Question about collimated display systems.

Neilh has a good point about a WAC style collimating display being easier to make. Here's some hopefully useful information.

WAC stands for wide angle collimating display. The name is something of a misnomer today, but when introduced it was a step up on the competition.

Imagery is produced on a CRT mounted above an angled beamsplitter. Light reflects off the beamsplitter to the collimating mirror before finally getting to the viewpoint.

http://www.mycockpit.org/photopost/data/536/wac.jpg

Depending on design choices, this approach provides on the order of 50 degrees horizontal field of view and 35 degrees vertical field of view. Adjacent units, angled in toward the viewpoint can provide panoramic imagery.

The mirror has a spherical curve. It should be much easier to make (or buy) than the film mirrors used in cross-cockpit collimated displays.

Glass Mountain Optics makes large optical elements for simulators. Several years ago they advertised on their website that they made lower cost mirrors for museums and hobbyists. Since then they were acquired by FlightSafety International. I don't know if GMO still offers products to the general public, but maybe they do.

A few years back I ran across a posting about having a spherical section mirror made by a firm that makes the hallway safety mirrors. It was a special order, but all they did was put the reflective layer on the other side, i.e. making a concave mirror rather than their normal convex products. If I can find more details I post them later. [edit: I think this was the company: http://www.campuscrafts.com/]

US patent #3,549,803 "Virtual Image System for Training Simulator" describes a method for making spherical film mirrors using reflective Mylar, urethane foam and fiberglass.

US patent #4,822,155 "Mirror Assembly with Flexible Membrane" describes the construction details of a large diameter, spherical-section film mirror. University of Strathclyde built a number of these mirrors for their use in research projects. They use reflective Mylar. As Matt pointed out, Mylar is tough stuff that takes a lot of force to stretch. The method of mounting the film is an important factor in the optical quality of the finished mirror. This patent details one approach which has been shown to work well.


It may be possible to use sun control window film as a low cost beamsplitter. A more expensive option is available from Teleprompters.ws They sell properly coated glass beamsplitters with the partially reflective coating on one side and an anti-reflection coating on the other to reduce ghost images.

To avoid image distortion and other problems, the faceplate of the CRT should has a spherical shape of one half the radius (or slightly more) of the collimating mirror. It also needs to put out a lot of light. CRTs of this shape are no longer available. A rear projection screen is the contemporary replacement. If you project on-axis with the proper throw ratio, the majority of the projection-related distortion vanishes.

Because of the multiple reflections and the 50% beamsplitter, at most 25% of the light from the screen makes to the viewpoint. 10~15% is more likely. You need a lot of light. The old CRTs were special units pushing 50 to 100 foot Lamberts. A pico projector won't do, but a regular projector should.



I think the value in a collimated display is the promise for greater immersion. It makes the world outside the sim window look bigger, and adds realism. I think the cross-cockpit off-axis collimated system would be best because it provides an unobstructed view with considerable allowable head movement. The WAC system can provide a panoramic view by sticking multiple units side by side, but this leaves structure in view and limits allowable head movement. That said, the WAC approach may be the best, workable approach available at present.

Acrylic mirror material does not tolerate heat well. Once you get it hot enough to bend, the reflective coating is toast. I've tried.

Quote:

---

Originally Posted by Mike.Powell

A few years back I ran across a posting about having a spherical section mirror made by a firm that makes the hallway safety mirrors. It was a special order, but all they did was put the reflective layer on the other side, i.e. making a concave mirror rather than their normal convex products. If I can find more details I post them later. edit: I think this was the company: http://www.campuscrafts.com/

---

Yes - their rectangular convex mirror looks like a section of a sphere; get a 36" one of those and have them silver the other side, you might be in business. That'd create a WAC unit wide enough for one window of a B737 sim. Make the join directly at the edge of the mirror, assemble a few units together, and you'd have something useful. Not much vertical FOV, though.

I'll see if I can find someone similar in the UK. Got to be worth a try, even just as an experiment.

Please keep in mind, this works well with curved monitors not flat screen. Flat screen will cause blurry edges. Also, some CRT's do not work well when the screen is facing down.

Matt Olieman

True, but the standard replacement for the CRT in a WAC (or cross-cockpit display) is to use a projector + curved rear-projection screen as the picture source.

