Dang. Now I'm going to be up all night again.

I'm anxious to see how much it works with an lcd and a halide.

It's pretty basic optics. Every lens has a focal plane, not just a focal point. The question is whether it is sufficient when the triplet is closer, as is needed for a bigger projection.

Great, good effort.

I've been testing this and see the exact same things, a bit of coma, some softening at the top corners with the triplet level with the top of the lcd. Full 4:3 focus will be a problem if I go this high, my goal is ~100" diagonal with bottom of projected image level with base of projector - about 12° of axis tilt. Seems like it will work.

A while back I saw a picture for testing focus that had a bunch of really thin concentric circles (or was it squares?) real close together, to use as a focusing image. Anyone know where that is?

Well i'm not sure what's going on now...

I managed to get my lamp/collimating fresnel (CF)/LCD/field fresnel (FF) and triplet all aligned on the same axis, based on an 87.5mm lens shift.

I'm not sure what i've done, but the top portion of my picture is now slightly bowed. I have a feeling this has to do with the lamp/fresnel alignment changes that I made. After trying the 87.5mm lens shift, I tried 70mm, 60mm and then the 50mm lens shift that I had done the last time (before moving the lamp), and in all cases the top of the image is slightly bowed. The last time I did a 50mm shift I did not have the bow, I can only assume that it is my lamp/fresnel setup. Any suggestions here ?

The good news is that the lamp and fresnel movements are certainly getting light into the triplet, however if this new lamp/fresnel arrangement has caused the bow at the top of the image, this is not going to work unless I can fix this. I need to see if I can make my lamp setup more adjustable. Not sure about the fresnels either. Get them on the axis or is just close enough going to work ? Need to figure out the effect on the projection.

Focus in the top two corners appears to not be so good either. On the 50mm shift, it appears to be worse with this new lamp/fresnel position. Again, this may be part of my lamp/fresnel setup...

The other possibility is that we are getting close to the limits on the FOV. The only thing that doesn't make sense is the difference in the images for the same 50mm lens shift....

Although (thinking out loud here...) perhaps we are indeed at the limits on the FOV. The more I think about it, the lamp stays on the same axis as a 87.5mm shift even though i'm moving the triplet around, hence, the bow at the top of the image, even when I go back to the 50mm shift. ?? I need to work on a lamp adjustment. I'll find a happy medium for the fresnels and leave them there and just adjust lamp and triplet position... next test this weekend ?? (folks visiting and a wedding to attend, not to mention a cookout on sunday... ahhhhh, I just need more time.... )

Some more good news. The PJ is sitting on the floor. Projected image is now 120" diag (8'x6' screen size). Things indeed look good for lens shift. I can raise my PJ up about another 9" before it starts interferring with the view from the couch so I can reduce the req'd lense shift if necessary.

Some pics:

My setup:

Click to view attachment

Click to view attachment

Click to view attachment

Click to view attachment

87.5mm lens shift

Click to view attachment

70mm lens shift (photo exposure modified)

Click to view attachment

60mm lens shift (photo exposure modified)

Click to view attachment

50mm lens shift (photo exposure modified / camera reframed)

Click to view attachment

last one... widescreen shot with 87.5mm lens shift. looks good except for the bow at the top of the image. General brightness in all the photo's is off from the previous set as I have removed my reflector until testing is over and the bulb position is finalized.

Click to view attachment

I also took some video of this test. Moving the image to the top and bottom of the viewable area using ZoomPlayerPro. I will check this out (to see if the footage is even any good) and post if there is any interest.

cheers,
gs

Some thoughts.

I think you are seeing some pin cushion distortion caused by the front fresnel.
(assuming you have this layout.) I think that in the absence of any compelling reason for it to be there (keystone correction), it should be moved (back) to the lamp side of the LCD. A fresnel in the imaging path is just wrong.

That said, obviously there are limits to what the triplet can do FOV-wise. The quality of the triplet is not superb, just adequate. This is pushing it.

