60 fps -- Production verses Playback Advantage

[bnhnet]

I am shooting a repetitive motion, something like waving your hand back and forth, where the clarity of the image at certain points in the cycle is important. The motion, when shot at 30 fps, is blurry at key points where the speed of the object in motion increases, and I am considering shooting 60p. That's a lot of file-size to absorb and I don't know if it's worth it. I normally shoot motion jpeg in a mov container at 720p 30 fps.

Are we just kidding ourselves with the 60p video frame rate? It's obvious that if we shoot 60 fps we reduce the blur caused by motions which cannot be captured clearly at 30 fps. Fine. But in the world of video, do we really see 60 fps when it is displayed on the monitor, or are the advantages of 60p limited to improvements during post-production where problems generated by blurry images foil special effects such as green screening, slow motion and motion tracking?

Do monitor cards and the monitors themselves allow/influence/control the frame rate of video? I recall when LCD monitors first came out there was a re-fresh rate problem which was causing additional blurring. Are we now to the point where 60 fps generates no artificats when viewed on LED and LCD monitors? In fact, do we actually see 60p fps during playback? Or, after a final rendering, do those 60 frames get interlaced and treated as 60 fields (as do the 60 fields in 30 fps NTSC video) and the resulting image we see is actually running at 30 fps?

If we shoot 60 fps and playback 60 fps I would expect to see a difference. But has anyone actually seen a side-by-side comparison test of a car traveling down the highway shot at 30p and 60p? Is the difference noticable? If the subject is realtively slow moving, e.g. a person walking, I wouldn't expect to see a difference, no matter what frame rate the production and playback. But only a slight difference. Is that correct?

In sort, does 60 fps (p or i) simply refer to the production media and has no bearing on the playback media, which means if the advantage isn't in post, it's little if no advantage at all?

[newpball]

You seem to omit one important other factor and that is shutter speed.

Take the lowest extreme, 24p. 24p generally looks great provided you use an applicable shutter speed.

However if you want a fast moving sports video 60p by itself is not enough you also need an appropriate shutter speed.

Also an increased shutter speed will come with a price as you obviously need faster lenses to keep the same amount of light reaching the sensor(s).

All that may be a non-issue if you use a nice Red camera with a $4,000 lens but if you have the latest consumer camcorder or consumer DSLR it may make a big difference.

Interlaced video takes a compromise between capturing more moments in time with the downside that those moments do not line up.

[racer-x]

If you have to ask this simple question, then it shows you have limited knowledge in film making of any kind. anything with speed will benefit from higher framerate. Lower framerate will help with compression, but will only look good with slow movement.

[bnhnet]

Thanks for the reminder, racer-x. I'm at an age where I'm not surprised by how little I know. And when combined with how much I have forgotten over so many different fields, I'm surprised I can think at all.

But the quick reply from newball jogged my memory. Shutter speed. Of course. How could I have forgotten about shutter speed? A more expensive, professional video camera. Of course. My consumer camera doesn't have control over shutter speed, so I wondered if there would be any noticeable advantage to shooting at 60 fps. And when I started reading about the artifacts that can creep into this combination of frame rate and shutter speed, I wanted to refresh my understanding. Now I think the answer is clear: don't shoot 60p but get a better video camera with shutter control. File size stays the same. Clarity improves. Wonderful.

[newpball]

Lower framerate will help with compression...

Actually the opposite is generally true, high framerate video compresses better than low framerate video.

[_Al_]

I just cannot resist, why don't you just point, shoot 60p and look at it on your monitor? You really get 60p on screen today.

Even very short shutter speed (cheap camcorder) could produce image that would be very fine with you. If not shooting directly at sun light, when auto feature could set really short time, and where it could introduce stutter. Perhaps not your case because you shot 30p in auto mode in the first place, those 60p could not be worse. If camcorder really shoots 60p and not some fake 60p (30p with every other copied frame) like some old cheap camcorder models could produce.

60p is advantage comparing to 30p, it feels much more real and fluent ...

[poisondeathray]

Yes there is a large real world difference in the final result , and the difference is more noticable the more drastic the motion. Do what _Al_ says and run a quick test.