Alternatively, perhaps a correctly-focused collecting fresnel lens in front of a TFT could create the same optical effect as having a curved CRT screen.

I've been reading through several patents related to design, manufacture and installation of cross-cockpit collimated displays. A few things I've learned:

• The mirror is typically made from Mylar film (we knew that already)
• The film is held in shape by vacuum. It is supported only at the edges, and is not glued to any backing.
• The applied vacuum is typically dynamically controlled via a feedback loop and position sensors located behind the mirror.
• A certain amount of distortion will be present due to uneven stress within the mylar film. Most of this distortion occurs near the corners, and can be partially mitigated by methods described in several of the patents.
• While high-school level physics usually teaches that "the focal plane of a sphere is at R/2", this only applies for significantly on-axis rays. In a cross-cockpit display, the eyepoint is far off-axis, and aberration must be taken into account. Due to this, the radius of the projection screen is significantly larger than R/2.
• Due to the details of the optics, vertical FOV over 40 degrees is difficult to achieve. The vertical FOV distribution (ex. 20 deg up/ 20 deg down or 10 deg up/ 30 deg down) can be set by design choices..
• Horizontal FOV can be as wide as you like, limited only by the available width (length is practically unlimited) of the mylar needed for the large mirror.
• Mylar film in widths greater than 56" is hard to find.

Excellent article by Mike Powell on the home page regarding Collimated Displays.

Thanks Mike :) :) :)

Matt Olieman

Quote:

---

Originally Posted by wledzian

I've been reading through several patents related to design, manufacture and installation of cross-cockpit collimated displays. A few things I've learned:

• The mirror is typically made from Mylar film (we knew that already)
• The film is held in shape by vacuum. It is supported only at the edges, and is not glued to any backing.
• The applied vacuum is typically dynamically controlled via a feedback loop and position sensors located behind the mirror.
• A certain amount of distortion will be present due to uneven stress within the mylar film. Most of this distortion occurs near the corners, and can be partially mitigated by methods described in several of the patents.
• While high-school level physics usually teaches that "the focal plane of a sphere is at R/2", this only applies for significantly on-axis rays. In a cross-cockpit display, the eyepoint is far off-axis, and aberration must be taken into account. Due to this, the radius of the projection screen is significantly larger than R/2.
• Due to the details of the optics, vertical FOV over 40 degrees is difficult to achieve. The vertical FOV distribution (ex. 20 deg up/ 20 deg down or 10 deg up/ 30 deg down) can be set by design choices..
• Horizontal FOV can be as wide as you like, limited only by the available width (length is practically unlimited) of the mylar needed for the large mirror.
• Mylar film in widths greater than 56" is hard to find.

---

Excellent summary. In particular, you're spot on about the radius of the screen in an off-axis display being greater than R/2. I got my facts scrambled in the latest Mike's Tips article and will have to get with Matt to correct it.

wledzian,

Great post!

Can you post a link to the patent site that describe the methods for correcting corner distortion?

I'm also interested in the method of vacuum regulation via feedback. I wonder if this is critical for maintaining of the image on a flight on an ongoing on demand basis or for keeping the Mylar tension and thus the image from degrading over a longer time. If it is the later, I wonder if one can build in some mechanism to regulate tension, like tuners on a guitar. I imagine one would need to have an idea of the practical requirements for keeping that Mylar in tune and set these adjustments at the correct intervals.

Thanks,
Mike

Great posts from Mike and wledzian. I was wondering about that R/2 question as well.

The original video from the discovery segment seemed to show a projection screen that was (at least based on a visual observation) larger than the R/2 ratios and extended beyond the top of the cabin windows so as to avoid any shadowing problems.

If you go back to the basic physics of optics and collimated mirrors, anything inside the focal point will appear upright and larger than the object, in this case the projection screen, and will be a virtual image.

As Mike noted, the need to warp the image to handle the distortion requires aspheric lenses and those are not cheap, as in thousands of dollars. This is going to be a major hurdle.

