Global Shutter - is it possible for RED?

[Gavin Greenwalt]

I've seen footage corrected for rolling shutter and I don't remember thinking "Hey this looks a lot less sterile and more filmic."

My money is on "Global Sensors have other compromises such as streaking that is worse than the rolling shutter for most applications."

But I would be fine with "virtually global". I don't care about strobes you can always use something like FrameFixer to take care of that http://www.framefixer.com/downloads/framefixer_demo.mov. I just care about tracks.

 

[Rick Burnett]

Maybe the issue is, with a global shutter, you'd need to block incoming light for an entire read of the frame. This means all that information would be lost. You can't get that back. Does that cause strobbish type images?

Conversion utilities for fixing rolling shutter issues are dealing with something different, mostly skew in the image on left/right movements and compression expansion for up down movements. I don't think that is the sterile part of the image IMO.

I do have a question though, do all rolling shutters read at a constant rate? To me this would make no sense. With such a fast read/reset, I'd think you'd read in bursts at the end of exposing a frame instead of spreading that time between each row read that would exacerbate the rolling shutter problem.

 

[Charles Angus]

I have shot with global shutter CCD camera's, as have many others, and not had any problem with the images looking sterile due to lack of skew.

I don't really think there can be an aesthetic argument made in favour of rolling shutter. Engineering arguments, perhaps.

I for one am a little disappointed that the optional global shutter seems to have gone by the wayside.

 

[Jannard]

I have shot with global shutter CCD camera's, as have many others, and not had any problem with the images looking sterile due to lack of skew.

I don't really think there can be an aesthetic argument made in favour of rolling shutter. Engineering arguments, perhaps.

I for one am a little disappointed that the optional global shutter seems to have gone by the wayside.

I'd love to hear what sensor you wish RED could emulate. Remember, you have sensitivity, noise level, frame rates, and image size to consider...

BTW... a global shutter never was on our roadmap. Only faster read-reset sensors.

Jim

 

[Colin Hubick]

Rolling shutter for us is a MAJOR problem when it comes to plate work and vfx tracking shots.....really REALLY need this issue to go away. it takes so much time trying to get the "wobble" out of shots that it's best to forget handheld work for VFX shots.....I hope it's a faster read/reset time soon.

I echo this comment.

 

[Jannard]

I echo this comment.

Understanding the problem... what sensors of the right size, frame rates, etc. do this? Everything is a compromise. We fully understand that faster read-reset times (as we have done on several occasions) are important... so is everything else.

Saying you want zero read-reset time (global shutter) but not lose anything else is like saying you want 5K at 500fps, no noise to ISO 12,000 and 20 stops of dynamic range. Someday you'll have that. Until then, we are picking up each of these pieces as technology allows.

Jim

 

[Colin Hubick]

Understanding the problem... what sensors of the right size, frame rates, etc. do this?...

Jim

None, no camera does this. This is why we push it here, because finally there is a camera manufacturer that is making "that" camera. Is it here yet, not totally, but it progresses, and that's important.

Red as a camera system has only been around for 3 years (give or take) and film for 100 years. It takes time to develop the technologies, and I understand that. IMO, as long as there is a forum to discuss with the manufacture, ideas and issues, Good Bad, and Ugly. I will use it. I love seeing the tech progression, and the "positive" discussion and debate over these topics.

cheers
Colin

 

[Rick Burnett]

Did I miss something? People keep implying that the skew is the characteristic that makes things less sterile, but I didn't get that from Jim's post. I think it's safe to say that the absence of skew/wobble on the global shutter approach is a benefit completely. Skew and wobble in my opinion add NO desirable characteristic.

 

[Jannard]

Did I miss something? People keep implying that the skew is the characteristic that makes things less sterile, but I didn't get that from Jim's post. I think it's safe to say that the absence of skew/wobble on the global shutter approach is a benefit completely. Skew and wobble in my opinion add NO desirable characteristic.

Skew has become an accepted "look" for 100 years with every film camera. It is an anomaly not unlike anamorphic... an embraced imperfection. It doesn't mean that it can't disappear in the future... and you will probably have a choice either way pretty soon. But what you have seen in theaters all of your life is one with a certain amount of skew produced either by a mechanical shutter or read-reset time.

Jim

 

[Rick Burnett]

Interesting. Honestly, I have NEVER noticed it before. I guess with the extent of how horrible it is on all 4 CMOS cameras I own, it drew attention to it with my eyes. Or maybe because that is always how it has been, it's what I come to expect.

