Linus Tech Tips Video Production

[Sparky49]

At this point I am worrying that Simon is not _actually_ trolling people....

[Nominal Animal]

Yes, Sparky49, anyone disagreeing with you and the holy knowledge you have, must be a troll.

It never ceases to surprise me how stupid human beings can be, no matter how knowledgeable or thoughtful otherwise.

Please, do look up the Dunning-Kruger effect.

[Sparky49]

Yes, Sparky49, anyone disagreeing with you and the holy knowledge you have, must be a troll.

What makes you think this? There have been multiple times where I have acknowledged flaws or lack of depth in my knowledge, unclear writing, etc. What makes you think I think my knowledge is 'holy'? I have even thanked you for your input during this thread, when you posted the link for a comparison which Simon was suggesting. He had said that if one were to fill a 43" 4k display with alternating black/white pixels, one could not observe them at 1m. After your helpful link, I rebutted saying that whilst I have no 4k display, I could very easily see the dots on a 1080p, 22" (much smaller) monitor and that effects such as aliasing were noticeable. I then suggested that it is therefore not "totally pointless" to consider people may want to invest in higher resolution displays, especially for a large monitor.  Since then, having tested the same on a 4k monitor in my lab, I can still distinguish details. Who is to say that people may not want a higher resolution in the future? Just because one person, cannot notice a difference does not mean that other people can't, especially as how we consume and interact with information has/is/will change. Coming up with more hyperbole like "We live in a brain dead society", or talking about people posting political rants online, or looking down on people who are blinking an LED with an arduino seems to be dragging the thread around wildly rather than discussing how this youtube channel may use its staff for video production...

Please, do look up the Dunning-Kruger effect.

I am well aware of this, hence why I try to reflect on what is being said to during a debate. To try to avoid such traps I always consider that others may have a difference perspective, view or feeling to my own. It is interesting to consider that freely utilising quotes like "bread and circuses" to put down other people because of a particular style of show they watch, might well be an indication of Dunning-Kruger. Some of the smartest professors I have worked for, enjoy consuming 'simple' media.

I am genuinely sorry, and apologise if I've frustrated you.

[Nominal Animal]

Yes, Sparky49, anyone disagreeing with you and the holy knowledge you have, must be a troll.

What makes you think this?

You implying that him disagreeing or making unexpected arguments or observations, makes him a troll.  I am not saying he is right, because I do not know whether he is right or wrong, and am interested in arguments either way -- well, not the arguments themselves, but the basis, the observations, that people back their opinions with.

What effect do you think a comment calling one party a troll has on the discussion?  Not positive, that's for sure.  (It is a sensitive point for me, because I don't mind posing as the devil's advocate just to see the basis for the arguments, as those are what I find interesting.  Instead, I've gotten ridiculed for even entertaining some positions -- including in physics; the mechanism of how hot water freezes faster than cold water being a particular sore point.  I got laughed at about a decade ago for that; but now, current best understanding shows I was not that far off with my opinion.)

For what it is worth, I do not know the truth of the matter here, and do not consider myself any kind of an expert, even though I have read a few dozen peer-reviewed articles on psychovisual modeling, human perception, and so on.  My own fault is, I have very little practical experience.  When I've experimented on other people, they've been either friends or colleagues, so may have provided artificially positive feedback.  So, aside from my own experience, my opinions are based on theories, not tested in practice.  And practice trumps theory.

Now, if the comment was made in a joking manner (say, to highlight the fact that even though he is a moderator here, he occasionally gets corkscrewed into the topic just like everyone else), it completely escaped me, and I apologize.  Me fail English often.

[Sparky49]

You implying that him disagreeing or making unexpected arguments or observations, makes him a troll.

Sorry, I was more frustrated by what I thought were ignoring arguments, diverting topic, ridiculous hyperbole, etc.

Now, if the comment was made in a joking manner

In all honesty, it was not. But I agree it's hardly conducive to properly discussing the topic of the thread. My bad, I normally try to restrain myself but snapped here. Apologies.

[Nominal Animal]

IMHO, anyone who can admit personal fault/error/miscommunication, is okay in my book.  So I withdraw my accusation above, and replace it with something like "well, the matter at hand is complicated, and not at all clear-cut as far as I can see; so, a certain amount of bifurcation of the topic is par for the course, and not trolling."