What is your LCD-FF distance? reducing this will reduce your required FOV by a small amount. The LumenLab plans say 20mm, this makes the triplet see an LCD of height 243mm instead of 228mm, on paper at least. If this is higher, say 40mm it'd see 260mm, albeit slightly further away but the apparent height increase grows faster than the apparent object distance. The Fresnel distortion gets worse in this case, I don't think they were ever intended to be used for this purpose (imaging instead of illumination).

i have checked it and found it is not true. The used field angle on both cases is the same. (20 gap and 40 gap). (Could be that i did a mistake as well )

20 gap with 228mm wide TFT; 243.3mm wide virtual tft at 21.3mm from the FF.
40 gap with 228mm wide TFT; 260.9mm wide virtual tft at 45.7mm from the FF.

if we take the same reference point at same distance from the FF, the angle is same for two cases; (243.3/2)/(320mm+21.3)=(260.9/2)/(320mm+45.7)

but maybe this 320+x distance i did is not fine, i see that the trhow would be different for each case... i don´t know...

Indeed it is different. Maintaining the same throw distance at about 2.5m gives a required FOV difference of about 1° from axis - as I said, a small effect, but could be quite significant when you are near the limit.

but, if you mantain the trhow, the image would be larger on the second case.

Let's see what happens with field angle, if we manain the image size;

250cm trhow needs 36.69cm lcd to lens (asume 320mm triplet).
this is 250/36.69=6.8125 magnification. So let´s define 243.3*6.8125=1657mm wide projected image (20 gap case)

now, if we would like same image size, but we have 260.9wide lcd (40 gap) then we need 6.3529 magnification for the new setup.

so same triplet; 1/320=1/D + 1/6.3529*D D=370.37mm (lcd to lens) the new throw wouldbe D*6.3529=2.352M.

so finally we got;
20 gap; 243.3 virtual lcd size and 366.9 lcd to lens dist; 18.34 degrees hafl field angle.

40 gap; 260.9 virtual lcd size and 370.37lcd to lens dist; 18.18 degrees half field angle.

conclusion; the field angle is practically the same (i don´t think the 0.16 degrees could be noticeable). But the trhow has been reduced (this is very noticeable, and would say not desired also )

nice discussion

Ah, yes I have thought of this, but wasn't sure. I'm already trying to purchase a set of the old fresnels to get a 330mm in order to move to the unsplit configuration. Being able to go to the unsplit fresnels is really what is so nice about lens shift, IMO. I can try to move my existing fresnel behind the LCD, but I know light output will drop again... not to mention it won't be easy as i've glued the fresnel into a dowel hanger (when I thought keystoning was going to be used)

I may try aligning the fresnels again to see if this will help eliminate the problem (right now they are on the angled light axis)... just for SAG.

My FF distance is 20mm from the LCD. Both tests were done with this configuration. I'm not sure I want to get much closer, but I suppose I can move it up until I see the fresnel rings on my projected image. ?

Good point. I'm working on the assumption that the distance to the screen is the constant. That might not be the important parameter in all cases.

Depends how you look at it I guess, it might be a worthwhile optimisation in a few cases but really I think the solution is to move the Fresnel back behind the LCD.
Edit: of course that won't solve the problem, the FOV is the real issue.



Nice to see someone able to work the theory.

You may not need to. Move the light source closer to the LCD. This will move the lamp image foward. Might actually gain some light, might not, depending on the "point source-ness" of the lamp and how much illumination non-uniformity is tolerable.

