Unknown light-based Phenomenon

[AFG]

Can you help me figure out how this light-based phenomenon works? Please explain with an intuitive answer using the simplest principles possible.



Note how the location and width of the dark horizontal bands is constant with respect to the edge of the video frame.

The LED light looks normal enough to the human eye, but this effect is viewable through a phone camera and the effect was reproduced on another person's phone camera as well. I can think of 3 things that might contribute to this effect.

1. LEDs are usually dimmed by cycling the LEDs on and off quickly so that the human eye sees something like the average light intensity being emitted. The cycling frequency seemed very high, or at least smooth- I didn't see any strobe effects through my eyes when I shook my head from side to side. If the cycle frequency was low I would have seen some bright afterimage trails "burned" into my eyes.

2. I imagine that your average phone camera acquires frames at a certain frequency. This and perhaps the physics/protocol of light sensing might have an effect.

3. The software post-processing of the video could contribute to this effect. This post-processing effect would have to be common across different phones for this to be the case though.

Thank for any input!

 

[Grinkle]

I wonder if the LED light is polarized and that particular dark band is somehow perpendicular-polarized to the LED light?

 

[jedishrfu]

Perhaps its the brightness from outside that is affecting how the camera.

It could be related to LED flickering as you mentioned:





Folks recommend changing the shutter speed to match the flickering.

 

[jedishrfu]

I wonder if the LED light is polarized and that particular dark band is somehow perpendicular-polarized to the LED light?

That's interesting you could use polarized sunglasses to check this theory.

 

[Grinkle]

That's interesting you could use polarized sunglasses to check this theory.

@AFG To test, take a pair of polarized sunglasses and rotate them while looking through the lens at the LED light. If you see the LED light go dark at some rotation of the sunglasses, the LED light is polarized.

 

[jedishrfu]

@AFG To test, take a pair of polarized sunglasses and rotate them while looking through the lens at the LED light. If you see the LED light go dark at some rotation of the sunglasses, the LED light is polarized.

You see this effect on computer flat screens. Some screens are polarized diagonally (45 degree sunglass tilt) and others vertically ie they go dark when you tilt the sunglasses by 90 degrees.

 

[mfb]

The dark band seems to be present to the left of the lights as well, although less prominent.

If it comes from cycling the LEDs (+rolling shutter effect), it would be surprising that the frequency matches the recording frequency.

 

[davenn]

If it comes from cycling the LEDs (+rolling shutter effect), it would be surprising that the frequency matches the recording frequency.

did you mean "wouldn't be surprising"

It was also my initial thought when I watched the video before reading any responses
That there is an effect occurring because of the interaction of the LED power/light osc. and the shutter activationDave

 

[mfb]

did you mean "wouldn't be surprising"

No.
Why should the frequencies match?

 

[davenn]

Why should the frequencies match?

they may not, but if they did or were close then there could easily be a beat signal created from the mixing
that is giving rise to the bandsD

 

[Grinkle]

they may not

The bands in the image seem fixed relative to the camera image sensor edges. If the frequencies are not matched, I think the bands would move relative to the edges of the image sensor.

 

[davenn]

The bands in the image seem fixed relative to the camera image sensor edges. If the frequencies are not matched, I think the bands would move relative to the edges of the image sensor.

indeed, which further supports my idea that the 2 freq's are matched

this phenomena can also be seen aiming a video camera at a TV screen ... don't know if the new LCD screens show it, but the old CRT screens did

 

[Grinkle]

indeed, which further supports my idea that the 2 freq's are matched

If the bands are caused by frequency matching, then the frequencies match so closely that the eye can detect no mis-match. This statement makes logcical sense to me.

If the bands do not move, then it is likely that the bands are caused by frequency matching. This statement does not make logical sense to me - not sure if this is what you meant.

As far as I know, the clock in the camera is free-running with respect to the clock in the LED controller. It is very hard to impossible to make free running clocks match each other exactly.

this phenomena can also be seen aiming a video camera at a TV screen

The bands move because the frequencies are not exact integer multiples of each other.

http://electronics.howstuffworks.com/mpeg/q336.mpg