JW

I've been using TurboCAD to do ray tracing. It's a little cumbersome, but I believe it to be accurate. I'm not using the paraxial approximations. Rather, I trace reflections based on the law of reflection (angle of incidence equals angle of relfection). The idea is that when a ray hits the mirror surface, I draw a line from the mirror center to the point where the ray intersects the mirror. I then draw the reflected ray to match the incident ray angle. It requires a bit of middle school geometry construction, but it appears to work. The results I'm getting are similar to the optical configuration in the 1982 NTIS report "Wide-Angle, Multiviewer, Infinity Display System".

For an off-axis system, the screen surface shifts closer to the mirror with the top tilting even closer. While a vertical cross section through the screen is still a circular arc, its center shifts toward the screen while the vertical axis of revolution stays with the mirror center, so the screen surface becomes toroidial or oblate (slightly squashed)

As far as image distortion goes, I'm hopeful that image warping software like Nthusim will be the solution. A requirement for color-corrected, aspheric projection lenses would make the project impractical.

A possibility is front projecting the screen using a second spherical-section mirror as a fold mirror. Possibly, this would resolve some or most of the projection distortion.

Just some pics you may find interesting: :)

Matt Olieman

http://mycockpit.org/images/collomating-mirror/B1sm.jpg
http://mycockpit.org/images/collomat...irror/B1sm.jpg[/IMG]
http://mycockpit.org/images/collomat...ing Layout.jpg
http://mycockpit.org/images/collomat...collimated.jpg
http://mycockpit.org/images/collomat...s retrofit.jpg
http://mycockpit.org/images/collomat...s_cut_away.jpg
http://mycockpit.org/images/collomat...X_Citation.jpg
http://mycockpit.org/images/collomat...m 3368f2-r.jpg
http://mycockpit.org/images/collomat...-46E_domes.jpg
http://mycockpit.org/images/collomat...I_OFT_dome.jpg
http://mycockpit.org/images/collomat.../NASA rfd7.jpg
http://mycockpit.org/images/collomat...ion System.gif

I don't have a list of the patents handy and time is short at the moment, but I'll pull the list together and post it tonight.
The patents don't give any real insight into the shape of the screen. Most of them state something along the lines of "this invention comprises a rear-projection screen of spherical shape, generally 1/2 the radius of the collimating mirror" or similar. Several patents even go so far as to specify non-spherical shapes (sections of ellipse or toroid), but continue to perpetuate the falsehood of a radius substantially half of the mirror.

I've also put together an Excel spreadsheet to play with geometry. In essence, I cast several sets of parallel rays from a prospective eyepoint, and let Excel do the math to see where each set converges. These points form the shape of the mirror surface. Ponts on this line will be collimated at infinity, at the selected eyepoint. Points on a surface of larger radius will appear to be closer to the pilot. As long as the virtual image is more than 30 feet out, monocular depth cues will dominate.

The shape of the screen is strongly defined by the chosen eyepoint and size of the spherical mirror.
The bottom edge (and therefore lower limit of vertical FOV) is limited by the highest sightline from the eyepoint. This is the limiting factor in the restricted vertical FOV.
The TOP of the screen has no such restrictions, and extending the top of the screen allows the pilot to lean forward and look upward. As long as the cockpit structure has a more restrictive lower vertical limit than the display, the pilot won't see the bottom edge of the mirror, and the illusion will be maintained.

Mike,

The image when the object is between the concave mirror and focus is a virtual image formed behind the mirror. Does the program allow you to extend the ray to a point behind the mirror where the image "exists"? Supoose you could do that manually but so much nicer if the program did it. ;-)

JW

"I've been reading through several patents related to design, manufacture and installation of cross-cockpit collimated displays. A few things I've learned:
• The bottom edge (and therefore lower limit of vertical FOV) is limited by the highest sightline from the eyepoint. This is the limiting factor in the restricted vertical FOV." from wledzian



Thanks again wledzian. I really appreciate your insights in this.

In regard to the mirror's bottom edge, I am trying to figure out the radius to use and also the lower edge location. I want to place the lower edge higher than the sim floor, perhaps a foot or so below the cockpit windows. This would set the bottom of the arc, or bottom of sphere above the cockpit floor. The rational is to have the mirror set close to the sim, as space is an issue in my room, while capturing a full range of view.

I have cut a couple of sections of arc, 5 and 6 feet respectively that are 30 degrees each. I want to get a general idea how much space will be used by placing these cut arcs next to the windows, look outside and make sure I don't see the borders.

