View Full Version : Film Rolling Shutter
02-09-2008, 07:47 PM
Ok, I was just watching a documentary series about masters of photography, and there was Jacques-Henri Lartigue showing his albums when he turned the page and I saw this.
Very interesting image: there's skew in both directions because there was movement in both directions and it was shot definitely on film.:)
So it'd be interesting if anyone with more experience explained what type of camera/shutter/technique caused this artifact.
It also serves as an example for those who think that rolling shutter skew can be easily removed in post, and for those who think it will only happen in Bourne/Cloverfield/shakycam style shots... and prove them both wrong. :sad:
I could also be argued that this could be use as a tool to build an interesting stylized image... those bent-forward tyres look cartoony. :clown2:
Excuse my English, if it doesn't seem clear.
02-09-2008, 08:44 PM
The photographer in this case used a camera with a "slit" shutter, right on the focal plane at the film (or plate I suppose) It's basically a mechanical version of our troublesome rolling shutter with Red. It "scans" the frame from top to bottom and so has trouble with pans or other horizontal movement.
Here is an interesting article in a large-format photo web forum:
I bet that this is why medium and large format cameras went to a bladed shutter design in the lens, and SLRs the faster focal plane curtain type shutters that use an "opening" blade and a "closing" blade.
Just something we'll have to learn about and be careful with...
02-09-2008, 09:12 PM
It was shot with a camera with a focal plane shutter. I have some Speed Graphic 4x5 press/view cameras with focal plane shutters, and it works in very much the same way as a CMOS rolling shutter. On the Speed Graphic, the shutter is a roll of black cloth with a slit in it that gets rolled up onto a spring loaded roller when the shutter is cocked. When the shutter is released, the cloth unrolls, pulling the slit over the film plane from top to bottom, exposing the negative(or positive). Since the spring mechanism pulls the cloth over the film plane at a fixed rate, there are actually 3 different sized slits, from wider to very narrow, that correspond to different shutter speeds. The smaller the slit, the faster the effective shutter speed since only a tiny part of the film is being exposed at a given moment. But the extreme skew on these cameras is caused by the speed, or I should say lack of speed, of the shutter traveling across the frame, coupled with the large size of the negative.
I also have a Canon AE-1 35mm still camera which has a focal plane shutter, but skew is not a problem with these cameras since a) the shutter mechanism moves much faster than on older cameras like the Speed Graphic, b)the format is much smaller so it takes much less time to "swipe" the frame, and c) the shutter swipes from side to side instead of top to bottom, so if there was a horizontal motion artifact, it would be compression or expansion, not skew. Also, on longer exposures, the shutter actually swipes fully open, then swipes closed, so skew(or compression/expansion in this case) would only be possible at the top and bottom ends of the exposure, similar to a the way a mechanical motion picture camera shutter works.
I'm sure that Jim and the Red team and others are sick of hearing me say it, but thats why I've suggested trying the CMOS sensor read/reset swiping from side to side instead of top to bottom. Since most quick motion in motion pictures is horizontal, I think that some slight expansion or contraction would be a less noticeable or unpleasant artifact than skew. I'm sure that they'd thought of that and there's a reason that it wasn't already that way, though, so... I'll stop pretending like I know what the hell I'm talking about now.
02-09-2008, 10:03 PM
wow thanks for your insights... this is a great forum. I didn't know about those medium format shutters, etc.
FWIW, I've read you suggestion of changing the orientation of the shutter some times before and here's what I think:
1) I wouldn't help as much as you think... wobble is wobble anyways.
2) This is speculation on my side: It can't be changed by the software. The chip is designed to work the way it is. The columns and lines are connected to work on a certain order and it's hardcoded that way and can't be changed by the firmware.
3)Anyone's still free to work with the camera sideways and crop the center 2K's. :) It would still be better to change to 2K cropped, as it's already been stated that the readout speed depends on the number of vertical lines being used.
Anyways let's wait for build 15 and cross our fingers. There's hope.
02-10-2008, 08:47 AM
Yeah, I'm planning on going to the subway when my camera comes in a few weeks and shooting fast moving trains with the camera both on it's side and right side up to see the difference. I'll post results when I get them. Reorientation still does nothing for the strobe issue, but baby steps : )
02-14-2008, 01:06 AM
So, it's not a clown's car driving down a steep hill with the camera on a dutch tilt:)
02-14-2008, 03:43 AM
Well I am sure RED is working a solution to the CMOS problem. I cannot image them letting that one issue screw up their baby.
Heck, I even rejected the idea of buying a cheapo Sony handcam that had a rolling shutter. I shot an event at a nightclub just to test it out and all the footage came out ruined because so many people were taking pictures.
Can you imagine spending cash on a pro camera not being able to shoot footage of people taking photos?
I bet within a few months the RED team will come up a fix.