Originally Posted by Richard Paul
I have read many technical explanations on motion judder and I have read books on video such as "Digital Video and HD" by Charles Poynton. Based on what I have read I am very skeptical of the idea that even frame repetition could cause motion judder which is why I have asked for evidence.
It is one of the great books, definitely, but some things go beyond the scope of the book, such as the lack of coverage of true native 120Hz refresh rate displays (including distinction between impulse-driven and sample-and-hold), as well as tests done on brand new refresh-rate-multisync sample-and-hold displays (e.g. modern LCD's that sync to all 24Hz/48Hz/72Hz/96Hz/120Hz that now exist, or supporting most rates). For example, my 120Hz computer monitor syncs to 48/72/96/120, while the SEIKI 4K syncs during 1080p to 24/48/72/96/120. These refresh-rate-multisync sample-and-hold displays present an excellent opportunity to test different refresh rates for film-based material, in a "see-for-yourself" manner.Situation #1For sample-and-hold (common LCD)
, 24fps looks identical on 24Hz, 48Hz, 72Hz, 96Hz, 120Hz, when excluding other subtle differences (e.g. skew effects from slow-scan, motion blur differences caused by pixel response differences, color degradation effects caused by fast scanning, other factors, etc). Judder is identical. Many LCD"s, such as the SEIKI 4K LCD are able to be slow-scanned at 24Hz. Likewise, SEIKI 4K LCD is also fast-scanned (at lower resolution, 1080p), at true native 120Hz, too. It syncs to all refresh rates in between, too! Film on such a refresh-rate-multisync LCD looks just as juddery at 24Hz, 48Hz, 72Hz, 96Hz, 120Hz. No more. No less. Actual test done on such refresh-rate-multisync LCD's that did not exist until recently. There are some LCD's that can sync to all these refresh rates, for testing, and tests have shown that judder (on continuously-shining displays) is exactly identical at all exact multiple of refresh rates. I'm talking about LCD's without strobing (scanning backlight, strobe backlight, accidental effects from PWM dimming, etc).
24fps@24Hz LCD (1:1 pulldown) continuously display the same frame for 1/24sec, showing a 1/24sec sample-and-hold effect
24fps@48Hz LCD (2:2 pulldown) continuously display the same frame for 1/24sec, showing a 1/24sec sample-and-hold effect
24fps@72Hz LCD (3:3 pulldown) continuously display the same frame for 1/24sec, showing a 1/24sec sample-and-hold effect
24fps@96Hz LCD (4:4 pulldown) continuously display the same frame for 1/24sec, showing a 1/24sec sample-and-hold effect
24fps@120Hz LCD (5:5 pulldown) continuously display the same frame for 1/24sec, showing a 1/24sec sample-and-hold effect
Assuming non-strobe-driven LCD, displaying at 100% brightness (to eliminate PWM dimming from affecting the motion blur math, to preserve the unity between subjectve=objective)Situation #2For impulse-driven displays (CRT) that flicker:
24fps@24Hz CRT (1:1 pulldown) completely eliminates motion blur and judder for movies, but is super duper flickery.
24fps@48Hz CRT (2:2 pulldown) has a double image effect, with two flickers per frame creating a 1/48sec sample-and-hold effect.
24fps@72Hz CRT (3:3 pulldown) has a triple image effect, with three flickers per frame creating a 2/72sec sample-and-hold effect.
24fps@96Hz CRT (4:4 pulldown) has a quadruple image effect, with four flickers per frame creating a 3/96sec sample-and-hold effect.
24fps@120Hz CRT (5:5 pulldown) has a pentuple image effect, with the five flickers per frame creating a 4/120sec sample-and-hold effect.
The larger the number of frame repeats, the multiple-image effect becomes harder and harder to see, and eventually the repeated image effect are so numerous it simply blends into simple motion blur (the sample-and-hold effect). The more it looks like the film playing back on sample-and-hold displays.
As a rule of thumb the higher X:X pulldown on a CRT display, the more motion blur occurs because of the sample-and-hold effect that repeated flickers creates (as long as the flickers are hitting above flicker fusion threshold). This also assumes no interpolation is being done. Also, old film projectors often does the equivalent of 2:2 pulldown, strobing each frame twice.
Observe 1/48sec < 2/72sec < 3/96sec < 4/120sec < 1/24sec, the repeated flickers increase motion blur, so the higher the number of repeated frames tends to add motion blur, until there are so many flickers that the flickers blend to a complete sample-and-hold effect (1/24sec sample-and-hold effect per frame).