[rs20]

As (mostly) engineers, I think we have a tendency to want to ascribe digital-camera-like specifications to our eyes; as if there's some sort of Nyquist cutoff style criterion we can express about the "resolution" of our eyes or a hard "frame rate" limit. I think it's interesting to look at frame rate as an analogy. Consider videos playing at different frame rates:

10 fps: Looks obviously choppy to any viewer
24 or 30 fps: "This doesn't look choppy any more. This seems fine."
60 fps: "Woah, OK this actually looks like real life now rather than just a video"

I remember way back in the day when 60 fps was just a dream (at least as far as consumer hardware was concerned) that people got trapped in the fallacy that 30 fps doesn't look choppy, therefore we've already passed our human nyquist threshold or whatever (this part could be explicitly stated or just implicitly assumed), therefore there's no point in 60 fps. Such advice was, as is obvious in hindsight, profoundly wrong. And even at 144 fps, moving images on the screen look blurrier than they could, leading to the introduction of (optional) backlight strobing. I'm not saying the latter is a good idea or a sensible compromise, but my point is this: in some regards, the "frame rate" of the eye is less than 25 Hz. In other regards, even 144 Hz is insufficient for a completely lifelike experience. I know we're not talking about framerate in this discussion, but my only point is that y'all better have a damn good reason for equating "I can't explicitly distinguish the pixels" to "this screen is utterly perfect"; because analogous logic proved woefully ridiculous in the case of frame rates.

My 2 cents: The only other thing I'd mention is that for many applications (PCB design being a leading candidate), the fact that I can move my head back and forth is relevant. Having a single screen that I can get in close and peer at details using my head + eye movements rather than using a mouse is a very attractive prospect. Any argument which presupposes that the viewers' heads are bolted at a precise distance from the screen seem absurd from the start for that reason. Arguing that you should always be able to see the entire screen in your field of view is less absurd-sounding, but equally unjustified. Especially for computer monitors, which are used in a much more dynamic and interactive way than an TV screen displaying a movie.

[technix]

Consider videos playing at different frame rates:

10 fps: Looks obviously choppy to any viewer
24 or 30 fps: "This doesn't look choppy any more. This seems fine."
60 fps: "Woah, OK this actually looks like real life now rather than just a video"

If the Linus Tech Tips video validating this has any merit, for an untrained eye 60fps and 120fps in games can be differentiated. If you throw in some trained eyes they can see 144fps and in some cases even 240fps.

As i explained it is expected that a person with good sight can distinguish two dots 1mm apart at 1m

That is an angular resolution of no more than 35 pixels per degree, a far cry from, for example 59 pixels per degree for iPhone 4s at typical phone distances.

[wraper]

If the Linus Tech Tips video validating this has any merit, for an untrained eye 60fps and 120fps in games can be differentiated. If you throw in some trained eyes they can see 144fps and in some cases even 240fps.

IMHO it's more noticeable because of lower lag in games as a consequence of higher FPS rather than because of FPS at such.

[PlainName]

lower lag in games as a consequence of higher FPS

If I might diverge this thead a little, I wonder how significant that is. I am a game player (used to be pretty OK in UT) so kind of automatically didn't care to refute it, but thinking about this now, how significant is it?

The issue would seem to me two-fold. First, games used to 'tick' on vsync, so clearly having a faster refresh would make the game run faster. However, that surely doesn't apply nowadays and the screen is merely a snapshot window into where the game is at. So I would assume that with modern games and computers, we are talking pure reaction times - you see something on-screen and react to it, so the faster the screen can show something, the sooner you can react.

OK, so worst case, lets say we have a 60Hz refresh - that is up to 16ms before we could react to something. How bad is that? Fortunately, the interwebs lets us find out

https://www.humanbenchmark.com/tests/reactiontime

Seems the best we can hope for is 150ms, nearly an order of magnitude more than the 60Hz monitor might add. I wonder if 'gaming needs fast refresh' is a case of a workman blaming his tools.

 

[Nominal Animal]

So I would assume that with modern games and computers, we are talking pure reaction times

No; that applies only to jump scares.

It has much more to do with rhythm, jitter, and anticipation.  A much better test is to have an animation, similar to a shooting track with moving targets, and then add jitter or glitches.  It is surprising how small changes are noticeable (and actually cause frustration).

[PlainName]

It has much more to do with rhythm, jitter

I can see the potential for that, but wonder if it is actually significant. I think the amount of jitter, assuming the frame is completely free-running from the action code, would be a maximum of half the refresh rate, so 8ms or so.