Even with a person walking , you will notice a significant difference in the motion fluidity. 60p will look more "real" or a "soap opera" look. 30p will look choppy. Just watch a "60p" video and then convert it to "30p" by throwing 1/2 the frames (that's not quite correct, because the shutter speed will be different, probably 1/120 for the 60p video, 1/60 for the natively shot 30p video if you used "auto" mode)

In the USA, most LCD/LED most monitors run at 60Hz (some run at higher multiples , or some gaming displays run at different refreshes like 144Hz), so yes you can "see" the benefit and the difference. Do you ever watch TV? Do you notice how "smooth" something like a baseball or some other sports channel looks. That's because of 59.94 motion samples / second (some stations broadcast 720p59.94, others 1080i29.97, which is 59.94 fields / second)

p vs. i refer to progressive vs. interlaced . "60p" is 59.94 samples per second, but "30i" or "60i" (they actually mean the same thing, just different naming conventions, both mean 59.94 fields per second or 29.97 frames per second) . The "30" and "60" are just lazy abbreviations. The actual numbers are 30000/1001 and 60000/1001. The importance in terms of this topic is they both represent the same motion samples per second

The faster the shutter speed, the less motion blur, the sharper the images. This can lead to "strobey" look, as _Al_ metioned if you're not careful. Humans are used to motion blur. We've been accustomed to it because everything we've seen on TV, cinema, has natural motion blur. We're used to it. If everything else is equal, the more blur, the smaller the filesize. Sharper pictures in general take more bitrate to look a certain quality level

MJPEG uses intra only encoding . It encodes each frame separately, independent of one another. It doesn't use temporal compression, which take similarities and redundancies from frames before and after to reduce the bitrate. So doubling the frames/sec might roughtly double the filesize. But with temporal compression, using something modern like AVC, it might only be 10-30% larger on average. To give you a rough idea of a modern consumer camera that shoots 1080p59.94, they usually record at ~28Mbps . The 1080p29.97 mode on the same camera of many models usually records around 24Mbps, so that would only be about 16% more. Ignoring the (large) motion quality differences, on most cameras, the 1080p59.94 mode actually looks better per frame in terms of quality when you look at it frame by frame. A longer GOP size and higher compression settings are usually used.

[racer-x]

Here is a side by side comparison of 60p vs 30p. I shot the footage in 1080/60p. The left is original 60p, the right is converted to 30p via decimation by 2.

[_Al_]

That is artificially made 30p, if 30p is shot with camcorder, then shutter speed would be slower than recording 60p, thus different than 60p to 30p conversion, which seems to always be kind of strobey, more than as oppose it was recorded 30p right away.

But who knows, cheap camcorders can produce exactly that or even worse. Meaning shorter shutter speeds for your 60p original even calculated 30p.

[jagabo]

Here's an extreme case example of the difference bewteen 24p, 30p and 60p. The top row is 24p with 3:2 repeats to make 60p. It shows judder in addition to flicker. The middle row is 30p with 2:2 repeats. It shows flicker. The bottom is 60p. If you're viewing on a 60Hz TV or monitor it should be perfectly smooth. Watch it full screen for full effect.

30p can look a little better with motion blur but it's never as smooth as 60p.

[newpball]

Here's an extreme case example of the difference bewteen 24p, 30p and 60p. The top row is 24p with 3:2 repeats to make 60p. It shows judder in addition to flicker. The middle row is 30p with 2:2 repeats. It shows flicker. The bottom is 60p. If you're viewing on a 60Hz TV or monitor it should be perfectly smooth. Watch it full screen for full effect.

30p can look a little better with motion blur but it's never as smooth as 60p.

Sorry but the 24p example is absurd, nobody who knows what he is doing would render it that way.

With the proper shutter speed 24p looks just fine.

I suppose many people on this forum simply never watch 24p because it must be horrible for them.

Here is some basics for you folks, it's called motion pictures and the shutter speed of those pictures is a very important factor.

[_Al_]

WE are talking about camcorders here, possibly auto modes, not freaking motion picture. They have pro equipment and time to set scenes to make it work, set proper shutter speeds, and even with this, they cannot shoot whatever they want.

[newpball]

WE are talking about camcorders here, possibly auto modes, not freaking motion picture. They have pro equipment and time to set scenes to make it work, set proper shutter speeds, and even with this, they cannot shoot whatever they want.

Oh boy.....

[_Al_]

They need you on facebook man, here you are just retarded.

[bnhnet]

This is most intriguing.

To riser-x: I looked at your footage and want to make sure I understand what I'm seeing. Attached is frame 16. I'm looking at a split-screen of a waterfall where one camera shot at 60p, then the same camera switched to 30p. I am not seeing the same motion shot by two different cameras at the same time with different frame rates. The 60p footage was then converted to 30p using a process described by poisondeathray. Correct?


You are all correct in that there is a sense of smoother motion in the 60p, but the reason I am asking about 60p has additional ramifications, which I'll attempt to explain. Here's the short version, and you can tune out at any time.

