You’ve probably encountered the phrase ‘3:2 pulldown’ in a description of an audio video product. DVD players perform it, projectors perform it and even televisions perform it. So what exactly is 3:2 pulldown?
To fully understand what 3:2 pulldown is, we first have to understand why we need it.
Movies are shot using 35 mm or 70 mm film. The film is a series of still pictures called frames. When these frames are projected at a rate of 24 frames per second at the theatre, you get a moving picture. A movie!
Most of us aren’t George Lucas or Steven Spielberg so we don’t have a film projector and a theatre at home. We use televisions, be it plasma, LCD, or CRT. Even those of us who do have a projector at home, certainly aren’t feeding it film. What our home displays and projectors are expecting through their inputs is NTSC video.
NTSC video is displayed at 30 frames per second. Each frame is made up of 525 horizontal scan lines. In our wonderful DVD format, 480 of those scan lines are available to contain picture information. NTSC video is interlaced. In other words, even though the video is shown at 30 frames per second, only half of these scan lines are displayed at a time, creating two video fields per frame. The first video field is composed of all the odd horizontal scan lines and the second is composed of all the even horizontal scan lines. To complicate things further, when you put these video fields together they don’t give you a picture like the frames did in film. Each video field represents a moment in time, 1/60th of a second. If the video is of a still object and the camera is not moving there is no problem, you interlace two fields together and get a crisp picture. But, as soon as the object or the camera moves, putting two fields together creates distracting interlace artifacts on the display.
So to reiterate, film is made up of 24 frames per second and video is made up of 30 frames (or 60 video fields) per second. Simply transferring film to video would result in the film being displayed 25% faster than intended. Instead of 24 frames being displayed per second, 30 frames would be displayed. The solution is the ‘telecine machine’ used to transfer film to NTSC video which performs the 3:2 pulldown process. The 3:2 pulldown process uses a mathematical approach of repeating film frames to prevent the speedup of the film when shown on your television screen. Simply put, it converts 24 frames of film to 60 video fields by repeating the film frames in a 3-2 pattern.
For example: 4 film frames ABCD would each be divided into 2 interlaced video fields. So we get A1A2, B1B2, C1C2, D1D2. These fields are then repeated as A1-A2, A1-B2, B1-C2, C1-C2, D1-D2 (see figure 1). Notice the 3-2 cadence. If we repeat this 6 times, we get 30 frames of video from 24 frames of film.
So the 3:2 pulldown process works by restoring proper timing by generating redundant image information from four film frames to construct five NTSC video frames (as in figure 1). If we were to store five frames instead of four on a DVD, we would be wasting 20% of the storage space. Fortunately the MPEG-2 standard (in which all DVDs are encoded) addresses this issue. When a film is encoded in the MPEG-2 standard on a DVD, it is stored at 24 frames per second in the interlaced format. Each video frame contains exactly the same information as the original film frame. There is no redundancy. Each frame on a DVD is written as a 720 pixel wide by 480 pixel high interlaced frame (where each frame contains two 720 by 240 fields). A DVD encoded from film contains a special 3:2 pulldown flag within the MPEG-2 data stream that instructs the DVD player to repeat certain fields to construct an interlaced picture at 30 frames per second. This process is performed by DVD players in real-time (as they are reading the DVD).
As peachy as everything seems, it’s not. That’s because the scan lines from field B2 will be displayed right after the scan lines from field A1. This can lead to some serious visual distortions.
But many of today’s displays are capable of displaying a progressive picture which is much simpler to construct from a DVD that was encoded from film. Progressive scan ‘de-interlaces’ the video fields and weaves them together to assemble a full frame much like the original film frame. But now we run into the ’24 frames into 60 fields’ problem once again. To get around it this time, the DVD player triples and doubles its frame output in an AAA, BB, CCC, DD sequence. We once again we encounter the 3-2 cadence.
This leads to another problem. Frame A is displayed each time for 1/30th of a second while frame B is displayed for 1/20th of a second each time. This temporal imperfection is called ‘judder’ and could be eliminated if televisions were capable of displaying 72 frames per second. Unfortunately that would only work for film and would leave video in quite a bind.
De-interlacing video is more of a challenge than film. Each video field is a unique moment in time and unlike film, weaving 2 fields together doesn’t yield a full frame. The odd scan lines are 1/60th of a second different from the even scan lines. Different motion adaptive and compensating algorithms are needed to correct the resulting flaws in the picture.
If you don’t think that’s a problem when watching DVDs, think again. First, DVDs are designed for interlaced displays. Second, there is a lot of DVD content that originates from video, or was shot on film then edited using video (because video editing is much cheaper than film editing), or has video special effects superimposed over the film (even something as simple as subtitles or title sequences).
DVD players must know when to switch from film-mode de-interlacing to video-mode de-interlacing and back. A documentary shot in video might have a film clip in it. At this point a good DVD player will immediately switch to film mode to provide the best picture quality. Sometimes the 3-2 cadence gets interrupted and can trip up the DVD player. You may not have realized it, but your DVD player works feverishly to decode the content.
You may have DVDs at home that contain certain scenes that always display artifacts. You have probably attributed this to a bad DVD, and it probably is a poorly encoded DVD but a good DVD player will compensate for this and reduce all sorts of problems while outputting better picture quality.