The Hobbit shot on Epic at 48fps

[kenneth gruyaert]

1 more question : shooting 48fps ...what happen for home video like bluray ??there is some sort of pulldown that create 60P off the 48P ?

I suspect they will simply release a 24p version for bluray since current bluray players do not support higher framerates. They would have to release new players and possibly firmware updates to allow for 48fps special edition bluray's which might actually have to be transmitted at 60fps to the tv since im not sure the tv's can handle 48fps input? I'm not sure all support 1080p60 input either so this may not be supported at all by some tv's.
The pulldown to 60fps would be far from optimal but still better than 24fps, they might simply do it like they often send 24fps over 30p broadcasts now: repeat every 4th frame.

I wonder why they are not shooting this movie at 120fps? Surely they have the budget for the memory cards and the processing power required to work with it.
120fps with 360 shutter would convert easily to 24fps 72 shutter by taking one out of 5 frames or to 24fps 144, 216, 288 or 360 by combining 2, 3, 4 or 5 frames. For old cinema's or standard blurays this will be as good as shooting at 24fps. 48fps would be less than ideal but the advantage of being able to do 60fps optimally at 180 or 360 shutter will more than compensate for this since most displays and projectors run at this speed. I dont know if 60fps projection in cinema's is likely to be as widely available as PJ is predicting 48fps to be but both are just temporary and they should start installing 120fps projectors as soon as possible anyways.
120fps is a genuine improvement the cinema's can easily offer that wont be available for home users for some time so why would they not take advantage of this opportunity?
Let's use these camera's to their full potential, framerate in combination with resolution is the main technological limitation which sets current camera's apart from future camera's and its pushing this limit that is largely responsible for the high prices, it is simply absurd to then only use 40% of it.

 

[David Mullen ASC]

Even with a 360 degree shutter, shooting at 120 fps would be nearly 2.5-stops more light needed over 24 fps with a 180 degree shutter, and probably their 3D mirror rig robs another stop of light, so that's almost 3.5-stops -- you'd be turning your 800 ASA Epic into the equivalent of maybe an 80 ASA camera. And simply dropping frames to get back to 24 fps would still give you the look of a 1/120th shutter speed, something in between the look of a 90 and a 45 degree shutter angle, unless you did some frame blending to increase motion blur. But I think that's less of an issue than the light loss problem. Besides, Jackson & Lesnie might not want to go that far away from the 24 fps look -- 48 fps might be enough for them to improve the motion characteristics in 3D without drifting too far from a traditional film look. The other problem is slow-motion shots - if 120 fps is your normal-motion frame rate, then you need to be able to get up to well-over 300 fps for truly slowed-down imagery.

 

[kenneth gruyaert]

Even with a 360 degree shutter, shooting at 120 fps would be nearly 2.5-stops more light needed over 24 fps with a 180 degree shutter, and probably their 3D mirror rig robs another stop of light, so that's almost 3.5-stops -- you'd be turning your 800 ASA Epic into the equivalent of maybe an 80 ASA camera.

What you are saying is true for the quality of the individual frames and for slowmotion video but for video played at real speed this is similar to saying that using a 5k sensor instead of a 2.5k sensor turned the EPIC into an 800 ASA instead of a 3200 ASA. If you look at a 2.5k crop on the 5k sensor this is true but not if you look at the total image. With a higher framerate camera you see the same total amount of light in the resulting images received by your eyes, perhaps because 24fps might be slower than the human eye there is a slight increase in noise going to 30 or even 48fps but after that it should stay constant. I think the sensor might cause more noise relative to the photon noise at lower exposures causing a slight reduction in quality but this would then also be happening with increased resolution and i think this reduction is very small compared to the reduction you suggest.
Ofcource if you want to pause the video in this form you would see a much lower quality image but something that might be called digital image stabilisation would resolve this (i started a topic about that: http://reduser.net/forum/showthread.php?60017-Digital-image-stabilization) I think that would be a necessity anyways since i do not think individually encoded frames is the future for video, i expect a video codec based on moving (and morphing) trackable image patterns (which could be surfaces or a shadow or lighting pattern etc...) which would be framerate (and resolution) independent. The idea that 120fps or even 1200fps video would require higher bitrates than 60fps is flawed for this reason, we do not see higher temporal detail than about 60fps so it does not need to be encoded. As i said in the other thread movements relative to the camera are irrelevant so it should not require more data to encode moving objects than it does to encode stationary objects (which are rendered perfectly at low framerates = low bitrates).

