Article: CMOS, CCD, Rolling Shutters, and more

Ok... I'm curious about two things.

#1. How come there are no Global shutter CMOS cameras? Meaning, is there some benefit to using a rolling shutter with CMOS that I didn't catch in the article or otherwise?

#2. What causes CMOS to be immune to vertical smear? If the CMOS sensor utilized a global shutter, would this still be the case?

Seems a global shutter, CMOS sensor would be the best of both worlds, so I'm wondering what complication there is, or what is cheaper/easier, or whatever about using a rolling shutter instead of a global one in CMOS sensors, that has prevented this from happening.

Thanks,
Jason
 
#1 is better answered by the manufacturers themselves as to why they keep choosing rolling vs. global.

As for #2, the type of shutter is irrelevant to the smear. In a CCD an entire line of pixels is read out at once, and if one of the pixel wells blooms over it affects the entire line. In a CMOS chip each well is read individually, so what happens in one well can't affect any other.
 
Hmmm..... if you shoot with a very narrow shutter angle on a Film camera, isnt that like a rolling shutter?

PS, i shot my first 35mm Today!!!! Arri BL4, 1000ft Mags... Zeiss primes the works.. I love it...
 
TimurCivan said:
Hmmm..... if you shoot with a very narrow shutter angle on a Film camera, isnt that like a rolling shutter?
Click to expand...
You'd have to use an extraordinarily narrow angle, like maybe 10 degrees, to simulate the rolling shutter effect on a film camera.

In most circumstances you're using a 180-degree shutter, and so a film camera will exhibit properties more like a global shutter, with a very little hint of rolling shutter-ness in it. But it's quite different.

With a film camera, the entire frame is exposed at once for a good amount of time. So any motion of the frame will result in blur. With a rolling shutter, only a certain small portion of the frame is exposed, so the amount of skew and the amount of blur is very different than a film camera.

Under normal shutter circumstances I'd say it's around 85% like a global shutter, maybe 15% like a rolling shutter. The narrower you make your shutter angle, the more like a rolling shutter it'll become.
 
Barry, like your previous article on CA, this one will be referenced by us often. Thanks for making the effort to compile it.
 
... and then you throw interlacing into the mix, and you can get som *really* interesting unwanted effects! :eek:

- Mikko
 
Basically the two problems from Jason are the same. CMOS sensor works in a row fashion. All commands, like reseting the sensor, enabling the integration (ie, open the shutter) and sampling, are performed row by row. This makes it easy to control. Furthermore, it matches the way the memory works. Since you can only read/write one row of the memory at one time, sampling the whole image at the same is useless because you have no place to save it.

There exist some ways to 'hold' the image inside the CMOS sensor array. In this way the global shutter can be implemented. However it inevitably reduces the light-sensitive area of the sensor, so most companies do not like it.

BTW I'm Chia-Kai Liang who mentioned in the article...this project travels far beyond my imagination. We are still working on that problem and having some new results. I will try to present them when it is mature:)
 
Hey, thanks for stopping by and commenting! Your work was very interesting, and we look forward to hearing what you've come up with next.
 
so if i understand this correctly - a rolling shutter CMOS and a global shutter CCD both with the same settings may render motion differently - correct?
 
Most definitely correct. A CCD is going to blur the motion of the overall frame. A rolling-shutter CMOS is going to expose as it goes. So if something began moving late in the frame, the CCD will show the blur, the CMOS will show much of the frame clear, and the rest blurred.

If there's constant motion, then the CCD will show constant blur whereas the CMOS will actually change the shape of the motion because different portions will be exposed in different areas of the frame, sometimes being duplicated (if the motion is racing down the scan) or perhaps even being skipped (if the motion is moving against the scan).

The faster the readout, the less of an issue this will be. Red claims to have fixed the rolling shutter issues they had; don't know how Sony's EX will perform. The V1/FX7 has it, and all the less-expensive cameras have it.

How does film perform? A rotating mirror shutter has some characteristics in common with each, but I'd say that it's probably 80% more like a global CCD shutter. Depends on your shutter angle, of course, the tighter the shutter angle the more film exposure would resemble a rolling shutter. But for normal exposure there's huge swaths of time where the whole frame is exposed simultaneously, thus more closely mimicking the effect of a global shutter.
 
Hopefully not already asked or maybe not relevant, but how do the CMOS and CCD type of image gathering relate to the way the human eye works? Are we more like one than the other?

Thanks,

Michael Erlewine
Heart Center Studio
 
I´d like to ask if the problem with writing whole image to memory couldn´t be solved by implementing some kind of CACHE memory. (I´m just wondering if there are other problems than amount of space consumed by this solution and it´s price.)
 
You must log in or register to reply here.
Back
Top