GadgetSmith, the effect you are seeing is indeed pincushion distortion but it is not caused by your fresnels it is caused by reaching the limits of your triplet.
About 10 months ago I did a simular set up using a 17” LCD with both fresnels behind and got quite good results. I haven’t pursued it very much as the fresnels I got from Diy Projector Company are incorrectly aspherically corrected and I was unable to focus the light to a small point to fit through the triplet. Since reading this thread I’ve got interested in this again all I have to do is find some large fresnels. Has anyone got the large fresnels from 3dlens.com? If so let me know and I’ll tell you how to test them to see if they have been corrected correctly.
Any way back to the pincushion distortion. The FOV limit doesn’t only make the outer edges of the image blurry, the limit can be pincushion distortion or chromatic. The problem is that the angle that the triplet has to resolve to get the bottom of the LCD into focus is much greater than the top of the LCD. The solution I found was to tilt the triplet so that it points more towards the bottom of the LCD, this brings the LCD back into the fov of the triplet.

DJ

i don´t like the idea of tilting the triplet or the lcd (is the same, it depends where do you look from)

tilting any of them, would change the focusing plane. Let me see if i can draw a pic.

here are two straight cases drawn, then a tilted one is drawn. since the intersected points are well known on the projection, we can see what shape would have the tilted projection and where is it located.

check that the tilted lcd's projected image is tilted as well, so we need a tilted wall or we can project from the ceiling or from the floor with the required angle so the wall looks like if it was tilted.

i hope it is easy to understand just looking the pic:

The idea would be to tilt both the LCD AND the projection lens, thus also tilting the focal plane. The result is that the LCD, triplet and screen are all parallel, and you get complete focus.

need a pic supraguy.

Hi DJ,
The problem I see is that before I moved the lamp I had done a 50mm shift and the top of the image was not bowed. post 345 (the dim pic is the result of the triplet no longer on the light axis) Now that I've lowerd the bulb the same 50mm shift is giving me a bowed image at the top ?

I certainly see what you are saying about tilting the triplet, however I think Rox is right, in doing this we certainly will effect focus, which is what we are trying to avoid. We need a solution where the LCD, Triplet and Screen are all parallel, this will give us the sharpest focus possible. That said, I am going to try your suggestion as it is easy to do and I am interested to see the results on focus first hand... and to see if it helps with the bow at the top of the image.

Incoherent, I can easily adjust the bulb to fresnel distance so i'm going to try that too... maybe all I need is a little tweek here or there... I hope. At any rate I need to come up with a method for adjusting the height of the bulb more easily. (right now it's bolted to the side of the box... maybe i'll just make some long slots so I can put it at any height I want.)... my box is sure to look like swiss cheese by the time i'm done. Thanks (if I hadn't said it already) for your suggestions and insight thus far. (thanks to broken records...)

cheers,
gs

Very interesting, gadgetsmith. I've got my lcd in place and just need to do the wiring now. But last night I through a standard lightbulb in my projector and tried lens shift on the floor as before, and managed to achieve crisp focus. It was still slightly sharper on the bottom then on the top, but when you sit back a bit (since we can do that now ) you don't see the differance. Funny. I was projecting at about 7 feet and noticed no barrelling. Lemme take another look.

Anyway, heres my bulb adjustment system, gadgetsmith. Easy to make.

Below, looking from the inside of the box, you can see I have the bulb mount with a horizontal slot in it. Bolted to the box.

Now looking at the same spot, on the outside of the box, you can see I have a vertical slot and a wing nut. This allows me to quickly adjust up down, forward and back.

Rox here is the correct way to ray trace lens shift.

DJ

Rox you should also look into depth of field. Here is a link describing it.
http://luminous-landscape.com/tutorials/un...eries/dof.shtml
DJ

What you guys are talking about here is called the Schleimpflug adjustment system. Schleimpflug noted that if the object plane (the lcd) the lens plane, and the screen plane all intersect in the same point you have a sharp image. This is true. But it always produces a trapazoid. Picture A below shows what I'm talking about. It should be noted that you should not simply tilt the triplet as shown in picture B. Either tilt the lcd or raise the triplet and tilt, or you will simply pull out of your triplets field of view.

