Rolling shutter reduction

[Ryan Patrick O'Hara]

Hello everyone. Just wanted to run this by and see if you can confirm a theory of mine.

Rolling shutter is created when a frame, instead of being captured all at once, is scanned from frame top down to frame bottom.

Rolling shutter artifacts happen when the frame is moving faster then the rolling shutter creating top of frame to be noticeably different then bottom of the same frame.

I'm sure we are all aware.

My question to you is this: Would increasing shutter speed reduce the effects of rolling shutter?

The camera is still shooting 24fps, but each frame is a smaller moment in time. This is in essence forcing the frame to expose at a smaller fraction, making the rolling shutter scan from top of frame to bottom more quickly. This should create more snapshot like exposures minimizing motion blur and leaving a smaller amount of time between top of frame and bottom, therefore minimizing the possible discrepancy/skew.

What say you>? Does this seemingly flawless path of logic work?

-rpo

 

[GuyB]

This has been discussed before (maybe on another forum) but the essence was that increasing shutter speed didn't change the rolling shutter effect.

What I find strange is that I am sure the way a rolling shutter reads the frame out is exactly the same way as film is exposed by shutter. I am sure I have seen examples of how a shutter can skew a fast moving with still photography because of the way the shutter exposes the frame.

 

[Barry_Green]

This has been discussed before (maybe on another forum) but the essence was that increasing shutter speed didn't change the rolling shutter effect.

Correct, increasing the shutter speed only means that the individual line you're scanning gets exposed for a shorter time, but it does not speed up the "sweep" speed of the shutter down the frame.

What I find strange is that I am sure the way a rolling shutter reads the frame out is exactly the same way as film is exposed by shutter.

Sort of. In a still camera, maybe. In a movie camera, not really (unless you're talking about an extremely short shutter speed). A movie camera exposes the entire frame for a goodly portion of its exposure time; a rolling shutter never exposes the entire frame; it only exposes a comparatively small slit of the frame.

I am sure I have seen examples of how a shutter can skew a fast moving with still photography because of the way the shutter exposes the frame.

Yes, from old cameras; I don't know if there's many examples from a modern camera but if there are you'd have to use a very long shutter speed to introduce rolling-shutter skew/stretch in still photography.

 

[matthew77]

Barry,

a few corrections:

Correct, increasing the shutter speed only means that the individual line you're scanning gets exposed for a shorter time, but it does not speed up the "sweep" speed of the shutter down the frame.

This is correct. In the EX1, the leading and trailing edge of the shutter scans the image vertically in 1/60 second, regardless of camera settings. Skew is the same at all shutter settings.

Sort of. In a still camera, maybe. In a movie camera, not really (unless you're talking about an extremely short shutter speed). A movie camera exposes the entire frame for a goodly portion of its exposure time; a rolling shutter never exposes the entire frame; it only exposes a comparatively small slit of the frame.

Not true.

At many settings, the EX1 will expose the entire frame at the same time, if for an instant. At any shutter time over 1/60 second, there is definitely a moment when all of the pixels on the sensor are being exposed to light at the same time.

Yes, from old cameras; I don't know if there's many examples from a modern camera but if there are you'd have to use a very long shutter speed to introduce rolling-shutter skew/stretch in still photography.

In still photography, on old cameras with a slower curtain shutter, the shutter operation was very similar to the CMOS rolling shutter in that no matter what the shutter speed, the curtain and its "slit" would always move the same speed. Only the width of the exposure slit would change. So skew amount (or angle) would be the same at all shutter speeds for a given motion.

Ryan,

In general, lower shutter speeds will tend to reduce the appearance of the rolling shutter because the amount on time "rolling" across the image compared to the amount of time that the entire image is being exposed is lesser.

In motion picture cameras, the amount of time when the shutter is whipping open and closed is quite small compared to the amount of time that the entire aperture is open and collecting light. Hence, much less visible skew than on the CMOS camera.

 

[Ryan Patrick O'Hara]

Thanks everyone. I was discussing this in the d-90 thread in the 35mm adapter section. I got this response which made sense to me.

http://www.dvxuser.com/V6/showthread.php?t=145435&page=55

 

[Dan Vance]

At any shutter time over 1/60 second, there is definitely a moment when all of the pixels on the sensor are being exposed to light at the same time.

Just to be more clear, ALL pixels on ALL CCD and CMOS video cameras are of course always exposed to the same light at the same time, because there is no mechanical shutter. (Exceptions are those with a mechanical shutter.) It would be more accurate to say that "there is definitely a moment when all of the pixels retain a portion of the same light at the same time." A subtle but important distinction.

 

[matthew77]

Just to be more clear, ALL pixels on ALL CCD and CMOS video cameras are of course always exposed to the same light at the same time, because there is no mechanical shutter. (Exceptions are those with a mechanical shutter.) It would be more accurate to say that "there is definitely a moment when all of the pixels retain a portion of the same light at the same time." A subtle but important distinction.

Yes. You are correct.

A good clarification.

 

[Ryan Patrick O'Hara]

Just to be more clear, ALL pixels on ALL CCD and CMOS video cameras are of course always exposed to the same light at the same time, because there is no mechanical shutter. (Exceptions are those with a mechanical shutter.) It would be more accurate to say that "there is definitely a moment when all of the pixels retain a portion of the same light at the same time." A subtle but important distinction.

Thanks for trying to be more clear, but the definition of a rolling shutter is the opposite of all CMOS pixels being exposed at the exact same time.

Clearly when panning, if a vertical object is skewed, the top of the frame was 'captured' first... no?

Not trying to insight an argument or anything.


All of what I know regarding CMOS rolling shutter was from Barry's demonstration located somewhere in the DVXuser articles.

 

[matthew77]

Thanks for trying to be more clear, but the definition of a rolling shutter is the opposite of all CMOS pixels being exposed at the exact same time.

Clearly when panning, if a vertical object is skewed, the top of the frame was 'captured' first... no?

Not trying to insight an argument or anything.

Each pixel is exposed for a certain time interval - the "shutter speed" or "exposure time"

These pixels are exposed in succession, as the "shutter" rolls down the image.

There is lots of overlap in this process. At 1/60 second exposure, when the last pixel is beginning its exposure the first pixel is just ending its exposure. At that instant, usabale photons (ones that are not discarded by the reset process) are being collected on the entire sensor at the same time.

The temporal "center"of exposure is different for each pixel, but there are definitely moments when usable light is being collected by the whole sensor at once.

This insight will not incite any arguments from me.

 

[matthew77]

All of what I know regarding CMOS rolling shutter was from Barry's demonstration located somewhere in the DVXuser articles.

Perhaps it's time to expand your horizons?

 

[Ryan Patrick O'Hara]

This insight will not incite any arguments from me.

Haha, yes my dumb moment. I meant incite.

Thanks!

Perhaps it's time to expand your horizons?

That's what I'm trying to do!

 

[Dan Vance]

Each pixel is exposed for a certain time interval - the "shutter speed" or "exposure time"

These pixels are exposed in succession, as the "shutter" rolls down the image.

There is lots of overlap in this process. At 1/60 second exposure, when the last pixel is beginning its exposure the first pixel is just ending its exposure. At that instant, usabale photons (ones that are not discarded by the reset process) are being collected on the entire sensor at the same time.

The temporal "center"of exposure is different for each pixel, but there are definitely moments when usable light is being collected by the whole sensor at once.

Excellent clarification to mine! Thanks.