Rolling Shutter (Again) at DVinfo.net

Paul Curtis · January 14th, 2008, 10:05 AM

I hate to bring this up but i don't have one in front of me to check.

http://www.freshdv.com/2008/01/xdcam...-examples.html

the link shows some footage, overcranked at night time and a pan along with someone running. The rolling shutter is very visible, in fact to my mind unusably so.

Can some tell me whether this clip is particularly unusual, is the effect so pronounced because of the overcrank? I always believed the amount of skew was constant because it's based on the native read times of the sensor regardless of shutter speed.

I don't believe this type of camera move is unusual, views out of cars, panning with action the list is probably endless.

In the various samples i've been watching i've not seen anything quite so objectional as this. Does anyone have any real life footage out of car window, or shots like this?

The F335/350 et all are all CCD based aren't they?

thanks in advance
paul

Steven Thomas · January 14th, 2008, 10:59 AM

Paul, there will be skew.
I shot some soccer footage at 720 60P that was posted here for a while. During real fast pans, and slowed down, you could see about a 10 degree skew on poles in the background. If I had shot it in 24P, you probably would not be able to tell.

That night running footage appears to show the same amount of skew I've witnessed under those conditions.
To make that shot, you would of had to change the angle of where the camera sits; therefore minimizing the quick pan on top of the quick motion.
This is a good example in showing this problem.

Rolling shutter skew artifacts are there, but can be controlled to some extent. As you know, it's all about the speed of motion across the sensor.

As it has been mentioned, the EX1 looks a lot like the rolling shutter artifacts seen on RED ONE.

If you don't like the thought of this, stay with what you're using, or wait it out for something new to come along.

All cameras have their good and bad. It's really about the ratio of the amount of time these problems will be viewable. Having low noise and high rez is needed 100% of the time. If I get an occasional rolling shutter artifact, I'll live with it.
But, of course, I will do my best to avoid amplifying the issue.

If you're shooting 24P, real fast motion crossing the sensor will have more image blur, which will hide "some" of the artifacts.

With all it's quirks, I really think the EX1 is quite the camera! I'm really looking forward for a camera with this image quality to arrive without the rolling shutter issues.

Now, only if there was a way to increase the CMOS sampling time via software upgrade.... OH well, wishfull thinking.
I don't know much about this particular technology, but I'm under the impression that it's a problem with some sort of propagational delay. It can only be clocked so fast?

Paul Curtis · January 14th, 2008, 11:28 AM

steven,

there seems to be some confusion as to whether the skew gets better or worse with different settings, that's really the crux of the question. I believe it would be the same all the time because it's based on the read out speed of the sensor (there's talk about the SI camera getting around it by dropping every other frame, but i don't see how that would work either!)

i can understand the skew being hidden by more blur of 1/24th, but does the amount change with different settings too?

i remember the original sony marketing blurb talking about how they've minimised it (separate AD convertors and all sorts of technical jargon), but it looks just the same as all other cmos cameras! (includes the red too).

thanks
paul

Steven Thomas · January 14th, 2008, 11:34 AM

I agree...
I only mentioned 24P since motion blur tends to hide a lot with objects crossing the sensor at a fast rate.

I added more to my last post.. I was still writing as you just posted..

Eric Pascarelli · January 14th, 2008, 11:47 AM

I did some skew tests on a motion control pan head. It seems that skew is the same at all shutter settings. I'll post some grabs in a bit.

Paul Curtis · January 14th, 2008, 12:02 PM

>It can only be clocked so fast?

I don't know but it strikes me as the achillies heel for the EX1 (from my personal perspective with what i need it for -- i know for a lot of people it's fine).

Perhaps it's a great way for the manufacturers to break down their prosumer and pro lines and maintain the 'high end' costs.

I suspect it must be solvable up to a certain point, i mean the Arri D20 is cmos and there's been no mention of any issues with that anywhere. When there are more than a few hundred Reds out there it will be interesting to see what the real world results are like. The SI seems pretty good at minimising the effect so far, but i've not seen that much footage so i can't really say for sure.

paul

Noah Yuan-Vogel · January 14th, 2008, 01:09 PM

the d20 has a mechanical shutter and optical viewfinder like a film camera. the mechanical shutter removes the rolling shutter issue. it is interesting that none of these other cameras decided to use mechanical shutters, it is obviously more complex as it must be synced and controlled in software, but it would help fix a pretty significant issue in all these cmos cameras.

