The Hobbit shot on Epic at 48fps

[kenneth gruyaert]

However, should you make any 24 fps version from the 120 fps original without resorting to blending frames, then the noise in the individual frames is going to become visible again.

If you blend 3 frames together you would have the same result of shooting 24fps with 216degree shutter, if you want 180degree shutter you would have to capture at 96, 144 or another multiple of 48fps to be able to use half the frames. Why would there be anything wrong with blending frames? It would give you the same motionblur and the same exposure.

There is the possibility that current sensor technology does cause a slight reduction in quality by shooting a given exposure time split into many shorter exposures, in theory with an ideal sensor though there should be no difference at all.
But one of the reasons that i think it would be very close is that increasing sensor resolution is also possible without significantly reducing low light quality suggesting the smaller pixels are capable of capturing the lower light levels with matching lower sensor readout noise which should mean the same would hold true when reducing exposure time. Also it simply is very obvious that if you combine many short exposures the end result will at least be much better than the individual short exposures even if it may not be exactly the same quality as a single long exposure.

Even if there is a slight reduction in quality the benefit of the higher framerate would be more than worth it. But considering that 24fps is usually shot with exposures using less than a 360degree shutter, the fact that the higher framerate capture could always use a 360degree shutter means it would capture more light which would compensate for any reduction in quality with the same amount of light, resulting in the same or possibly even higher image quality.

 

[David Mullen ASC]

You gain a stop of exposure by going from 180 degrees to 360 degrees, but that doesn't completely compensate for any exposure loss once you pass 2X 24 fps, i.e. 48 fps. Anything above 48 fps is more than a stop of exposure loss compared to 24 fps even once you switch to a 360 degree shutter angle. To simplify the math, shooting at 96 fps with a 360 degree shutter angle is still one-stop underexposed compared to 24 fps with a 180 degree shutter angle, which could be compensated by rating the sensor at 1600 ASA instead of 800 ASA.

But going back to "The Hobbit" they still have to also deal with another stop loss from shooting 3D with mirror rigs.

 

[SeanBrown]

So a solution for all the light lost on the camera end is more lights? Brighter lights for 3d?

 

[David Mullen ASC]

It's every thing... any combination of increasing ISO, opening up the iris (though you don't want to get too super shallow-focus with 3D), using more light, using longer shutter times, etc. For example, "The Hobbit" is using a 270 degree shutter angle instead of a 180 degree angle, so even though they are losing one-stop with 48 fps, they are gaining a half-stop by switching to 270 degrees, but then they are losing another stop by using a 3D mirror rig, but I don't know what ISO rating they are using. But also remember they are using the Epic, which can be shot at 800 ISO and higher, whereas the previous Tolkien films were shot on 35mm film, 200T and 500T Kodak inside, and probably the 500T was rated at 400 ISO at least, so they already have a camera that is practically twice as fast as what they used before. In the end, the light levels, despite the combination of 48 fps + 3D, may not be much different than with the previous movies.

 

[Matthias Hutter]

That doesn't negate the need to compensate for the exposure loss from using higher frame rates, it just suggests that underexposing the sensor -- i.e rating it at a higher ISO -- is one solution because noise may become less visible at faster projection rates (that is true of grain because it is in a physically different place on each frame, I'm not so sure it is true of noise, especially fixed pattern noise, but for now, I'll take your word for it.)

I agree that this only applies to random noise (e.g. shot noise) and precludes other sensor effects like banding or pattern noise. I haven't tested it side by side yet. I can only say that faster noise is less visible, and it for an ideal configuration 24p 1/48 and 48p 1/96 should be similar noise wise. But it's easy to test, and I may do that later:

1. Shoot steady scene
2. Same a stop underexposed
3. Play 1 normal and 2 with a stop gain at 200% speed and compare side by side.

 

[Julio Quintana]

I don't want to drag this debate out, but I would like to attempt to explain why I think David is right. If you properly expose a white wall and film it at 24 fps with a 180 degree shutter, when you project it at 24 fps the screen will look white. Now if you switch the camera to 120 fps with a 360 degree shutter, but don't open the lens or increase the light on the wall, the recorded image will be underexposed by a stop and a half, so the projected image will now look grey, even when projected at 120 fps.
Kenneth, it sounds like you're arguing that our eyes will somehow add up all the light from the dimmer individual frames and the sum exposure reaching our brains will be the same, but in reality a dim flashlight blinking at 120 fps will not be as bright as a bright flashlight blinking at 24 fps.

