Rolling Shutter Suggestion

[Phil Bates]

I briefly mentioned this in another thread, but it didn't get much traction so I thought I would post it here.

Here's the problem, and it's not skew (at least for me): Flashing subjects often produce banding artifacts because of the hard cut-off from one frame to the next. This can affect shots with lightning, fireworks, explosions, camera flash, muzzle flash or any other random (or not random) flashing subject. Shots of these subjects are at risk of banding, and I personally have lost many good takes because of this.

Here is an example of the hard edge I'm talking about:

Film cameras don't have this problem because of the motion blur or "out-of-focus" shutter as it swings by. My suggestion is that Jim and his brilliant crew is to match this effect from film cameras build in the option for the user to feather this hard edge. I guess this would mean slowing down the switch of sensor pixels from one frame to the next. And I don't know if it's possible, but please consider adding this feature to Epic or to a future build.

Here is an example of a feathered edge 180 pixels wide:

Here is an example of a feathered edge 300 pixels wide:

The above image looks more natural to me and I believe I could live with that. A wider feather would still be better.

I realize this is causing a brief double exposure between frames, so I am only asking that this be an option in the menus to switch on if we need it. Any remote chance that this could be a possible feature of Epic?

Thanks for listening,
Phil

 

[Stephen Aldridge]

Wow! I really like that! Good Idea

 

[Gavin Greenwalt]

Doing this the only way I can think how would mean losing a stop of sensitivity.

Might be worth the trade off if it were possible. You would also need analog gain on each line of photosites that could be adjusted during the exposure.

 

[Phil Bates]

I thought I would bump this back to the top. Rather than let this idea die, I would love to give the Red team some chance to respond.

Thank you,
Phil

 

[Marc Wielage]

My only suggestion is to fix this in post. I don't think it's something the camera could anticipate and compensate for on the fly. Lightning as you show in your still is fixable during color correction; the hard stuff would be fast-moving multiple flash cameras in a crowd on a city street, stuff like that. That's a tough test for any CMOS camera.

 

[Mark Andersen]

I'm really surprised that you got the flash edge effect from a lightning strike. I would think lighting isn't nearly fast enough to cause this effect.

 

[Chi Ying Chan]

"Film cameras don't have this problem"


Film cameras DO.

Regards
Chan Chi Ying
DP HK

 

[Peter Majtan]

I'm really surprised that you got the flash edge effect from a lightning strike. I would think lighting isn't nearly fast enough to cause this effect.

Mark, I don't believe this is an actual shot from RED. I think he was just using Photoshop to illustrate the issue and possible solution. I wouldn't read too much into it...

:thumbsup: Peter

 

[Phil Bates]

Mark, I don't believe this is an actual shot from RED. I think he was just using Photoshop to illustrate the issue and possible solution. I wouldn't read too much into it...

:thumbsup: Peter

Peter,

I used techniques to illustrate a real problem. There was no way to duplicate a feathered rolling shutter(the solution) without generating it, thus the need for an illustration. However, the effect is real, it happens all the time when shooting lightning. Here are actual examples (frame grabs) of this problem shot on RED three weeks ago:

Phil

 

[Phil Bates]

My only suggestion is to fix this in post. I don't think it's something the camera could anticipate and compensate for on the fly. Lightning as you show in your still is fixable during color correction; the hard stuff would be fast-moving multiple flash cameras in a crowd on a city street, stuff like that. That's a tough test for any CMOS camera.

Mark it is not easily fixed in post (please see the actual examples above). The light is not even, so it would involve generating complex gradations. In some cases you would have to re-draw the lightning strike. Also, trying to use info from adjacent frames doesn't work either, because nothing lines up.

To respond to your other point, I am not asking RED to generate a system that can anticipate and compensate on the fly. I am asking for a user-adjustable setting that forces a feather on the shutter, so that we can anticipate scenarios where there will be flashing elements.

Phil

 

[Phil Bates]

"Film cameras don't have this problem"


Film cameras DO.