But it makes sense. I've never shot film so I had never looked at how the film camera is working, only CCD previously, but the image on this page:

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

Certainly looks like it operates the same as a rolling shutter. (Well, closely, the film camera completely blocks the frame while the frame advances, the whole 180 thing. The RED could shoot at rates up to the read/reset time of a single row. Not that it is desirable, but available especially if you will trade some of the image difference for more exposure time).

Thanks for pointing that out.

I guess this is a good illustration of people like me coming from the "VIDEO CAMERA" world in to the "FILM CAMERA WORLD" and realizing the differences. No one really explained that before (or I missed it)

This would seem to imply that film cameras in fact could get jello, have problems with strobe lights, and also, I notice the shutter is not square with the frame at all times, it is angular as it passes through the frame, which surely must have some impact in certain conditions. Is this true?

Regardless, I CANNOT WAIT to get a camera with reduced read/reset. I know the difference in skew in my D90, 7D and HV30 are VERY different. I'd be happy to even have the performance of the HV30. The D90 is SO bad it's not even funny. I tried to find any sort of comparison on read/reset on those cameras I mention out of curiosity but that information seems to not be shared or evaluated.

Will the Scarlet 2/3 Cinema be looking at 5ms as well?

 

[Jannard]

Interesting. Honestly, I have NEVER noticed it before. I guess with the extent of how horrible it is on all 4 CMOS cameras I own, it drew attention to it with my eyes. Or maybe because that is always how it has been, it's what I come to expect.

But it makes sense. I've never shot film so I had never looked at how the film camera is working, only CCD previously, but the image on this page:

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

Certainly looks like it operates the same as a rolling shutter. (Well, closely, the film camera completely blocks the frame while the frame advances, the whole 180 thing. The RED could shoot at rates up to the read/reset time of a single row. Not that it is desirable, but available especially if you will trade some of the image difference for more exposure time).

Thanks for pointing that out.

I guess this is a good illustration of people like me coming from the "VIDEO CAMERA" world in to the "FILM CAMERA WORLD" and realizing the differences. No one really explained that before (or I missed it)

This would seem to imply that film cameras in fact could get jello, have problems with strobe lights, and also, I notice the shutter is not square with the frame at all times, it is angular as it passes through the frame, which surely must have some impact in certain conditions. Is this true?

Regardless, I CANNOT WAIT to get a camera with reduced read/reset. I know the difference in skew in my D90, 7D and HV30 are VERY different. I'd be happy to even have the performance of the HV30. The D90 is SO bad it's not even funny. I tried to find any sort of comparison on read/reset on those cameras I mention out of curiosity but that information seems to not be shared or evaluated.

Will the Scarlet 2/3 Cinema be looking at 5ms as well?

Many CMOS are well above 30ms (jello). The RED ONE is about 9ms. Mysterium-X (including Scarlet 2/3") is 5ms. Film is 4ms.

Jim

 

[Rick Burnett]

Oh wow, that is quite a bit more. That's excellent to hear about the Scarlet that it is the same!

I've seen mention that the rolling shutter is a constant speed (mentioned this before), but to me, that doesn't seem to make sense. I'd think you'd expose the frame, then zip through the rows as fast as possible to minimize skew between each line. Given that exposure time is selectable, does the sensor just go through another sweep of resets during the time when the image should be ignored just before you start again for the next frame?

Another thing with using a global shutter is that it would seriously limit the exposure time at the faster frame rates. If you were shooting at 120fps, for 1/240 you'd need to expose a frame for 4.16ms and have up to 4.16ms to capture the data. Clearly, this wouldn't be possible without a rolling shutter with this read/reset time. Since each line I'd imagine would take around ~1/1600 for the vertical row count of the scarlet, that's like 2.6uS of down time for the row with a rolling shutter. That's a HUGE difference.

And of course, 150 burst. Is there an explanation somewhere of what 150 burst is and what that means in terms of how much time? No other camera I have used has that feature so I wasn't sure what to expect of it.

 

[Steven Caesare]

Many CMOS are well above 30ms (jello). The RED ONE is about 9ms. Mysterium-X (including Scarlet 2/3") is 5ms. Film is 4ms.

Jim

Jim...

Given that the sensors in the Scarlet and Epic are both of the M-X variety (but of different sizes), is there a reason that the sensor with 40% fewer rows to read out wouldn't have a faster R/R time?