 

[Matthias Hutter]

@Julio: I'm not sure you understand completely. I think it's obvious that you need gain to get the same brightness level at higher frame rates (all else being equal).

Think of it this way: 1st camera is 24p 1/24, second camera 48p 1/48. If exposed for the first, the second is one stop underexposed. If you add every two frames of the second camera together, you get a 24p 1/24 with similar noise to the first. If on the other hand you gain the second camera by 1 stop, you get double the noise per pixel exposure, but it's only visible for the half amount of time.

 

[Julio Quintana]

@Julio: I'm not sure you understand completely. I think it's obvious that you need gain to get the same brightness level at higher frame rates (all else being equal).

Think of it this way: 1st camera is 24p 1/24, second camera 48p 1/48. If exposed for the first, the second is one stop underexposed. If you add every two frames of the second camera together, you get a 24p 1/24 with similar noise to the first. If on the other hand you gain the second camera by 1 stop, you get double the noise per pixel exposure, but it's only visible for the half amount of time.

Your explanation seems logical. Have you ever tested this?

 

[Gavin Greenwalt]

Yes you would have the same amount of light theoretically but you would need a sensor with a higher sensitivity to compensate. The "problem" with a real sensor instead of a mathematical sensor is that the 5 underexposed vs 1 properly exposed frames is that they will probably fall *outside* the sensor's sensitivity. There is a noise floor on sensors and it's not linear... hence "floor" below which you'll just get noise and nothing else. And as David mentioned you start getting fixed pattern noise a lot earlier than that.

For example if you had 5 stops of latitude below middle gray on your hypothetical sensor, and you lost 2.5 stops then you wouldn't still have 5 noisier stops you would have 2.5 stops below middle gray. So practically speaking, no you no longer have the same data just stored in a different manner, you would need to increase the brightness of your scene by 2.5 stops so that the dark range now falls where it used to to get the same dynamic range.

Just ask someone like Tom whether blended under-exposed frames are equivalent to a good time-lapse long-exposure feature on a camera.

 

[Matthias Hutter]

@Gaevin: As I said before, this is obviously only in an ideal situation, i.e ignoring other capture device effects (pattern noise etc ). I haven't experienced the noise parameters of a RED camera, so I cannot comment on that.

@Julio: I haven't ABX tested it yet. But it's very easy to do if you look at post 86.

Edit: So I did some testing. First thing I noticed that it's only reasonable to do it with RAW files without post processing.
Lacking a RAW video camera, the only thing I could do is compare stills.

All Images 450D RAW files with no post processing other than noted below. All ISO 200.

1. Native (100% crop) 1/5 s:
2. Two stop underexposed (1/20 s), 4 frames added together:
3. Two stops under + 2 stop post gain.

 

[kenneth gruyaert]

Nice comparison, thanks for sharing. At least the multi exposure picture is much closer to the long exposure than to the short exposure picture but i had hoped it would be even better.
I would like to see if doing the short exposures at iso 800 offers an improvement though before drawing any conclusions.

 

[Eki Halkka]

Nice comparison, thanks for sharing. At least the multi exposure picture is much closer to the long exposure than to the short exposure picture but i had hoped it would be even better.
I would like to see if doing the short exposures at iso 800 offers an improvement though before drawing any conclusions.

That's a good example - and i'd say that this small increase in noise is well worth the two stop increase in highlight protection, if such a thing is needed. It would be interesting to see what happens near pitch black - the longer exposure may have more advantage there, i'm not sure do the added frames actually bring back detail from the near black noise that is undetectable from a single frame... they could, in theory.

 

[Christoffer Glans]

Won't 48fps look alot like the 48p 720p mode in the 7D? Not talking about image quality I'm talking about motion. Personally I think it looks like "home video" speed, a "too fluid motion" which is more common in home video cameras.

Although I cannot say how this will look in 3D, where it might look awesome (24fps in 3D feels like it stutters a little). I also don't know how it will look on the big screen.
But if it's 48 fps on my TV in 2D, it's most certainly gonna have the fluid feel of home videos, it will look like the "120 Hz" televisions without the awful artifacts.

So I don't know how else it will look, how much different can the "motion" feel from something else shot in 50 fps? I recon you could just test shooting with the epic or red with a timebase of 48fps to check it out, but I feel quite sure it will look as I described, at least on a television set.

