Hardly a month goes by when we don’t receive a reader letter or e-mail taking us to task for discussing the “2:3 pulldown” functions of DVD players instead of the more commonly found “3:2 pulldown.” Let me immediately put the confused at ease by saying that both terms refer to the same thing. Why we use 2:3 instead of 3:2 requires a short tutorial on “pulldown.”

The term itself comes from film. In a motion-picture camera, a “pulldown claw” engages the film’s sprocket holes and literally pulls the film into position for exposure. Since most film cameras have the reels positioned vertically, with the feeding reel on top, the action is downward, hence pulldown. (Imax equipment has horizontally mounted reels. “Clawsideways,” anyone?) The pulldown action occurs 24 times a second during normal film shooting (that is, other than time-lapse or slow-motion sequences). This produces a film frame rate of 24 frames per second (fps), each frame lasting 1/24 second. 

By comparison, the video frame rate for standard analog TV is 30 fps (rounded off from 29.97 fps), with the scan lines of an entire 1/30-second frame transmitted in two alternating (“interlaced”) sets, first the odd-numbered lines in a 1/60-second video field and then the even-numbered lines. Now, clearly 24 fps doesn’t equal 30 fps, nor does 1/24 second equal 1/30  second. So to show a film on television you have a problem. You can’t simply point a video camera at a running film and have it come out right.

The solution is to use an uneven pulldown sequence, as shown in the diagram. A telecine machine used to convert film into video signals takes two sequential film frames, A and B, and converts frame A into two video fields and frame B into three video fields. The whole sequence is then repeated for the next two film frames, and so on. The resulting signal gets recorded on DVD, usually with the third B field (marked X in the diagram) omitted, as it duplicates the first B field. Two fields, then three fields — 2:3! It all comes out even at the end since the duration of two film frames is the same as that of five video fields: 2/24 = 5/60.

PDF: Diagram

The “3:2” terminology comes from how telecines were operated years ago. But nowadays when you load up the film and the blank videotape into a telecine and press start, the first video you get is a two-frame group. The “2: 3” terminology, therefore, is used in professional video circles and in reference books such as the authoritative and recommended DVD Demystified by Jim Taylor. Actually, Taylor uses the supremely accurate “2-3” designation, since the operation is a sequence rather than the ratio implied by a colon.

From a 2:3-pulldown signal recorded on a DVD, a player’s progressive-scan output must produce a frame rate of 60 fps, twice that of interlaced video, with each progressive frame containing a full set of scan lines, even and odd. There are two primary ways of doing this, each having its own problems.

Weaving takes two successive interlaced fields and reinterleaves them to create a progressive frame. This is the process illustrated in the last row of our diagram, and it works well if the original material was progressive, like a film that has undergone the 2:3-pulldown process, and provided the disc tells the player that it was. How a player interprets a disc’s data stream if it has been misidentified is an important aspect of its progressive-scan performance, in addition to how it handles programs that mix original video material with film-based material that was converted to video using 2:3 pulldown.

These film-derived images show some of the artifacts that may result from progressive-scan conversion. On Top, artifacts from "weaving," on bottom, those from "bobbing." A variety of more advanced conversion techniques are becoming available.

If the original material was interlaced video, not 2:3-pulldown film, weaving can cause the artifacts shown in the first photo above. The existence of these faults is why progressive-scan conversion of video material is more difficult than movies and why some DVD players that look great on films stumble when reproducing live concert videos.

“Bobbing,” the normal process for generating a progressive signal from interlaced video, takes a single field and creates a full progressive frame from it by line doubling or line interpolation. It can produce the artifacts shown in the bottom photo, some of which resemble those produced by weaving. The most significant of them are stepped diagonal lines and a loss of vertical resolution.

More advanced progressive-scan con version systems combine bobbing and weaving, depending on the content of the image. Weaving is performed in areas of the picture where nothing much is moving over the course of several frames, increasing the vertical resolution in those areas. Bobbing, or even a form of time-averaging (smearing), over several frames is used in areas with a lot of motion.

Such advanced technology can be found in some standalone line doublers and re­scalers, and it’s slowly finding its way into DVD players. It’s not enough simply to look for a DVD player with 2:3-pulldown capability — all progressive-output players do it in some fashion. What you want to see in DVD data sheets and advertisements is some indication that something more than simple bobbing and weaving is going on. While you can look for the phrases “field-adaptive” or “motion-predictive,” you won’t always find them since there’s no standardized terminology for the various advanced conversion techniques. That some standard terminology would emerge is, as the Bard would say, a consummation devoutly to be wish’d.