Multiple exposure question

[Gavin Greenwalt]

I've got a question which arose in my mind last night regarding multiple exposures.

I'm not sure how CMOS sensors work so this might all be my ignorance giving me bad ideas.

If I were to expose one frame at 1 fps with a 360* shutter would I get the same image as if I exposed 120 frames at 360* shutter all added together in post?

If so couldn't you use that single frame sequence to reconstruct a 7 stop bracket without adjusting shutter from frame to frame and even re-using exposures in order to speed up the process considerably?

Aka one 1/120th, One 1/60th, one 1/30th, one 1/15th, 1/7th, 1/3 and 1/1 second?

Seems like this would be perfect for the "burst" option on scarlet/epic. On epic it would also let you choose the amount of frames you want mixed back in together for selective shutter in post for DSMC. And of course this could be applied to REDOne right now.

I can only assume I'm not thinking this through correct otherwise RED would already be doing it.

 

[David Mullen ASC]

Ignoring the idea of different exposures for a moment...

You're asking if one second for footage at 1FPS with a 360 degree shutter (so one frame with a one second exposure time) can look the same as blending 120 frames shot at 120 FPS with a 360 degree shutter (so 120 frames with a 1/120th shutter time per frame blended into one frame). Good question -- I'd make a guess and say that you could get a similar result but not exactly the same, because the moving object would be solid but in a different spot in each frame of the 120 FPS version, with a minimal amount of blur trailing behind it, whereas the 1 FPS shot would be in one spot with a lot of blur behind it.

Getting back to the idea of varying the exposure per frame...

But if you start to vary the exposure per frame on the 120 FPS version, the moving object would be in different positions on each frame and with different exposures, so while the non-moving background (if a static shot) could combine those different exposures to get an HDR shot, the moving object would probably be flickering due to variations in brightness, or its blur would be flickering, even after you blended the frames.

 

[Gavin Greenwalt]

But the blur wouldn't be "Behind it", it would be spread out evenly through the entire time frame. So theoretically they should look identical unless there is a short pause between frames in which case there would be a minuscule amount of onion skinning -- but I don't know of anyone who uses 120 pass multi-sample motion blur so I imagine it would be invisible. Even 32 samples is just about overkill in 99% of all case. I'm most concerned that maybe the noise characteristic of the sensor hits a hard floor of sensitivity and the 120 noisy images wouldn't add up to one well exposed, clean image.

I've done some simulations before using optical flow of what it would look like and it wasn't pretty with even moderately moving subjects but I'm thinking more along the lines of motion control or just scenic panoramas free of lots of motion. But this approach would seem to be an improvement on most HDR Modes (which I see RED just announced) in which it takes consecutive stills: 1/1, 1/2, 1/4, 1/16, 1/32 which takes takes a full 1 2/3 seconds (if my calculus is sound off the top of my head ). This would shave off of a full 2/3 seconds off the process and result in a better HDR. Of course RED is one of the only cameras which can do this since no still cameras can shoot full resolution at > 100 fps.