Ok. Now I'm not sure exactly what you guys are suggesting with this parallel alignment thing but if this is what you mean, I see a problem with it.

Picture 1 shows lens shift with the lcd, triplet and screen all paralell.
I believe what you are suggesting is tilting the triplet towards the offset lcd to stay inside its field of view. In picture 2, the triplet was tilted. You can see on the blue line where the focus plane ends up. Tilted. In picture 3 we tilted the lcd. The focus plain is now paralell to the lcd and the triplet. In picture 4 the screen is then tilted to match the focus plain. Now everything is parralell. Um..do you see what I see? We might as well have tilted our projector up and tilted the screen. We really just went in a big circle.

I can’t explain this. Only that maybe your field fresnel is bowed if you are using it in front of the LCD. With my set up the pincushion distortion was even more noticeable than in you pics, probably because of the 17” LCD and I was able to correct it by tilting the triplet. Although it was difficult to get an even brightness because of the faulty fresnels I used.

DJ

Well according to the schleimpflug system, tilting the triplet that much should really throw off the focus plane. See top pic.

Tilting the lcd you should be able to correct the focus again, but the tilt would be significant and would produce a trapazoid. See bottom pic.

This is not exactly Schleimpflug principle. I thought it was when I first started playing around with my test set up. Although some of the Schleimpflug principle may be having an effect. It is more lens shifting and tilting the triplet to limit the amount of pincushion distortion. That said, lens shift and Schleimpflug principle are very simular. Maybe some where in between
Just a note on the image plain; depth of field plays a big role here.

Tilting the triplet has little effect on the shape of the image, the only thing that I notice when I tilt the triplet is the introduction of pincushion distortion. I could not see any keystone distortion.

DJ

From what I have observed, the schliempflug principle is true. I've tested it many times. And also confirmed it using the lumenlab focal length calculator and plotting various points.

Tilting the triplet does not effect the shape. But it effects the focus. The schleimplug principle states only that the image will be in focus. It doesn't say the shape will not be distorted.

Just tested tilting the triplet as you suggested. Doesn't work. Its impossible to get the image into even focus. From what I can see keeping it parallel produces the sharpest image.

Yea, exactly it doesn't make sense. My fresnel is the same for both tests. I purposefully mounted the FF to a piece of 1/8" glass to keep it from bowing. (which it was prior to the first test). I'm going to change my setup and move some things around. I really think i'm close... just a few things and whamo, i've got it. Then if I can get a 330mm fresnel for going behing the LCD... Hummm.... that would be even better. (I think...)

gs

yes dazzla, tilting the triplet is the same as tilting the lcd, then schliempflug principle needs to happen there. (i didn´t heard about this principe but my pictures looks like it is right, need to check it matematically, will post my conclusions )

Your idea of using full field angle is wrong since the schlienflug principe would be noticeable.

about the field depth you talked about, it is only a distance where the focus is somehow still in a range of aceptable focus. So if you try to use this "tolerance" of distortion to hide the problem, it shouldn´t be desired. just try to keep as focused as posible, do not force it.

Anybody besides me think the name of the thread needs changed?

"Keystone Correction Soon to be Obsolete"

Great job with the testing Mikau and GadgetSmith!

Here's my theory on what is needed.
Add a chromatic doublet before the projection lens. This will help in 2 ways. Better color alignment and a more cullimated light entering the projection lens.
Then shift away!

I am aware that the focal length on many of the lenses may need to change.

i have checked the Schleimpflug principle and found it is right. there is a point always where the focused planes intersect with the lens plane.

so the conclusion i have form there is that if we were doing things in the perfect world (ideal maths) then if we did the keystoning using the field fresnells tilting method, we should tilt a litle bit the triplet as well. (but this tilting the triplet is soo small amount that could be not neccesary, but i repaeat it, if talking in ideal world, we should do it)

I fired up the PJ last night to watch a movie. Still had the 87.5mm shift and everything setup as before. I tried tilting the triplet to see if this would effect the bow in the top of the image. All it did was make the upper part of the screen out of focus (bottom stayed in focus). No effect on the bow.

