New theories for keystone correction, merged topic Options

[charlie10]

Ok you guys are gonna think I'm crazy. I seriously geeked out on this one. I hope it's somewhat correct...

This one is ceiling mounted, sortof to scale (14" LCD, 10mm lamp arc, 220mm rear fresnel, 330mm front fresnel).

I drew it pretty wide. You might have to scroll to see the screen on the right. ---->

Attached File(s)

 ray_7_5.gif ( 96.21K ) Number of downloads: 6

 

[charlie10]

LONG STORY SHORT, it looks like Mikau's real-world result earlier in this thread...
"Perfect! Tilted screen to correct trapezoid, adjusted lens slightly to keep in focus"
... can be successfully modeled using RAYTRACE!

Details:

The animation is six frames, 10s each:
1. initial setup. this should show a "normal" projector. I traced three LCD pixels.
2. tilt the projector 10deg. (actually, I tilted the screen / room).
3. tilt the fresnel 10deg. to correct (definitely improves the image)
4. tilt the fresnel a little more, to 10.9deg (fully fixes trapezoid, BUT not focus)
5. tilt the fresnel way out to 21.3deg. (fully fixes focus, BUT not trapezoid)
6. tilt the fresnel back to 10.7deg. and tilt lens 3 of the triplet (PERFECT IMAGE).
7. I was gonna try tilting the whole triplet, but raytrace crashed on me. maybe tomorrow.

The triplet has a rear positive lens which I focused on the LCD. For the middle (negative) and front (positive) lenses, I just guessed. I played with their focal lengths until I got a reasonable focus, approx. 1m in front of the projector.

I pretended the LCD was black except for three pixels: one red, one green, one blue. I only showed two rays per pixel: a center-ray and a side-ray. Pretty eh? Now let's hope it's correct!...

The lamp rays are shown, cut off at the LCD. I used three lamp raygroups, to represent a 10mm bulb arc. Yes this means non-parallel light hitting the LCD, well I used RAYTRACE's "protractor" tool to figure out, exactly how much non-parallel? 1.302deg, apparently. That's the angle between the "perfect" ray (coming from the middle of the bulb arc) and one of the side rays (coming from the side edge of the bulb arc). Actually that number doesn't really matter for this keystoning discussion, but, it was fun to calculate :-). NOW I see why a perfect point-source bulb wouldn't even need a projection lens... the lens is there to intercept that imperfect 2.604 degree wide cone of light, and squeeze it down onto a single spot on the screen! (Or something like that, anyways :-)

[Squalish]

That's actually really, really helpful in describing how the triplet works.

[Squalish]

Question: Does one of the fresnels(half fresnels, whatever) used backwards = a diverging thin lense?

[charlie10]

Good question. My guess is, if it's converging one way it's also converging the other way

Here's something else from the BenQ website

"Each projector has an individual, optical axis that sets how high an image can be projected above the lens axis. A relationship of 19:1 means that the image will be divided up into 20 parts, 19 of which will be projected above and 1 below the lens central axis.
"With a relationship of 10:1 we talk about 'Con axis', i.e. the projection's lower edge is at the same height as the lens".

http://www.benq.us/serviceandsupport/libra...404,417,425#426

I don't understand it. but it sounds relevant ...

[eudaimonia]

The triplet has a rear positive lens which I focused on the LCD. For the middle (negative) and front (positive) lenses, I just guessed. I played with their focal lengths until I got a reasonable focus, approx. 1m in front of the projector.

Charlie10,

Nice work. I'm curious though, if you wouldn't mind playing with your raytrace some more to investigate what happens when the triplet lenses are shifted. By this I mean moved up or down, left or right, not tilted and not front to back (keeping all other alignments as if they are straight on projections). I would be curious to see what happens when the entire triplet is thus shifted and what happens when individual lenses within the triplet are shifted. The reason I'm asking is that it seems to me the lens shift mechanism that allows a projector image to hit the screen square, but above the center line, involves moving the triplet and/or individual lenses. I sent Mikau some info from patents involving lens shift and they seem to describe just moving the triplet or one of the triplet lenses up/down/left/right in the same plane as the (untilted) screen.

Thanks,
Chris

[charlie10]

So just shifting up/down and not tilting. Sounds good.

The results would be more valid if I had a better more of a clue about the triplet. Pretty sure the middle lens should be negative and the other two positive, but what should the relative focal lengths of the three lenses be?

All three roughly the same?

[eudaimonia]

Charlie10,

You know, I'm not sure what the focal lengths are. If I had to guess, I would say that two of the 3 are probably focused on the lens you call the "rear positive lens" or the other way around and the rear/middle lenses are focused on the front lens. But that's more of a wild guess than anything. Brain would probably know given he originally played with a single throw lens before moving to a triplet that corrects the barrelling of a single lens.