Given the examples that I have seen on pro simulators, the mirror edge seems close or at cockpit floor level. I wonder if by raising the sphere vertically as I have described, I would be making the mirror unworkable. Best to know ahead of time before plunging in.

Few interesting links:

Matt,
How can I upload an image (jpeg) here?
Mike

http://www.q4services.com/links.php
http://www.simulation.org.uk/register.php

Quote:

---

Originally Posted by castle

Mike,

The image when the object is between the concave mirror and focus is a virtual image formed behind the mirror. Does the program allow you to extend the ray to a point behind the mirror where the image "exists"? Supoose you could do that manually but so much nicer if the program did it. ;-)

JW

---

TurboCAD allows line extension. Select the line and drag as far as you want. TurboCAD also provides a means to measure the angle between two line segments. It draws a circular arc centered on the intersection point even if the segments have not been drawn long enough to intersect. It's a quick and dirty way to see where the image falls, as well as, what the angular collimation error is.

This is undeniably clumsy, but it seems to work. I've been developing graphical techniques for locating the top and bottom edges of the mirror given the field of view requirements and viewpoint location. It also lets me know where the screen should be and what the shape is. I've been using the pro version of TurboCAD, but I expect the free version of DoubleCAD XT would work as well.

OSLO is a professional optical design tool available in a free student version. It's another possibility, but I haven't taken the time to learn how to use it.

Quote:

---

Originally Posted by mikesblack

Given the examples that I have seen on pro simulators, the mirror edge seems close or at cockpit floor level. I wonder if by raising the sphere vertically as I have described, I would be making the mirror unworkable. Best to know ahead of time before plunging in.

---

So much to say, so little time during lunch break...

Here's a quick sample from my raycasting simulation.

• The dark blue circular line is the mirror. In this case, radius is 48 inches.
• Magenta lines represent the cast rays. Vertical FOV is 10 deg up, 30 deg down. Eyepoint is located 20 inches below the center, offset 4" forward to represent the offset of my eyes from the vertical rotation axis of my head. For this simulation, I've used a parallel ray offset of 1". Due to the shape of the mirror generated by these parameters, this geometry may not be achievable for a full 180 degree horizontal FOV unless I can find a reasonable source of mylar film in larger widths.
• Black dots represent the screen for convergence at infinity. A screen surface inside these points will not properly form a virtual image when viewed from the design eyepoint.
• Blue dots represent the screen for convergence at 30 feet. A screen surface outside these points will produce a virtual image closer than 30 feet, partially defeating the purpose of collimation.
• The yellow line represents a spherical screen at the minimum radius required to form an image over the full field of view. As you can see, a screen of this shape would further restrict the achievable vertical FOV and would cause the lower half of the image to appear too close. It also serves to illustrate the oblate spheroid nature of the 'proper' screen shape.

http://www.mycockpit.org/photopost/data/500/raycast.JPG

Quote:

---

Originally Posted by mikesblack

Given the examples that I have seen on pro simulators, the mirror edge seems close or at cockpit floor level. I wonder if by raising the sphere vertically as I have described, I would be making the mirror unworkable. Best to know ahead of time before plunging in.

---

The commercial sims typically have a mirror radius between 9 and 11 feet, and have a much larger horizontal offset applied to the design eyepoint. If you provide a bit more information about your sim, I can crunch some numbers for you.

In designing my sim (still in the design phase), I started with the space I could get away with taking up, chose a field of view based on desired over-the-nose visibility in the pattern and on approach, and designed the cockpit interior based on the space remaining. My cockpit will be a single-pilot generic pit, as I'm finding that the bottom edge of the 48" mirror is driving the location of my instrument panel, and would not be usable for a cockpit sized for two pilots.

I won't be home until rather late tonight, I'll get that list of patents put together tomorrow morning.

Quote:

---

Originally Posted by mikesblack

Few interesting links:

Matt,
How can I upload an image (jpeg) here?
Mike

---

You can either attach it to your post or what would be better; post it in the Photo Gallery and link the image to it there. When you look at a image in the "Photo Gallery" you'll see near the bottom "Direct link:" That would be the link to your image.