1/48 = time period from T+0/48 to T+1/48sec (two flicker)
2/72 = time period from T+0/72 to T+1/72sec to T+2/72sec (three flicker)
3/96 = time period from T+0/96 to T+1/96sec to T+2/96sec to T+3/96sec (four flicker)
The count of X flickers per movie frame, creates a sample-hold effect of (X-1)/(X * 24) seconds, assuming the length of the flicker is insignificant. Otherwise, the math becomes: ((X-1)/(X * 24) + length of flicker). For CRT, especially short persistence, the length of flicker is an insignificant variable (far less than 1ms).
It's simple mathematics.
Historically, many people's definitions of "judder" / "stutter" can vary, so to clarify, for the purposes, of this *specific* reply:
As an extreme case, I define plain movie "judder" as the sense of motion not looking as smooth/fluid as 60fps@60Hz (e.g. sports on CRT -- perfectly fluid motion, no blur, no stutter, no judder).
Regular 2:2 pulldown obviously has less "judder" than 3:2 pulldown, but even consistent frame repeats are still a sensation of judder, by the specific definition of judder as defined by this specific followup (movies not as smooth as sports on 60fps@60Hz CRT). Also motion blur is not the same thing as judder, but motion blur can make judder harder to see, so a converse argument occurs -- 24fps@48Hz CRT would have more judder than 24fps@120Hz CRT because of the reduced of motion blur of doing 24fps@48Hz CRT. However, that depends on how one interprets the word "judder" which I've seen defined differently by different people (aka 3:2 pulldown judder is completely gone from 2:2 pulldown -- yes, THAT flavour of judder). But the dictionary definition of judder is consistent with the judder I see even in consistent pulldown (e.g. 2:2 pulldown), even on sample-and-hold displays. It just isn't as fluid as 60fps@60Hz CRT; and it's not just because of motion blur; and there is still a vibrating-moving-edge effect, so for the purposes of this specific reply, "judder" is the correct word.
IMPORTANT NOTE #2:
I stuck to impulse displays that do one flicker per refresh. I excluded mentioning plasmas because comparing objective vs subjective is more complex with them. Motion blur math is far simpler with LCD's and CRT's and easier to correspond with subjective observations. Plasmas are more complex to interpret (in terms of math and physics) due to the subfield refreshes, which also essentially behave as a form temporal dithering too as well. The earlier talk in this specific post also generally applies to other flicker display but the math can get complex (e.g. plasma subfield refreshes are tantamount to repeated flickers, BUT some plasmas also motion-interpolate the subfield refreshes, BUT on many plasmas some subfields are brighter than others, AND some of them compress the brighter subfield pulses into shorter time periods - e.g. Panasonic FFD plasmas). As a rule of thumb, plasma flickers are "on average" a somewhat longer sample-and-hold effect than CRT, but a far shorter sample-and-hold effect than most LCD's. This is because clusters of subfield refreshes are brighter than others, and those specific subfield refreshes will be the dominant determinator of motion resolution for that specific said plasma display for the specific said material that is currently being displayed at that specifically said exact moment, as the number of bright subfield refreshes per actual refresh, can vary very widely from refresh to refresh. Plus, the red/green phosphor decay is worse than CRT. Thus, it greatly complicates motion blur mathematics. The judder / stutter / motion blur mathematics of CRT versus LCD is far, far simpler than with plasma as a result, due to the extremely complicated differences in motion quality between different plasmas (e.g. the advanced concept of motion-interpolated subfield refreshes used by Panasonic VT50, combined with clustering of bright subfield refreshes by its Focussed Field Drive). For simplicity, I limit to pure sample-and-hold displays (e.g. LCD at 100% brightness), and pure strobe displays (e.g. CRT's, LightBoost, Trimaster OLED, etc) as the motion blur mathematics on these displays are relatively simple where the objective math/measurements virtually exactly corresponds to subjective motion blur observations.
Originally Posted by Richard Paul
The duty cycle of a display affects motion resolution but I don't think it affects the even repetition of frames.
Repetition of frames create a sample and hold effect on impulse-driven displays. The fewer the repeats (down to showing the frame only two times -- a single repeat), the less motion blur and the judder amplifies (by this post's specific definition of judder). The more repeats, the more motion blur occurs (and its somewhat masks judder, by this post's specific definition of judder). Once you go down to 1 flicker per refresh (24fps @ 24Hz on a CRT), the motion blur and judder is completely gone but it is an eye-annoying 24Hz flicker.Edited by Mark Rejhon - 7/6/13 at 9:25am