Apparently, you need to learn that skew and rolling shutter are not the same.
GH2 = No rolling shutter/jello ?
- Thread starter BWyldeck
- Start date
Apparently, you need to learn that skew and rolling shutter are not the same.
Rick Burnett
Section Moderator
The concept of the spinning physical shutter of a 35mm film camera is the same as the concept of a rolling shutter. If it takes 4ms for the edge of the shutter to pass all the way down the face of the film, the difference in exposure from the top of the film to the bottom of the film is going to be 4ms, it's that simple.
if a CMOS sensor takes 4ms to scan ALL the lines, the difference in time from the top line to the bottom line is 4ms. Same concept. The bottom line is still recording for 4ms more than when the first line started to get read.
It's not like a single slit being read at a time because the sensor is not recording for the whole time. If your framerate is 24fps and your exposure is 1/48 then clearly the sensor isn't being used for 1/48 of the time. I imagine it scans down the whole sensor, waits then scans down a reset to work just like a spinning shutter would.
If the film and CMOS sensor are both rated at 320ISO, then the results would be very similar. Sure, there are other physical differences like how the edge of the physical shutter diffracts the light at the edge and the fact film doesn't record light linearly at the top and bottom of the curve. AND, the fact that you can use film with different speeds which you cannot do with a CMOS sensor. It is what the sensor is rated at. But these concepts are pretty simple. Of course, no CMOS camera reads EVERY line as fast as the edge moves across the film, but Red claims to be at 5ms now with the Epic. (Without actual testing or proof, this is something that no one knows for sure outside Red).
I also read that Arri is using a physical shutter in some of their cameras. I am looking forward to seeing how this works.
Not all rolling shutters are created equal, that's for sure. In terms of worst to best, iPhone 4, D90, 7D and HV30 are all VERY different. This is something that almost all camera providers DO NOT rate. Unfortunately.
I agree with Barry that eventually as the read/resets get fast enough, eventually rolling shutter won't be an issue anymore, or at least, barely perceivable.
The 35mm shutter speeds I mention are not from my own experience but paraphrasing Graeme at Red. Personally I am in talks with a production at SCAD at the moment to buy some time on a current project happening there. Specifically I am trying to get access to some time with their 35mm camera AND Red One (with original sensor) along with my 7D to do some skew tests since this information is not readily available and I cannot get anyone to do the test for me with a 35mm camera.
I have read assertions from many people that 35mm has skew as well, but have not been able to find any studies on it. I'd be glad for anyone to point me to work already done so I don't have to spend money on this as well. I'd like to end this debate once and for all. (either way)
if a CMOS sensor takes 4ms to scan ALL the lines, the difference in time from the top line to the bottom line is 4ms. Same concept. The bottom line is still recording for 4ms more than when the first line started to get read.
It's not like a single slit being read at a time because the sensor is not recording for the whole time. If your framerate is 24fps and your exposure is 1/48 then clearly the sensor isn't being used for 1/48 of the time. I imagine it scans down the whole sensor, waits then scans down a reset to work just like a spinning shutter would.
If the film and CMOS sensor are both rated at 320ISO, then the results would be very similar. Sure, there are other physical differences like how the edge of the physical shutter diffracts the light at the edge and the fact film doesn't record light linearly at the top and bottom of the curve. AND, the fact that you can use film with different speeds which you cannot do with a CMOS sensor. It is what the sensor is rated at. But these concepts are pretty simple. Of course, no CMOS camera reads EVERY line as fast as the edge moves across the film, but Red claims to be at 5ms now with the Epic. (Without actual testing or proof, this is something that no one knows for sure outside Red).
I also read that Arri is using a physical shutter in some of their cameras. I am looking forward to seeing how this works.
Not all rolling shutters are created equal, that's for sure. In terms of worst to best, iPhone 4, D90, 7D and HV30 are all VERY different. This is something that almost all camera providers DO NOT rate. Unfortunately.
I agree with Barry that eventually as the read/resets get fast enough, eventually rolling shutter won't be an issue anymore, or at least, barely perceivable.