[PlainName]

here is a interesting study done using starcraft

It's an interesting study, but I'm not sure how it is relevant - please quote the appropriate part (there are no page numbers on the link destination). What it does show, though, is that the game tick is more or less 10ms, and that players can compensate for loss of skills (thus, presumably, also reduced feedback from the game).

so what provides a better perception of an object on screen?

I think that might be going down the wrong tunnel. For me, increased resolution of a screen doesn't mean I can see, say, a photo in higher quality but that I can fit more windows and stuff onto it (and still be able to read the text). I would love an 8K screen just so I can replicate my three-monitor layout without having bezels get in the way. I accept that this is another way of saying 'more detail' but I think the end desire is important - a widescreen monitor, for instance, would be fab for seeing individual nose hairs in a movie but shit for multi-windowed code editing.

[Nominal Animal]

It has much more to do with rhythm, jitter

I can see the potential for that, but wonder if it is actually significant. I think the amount of jitter, assuming the frame is completely free-running from the action code, would be a maximum of half the refresh rate, so 8ms or so.

Assuming there is enough processing power to do that, and no external causes for jitter, sure.  (But the emerging theme in this thread is, IMHO, that maximizing any single feature like resolution, or examining any single facet out-of-context, is not realistic; we need to examine the entire situation before we can make any sensible inferences.)
In practice, the kind of jitter we experience, is an occasional, say 50ms or longer delay or missing frame, because the machine was busy doing something else; often I/O.

You might find Player Perception of Delays and Jitter in Character Responsiveness (PDF) by Normie Normoyle, Guerrero, Jörg interesting.  (It's the first interesting PDF paper that seems to be completely open access that I found after quick DuckDuckGoing.) Although their sample was small (just ten victims test players per scenario), I kinda agree on their conclusion on player performance: "... higher latencies affect path following while jitter does not. Conversely, jitter interferes with the player’s ability to time jumps or avoid lasers resulting in more deaths and health loss." with the latencies they tested in the 0 to 500 ms range, and jitter in the 0 to 150 ms range.

[PlainName]

Assuming there is enough processing power to do that

No, that's the worst case - 60Hz refresh and unstressed CPU. If the CPU is struggling then the jitter caused by the screen refresh - which this is all about -  would be less since there would be a higher number of refreshes per game tick.

the latencies they tested in the 0 to 500 ms range, and jitter in the 0 to 150 ms range

Order of magnitude greater than what we are looking at here. I think it's important to realise that when they discuss how jitter affects things, that's with jitter bad enough TO affect things. Do they say how bad it has to be before it has a measurable impact? Our theorised 8ms is barely off the bottom of their scale.

[thm_w]

If I might diverge this thead a little, I wonder how significant that is. I am a game player (used to be pretty OK in UT) so kind of automatically didn't care to refute it, but thinking about this now, how significant is it?

The issue would seem to me two-fold. First, games used to 'tick' on vsync, so clearly having a faster refresh would make the game run faster. However, that surely doesn't apply nowadays and the screen is merely a snapshot window into where the game is at. So I would assume that with modern games and computers, we are talking pure reaction times - you see something on-screen and react to it, so the faster the screen can show something, the sooner you can react.

OK, so worst case, lets say we have a 60Hz refresh - that is up to 16ms before we could react to something. How bad is that? Fortunately, the interwebs lets us find out

https://www.humanbenchmark.com/tests/reactiontime

Seems the best we can hope for is 150ms, nearly an order of magnitude more than the 60Hz monitor might add. I wonder if 'gaming needs fast refresh' is a case of a workman blaming his tools.

Doesn't matter what our reaction time is, lower input lag (via monitor, mouse, kb, etc.) will always be better. Its just a question of how much. Linus' small test showed significance, but was not terribly scientific, I'm sure other tests exist.
Think about two people see each other, they both have the same 150ms reaction time then click their mouse, but one mouse has a 20ms input lag. Its obvious who will have the advantage.

[PlainName]

but one mouse has a 20ms input lag

We're not talking input lag (and I chopped some discussion of why not from here - happy to post it if you're bored ). But if we were, I think you've missed that these would be separate machines and, hence, not synced to each other. As we saw from the previously posted paper, a typically useful (for testing) game has a 10ms tick, and a 60Hz refresh gives less jitter than that (I assume - anyone disagree?). That's ignoring stuff like propagation delay and round-trip times, etc.

I am talking here as someone who accepts that a higher monitor refresh can be noticed, by the way. I just can't get my experience to agree! I can remember when CRT with TV-rate refresh was pretty terrible, but my 60Hz monitors seem perfectly OK. I don't feel the need to splash the cash on changing them just for higher refresh. Maybe I am insensitive, but I've already mentioned my loathing of non-high refresh LED displays/indicators.