For the near term value of 120fps movies codecs based on this new way of encoding video are not required since its easy enough for cinemas to receive these high framerate films using standard codecs at very high bitrates.

And simply dropping frames to get back to 24 fps would still give you the look of a 1/120th shutter speed, something in between the look of a 90 and a 45 degree shutter angle, unless you did some frame blending to increase motion blur. But I think that's less of an issue than the light loss problem.

If you simply blend frames to end up with the same motion blur and total exposure as 24fps capture you would not lose any light.

Besides, Jackson & Lesnie might not want to go that far away from the 24 fps look -- 48 fps might be enough for them to improve the motion characteristics in 3D without drifting too far from a traditional film look.

I really dont like the idea of a film look and i dont expect it to stay to be honest, but something like that could be used to convey feelings like nausia or disorientation like is being done now with exagerated motion blurring or split ghost images and other effects. Another use might be intentional motionblurring on background imagery to shift attention to the desired object in the same way shallow depth of field is being used (there is the possibility for the object of interest to be moving across the screen which means it would require a high framerate and negligible motionblur to be sharp and the motion blurred background might actually be stationary on the screen which is the opposite of what a low fps camera captures natively). This is just to show i understand it is not always desired for everything to be in sharp focus all the time but i dont agree that this means these effects should be baked into the camera to be on constantly and i do not expect that to remain popular when the alternative is available.

The other problem is slow-motion shots - if 120 fps is your normal-motion frame rate, then you need to be able to get up to well-over 300 fps for truly slowed-down imagery.

There are faster camera's available now at similar prices as the top priced cinema cameras right? I think the main difference is that they are not intended to shoot long video's or to be very portable which makes processing and storage easyer. I would expect the speed of the high speed camera's to scale similarly to Cinema camera's in the future. Specialized camera's have always been required for slowmotion video so nothing would change it seems?
I think the EPIC might even have the potential for slowmotion full resolution video since the sensor can run at over 300fps so it might be possible to adapt it to be able to send that data to a separate storage system. Has there been any information about that?

Even if you dont have a camera to shoot slowmotion video for 120fps playback, that doesnt mean you should restrict the realtime parts of the movie to a lower framerate as well.

 

[David Mullen ASC]

Higher frame rates reduce exposure time, period. The resolution or sensor size isn't the issue. It's simple math. 120 fps + 360 degree shutter = 1/120th exposure time per frame, regardless of the sensor size. I don't think Jackson increasing the frame rate from standard 24 fps to 48 fps is an example of restricting the frame rate to something "low", it's already something twice as fast as normal!

There are practical issues involved in shooting 120 fps for 3D photography due to the increased light levels needed. Unless you started rating the Epic at 1600 ASA or 3200 ASA, the problem is needing to compensate for over 3-stops of light lost from the higher frame rate and the mirror rigs.

This makes no sense to me:

What you are saying is true for the quality of the individual frames and for slowmotion video but for video played at real speed this is similar to saying that using a 5k sensor instead of a 2.5k sensor turned the EPIC into an 800 ASA instead of a 3200 ASA.

Look if you are lighting a set and you switch your 800 ASA camera that normally runs at 24 fps with a 180 degree shutter to 120 fps with a 360 degree shutter, and then put a 3D mirror rig on it, you're going to have to increase the light levels by over 3-stops to shoot at the same f-stop. That's what I meant by turning your 800 ASA camera into a 80 ASA camera, because in practical terms, that's the sort of light levels you'd have to aim for.

 

[kenneth gruyaert]

Higher frame rates reduce exposure time, period. The resolution or sensor size isn't the issue. It's simple math. 120 fps + 360 degree shutter = 1/120th exposure time per frame, regardless of the sensor size.

Per frame the exposure is reduced but over the same time period it remains constant. Our eyes receive trillions of photons per second at random intervals, they do not come as frames, yet we see a high quality final image... Now if our eyes and brain would see at 10000fps that would be another story.
The point is that 1200fps noisy frame video should accumulate in our eyes to the same quality as 60fps video of which the individual frames had the light accumulated on the sensor.
This does mean its impractical to distribute the raw images of 1200fps video since the detail that accumulates over many frames is actually contained in the noise of the individual frames meaning they would have to be stored with near lossless compression.
I did explain in the other post that i do not think they should be distributed in this raw form except possibly initially for cinemas only.

