General Question About CMOS Rolling Shutter

[Terry VerHaar]

I have been busily educating myself on the process of the rolling shutter in a CMOS sensor and became curious about one aspect... in descriptions of the rolling shutter, the typical process suggests that one line of the sensor "fires" at a time, with each successive line firing in sequence as the exposure "rolls" down the length of the sensor from top to bottom.

My question is this: is it possible - and, in fact, are there actually sensors - where more than one line "fires" at once. For example, could lines 1 and 2, or even lines 1,2 and 3 all fire together thus rolling the exposure in pairs or groups rather than single lines? (BTW - I understand that successive lines don't have to wait for the previous one(s) to be read and that several lines remain "on" at the same time. My question relates to the actual "firing" instant.)

I am not specifically asking if RED sensors work this way (although it would be way cool to have that information) but am mostly just wondering if this exists, or is actually possible.

Would love to hear from the sensor gurus at RED or elsewhere on the forum. Thanks in advance. :cheers2:

 

[AnthonyFlores]

I have been busily educating myself on the process of the rolling shutter in a CMOS sensor and became curious about one aspect... in descriptions of the rolling shutter, the typical process suggests that one line of the sensor "fires" at a time, with each successive line firing in sequence as the exposure "rolls" down the length of the sensor from top to bottom.

My question is this: is it possible - and, in fact, are there actually sensors - where more than one line "fires" at once. For example, could lines 1 and 2, or even lines 1,2 and 3 all fire together thus rolling the exposure in pairs or groups rather than single lines? (BTW - I understand that successive lines don't have to wait for the previous one(s) to be read and that several lines remain "on" at the same time. My question relates to the actual "firing" instant.)

I am not specifically asking if RED sensors work this way (although it would be way cool to have that information) but am mostly just wondering if this exists, or is actually possible.

Would love to hear from the sensor gurus at RED or elsewhere on the forum. Thanks in advance. :cheers2:

+1 Thanks Terry for this ... researching sensors myself and would love to know more about this.

 

[Terry VerHaar]

David - thanks for weighing in here but I am not sure your explanation addresses the question. Or maybe I don't quite understand what your concluding.

One reason the flash in your example could illuminate several rows at once is because they are all "active" at the same time - but that does not mean they fired at the same time. It just means their "active" or "on" states overlapped.

I totally get the overall process - and also how mechanical shutters work with film to expose it, effectively, all at once. I may not have been clear in identifying my specific question earlier - which is: Can more than one line of a CMOS sensor "activate" (and "-de-activate or "read") at precisely the same time, in pairs or groups, with the exposure "rolling" down the shutter in these multi-line groups rather than one line at a time. Hope that is more clear. :001_smile:

 

[Randall Weatherington]

Terry, I do not know the answer to your question about currently functioning CMOS sensor chips, but to clarify what you are saying, although you are referring to just a few lines, it would be nice to have a CMOS sensor where all of the pixels would be triggered to "expose" simultaneously, and retain that exposure until the contents are read out into memory. That way there would be no "rolling shutter" effect.

I would say that someone, somewhere is working on this, but just has not perfected it, yet. If they have achieved it, it is still a very closely guarded secret.

 

[Lauri Kettunen]

it would be nice to have a CMOS sensor where all of the pixels would be triggered to "expose" simultaneously

I guess you mean a global shutter, such as

http://www.dalsa.com/mv/products/cameras/family.aspx?fam=Falcon

http://www.dalsa.com/mv/knowledge/globalshutter.aspx

 

[Terry VerHaar]

Terry, I do not know the answer to your question about currently functioning CMOS sensor chips, but to clarify what you are saying, although you are referring to just a few lines, it would be nice to have a CMOS sensor where all of the pixels would be triggered to "expose" simultaneously, and retain that exposure until the contents are read out into memory. That way there would be no "rolling shutter" effect.

I would say that someone, somewhere is working on this, but just has not perfected it, yet. If they have achieved it, it is still a very closely guarded secret.

I agree, that is the "holy grail." To "flash" the whole sensor on at once, leave it in this state for the entire duration of the intended exposure, and then read the accumulated photon data in one instant, would be great and eliminate all CMOS "issues." But, from what I have been able to determine, that is a very difficult thing to do with today's CMOS sensors and such a device is not yet available.