 

[David Mullen ASC]

I guess we will have to wait and see. Perhaps the 2D versions will be 24 fps conversions and the 3D versions will be 48 fps, so any of that "video-ish" smooth motion will help reduce stutter and perhaps even ghosting problems in 3D, while one could go and see the 24 fps 2D version for a more traditional experience. I've seen 2D 35mm film shot and projected at 48 fps and it wasn't as nice-looking as I had hoped it would be -- as you say, it suddenly made 35mm film look like interlaced-scan betacam footage. But maybe in 3D, the benefits will outweigh that issue.

 

[Deanan]

Although I cannot say how this will look in 3D, where it might look awesome (24fps in 3D feels like it stutters a little). I also don't know how it will look on the big screen.

The whole point behind 48 ( or greater ) is specifically for 3D and not 2D. The intention is to improve the immersive 3D experience
where 24fps strobing is much more disarming/distracting than in 2D. IMHO, in 3D, 48 becomes more dreamlike because you're able relax and get absorbed into the story rather than be jostled by the 24fps strobing which is doubled with stereo.

 

[Matthias Hutter]

Yeah I'm curious too. The high fps footage I've seen always felt 'faster', I don't know how to better describe it.

 

[kenneth gruyaert]

I think its ridiculous that something that is clearly better is considered inferior just because it was available on cheaper equipment for small budget productions long before it has finally come to big budget cinema production equipment. For some reason many people seem to have to rationalize even the flaws of something as being desireable just because it is expensive. Both the 24fps speed and the grain make film cameras clearly far inferior to new digital cameras, i cannot understand why so many people still consider immitating film camera's as the ultimate target for a digital camera. Film cameras are antiques, nothing more.

 

[David Mullen ASC]

Kenneth, you're ignoring personal taste -- it's like you're telling someone that they are wrong to prefer one thing over another thing, that their tastes are wrong.

There are a LOT of people who have enjoyed the 100+ years that cinema has given us, and that includes enjoying the look of those movies, and it is silly to tell people that now they should stop having a pleasurable reaction to these types of cinematic images just because there is new technology that is superior technically. It's like telling someone that they are wrong to like something hand-carved in wood when something machine-made is technically more flawless. We don't judge whether one process used in the creation of ART is better than another by whether it is technically superior -- there is an aesthetic component to the enjoyment of art as well.

Your comment about people "rationalizing" something to be desirable implies that they actually cannot possibly find it desirable, they somehow have to convince themselves to like it, when the opposite is true -- they have a natural "gut" reaction to 24 fps film photography that has nothing to do with thinking about how expensive it was or anything like that. They don't rationalize anything, they simply respond emotionally, immediately, directly to images that remind them of other cinematic images that have brought them pleasure in their lifetimes.

 

[Ketch Rossi]

The whole point behind 48 ( or greater ) is specifically for 3D and not 2D. The intention is to improve the immersive 3D experience
where 24fps strobing is much more disarming/distracting than in 2D. IMHO, in 3D, 48 becomes more dreamlike because you're able relax and get absorbed into the story rather than be jostled by the 24fps strobing which is doubled with stereo.

Agreed Deanan, and exactly why we are shooting "Rape of a Beauty" 3D @ 48fps...

 

[Deanan]

Kenneth, you're ignoring personal taste -- it's like you're telling someone that they are wrong to prefer one thing over another thing, that their tastes are wrong.

There are a LOT of people who have enjoyed the 100+ years that cinema has given us, and that includes enjoying the look of those movies, and it is silly to tell people that now they should stop having a pleasurable reaction to these types of cinematic images just because there is new technology that is superior technically. It's like telling someone that they are wrong to like something hand-carved in wood when something machine-made is technically more flawless. We don't judge whether one process used in the creation of ART is better than another by whether it is technically superior -- there is an aesthetic component to the enjoyment of art as well.

Your comment about people "rationalizing" something to be desirable implies that they actually cannot possibly find it desirable, they somehow have to convince themselves to like it, when the opposite is true -- they have a natural "gut" reaction to 24 fps film photography that has nothing to do with thinking about how expensive it was or anything like that. They don't rationalize anything, they simply respond emotionally, immediately, directly to images that remind them of other cinematic images that have brought them pleasure in their lifetimes.

Agreed. Many of the very same people that like 48 in 3D also love 24 in 2D. This is about what's best for a 3D experience independently of a 2D experience.