Also tried moving the lamp position towards and away from the collimating fresnel. (I did this by moving the LCD sled and refocusing) No effect on the bow. Optimum light still at around 210mm from lamp to collimating fresnel.

Still happy with this setup (PJ on floor) when watching movies. The bow is less noticeable if I drop the screen lower (using ZoomPlayerPro). Tomorrow i'm going to play with the fresnel heights to see if I can get rid of the bow.... i'm really starting to think this is the problem... I also need to try the 317mm fresnel behind the LCD... pretty much guarentee's lower light output (I think), but i'm interested to see the clarity improvement by not having anything betweent he LCD and triplet.

cheers all,
gs

Maybe not renamed... maybe just relocated... as in Keystoning vs. Lens Shift thread.

Thanks, although i'm more of a lab rat around here implementing the good ideas of the mad scientists

I believe in the end, whatever we've arrived at a for a solution(s) for lens shift, we will need additional or newly designed lenses. I guess I haven't thought too much about an additional doublet, only about how to get the most out of the lenses I do have and a solution that will need a minimal of changes to properly implement. (research funds here are not a limitless supply... i only wish....) Perhaps wiping the slate clean and looking at this "from the ground up" would be an interesting thing to do as well...

-gs

I've been playing around with the doublet idea and IMO it looks to be the way to go, on paper anyway. I'm building a PJ based on the PSOne LCD right now, actually 3 of them. One for my son's room and 2 other for coworkers. All are wanting about a 60" max screen, maybe a little less. And price was a factor in all 3 (I was able to get 3 screens for less than $100). I found where to get fresnels that are 3.5"x5" for $1.25 each that are good grade and not cheaply made "page magnifiers". I'm trying to get more specs on them. Should fit the build great pending the exact specs.
I'm still working on the extra doublet and projection lens seletion to get the 50% lens shift I want. Trying to work with what is readily available cheap. Going with a straight design both to save money and for ease of alignment since I am testing some new ideas. Adding the doublet means changing the FL of most of the lenses to keep from making a box that ends up so long it looks like a bazooka.

I already have a brand new 17" Samsung LCD and controller kit for my next full size PJ. Have to finsh the three mini PJs first. I plan to implement all the new things from the PSOne builds that work best. That's one thing I like about the minis, they are cheap to experiment with.

I like lens shift method vs fresnell tilting to achieve good keystone effect (square shape and focused image on the non centered projection wall)

first of all, we are speaking about the field angle of the standard lens here and there, but yet nobody knows it. I have a standard lens model on OSLO and looks like it is 50 degrees full angle lens. Out of this angle, the performances fall down drastically, but inside this angle, performances are better on the center as well, it is linearly decresing performances curve until it is suddenly a very bad lens (out of 50 degrees).

So what i mean is, even if we are inside the fov, the performances are not even in full fov, so should be specified what we understand by usable fov.

I think lens shift is much better than any other way. (i mean just based on the maths behind it in a ideal world)

Now let me talk about field fresnell tilting method;

Here the filed fresnell will distorse the image, we all know this, this is because it starts interacting with the image optical system.

have a look to the first drawing;

Click to view attachment

this is a straight lcd and a tilted field fresnell, check that the virtual image (it id the lcd viewed by the projection lens) is trapezoidal shape and it is tilted in the oposite direction as well, this is because of field fresnell effect.

Click to view attachment
now since we have a virtual tilted lcd, schleimpflug principle says that there is a point for the virtual lcd plane intersecting with projected image plane and triplet plane, the three planes interesect in 1 point (well, line ), so we should tilt the triplet that way always, not only for keystone in more than 10 degrees. It should be done always, more or less, the maths will tell us how much is needed.