1. I have developed a system for composing and performing music constructed with independent, layered tempi and meter. I know this means nothing to you at the moment, but the performance phase involves video. Composing and performing independent, layered tempi and meter is kind of the holy grail (at the moment) in the evolution of contemporary music. It has been done before, but with limited success and precision. I hope to push that development further with my process. I have tested my system. It works, and I am now in the process of perfecting it. A local university has provided the musicians for the tests. I don't know if any of you have attempted a multidisciplinary project, but it is loaded with learning curves and tends to suck up money and time like crazy.

2. In order to control the variable tempi and meter (hence the term "independent") during performance, I am synchronizing X number of computers using video. (Using a click track -- if you are familiar with that process -- is not an option for several reason.) If there are 4 musicians in the group, there are 5 computers synchronized via a hardwired LAN, each computer containing the video file for one of the musicians. The 5th computer serves as the Master/Server, the other 4 as Slaves/Clients. So, each musician is watching a video of their own part. No two videos are exactly alike in terms of tempo and meter.

FYI: Heretofore, complex music involved the subdivision of a common beat. You can think of this beat as the tick of a clock. That beat is either inherent in the rhythm being played or conveyed by the conductor. In this new system, there is no subdivisions of a common beat. Hence the term "independent." If there are 4 musicians performing, there are 4 independent clocks at work. And while each clock is independent of the other, they are all synchronized into achieving the desired effect. The possible variables with just 4 musicians/clocks gets very intense.

3. I won't even touch on the complexity of the composition phase, other than to say that all of the musical parts are controlled, or synchronized, to a resolution of 1/1000 of a second. That is, the composer must know where each note will be at any time to 1/1000 of a second, and can then join various musicians in the same tempo at any time, or separate their tempi at any rate at any time. And that's just the application of tempo. I won't even get into the additional complexity of independent meter.

In the case of 4 musicians performing -- and it could be 2, 10 or 50 musicians in the group -- the system involves the control of the 4 clocks working independently, yet synchronized and under the control of the composer, as then interpreted and presented to each musician in the form of 4 (or X number) synchronized video files.

4. To synchronizes these 5 computers with their 4 musicians, I produce a video which shows a graphical representation of the tempo and the meter. Although I'm now in my 3rd variation of this element, you can think of the video subject as a metronome with its cyclic arm, sort of like the wipers on a car window. Actually, the visual doesn't look like this at all, but that will give you an image with which you can grasp the concept.

5. Video of the metronome shot at 30p means that the resolution, i.e. the precision during the performance phase, has been reduced from 1/1000 of a second to 1/30 of a second. That's a big loss. I'd like to get it closer to 1/60 of a second. It would make synchronizing the musician's parts more precise. Hence my interest in 60p.

6. Here's the rub. Several packages of software/hardware are available which allow the synchronization of X computers using video moving at 30 fps. Some developers prefer Mac OS because of the different way the OS handles the media. (How this makes a difference I am not yet clear on, but I can tell you that cross-platform developers constantly complain about the Microsoft-compatibility bugs inherent in this process.) Other developers have gone the PC route and developed external pieces of equipment to be attached to the computers to improve the reliability. I think you can imagine that 100%, rock-solid frame lock, lasting at least 10 minutes, is mandatory for this system to work.

So now I have the problem of keeping the video file size small enough, and the decompression fast enough, to ease the burden on 5 computers talking to each other and adjusting their frame rates while working at 30 fps.

7. Shooting and editing at 60p is, technically, not a big problem since we now have video cameras and editing software that will accommodate this speed. But unless poisondeathray's comment about using gaming monitors proves feasible, the final product would still have to be reduced to 30p. When i first started this project, I thought that a resolution of 1/30th of a second would be fine enough and would actually exceed the perception of a musician to hear and react. It turns out that 1/30th of a second is at the very minimum threshold and should be moved to 1/60th of a second. This threshold was mentioned by a 20th century composer working at Princeton by the name of Melton Babbitt.

8. In my development of this system, I was surprised by the amount of processing power it takes to keep just 5 computers synchronized with 100% reliability. I kept throwing more money, bigger CPUs (multi core dual i7's), and more RAM at the problem, and still couldn't get 100% performance reliability, and that was using video playing back at 30 fps. It turns out that the different synchronizing packages have different preferences for compression. For example, one system forbids using temporal compression to reduce file size. Spatial compression, e.g. ProRes, is the only format recommended. Other packages have no problem with using temporal compression schemes, and that's one I am investigating now. In order to keep file size down, I reduce 1280/720 to 640/480 and then apply the compression codec.

