New theories for keystone correction, merged topic Options

[Mikau]

What we need to is set the screen to display widescreen all the time, and when you watch/play something in fullscreen, black bars appear on the sides, instead of top and bottom. A 1080x576 image can fit a 640x480 image inside it, so that would work. It would be smaller, but it would still work.

[Mikau]

Ok. I did a little more research on this.

It is true that the full screen formats use lower resolutions (480p) and the high def formats are widescreen, this works well, you can scale down all your images to widescreen and pan it to the buttom of the screen. Use your scaller to play the full screen formats inside the widescreen box, with black bars on the sides. This will make full screen formats smaller but you will not sacrifice any resolution if the box is big enough to fit 480p (640x480 pixels).

The only thing that throws a wrench in this mechanism is fullscreen dvds, and dvds that use a 1:1.85 ratio. These are dvds using the same high def resolutions made for widescreen tvs only the picture is zoomed in. This makes a bigger picture and better resolution (assuming your watching a dvd that displays higher resolution then your lcd such as a 720p or 1080i dvd). But some of the picture is choped off. While full screen format dvds can generally be avoided, about 50% of dvds use the slightly zoomed in 1:1.85 format. While you could keep it the same size, you might want to make it slightly larger so you can appreciate the closer look and better resolution since your sacrificing a little of the sides of the image in that format.

I need to experiment with scallers to find out how I can get these three formats to work together without stretching or shrinking the lcds image manually everytime you change formats. With the widescreen format you should be able to place the projector almost centered on the top of the projection screen, same on the full screen image, and on the 1:1.85 image the projector would be about halfway between the center and the top. This might be enough so this could work.

[this-is-me]

Since this is the keystone thread anyway, what would happen if the front lens were tilted instead of the front fresnel? Would there be any ill effects?

[Mikau]

Well tilting the entire triplet doesn't work. However if you take apart the triplet and tilt the front lens, it produces a trapazoid in perfect focus, tilting the screen seems to correct this though. It does work to some extent, but I'm waiting to finish my lumenlab projector to put this idea to the test on a lumenlab projector. So I'll keep you posted.

[Mikau]

The absolute best thing would be if we just had a slightly differant triplet and collector, a triplet with more magnification so we can place it farther away, and 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.

In my oppinion. This should be the standard lens setup. Because who doesn't want to be able to place their projector on the ceiling or floor, completely out of the way?

Attached File(s)

 ketstone_triplet.JPG ( 26.22K ) Number of downloads: 5

 

[charlie10]

I agree, I'd like to build the projector to go on the ceiling or floor, completely out of the way. I like the idea of a collector that points light towards the bottom. I'm trying to design a widescreen PJ so cutting a large fresnel in half (or at least somewhat off-axis) seems to make sense.

Here are some raytraces (animated gif) of keystone correction. The first shows an on-axis collector fresnel that is tilted to correct the keystone. These are 2-d diagrams, please refer to the relative spacing between the rays on the top and bottom ("squashed" vs. "stretched"), I think that demonstrates the keystoning.

Unlike the last post this assumes pj on the floor not ceiling.

Ignore the projection lens in all these diagrams.. I don't think I understand much about that lens yet. I wish I'd done the raytraces (esp. #3) without the lens. Please let me know if there's anything else I did wrong, I'd be happy to post corrected raytraces.
Charlie

Attached File(s)

 ray_keystone_onaxis.gif ( 27.34K ) Number of downloads: 5

 

[charlie10]

Trace #2 This tries to trace an offaxis collector fresnel. I kept the projection lens at the focus of the fresnel, for what that's worth.

It seems very effective at raising the image, but there is the same keystone problem. In trace #3 I tried to correct it with tilting.

Attached File(s)

 ray_keystone_offaxis.gif ( 84.31K ) Number of downloads: 5

 

[charlie10]

Here's trace #3. To me it looks like a combination of tilting the collector fresnel and designing it to be off-axis might be effective.

For this trace I started to play with the position of the projection lens. I actually have no idea what I'm doing here, and this raytrace is basically incorrect in many ways. I'm currently working on how to raytrace a triplet. Only then could I start to model the diagrams Mikau has done in MS paint.

