One VERY exciting software program

[andyb]

Good point magicaldr!

I've just been playing with a system we've just put together .... just a single 1280x768 projector, on a 130" screen.



If you sit at the center of curvature, then you'll get 100 degrees horizontal field of view.

[BHawthorne]

Good point magicaldr!

I've just been playing with a system we've just put together .... just a single 1280x768 projector, on a 130" screen.

If you sit at the center of curvature, then you'll get 100 degrees horizontal field of view.

Nice! I have a piece of MDF white glossy 4x8 ft paneling (108" diagonal) I want to try this on. Bought it over a year ago at Home Depot for a few bucks to experiment with. It should prove sufficiently "bendy" enough for me to make a circular screen area. The glossiness of it might give me some bad hot spots in the projection though. Are there any noticeable focus issues projecting onto a curvature? I'm assuming you just focus it to whatever looks sharpest.

[andyb]

There aren't too many specific issues when projecting onto curved surfaces (apart from the obvious geometry correction). Most projectors have pretty good focus ranges, so focus isn't a particular issue. You may need to spend a couple of extra seconds mid-positioning focus to ensure it's even over the entire surface, but that's about it.

The glossiness of screen material might be a bit more noticeable though. For home systems, I use a matte white (ceiling paint) with a small amount of talcum powder ..... a little secret to improve the texture on the surface. You can also add a tough of grey to improve contrast if needed.

Nice! I have a piece of MDF white glossy 4x8 ft paneling (108" diagonal) I want to try this on. Bought it over a year ago at Home Depot for a few bucks to experiment with. It should prove sufficiently "bendy" enough for me to make a circular screen area. The glossiness of it might give me some bad hot spots in the projection though. Are there any noticeable focus issues projecting onto a curvature? I'm assuming you just focus it to whatever looks sharpest.

[Trevor Hale]

Hi Andy,

Forgive me but I am very interested in your test image above. If you were to use your software and add keystone correction to the image above, essentially bringing the top left and top right corners closer to the field of view, how does your software give you the added scree depth without distorting or stretching the image forward to the corners? (something from nothing?) Maybe I am not quite understanding the possibilities of your software but I really find this interesting.

Will your software work on a rear projected image with the same result with a convex curve rather then a concave curve?

Again, thanks for taking the time to post here.

Best regards,

Trev
Mycockpit Admin.

[andyb]

Hi Trevor,

If you're using Sol7 to control the distortion correction, then there's no real need to use the projector's keystone .... all of this can be done using Sol7.

Because we curve the screen towards the viewer in the pic, the projector will over scan the screen at the edges and will also introduce some for of barrel distortion for front projection - the exact distortion will depend upon the projector's lens characteristics.

We then take the original image and remap this to match the screen size and shape. Therefore, we effectively resize the original image so that it's correctly aligned onto the curved screen.

This process supports whole host of screen shapes including cylindrical and spherical etc, as well as concave and convex (front pro / rear pro) set-ups.

So, the something for nothing part relates to the projector's pixels that over scan the screen. These are effectively lost (not used in the distortion map).

As an example below I've included an image of an exaggerated Sol7 config. The black areas at the sides and top would be the area of the projector that would be lost in pixels. The map is intended to highlight projecting onto a spherical screen from a side offset angle.



The loss of some pixels in this process is a general issue of geometric correction with all fixed matrix panel projectors (DLP, LCD, DiLA etc). Generally the numbers are not very significant.

I hope this makes some sort of sense

[Trevor Hale]

It makes lots of sense. Thanks for taking the time to respond. I am just interested in the concept. So by altering the image prior to the projection stage your not distorting the image after the projection your rearranging it prior to the projection allowing the image to end up asymmetrical once on the screen. In essence adjusting the image perfect on the curved screen then if you were to project the readjusted image to a flat screen would look very strange indeed.

Your exaggerated picture helps me understand.

I currently use a single rear projected image onto a curved screen facing the pilots. (needed to do this to get the image size I required) I get a little stretching of the image in the high corners at the top corners of my screen, as the image rolls off the curve. I gather your software would reduce this stretching effect.

If I could convince the wife, it may be a great idea.