I wish companies would be a bit more transparent about things like the clock speed of the sensor. SI does this, they make it clear that there is a rolling shutter, but they run the camera at twice the working framerate to minimize artifacting. I dont believe RED has mentioned what their pixel rate is, they just say they are working on it and optimize for it in new firmwares.

i was wondering (might have posted about this elsewhere, dont remember) if there is a difference between the way the EX1 sensors are clocked in different modes. it seems possible, for example, that skew in 720p mode could be less pronounced, but that may not be the case at all depending on how it works. it would be great if someone could do some tests on this. who knows, maybe running the camera at 720@48fps and removing every other frame will help the issue a bit. hard to say till i get my hands on the camera, though.

Steven Thomas · January 14th, 2008, 01:20 PM

Jim Arthurs has posted some pretty clear cut examples of the RED ONE.
This is for real... nothing else needed. For skew, he determined if you keep the full frame pan time greater than 3 seconds, there should be no noticeable skew.

http://www.reduser.net/forum/showthread.php?t=7500

I haved a feeling the EX1 will be VERY close to these numbers.

If anyone feels this is a show stopper, don't board the bus!

It's a bummer no doubt, but If you plan on shooting a lot of 24P you're already thinking about 24P motion and how to minimize the judder.

The partial exposures on strobes, are another story...

Evan Donn · January 14th, 2008, 09:15 PM

Quote:

Originally Posted by Noah Yuan-Vogel

the d20 has a mechanical shutter and optical viewfinder like a film camera. the mechanical shutter removes the rolling shutter issue. it is interesting that none of these other cameras decided to use mechanical shutters, it is obviously more complex as it must be synced and controlled in software, but it would help fix a pretty significant issue in all these cmos cameras.

I was under the impression that mechanical shutters are prone to the rolling shutter issue as well because it's not physically possible for them to reveal the entire frame instantly - in this sense a CMOS should actually be closer to the characteristics of a mechanical shutter than an electronic one on a CCD.

Eric Pascarelli · January 14th, 2008, 09:29 PM

Evan,

This is true, but the farther the rotation shutter is away from the film plane, and the closer it is to the lens nodal point, the more it will approximate the global shutter ideal.

Many film cameras have a focal plane shutter and can have "rolling shutter" distortions and partial exposure of flashes, especially at narrow shutter angles.

But usually those focal plane shutters are rather large and would have the effect of "whipping open" and "whipping closed" with the film gate being only a small window of their operation, so the percentage of time that the shutter is moving over the gate is small compared to the time that the gate is fully open or closed.

January 14th, 2008, 10:05 AM	#1
Paul Curtis Major Player Join Date: Mar 2005 Location: canterbury Posts: 411	Rolling Shutter (Again) I hate to bring this up but i don't have one in front of me to check. http://www.freshdv.com/2008/01/xdcam...-examples.html the link shows some footage, overcranked at night time and a pan along with someone running. The rolling shutter is very visible, in fact to my mind unusably so. Can some tell me whether this clip is particularly unusual, is the effect so pronounced because of the overcrank? I always believed the amount of skew was constant because it's based on the native read times of the sensor regardless of shutter speed. I don't believe this type of camera move is unusual, views out of cars, panning with action the list is probably endless. In the various samples i've been watching i've not seen anything quite so objectional as this. Does anyone have any real life footage out of car window, or shots like this? The F335/350 et all are all CCD based aren't they? thanks in advance paul

January 14th, 2008, 10:59 AM	#2
Steven Thomas Trustee Join Date: Sep 2005 Location: Gilbert, AZ Posts: 1,896	Paul, there will be skew. I shot some soccer footage at 720 60P that was posted here for a while. During real fast pans, and slowed down, you could see about a 10 degree skew on poles in the background. If I had shot it in 24P, you probably would not be able to tell. That night running footage appears to show the same amount of skew I've witnessed under those conditions. To make that shot, you would of had to change the angle of where the camera sits; therefore minimizing the quick pan on top of the quick motion. This is a good example in showing this problem. Rolling shutter skew artifacts are there, but can be controlled to some extent. As you know, it's all about the speed of motion across the sensor. As it has been mentioned, the EX1 looks a lot like the rolling shutter artifacts seen on RED ONE. If you don't like the thought of this, stay with what you're using, or wait it out for something new to come along. All cameras have their good and bad. It's really about the ratio of the amount of time these problems will be viewable. Having low noise and high rez is needed 100% of the time. If I get an occasional rolling shutter artifact, I'll live with it. But, of course, I will do my best to avoid amplifying the issue. If you're shooting 24P, real fast motion crossing the sensor will have more image blur, which will hide "some" of the artifacts. With all it's quirks, I really think the EX1 is quite the camera! I'm really looking forward for a camera with this image quality to arrive without the rolling shutter issues. Now, only if there was a way to increase the CMOS sampling time via software upgrade.... OH well, wishfull thinking. I don't know much about this particular technology, but I'm under the impression that it's a problem with some sort of propagational delay. It can only be clocked so fast? Last edited by Steven Thomas; January 14th, 2008 at 11:31 AM.