But the question is, how much tilt does it need on each element for a especified setup? We can work out the exact amount of tilting (in ideal world, i just understand the ideal world, i think it is a good starting point) just let me know some axioms (trhow, image size, keystone degreess...) and lets see is we can work out.

Rox that is nearly the exact post I was about to post, thanks, saves me having to do the pics LOL. One thing that I didn’t think about that you mentioned was tilting individual lenses in the triplet. This could make it possible to get more than the 12-15° keystone correction that is normally achieved.

Just a note on depth of field: You’ve already stated that schleimpflug principle should be used for best results, this is true, my point is that tilting the triplet is not that necessary on smaller angles and in this situation the system is relying on the depth of field of the triplet.Plenty of people here have used fresnel tilting and no triplet tilting with good results.

DJ

Dazzla, I have been working on the keystone math for a while, but havent found a solution to that question of howmuch would the triplet tilt for a certain keystone degreess, i have seen there are much variables implied though, but the statement of low keystone degrees not needing the triplet tilt, i don´t like it . (I told you i like ideal world, i know that this angl ecould be 1-3 degrees of tilting the triplet, maybe nonpractical, but if ideal world says it should be done, i would try to do it. Deph of field tolerance is an easy excuse .

you said higer keystone degrees need the triplet tilting, whats your criteria on when to tilt it or not? My pont of view is that it is always needed, doesn´t matter how much keystone you need. Need to work on the maths behind it yet.

Anyway i prefer lens shift method, it is much easyer.

sorry but i have been reeding the previous posts and found you were talking about lens shift + tilting the triplet (to use better fov on lens).

well, i don´t agree with this, i said triplet tilting is needed on fresnell tilting method only. Since the lens shift method, the real and virtual lcd are paralell to the projected image on the wall , ther eis no need of tilting the triplet, just place it paralell as well.

IF you introuce the tilt on the triplet, any (or both) of the other things need to tilted as well. That will change the shape on the image, so good bye to a propper keystone correction.

This is not an excuse. It is a fix for a problem


When you can see that the image is out of focus.


This is where pure theory and real world don’t match. If we had unlimited funds and were able to design what ever we needed, near ideal theory could be reached. But we have to utilise what we have available to us, in this case a triplet FOV that is being pushed to the limit. GadgetSmith experienced some pincushion distortion on the top of his image, This indicated to me that he was pushing the limits of the triplet, my suggestion was to tilt the triplet to try to correct the pincushion. And yes this doesn’t agree with your perfect theory, it is what it is, a fix. It’s not perfect and I’m terribly sorry if you don’t like it. There is nothing wrong with working in theory, just don’t get hung up on it. Do some practical experiments to see how much of a problem that theory says there will be and make an empirical judgment.

DJ

GadgetSmith, Its strange that it didn’t fix it, maybe because you are using the field fresnel in front of the LCD. Here are some pics of the effect I am talking about. They are of a 17” LCD plastic frame. Just a note, these are crappy pics and done very quickly. My fresnels and clear ohp sheet have been sitting around for quite some time collecting dust and getting scratched but it should show what I mean.

DJ

I think that i'm going to have to move my FF behind the LCD and try this method again. Today I plan to re-adjust the fresnels to see if anything will improve.

In your picture to the right, the very upper corners are dark. I assume that this is the type of thing to be looking for as we approach the limit on the FOV of the lens ? ...this I would think occurs just after the pincushioning that i've been seeing if I was to push the lens shift any further ?

Plenty of things to try.... I just need more time... Anyone got some for sale ?

gs

I just am triyng to understand the theory beind it (idealy) as a real world starting point. I agree that real world never is going to be as the ideal world, but since we are discussing the teory behind it, we need to work on the precision of the ideal world.

with the field deph excuse, what i was triyng to say is that you don´t need to tild the triplet because of the field deph doing this work for you. (But think in ideal world (field deph=0) we need to be as accurate a sposible there) But we are speaking about 2-3 degrees of diference anyway.