+-------+-------+---+---+ Field motion| No | Yes | Yes | No | No | No | Yes | Yes +-------+-------+-------+-------+-------+-------+-------+---+---+ LD frame # | 1 : 1 : 1 | 2 : 2 | 3 : 3 : 3 | 4 : 4 | 5 : 5 : 5 | 6 : 6 | 7 +-------+-------+-------+-------+-------+-------+-------+---+---+ White flag |W | | |W | |W | | |W | |W | | |W | |W .-------.-------. .-------.-------. .-------.-------. Laser still |< 1+2 >|< 2+3 >| |< 3+4 >|< 4+5 >| |< 5+6 >|< 6+7 >| `-------'-------' `-------^-------' `-------^-------' 2. Backward field shift White flags are placed at 3-2 field intervals, but are one field too early. Results in every OTHER LD frame having field motion. |<------------------------1/4 second----------------------->| +=========+=========+=========+=========+=========+=========+==== Film frame | 1 | 2 | 3 | 4 | 5 | 6 | +=========+=========+=========+=========+=========+=========+==== Transfer / \ \ / \ / \ \ / \ / \ \ / \ / .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---. Video field | 1 | 1 | 1 | 2 | 2 | 3 | 3 | 3 | 4 | 4 | 5 | 5 | 5 | 6 | 6 | 7 +-------+-------+-------+-------+-------+-------+-------+---+---+ Video frame | 1+1 | 1+2 | 2+3 | 3+3 | 4+4 | 5+5 | 5+6 | 6+7 +-------+-------+-------+-------+-------+-------+-------+---+---+ Field motion| No | Yes | Yes | No | No | No | Yes | Yes +-------+-------+-------+-------+-------+-------+-------+---+---+ LD frame # | 1 : 1 : 1 | 2 : 2 | 3 : 3 : 3 | 4 : 4 | 5 : 5 : 5 | 6 : 6 | 7 +-------+-------+-------+-------+-------+-------+-------+---+---+ White flag | | |W | |W | | |W | |W | | |W | |W .-------.-------. .-------.-------. .-------.-------. Laser still |<--1-->|< 1+2 >| |<--3-->|< 3+4 >| |<--5-->|< 5+6 >| `-------'-------' `-------^-------' `-------^-------' 3. 4th frame repeat. There is at least one report of an LD having sequences done this way. > In particular, we can see this on "Elephant Parts" on those sections where > the material was filmed rather than video taped. However, on one section > of "Elephant Parts", 4 out of 5 jitter rather than 2 out of 5. Sounds like it was transferred using the primitive 4th-frame-repeat, and the white flags have (erroneous) even field dominance, as follows: |<------------------------1/4 second----------------------->| +=========+=========+=========+=========+=========+=========+==== Film frame | 1 | 2 | 3 | 4 | 5 | 6 | +=========+=========+=========+=========+=========+=========+==== Transfer / \ / / / / / / / \ / \ / / / / .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---. Film-field | 1 | 1 | 2 | 2 | 3 | 3 | 4 | 4 | 4 | 4 | 5 | 5 | 6 | 6 | 7 | 7 +-------+-------+-------+-------+-------+-------+-------+---+---+ Even Field Dominance: ----+-------+-------+-------+-------+-------+-------+-------+---+ LD frame | 1+2 | 2+3 | 3+4 | 4+4 | 4+5 | 5+6 | 6+7 | 7+ Field motion | Yes | Yes | Yes | No | Yes | Yes | Yes | Yes ----+-------+-------+-------+-------+-------+-------+-------+---+ Field motion in 4 out of 5 frames. Correct flagging of 4th-frame-repeat material. Odd Field Dominance: +-------+-------+-------+-------+-------+-------+-------+---+---+ LD frame | 1+1 | 2+2 | 3+3 | 4+4 | 4+4 | 5+5 | 6+6 | 7+7 Field motion| No | No | No | No | No | No | No | No +-------+-------+-------+-------+-------+-------+-------+---+---+ No field motion, but every 4th LD frame is a repeat. EOF