Hope that helps. :)

Matt Oleiman

Thanks Matt. That does.

By the way, has anyone seen this ?

http://www.doti-optics.com/PERMA%20%C2%AE%20Mirror.htm

Thank you very much indeed. Your help and advice is most appreciated.

The following are the dimensions that I am working with.

Width of sim at widest point(rear) :11 feet.

Measurement from port rear of cockpit to wall: 23 inches.
Measurement from starboard rear of cockpit to wall: 18 inches.

Sim frame slopes in from rear to nose and so the above measurements are the most restrictive.

Height of sim 73 inches at rear and slopes to 57 inches at the apex of the front windows.
Height from top of sim to ceiling is 41 inches at rear.

Length from back of sim to front( measured at apex of front window) 68 inches.
Length from throttle location to apex of front window : 24 inches
and 48 inches to the most forward part of front window.

- Note the recent photos I just submitted. There is a rope that is attached to the throttle location and held out at 5 feet. The curve that I have in the photos is 30 degrees and 5 foot radius.
http://www.mycockpit.org/photopost/s...-21-10&cat=690

http://www.mycockpit.org/photopost/data/690/0231.jpg

Thought I posted this earlier, but don't see it. Apologies if this winds up as a double post.

In all the discussions on ray tracing and FOV are we actually considering what the pilot will see; i.e., where is the image, amount of distortion, image warping required, and where and how to actually position the mirror and projection screen?

Seems we need to also consider the two rules of reflection for concave mirrors:

Any incident ray traveling parallel to the principal axis on the way to the mirror will pass through the focal point upon reflection.
Any incident ray passing through the focal point on the way to the mirror will travel parallel to the principal axis upon reflection.

In order to do that we also need to specify the principle axis which sets the vertex of the mirror, relative position, orientation, and shape of the object (the screen), and location of the focal point which is on the principle axis at 1/2 the radius.

Need to think about it a bit more....

JW

In addition to the FOV another important calculation is the determination of the image formed by the screen and mirror geometry based on their relative positions to each other, the principal axis of the concave mirror that defines the vertex, the focal point, and radius of curvature.

Beat me up and call me crazy, but don't think the examples shown provide that solution. For that we need to apply the two rules of reflection for concave mirrors and incident rays;

* Any incident ray traveling parallel to the principal axis on the way to the mirror will pass through the focal point upon reflection.
* Any incident ray passing through the focal point on the way to the mirror will travel parallel to the principal axis upon reflection.

Performing that calculation(s) will help determine, in addition, the amount of image warping the software has to handle. Also assuming that the principal axis has to intersect the physical mirror and the horizontal axis will be symmetric for the entire 180 degrees and then one only needs to calculate the geometry for a single set of points on the screen in the vertical.

An important item to define is the vertex of the mirror which, in turn, lays out the principal axis and focal point.

This is really neat stuff. Kudos to mikesblack for starting the thread and to all who have contributed.

JW

Film collimating mirrors are spherical-section. Those rules are accurate only for parabolic mirrors or for very shallow spherical mirrors.

@mikesblack: (Thanks Mike.Powell for the link to the photos) A full-size 767 cockpit - color me green with envy. When I get to work on Monday, I'll pull up the 767 ops manual and check out the eyepoint locations and field of view definitions. I will say that fitting a wide collimated display around that may not be possible in such a tight space. In the meantime, I'll work with guesstimates and see what I can come up with. ...And I haven't forgotten about the patent list.

@Castle: Those two rules only apply to an optical component that actually has a true focal point. A sphere does not have a focal point. While parallel rays very near the axis will converge to a small region around R/2, this does not hold true for rays significantly off-axis. Rather than a focal point, a spherical mirror has a "caustic focal surface". Parallel rays which are close enough together, say the diameter of your pupil, will converge to a small region somewhere along that caustic surface.
http://upload.wikimedia.org/wikipedi...circle.svg.png

Rather than using the simplifying assumptions of ideal optics, my calculations do in fact calculate the reflected ray vectors for sets of parallel rays (from a hypothetical virtual image at infinity), and calculates the focal point to be the average of their intersections. I can also set a specific image distance for the ray sets so that the cast rays from the eyepoint converge, and generate a focal point for a virtual image at a non-infinite distance. Those calculations are used to generate the points for the limiting surfaces for images at infinity and 30 feet in my earlier post.

