AVS Forum banner
Status
Not open for further replies.
1 - 7 of 7 Posts

·
Registered
Joined
·
56 Posts
Discussion Starter · #1 ·
There are some new projectors coming out this year with 4X color wheels (e.g., Plus Piano). The reason for the increased color wheel speed is to eliminate rainbow artifacts. However, I do not think increasing the color wheel by 4X will completely eliminate rainbows, though it will reduce them significantly.


Consider how rainbows are observed. For a static white object on a black background, when the eye darts across the object fast enough (say with angular velocity A), the individual color wheel colors (RGBW) are seen by the observer as a streak of colors. By increasing the color wheel by 4X, the eye will have to move with angular velocity 4*A to see the same rainbow length as seen with a 1X color wheel. Alternatively, when the eye moves with angular velocity A, a rainbow streak of length L can be observed; for a 4X colorwheel, a rainbow streak of length L/4 will be observed.


I wrote a simple application to measure the rainbows observed on my LT150. Ideally I'd like to measure the length of rainbows as a function of the eye's angular velocity. Since measuring this requires expensive equipment I didn't have handy, I used a less direct (and accurate) method. By translating an object across the screen and visually tracking it, the eye will make a series of jumps (saccades). By changing the velocity of the object, the angular velocity of the eye will also change.


On a 1024x768 black screen, I programmatically translated a 10x10 white square across the screen horizontally. By visually tracking the square with a speed of 500 pixels/sec, I could see rainbow colors on the trailing edge of the square. When I increased the square to 1000 pixels/sec, the rainbow colors were significantly larger (about the width of the square); at 2000 pixels/sec, the rainbow colors were about the width of two squares.


I predict I will still be able to see rainbows on a 4X color wheel projector, though the rainbows will be 1/4 the length as measured on a 1X color wheel projector. To eliminate rainbows, the color wheel needs to spin fast enough to make the rainbows streaks subpixel in length. I believe this is possible, since there is a physiological maximum angular velocity the eye can move.


I hope I'm wrong in my prediction, but note that at least one observer of the Plus Piano at Infocomm still saw rainbows on this 4X color wheel projector.


Ross.
 

·
Registered
Joined
·
2,027 Posts
I hope it would eliminate them. I find them pretty annoying on the LT150. If it only made them 1/4th as "thick" as they are now, I'm not sure that's a final solution.
 

·
Registered
Joined
·
5,132 Posts
Anyone know the MTBF (Mean Time Between Failure) of color wheel? Since it's spinning constantly, it'll eventually fail (all fans eventually fail or get very noisy). Is this a concern for DLP color wheel? I know hard drive spins quickly and constantly too and MTBF is hundred of thousands of hours. Can we assume the same for color wheel? Just curious.


------------------

Huey ;-]
 

·
Registered
Joined
·
578 Posts

·
Registered
Joined
·
56 Posts
Discussion Starter · #7 ·
I've done some computations on how fast a single chip DLP color wheel has to rotate in order for the user not to see any rainbow artifacts. My configuration has a 6' high screen and I sit about 12' away from the screen. The maximum angular velocity for eye saccades (and glances) is about 830 deg/s, so the max eye motion is 22250 pixels/s. With a 3 segment 180 Hz color wheel, the max rainbow streak observable is about 124 pixels. This agrees with what I've seen if my eye really moves fast while viewing a white object on a black background, though I typically see much smaller rainbow streaks.


Assuming a rainbow artifact is not noticeable if it is less than or equal to 1 pixel, then a 3 segment color wheel needs to rotate at 22250 Hz to eliminate rainbows. If a 300 segment color wheel can be made, then only a 222.5 Hz rotation is needed.


It is possible that rainbows are not noticeable even if the streaks are larger than 1 pixel, since motion blur could hide them. I'm sure TI has done the psychophysical studies to determine the minimum color wheel speed needed to eliminate rainbows. Perhaps someone from TI can give us these results (if they are public).


Ross.
 
1 - 7 of 7 Posts
Status
Not open for further replies.
Top