I don't think Jackson increasing the frame rate from standard 24 fps to 48 fps is an example of restricting the frame rate to something "low", it's already something twice as fast as normal!

I do not compare things to how they used to be, i judge things objectively by thinking about how they should be. For now we can just aim to get as close to that as possible which unfortunately they are not doing. The EPIC is capable of 120fps which means that by today's technology standards 48fps is low. Like i said i think it is a real shame to waste the potential of the EPIC like this.
This is the end of longterm standards in cinema and television, we will have continuously improving formats as technology permits.

There are practical issues involved in shooting 120 fps for 3D photography due to the increased light levels needed. Unless you started rating the Epic at 1600 ASA or 3200 ASA, the problem is needing to compensate for over 3-stops of light lost from the higher frame rate and the mirror rigs.

I never mentioned 3D mirror rigs so thats another issue. I was only explaining why high framerates do not cause any light loss.


I explained the equivalence of increased resolution with increased framerate in a different way in the other thread, its something that requires some thought to understand, i dont see what else i can say about it.

 

[David Mullen ASC]

I never mentioned 3D mirror rigs so thats another issue. I was only explaining why high framerates do not cause any light loss.


I explained the equivalence of increased resolution with increased framerate in a different way in the other thread, its something that requires some thought to understand, i dont see what else i can say about it.

But increased frame rates DO decrease exposure -- that's basic photography. Doesn't matter if resolution is increased, or temporal resolution, or projection speed, whatever, the exposure is decreased. If you switched an EPIC from 24 fps at 180 degree shutter angle to 120 fps with a 360 degree shutter angle and do not change the ASA rating or the f-stop, the 120 fps footage will be darker, no matter what frame rate it is projected at. You seem to be implying that a filmmaker can switch from 24 fps to 120 fps photography without compensating for the exposure loss, which is not true.

 

[Ravi Kiran]

I don't know if this has been answered already, but how will projectors project 48fps? Do many projectors in use already have this capability? Is it simply a matter of a firmware upgrade or replacing a part or something?

 

[Jason Svatos]

I don't know if this has been answered already, but how will projectors project 48fps? Do many projectors in use already have this capability? Is it simply a matter of a firmware upgrade or replacing a part or something?

From what I've read it's only going to require a firmware upgrade.

 

[kenneth gruyaert]

But increased frame rates DO decrease exposure -- that's basic photography. Doesn't matter if resolution is increased, or temporal resolution, or projection speed, whatever, the exposure is decreased. If you switched an EPIC from 24 fps at 180 degree shutter angle to 120 fps with a 360 degree shutter angle and do not change the ASA rating or the f-stop, the 120 fps footage will be darker, no matter what frame rate it is projected at. You seem to be implying that a filmmaker can switch from 24 fps to 120 fps photography without compensating for the exposure loss, which is not true.

I do not understand how it is possible that if you read my post you would think i dont know that the exposure per frame is reduced?

Per frame the exposure is reduced but over the same time period it remains constant. Our eyes receive trillions of photons per second at random intervals, they do not come as frames, yet we see a high quality final image... Now if our eyes and brain would see at 10000fps that would be another story.
The point is that 1200fps noisy frame video should accumulate in our eyes to the same quality as 60fps video of which the individual frames had the light accumulated on the sensor.

I am aware that if you switch from 24fps 180d shutter to 120fps 360d shutter you will get only 40% of the light per frame which results in a noisyer lower quality image, but over the same time period of a single 24fps frame you get 5 of them giving you twice the amount of light and as a result the image quality would actually be improved. Youre capturing all the available light with a 360d shutter instead of throwing away half of it with a 180d shutter.

The point of my trillions of photons per second argument was that real life vision could essentially be thought of as single photon frame video at trillions of frames per second, they can be grouped together in any framerate for video playback, that could be a million fps with millions of photons per frame or a thousand fps with billions of photons per frame, it makes no difference if the technology is good enough to make it happen.

What you are saying is true for slowmotion video only which is what i said in my first reply.

 

[Steven Caesare]

... but over the same time period of a single 24fps frame you get 5 of them giving you twice the amount of light and as a result the image quality would actually be improved.

It doesn't work that way...

David is right.