Dug out my old drafting set and went crazy trying to come up with some diagrams using the aforementioned reflection rules. Nothing seemed to work and what worked wasn't buildable or useful.

So my understanding is you are working from where you want the image to be and then defining the projection screen (location and position) based on the mrror size.

Thank you for the explanation.

JW

Some of the applicable patents:

Aspheric screen for collimated visual display apparatus
http://www.google.com/patents/about?...BAJ&dq=6944581

Method of constructing a thin film mirror
http://www.google.com/patents/about?...BAJ&dq=6758569

Apparatus for constructing a thin film mirror
http://www.google.com/patents/about?...BAJ&dq=6945659

Collimated visual display with elliptical front projection screen
http://www.google.com/patents/about?...&dq=12/273,053

Method and apparatus for modifying aircraft simulator wide-angle infitiy display equipment mirror to enlarge field of vision and for reskinning aircraft simulator spherical mirror cell to minimize mirror stress and distortion
http://www.google.com/patents/about?...BAJ&dq=7708561

Quote:

---

Originally Posted by castle

So my understanding is you are working from where you want the image to be and then defining the projection screen (location and position) based on the mrror size.

Thank you for the explanation.

JW

---

You're welcome.
I'm doing exactly that, point-by-point, for enough points to sufficiently define the screen shape.

@Mikesblack:
I took a stab at putting a mirror around your cockpit. I figure you could shoehorn an 84" radius mirror in the space you described, so I started there. I located the eyepoint in a freeware model of a 767 in Sketchup, and set the mirror in place to get a first-shot eyepoint. This first try gives a 180° horizontal FOV and nearly 75 degrees vertical. The mirror geometry intersects the aft windows, so I couldn't swing the HFOV wider.
http://www.mycockpit.org/photopost/d...irrorStudy.JPG
http://www.mycockpit.org/photopost/d...y_firsttry.jpg

If you're willing to sacrifice some of your vertical FOV (a permanently installed "sunshade" should suffice), you can move the whole mirror assembly aft and get the HFOV to cover all the windows. I haven't worked up the geometry for that yet.

So far, that covers the "is it possible" aspect. There are, however, several large roadblocks:

• The forward structure of the cockpit below the windows interferes with the lower mirror edge. This makes mounting of the mylar challenging.
• The mirror would require a single piece of mylar sheet over 13 feet wide and almost 30 feet long. I haven't been able to find a supplier willing to sell me less than a full production roll for sizes over 56" wide.
• The screen itself is over 10 feet wide, posing its own challenges simply finding raw material for forming.

I can play with mirror placement if you'd like, but the long and short of it is that the optics for a jumbo jet simulator are going to be a bit large to fit in a bedroom.

There are a few other issues to keep in mind as well.

Mylar is elastic, but with a high spring constant. To stretch it, you must pull hard, and it will pull back. With a large mirror, you have a lot of Mylar putting stress on the frame. You need a structure capable of accurately maintaining its shape while supporting perhaps several hundreds pounds force.

Mylar is elastic up to a point then it deforms permanently. If the mirror is too deep, the Mylar must be stretched more and is more likely to be pushed outside its elastic region. When this happens, it won't happen uniformly across the film. Over-stretched Mylar has different properties than Mylar that has not. The distribution of tension across the Mylar is very much not uniform, and the shape is no longer spherical. The optical properties of the mirror are damaged. Much of the recent work on film mirror seems to have been focused on dealing with this.

That's good info Mike. I'll add a strain estimate to my spreadsheet to display a warning if the strain is outside the elastic region of Mylar's stress/strain curve.

Wow,
That is the most awesome thing you have done. Thank you very much for that. The diagrams are great.

I will need to study the diagrams for a while to get a better understanding and will reply in time.

Also, those patents are most helpful and have provided me with good sense for the construction. I will need to study these further and then will most likely experiment with smaller sections.

Again, most appreciated.

Mike

Hi,
didn't see this thread before just now, so I'm still reading through.
I have experimented a lot with plexiglas and mylar to make collimated displays. Some year ago I also made a little beamsplitter visual with a normal plexiglas as beamsplitter and a normal mirror on the backside. That works when it's dark enough in the room. offcourse it will work bether with a real beamsplitter. The best company I found for such mirrors in europe is seco-sign. You can choose how much the mirror should reflect from 5 to approx 20%.

