Also, keep in mind commercial projectors use much smaller lcds, so it would be much easier to offset the panel if thats how they do it.

I just stumbled across a diyaudio post from xblocker on 5/16/2002 :

"Here comes the Schleimpflug condition into the game. Scheimpflug discovered some laws concerning Off-axis imaging, which is also usable for projecting. He statet, if the object-plane, the lens-plane and the image-plane are meeting in one point, then you have a sharp image".

It certainly seems to be relevant to this thread, even to your most recent posts Mikau! Googling for this term pulls up a wealth of interesting stuff. From a quick first take, is seems that if the LCD and screen are not parallel, focus can be corrected by tilting the triplet to an intermediate angle.

Weird, since what I was looking for something totally unrelated!


http://images.google.com/images?q=Scheimpflug

edit: correct xblocker link. edit 2: of course, now that I read, this is totally well-known on the forum... but the name seems new :-)

Thats extremly interesting news. I only hope this works for triplets. Then again he said if the object plain is flat, which it is, so it should work right?

This would explain why some people here claim they can achieve keystone correction by tilting the triplet, and why another guy claimed he could achieve keystone correction with a mirror system behind the triplet. After analizing I quickly realized this was the same as tilting the triplet.

But this would probably still be limited by the size of the triplet. Therefore I suppose it might be better to have the collector a little farther back to focus more light closer to the center so it takes up less space inside the triplet allowing for a more extreme angle.

So who wants to test this? We need to try projecting onto a 15 degree screen by tilting the triplet at an angle so that its plane lines up with the the intersection of the planes of the screen and the lcd.

From personal testing of tilting the triplet this is what I have found.
Triplet tilt results

Let me add that the center is always in focus. The only parts affected are the edges.

Hmm.. did you try doing it using this exact method? Other tilting angles may come close but maybe this perticular one is exact.

Well as an example, I said lets say we have the projector 9 feet from the screen and lets tilt the screen to an insane 45 degrees. Using this method, I calculated the triplet only tilting about 5 degree's! That seems awfully low for such an extreme angle of correction. What would 10 degree's do?! 90 degree's?

Somehow I feel that can't be right. Though I fervently hope it is.

Don't quote me on that yet, I'm going to recheck my calculations.

This would mean you can achieve keystone correction to almost any desired degree by simpy tilting the triplet plane to face the point where the planes of the screen and the lcd intersect. If this is true why do commercial projectors use things like lens shift and digital keystone correction instead of a tiltable triplet mechanism with a relativly simple trigonometric tilt calculation?

I'm begining to doubt this theory is 100% true.

Yes, a 330mm lamp -> collector length would help the "arc image." It would also greatly assist fixing the dark corner effect in a box where everything is properly aligned.

But it would drastically reduce brightness.

This weekend I'll mess w/ Raytrace and Excel a bit to try and show what I'm describing.

If you change the lamp->collector distance by 1.5x, the collector fresnel "appears" to have 1/(1.5^2) as much area, from the perspective of the lamp. The difference is that the edges were already getting less light, and cutting them off has less of an effect than just reducing the light to 1/2.25 of what it used to be.

Squalfish you take no notice of this newly arisen topic?

I think this would just affect the focus of the projected image, not its overall shape?

A possibilty but thats not what I've heard from triplet tilters. Lol. A weird bunch they are.

Absolutely! They're a bunch of total freaks.

Anyways, so what about this triplet tilting thing?

For the 45 degree insane tilt example. I'll wager one of those tilter type people could get your insane lcd into perfect crispy focus. It would just be shaped wrong.

Regarding commercial pj's. Check out http://www.christiedigital.com/products/DS...cifications.asp . This high-end DLP projector ($18995 on pricerunner) lists "Scheimpflug (tilt) adjustment" as a feature. (for what that's worth!)

I wonder if its not just the triplet itself. Maybe theres an additional lens as well for the shleimflug adjustment. Who knows.

I should be finished my projector soon. Then I'll give it a try.

That dlp projector sounds amazing.

What does dlp stand for anyway?

Nevermind I looked it up.

Anyway, last night I tried testing the schleimpflug theory with a magnifying glass and a tv.

The results were, as charlie10 guessed, a clearly focused trapazoid. But theres one discovery I made which would explain why dergrin claims to only get partial focus, unless he used the following method already.

