DLP Rainbows and PJ resolution - a link?

[nelliekins]

Hi all.

I've been giving the matter of rainbows some thought recently (now demo'd a number of DLP PJs with a variety of colour wheels and resolutions), and I have a question for you wise people.

Is there any connection between the native resolution, and the effect of the rainbows seen? It seems to make sense to me that the physical size of the rainbow has to be related to not only the color wheel, but the precision of the light being reflected off of the DMD as well.

So, given

1) an 854x480 panel and a 1280x720 panel
2) the same type/speed of colour wheel
3) the same size of output picture

then I would expect significantly lower rainbow effect on the higher res PJ. This is because the total illuminated area for each pixel (given the same output picture size) is proportionally smaller, which should in turn reduce the area affected by rainbows.

This seems to be backed up by a number of the threads reporting upgrades from X1/4800 to XGA panel PJs with 2x colour wheels (including Optoma 739, BenQ 6200, etc).

Having said all that, I have not yet bought my first PJ, and I know that most of you reading this have a LOT more experience & knowledge about the matter than me.

I would welcome your comments (even if they contradict me completely!).

Neil.

[KraGorn]

I don't see any connection .. a higher resolution will simply mean that more pixels are used to render any particular image component, the shiny edge of the sword in front of the dark background which is likely to cause the rainbow effect in someone susceptible to them. However, if there are enough pixels to render something then the image they produce will have the potential to cause a rainbow, the number of pixels making up that part of the image I don't see being relevant.

Lumen output per unit area will be similar for similarly bright projectors, lower resolutions will have brighter pixels.

The only issue I think where size would play a part is screen size .. a smaller image will have smaller highlights to cause rainbows and so they'll arguably be less noticeable, but for a given screen size I can't see that pixel packing density will affect things.

[Gary Lightfoot]

It's colour wheel related - you'll find modern 800 x 600 multi-speed pjs have less rainbows than the older 800 x 600 pjs that were single speed.

Gary.

[LV426]

I can't see how there'd be any connection. Regardless of how many mirrors (pixels) they are all (in effect) sequentially red, then green then blue (or whatever the sequence is). The effect is cause by the entire projected image being (say) red, then blue, then green; not by the behaviour of or quantity of mirrors.

Or, to put it another way......

Suppose you just shone the lamp through the colour wheel at a screen. No DMD device, just a lamp and colour wheel.

Suppose you took a large black cloth with a straight vertical edge, and quickly revealed the screen by drawing the edge across it. And followed the edge with your eye. You'd see the rainbow.

[nelliekins]

Hi KraGorn,

So, DLP works by redrawing the entire image in red, then green, then blue - correct?

And the reason that we only see rainbows along the edges is presumably because objects producing rainbows are > 1 pixel "thick", and therefore the colour for the other pixels are effectively overlapping where the colour deficiency occurs in interim pixels due to the rainbow effect?

I think my understanding of DLP is fundamentally flawed... Let me draw a (really poor!) picture of what I _think_ is happening, using your sword edge as an example.

Assume the "edge" consists of 10 white pixels (measuring in the direction of the motion).

Source Line 1 of the sword edge is white. Therefore Output Line 1 of your sword edge _should_ be white, but the motion (which is the cause of the rainbow) means it produces the rainbow effect, meaning that what is output is:

Output Line 1: RRRRRR
Output Line 2: GGGGGG
Output Line 3: BBBBBB

However, this isnt _seen_, because motion is causing the same thing to happen to _source_ line 2, meaning that it effectively fills in the Red component on the _output_ line 2, giving R+G for output line 2, and G+B for output line 3. And of course, the 3rd _source_ line does exactly the same thing, i.e. filling in from _output_ line 3. Which gives

Output Line 1: RRRRRR
Output Line 2: YYYYYY (R+G=Y)
Output Line 3: WWWWW (R+G+B=W)

and of course you get the same sort of thing along the trailing edge as well, except it would be:

Output Line 10: WWWWW
Output Line 11: CCCCCC (G+B=C)
Output Line 12: BBBBBB

Is this right?

If so, then the size of the rainbow _is_ directly proportional to the pixel density.

In which case, the way I must be wrong is in assuming that the light spill between R, G, and B, is related to the pixels at all, when it is purely a function of colour wheel, contrast and lumens...

...Let me guess - I just answered my own question!

Of course if it _is_ just CW, contrast and lumens, how come nobody has devised the _formula_ for rainbow intensity? One for the moderators, methinks!

Edit:

Hi Nigel,

I think I follow what you are saying - thanks!

Gary,

Surely more than just colour wheel is involved - otherwise how come everyone and their Granny keep asking "how good/bad are the rainbows" every time someone says they have a new model PJ? If it was _just_ the colour wheel, there'd be no point!

Seasonal wishes to all,

Neil.

[nelliekins]

Hi all,

This is probably another _really_ silly question...

... But if a framebuffer stores the image information for the duration of the frame being drawn for RG+B colour components, how come you get rainbows at all? Are you saying the colour wheel is misaligned? Certainly the wheel spins fast enough to draw every frame at least twice per frame of input video signal (NTSC = 60Hz, PAL=50Hz, 2x colour wheel = 120Hz? 240Hz? Never can remember!) Given that, why is there even a need to see rainbows? If the framebuffer is used to "set" the DMD, its the same picture information for each colour...and I don't see how it can go wrong...

Regards,

Neil.