-sc

 

[kenneth gruyaert]

The reason for high framerate video beyond 48-60fps is a very simple concept, if you capture blurred frames how could you possibly see a high resolution image? I am sure you are aware that you need shutter times much faster than 1/100th second to capture even slowly moving objects with good sharpness. Basically the only argument against needing higher framerates would be that it is impossible for a human to see high resolution on moving objects. If you look around in real life you should quickly realize that this is not true. Simply move around on your seat and see if you can still read text on your display? Or look at the cars passing by in the street and check how much motion blur you see on them. If you are simply walking around and look to the side can you not see objects clearly? It is possible to rotate your head or move your eyes to track an object that is moving relative to your body, our head and eyes are not locked into place causing blurring on anything that moves past us! This strange assumption that our eyes are always locked into place unable to move relative to the screen is what causes the whole idea that low framerate blurred video is good enough because our eyes would blur it the same way anyways. I have seen that argument so many times to justify even 24-30fps as perfection, it should be clear to anyone who seriously thinks about it how flawed the reasoning behind that is.

I am looking forward to moving imagery that is as perfectly solid and sharp as a very high quality photo, why should we settle for anything less? Right now anything that moves even at slow speeds simply is not hd, not even close, there happen to be a lot of static scenes with fixed camera's in movies to circumvent the problem allowing high quality images to be shown. This also seems to have masked the issue for most people.

 

[kenneth gruyaert]

It doesn't work that way...

David is right.

-sc

What reasoning or evidence do you think you have to come to that conclusion?

 

[kenneth gruyaert]

http://en.wikipedia.org/wiki/Direct_Stream_Digital

This is based on the same principle but for sound, for the exact same reason that this works even 1bit per channel video at very high framerates would also work. Although this is completely unrelated to the reason i want higher framerate video, it shows how noisy subframes can combine to create a high quality result.

 

[David Mullen ASC]

What reasoning or evidence do you think you have to come to that conclusion?

It's simple: higher frame rates require more light at the same ASA and f-stop because of the decreased shutter time. The speed of the playback isn't a factor, it doesn't matter if you project the 120 fps footage at 2 fps, 24 fps, 120 fps, 1000 fps, you still have to expose based on the shutter time, which is 1/120th of a second per frame at 120 fps with a 360 degree shutter angle, which is less exposure time than 1/48th at 24 fps with a 180 degree shutter angle.

Doesn't matter if you project at 120 fps, you still have to compensate exposure for the higher frame rate when shooting. You need more light due to the shorter shutter time -- doesn't matter if you do it by increasing the light levels, ASA rating, or opening up the lens iris, you need to compensate for the higher frame rates whether or not the footage is played back for slow-motion or normal motion. Speeding up or slowing down footage in playback doesn't change your exposure.

If you want to shoot a movie at 120 fps with a 360 degree shutter angle, you have to find more exposure or light somehow compared to shooting at 24 fps with a 180 degree shutter angle because each frame is only being exposed for 1/120th of a second, not 1/48th of a second, regardless of the playback speed.

 

[Jay A. Kelley]

What reasoning or evidence do you think you have to come to that conclusion?

Yeah Kenneth. This is pretty basic stuff.. Listen to David.
Jay

 

[kenneth gruyaert]

It's simple: higher frame rates require more light at the same ASA and f-stop because of the decreased shutter time.

You keep replying with this same argument while it should have been obvious from many of my previous posts that i agree the individual frames would receive less light so obviously they do not have to be captured at the same ASA setting.
What is happening here is that you decided i was wrong and didnt have a clue what i was talking about from the first post and refuse to take the time to even think any of this through because it is obviously a waste of your time.

 

[Álex Montoya]

Relax, Kenneth.

 

[Jay A. Kelley]

Agreed.. Play nice.

 

[Matthias Hutter]

I agree with Kenneth. Let me try to explain:

All else being equal (lightning, sensor technology etc):

With higher spatial resolution, you get more noise per pixel exposure...
but finer noise and integrated over a given area the amount of noise remains constant.
With higher temporal resolution, you get more noise per pixel exposure....
but faster noise and integrated over a given time the amount of noise remains constant.
('faster' meaning being visible for a shorter time, so I'm not counting slow-mo or frame grabs obviously)

What I'm saying is with higher resolution the total perception of noisiness remains similar.
Personally, I prefer finer and faster grain.

Besides, I hate the fact that we have 1080p+ but a huge portion of the sharpness gets lost thanks to motion blur.

 

[David Mullen ASC]

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.)

You are still underexposing if you don't compensate with more light and/or a faster lens, it's just that Kenneth is suggesting that the extra noise will become less visible due to temporal resolution.

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.

The exposure compensation needed for faster frame rates is real, that's all I'm saying.