Seems to me that simultaneous multiple line read-resets would be a logical step in that direction. I am just wondering if there is a technological reason why this can't/won't happen.

 

[Terry VerHaar]

I guess you mean a global shutter, such as

http://www.dalsa.com/mv/products/cameras/family.aspx?fam=Falcon

http://www.dalsa.com/mv/knowledge/globalshutter.aspx

Lauri - Thanks for pointing that out. I do know about the concept of a global shutter but was not educated about the Dalsa.

Of course, it would appear these are lower resolution sensors than even the Scarlet. So, given the specs of the Scarlet (and Epic), my one simple, very specific question remains: is it possible, or even being done, that multiple lines are simultaneously being read-reset in a "rolling shutter" manner as a logical (?) first step toward a global shutter or perhaps as a "compromise" to get less skew (etc.) in a higher resolution sensor.

To be sure, I am not trying to break new ground on sensor design here, just hoping to satisfy my curiosity as to the current range of existing sensor behavior. (Actually, I am writing a short article on the matter and wanted to be as accurate as possible in terms of the state-of-the-art.) :cheers2:

 

[Brian Merlen]

now let me preface this with: i am no engineer(unless it some how can go through genetics)....but

http://en.wikipedia.org/wiki/Rolling_shutter

told me that CCD's use a system that takes the whole image at once, and that is why CCD's didn't have rolling shutter issues...

so to answer the question simply: yes such technology exists,

I do notice rolling shutter can be worse on cheapo cameras than on higher end ones due to a faster RR on higher ended models.

http://dvxuser.com/jason/CMOS-CCD/

has a real in depth detailed comparison that you may find helpful. one of the first questions i brought up in class is why would high end models use CMOS given the rolling shutter effect? i asked this about a year ago to the head of my video dept as a Jr and i asked specifically about the red one, of which he told me the issue was almost non existant due to their RR speed.

smears dont occur with CMOS, less energy consumption, and with a decent RR the rolling shutter issues will be non existant. i think that is why they took over and CCD's aren't used now.

and far be it from me to really know much more than basics about any of this... so please dont anyone jump on me if i got something wrong, i only graduated a couple months ago and it was a state school, not even a film school that i got my video knowledge from... so i probably have no idea what i am talking about... lol

 

[Randall Weatherington]

I guess you mean a global shutter, such as

http://www.dalsa.com/mv/products/cam...spx?fam=Falcon

http://www.dalsa.com/mv/knowledge/globalshutter.aspx

Lauri, as best as I can make out, Dalsa is using a hybrid sensor that alternates lines with CMOS and CCD type pixels. After the read-out, there are probably some algorithms applied to eliminate blur and smear, using the CCDs as a guide, and correcting the CMOS pixels to augment the CCDs.

Based on their suggested usage, the Dalsa sensors are not likely to be acceptable for DSMC usage, as it appears they have quite a few limitations.
Dalsa had a digital-cinema camera at one time. If the Falcon sensor was good enough for cinema, they might still have a digital-cinema camera.

 

[Terry VerHaar]

I guess I am really bad at asking questions!! :w00t:

Yes, I know CCDs read all at once. And yes, there are global shutters. I hate to be obstinate in my relentless pursuit of one small piece of information but, SPECIFICALLY, I was wondering if there are CMOS (!) sensors that USE ROLLING SHUTTER technology, where MORE THAN ONE LINE WILL READ-RESET IN UNISON. :mad2:

For context here, RED has chosen NOT to use CCDs for many good reasons since they compromise the process in other (apparently unacceptable) ways. And, it would appear that a global shutter is difficult or impossible to implement at the resolutions and frame rates targeted by RED. SO, I was wondering if, as a sort interim technology to a global shutter, this multiple line CMOS read-reset capability was being employed. Am I being articulate yet? :bigear:

EDIT: I should say, I do appreciate the attempts to step up and add information here. I don't mean to sound ungrateful! :seeya:

 

[Brian Merlen]

hmm

hmm

your multiple line theory keeps reminding me of interlaced vs progressive scanning... i remember way back when (crt days) reading that people who dealt with high speed work prefer interlaced scanning specifically because of the rolling shutter/blur issue... and that with double the scanning vs progressive it worked better for high speed movement...

from wikipedia: "interlace can provide a video signal with twice the display refresh rate for a given line count"

but i am sure you already know this....

 

[Randall Weatherington]

Yes, I know CCDs read all at once. And yes, there are global shutters. I hate to be obstinate in my relentless pursuit of one small piece of information but, SPECIFICALLY, I was wondering if there are CMOS (!) sensors that USE ROLLING SHUTTER technology, where MORE THAN ONE LINE WILL READ-RESET IN UNISON.

Staff???
Umm.. Uh... Well...

Poor Terry.
Ask a simple question...:mad2:

In all seriousness, Terry, I know of no sensor at the present capable of doing what you, and probably all of us would like.

 

[Peter Moretti]

Randall, I don't see why anyone would want this. It would create a stair stepped rolling shutter distortion which would be much much much more difficult to correct.

Terry, I think the answer to your ? is no. Simply b/c we've never seen the rolling shutter effect as having anything but a slanted line w/ no stair stepping.


Here is a crude graphic:

Rolling shutter as we now see it

\..............\***
*\..............\**
**\..............\*
***\..............\


Rolling shutter as it would appear if groups of rows fired at the same time

|..............|***
*|..............|**
**|..............|*
***|..............|

(Reminds me off typing class when we made a bunny, LOL!)

 

[Andrew M.]

Terry, here is a good explanation

http://www.caspegroup.com/How an electronic shutter works in a CMOS camera.pdf

And if you would like to go into details.

http://www.sony.net/Products/SC-HP/cx_news/vol47/pdf/featuring47.pdf

 

[Peter Moretti]

Andrew, thanks for the links, but I think Terry understands all that. He's asking a very specific ? about if all sensors w/ rolling shutter roll smoothly or do some roll in groups of lines.

 

[Terry VerHaar]

Andrew, thanks for the links, but I think Terry understands all that. He's asking a very specific ? about if all sensors w/ rolling shutter roll smoothly or do some roll in groups of lines.

Peter - Yes. Thank you very much for your clear understanding and your thoughts. Of course, I feel stupid once you pointed out the logic that we have not observed this type of stair-stepping which would, obviously, be evidence of the "batch" approach. I think you have solved my mystery and I can sleep better tonight!

I do thank everyone for your comments and the effort you went through to pull up links and such. I try not to ask questions on here until I have done my homework so, obviously, I just couldn't pin this one down with any prior research.

:grouphug:

 

[Peter Moretti]

Glad to have helped you sleep. :beer:

 

[Randall Weatherington]

Peter, I was not very specific, so let me clarify my last statement: It is s CMOS sensor that fires all of the pixels simultaneously that I'm sure we would all like to see. Not sectional stair stepping.

Your graphic also reminds me of my career shooting end title rollups, and the issues I faced shooting stop motion vs motion blur of titles on film. Fodder for another thread.

Terry, have a good night's sleep. Based on what we are seeing from RED, this is pretty much a non-issue.

 

[Rick Burnett]

The answer I believe is no, you cannot read more than one line at once because you don't have multiple sets of A/D converters from what I have seen. Granted, maybe some designs have more than one set of AD converters and so they can read out faster that way (offsetting the second line from starting till half the time of the line before it to keep from getting stair stepping distortion in the image).

But, from all the reading I have done, when a line is read, every pixel in that line is sent to an A/D converter in parallel that determines the level of charge for that pixel.

With a global shutter, it still reads all the lines out in rows, but a mechanism is used to prevent capturing additional photon information.

 

[Gerald Menclik]

well Terry, I do understand you question, but can only try to assume *ggg* reading more than one line per cycle would mean a hell of an algorythm to "finish" the picture after your rr is through the complete sensor ..... maybe the asics at this time aren`t fast enough to proceed the electric information correct when you read let's say 5 lines per cycle ..... just my 2c.
I don't know a sensor doing this.
... and you don't sound ungrateful ... *ggg*