I'll see if I can find info on it elsewhere but if anyone know the answer please do post it!

Thanks,
Chris

[charlie10]

Ok I have come up with the following solution:

Attached File(s)

 lenspr17.gif ( 6.14K ) Number of downloads: 5

 

[charlie10]

Joke... you know I did not come up with that. I couldn't resist since the images looked kinda similar. Sorry :-) Anyways that is a cool looking raytrace of a "reverse telephoto" lens I found via google on the following web page:
http://www.cartage.org.lb/en/themes/Scienc.../Lens/Lens2.htm

If you look there is a "Cooke triplet" lens on that page too, which seems closer to what we're doing. So I'll try to get some info from that.

By the way, Google rocks, and so do these forums!

[charlie10]

I found a 50mm Cooke triplet design from another program called OSLO. So I scaled this by 7 to get a 350mm triplet, and entered that into RAYTRACE. Well I am not sure about this lens, either that or I got something totally wrong. Because the focal plane came out very distorted-looking, and there's lots of chromatic/color aberration.

Anyways I tried sliding the first and second lens elements up and down, together without tilting, to see what effect this would have. It does appear to have some kind of keystone type effect. However it doesn't seem as compelling. See pic.

(the black lines represent the focal "plane", basically where the screen would go).

edit: tilting the fresnel while sliding just the second lens, seems to be alot more effective. Sorry I don't have the energy to make another GIF of it .

Attached File(s)

 ray_9_4.gif ( 82.95K ) Number of downloads: 5

 

[charlie10]

I can't help going back to the initial idea in this thread. Basically sliding/shifting the triplet AND the fresnel up/down by the same amount. It just seems so much simpler than breaking apart the triplet.

For a 330mm triplet, a vertical shift of 80mm / 3" should achieve the equivalent of 15 degrees tilt+keystone-correction.

a collector that points light towards the bottom, like the picture below. The triplet has the same viewing angle, but higher magnification so it can be placed farther back and offset. The fresnel is either a special made one, or a large one cut in half.

from http://www.lumenlab.com/forums/index.php?s...indpost&p=57904
please refer to Mikau's upper diagram.

So... do we even need the higher magnification? If the existing triplet has just enough FOV for a 15" LCD, then I guess the extreme corners would almost immediately darken with this type of shifting. However even a slightly smaller LCD should have alot more wiggle room... as long as the shifted rectangle still fits into the 15" circle? Which should be easier the more widescreen the LCD is.

You could even put the whole thing (fresnel+triplet+focus box) on locking sliders, to allow adjustment.

Attached File(s)

 ray_10_1.gif ( 32.48K ) Number of downloads: 5

 

[Squalish]

Shouldn't be too difficult to test... anyone up for it?

[eudaimonia]

Nice work Charlie10, that does appear to be "it"! I've seen some 14 and even 12 inch LCDs with the same specs as the usual 15 inch models so hopefully someone has one of these and would be willing to do a little experimenting. Either that or someone with the pro lens and a 15" LCD might try it (although I imagine if they have the pro lens they probably have the larger LCD)....

[charlie10]

I tried this out with a single Staples fresnel (page magnifier) and a Surplusshed copy lens triplet. I put a black&white grid pattern on my laptop for an image source.

I can't really verify until I have the proper parts, but even the above crude setup seemed to do a good job of shifting the image up/down/left/right while keeping focus and avoiding keystoning. As long as I moved the lens and the triplet by the same amount. So I'm guessing it could work for smaller LCDs. I can't find many that are <15" though... just laptop LCDs, and nobody seems to use those.

Mikau, what did you use to break apart your triplet? I got the one from surplusshed but it is built pretty solid, I don't think a hacksaw would cut it. The lens retainer ring is threaded, but I don't know how to get enough grip to unscrew it :-/

[Mikau]

I tried this out with a single Staples fresnel (page magnifier) and a Surplusshed copy lens triplet.  I put a black&white grid pattern on my laptop for an image source.

I can't really verify until I have the proper parts, but even the above crude setup seemed to do a good job of shifting the image up/down/left/right while keeping focus and avoiding keystoning.  As long as I moved the lens and the triplet by the same amount.  So I'm guessing it could work for smaller LCDs.  I can't find many that are <15" though... just laptop LCDs, and nobody seems to use those.

Mikau, what did you use to break apart your triplet?  I got the one from surplusshed but it is built pretty solid, I don't think a hacksaw would cut it.  The lens retainer ring is threaded, but I don't know how to get enough grip to unscrew it  :-/

Hi guys! Sorry I haven't been here for much of this. My computer is in the shop. Its got some new uber spyware. Appearantly a lot of people have it. Not even the people at nortan antivirus have found a remedy yet :angry: blasted spyware!!!

