Hmm..I may be way off on this but I was thinking more of a strobe light effect. There is another word that describes this effect but it escapes me and I guess its similar to the strobe light effect. As for the curving I think its due to the different angular(?) speed of the blade. If you do an analysis on a prop blade, the point furthest fr the center moves fastest, that makes it look curved and possibly the part that looks like its detached is because your camera shutter is moving at a speed similar to the other 'points'(the parts you captured) but not the points thats 'invisible'. It doesn't pick it up hence making it look like its not there. Hope this helps!..or perhaps it confused you more, in that case, sorry!! :)

The light reflecting off of the moving propeller is deflected and passes through the glass at an oblique angle. I'm guessing that you would be able to see it without electronic capture devices (if you could process the image fast enough).

I think I know. It must be that the phone takes pictures by scanning from top to bottom or visa versa. It must be a very fast scan, so that by the time the scan line has has moved up, the propeller has moved up and gets rescanned.
I can't explain exactly why it makes that pattern.
A long time ago I used to use an old computer camera that had a very slow scan and was able to get some wierd effects.

The aircraft has a stobe light either on top or underneath th fuselage. It gives the impression of stop action to a relatively poor camera.

Thanks, nex!

Also, of course! The flickering of the fluorescent tube makes some nice sense to me. I can't recall if I was still on the old CRT for that picture. I'm on a flat panel these days, but the upgrade was somewhere about the time I took the picture. Thanks for the help.

Gary, you got that animation to fit in pretty much perfectly, I'm very impressed!

As to your question about the multiple monitor images: No, the camera certainly doesn't scan multiple times. The picture would look very different in that case! It depends on which kind of monitor you have. If it's an LCD with a backlight, then that backlight is a flourescent tube, which naturally flickers, albeit _very_ fast in this case. CRTs are only ever lit at a single pixel, and while the electron ray is moved back up to the top, the screen really is unlit for a while, and this interval is indeed called the blanking interval. However, the afterglow of a TV's phosphor layer is quite long, so the tube doesn't really go completely dark. It can, however, be so dark that it looks completely blank with some exposure settings ... it's just unlikely that you'll find the entire screen completely dark, as the blanking interval isn't that long compared to the afterglow duration. (I'm guessing a bit here.)

from the fencecheck.com article linked above, there are some other
cool distortion examples:

http://www.rit.edu/~andpph/text-slit-scan.html


Warren

Edited by author 02-23-2007 11:05 AM
It's the scan line effect as Gary Fixler demonstrated. I've used this effect in some art by moving hard-copy portrait photos of people around on a flatbed scanner.

Here's a link to a simple study in lateral transfer using a Bell phonecard.

http://img528.imageshack.us/img528/6763/bellcallingcardnz7.jpg

Edited by author 02-23-2007 08:33 AM
How about a nice visual? I just mocked this up in Maya. I didn't get everything exactly matched up, but that's to be expected with the ambiguous angle, and position of the prop housing, the probable differences between the original camera, and my camera's focal lengths, the contours, and lengths of the blades, and even the number of them. This was the best fit I could get fudging some things around in about 15 minutes, and I ended up needing 5 props to get both the camera scan-line, and the prop rotation to work for ideal linear movements. This doesn't necessarily mean the plane's prop had 5 blades. It's just what I could get to work for me.

That said, the yellow line represents the camera scanning in the image row-by-row, just like an old TV scanline, at what would in real life be a very fast rate, but up against the speed of the prop spinning is relatively slow. You can see that each row that's recorded takes a single-pixel tall slice of the current state of the blades as they spin. Note in particular that the yellow tip of the propeller blades actually makes its appearance on both ends of the detached blade. That's because your seeing the same thing get scanned 2x, as the blade dips into the retreating scan-line, and pulls back out again. This should be more clear in my animation, each version being about 1.4MB

http://www.garyfixler.com/prop/prop.avi
http://www.garyfixler.com/prop/prop.mpg

Also, feel free to poke around with the 200 source images:

http://www.garyfixler.com/prop/

Another interesting version of this effect can be seen with camera tosses. In the following, I set my camera to take a picture after a short delay, and heaved it spinning into the air in my room. You can see my TV and computer monitor, and if you notice, they're heavily bowed in each instance. This is because both displays create their images with scan lines, so while the camera's moving, the rows are being drawn, meaning the camera keeps pointing in a slightly different direction for each row, creating warped images.