I discovered that while the theory seems true, it should be noted that you must keep inside the triplets angle of focus. Which means the triplet should always be pointing towards the middle of the lcd, not just tilted. Which would mean not just tilting, but also moving the triplet, see pic below. I should also add that this of course has to be at a working focus point for the lens so for a specific tilt at a specific distance the lens distance and angle from the lcd needs to be readjusted, and the lcd angle must be readjusted (or the projector itself)

NOW digital keystone correction would work but I'm not willing to go that rout since that means sacrificing resolution. We just need a method of correcting the trapazoid without affecting the focus. I think tilting the fresnel also effects the focus so if thats true, that won't work. Perhaps a curved anamorphic lens of some sort.

I have to think about this one.

As for getting the light in the triplet when its in this position, only way we can do that is by placing the collector behind the lcd and tilting the lcd. (Top pic below)
If the collector doesn't destroy focus you could just tilt and lower the triplet. (bottom pic below)

An anamorphic lens is capable of shrinking and adjusting the vertical and horizontal ratio ratio of the screen. So there must be a way to make it shrink one side more then the other. I need to figure out how anamorphic lenses work.

Heres the link for building an anamorphic lens.

http://home.c2i.net/ahustvedt/lens.html

What would need to be done is you would need to find out how the anamorphic lens is adjusted to correct for differant sizes, then we'd need to figure out how to build one to work at a specific uneven pace. This would probably require bent plexiglass or maybe just tilted prisms if we're lucky (which makes it unlikely). Anyway, that I think is our next course of action.

When I finish my projector, I'll try to build an anamorphic lens and see what I can learn from it.

Yes I enjoy talking to myself.

Anyway, another test, seems to work as well. I'm trying to determine the range and domain of this shcleimflug function. Being that the lens is 10 times closer to the lcd then the lens is to the screen I think we can achieve some pretty extreme tilting.

I'm going to see if I can reassemble my slide projector triplet and test it with that.

Great results! I will try to verify these two with my test-bench and post pictures.
Are you considering the possibility of shifting the collector fresnel?

Shifting the collector lens won't work much unless it is bigger. Perhaps we could use one of the profield fresnels but that would only work behind the lcd, but I'm not much concerned with that at this point.

I printed out the anamorphic lens instructions I'm going to read it in detail tonight and see what the differance is when changing sizes, then see if I can estimate what sort of modifications could be made.

Come to think of it charlie10 I believe we've struck gold. Do you remember my thread "resolutions and scaling- how much is necessary?"? I talked about how fullscreen formats use 480p at maximum. (except in pc games) and how you can shrink the fullscreen image size down a great deal before you get less then 640x480 when your using a 768x1024 screen. Therefore if we used those keystone correction programs that come with those graphics cards, we really wouldn't loose any resolution at all if we shrunk one side a little smaller then the other, it will still be over 480 pixels vertically. And with widescreen letterbox format, you can simply increase one side into the black bars instead of shrinking one side. Therewere assuming the schleimflug theory is correct and these functions can be performed with this so called keystone correction program, we have a solution for keystone correction with zero distortion and no sacrifice in resolution (except with pc games )

Sadly my slideprojector bulb just blew :angry: so I won't be able to confirm the schleimflug theory is correct.

But even if this does work it still offers no solution for people without pc's. Unless there is some known image scaler that has a built in keystone correction program which would make little sense.

Therefore I'll see what I can do with this anamorphic lens.

Isnt normal keystoning is accomplished by tilting the collector but yet the triplet isnt lowered? When the fresnel is tilted does it not still direct all the light directly forward and not down at an angle? By moving the tilted triplet down would you not lose a lot of brightness?

Maybe. I'm not sure. If what you say is true then all you need to do is tilt the lcd.

Read about the anamorphic lens last night. And I really have no idea how it works so I'd probably just have to use manual experimentation. I'll try it but I doubt I'll be able to find it randomly.

Does anyone know if the fresnel corrects the focus or is it just corrects the keystone the way digital keystone correction does?

I tried to set up a few experiments to validate these diagrams, and ran into a problem. Basically it was really hard to guess the right focal length and adjust things by tweaking. The biggest problem seemed to be that the focal length on both my triplets is too long for my test setup, which has a screen quite close to the lamp. At one point I had to put my "LCD" (a transparency mounted on a piece of scrap lexan) behind both fresnels to get it to focus! That was interesting.

So, still working on it.

I got the calculations going which was a huge help. 1/f = 1/d + 1/D.

Right now my setup is:
halogen lamp (7mm arc)
- 12 1/4 " -
220mm fresnel (grooves facing back)
- 21" -
290mm Fujinon triplet
- 28" -
screen

My 'LCD' really transparency is 1" in front of the fresnel, and tiltable.
The projected image on the screen is a manageable 18" across.
I'm using just the single fresnel, since I'm cramped on space and want a small projection for now.

This setup allows me to keep focus up to about 15 degrees of LCD and screen tilt (they're parallel); however it wouldn't focus on any part of the image at any greater tilt than that. It definitely seems like the LCD is no longer inside the triplet's angle of focus. So I think I've confirmed what Mikau said in an earlier post about the triplet's angle of focus.

To get some more interesting results, I think I need a fresnel in front of the LCD. Unfortunately the focal length of my fresnels is only about 10% more than the focal of my triplets, which seems to force my to place the screen at least 7' away. My screen is small (4'x5') so this is not ideal. I'll try to figger something out...

I calculate with a larger projection you should be able to do more.

Still trying to figure out a way to correct for the trapazoid without affecting the focus. I'm thinking acrylic mirrors might work if you bend it to correct the shape, but I'm afraid that might mess up the focus, see pic below. The black line shows the point where the screenshould be for a sharp focus. If you use a bent mirror, it may just pull certain area's away from the screen producing a curve. Then again it might not.

I'm going to try experimenting with that anamorphic lens as soon as I'm done my projector.

The curved mirror idea sounds interesting. I could see using reflective mylar and some sanded wooden forms to make a mirror curved in one dimension. It would be really tricky but based on some of the woodwork I've seen on this forum, definitely possible. The focal plane might become a curve as you show, but maybe the focus is forgiving enough for this.

Tilting the front fresnel does seem to correct the trapezoid, and to some extent alter focus. It seem tilting the lcd can recorrect the focus. As reported by dozens of LL pj builders.

Still waiting on my 15" diypjco fresnels, apparently they've arrived in US customs. I still haven't really looked into anamorphic lenses.

The people here do have great building skills, but there is a huge difference between making a nice enclosure and making an optical grade curved mirror, this is like making the mirror for the Hubble. Close just won't work, it would have to be absolutely perfect and everyone has different projector placement requirements so every on would be a custom job.

The problem is I only have one example for an anamophic lens, I need several. Then I can see what you need to do to change the size of the image without affecting the focus.

Ok charlie, I just made an accidental but fruitfull discovery.

I hypothosized that if you used two perfectly round mirrors (meaning a piece of a large circle) one concaved and one conxex, you could use the first mirror to shrink the image and the second to stretch it back the way it was, the idea was basicly making an anamorphic lens only with mirrors.

Well drawing light path reflections on cad and reflecting off perfectly round objects of the same size didn't work so well. I tried various sizes, tilts and alignments and they all produced wacky curves with proper focus distances scattered in bizarre irregular patterns. Then I theorized that the light beams had to reflect off of the outside of the convex mirror at the same angle it came in on the concaved mirror, thus reversing the stretch effect. Well I found that would require an irregular mirror so I abandoned that. Finally I decided instead of using mirrors of the same size and curve, use a smaller mirror with the same curve and place its radius point over the same radius point as the larger mirror (concentric circles). Tracing the reflections I finally produced something orderly.

The result was picture A below, you can see how the curvature and location of the mirrors lies on two concentric circles, the focus points (where the screen should be for a sharp image) ended up producing a very mild curve, virtually flat. (picture B shows the original cad document, you can see by the size of the circle how mild the curve is) However much to my suprise, the spacing between the three beams was uneven. BOO YA! You know what that means? That means the resulting image would be a perfectly focused trapazoid. With this you could in theory, project a trapazoid, and then use the schleimpflug tilt adjustment system to correct it. (obviously the mirrors would be smaller then the ones used in my drawing, it was just to see a close up)

There are three kinks in this theory.