Anyway, nice job charlie, you've done a lot of work. We just need a triplet with double the viewing angle and a matching triplet. Easier said then done..

Thats up to brain.

Anyway, I used a hack saw to take apart my triplet. But mine was plastic. At first I tried to melt them off with a hot knife and almost killed myself with the toxic gas it produced.

[shivers20]

 f04.jpg ( 37.38K ) Number of downloads: 70
Has anyone tried using a vesta pocket camera design for their focusing mechanism. This would allow the triplet to be moved around from top to bottom, side to side making. A pic

This post has been edited by shivers20: Apr 6 2005, 05:35 PM

[jesusjj]

hello light bulb, fresnel united (they do not incline), tft (if circuiteria inclines next to his) and the objective in its geometrico center. the projector inclines whole, tft arranges parallel to the screen to prove it and to count results.

[DAZZZLA]

hello light bulb, fresnel united (they do not incline), tft (if circuiteria inclines next to his) and the objective in its geometrico center. the projector inclines whole, tft arranges parallel to the screen to prove it and to count results.

?????

[DeathRay64]

I'm thinking english isn't his first language. I think that he's saying just tilt your panel.

[jesusjj]

I'm thinking english isn't his first language.  I think that he's saying just tilt your panel.

[rlwoodjr]

What would happen if you moved the light down and the triplet up? I have tried this with a framed transparency and an ohp fresnel and it apperars to work.
I will be experimenting some more. I am rebuilding my projector. I am trying to put both fresnel lenses in back of the LCD.

See attached picture...

This post has been edited by rlwoodjr: Apr 10 2005, 09:27 PM

Attached File(s)

 pj.bmp ( 106.37K ) Number of downloads: 65

 

[jesusjj]

it happens the same. it is the same system, it requires to move the light bulb and the objective. my single system is necessary to incline tft few degrees. I have not calculated it, about 7 or 10 degrees at the most. I see the advantage him my system no that there is to move the light bulb above and down. the axis of the single light varia. is not necessary to incline a little tft. your system also studies it like alternative, but a little convinces to me but to incline tft. I at the moment cannot test you practice. I recommend to you that you prove systems both. I am almost safe that they work both perfectly, but I have not been able it to prove definitively. my Web so that you see photos it is www.morcillon.tk it inside has carperas to the left with photos. a greeting, luck and place photos of the result.

My Webpage

[charlie10]

What would happen if you moved the light down and the triplet up? I have tried this with a framed transparency and an ohp fresnel and it apperars to work.
I will be experimenting some more. I am rebuilding my projector. I am trying to put both fresnel lenses in back of the LCD.

See attached picture...

I've gotten my LL fresnels and triplet, and set up a testbed of sorts. Using an Ikea "Flygel" 50W halogen lamp as a test light source.

Til now I hadn't tried moving the light source. So I tried it. It does seem to work, but maybe at the expense of brightness / uniformity. The beam of light cast after the first fresnel is no longer rectangular in cross section; it's a little squashed on one side, like a twisted piece of paper. The shape of the light hitting the triplet is more quarter-moon than circular. The LCD image was not distorted, but it seemed harder to position the triplet in a way that cast even light. My initial guess is that moving the lamp in the way you describe, could accomplish about 5-10 degrees of shift before it the picture started to degrade... just a guess.

Here's a pic of my most recent setup setup , showing fresnels parallel but offset. A laser level indicates the primary optical axis of the setup, and that laser light gets angled by the second fresnel, hitting the triplet. I got the best results orienting the triplet perpendicular to this light, as shown. The screen is still parallel to the second fresnel.

From what I can tell, the standard LL triplet is able to handle the light path well, even at very significant offset. What I think I need is bigger fresnels.

Attached File(s)

 offset_fresnels.jpg ( 82.08K ) Number of downloads: 5

 

[Mikau]

BO YA! GUESS WHO'S BACK!

I'm pleased to see that this thread stayed alive during my absense, and I think I owe it most to charlie. Thanks, dude!

Well charlie, brain said that the standard triplet is capable of getting about a seventeen inch lcd into focus when the edges will begin to dim. 15 inches is a safe distance from that, but you could probably get away with at least an inch. But thats not much really. What you could do is use a standard 15 inch collinator, and the profield collector, with the pro zoom triplet. The small collector would be better for a 15 inch so no light is wasted, and directly infront you have the pro collector offset as much as possible. The pro collector will cover a 21 inch lcd so you should be able to shift it about 25% which isn't much really, but not too bad when you sit close to the projector and are using a larger projection. (The bigger the projection, the bigger a 25% shift is.)