sample of unreasonably shakey footage showing rolling shutter distortion

EDIT: also reference this new post: http://dvinfo.net/conf/showthread.php?t=109613

because people asked me to post this footage:

even if there wasn't any rolling shutter distortion, this footage would look terrible on any camera, video or film.

i repeat: even if there wasn't any rolling shutter distortion, this footage would look terrible on any camera, video or film. please be fair and remember that!

http://www.bakraa.com/ex1supershakey/
(17 seconds and 12MB in wmv format)

the file is not the raw mp4, which was quite a bit longer and bigger... i couldn't figure out how to get vegas to chop it up for me without recompressing it. so i gave up and recompressed the clip to wmv at 6mbps/10mbps avg/peak.

settings:
720p24 (HQ-35MBPS VBR), probably f1.9 or close, probably around 1/120th sec. shutter (might have been 1/250th). it was in cinegamma1, colormatrix=cinema, i turned off detail (like any sane person :) ). and of course image stabilization was off. manual focus (focus assist off too). manual iris. its not cc'd. i don't remember if there was any gain added, probably not. it was around sunset and getting dark.

i zoomed in to the end of the range, and jiggled my wrist to shake the camera around it's axis of rotation. to get the most pathological looking distortion you need:

(1) to use a fast shutter speed (to get the least motion blur distortion of the distortion :)),
(2) to keep the same subject in the frame, and
(3) to accelerate the image in one direction then the other.

this requires small, very quick accelerations. if you wildly move the camera around, the distortion is less visible, because you're not seeing the same object squish and then unsquish. it's hard to accelerate the image like this by hand and keep the same subject in the frame by merely translating it in space: rotation helps; and you cannot accelerate the image this quickly by hand without being on the long end of the zoom for leverage.

i don't want to be the cause of people bashing this camera. so i feel compelled to say that:
if you somehow require footage to look terrible in the same way shakey film would look, then you should rent a film or global-shuttered video camera for that niche purpose. the image from this camera is stunning and richly detailed, the codec shows little apparent artifacting, and i can cc the snot out of footage at full-resolution and it holds up. compared to the cameras available just one month ago, it is a minor miracle. you should try it out for yourself on a real project and see what you think.

Thank you Ali,
Actually, I was surprised it looked that good!

For those conditions, it looks good to me.

Ali, that's great of you, thanks

Won't be mounting it on a helicopter any time soon :)

Have you tried just normal shaky handholding, something more useful, running with the camera, documentary style?

I think the camera is fantastic in general, but i like to know when it won't work first...

paul

I was wondering if you could perform the same sort of torture test with everything the same except for turning on image stabilization. That one feature alone might be part of the solution for eliminating this effect.

-gb-

Have you tried just normal shaky handholding, something more useful, running with the camera, documentary style?

I took a few hand-held shots at a wedding this weekend including following some groomsmen up a stairway and along a hallway, and didn't see anything objectionable in the resulting footage. I did have one shot panning up the wall of a church which might exhibit a little of this effect, but it's hard to tell viewing the clip on my laptop. Another clip a friend shot involving both spinning and panning the camera looked fine, so I'd say this is mainly a case of how violently you move the camera - back and forth in rapid succession appears to be the worst example.

On a related note, I haven't seen any sign of vignetting in any footage shot under normal conditions, but did see a hint of it following the test procedure recommended here. So for both of these issues you have to go looking for the problem to find it, which should be kept in mind when assessing the significance of such reports. After looking at all of my footage from this weekend I'm thinking this is a great camera which is going to be very popular, and like any piece of gear you'll just have to learn to respect its limitations.

Thanks Ali for sharing.
Frankly, I find the the rolling shutter effect much worse than expected. I really hope that in normal movement shots these artifacts are completely gone.

Like Greg, I would be interested to know what people's experience is with the OIS on. How much does it help reduce these kind of distortions...10%....80%?

My sense is that if you can create the same image with the OIS on - if you can shake the camera so much that the shaking overwhelms the OIS and the image looks like Ali's clip, then the distortions will be the same.

