[mlivesay]

Hi Guys,


I was just reading someone's comments about the rainbow effect on DLP projectors and had an idea of what might be causing this and why some people see it and others don't. Please note that I might be totally out to lunch on this, but here goes:


Apparently the rainbow effect appears as multi-colored artifacts on the edges of moving objects. The fact that it seems to appear only on moving objects is an important clue.


My understanding is that a single panel DLP projector constructs a given video frame by flashing each of the primary color components on the DLP panel in sequence as each of the red, green, and blue segments of the color wheel pass in front of it. The refreshing of the DLP panels must be in perfect sync with the color wheel segments in order for the whole thing to work properly.


Assuming a video frame rate of 30fps, the color wheel should be spinning at 90 revolutions per second so it can pass three color segments in front of the panel in 1/30th of second for a single frame. Also, the DLP panel will need to be refreshed at the same rate (90 images/sec).


So assuming the above is correct, then here's what might be causing the rainbow effect:


1) The DLP projectors on the market are spinning the color wheel and refreshing the DLP panels at speeds less than 90rps or at rates that are not a multiple of the video frame rate. If the projector is refreshing the panels less then 3x the video frame rate, or in a non synchronous manner, you could have a situation when the video frame changes in the middle of an R,G,B color wheel sequence (ie, you get say the red and green images from frame N, followed by the blue image from frame N+1). On a stationary image, this wouldn't be that noticable, but on a moving object you would see color artifacting along the trailing edges of moving objects.


2) The persistence of the DLP panel might be longer than 1/90th of a second. In other words, the DLP panel might not be able to fully reset and refresh at 90fps. Since the DLP panel is a solid state device, I doubt this would be the problem, but if it was, you might see a strange blending of colors, especially on the edges of moving objects. Actually, the max refresh rate of the DLP panel would dictate the speed of the color wheel and might be the underlying reason for #1 above.


As for the reason some people see the rainbow effect and other don't: the sensitivity of the HVS (human visual system) to the above is kind of a perceptual thing. For example, I imagine some people might not be able to tell the difference between video at 20fps and 30fps. People's sensitivity to things like color and contrast also vary by some margin. Thus the "perceptual persistence" of some people's eye might be high enough to mask the rainbow effect completely.


If my assumptions are correct, I would expect that the rainbow effect would be most noticable on B&W video, especially in high contrast seqeuences. Also, DLP projectors with no color wheel (and 3 DLP panels) would not exhibit the rainbow effect.


I don't own a DLP projector and have actually never seen the images they produce. I do have some background in image and video compression, and have messed around with this kind of stuff quite a bit. Perceptual coding seeks to exploit the imperfections of the HVS and I've learned that the eye can be fooled quite a bit. Its possible the engineers that have worked on the DLP systems out there built their systems on the assumption that most people's HVS will mask the rainbow effect, or won't find it that objectionable.


Anyway, just some food for thought. If there are any engineers out there that have worked on the development of DLP projectors, I'd love to hear your thoughts on this. Its funny that the industry has been pretty quiet on the rainbow effect and its causes - they don't really seem to acknowledge its existence.


Anyone else out there have any thoughts on this?


Thanks!

Mike

 

[ckolchak]

Mike,

interesting theories. as one of the people who noticed this phenomenon w/in his first 5 minutes of DLP exposure, this is a subject i'm eagerly waiting to hear others input on.

maybe it could be tied in to minor flucuations in voltage?

or perhaps the longer a DLP is used the less well synched the refresh, etc.

maybe, someone will come across a photographic filter that somehow, magically cuts down the effect.

who knows.

hopefully,

the more hands this unit (we all know which one i'm talking about) gets into, the greater the chances creative tweaks will be found to squeeze every last drop of performance out of it. just look over to the VW10HT- gamma settings to filter tweaks to grey screens, there must be some way to deal with the rainbows. perhaps a HTPC with a refresh rate seasoned to taste is only true way to go.

but it will be sooo much better if 1,000 people are pooling their brains and playing with them, than just a handful of true-believers.

