Reading mode: Light Dark

video jello

Go to last post
Greg Pavlov said:
What does CMOS have to do with this?
Click to expand...
Nothing really, other than the fact that it can occur in CMOS imaging chips - it doesn't have to though.

The effect is actually a mismatch of imaging and display technologies, and is not really a problem with one or the other, it is the combination of both.

30 years ago, when I first started in the sensor design business this same effect was actually a criticism of early CCD cameras. How come?

Well, the predominant display technology of the time (in fact almost the only image display technology) was the CRT and that had a sort of "rolling shutter" display. The image was written on the CRT in a raster fashion from top to bottom. This hadn't been a problem in the previous 40 years of television because until then all earlier cameras had also been based on CRTs: the vidicon and its many derivatives with the image being read off the faceplate in the same raster fashion. The image built up on the faceplate between each read by the electron beam, so this was a rolling shutter with a shutter speed exactly the same as the frame rate.

Using both of these CRT based image and display technologies the time delay between the top of the frame being captured and displayed was exactly the same as the time delay at the bottom of the frame, so everything "looked" right.

Of course, if you recorded the video and played it back in slow-mo or freeze framed then you could see the time delay in the vidicon tube between the top of the image being read and the bottom of the image being read - so panning across verticals caused them to tilt backwards in still frames: the top of the image was recorded before the bottom of the image, which had panned a little further in the delay between the two. However, this was perfectly compensated on the CRT display during normal playback so that the tilt on each vertical appeared upright again due to the delay as the CRT wrote the image on the screen.

Then, along came CCDs in several different varieties: Interline transfer, frame transfer etc. One thing that was common to all of the CCD designs was that the image was captured on all of the frame simultaneously. So, when panning across a vertical edge, that edge stayed upright on every single still frame. However, when displayed on the raster updated CRT display, the top of the image was written an entire field period before the bottom of the image. The result was that continuous video of pans appeared to tilt vertical lines forwards.

Initial CMOS image developments for video sensors attempted to correct this problem by reading the image in the same sequence as the old vidicon tubes, sequentially with a rolling shutter scanning from top to bottom. That worked well when the images were shown in standard time on displays which also updated in raster fashion, and most video displays were CRTs so the "problem" went away for a while. As with the old vidicon images, still frames from mid-pan shots still showed the same tilt, but the CRT display fixed that when the movie ran in normal time.

Nowadays, many flat screen displays such as LCDs and plasma, update the entire image simultaneously, with the data first being loaded into a frame store. That is why the rolling shutter CMOS design, which was originally implemented to correct the same deficiency in CCDs, causes this same visual effect - in fact, it is the opposite effect and the tilt is reversed. And the effect gets worse as the frame or field rate is reduced - so you see less leaning verticals with 60i than with 30p.

So, when you combine the "problem" of the CCD with the "problem" of the flat panel display you get the best of both worlds - vertical still frames that stay vertical when panned in moving footage.

There is no reason why CMOS sensors cannot be designed to have a similar "snapshot" mode of operation as CCDs and, indeed, many do. Some even support both modes. They can do this while the exposure time is a small fraction of the frame time - all it needs is the exposure control line for each row taken out to a shift register at the edge of the chip rather than globally connected.

However, to operate in low light both sensor types need to capture the image for as large a proportion of the frame time as possible. Neither sensor type can image while it is being read out. With CCDs, this was resolved using interline or frame transfer structures - blind areas of the sensor that the image data was quickly transferred to before being read out while the next frame was being captured in the active area. CCDs without these storage areas are suitable for still frame cameras but not for video. With interline transfer CCDs, the blind areas are in between each active column - this limits the resolution achievable in that axis. With frame transfer CCDs, the blind area is an adjacent area next to and the same size as the active area - this doubles the size of the chip. To implement snapshot operation and achieve low light sensitivity, the CMOS would have to follow similar architectures to video CCDs, which means additional circuitry and restrictions.

So it isn't so much a problem of CMOS. Its more a problem that has gone away with CCDs as display technology has evolved.
--
Its RKM
 
Its RKM said:
There is no reason why CMOS sensors cannot be designed to have a
similar "snapshot" mode of operation as CCDs and, indeed, many do.
Click to expand...
Do you need a snapshot mode or do you just need a fast enough rolling shutter to minimize the effect?

Don't film movie cameras still use very fast mechanical rolling shutters?
 
I knew this days ago -- because Vincent Laforet's mind numbing sample had ZERO horizontal panning.
markowp7 said:
i tried te 5D mark II today in canon's tokyo shinjuku demo area. nice
camera, good feel (it was new for me, i have never handled the 5D
before). the nice lady let me try out the video function and i
immediately tried out whether it produces the jello effect. it turned
out that is was easy to get, just by (not even very fast) panning. i
made a column in the shop wobble by just panning left and right
around it. no miracles there, CMOS is CMOS after all.
unfortunatly i could not take shots on my CF card, the CF door was
taped....
i am sure that it will be a great camera.
peter
Click to expand...
 
spencermcchester said:
I knew this days ago -- because Vincent Laforet's mind numbing sample
had ZERO horizontal panning.
Click to expand...
What about that scene where the car is zooming through a tunnel. That is in effect a pan, for the sake of this discussion. It's a horizontal pan in one direction on one side, in the opposite direction on the opposite side, and a vertical pan at the top. It was also very fast. Any jello should be very apparent there, at least if the clip is paused (haven't tried that yet).
--
http://www.pbase.com/victorengel/

 
Maybe, but we don't know how the LCD is being updated. Still would want to see the actual video to know if the effect is really a problem.
--
Chris Rake, Austin Texas
Personal Gallery and Central Texas Photo Guide
http://crakephoto.com
 
Not that I know much about video but I have seen enough professional movie making to know that there is a lot more to it than just the camera.

It's a major undertaking and a lot of additional equipment needed to do it right.

From what I could see from the samples of video so far, the camera was fixed and the focus was manual and fixed. Under those conditions, there is a lot of beautifully high quality video that can be made with the 5Dm2.

I just want to take pictures and maybe play with the video.

Duane
 
phauksson said:
Canon 5DII does have jello vision like the D90, sorry to burst your
bubble guys.

Watch this http://vimeo.com/1815853

Lafortes footage was stable because he compensated for it with gyros
and rigs.
Click to expand...
And how many movies have you seen where the camera operator swings the camera around in such a violent fashion? Maybe it will prevent any more Blair Witch derivatives, but for those who know what they're doing (and also use gyros - pretty standard stuff in flim making) I doubt if it will be the deal breaker.
 
Several people have mentioned the use of gyros to stabalize video footage. Is this really what you mean? A big (and heavy) metal disk rotated at high speed to produce directional stability?

Do you maybe mean a steady cam?

Anybody have a link to a camera gyro that's not a specialized 50k heli mount?

Stabalization is common in films and TV, but it's mostly by steadycam (stabalization by increased polar moment) not gyro (stabalization by gyroscopic precession and servo compensation).
 
delftkool said:
Several people have mentioned the use of gyros to stabalize video
footage. Is this really what you mean? A big (and heavy) metal disk
rotated at high speed to produce directional stability?

Do you maybe mean a steady cam?

Anybody have a link to a camera gyro that's not a specialized 50k
heli mount?
Stabalization is common in films and TV, but it's mostly by steadycam
(stabalization by increased polar moment) not gyro (stabalization by
gyroscopic precession and servo compensation).
Click to expand...
There is a gyro device attached to the car in the clip. This isn't a steadicam. You can pick up the 'poor mans's' steadicam for less than 50k ($ or £?).

Google for Steadicam Merlin.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top