The 35mm shutter speeds I mention are not from my own experience but paraphrasing Graeme at Red. Personally I am in talks with a production at SCAD at the moment to buy some time on a current project happening there. Specifically I am trying to get access to some time with their 35mm camera AND Red One (with original sensor) along with my 7D to do some skew tests since this information is not readily available and I cannot get anyone to do the test for me with a 35mm camera.
I have read assertions from many people that 35mm has skew as well, but have not been able to find any studies on it. I'd be glad for anyone to point me to work already done so I don't have to spend money on this as well. I'd like to end this debate once and for all. (either way)
Barry_Green
Moderator
In the BTS, there's a shot of a 35mm camera.
It's entirely possible that parts of it were shot on one camera, and the rest on another.
It would be infinitely more informative if we could limit the film candidates to those shot before CMOS became prevalent, so -- anything made in 2005 or earlier, for example, should be CMOS-free.
So, let's eliminate any variability and ask for film examples of skew from the pre-CMOS era. Do you know of any examples of films from before 2005 that exhibit skew?
Barry_Green
Moderator
In flashes, there is a potential for partial exposure, yes. But not skew. Not jello.
You can get partial exposures on a flash on a movie camera because it has a rotating mirror shutter, and it's entirely possible that the flash will go off when the mirror shutter has rotated to a position such that the shutter is blocking part of the frame. But the issue is far less likely to occur on film than it is in a CMOS camcorder; the CMOS camcorder will show the effect on every flash ever shot; the movie camera will not. It will only happen when the leading or trailing edge of the shutter is obscuring part of the frame.
Again, it's just not reasonable to think that movie film is going to go showing skewing and jello like a CMOS camcorder does. The shutters don't work the same way. The movie camera exposes its frame globally, or mostly-globally, for a high percentage of the time, and it is that very difference (global exposure vs. rolling-line exposure) that makes it different.
Rick Burnett
Section Moderator
Absolutely agree. The difference between rolling shutters on all the CMOS cameras I have used have been VERY different. I absolutely HATE the video camera on my iPhone the rolling shutter is SO bad.
And i did find a good example of film skew:
http://www.youtube.com/watch?v=cD-uQreIwEk
Around 1:11 in the trailer, there is a VERY FAST pan across an office. While the shutter speed is pretty open, so there is a lot of blur, watch it a few times, you will DEFINTELY see the leaning and straightening as it comes to stop.
Also, at 1:26, there is a very fast pan where the woman is on the left edge of the screen. Again, the skew is very noticeable.
The issue at hand of course is that we are used to these spinning shutter speeds, and this is what we expect. Just like most 35mm operators are going to know how to get the best looking shots, how fast to pan, and also, what shutter speeds to avoid. We do this know with our 7D and we keep skew out of our productions pretty well.
It's like people who don't recognize juddering on movies that have been converted to NTSC frame rates. I see it, but other people don't because they are so used to pan juddering from watching years of movies on television.
Barry_Green
Moderator
Studies? I haven't been able to find a single EXAMPLE of it.
The only thing that's ever been shown, to reinforce that assertion, is an old still frame of an ancient race car, and it's skewed because the still camera used a focal-plane curtain shutter.
Agreed. It's simple enough to put a film camera next to a rolling-shutter camera, set them to equivalent exposure shutter angles, equivalent field of view, and point them at something that's known to cause skew. I used to subscribe to the idea of waving them back and forth at a known speed, but recently I've decided that's still just too variable. Instead, the way I did it in the HPX370 vs. EX3 comparison is my preferred method now -- put a big telephoto lens on 'em and point them at a 90-degree angle to a freeway and capture some big semis barrelling past at 70+ MPH. If you get the cameras far enough away from the freeway and use long telephoto, you'll minimize any parallax differences, and will be getting an effectively straight shot at a known/contained/constant-speed object that should be rectilinear.
I'd do the test myself but I sold my 35mm camera off last year. I do still have some 16mm & super16mm cameras, and the priniciple is exactly the same as in 35mm, so I might try that, but I'd have to dig up a Red (unless people would be happy with the comparison against an HMC40 or GH13).
Last edited:
Ian-T
Veteran
I thought about that also. There were some scenes where window shots were "totally" blown out. That screamed video to me... And then came the skew. I thought to myself "this can't be 5Dll footage" especially since the movie was made in 2007/2008.
True. Unfortunatley I know of no example.