http://www.flickr.com/photos/garyfixler/42565941/

As to why there are many separate images of each screen, it's either that the scan lines take a break between each refresh, meaning the screens are actually dark for a moment before the next refresh, or the camera takes its long exposure by scanning multiple times, pausing in between each. I lean toward the former, as the red LED from my mouse creates an unbroken trail.

RaGe, you are correct as far as the imager technology is concerned. Pretty much all tiny, extra-cheap digital cameras measure the image line by line, which results in a distinctive distorsion when your subject is in motion. This is what makes the blades curved (AFAIK the actual propeller blades of that plane are really straight). It is the exact same effect that causes one blade to seem detached.

Here's how you could make yourself understand what's going on: Stick a piece of graph paper on a wall, put a fan in front of the wall, and some sort of light behind the fan. Use a small light, or put the light source in the other corner of the room, so you have a sharp shadow of the fan on the graph paper. Now trace the outline of the shadow, but only one line at a time. Manually rotate the fan (don't plug it in) by, say, 5 degrees between lines. If you experiment a little, you'll see that contiguous areas can easily become detached.

So from all the posts I've read here and from all the external pictures linked, I gather that :

* The sensor was more likely a CMOS than a CCD
* CMOS uses and Electronic Rolling Shutter, which produces effects similar to the old vertical shutters

This gives me a fairly satisfactory explanation for why the blades are curved. This curved effect is more pronounced in the pictures of Helicopter blades that some people linked to.

However I am still in the dark about why one of the blade pieces seems detached.. Help!

"has anyone ever created a 1-bit sensor?"

Combining multiple frames gives you more bits per pixel (higher quality, less noise), not more pixels.

(1-bit sensors aren't that uncommon, though. But you'll need to move either the sensor or put a moving or rotating mask or mirror in front of it to get something useful. Satellite instruments often use moving sensors or mirrors, missile navigation systems sometimes use a rotating mask in front of a single sensor.)

The strobe effect has been explained here sufficiently. Why do they propeller blades appear to be bent? Because they are, due to a differential in air pressure from the front to the rear of the blades. In the pre-jet years, airlines would never mention this because they didn't want to scare people
into thinking the blades might break off.

Edited by author 02-23-2007 12:23 AM
I have several pics showing this same effect:
- http://www.flickr.com/photos/edans/47661377/
- http://www.flickr.com/photos/edans/73256529/
- http://www.flickr.com/photos/edans/107268768/

One of my commentators, Herme Garcia, provided an explanation here (in Spanish):
http://itsp.typepad.com/voip/2005/12/ingeniero_de_gu.html

Translating from his text, the effect is a result of the Nyquist Theorem, there's a mathematical explanation here:
http://cnx.org/content/m10791/latest/

The theorem says that in order to recompose an analog signal from its samples, these samples have to be taken with a minimum sampling rate exceeding the double of its maximum frequency component (Nyquist frequency). If the sampling is done at a lower frequency, as is the case in the CCD of many digital cameras, you get "aliasing", the same effect that produces the wheels of a car to look like they are spinning backwards or the bands in a CRT monitor when you see it through a camera.

Hope it helps...

Enrique Dans
http://www.enriquedans.com

the end of the prop is moving faster than the parts closer to the rotor linkage. The ccd does not record all lines of information simultaneously therefore some information at the top of the frame actually occurs before the information recorded at the bottom. Solved.

Does this effect only happen with single-blade "clapboard" type shutters, or with telescoping multi-blade shutters, or both?