Best regards,

Trev

[BHawthorne]

Yep, in order to distort it to look correct on the curved screen geometry you do lose some pixels, but the tradeoff looks very impressive. This is some of the stuff I've been reading up on lately:
http://local.wasp.uwa.edu.au/~pbourke/projection/

I especially want to implement a variation of iSphere. Can I accomplish this in a 270 degree 3 projector setup (left, front, right only, no back) using Sol7? This would be pretty ideal for a fighter cockpit setup. I have the projectors, just would need to build the screen and setup the software to distort to those specs. Does projection off of a spherical security mirror require a different type of image distortion or can Sol7 also do that or be made to do that?
http://local.wasp.uwa.edu.au/~pbourk...ction/isphere/

My plan would be to implement the screen as a modular design (a screen module would be 22.5 degrees per panel) made of PVC piping and screen material from eBay. Both would be affordable materials. Another alternative would be I would manual layup the panels as fiberglass composite, as I have the knowledge to do that too. Fiberglass tooling for a mold might be the way to go on this for me. I make 1 tooling mold and then wet layup all the panels to the common mold. It would give me consistency on the panels that way. Although that would probably make it a lot less portable. I'll end up doing probably what costs the least to accomplish.


Edit: Added an old pic I had laying around I made in Illustrator awhile back.

[Bob Reed]

Yep, in order to distort it to look correct on the curved screen geometry you do lose some pixels, but the tradeoff looks very impressive. This is some of the stuff I've been reading up on lately:
http://local.wasp.uwa.edu.au/~pbourke/projection/

I especially want to implement a variation of iSphere. Can I accomplish this in a 270 degree 3 projector setup (left, front, right only, no back) using Sol7? This would be pretty ideal for a fighter cockpit setup. I have the projectors, just would need to build the screen and setup the software to distort to those specs. Does projection off of a spherical security mirror require a different type of image distortion or can Sol7 also do that or be made to do that?
http://local.wasp.uwa.edu.au/~pbourke/projection/isphere/

A security mirror? Have you tried this? I think they are far too imperfect for this and from what I have read would be way too much curve. But I know very little in this area.....

[BHawthorne]

A security mirror? Have you tried this? I think they are far too imperfect for this and from what I have read would be way too much curve. But I know very little in this area.....

In the links I gave he's implemented most of those ideas in some form. I've not seen any mention of the security mirror impairing the projection enough to note. I realize that first surface mirrors are preferred. What works, what is preferred and what is most cost effective are 3 different things though. I think this is one of those situations where I just have to go and implement it and see how it turns out. A security mirror is around $40. A bunch of PVC piping is around $50. I could go to a cloth store and pick up a bunch of black-out curtain cloth to be cut and sewn to use as screen material to test it an affordable way.

Only thing stopping me is the real-time distortion software. It would be ideal if Sol7 can do this.

I'm not afraid of doing silly experiments. Some that I've done has turned out pretty cool.

Edit: I found this info:

Where can I source a spherical mirror from?

If you want a standard spherical mirror do a google search for keywords "safety mirror" or "security mirror", you will find lots. Refine the search to your country domain and hopefully there will be a local importer. I suggest 60cm size [See question on mirror size]. These mirrors will not be first surface, that is, the mirror coating will be on the back of the mirror. The above will work for testing but the quality will be severely compromised, due to multiple light passes throught the clear substrate and refraction.

At the moment the best source of first surface mirrors is the following
Acril Convex Co Pty
Email: sales@acrilconvex.com
WWW: http://www.acrilconvex.com/halfdome.htm

The standard mirror I recommend is product 18535 HDM.600 60cm acrylic half dome.
The first surface mirror is product: 20534hB polymer coated half dome mirror w/first coat. It doesn't seem it is on their web page yet, so email them, say you were speaking to me and ask for the polymer coated first surface half dome mirror in 60cm size.

I normally suggest people get a standard (not first surface coated) mirror first from a local supplier (try to get the same 60cm version) and then once they are familiar with the technology buy the better mirror and reap the quality improvements.

There are various other companies looking at creating coating for the same acrylic mirrors but to date the ones I have seen are inferior.

http://local.wasp.uwa.edu.au/~pbourk...irror/faq.html

[BHawthorne]

It might just be easiest for me to go this route with Sol7:
http://local.wasp.uwa.edu.au/~pbourk...tion/cylinder/
http://local.wasp.uwa.edu.au/~pbourk...ion/cylmapper/