If OIS does its job and reduces the visible shakiness, then the distortion will be reduced accordingly.

The distortions happen in the way the chip is read, which all happens after all of the optics of the camera (including OIS) have done their thing. Camera movement is not relevant - it's image movement that causes the distortion.

Imagine a stationary camera with the world shaking around it. OIS would do nothing to change the movement of the image, whether on or off, and the image would still be distorted by the CMOS.

In my use of the EX1, the OIS is pretty good, but not as "powerful" as some others I have tried.

Wow, Ali... I think you need to cut back on the caffeine! Your hands are shaking! ;-D

Serious question, though... If you use a slower shutter speed (or shutter off) does it mask these distortions by motion blurring them away?

Thanks Ali, for providing a clear example of the wobble everyone's been talking about. Given how much shaking you had to do, I don't think this is a serious issue.
But you might take Alex's advice, and cut back on the caffeine.

Ken Hull

Thanks Ali.
Excellent earthquake simulation footage! You can really see the shockwaves coursing through the landscape. News at 11:00...
Wayne Z

Exactly why there need to be global shutter CMOS chips... Thanks for sharing.

Looks just like a bad cell phone video...

Exactly why there need to be global shutter CMOS chips... Thanks for sharing.

Looks just like a bad cell phone video...

LOL...
guess I better sell mine on ebay for $50.

give you 100$ right now! ;-)

I was wondering if you could perform the same sort of torture test with everything the same except for turning on image stabilization. That one feature alone might be part of the solution for eliminating this effect.

-gb-

hi greg, unfortunately i only had the camera for a few days on rental for a shoot.

Ali, that's great of you, thanks

Won't be mounting it on a helicopter any time soon :)

Have you tried just normal shaky handholding, something more useful, running with the camera, documentary style?

I think the camera is fantastic in general, but i like to know when it won't work first...

paul

i put up some footage like that and documented it here:
http://dvinfo.net/conf/showthread.php?t=109613

it's probably a bit more shakey than someone would normally shoot though (and the shutter speed is higher).

Wow, Ali... I think you need to cut back on the caffeine! Your hands are shaking! ;-D

Serious question, though... If you use a slower shutter speed (or shutter off) does it mask these distortions by motion blurring them away?

yes, the slower shutter speed helps a bit.

i put up some footage like that and documented it here:
http://dvinfo.net/conf/showthread.php?t=109613

it's probably a bit more shakey than someone would normally shoot though (and the shutter speed is higher).

Ali, that's so kind of you, a real help

It's pretty good, i can still see the effect but it's acceptable which is all i needed to know. I see this artifact as being one of the major differences between this and the other XDCAMs (which i believe are all CCD based)

I know i've been going on about the shutter but it's only because i know i'll have some very specific requirements next year (which i don't think it will cope with) but on the other hand what you get for the cost will deal with 95% of what i need.

thanks again
paul

This is a very interesting thread in terms of seeing just how well the camera can deal with extreme movement, however unless you were going for that kind of look explicitly (and perhaps the "nervous hands" look might be better done in post anyway ;-) I would think that you'd have the camera on something like a SteadiCam (Jr.) or one of those backpack-with-hanging-bungee-cord type rigs (I forget the name of the one I've got) to try to dampen the shakes due to running or not enough coffee/drugs in the morning.

Sure, if you screw the camera to the side of a helicopter rotor blade and then notice that the footage isn't pristine, that's not really the fault of the camera, you've just exceeded it's operational envelope just a tad ;-)

Thanks for all the clips though, it is good to understand the limits of the camera (and any gear that you may be using).

mike

Pretty good for shaking, but the back of the goat is blown out.

Knowing the limits of CMOS chips, and knowing that the shaking is intentional, a production team would think the process through and consider doing the shaking in post.

Two advantages: avoid the rolling shutter artifacts. And make the shaking controllable. You can adjust the shaking in post to get the desired effect. But if you do the shaking in-camera, even with a CCD camera, you're stuck with it.