 

[JoeFloyd]

A few things to keep in mind.


The DLP color wheel has four segments. R/G/B and a clear section. In one rotation all color segments will be used.


The rate of rotation of the color wheel is synchronized to the vertical refresh rate of the input signal. When displaying 480i it's likely that the project will set the color where to 60 RPM.


A new color wheel design with six color segments and no clear segment has been designed and is being used in a few hometheater projectors. The effect of 6 color segments is to double the effective rate of the color wheel spin. A six color segment wheel spinning at 60 RPM is like a three color segment wheel spinning at 120 RPM.


The effect that produces the rainbow effect is tied to a phenomena called persistence of vision. As the color wheel rotates the DMD mirrors modulate the intensity of light using a technique involving switching the mirrors on a off many many times while the panel is illuminated by a given color. This is called pulse width modulation. Even such, the short period of time a single color illuminates the panel is very short and is somewhat like a strobe light. There is a lot more to this, but I hope you get a sense of how complex a DLP projector is based on this simple discussion. TI has quite a few white papers that go into great detail about the DMD technology. It would be worth a few minutes of your time to track them down.

 

[kelliot]

deja vu


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

Ken Elliott

 

[mlivesay]

Thanks for the info, JoeFloyd.


Yeah, I was aware that most DLP projector have four segments - just wanted to keep things simple for the sake of the discussion The Yamaha DPX-1 will apparently have three segments, and I've heard about both the six segment and spiral wheels (although I get headaches trying to figure out exactly how the spiral wheel technology works).


I was unaware that DLP projectors used PWM to generate the intensity gradients (I assume you mean PWM in the sense that they toggle at a fixed frequency, but change the on to off ratios to generate various intensities, right?). I didn't realize that DLP pixels where two-state - thought they could vary intensity in a somewhat analog fashion.


It would be interesting to see what kind frequency they are toggling at. Given that so many people see the effect, they must be right at the threshold of the average HVS persistence.


I'm somewhat skeptical that they can vary the rotation speed of the wheel with such accuracy as to sync it exactly with the frame rate of the incoming interlaced or progressive video. Are they using some high speed stepper motor, or some sort of feedback loop speed control system? Anyway, I'm a software guy so I'll take your word that such a thing is possible


I'd like to take a look at those papers you mentioned - any links?


Anyway, thanks again, JoeFloyd.


Mike


Ken - Not sure I understand your remark - should I be insulted?

 

[hanawa]

The Rainbow Effect does easily be noticeable on B&W video.

For vivid video source on HTPC, or DreamCast game on VGA mode, I can not see the Rainbow.

 

[ckolchak]

Quote:

Originally posted by KeithR: I think the rainbow topic is going to come up big time if the Dell LT150 comes off....there are going to be some pissed people even considering the low price. I recommend everyone who is getting one to go to there local store and see a Runco (these are usually easy to find, that's all) and see if it bothers them. Luckily, Dell also has a no questions asked 30 day return policy, so one can get out if it bothers them. Keith

yep. as much as this board is filled with enthusiasm/anticipation now, in a couple weeks the bile spewed forth over this issue may get pretty deep.

it was Dells return policy, along with the great price that convinced me jump into FP and to give this projector a shot after all.

i had pretty much written it off because of this factor and was looking to LCD. if i could have found a VW10HT for under $2k, i'd be even MORE excited.

the thing is-screendoors, blacklevels, etc. they are all constants, they don't come and go throughout the viewing. once you've watched for a while, it seems to me, you can more easily put them out of your mind. but if they are constantly springing up out of the blue, and then going away, it's harder to ignore them.

 

[David Mendicino]

As was said,"Deja Vu".


Do a search and you will find a wealth of information regarding the rainbow effect. The cause and explanations are known.


Those who by the LT150 wihtout checking for it may be unpleasantly surprised.