Or is it perhaps the electronics package (D/A converters, etc...) in the less expensive Scarlet doesn't need to be as fast as the Epic because of the fewer pixels?

Hmm, I've never thought about it, but what is the R/R time on an R1 in 2K windowed mode...?

Thanks.

-sc

 

[Michael McVicar]

Jim...

Given that the sensors in the Scarlet and Epic are both of the M-X variety (but of different sizes), is there a reason that the sensor with 40% fewer rows to read out wouldn't have a faster R/R time?

Or is it perhaps the electronics package (D/A converters, etc...) in the less expensive Scarlet doesn't need to be as fast as the Epic because of the fewer pixels?

Hmm, I've never thought about it, but what is the R/R time on an R1 in 2K windowed mode...?

Thanks.

-sc

I've been wondering the same things. However with Scarlet we will see real world testing when we see it, not speculation. The R1 window mode could be tested now. Anyone done one by chance?

 

[Charles Angus]

I'd love to hear what sensor you wish RED could emulate. Remember, you have sensitivity, noise level, frame rates, and image size to consider...

BTW... a global shutter never was on our roadmap. Only faster read-reset sensors.

Jim

I appreciate your taking the time to answer our points on this thread.

I'm only talking about read-reset times here... Not talking about sensitivity, noise level, frame rate, or image size. I realize that these points likely can't be addressed in isolation, but RED is already hitting the all the points you mention out of the park with M-X.

As I said, however, I see no downside to a global shutter. Maybe some people think it looks sterile, but I would wager that all lay-people and most filmmakers would be unable to tell the difference in normal circumstances, and would prefer global shutter in "challenging" situations (strobes, flashes, vibration, fast panning, etc.).

No problem... I have resigned myself to having to de-bunk the "rolling shutter myth". People tend to hang on to soundbites. We could easily do a global shutter. It just isn't the right solution. If we made a global shutter sensor, everyone would wonder why the footage looked sterile.

One of the options we have is to give the user adjustable read-reset times... just so we don't hear that we are "marketing". From zero (global) to something (what we recommend and looks right).

Jim

I think the upshot of threads like this is that that's an option people would like to have - even if you folks feel like it's not necessary.

That's all.

 

[Charles Angus]

Another thing with using a global shutter is that it would seriously limit the exposure time at the faster frame rates. If you were shooting at 120fps, for 1/240 you'd need to expose a frame for 4.16ms and have up to 4.16ms to capture the data. Clearly, this wouldn't be possible without a rolling shutter with this read/reset time. Since each line I'd imagine would take around ~1/1600 for the vertical row count of the scarlet, that's like 2.6uS of down time for the row with a rolling shutter. That's a HUGE difference.

The Phantom v640 can do 1500fps and has a global shutter, so I'm not sure why high frame rates would necessitate a rolling shutter.

 

[Jannard]

I appreciate your taking the time to answer our points on this thread.

I'm only talking about read-reset times here... Not talking about sensitivity, noise level, frame rate, or image size. I realize that these points likely can't be addressed in isolation, but RED is already hitting the all the points you mention out of the park with M-X.

As I said, however, I see no downside to a global shutter. Maybe some people think it looks sterile, but I would wager that all lay-people and most filmmakers would be unable to tell the difference in normal circumstances, and would prefer global shutter in "challenging" situations (strobes, flashes, vibration, fast panning, etc.).



I think the upshot of threads like this is that that's an option people would like to have - even if you folks feel like it's not necessary.

That's all.

The point is that we (or seemingly no one else) can do a sensor with all the specs listed above AND a global shutter. Pick your poison. Can't have them all... today.

Jim

 

[Charles Angus]

Understanding the problem... what sensors of the right size, frame rates, etc. do this? Everything is a compromise. We fully understand that faster read-reset times (as we have done on several occasions) are important... so is everything else.

Saying you want zero read-reset time (global shutter) but not lose anything else is like saying you want 5K at 500fps, no noise to ISO 12,000 and 20 stops of dynamic range. Someday you'll have that. Until then, we are picking up each of these pieces as technology allows.

Jim

Just wondering what some of the trade-offs would be?

 

[Jannard]

Just wondering what some of the trade-offs would be?

As stated above... sensor size and frame rates, sensitivity or DR. In the future you will have more of everything, but right now it appears that M-X is the best compromise (in our eyes).

Only proof I have? ... name a global shutter sensor that comes close.

Jim

 

[Charles Angus]

Wow - thanks for the fast response!

Communication like this is one of the things that really makes me happy to shoot RED.