Meanwhile, I read in another post http://www.diyaudio.com/forums/showthread....10&pagenumber=6

What happends is when you go too far off axis of the frensels contoured rings angle they get a negative effect. The rings that are missalighned to your light will go dark and show in your image. Another thing that will happen is the grassing affect, (same thing as having the rings missaligned from one frensel to another)."

Maybe the idealized thin-lens that RAYTRACE here uses, can't really simulate off-axis fresnels? I don't have any fresnels yet so can't experiment...

Another comment: I used an Infocus X1 for awhile and it projects "up" with keystone correction by default. It is fairly short-throw but it did a perfect job of optical keystone correction, somehow. How---?


This post has been edited by charlie10: Mar 7 2005, 04:41 AM

Attached File(s)

 ray_keystone_offaxis_crrct2.gif ( 12.03K ) Number of downloads: 5

 

[charlie10]

edited down, dam these raytraces turned out big

This post has been edited by charlie10: Mar 7 2005, 04:45 AM

[Mikau]

That is the sexiest thing I've ever seen!

Very very interesting. But notice the extreme angle its not hitting the lens. That may be too much.

[Mikau]

Ok. Heres some analysis. Moving the lens farther away from the lcd results in a smaller picture, closer up. Moving the lens closer to the lcd results in a bigger picture, farther away. That said, you would think that you could achieve keystone correction by tilting the lcd, one side is closer so the projected image will be smaller and closer, the other will be farther and bigger. The problem with this theory is that the ratio of the distance of the lens to the ditance of the screen changes depending on how close or far it is.

So if you tilt the screen, the proper focus points for where the lcd should be would be curved. As shown in diagram A below.

Now I suspect tilting the fresnel helps to compensate for this BUT the fresnel is still flat. If you could mathematicly calculate exactly where the fresnel needs to be in each location you could create a curved frame and BEND the fresnel on these points exactly, that would in theory produce a perfectly focused image.

Moving the fresnel is differant then just tilting the lcd, and the fresnel distance and angle obviously has an effect on the lcds focus distance (since it works for keystone correction) But obviously there are some aberations produced. The fresnel may have the same issue with irregular magnification so it may need to be first curved to compensate its own focus, and then curved from there to compensate for the triplets focus.

My only worry is I have a sneaking suspicion that this will produce a trapazoid, in perfect focus. However, maybe this can be compensated for by tilting the lcd as well as the fresnel. The other is we can't bend the fresnel too much without making it too small to catch all of the lcd.

You would be lost simply trying to bend your fresnel. There are an infanite amount of combinations and the curvature varies based on projection size and screen tilt. Also if the starting point of the fresnel would have to be tilted, and you had to adjust the lcd as well to compensate for the trapazoid, there are an awful lot of variables.

In otherwords, I think the exact curvature would have to be calculated mathimaticly, you couldn' t just manually adjust to find it.

How precise would the curveture have to be? I'm not sure exactly, but I'm sure it would work better then a flat fresnel. The accuracy would depend on howmany points you plotted and how precisly you bent the fresnel around that point. You could use push pins, a curved grove or maybe attached to square fram with a curved front.

How would you calculate the differant focal distances of the lcd? You would have to calculate that based on the distance of the screen in various points. The more points you plot, the more accurate it will be.

To figure this out mathematicly, we'd need to know exactly how the fresnel effects the image based on distance and tilt. This may also vary in differant area's of the fresnel. This would require a lot of experiments to calculate a formula to figure out exactly how the image is effected. This formula would have to be used with the focal distance formula for the triplet.

In otherwords, it would be a highly scientific mathimaticle procedure to figure this out, and create a formula to plot the points of the curveture. But once you had it, you could just punch in the distance and angle of the screen your projecting on, and it produces the points to bend your fresnel around and how far to place it from the lcd AND what angle to tilt the lcd (if thats required).

We need that guy from N3mbers!

Attached File(s)

 Plan_A.JPG ( 38.29K ) Number of downloads: 5

 

[Mikau]

By the way, bending a fresnel one way is very easy. Two ways is whats hard. This should only require one way.

Keep in mind that the fresnel may not (and most likely will not) simply have to be curved around the points where the projection focus is.

There are an awful lot of things to be considered here, so we're going to need a math genius.

But I have confidence that it will work, if you can find the right curveture. But I doubt that can be done manually, So we need it to be done mathematicly.