January 14th, 2008, 11:28 AM	#3
Paul Curtis Major Player Join Date: Mar 2005 Location: canterbury Posts: 411	steven, there seems to be some confusion as to whether the skew gets better or worse with different settings, that's really the crux of the question. I believe it would be the same all the time because it's based on the read out speed of the sensor (there's talk about the SI camera getting around it by dropping every other frame, but i don't see how that would work either!) i can understand the skew being hidden by more blur of 1/24th, but does the amount change with different settings too? i remember the original sony marketing blurb talking about how they've minimised it (separate AD convertors and all sorts of technical jargon), but it looks just the same as all other cmos cameras! (includes the red too). thanks paul

January 14th, 2008, 11:34 AM	#4
Steven Thomas Trustee Join Date: Sep 2005 Location: Gilbert, AZ Posts: 1,896	I agree... I only mentioned 24P since motion blur tends to hide a lot with objects crossing the sensor at a fast rate. I added more to my last post.. I was still writing as you just posted..

January 14th, 2008, 11:47 AM	#5
Eric Pascarelli Major Player Join Date: Nov 2007 Location: Los Angeles, CA Posts: 498	I did some skew tests on a motion control pan head. It seems that skew is the same at all shutter settings. I'll post some grabs in a bit.

January 14th, 2008, 12:02 PM	#6
Paul Curtis Major Player Join Date: Mar 2005 Location: canterbury Posts: 411	>It can only be clocked so fast? I don't know but it strikes me as the achillies heel for the EX1 (from my personal perspective with what i need it for -- i know for a lot of people it's fine). Perhaps it's a great way for the manufacturers to break down their prosumer and pro lines and maintain the 'high end' costs. I suspect it must be solvable up to a certain point, i mean the Arri D20 is cmos and there's been no mention of any issues with that anywhere. When there are more than a few hundred Reds out there it will be interesting to see what the real world results are like. The SI seems pretty good at minimising the effect so far, but i've not seen that much footage so i can't really say for sure. paul

January 14th, 2008, 01:09 PM	#7
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	the d20 has a mechanical shutter and optical viewfinder like a film camera. the mechanical shutter removes the rolling shutter issue. it is interesting that none of these other cameras decided to use mechanical shutters, it is obviously more complex as it must be synced and controlled in software, but it would help fix a pretty significant issue in all these cmos cameras. I wish companies would be a bit more transparent about things like the clock speed of the sensor. SI does this, they make it clear that there is a rolling shutter, but they run the camera at twice the working framerate to minimize artifacting. I dont believe RED has mentioned what their pixel rate is, they just say they are working on it and optimize for it in new firmwares. i was wondering (might have posted about this elsewhere, dont remember) if there is a difference between the way the EX1 sensors are clocked in different modes. it seems possible, for example, that skew in 720p mode could be less pronounced, but that may not be the case at all depending on how it works. it would be great if someone could do some tests on this. who knows, maybe running the camera at 720@48fps and removing every other frame will help the issue a bit. hard to say till i get my hands on the camera, though.

January 14th, 2008, 01:20 PM	#8
Steven Thomas Trustee Join Date: Sep 2005 Location: Gilbert, AZ Posts: 1,896	Jim Arthurs has posted some pretty clear cut examples of the RED ONE. This is for real... nothing else needed. For skew, he determined if you keep the full frame pan time greater than 3 seconds, there should be no noticeable skew. http://www.reduser.net/forum/showthread.php?t=7500 I haved a feeling the EX1 will be VERY close to these numbers. If anyone feels this is a show stopper, don't board the bus! It's a bummer no doubt, but If you plan on shooting a lot of 24P you're already thinking about 24P motion and how to minimize the judder. The partial exposures on strobes, are another story...

January 14th, 2008, 09:29 PM	#10
Eric Pascarelli Major Player Join Date: Nov 2007 Location: Los Angeles, CA Posts: 498	Evan, This is true, but the farther the rotation shutter is away from the film plane, and the closer it is to the lens nodal point, the more it will approximate the global shutter ideal. Many film cameras have a focal plane shutter and can have "rolling shutter" distortions and partial exposure of flashes, especially at narrow shutter angles. But usually those focal plane shutters are rather large and would have the effect of "whipping open" and "whipping closed" with the film gate being only a small window of their operation, so the percentage of time that the shutter is moving over the gate is small compared to the time that the gate is fully open or closed.