1: The centerline of the image is off centered thus I assumed the image would be a trapazoid. However it may actually turn out to be a trapazoid with curved edges, (since I only checked the spacing between three beams) however this might be fixable using the schleimpflug adjustment to a specific angle. I'm going to try it using more beams later.

2. I'm not sure exactly how we can calculate what level of tilt a given mirror set up will procude. It would probably be better to make the mirror system first, and then use the schleimpflug adjustment to tilt the screen untill the trapazoid was fixed, but who knows how much or how little tilt would be required? Perhaps adjusting the direction the light shines onto the concentric mirrors would alter the trapazoid, perhaps nots.

3. Papercam is right. Producing an accurate curve by bending is not going to be that accurate and optics are unforgiving when it comes to imperfection. While its 1000 times easier to bend it to an even curve then to exponential curve, its still not going to be flawless. Perhaps we can obtain perfectly round mirrors from an optics company. If not, we still may be able to do it, the larger the mirrors, the less precise they must be.

Just checked rolyn optics. Found a bunch of concaved mirrors but no convexed ones. Convexed are often used in supermarkets but I doubt their very precise curves.

Whoops! I forgot! We need mirrors that are curved one way, not both ways. Hmm..ok scratch the supermarket mirror. Also the rolyn optics mirrors. Still I think we may be able to find them somewhere.

Charlie, you know those raytrace demos you were making? Would it be possible to make one out of this?

I was just looking at your triplet diagram a few pages back on this thread, and I noticed how the triplet takes each pixel and focus it to a point, I'm wondering if a bent mirror effects that focus or if it just redirects the beam.

Yeah, I have to do another trace to find out where the focus points end up. But I have a feeling it will still work.

WOULD SOMEBODY PLEASE SAY SOMETHING? :angry:

Very simple means of doing it: use bent acrylic mirrors. Forget the sphericals.

If you bounce it at an angle off a 1-dimensionally bent mirror (think cylinder subsection), the top will stretch at a different rate than the bottom, and you will have your trapezoidally-distorted (keystoned) image. The amount that it's distorted is proportional to the angle of attack, the amount the mirror is bent, and the distance of the center of the beam from focal point to mirror - it can be adjusted easily.

If you're having trouble visualizing it, picture the same 1 mirror design that lots of people are building, but replace the flat mirror with a cutout of a shiny cylindrical trashcan.

By all means squalfish, but I was talking about a 1d bend, when I said perfectly round, I meant bent at an even curve (a piece of a circle)

Ahh, okay, thought you were referring to a sphere for some reason.

Is your diagram a view of the PJ from the side?

I thought of this originally (read what I first said about the mirrors) but wasn't sure if it would effect the focus points. I thought it would cause them to bow in or out. Now I'm not so sure because of the fact that I was just tracing straight beams instead of rays focused to a point. I need to redo it and see what happens.

Side or top depending on if your doing vertical or horizontal keystone correction.

My idea was to use a single mirror shaped cylindrically, but orient it so that the top-bottom centerline of the mirror appears to the image as a straight line.

In place of a 45 degree FS mirror, a 'la OHP style:

what?

If a single very slight circular curved mirror is used at an angle, then the bottom of the beam should be "squeezed" slightly less than the top of the beam, without distorting the picture in any other way.

= keystoning

Now I get it.

But can you varify thats true?

Unfortunately, not without a 3d raytracing program and the expertise to use it, or physically building the thing (which is an option when I finally get everything together, but not yet). The principal should work regardless of mirror angle, so you can use whatever's practical. I derived this from a realization that you couldn't use a mirror as a replacement for an anamorphic lense without problems in the other dimension.

Hmmm...well what if we did two 2d traces? one for an xy plane, the other for an x,z plane?

That woul be tricky

I'm not 100% sure, but I don't think you could combine it like that easily. If we could grab someone experienced with 3D raytracers, we could ask them for a simulation.

I have finals I need to be studying for, so I must take my leave of this discussion for tonight.

Ok thanks for the help, dude!

Ok. Last night I did a somewhat primitive drawing and the curve of the mirror shortened the focus point so much it crossed before it got to the second mirror. However, I think I drew the rays leaving the triplet a bit too big. Charlie, do you have any idea how wide a single pixel ray is when it first leaves the triplet?