You could try bending your fresnel by hand when keystoning to see if you can get any improvement I suppose, you would have to tilt it as usual and pull the middle in slightly, but again, it would be very difficult and you might ruin your fresnel.

[sjetski71]

By the way, bending a fresnel one way is very easy. Two ways is whats hard. This should only require one way.

Keep in mind that the fresnel may not (and most likely will not) simply have to be curved around the points where the projection focus is.

There are an awful lot of things to be considered here, so we're going to need a math genius.

But I have confidence that it will work, if you can find the right curveture. But I doubt that can be done manually, So we need it to be done mathematicly.

You could try bending your fresnel by hand when keystoning to see if you can get any improvement I suppose, you would have to tilt it as usual and pull the middle in slightly, but again, it would be very difficult and you might ruin your fresnel.

Just want to clarify a few things and perhaps add a note of caution.

Not trying to stifle innovation here but Mik where have you read about bending fresnels? (just asking because you sound fairly convinced) Maybe i've missed a good plog or two but i haven't noticed the method employed anywhere. I will admit the idea sounds intriguing and maybe possible with the "perfect bend and placement", otherwise...hmm.

Just a note of caution for the noobs before they begin trying to bend their fresnels with a karate chop or two

[Mikau]

How do i know bending works? Well to be honest I don't. But why not? The fresnel should bend most around the thinest area, which is the gaps between the triangles in the fresnel, so the triangles will not be bent, they will just be tilting.

[charlie10]

Here's a pic I found which seemed very encouraging. This looks just like the adjustment on some commercial projectors. I highlighted the red because it sounded to me alot like some of the stuff in this thread, ie. on the right track?

Attached File(s)

 keystone_navitar_xypc_lens.gif ( 15.07K ) Number of downloads: 5

 

[Mikau]

Nice! Where did you find that?

I'm gonna pm joechevy about this. We're doing some experiments with lens shift.

Any more info on that lens?

[Mikau]

HOLY SMOKES! I just found the page. Tell me that thing is an attachment you can put in front of any projector!!!!

WHERE DO YOU GET THAT THING???!!!!!!

[Mikau]

Hmmm...well the only place I found it available is a place where you can "rent it" for 25 bucks.

So that thing must be an expensive product only for commercial use. :angry:

[Mikau]

Yeah, just checked the pricing. Everything seems to be 1000 dollars and up. But at least we finally have some information on the lens shifting system.

Back to the fresnel bending idea. Where were we?

[charlie10]

i found it in the middle of a google fest, it showed up under "projector keystone optical correction". There might be some other useful stuff in there. 47000 results :-]

I'm pretty sure the Epson 500 has a similar pair of knobs to twist. They call it "manual lens shift". They list "keystone adjustment for fine tuning" as a separate feature.

I wonder if this type of lens only works well if the input is already a collimated / parallel beam? Or does the first lens of the triplet do that for us. I'm still trying to understand the basics of that triplet! Optics 101 (or maybe 201?)

[Mikau]

Not sure. I emailed the company of that gnk xypc keystone correction lens for info on pricing and dealers. And they said its been discontinued but might be available on ebay.

For now I'm going to think more about bending the fresnel.

[Mikau]

Charlie if you could do me a favor, could you post a still picture of the drawing of the keystone correction after its been corrected? I'm trying to compare the two and see if ALL of those rays have been corrected.

If not then we can try readjusting the collector, then see if other rays are corrected. That would prove that a curve is required.

Bending the collector might mess it up a bit, but it might mess it up precisly the way we want it.

The tricky part is when bending the collector is a lot differant then just tilting. Its like tilting each piece of the fresnel in differant spots. Only differance is they are connected, so tilting it effects the height of the area's of the fresnel. But again, these aberations may be what we need.

[Cold Steel]

Here is the pdf for the GNK.
http://www.navitar.com/download/gnk_keysto...nstructions.pdf

All the pricing I've found for it is call us.

[charlie10]

still workin on it...

[charlie10]

Here is the one I think you wanted -
The RAYTRACE demo doesn't allow saving, so I lost yesterday's model -
I still had the gif though, here is the 3rd frame and zoomed up the screen. Do they look even?
I'm working on another model using RAYTRACE's "measuring stick" tool.

Attached File(s)

 ray_keystone_corrected_b.gif ( 26.02K ) Number of downloads: 5