Frame rates, shutter and judder

The numbers, and where they come from

24, 25, 30, 50, 60. None of them is arbitrary and the history explains every annoyance that follows.

24 was the sound-film standard from the late 1920s: the slowest rate that gave acceptable optical soundtrack quality, chosen to save film stock. It remains the cinema rate and the reason “cinematic” and “24fps” have become entangled.

25 is European television, one frame per two cycles of 50Hz mains.

30 is American television, matched to 60Hz mains. When colour was added to NTSC in 1953, the frame rate was nudged to 29.97 to stop the colour subcarrier interfering with the audio carrier. That fractional rate is why American video timing is awkward to this day.

50 and 60 are the doubled rates, used for sport, for broadcast, and for anything where motion clarity matters more than the look of 24.

Drop-frame timecode is not dropped frames

At 29.97fps, one hour of video contains 107,892 frames, while an hour of timecode counting at 30fps expects 108,000. Over an hour, timecode drifts about 3.6 seconds ahead of the clock.

Drop-frame timecode fixes this by skipping certain labels: the numbers :00 and :01 at the start of each minute except every tenth minute. No frames are removed. It is a renumbering scheme so that timecode matches wall-clock time, which matters when a programme has to be exactly 28 minutes 30 seconds.

People lose hours to this. Nothing is missing; the counter skips.

Shutter angle, and what 180 degrees means

A film camera’s rotating shutter is a disc with a slice cut out. A 180-degree opening means the film is exposed for half of each frame’s duration: at 24fps, 1/48 second.

This is the 180-degree rule, and it is a look rather than a law. Each frame carries motion blur covering half the interval between frames, so consecutive frames overlap in what they record and the eye reads continuous movement.

• Shorter exposure (90 degrees, 1/96 at 24fps) gives crisp, blur-free frames that strobe visibly in motion. The opening sequence of Saving Private Ryan is the famous example.
• Longer exposure (270 or 360 degrees) gives heavy blur and a dreamy, smeared quality.

On a digital camera it is a shutter speed, and the arithmetic is: shutter speed = 1 / (2 × frame rate) for the standard look. 1/50 at 25fps, 1/60 at 30fps, 1/120 at 60fps.

The practical consequence outdoors: matching 1/50 in bright sunlight needs a very small aperture or a neutral-density filter, which is why ND filters are standard kit rather than an accessory. Going to 1/500 to control exposure instead gives a strobing look you did not choose.

Judder, and why 24fps looks odd on a 60Hz screen

To show 24 frames a second on a 60Hz display, something has to be repeated. 60 does not divide by 24, so the standard scheme (3:2 pulldown) shows one frame for 3 refreshes and the next for 2, alternating.

The result is that consecutive frames are displayed for different lengths of time. A smooth pan then advances in uneven increments, which reads as a stutter: judder. It is most visible on slow horizontal camera movement, which is why cinematographers have a rule of thumb about how slowly a pan may cross the frame at 24fps before it falls apart.

Modern displays running at 120Hz can show 24fps cleanly (120 is divisible by 24, five refreshes each) which is one real reason a film looks different on a high-refresh screen.

Changing frame rate, three ways

Conform. Play the frames at a different rate. 25fps material played at 24 is 4 per cent longer, with correspondingly lower-pitched audio unless it is corrected. This is the traditional film-to-PAL method, in reverse.

Duplicate or drop frames. Cheap, and it produces exactly the judder described above.

Interpolate. Generate new frames between the real ones by estimating motion. Modern optical-flow retiming is impressive and fails in characteristic ways: around fast-moving thin objects, at occlusion boundaries where something appears from behind something else, and on repetitive textures where the motion estimate is ambiguous. The artefacts look like melting or tearing.

For slow motion, the honest answer is to shoot at the higher rate. 120fps played at 24 is true five-times slow motion with real frames. Interpolating 24 up to 120 invents four frames out of every five, and on anything with complex motion it shows.

Mixed rates on one timeline

Dropping 60fps material into a 24fps project makes the editor decide what to do, and different editors default differently: some conform to slow motion, some drop frames. This is the source of a great deal of “why is this clip stuttering” confusion.

Set the project rate first, deliberately, and convert anything that does not match on the way in rather than letting the timeline improvise.

Free tools

ffmpeg handles all of it:

# conform: relabel 25fps as 24fps, no new frames, 4% longer
ffmpeg -i in.mov -r 24 -filter:v "setpts=25/24*PTS" out.mov

# interpolate 24 to 60 with motion estimation
ffmpeg -i in.mov -vf "minterpolate=fps=60:mi_mode=mci" out.mov

DaVinci Resolve’s free edition has optical-flow retiming that is good. Blender’s video sequencer and Kdenlive both do frame-rate conversion. RIFE and similar open-source interpolation models often beat classical optical flow on difficult motion and are free to run locally.

The limitation

There is no correct frame rate, only conventions and what an audience is used to. The reaction to The Hobbit at 48fps (that it looked like television or a stage play rather than a film) is the clearest evidence that frame rate carries learned meaning rather than an objective quality ranking. Higher is not better; it is different, and the difference is cultural as much as perceptual.

Interpolation also has a hard limit that no model removes: information absent from the source cannot be recovered, only invented plausibly. When the invention is wrong it is wrong in a way that draws attention, which is why broadcast and cinema still prefer to shoot at the rate they intend to show.