I can't think of any. It was this same movie I was thinking about when I made my comment earlier. But you could be correct about them using multiple types of cameras.
Barry_Green
Moderator
Not convinced. Because that looks like it's probably a steadicam shot, and it could very well be that the camera itself is what's leaning during the high speed pan, due to the motion of the steadicam.
Same thing.
The only way to answer the question definitively is to eliminate the variables. That's why I say it should be tested locked down, pointed at constant-speed moving objects.
Isn't it possible that the fact you can only think of one movie which you say exhibits skew could mean that the effect is stemming from something else entirely than the shutter system on a 35mm camera?
Rick Burnett
Section Moderator
The same principle is happening in BOTH situations. The shutter is doing the same thing, just current CMOS sensors have essentially a slower moving shutter edge.
Like I said before, if you are exposing for 1/48 that is 20.8mS. 4mS of that is 20% of the time your frame is being exposed. Shooting at smaller angles and you are DEFINITELY going to get a lot more noticeable skew.
The 4ms comes from Jim at Red, so I cannot say how accurate that is. But I don't see any evidence to refute it as well. Just like shooters of the 7D and GH1, people who shoot film know how to shoot well. They are going to shoot within the limitations of the film as well.
After looking at the Bourne footage closer, I can definitely see skew. I am so used to watching 35mm films that I have to actually look for it. But it's there. The faster the CMOS read reset gets, the more film like it will be. The difference between 5ms and 4ms can still be a lot.
Rick Burnett
Section Moderator
The girl isn't tilted in that shot. I would have pointed out the bus moving by, but the angle of the bus movement makes it difficult to know if the leaning is due to something else.
I agree that a locked down test is the BEST thing. Waiting to hear back. I'll also try different shutter angles as well just so it will be easy to see for some of the tests to compare between the other cameras as well. Actually, it would be best to test it with as small an exposure that I can match just so the angles can be measure easily. (since all I will be doing is removing the blur)
Ian-T
Veteran
Yes that is skew. It's very noticable on the computer monitors when I freeze it. It skews a lot and when stopped it is back straight up.
It's there also but not as much as the quick pan in the office.
Ian-T
Veteran
If the entire movie was shot on film then I'm not sure what else could have produced that type of skew. But it was noticible enough for me to raise the red flag.
Ian-T
Veteran
Barry, I agree with you about how to conduct the test. But wouldn't it work out the same way with film (to a lesser degree)? What I mean is...the more telephoto you go on film you should get the same type of artifact with quick pans.
On my old HV20 skew (and more so jello) was really more of a problem when I'm in telephoto mode.....but wide shots faired well. I would think film would be similarly affected during long telephoto shots but not nearly as much.
Martti Ekstrand
Veteran
I achieved image skew in whip pans with a 16mm camera set to a very narrow shutter angle. Had I thought to whip pan back and forth I have no doubt it would resulted in something similar to jello. Was back in the 80s so doubt I can dig up the material now and even if I did I don't think I want to pay for the telecine. 
The problem with photo flashes yielding partial exposures on film is not uncommon at all by the way.
The problem with photo flashes yielding partial exposures on film is not uncommon at all by the way.
Rick Burnett
Section Moderator
http://www.youtube.com/watch?v=kx7dFp0WhN4
5:29
Look at the bar handle that is spinning. If you stop it when it is first straight up and down, you can see the curve in it. More near the beginning and hard to pause there, but i was able to see it.
I think though the bigger difference between the two is compression/expansion. I've never been able to detect horizontal compression with my eyes like a CMOS sensor exhibits at the current speeds. Will definitely have to test that as well.
5:29
Look at the bar handle that is spinning. If you stop it when it is first straight up and down, you can see the curve in it. More near the beginning and hard to pause there, but i was able to see it.
I think though the bigger difference between the two is compression/expansion. I've never been able to detect horizontal compression with my eyes like a CMOS sensor exhibits at the current speeds. Will definitely have to test that as well.
Blackout
Veteran
Ahhhhh... but if you move it around with the shutter open low light ya kinda get 'jello trails' but I guess you are right it's not the same artifact.
But what about this, this was shot on 35mm film and it CLEARLY shows jello:
So don't tell me it's IMPOSSIBLE to get jello on 35mm film...
:2vrolijk_08: