Filming cine film from screens - optimizing DR - hmmm, I guess a little bit 'zoney'

[nick mulder]

Hi,

I want to shoot video footage off an LCD (or an HD CRT monitor if I can get one) with a 16mm camera. I'm trying to learn how to optimize the output so that it would appear as if the footage were shot on film in the first instance so I'm keen to learn how to optimize the DR/gamma/luminance distribution from screen to film.

Imagine I had a video of a contrasty outdoor sunny scene in which its max and min luminance values hit the actual %100 and %0 values of the screen (I'll make them do that) - I also have dialed in an %18 grey and have taken the following quick readings with a spot meter from the screen:

EV % 0 = 4
EV %18 = 7
EV %100 = 10

So that is a DR of 6 and I have my baseline exposure set with the %18 grey.

I think I'm going to be needing to push the film to get the required range of exposure values distributed on the film as it would have if I were to shoot the footage in camera in the first place - right ?

Is it as simple as that ?

What don't I know about ? Am I about to get the range correct but send all the nice visual data to peak or shadows ?

I do have some old old film here that was used in the 80's to shoot strictly off CRT monitors, I have noted it is very contrasty compared to standard films, perhaps I should start with that stock as a baseline (?)

It's not lost on me that I have control over both the file going to the screen (I can play with it 'in post'), the screen itself (brightness, contrast settings) aaaand photo chemically with the film itself - I'd like to nail this with the end goal being that I can have all my composite effects and CGI footage have a 'shot with cine and developed in a bucket in the backyard' aesthetic (well, there's more to it than that, but I don't want to get off topic)

Also - as an aside why isn't it %50 grey ??? How did %18 grey become the mid point here ? The readings show it follows er... 'properly' too :confused:

 

[Worker 11811]

Don't forget that you might have flicker problems if you are recording video footage with a 16mm movie camera.

NTSC video runs at 30 frames per second. (29.97, actually.)
So do HD 720 and HD 1080 but they can also run at 60 f.p.s. You also have to contend with interlacing.

Yes, you can do it but you need to be aware of this and account for it if you have trouble.

 

[nick mulder]

Thanks, I'm on top of all that carry on - I'll do it frame by frame ...

Easy

The question I need answered would apply for a single frame in any format - I'm really just keen to nail the DR/luminance part of it at the moment

 

[Worker 11811]

It's good to hear you are on top of that.

Where I work, I get so many people who come to me and ask about shooting video from movie screens asking why the image flickers that it's almost a knee-jerk reaction for me to try and explain.

As far as exposure and dynamic range, I think you are covering all the bases I would, plus a few more I didn't think of. My only concern is whether a video display CRT/LCD, etc. limits the DR of the image you will be able to capture. Many video screens today can not display all the levels of brightness/contrast that HDR images can contain. But I'm going to bet that you have already thought of that too.

The only other thing that I can think of would be to use a film recorder. However, that could be cost prohibitive. Film-out services aren't cheap.

 

[nick mulder]

As far as exposure and dynamic range, I think you are covering all the bases I would, plus a few more I didn't think of. My only concern is whether a video display CRT/LCD, etc. limits the DR of the image you will be able to capture. Many video screens today can not display all the levels of brightness/contrast that HDR images can contain. But I'm going to bet that you have already thought of that too.

Well, yip - that's the whole point of my question...

I'm only getting 6 stops from my screen - but then scene as shot had say 10 stops ...

 

[Worker 11811]

Dumb idea, here. Just spitballing. Probably won't work but, what the heck...

If your camera is exposing at, say, 1/50 and your display is at 1/24 that would mean one frame of film would be exposed roughly twice as long as the video. (Almost a stop.)

Would it be possible to make the video display two frames for each frame of film exposed?

If that is possible, would it be possible to make each of those video frames display a different contrast level which would be captured on the film as one composite frame?

If so, could you produce two versions of the video, each having different contrast settings, which are then interleaved, frame by frame (or field by field ?) and cause them to both to be displayed to the camera for each frame of film recorded?

The result would be a sort of "bass-ackward" form of HDR imaging similar to what we do with digital cameras when we take two pictures at different exposure settings and composite them together to make one image.

I know this seems a lot like flying an airplane just to cross the street but it's also so crazy it just might work.

I still think a film recorder like a Lasergraphics machine would produce the best results but the cost could be prohibitive even if there is still such a thing as a film recorder for 16mm. Or, is this more of a do-it-yourself project?

 

[nick mulder]

DIY for sure...

hmmm - I've thought about HDR yes, but then got distracted - but you've got me thinking again...

Ok out loud and off the cuff:

So for discussions sake lets just keep it simple and do two exposures, I'd find the average point of the footage (I can do this with a curves analyzer in Nuke) - I then split it luminance wise into two streams at this point then optimize both of these new videos so they are using the screens max white and black ...

So how much to offset the exposure for each stream ? - shadows would be underexposed from 'proper' exposure (original %18) - highlights overexposed... I think I'd need to go back to the original footage and do further analysis on it, like find the average luminance on the two streams or ... hmmm (?)

And exactly how to 'split it luminance wise' is a thinker also - imagine the shadow stream, so where do those...

uh oh

brain fart

ha ha - now I'm thinking its silly

gah!



more thinking required

I like where this thread is going tho - keep talking!

 

[nick mulder]

I'll do it frame by frame - so there are no issues re. exposure times and what not - nor do we need to think about fields etc...

If it had two streams per frame I'd simply go:

1s+1h, 2s+2h etc... per cine frame

# = frame number (on screen)
s = shadow stream
h = highlight stream

this can all be automated

 

[Worker 11811]

Let's just kind of puzzle this out. Shall we?

A camera and a projector are essentially two inverse devices. A camera captures reflected light from objects in its field of view and a projector throws light onto objects in its field of view. However both of these devices use light reflectively whereas our hypothetical setup uses light transmissively. Our "projector" (the video display) sends light directly to the camera.

This is where I think the problem lies. You see, if I was going to calibrate my movie projector I would run a loop of SMPTE 35-PA (AKA: RP-40) test film then use a light meter to test the picture on the screen and adjust the projector accordingly. (I am a cinema projectionist and technician.) The checkerboard pattern of the SMPTE test film simulates a gray card. The black/white areas of the test pattern approximate an average brightness of 18% IRRC.

We can't exactly use a gray card in our present situation because the gray card works reflectively and the projector/screen works transmissively.

So... What if, instead of using a video screen, you used a hi-def video projector?

We could find a suitable test pattern to project on the screen and take a few test shots. You could then calibrate your video to match the test shots and once you have the settings done, you should be good to go.

If you can get your hands on a good hi-def projector, especially a 3-chip DLP model, I bet you'll be able to get a lot better dynamic range than an LCD, a CRT or a plasma screen can produce.

Then, if we needed even more DR than that, we could do the double exposure thing.

 

[nick mulder]

I've used those SMPTE loops on the Phillips 35/70mm projectors at my old work (done a little projection myself, a very little)...

Hmmm, I can source a projector from the same folk - would be interesting to see the comparative DR.

Without a rear projection screen I'd need tilt/shift lenses to correct for the parallax - need a nice big and dark room, i.e. outdoors - no reason the 'HDR' thing if I can nut it out couldn't still apply too

 

[Worker 11811]

Those Phillips are nice, old machines! They sure don't make them like that anymore!

If you line up the projector vertically above/below the camera so that the film planes are exactly the same distance away from the centerline the keystone effects should cancel out. The same should be true if you align the projector and camera horizontally.

My cinema projector is set up with almost 10 deg. of down-angle. Yes, there is keystoning but I have the aperture plate filed with the vertical edges at an angle to compensate. The PICTURE is still keystoned but the edges of the image are straight. In nearly 15 years of operating this theater NOBODY has commented, let alone notice.

I am betting you will be able to zoom the camera in and crop the edges of the projection screen, slightly, to visually square up the picture and nobody will be able to tell unless they hold a ruler up to the image.

BTW: I also have a Christie HD-8K video projector where I work.
http://www.christiedigital.com/AMEN/Products/christieHD8K.htm

It produces a crystal clear image in 1080p and its contrast ratio is 2,000:1. There are some projectors that are even better than that.

 

[nick mulder]

Log base2 of 1600 = 10.6 stops of DR from that projector

hmmm ... getting there huh

What exactly is 8K about it aside from it's model number ?

 

[Worker 11811]

What exactly is 8K about it aside from it's model number ?

8,000 lumens. That will shoot about 50-something candelas per square meter or just shy of 15 ft.-lamberts on a matte white screen, if my math is right. We use a 1.5 gain screen so we get a little more apparent brightness than that.

This produces an image on a 12 ft. X 22 ft. screen that you can see in full room light with room to spare. In fact, the only thing that degrades the picture so much you can't see it is a spotlight shining directly on the screen.