Regards
Chan Chi Ying
DP HK

Chan, film cameras have a rolling shutter, but I've never seen one that has the specific problem of a hard edge cutoff between frames. It is always feathered due to optical reasons like focus or motion blur of the blade. I could be wrong, but I've never seen it when working with Arri, Bolex, and Mitchell cams.

Regardless, even if film cameras had the same problem, why not fix it in the next generation of digital cameras by giving the users the option of feathering the rolling shutter?

Phil

 

[Stefan Christou]

Hi Phill,

What you are suggesting sounds a little like the slow shutter mode as featured in some Sony ENG cameras.

I suppose in Epic you could just shoot at as high a framerate as possible then combine the frames. And since RED are all about preserving information, I think this would be the route they would suggest, rather than using a quick and dirty (in camera) frame combination feather as you have suggested.

S

 

[Stephen Pruitt]

I don't know much about trade-offs involved with the fixes, but it sounds like a great plan.

I would guess that the only way to kill this problem off entirely would be to put a mechanical iris shutter on the front of the camera. I don't even know if an iris shutter can fire that rapidly or not, and, even if it could, it sure would be expensive and quite a hassle to implement.

I'll be doing some tests Friday evening with police lights. I'm expecting a nightmare.

Stephen

 

[Phil Bates]

I don't know much about trade-offs involved with the fixes, but it sounds like a great plan.
...
I'll be doing some tests Friday evening with police lights. I'm expecting a nightmare.

Stephen

Stephen, I would expect you will see this same banding effect if there is any glow or reflection wash around the lights. The lights themselves are probably too small to catch the shutter mid-stream. If you do see the effect, would you mind posting an image here on this thread?

Thanks,
Phil

 

[Stephen Pruitt]

I certainly shall. . . and though I have never experienced it myself before, I've also never shot emergency lights before. Other people certainly have had the problems.

Stephen

 

[Jason Murphy]

This has been something I've been wondering about for years. One of the major problems with CMOS rolling shutters has been less about the speed of the read-refresh (though obviously, faster is better), it's been the hard edge cutoff. I remember some years back trying to capture lightning strikes on an old Sony A1U, and getting incredibly frustrated with this very thing. Admittedly, RED is far superior in pretty much every respect, but the hard edge of the rolling shutter produces the same artifacts in these situations.

If a feathered edge shutter were even technically possible (and I'm not sure it would be), it would be a fantastic to have this as a user-selectable mode for those times when uncontrolled strobes or flashes are being shot. 99% of the time, it's probably not necessary, but it would save some major headaches that other 1% of the time.

As Phil mentioned above, fixing this stuff in post is far from trivial.

 

[Ryan Fitzgerald]

great post

great post

I think this is one of the best posts I've seen on this sight. Rolling shutter is the thing that bothers me most about cmos sensors.
-Ryan

 

[Stefan Christou]

Well epic will have a much faster read/refresh rate so it will be a thing of the past soon. Scarlet 2/3" should also be absolutely ace in this department. Probably considerably better than Epic even.

S

 

[IAN SUN]

Well epic will have a much faster read/refresh rate so it will be a thing of the past soon. Scarlet 2/3" should also be absolutely ace in this department. Probably considerably better than Epic even.

S

Not fast enough to solve this problem, the faster readout will further minimize (the R1's already minimal) diagonal distortion of verticals moving fast across the frame or in the case of whip pans.

Gunshots, lightening strikes, strobes and the like will still result in unevenly illuminated frames.

 

[Daniel García]

I don't think this is possible. I don't know the specifics of CMOS design but I don't think that can be done the way current sensors work. The sensor reads one line at a time, top to bottom. It's binary, not progressive: a line is either already read or not read. The flash happens between one line readout and the next. And there's only one readout per line per frame. It doesn't go back to blend another exposure.

The film shutter is soft because the rotating shutter is not exactly on the film plane, thus casting a soft edged shadow on the exposed film.

But I believe the OVF got cancelled so that route is closed for RED.