What affects a camera's "motion cadence"?

[Jonesy Jones]

So in another thread about the newly released EVA1 footage, discussion came up about it's "motion", in that it doesn't have very cinematic motion. First impressions suggest that it has great color, but poor motion. Which led me to start wondering, what specific tech affects motion?

Long GOP vs All-I. My understanding would be that this probably affects motion as much as anything. However, to my knowledge, the EVA1 is shooting ALL-I. 

Secondly, rolling shutter. I have no idea on the rolling shutter specs for EVA1, but to my knowledge the GH5 is somewhere around 15ms, and I can't imagine the EVA1 would be worse. I didn't notice any poor rs problems in any of the footage. 

The original BMCC cam for instance, has ALL-I prores and I believe even worse (than GH5) rs, but is popularly described as having very cinematic motion cadence. 

So what else is there? How can we more clearly put our finger on the mystique of motion tech?

EDIT: I failed to mention fps as another factor. Seems so obvious it skipped my mind. Let's assume 24 or 25 fps. 

[tugela]

Rolling shutter can't be worse than the frame rate, its physically impossible.

[cantsin]

"Motion cadence" is a term that only exists in camera forum discussion threads - it's neither a technical, nor a scientific term.

[jcs]

It would appear the primary factor in motion cadence would be the clock / sampling interval. If the sampling interval is perfect, e.g. each frame sampled at very close to 1s/23.976, that will have different a perception vs. a system with temporal jitter. It would seem film would have some temporal jitter, vs. a digital system would could be a lot more precise. The question is what 'looks better' and if temporal jitter is helpful, how would a digital camera implement it? (statistical sampling: average interval, variance, magnitude etc.). Likewise, if temporal jitter is pleasing, why aren't there post tools which specifically address this? (I've seen film effects which 'damage' footage, however nothing so far that subtly creates temporal jitter based on specific camera systems in the same way e.g. Film Convert works for color/grain).

With a precise motion pattern, cameras could be measured for temporal jitter / motion cadence (it would appear that cameras with genlocks could be jittered live).

[cantsin]

Oh dear, please don't introduce another un-/pseudoscientific term like "temporal jitter". (Jitter, btw., is temporal by definition.)

EDIT: Maybe we need to write a present-day "Dictionary of Received Ideas" for camera forum culture, with entries such as "motion cadence", "3D-pop", "full frame look" etc.

[Ed_David]

Well, I mean,

has this group ever been the cutting edge bastion of science? Or is it, just a bunch of filmmakers and shooters trying to help each other with tips and tricks to improve?

And if one person asks the question, and another person tries to help and answer it, should you, dear friend, Can'tSin, be upset?

These are great mysteries of life.

If you want true, science-based research and answers, go on cinematography.com and ask Art Adams or David Mullen.

If you want to have a little fun, stay here.

[Jonesy Jones]

50 minutes ago, jcs said:

Likewise, if temporal jitter is pleasing, why aren't there post tools which specifically address this?

This is exactly what I was wondering. Might be a simple expression  for After Effects like wiggle with very low values. I'm going to experiment. 

[cantsin]

8 minutes ago, Ed_David said:

If you want to have a little fun, stay here.

Good point, Ed! All is fine if this is clear enough to people who google "motion cadence" and get two links to this forum as their top hits.

[jcs]

Jitter also applies to positional variance... Temporal jitter makes it clear, and is something we can measure or create in post WRT to the idea of motion cadence.

[dbp]

CMOS vs CCD as well? For all of their flaws, I remember the HVX200, HMC150 Panasonic cameras having nice motion cadence.

 

[Jonesy Jones]

3 minutes ago, jcs said:

Jitter also applies to positional variance... Temporal jitter makes it clear, and is something we can measure or create in post WRT to the idea of motion cadence.

 Good idea. That's one more thing for me to experiment with. 

 But also begs the question as to why some cameras, take black magic for instance, have such a good feel of motion, when it's  very doubtful that they are playing with temporal values in camera. 

