"Rolling Shutter" Skew Tests and Grabs

[Ali Husain]

all right. i edited my original comment to be more fair. and i gave myself a todo to look through all the ex1 footage i have and study the motion blur, plucking out offending examples. good comments! thanks!

[Noah Yuan-Vogel]

I saw a presentation about the EX1 given by Juan Martinez from SONY at VCA in new york, and he did mention something about the exmor sensor reducing noise by scanning more than once per frame... although i cant seem remember specifically at the moment. i know he also said it reduces noise by taking a black voltage level for each pixel per exposure and subtracting it from the voltage output from the exposure on that pixel. i imagine one the former reduces analog noise and the later reduces fixed pattern noise? but i wonder if any of that is related to the frame accumulation motion blur effect you were seeing?

[Steve Mullen]

Originally Posted by Ali Husain

one thing i noticed(*) is that the motion blur on the ex1 has a strange triplicate-image look... as if someone took three exposures in quick succession then merged them together.

Perhaps you were in the mode that accumulates images to increase sensitivity. Any system that accumulates images can create either blur or image "copies" depending on the type of lighting. For example, flourescent lights might yield images.

PS 1: I'm increasing less sure about my claim that there is a difference between the way EXMOR and 3ClearVid work OTHER than the EXMOR has all digital output which reduces noise and, I suspect, decreases read-out time. If the read-out time is less, then indeed EXMOR may help reduce rolling shutter. But, I honestly do NOT remember Sony claiming rolling shutter is reduced.

PS 2: I'd love to the test done on a V1 to see if there is any difference.

[Christopher Barry]

On another forum, NYC night footage was posted. I d/l the clip, however, I can not find my way back to the post, so I just frame grabbed where I think the raised issue has occurred.

I will take this frame down once I locate the original post with a link to d/l. Anyway, I am not sure if I would be within DVinfo forum rules to link to another forum?

http://www.siliconcine.net/temp/Motion_Blur_EX1.jpg
Size: 200%

The rest of the footage in the clip is fine, actually amazing IMO, just this fast moving, close to the camera, taxi passing did register the stepping motion blur the first time I saw the footage. I note the specular highlights and edge detail appear to be making the stepping apparent in this instance. Other smoother tones on the taxi do not display any issue in earlier frames.

All said and done, I can not wait for the EX1's to arrive next week.

[Eric Pascarelli]

That's what i was talking about - a flickering light source. NYC (and most every city) is lit with fabulous sodium vapors which flicker 120 times per second.

I would guess that that clip was shot at 24p with shutter off - hence five cycles of the light per frame. Notice that the cab's taillight, at 12VDC does not exhibit the same stepping (because it does not flicker).

This would happen on any video camera.

[Christopher Barry]

Found the link: http://www.8112studios.com/nyc.mov

The thread: http://www.dvxuser.com/V6/showthread.php?t=121427

Look forward to further comments, now that the full clip is available to assess. I will remove my frame grab, in due course.

[Eric Pascarelli]

I won't change my comment except to say that looking at that shot, it might be an even harsher flickering source that is creating the multiple exposure - perhaps old fluorescents from a store.

Also, it is clear from the footage that the shutter is indeed turned off in the camera, because the blur from one frame connects with the blur on the next.

And so the 120Hz/5 images per frame flicker explanation makes sense.

[Christopher Barry]

Eric, this is very interesting information and discussion. Thank you.

[Christopher Barry]

Originally Posted by Eric Pascarelli

That's what i was talking about - a flickering light source. NYC (and most every city) is lit with fabulous sodium vapors which flicker 120 times per second.

I would guess that that clip was shot at 24p with shutter off - hence five cycles of the light per frame. Notice that the cab's taillight, at 12VDC does not exhibit the same stepping (because it does not flicker).

This would happen on any video camera.

Eric, the following Cloverfield article affirms your position that sodium-vapour lighting is the cause, and I refer you to the penultimate paragraph of the article:
http://www.videography.com/articles/article_15469.shtml

"When you do a fast pan with the Viper or the F23 under sodium-vapor lighting, you get a triple image," the cinematographer notes. "You get the image itself and then two trails."

Indeed, Viper and F23 suffer similarly. With the hand held approach in that movie, I can see a lot frames suffering with this issue. Cloverfield 'team roto' post nightmare for VFX crew.

[Eric Pascarelli]

Well there you go.

The reason they get three images (really 2.5) instead of 5 is that they were using 1/48 second shutters, as any feature filmmaker would.

[Bill Ravens]

So, armed with this knowledge, one could use 1/125 shutter speed and eliminate any multiple images.

[Eric Pascarelli]

Hmm. 1/125 sounds dicey. It's worth a test, and would depend on the duration of the light's "on" cycle vs. "off" cycle

If there are any incancesdents at all in the shot, I would stick to 1/48 to get the natural blur from those and let the flickering lights to their steppy thing.

[Sean Donnelly]

Eric I think you are right, it is pretty much impossible to predict the exact flicker rate of a light source (especially in a place with so many different types) so there is no arbitrary safe speed, but if you're in a room with a single source you may be able to phase it out using the ECS function. Also, I would go beyond saying that it would happen with any video camera, it will happen with film, both motion picture and still.

-Sean

[Bill Ravens]

Are any of you guys musicians? If you are, then you must understand harmonics. Harmonic principles in physics are what causes resonance and can generate amplification.

The same happens with any frequencies when they beat against each other. Two matching frequencies provide the most noticeable results(harmony)...as in the frequency of the fluorescent matching the frequency of the scanner. If you can get sufficiently above the nyquist frequency, which is 2x the base frequency,the beating goes away....go figure. In this case, the base frequency is 120Hz. Nyquist is then 240 Hz. I think 1/250 will do quite nicely. Of course, then you start experiencing other effects like wheels appearing to rotate backwards.

[Eric Pascarelli]

Bill,

I think you are confusing shutter speed with frame rate. Since frame rate is constant, frequency (of the camera) is constant.

If a wave metaphor applies at all to a camera, changing the shutter speed changes the shape of the wave, not its frequency. Like changing the duty cycle of a square wave perhaps?

1/250 would certainly eliminate the multiple images, but it would also render the exposure unpredictable because it could be photographing the lights on or the lights off, depending on how the camera is phased to the light source. But as Sean said, you could dial it in, just like when you shoot a CRT television.

Edit: But all of this goes to hell with the rolling shutter as Bob mentions below.

Sean,

Of course you are correct. This would happen on any camera, conditions being equal.