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

David Mendicino

Sharp xv-s55u (Don't laugh)

 

[Randall Morton]

Mike,

I can see rainbows on the edges of nonmoving objects just as easily. I can pause a DVD on a contrasty frame and quickly scan my eyes across the screen and see rainbows the same as if the objects were moving.

 

[Morbius]

A single chip DLP relies on the persistence of vision

effect.


At any instant in time, the projector is displaying a single

color image - or a white image, if the color wheel has a

clear section.


It is relying on the persistence of vision to mix the three

primary colors into the appropriate color.


As Randall points out - you can get rainbow effect not only

with a moving image - but a still image - if you move your

head.


As you move your head, the red, blue, and green images are

formed on slightly shifted areas of your retinas. Since

the 3 color images don't align on the retina of your eye -

the persistence of vision effect doesn't blend the images

properly, and you get rainbow effect at the edges of

objects.


You see the same effect on color images in newspapers if

the printers are off in aligning the plates that print the

primary colors.


Greg

 

[ckolchak]

Quote:

Originally posted by Morbius: ... You see the same effect on color images in newspapers if the printers are off in aligning the plates that print the primary colors. Greg

are you talking about registration? that analogy would translate better as a convergence issue, i would think. something that doesn't apply to these projectors...or does it?



what makes me curious is that one person (Alan, for instance) will recommend a gray or low-gain screen to cut down the effect, while another (Grant, for example) will swear by a Hi-Power which is a 2.8 gain screen?! the same thing seems to apply w/ refresh rates.

it just sounds to me like to get the best, most rainbow free image from this thing, you really have to 'season to taste'

solely conjecture on my part since i don't have one yet, but if this is the case, meaning that no rosetta stone for optimal settings exists for every single viewer, maybe there might be other, so far un-thought of ways of manipulating the optics to also cut down or eliminate this effect...for some people.


i just amazes me that one person came up with the idea of using a simple inexpensive colored photo filter to significantly improve the image from a $5500 projector, after many people had written it off 'as is'. a simple a concept as that and only one person thought of it.


of course i could be totally off base here...

 

[XrayEyes]

The rainbow effect is especially annoying because you cannot anticipate it and becomes very distracting. I found this effect particularly bad in many scenes of Gladiator, especially dark scenes. This caused me to return my Infocus unit and get an LCD projector - I am much happier.

 

[Daniel Hutnicki]

I recently purchased and returned the Dwin DLP. At the show room, I didnt see any of the rainbow artifacts. However, 2 days after getting it home, I saw them everywhere. On Crouching Tiger, I saw them as the subtitles would change, I saw them off the reflection of the swords, I saw them on the main male characters reflection of his head. Anywhere where there was a white or high contract colors, I saw them.


In castaway, when Tom Hanks is in the plane, the daylight coming in through the windows had the rainbows.


A easy way to find them for those who never see it (maybe you shouldnt) is to watch the first Farscape DVD. On the menu selection, they have little white lights streaking across the background, if you watch the white streaks, you will see them. The retailers who I bought the Dwin from, who had never been able to see the rainbows, were able to see them there.


The only movie I couldnt see any rainbows was the FiFth Element. So there must be something to how the movie is made

 

[KeithR]

I think the rainbow topic is going to come up big time if the Dell LT150 comes off....there are going to be some pissed people even considering the low price. I recommend everyone who is getting one to go to there local store and see a Runco (these are usually easy to find, that's all) and see if it bothers them. Luckily, Dell also has a no questions asked 30 day return policy, so one can get out if it bothers them.


Keith

 

[Morbius]



Originally posted by ckolchak:

are you talking about registration? that analogy would translate better as a convergence issue, i would think. something that doesn't apply to these projectors...or does it?


Yes - I guess it is called registration. But it is not

only a convergence issue.


The projector can be perfectly converged - the blue image

overlaying the red image overlaying the green image

perfectly.