One more question, should we all be shooting at 24P and distributing at 23.98?

[jcs]

12 minutes ago, Jonesy Jones said:

 Good idea. That's one more thing for me to experiment with. 

 But also begs the question as to why some cameras, take black magic for instance, have such a good feel of motion, when it's  very doubtful that they are playing with temporal values in camera. 

One more question, should we all be shooting at 24P and distributing at 23.98?

Maybe those cameras have less temporal jitter? Or the variance is Gaussian random around the average/center, vs. non-random drift-and--snap. It's something that could be measured...

23.976 for everything except theater in 60Hz countries. Notice how motion looks much better on embedded devices vs. browsers on desktops (e.g Apple / Fire / Roku / HDTV). So called hard real-time systems guarantee consistent frame rates. So far I haven't seen desktop browsers come close. A dedicated desktop app using the GPU synced to the hardware  refresh, as with a video game, could come close to a hard realtime system.

Thus it's helpful when studying motion to understand where the temporal jitter is coming from.

[Drew Allegre]

I always thought it was the Phillip Bloom woodgrain pocket dolly that gave that authentic organic look to footage.

It makes sense to me that the playback device is a pretty major factor in perceived cadence.  Once you eliminate variables: playback device, frame rate, shutter speed/angle, rolling shutter, jitter (introduced at any point in the chain), scene and lighting, codec and compression, is there some "thing" that makes one sensor or camera subjectively better than another?  It almost seems like audiophile territory trying to define the marginally perceptible, but I think many of us would agree that there is definitely something there to perceive and define.

Gimme a good motion cadence mixed with that Canon color science.

[tugela]

3 hours ago, jcs said:

It would appear the primary factor in motion cadence would be the clock / sampling interval. If the sampling interval is perfect, e.g. each frame sampled at very close to 1s/23.976, that will have different a perception vs. a system with temporal jitter. It would seem film would have some temporal jitter, vs. a digital system would could be a lot more precise. The question is what 'looks better' and if temporal jitter is helpful, how would a digital camera implement it? (statistical sampling: average interval, variance, magnitude etc.). Likewise, if temporal jitter is pleasing, why aren't there post tools which specifically address this? (I've seen film effects which 'damage' footage, however nothing so far that subtly creates temporal jitter based on specific camera systems in the same way e.g. Film Convert works for color/grain).

With a precise motion pattern, cameras could be measured for temporal jitter / motion cadence (it would appear that cameras with genlocks could be jittered live).

Why would anyone want film to look flawed in the way it was in the early days due to imprecise mechanical effects? That has nothing to do with quality and is just an affection to some sort of antique appearance. Sort of like some furniture is deliberately "distressed" to make it look antique or rustic even though it is not.

[Jonesy Jones]

39 minutes ago, tugela said:

Why would anyone want film to look flawed in the way it was in the early days due to imprecise mechanical effects? That has nothing to do with quality and is just an affection to some sort of antique appearance. Sort of like some furniture is deliberately "distressed" to make it look antique or rustic even though it is not.

 My understanding of what JCS is saying is that all film, even modern, is not as precise as digital. In other words, film is mechanical, not electronic, and therefore there are some extremely subtle variances to the way it records.  These variances could be both temporal and positional. Unlike digital which will be perfect.  I don't think we're talking about massive skipping and stuttering, I think we're talking about nearly imperceptible flaws that make a film feel alive the way digital, video, often does not. 

[jcs]

1 hour ago, Jonesy Jones said:

 My understanding of what JCS is saying is that all film, even modern, is not as precise as digital. In other words, film is mechanical, not electronic, and therefore there are some extremely subtle variances to the way it records.  These variances could be both temporal and positional. Unlike digital which will be perfect.  I don't think we're talking about massive skipping and stuttering, I think we're talking about nearly imperceptible flaws that make a film feel alive the way digital, video, often does not. 