I have also asked several plexiglas specialists here in swizerland if they could form such a spherical shape. The closest I got was one company that makes "bubble-windows" of plexiglass. But they can make max 1m diameters. So to small for a visual. They told me that to make a shape in the magnitude for such a visual plexiglas would be around CHF 20'000.- (that's currently also about $20'000.-) That's just for the form. Mirror coating the formed plexiglas will probalby add another $ 5'000 to 8'000. Each.

So a spherical collimated visual will probably be to expensive or to technical difficult for me. Even just to make the backprojection screen spherical.
What I have started to experiment with some time ago is to make the "collimated" display conical shaped instead of spherical. It's much easier to shape. No heat needed. For the plexiglas mirror and the backprojection screen.
I use an opal white 3mm thik, with 90% light admission as a backprojection screen. The mirror is also 3 mm thik.
http://www.md80.ch/index.php?option=...lator&Itemid=1
This setup is still very experimental. Important is that, this will have a horisontal disortion that needs to be corrected. I use NEC NP901w projectors. They have a software to correct geometry dissortions directly in the projector. So no need for extra CPU consuming software on the pc.

I was wondering, has any one ever asked for prices at those companies like q4services or glassmountain ?
Especially glassmountain always stess their low-costnes. But I guess that very relative. But even if just the backprojection screen would be in a payaple price range would help a lot for a real collimated setup. http://www.glassmountain.com/Simulat...epointDisplays

Cheers,
Gery

wledzian and I got some good work done on Saturday. We're getting closer. :)

I'm going to try to get the framework I have finished off with interior supports and a sealed back this coming week or so.

g.

Quote:

---

Originally Posted by crashdog

So a spherical collimated visual will probably be to expensive or to technical difficult for me. Even just to make the backprojection screen spherical.
What I have started to experiment with some time ago is to make the "collimated" display conical shaped instead of spherical. It's much easier to shape. No heat needed. For the plexiglas mirror and the backprojection screen.
I use an opal white 3mm thik, with 90% light admission as a backprojection screen. The mirror is also 3 mm thik.
http://www.md80.ch/index.php?option=...lator&Itemid=1
This setup is still very experimental. Important is that, this will have a horisontal disortion that needs to be corrected. I use NEC NP901w projectors. They have a software to correct geometry dissortions directly in the projector. So no need for extra CPU consuming software on the pc.

---

So a non-spherical section mirror with correction does work - or at least provides enough focal distance to trick the eye into seeing through the plane. A conical section mirror like that is easy enough to do with acrylic mirror sheet. Fantastic that someone has already done it!

Looks like this may be the way to go with my project - my limited front projection room becomes much less of an issue and I could probably get 140 degrees out of my 16:9 projector.

It'll be fun experimenting :)

56" Mylar would provide a 45 degree FOV for a 6 foot mirror --- (45/57.2957)*72.0 = 56.548

OK, so we're a half inch short ;-) but close enough. You can get that size in very large lengths.

A 7 foot system would provide ~ 40 degrees vertical FOV. Still not bad for a DIY and you can't beat the price.

@wledzian. I would like to try and equip my 747 with a six foot mirror. What would you need from me in the way of data to crunch some numbers. Fortunately, have a little more room and a single available stall in my garage shop, but 6.5 radius is the upper limit and 6 leaves a little wiggle room.

Nice work.

JW

Here is a simple graphic approach to designing and analyzing an off-axis collimated display.

A key step is tracing a ray path as it is reflected by a spherical surface. This draws on some middle school geometry and works well with drawing packages like TurboCAD that have line and circle tools with "snap to intersection" options.