However, look at where these images wind up on the retina

of your eye. Because those images are not all projected

at the same time - if you move your head - or shift your

gaze - the images won't align or "register" properly on

your retinas.


The convergence of the projector is only part of the story -

you need to have "convergence" of your eyes.


Now three chip projectors, LCDs, D-ILAs, and 3-chip DLPs

don't suffer this problem - they project the three color

images simultaneously. Since your eyes receive the three

colors at the same time - they mix them properly - without

relying on persistence of vision.


Only the 1-chip projectors rely on the persistence of vision

to "buffer" the image while the projector takes a finite

amount of time to put all information - all three colors -

on the screen.


Greg

 

[srauly]

This has been discussed in another thread, but I'll chime in with what seemed to be the popular idea and which has been referred to in this thread. However, because I don't own one of these things, I can't state this as fact.


Refresh rates and screen gain appears to have no effect. Eye movement appears to be the cause. Therefore, you want to minimize eye movement. How do you do that? By maximizing the screen width to distance ratio. Common sense: the more of the screen that you can see in your area of vision, the less your head/eyes are going to have to move around. For example, someone sitting with a 120" wide screen sitting 15' away (or 1.5x screen width) should see more rainbows than someone sitting 20' away from the same sized screen (2x screen width). If you want/need to sit closer, you would use a smaller screen.


Under this theory, if you want to have a huge screen and a relatively short viewing distance, you may be unhappy with the number of rainbows you see. Grant has stated that he doesn't see rainbows. He's using a 120" 4:3 screen, but I'm not sure how far away he sits from it.


Scott

 

[David Mendicino]

I don't know the situation in which someone increased the rainbow effect by increasing refresh rate...but I must say that refresh rate ABSOLUTELY affects the rainbow effect.


While eye movement and screen gain cause you to see the rainbow effect, what is really behind it is the strobing of the colours. The faster they strobe, the less rainbow effect. Increasing the refresh on a DLP projector speeds up the colour wheel, in order to maintain colour synch. In doing so, the rainbow effect should be reduced...as there is less delay in colour change, which the human eye percieves as the rainbow effect.


Why do you think the DLP companies are coming out with 6 segment colour wheels, or wheels that rotate at twice the speed (which has the same effect, re: the colour wheel, as does increasing the refresh rate).

It is to alleviate the rainbow effect. Unless the person has some form of DLP that doesn't synch the colour wheel to refresh rate, increasing it would decrease the perceived rainbow effect.


Technically, the new colour wheels are alsmot at 240hz rotation.


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

David Mendicino

Sharp xv-s55u (Don't laugh) http://www.avsforum.com/ubb/smile.gif


[This message has been edited by David Mendicino (edited 07-24-2001).]

 

[Morbius]

David,


Additionally, the faster the refresh rate - the shorter

the distance that you can reasonably move your head in

the time between color frames. Therefore, the less the

shift - the less the rainbow effect.


In the limit in which the colors were refreshed at gigahertz

frequency - how far can you move your head in a nanosecond?


Greg

 

[hanawa]

I do believe increase the refresh rate will reduce the rainbow effect(That's why I see no rainbows on DC VGA mode), but when I did a test with same video source with P-Scan(31.5KHz) and S-video(15.75KHz), both appeard the same amount of rainbows to me. That's really confused me(Does P-Scan and S-Video has same refresh rate?). I do want to test with 63KHz(QuadScan) to see the improvement, but I can not find one in Taiwan. Is there anyone who can share your experience for me? Thanks in advanced!!

 

[ScottS]

All,


It is possible that adjusting the refresh rate may affect the visiblity of rainbows. Some projectors have a variable speed color wheel that syncs to the refresh rate. This is the case for the Davis DL 450 and clones (e.g. Viewsonic PJL 830). My 830, though rarely producing rainbows to my eye, does not appear to be sensitive to refresh rate. Other projectors with a fixed rotation speed color wheel may be.


I don't know if there are any remaining archived topics on this issue, but this topic was discussed in depth last spring and summer.


Scott