Where people prefer the motion cadence of one camera to another, the difference would be the sampling interval. Is it precise, right-on every frame, or does it jitter, and if so is it constant, variable, is the magnitude constant, variable, is it uniform/Gaussian random etc. It's interesting with all the tests folks have posted over the years, I don't recall seeing one on motion cadence / temporal and/or spatial jitter. Motion picture film has spatial jitter during playback in the theater, so even digitally acquired material transferred to film would have playback jitter. Via testing one could discover how much, if any jitter helps make the viewing experience more pleasing or not.

I'm personally am not a fan of large amounts of jitter, however perhaps small amounts might help create the illusion of being more organic. Or maybe the opposite is happening, cameras with little or no jitter are what people prefer when they say they like the motion cadence of a camera. A scientific test with perhaps a very accurate strobe could be used to test cameras. Another way would be to introduce jitter in a completely synthetic test with computer graphics/animation. I do know from video games running at constant- vblank-sync'd 60fps, a single frame drop is massively noticeable and used to be a criterion to fail a game in QA back when there were no online updates available. Similarly, if a game is running with constant variable jitter, a single frame drop or game slowdown can be barely noticeable.

Found research papers on temporal jitter for video: https://www.google.com/search?rlz=1C5CHFA_enUS508US508&q=temporal+jitter+video+quality&oq=temporal+jitter+video+quality. Quickly skimming the conclusions, I didn't see a clear idea presented beyond what I mentioned from video game experience: constant temporal jitter is better than perfect timing with occasional jitter/dropped frames.

Temporal jitter used to remove rendering artifacts:

24Hz judder can be an issue with panning, perhaps temporal jitter can provide a kind of temporal anti-aliasing, resulting in less apparent judder. This would make sense based on the 'perfect 60fps highly noticeable frame-drop video game' effect. This could be tested by adding jitter to juddering video to see if it helps hide the judder.

Another factor could be effective motion blur with a digital sensor (may not be precisely related to the camera's shutter setting). This too could be scientifically tested.

[Matt Kieley]

The only thing "different" that stood out to me about the motion in the first EVA1 video were the shots that used a higher shutter speed. I have only watched a little bit of the short films that just came out but the motion looked normal to me. The 1/24 shutter speed (or 360 degree shutter angle) can also give an awful digital looking motion.

[jonpais]

Several times I've delivered videos at 24p that were recorded at 23.976 fps and vice versa: not once has anyone pointed out that anything was amiss. The only time anyone noticed something was a video where I shot at the wrong shutter angle. And several of those videos even included speech! So I'd say in my list of priorities, it's pretty low.

Edit: Ditto for all the 24p videos I’ve shot at 1/50th sec. 

[sondreg]

A thought that popped into my head while reading this - I know for a while back I was trying to understand why scrolling texts in premiere had such issues with jitter.

I discovered that it was due to premiere using only whole numbers only in their positional values (in cc2015), meaning if moved a text 0.5 pixelsteps per frame, it would only move 1 pixel per 2 frames. Which is OK. However the problem arises when the value doesn't end up as a whole number consistently, which would cause it to jump an additional pixel.

Such as moving at 1.2 pixels per frame

1.2 = 1 pixel moved
2.4 = 1 pixel moved
3.6 = 2 pixels moved (an inconsistency)

to easily fix this you could just add a small blur

I'd assume it's the same with motion cadence where pixels can't be on decimal levels like vector graphics - and how the position of the pixels in the algorithm of the camera determine what causes the motion cadence

[Laurier]

My guess it s it have to do with how the electronic curtain sync operate , the rolling shutter and the encoding.

It s a subjective observation but for example the motion blur on blackmagic camera is much more ...somewhat contrasted and apparent, whereas on sony/panasonic mirrrorless it s less visible.In camera noise reduction also remove some of the motion blur smoothness making it more hard and less pleasing.

But the motion perception is affected by your stabilization and your focal as well, I guess hand holding a tele lense on a alexa or a red epic won t look very nice if you compare it to a 35mm on a7r2 on a gimbals.