[Nick_UK]

You are forgetting one fundamental fact - one rev of the colour wheel does not equal 50 or 60 Hz, because the colour wheel is showing 3 or more colours per revolution. Even if the colour wheel was spinning at 2 x 50Hz, you would still see the effect, because the refresh rate would be 100/3 or 33 Hz. The problem is not with the technology, the problem is down to persistance of vision, and this is why some people are affected more than others, because persistance of vision varies from person to person.

[LV426]

Let's try this again.......

The entire image projected onto the screen is completely red (for a fraction of a second). Then it's entirely green, then entirely blue, each for a fraction of a second.

The projector never projects a full colour image, or more than one of the three primary colours at any one moment in time.

The fraction of a second is refresh rate x actual wheel speed x number of segments.

You percieve a full colour image because your brain merges the sequence of colours together.

You see rainbows when something happens which allows your brain to perceive what's really on screen. Such as - scanning your eye across the screen (eg when following the trailing edge of a moving object). How well you percieve it depends on your own visual acuity - which fully explains why some people have no concept of how nausaeating DLP rainbows can be.

It's nothing to do with the projector adjustment - it's a fundamental artefact of the sequential display of colours. There's absolutely nothing the can be done to a single-chip device to prevent it - it can only be sped up to a point where few people will ever notice it.

[Gary Lightfoot]

Hi Neil,

It is more or less just colour wheel - some people see rainbows on single speed colour wheel models (3 segements - R, G and B), but not on models with double segment (RGBRGB) and faster rotation (4 x RGBRGB). The Davis DLS8 was single speed 800 x 600 and the new Optoma H30 has a faster colour wheel - (4 x RGBRGB) and less or no rainbows for many of those that could see them before.

Not everybody sees rainbows, and those that do will ask about them - if they're after a new pj and want to buy a DLP but are concerned about rainbows, they simply ask people if rainbows are visible. I never see them so cannot give them an answer, but you will find those that used to see them on older models no longer see them on some newer ones.

If it was resolution based, there would be no rainbows on 1280 x 720 pjs, but some still see them.

3 chip DLP no longer have rainbows because they don't have a colour wheel. Like Nigel said, you don't need anything other than a piece of card and a rotating wheel with a lamp behind it to see the rainbows. The same trick won't work with a 3 chip DLP though.

edit - a framebuffer can't be used to store an image because the way the image is made still relies on the rotating wheel. You can't store mirror activity and RGB images unless you had a phosphore screen maybe.

A tv makes up an image in a similar fashion, but because the tv has phosphor on the inside of the screen, the image remains on the screen for a short while while the image is being built up, so instead of seeing a line scanning across the tube, you see just the image. That, combined with your eyes persistance of vision is why you see the complete image and not an array of fast moving lines.

Gary.

[nelliekins]

Hi everyone,

Thanks for all your responses - it's really boosted my understanding of DLP. Makes you wonder why the manufacturers of PJs dont _all_ just use 4/6x color-wheels, doesn't it! I suppose its down to meeting a price point on the DMD (which has to change mirror state fast enough to keep up with the color wheel), and associated components.

Originally Posted by Gary Lightfoot

Hi Neil,

edit - a framebuffer can't be used to store an image because the way the image is made still relies on the rotating wheel. You can't store mirror activity and RGB images unless you had a phosphore screen maybe.

Gary.

Sounds like you've just come up with the ultimate solution for the several million DLP owners with 1/2x colour wheels! Something that acts like phosphor that can be used to coat a screen. I've heard about rear-silvered mirror materials, with partial opacity on front surface, but never anything like using fluorescent coatings!!!

Neil.

[Gary Lightfoot]

This might help you visualise how it actualy works on a physical level:

http://www.digitalhomemag.com/featur...bsectionid=935

The pic only shows a single RGB wheel, but imagine two segments of each colour and that will represent the later models. Some have 7 or 8 segments which can reduce 'dithering' (mirror activity) in darker areas, but that's another story.

The one drawback I think multi-segment colour wheels may have is increased possibility of colour banding - the less time the mirrors have to operate for each segment, the reduced range we may have in colour graduation. Some pjs with 8 segments seem more prone to banding than those with just 6 segments.

Gary.

[Tempest]

So you need to get 3 DLP chips and get rid of all the wheel mechanism.

Makes you wonder dosen't it.

Chips (etc) which DLP are generally not that expensive (In Real Terms) against physical moving/spinning things.

3 colour filters that don't move, a few mirrors and 3 dlp chips.

Why don't they stop ******ing about, get rid of the dam stupid wheels and just make them all this way?

[LV426]

Hear hear. Well, actually SIM2 have. But they want GBP25k or so for it. I don't think so.....

[Tempest]

Having read some thoughts on other forums on the very issue, it appears to be a common thought that the Single Chip with spinning wheel at a low consumer price and the triple chip at a astounding PRO price is mainly down to marketing and nothing to do with actual real manufacturing costs.

They could (and still could) produce low cost 3 chip dlp projectors but don't want to as they want to have 2 seperate products for two different markets.

If you make cheap ones the same you would destroy your high models.

Just think how much a DLP chip does cost?

$10 perhaps ?

Who knows.

So you could buy three whole single chip DLP machines, and be no where near the cost of one triple chip machine.

I guess it's the trouble when things are controller too much by some people at the top.

Ahhh, brings back memories of the old price fixing on RAM chips before it was blown apart. They held us all back for years on that one.

Ok, i have a question, if you had 3projectors, all the same make, same spec etc. and each one displayed a different colour projector 1- RED
projector 2- GREEN
projector 3- BLUE.
would you still get rainbows, obviously you would also have to get them all lined up correctly but still would this eliminate the rainbow effect?