[KL27x]

The resolution ones eye can appreciate is called visual acuity. With my 20/15 vision (I had to look up Wikipedia to translate my own vision acuity report to American notation - it is 20/15 not 20/20) I can see a lot more minute details than someone with, for example, 20/30 vision.

No, this is totally false and in the vein of what you are saying is actually backwards. You may or may not have better clarity of vision than average. That is a different measure. Having 20/15 vision just denotes that you are a bit farsighted. This makes you worse at appreciating the resolution of a cell phone or computer screen than someone with 20/20 vision.

Someone who is moderately near-sighted will have advantage viewing a monitor. They will be able to position the monitor closer to their eyes while remaining in focus. And they will be able to appreciate more detail by using a greater area of their retinas to view the same area of monitor. You would need reading glasses to achieve the same view.

[Nominal Animal]

Assuming there is enough processing power to do that

No, that's the worst case - 60Hz refresh and unstressed CPU. If the CPU is struggling then the jitter caused by the screen refresh - which this is all about -  would be less since there would be a higher number of refreshes per game tick.

No, you're thinking of latency there (roughly constant overload).  Jitter is the random component; in this case, say an unrelated sudden task that delays the game by a couple of dozen ms or so.  It can be repeated, but at random intervals.

Our theorised 8ms is barely off the bottom of their scale.

I was only pointing out that it is not about human reaction time; much smaller jitter affects performance.  Even the smallest jitter they tried in that study, 50ms (1/20th of a second), is much faster than human reaction times (about a quarter of the typical reaction time to visual stimuli, 190ms or so), and shows pretty clear performance degradation in games.

I'm not claiming anything about 8ms jitter, because I don't know; I do know that it depends on the situation.  The most sensitive case, I think, is when anticipating the movement of a target moving at a constant speed.  If there is jitter, even at the ten millisecond range, I believe it affects the accuracy of the human tracking it.  I do not know for sure, though, and have not yet seen any research into that particular case.

(Perception-wise, it is similar to testing how accurately you can distinguish the frequency of two tones.  The interval between the two tones affects the sensitivity to their frequency difference; the easiest case being when they are played continuously.  The sudden change in frequency is easier to detect than the difference in frequency of two separate tones.)

[PlainName]

No, you're thinking of latency there (roughly constant overload).  Jitter is the random component; in this case, say an unrelated sudden task that delays the game by a couple of dozen ms or so.  It can be repeated, but at random intervals.

Whilst what you say is correct, it's the wrong context for what we are discussing (effects of monitor refresh). So, to clarify my comments, monitor refresh rate is constant - it doesn't vary with load. Scene generation is the thing that loads the CPU, so if there is too much to do and the CPU can't keep up, the game scene rate will lower (that's game frame rate, which is NOT monitor refresh rate).

OK, now jitter, for us, is caused by the monitor refresh occurring non-coincident with scene change, so something changes in the game but we don't see it until the next monitor refresh. That time could be 8ms @60Hz and that's our range of jitter in when the scene change will be apparent to us due to monitor refresh.

BUT... I've made assumptions here  and would appreciate someone in the know putting me right (or applauding my insight, of course!). I've assumed that the game scene updates are NOT tied to monitor refresh. Once upon a time I believe they were, using the vsync interrupt as the game tick (which also prevented screen 'tearing' when a change was made mid-refresh).

[PlainName]

Even the smallest jitter they tried in that study, 50ms (1/20th of a second), is much faster than human reaction times (about a quarter of the typical reaction time to visual stimuli, 190ms or so), and shows pretty clear performance degradation in games.

I agree that things small enough not to matter do, often, have an effect (but  homeopathy is still bollocks  ), and I keep mentioning that I suffer from seeing refresh rates that shouldn't be seen (apparently).

Reaction times are not important here since the player uses prediction, and it's that prediction that is thrown off by jitter. Did the paper give an actual range at which jitter had an effect, rather than the range they tested? 50ms is big, and if that's the low point at which it shows up then I don't see the 8ms of monitor-induced jitter being an issue (and, indeed, they wouldn't if they quote 50ms).

[Nominal Animal]

So, to clarify my comments, monitor refresh rate is constant - it doesn't vary with load. Scene generation is the thing that loads the CPU, so if there is too much to do and the CPU can't keep up, the game scene rate will lower (that's game frame rate, which is NOT monitor refresh rate).