1 - Draw a line from the mirror center to the intersection point of the incident ray with the mirror.
2 - Draw a circle centered on that intersection point. The radius is doesn't matter.
3 - Draw a second circle centered on the intersection of the first circle arc and the line from the mirror center. Set the radius at the intersection between the incident ray and the first circle arc.
4 - Draw the reflected ray from the intersection of the mirror surface and incident ray to the intersection of the arcs of the two circles opposite the incident ray.

http://www.mycockpit.org/photopost/data/536/reflect.jpg

Laying out the positions of the collimating mirror, screen, and viewpoint starts with estimates. Because this is an off-axis system we know that the screen will be larger that the textbook, on-axis figure of half the mirror radius. We'll start with 65% of the mirror radius. We'll place the viewpoint directly under the mirror center of curvature down 50% the mirror radius. The viewpoint doesn't have to be directly under the center, but this position provides the most symmetrical viewing for a single person system, and presumably the smallest geometrical image distortion as you turn your head.

1 - Draw circular arcs representing the mirror and screen surfaces.
2 - Draw lines from the viewpoint representing the upper and lower limits on the vertical field of view.
3 - Reflect the lower field of view up to the screen surface. This is where the lower edge of the screen needs to be for the current estimated position of the viewpoint.
4 - If the screen edge is below the upper field of view limit, the viewpoint is too high. Move the viewpoint lower and try again. If the screen edge is far above the upper limit, move the viewpoint higher and try again. If there is enough space between the lower screen edge and upper FoV limit for the screen framework you haven't designed yet, proceed to step 5.
5 - Reflect the upper FoV limit to the screen. If the reflection misses the screen, the upper FoV may not be possible. Before giving up in dispair, move on to screen placement refinement.

http://www.mycockpit.org/photopost/d...ollimated1.jpg

The screen shape and placement is an estimate. Almost certainly you'll want to tweak it. Do this by using parallel test rays that bracket the viewpoint. Reflect them off the mirror to the screen. Their intersection point should be on the screen surface.

1 - Draw two parallel horizontal lines from the viewpoint to the mirror arc. One line should be above the viewpoint, the other below.
2 - Reflect both lines from the mirror. Extend the lines until they intersect.
3 - Draw another pair of parallel lines bracketing the viewpoint to the mirror. These lines should angle downward at the lower FoV limit. Reflect and intersect.
4 - Repeat the process with lines angled at the upper FoV limit.
5 - These three intersection points define a circular arc which is now a very good estimate to where the screen surface should be. The center of this arc is unlikely to be the same as the center of curvature of the mirror. The new screen center will probably be closer to the mirror surface and higher than before. The net effect is to produce a screen that bulges out more and has a top that leans in toward the mirror.

http://www.mycockpit.org/photopost/d...ollimated2.jpg

You will probably need to make several passes through the whole process to converge on a configuration that meets your needs.

Once you have a workable configuration, you can ray trace diverging paths from multiple points on the screen surface to check for collimation accuracy.
I used TurboCAD, but I think you can do the same thing with DoubleCAD XT, a free program.

@castle -
Could you provide a little more info regarding the numbers in that formula? Do you want a wraparound section of a cone? It's the flattening of the conic surface that pushes an otherwise workable mirror height past the available material width. If the conic surface works well, the issue of a monolithic surface may be moot, as joined gores would suffice without introducing issues of distortion due to uneven stress and potential weak points at the seams.

@crashdog -
I'm intrigued by your use of a conic surface mirror and screen. That would certainly be easier to fabricate than a vacuum-formed spherical surface. The horizontal movement of the camera seems to show a distant image, but I'm concerned that the optical properties of a conic surface would introduce significant astigmatism to the image. How does the effect hold up for vertical movement?

@geneb -
Thanks for having me over on Saturday. I have a sense that as many questions were raised as were answered, but it feels like there's some progress. If we can mange to get a mirror formed to your existing frame, it should work well for a 30 degree VFOV with a couple inches margin left over at the top and bottom to take up any residual distortion in the material (3 inch horiz offset, 18 inch vert offset, 75° top angle, 105° bottom angle, convergence at 30 feet). I have a feeling that the following two patents together may offer a solution for mounting the mylar without having to stretch it in the process.
http://www.google.com/patents/about?...BAJ&dq=6050692
http://www.google.com/patents/about?...BAJ&dq=6945659

I don't have any math. figures at the moment. The impression I got from my experimental setup is that vertical movement is no problem as long as the angle of the backprojection screen is the same as the angle of the mirror. But the setup has to be dainly precise. Otherweise the picture looks really weird. But I guess the seame issue would be for a spherical setup.

Gery