The mirror based collimated display, in a variety of forms, is the mainstay of large commercial simulators. In their most compact incarnation a single monitor is located above and forward of the pilot's position and is pointed down. The image is reflected away from the pilot by a partially reflective "beam splitter". (Basically, a partially silvered mirror that reflects about half the light, and allows the remaining to pass through.) The image is then reflected back toward the pilot by a large concave spherical-section mirror. This enlarges the image and makes it appear to be generated at a great distance from the pilot. Hence, stereopsis cannot generate distractive miscues. Glass Mountain Optics, CAE and SEOS Displays, Ltd. make reflective collimated displays and components. Glass Mountain Optics also sells to the hobbyist!
For theoretically perfect performance the mirror should be parabolic; however, these are quite expensive to make in suitable sizes so generally spherical mirrors are used. The distortion introduced by using a spherical mirror only appears off the central axis of the mirror. Because these systems are generally used with the pilot's head situated along or close to this axis, using a spherical mirror is a good tradeoff.
The monitor is placed so that the screen is at the focus of the mirror. The optical path from the center of the mirror to the center of the screen is on the order of three feet. Because the focal length of a spherical mirror is one half the radius of curvature, this makes the mirror's radius on the order of six feet. For best performance the monitor's screen should be a spherical surface with a radius of curvature equal to one half the radius of the mirror. This is bad news as CRT monitors are being manufactured with increasingly flatter screens.
This move toward manufacturing flat screen displays has created a small market for replacing obsolete spherical faceplate CRT displays used in military simulators. Diamond Visionics developed a rear projection system based on the TI moving mirror light modulator. The light modulator projects the image onto a spherical screen. The whole system is packaged like a CRT monitor. By today's standards it looks like the front bulges way out.
Collimated displays of this sort can be placed side by side to expand the width of the field of view. The mirrors enlarge the apparent size of the monitors. This causes the perceived images to overlap even though the monitors don't physically touch each other. With careful positioning of the monitors the adjacent images are moved into alignment creating a single blended image.
While you could make a very wide view display by positioning several of these monitor-based displays side by side, the big boys seem to have left this approach behind. Instead, a large, curved screen is mounted above the cabin and the exterior view is rear projected onto it. There will likely be several projectors carefully aligned such that the individual projected images blend into a single panoramic scene. This is viewed as a reflection from a very large curved mirror that surrounds the cabin and fills the entire field of view through any of the cabin windows. The mirror curvature is a type of spherical section and acts to collimate the image. As you can see from the diagram, the light paths into and from the mirror are somewhat off the mirror's central axis, and this might be expected to result in image distortion due to spherical aberration. In fact it does, but the spherical curve of the projection screen is designed to compensate for it. This approach delivers an impressive collimated, panoramic view. Unlike the smaller monitor based systems with their simpler mirrors, this approach allows for considerable latitude in possible viewing positions.
Collimating mirrors have been made from a variety of materials. Polished stainless steel, glass, acrylic, and thin films are a few of the common choices. Material selection depends upon many factors including size, complexity of shape, weight constraints, and required robustness. Given the large size, complex shape and a desire to minimize weight on the motion base, thin film is an excellent choice for simulator mirrors. Surprisingly, this is not new technology. Thin film mirrors have been around for more than 40 years.
While making a large, thin film mirror with a complex shape and good optical properties can be quite a challenge, making a simple thin film, spherical mirror is quite easy. Cover a circular hole with a piece of aluminized Mylar (or aluminised Melinex), pull a small vacuum behind it, and voilá, instant spherical mirror. Unfortunately it doesn't take long to discover that it isn't a circular mirror that you want.
You want a square mirror, or a rectangular one, or a trapezoidal one. You want a mirror that fills the field of view out all windows without having massive mirror area that cannot be seen. But you can't simply cover a square, or rectangular, or trapezoidal hole with a chunk of Mylar and end up with a spherical-section mirror.
The main issue is that the frame that holds the film must lie on the surface of the desired mirror shape. So, if you want a spherical-section mirror, the supporting frame must be curved as though it were draped over a sphere. The reason that a spherical-section mirror with a circular frame is so easy to make is that the frame just happens to be flat. No such luck when you want different shapes.
A secondary issue is establishing the proper tension in the thin film to accurately form the desired mirror shape. This creates yet another challenge when the frame has a complex shape. When the mirror is complete, the film is slightly stretched by the partial vacuum behind it. It is in this stretched condition that the film should be clamped to the frame if the proper shape is to be maintained. Proper assembly requires slightly stretching the film's edges as they are fastened to the frame. Just how must stretch and in which direction is the challenge. And those depend upon the specific shape and complexity of the mirror's frame.
Here's a video from the Science Channel program, How's It's Made. It describes the construction of a commercial flight simulator. You'll get a brief view of the collimating mirror abour mid way.
Given the simplicity of circular spherical-section mirrors and the low cost of aluminized Mylar (~US$0.45 per square foot), mirror based collimated display systems should be great projects for the flight sim hobbyist. At least on the west coast of the U.S., TAP Plastics is a source of the reflective Mylar.
Other sources for reflective Mylar, surprisingly, are suppliers to plant nurseries. The film is used to maximize the amount of light on the plants' leaves. One such firm carrying reflective Mylar is Nielsen Enterprises.
The pressure differential required for a 20-inch mirror is on the order of one psi, well within the capability of an ordinary vacuum cleaner. You may find that the real challenge is controlling the shape of the mirror. Even with a perfect seal (which you can't realistically expect) you will still see small transpiration through the film itself. Without an active shape control system the mirror will continually be going flat.
Learning more about mirror based collimated displays should probably start with some introductory geometrical optics. Try Edmund Industrial Optics for an on-line "Optics Primer". After that a trip to the library is in order. If you can't find a book specifically on geometrical or ray-tracing optics, thumb through an undergraduate college physics textbook. Generally, simple optics is included and it's too early in the college curriculum for the math to be too intimidating. (For example, try The Feynman Lectures on Physics, volume 1.) It's very doubtful that you'll find a book specifically on mirror based collimated displays; the optics is just too simple. The real complexity is in the details of implementing those optics. This is one of those cases where the devil really is in the details. However, if you check out the websites of the manufacturers, primarily SEOS, Ltd., CAE and Glass Mountain Optics, Inc., you should find at least some basic descriptions in their product literature.
There are several papers that address making membrane mirrors. An early one is "Aluminised Mylar as a Flux Collector" by Maurice Gavin. Originally published in the May, 1979 issue of Sky & Telescope, it is now available on the author's web site. Another is "The Varifocal Membrane Mirror" by L. Van Warren, also available on the web.
There has been some serious research at the University of Strathclyde into the use of large film mirrors to display 3D images. Two papers addressing their work, which are well worth reading, are, "Membrane Mirror Based Display for Viewing 2D and 3D Images", and "Stereoscopic Display Using a 1.2-M Diameter Stretchable Membrane Mirror", by Stuart McKay, Steven Mason, Leslie Mair, Peter Waddell and Simon Fraser. (My thanks to Scott Ashburn for bringing this to my attention.)
For information on more complex thin film mirrors you'll have to check out some patents. The basic patent for a "vacuum formed flex mirror" (pat. # 2,952,189) was issued in 1960. There are no surprises in it, but to see how mirrors with more complex shapes are made you might take a look at these: #3,973,834 "Mirrors having stretched reflective sheet materials and method and apparatus for their production", #4,592,717 "End retract device for completing spherically shaped reflective film" and #6,050,692 "Method of constructing a thin film mirror". These are U.S. patents available online from the Google patent search.