I was shooting some exposure tests with a RED and I turned my tungsten reference light off to look at something on the camera. I noticed green roll bars crawling up the monitor, and quickly figured out that it was the rolling shutter interacting with the overhead cool white Ace Hardware fluorescents in the shop. The problem was completely eliminated by going to 1/40th/sec. on the camera shutter at 23.98 fps.
This was build 14. I understand build 15 will introduce shutter angle control in degrees, so it’s probably good to know that the equivalent to 1/40th is 217 degrees.
1/48th at 24 fps is right on the edge of the 60hz flicker-free window, so I habitually shoot with other cameras at 1/40th or 217 degrees just to avoid flicker issues with house power or odd discharge lights in the background. It looks to me like this practice might be mandatory with this camera. Any additional motion blur is minimal.
Notice how green the image is. Most other cameras (three chip CCD’s) don’t see the green spike in the average uncorrected hardware store fluorescent fixture. This camera sees it almost as strongly as film would. Very interesting.
Cameras •
(Page 1 of 1 pages for this article )
Any rolling-shutter CMOS camera will show this artifact. Any global-shutter CMOS camera (Dalsa Origin) or a CCD camera will show a periodic variation of brightness and color over the entire frame, with the effect generally intensifying the shorter the shutter time.
Under 60Hz AC lighting, any shutter speed that’s a multiple of 1/120 will eliminate the problem, whether using a global or a rolling shutter (or even a film camera, for crying out loud!): 1/120, 1/60, 1/40, 1/30. Under 50Hz lighting, try 1/100, 1/50, 1/33, 1/25.
Some people feel that speeds around 1/32 smooth out overly staccato motion in 24p video, making it more film-like in motion rendering; try 1/30 as well as 1/40, and see if you like it.
Posted by Adam Wilt on 03/16 at 11:02 PM
I like 1/40 just because it’s right in the middle of the flicker-free window, and it’s a good combination of motion blur and safety under discharge lights.
I guess the RED is the first CMOS camera I’ve gotten to play with in depth, as I’ve never noticed this before. I’ve played a bit with the Dalsa in a fluorescent-lit room and never noticed this particular oddity.
CCD cameras will pulse slowly under certain conditions but I’ve rarely seen them react noticeably to 60hz flicker.
Posted by Art Adams on 03/17 at 07:44 AM
For a rolling shutter camera, the flicker (banding) amplitude depends upon exposure time, frequency remains constant (more or less) and phase change from frame to frame depends upon frame rate.
Hence, there are two dependencies, one on exposure time and one on frame rate (any vertical blanking intervening). However, the good news is that selecting appropriate exposure times removes the dependency on frame rate.
In addition, if exposure time is not among those that causes “flicker free” response, then longer exposure time will have less flicker banding amplitude, though, at the expense of more motion blur.
“Flicker free” case:
===================
The formula for best exposure time for “flicker free” is n / (2 * f), i.e., for 60Hz countries, 1/120, 2/120, 3/120, ...; and for 50 Hz countries,
1/100, 2/100, 3/100, ...,
“Worst case” flicker:
=====================
Worst case flicker will be seen at every other instance of (2n + 1) / (4 * f), for 60 Hz countries, i.e, for 3/240, 7/240, ..., and for 50 Hz. countries, 3/200, 7/200, ...
Posted by on 03/18 at 01:47 PM
