2-3 pulldown

Except for certain very unusual film formats, motion picture film runs vertically through the camera and the projector. Be wary of anyone who presents a diagram like Figure 34.1 but rotated 90°.

Motion picture film is intended for display at 24 frames per second. Many television programs, including a large number of prime-time programs, historically originated on film at this rate (or at 1000/1001 of 24 Hz, that is, 23.976 Hz). This chapter discusses the conversion of film to frame rates higher than 24 Hz.

Film has historically been transferred to 29.97 frame per second video using a technique called 2-3 pulldown, whereby successive film frames are scanned first twice then three times to produce five video fields. The process is then repeated, reversing the roles of the first and second fields. The scheme is sketched in

Figure 34.1.

Figure 34.1“2-3 pulldown” refers to transfer of film at about 24 frames per second to video at about 60 fields per second. The first film frame is transferred to two video fields; the second frame is transferred to three. The 2-3 cycle repeats. SMPTE

RP 197 denotes film frames as A, B, C, and D. The A-frame is unique in being associated with exactly two fields – first, then second – of a single video frame. (“A-frame” denotes both a film frame and a video frame.) According to SMPTE RP 201, the A-frame timecode’s frame count should end in 0 or 5.

A

B

C

D

B2

D2

D1 :x4, :x9

D2

Telecine is usually pronounced tell-e-SIN-ee.

What I call 2-3 pulldown was historically called 3-2 pulldown. SMPTE standards assign letters A, B, C, and D to sets of four film frames. The A-frame is associated with the frame without a duplicate (redundant) field, so the sequence is best described as 2-3, not 3-2.

EBU Tech. R62, Recommended dominant field for 625-line 50-Hz video processing.

A piece of equipment that performs this film-to- video conversion in realtime is called a telecine. (The term film scanner ordinarily implies nonrealtime operation.)

In 29.97 Hz systems, the film is run 0.1% slow, at about 23.976 Hz, so that the 5⁄2 ratio of 2-3 pulldown results in a field rate of exactly 59.94 Hz. Figure 34.1 sketches four film frames; beside the set of film frames is the sequence of video fields produced by 2-3 pulldown. The labels 1 and 2 at the right indicate the first and second fields in an interlaced system.

When a 2-3 sequence is associated with nondropframe timecode, it is standard for the A-frames to take timecode numbers ending in 0 and 5.

In a sequence containing 2-3 pulldown, cadence refers to the temporal regularity of the A-frames. Careful editing preserves cadence; careless editing disrupts it.

In an interlaced sequence, field dominance refers to the field parity (first or second) where temporal coherence is susceptible to interruption due to editing. Video edits can, in principle, be made at any field; however, it is poor practice to make edits anywhere except the beginning of the first field.

Film is transferred to 25 Hz video by the simple expedient of running the film 4% fast, scanning each film frame to two video fields. This is somewhat jokingly referred to as 2-2 pulldown. The 0.1% speed change of 2-3 pulldown has no significant effect on the accompanying audio; however, the 4% speed change of 2-2 pulldown necessitates audio pitch correction.

I have described transfer to interlaced video. The 2-3 pulldown technique can be used to produce progressive video at 60 Hz. In this case, what I have described as first and second fields are first and second frames.

Engineers are prone to think that better-quality motion results from higher frame rates. Proposals have been made to shoot film for television – even to shoot movies – at 30 frames per second instead of 24. The 24 Hz rate is obviously the world standard for cinema, but it is also uniquely suited to conversion to both 50 Hz systems (through 2-2 pulldown, 4% fast) and 59.94 Hz systems (through 2-3 pulldown, 0.1% slow). Choosing a film frame rate other that 24 Hz would

Figure 34.2 “2-3-3-2 pulldown” is identical to

2-3 pulldown except that the eighth video field represents film frame C instead of D. The advantage over 2-3 pulldown is that film frames can be reconstructed from intact video frames; no “stitching” of fields is necessary. 35 mm motion picture film runs vertically through the camera and the projector. The scheme is also known as 24pA. It is not intended for broadcast; however, if broadcast, the motion impairment is not much worse than 2-3 pulldown.

A

B

C

D

Video fields

A1

A2

B1

B2

B1

C2

C1

C2

D1

D2

A-frame

Timecode

:x0, :x5

:x1, :x6

:x2, :x7

:x3, :x8

:x4, :x9

2-3-3-2 pulldown is denoted

24pA (for advanced) by some manufacturers. When used with DV recording, it is an advantage of 2-3-3-2 that film frames can be reconstructed without decompressing the video bitstream.

compromise this widely accepted method of conversion, and make it difficult for film producers to access international markets.

2-3-3-2 pulldown

When film at 23.976 frames per second is subject to 2-3 pulldown, the video “midframe” (timecode x2/x7) has film frames B and C mixed; the following video frame has film frames C and D mixed. (A video frame containing elements from two film frames is sometimes called a blur frame.) Although film frames A, B, and D can be simply reconstructed from single video frames, there is no single video frame that contains film

frame C.

This disadvantage is overcome by a minor adaptation to the pulldown sequence, denoted 2-3-3-2 pulldown, also known as 24pA (24-frame, progressive, advanced), diagrammed in Figure 34.2. Film frame C can now be reconstructed from the video frame having timecode ending in 3 or 8. A video sequence with 2-3-3-2 pulldown exhibits slight degradation in motion rendition compared to straight 2-3, because the 6 Hz beat frequency of 2-3-3-2 is slightly more visible than the 12 Hz of 2-3 pulldown. Nonetheless, the scheme is popular for desktop video editing.

FILM FRAMES

VIDEO FIELDS

Figure 34.3 Vertical/temporal relationships of 2-3 pulldown. Time is on the x-axis; vertical displacement is on the y-axis. 2-3 pulldown, shown on the left, introduces temporal discontinuities. The “dual-port” approach, on the right, introduces spatial discontinuities.

2-3 PULLDOWN

AB

0 1 2 3 4

“DUAL-PORT”

AB

0 1 2 3 4

FILM

A

A

VIDEO, 2-3 PULLDOWN VIDEO,“DUAL-PORT”

Figure 34.4 2-3 pulldown, spatial view. These sketches show the effect on the picture of two schemes to transfer film frames at 24 Hz, shown in the left column, into five video fields at 60 Hz. The center column shows the result of 2-3 pulldown. The right column shows the naive approach of writing into

a framebuffer at film rate and reading at video rate.

SMPTE RP 197, Film-to-Video

Transfer List.

SMPTE RP 201, Encoding Film

Transfer Information Using Vertical

Interval Time Code.

duce exactly the same temporal stutter as 2-3 pulldown!

Although 2-3 pulldown introduces a temporal artifact into the video stream, acceptable motion portrayal is obtained when the 2-3 video is displayed at video rate. However, if the frame rate is altered a second time, temporal artifacts of the two cascaded conversions may become objectionable, especially when the image is displayed with a wide picture angle.

In video that carries a carefully done film transfer, video pictures ending with timecode digits 0 or 5 correspond to A-frames of the film source. Apart from A-frame-locked timecode, there are no standards that convey, along with a video signal, information concerning film origination. Absent locked timecode,