Sorry to bother you with all these details relating to some obscure field in which you have no interest, but I can assure you that each piece of knowledge you impart is valuable to me because I have to apply it down the line toward a process that is filled with deadly variables. So, clarity and smoothness of the metronome's waving arm is important to my process. Since I may not be able to improve the resolution of playback accuracy to 1/60 of a second, I am looking for ways to at least clear up that moving arm at 30p while not losing sync between the 5 computers. It's as simple as that.

Yes, I will have to experiment with a camera that can shoot at 60p and has shutter speed control. This project has been a money pit and I'm trying to use as little of that resource as possible. Any other suggestions for me to investigate would be appreciated.

[poisondeathray]

You are correct - It looks like racer-x didn't align the left/right views in terms of time, it looks like they are offset by 30 frames. Here is an "aligned" version attached below. To be clear, that is 1 camera shot at 60p. The 30p view is "faked" by dropping half the frames. The shutter was not adjusted (same shutter setting as 60p) so it's actually "more clear" than it should be with a native 30p shoot

What software are you using to take screenshots ? The reason I ask is it looks "blurred" compared to what it should be. Perhaps your viewing software is contributing to your "blurriness" considerations? This is what it should look like on frame 16. Notice it's less blurred compared to your screenshot ? In fact, nowhere in his original clip could I find your screenshot. It looks like your screenshot was blended some frames together


If you're limited by CPU/ hardware/ sync software limitations and must record 30 fps, the only way to reduce the blur is to use a faster shutter speed. It's the exact same concept as shutter speed photography, so you can look at some photography examples to see the difference e.g. here is a random one (just look on google for shutter speed comparisons) . The problem is most consumer VIDEO cameras won't have as many manual settings as photography cameras; in fact most consumer level video cameras won't even let you control shutter at all. But indirectly, a higher frame rate will have a proportionally higher shutter speed on "auto" mode on those consumer models



Can you list some of the sync software applications titles you are using ? I'm just curious

[racer-x]

Yes I quickly put two versions together without properly aligning them, I didn't think it would matter at the time.

After reading your intentions, I agree that a high shutter speed will help in your situation. Add as much light as you can to achieve a high shutter speed. The images will be clearer and allow you to focus better on movements. You can also look at the audio waves to better time the synchronization.

Sounds like you have a tough road ahead.

[bnhnet]

This is in response to poisondeathray's question about syncing software.

Zach Poff has a free program called Multiscreener available for download on his website. Just google his name or the product. Poff recommends shooting with the mov container and using spatial compression only. You can read his detailed instructions on his website, and I have followed them religiously. I have used Multiscreener on a couple of projects and it worked well. But lately, especially when synchronizing more and more computers in the single LAN, it started exhibiting a hiccup (my term) where one of the computers would lose sync and have to stop a few seconds before it could re-sync. Poff mentions this chase-and-catch-up feature and it does work exactly as described. However, in my application, I can't afford any disruption to the synchronization. Poff's last comments were that improvements to his system would be stalled until MAX/MSP improves its efficiency. Since then I have been playing with ArraySync. The developer indicates he's going to switch to Windows Media for his PC version rather than use QuickTime. I'm eager to see that version. There's also a nice product under the name of Dalton Watchout and some others that get very pricey like CoolLux. Go to Poff's website. He's got a lot of interesting stuff there, and most of it is free.

[jagabo]

Sorry but the 24p example is absurd, nobody who knows what he is doing would render it that way.

As usual, you have no idea what you're talking about. Everybody who watches 24p material on a 60 Hz TV or monitor sees exactly that. Obviously, that video was designed to accentuate the problem. But you can see it in movies all the time.

[newpball]

Sorry but the 24p example is absurd, nobody who knows what he is doing would render it that way.

As usual, you have no idea what you're talking about. Everybody who watches 24p material on a 60 Hz TV or monitor sees exactly that. Obviously, that video was designed to accentuate the problem. But you can see it in movies all the time.

Well then, that's funny, I am watching this pure 24p video with moving circles and I do not see anything like your top example, not at 48hz and not even at 60Hz.

[budwzr]

Phew! Ya'll "smartfellers" are extrapolatin' out to infinity. hehehe. "Smartfeller", hehehe, you get it? It's a anagram.

[jagabo]

In any case, as has already been stated several timesw, frame rate isn't the OP's primary issue. Shutter speed is.

[Cornucopia]

Yet with video, shutter speed is inextricably related to framerate. Seems to me long-term synchronization and jitter reduction are the primary working points. This kind of thing shouts for hardware assistance.

Scott