OK, now jitter, for us, is caused by the monitor refresh occurring non-coincident with scene change, so something changes in the game but we don't see it until the next monitor refresh. That time could be 8ms @60Hz and that's our range of jitter in when the scene change will be apparent to us due to monitor refresh.

In practice, jitter is caused by unexpected delay in scene generation, typically because of storage I/O.  That scene will be eventually displayed, but much delayed from the point in time it was supposed to represent.  The study I linked to shows that even at 50ms (one twentieth of a second), the smallest jitter they tested, such a jitter affects the player results.

(Why didn't they test smaller jitter?  I'd say because it would have been very hard, because they used a game for the study.  If zero-jitter games were possible, that's what we'd have, right?)

If we look at the rendering pipeline, each scene represents the game state at some time in the past, unless the physics engine uses a predictor/estimator to estimate the real-world duration from the moment the previous frame became visible, to the moment the frame being drawn becomes visible; I do not believe any widely-used physics engines do this, since they tend to rely on fixed-duration game time steps for simplicity.  If there are no unexpected delays and the scene can be rendered within one frame, the expected jitter from the frame update along is about half a frame, so about 8 ms at 60 Hz update rate.  Almost all games use double or triple buffering, where the scene is rendered to an unseen canvas, and the displayed canvas switched at the next refresh interval.  This is also used when playing video without tearing.  This adds latency, but not jitter.

Current computers and gaming consoles are complex machines, and completely predictable timing is no longer possible. Although individual events are so short their effects to timing are insignificant, they often cause cascades that can lead to unexpected delays in scene generation.  I'm most familiar with Linux system, and there, storage I/O (disks, network) tends to be the main cause for these.  These random events cause any measured timing to have so much noise, that it is very hard for a predictor/estimator engine to yield a reliable estimate, probably explaining why such are not used in practice.

In games, scene changes (like rounding a corner that reveals new scenery) are also a cause for additional jitter, although there are various ways to mitigate that -- basically, preloading things in staggered order, so that there are no sudden scene changes, only incremental ones.

It is, in my opinion, silly to worry about that 8 ms, when there are larger factors in play.  Just like I claim that when creating videos, maximizing the camera resolution is not sensible, and that one needs to consider the whole situation and the effects each choice can have; including the content to be videoed.

I agree that things small enough not to matter do, often, have an effect

No, my point is completely different:  I'm saying you are concentrating on the wrong issues; that in reality, other things have larger effects.
First, that human reaction time is not involved, because this is about a different part of the visual perception part of the brain; something in the brain architecture, if you will.
Second, that issues with jitter in practice are not due to the display frame rate, but due to unexpected delays in scene generation.

If we were to look at why some players prefer high frame rates, I'd say that it has more to do with that motion anticipation part of the human brain working better with more frames per second, and nothing to do with how many frames per second they can perceive. Perhaps the extra information mitigates the effects of the unavoidable jitter?  Or perhaps it is completely psychological, a placebo effect?  I don't know.  I do know that motion anticipation and tracking occurs in a separate part of the brain than reaction, so I'd say it is plausible, but I'm definitely not sure.

[PlainName]

this article
http://news.mit.edu/2014/in-the-blink-of-an-eye-0116
talks about something along the lines of image recognition speed

This can be fascinating stuff indeed.

[PlainName]

a typically useful (for testing) game has a 10ms tick, and a 60Hz refresh gives less jitter than that (I assume - anyone disagree?)

Yes, I disagree 

On the one hand I say that 8ms is less than 10ms so the jitter is hidden within the tick period. But another way of looking at it is that 8ms is very nearly 10ms and could be interpreted as missing a complete tick. I doubt if that would happen because things would drift gradually rather than jump from one extreme to the other, but the possibility is there.

There is also an assumption that a monitor refresh is dead accurate, but nothing is. I can't recall ever seeing an error range stated, though.

OK, back to Simon for a new rant, I think

[Simon]

a person with good sight can distinguish two dots 1mm apart at 1m

That's a target, not a limit (as they tend not to say in speed awareness courses). And having just checked, it's a very easy to reach target too.

Whoa, I stopped getting notifications for some reason.

No it's not a limit, it's average vision. It was something I was told a long time ago and it was about the ability to distinguish the dots not neccessarily focus on the dots. I suppose the other way of expressing it is the shortest break in a line that one can clearly see. so at 4 PPmm I think we are fairly safe to say that that is a good resolution and much more is just an overuse of technology.