Question about collimated display systems.

[wledzian]

Salire, the increase in immersion really is one of those things that has to be experienced to be believed.

When the mirror is drawn down just enough to get rid of most of the wrinkles, an image is formed at a little more than the total path length - that is to say, if the mirror is 4 feet from your eyes and the screen is 2 feet from the mirror, the image feels like it's a little more than 6 feet away or so. This is probably similar in feel to a large 'normal' screen.

As the mirror draws down to its proper position, the image formed becomes far away, and it feels far away. It no longer feels like you're looking at a screen; Your eyes relax and focus off at a distance, as if you were looking out a window. If you've built a hardware cockpit, you'll find yourself moving your head to look around things.

Once the image is far away, you no longer have binocular cues to tell you what is near and far; other cues take over, primarily perspective and contrast. Without the binocular cues screaming "MONITOR THREE FEET AWAY!", your brain can pay attention to the other cues, and the image actually feels 3-D.

In addition to the feel of the distant image, the seamless wraparound filling your peripheral vision adds alot - you really feel like you're moving, even though you're sitting completely still. I've piloted demo flights for several people; one had to quit because she started getting motion-sick, and two almost fell out of their seat.


EDIT to capture your post made while I was typing:
X-Plane doesn't do multiple visual windows on a single computer. To get past ~120, you'll need multiple computers.

[Salire]

Thanks Wayne. Yes, I think I'm sold. I think the Citation sim I had a chance to fly years ago at AA training center near DFW had a WAC type display and I was bowled over.

I haven't tried this yet, but I *think* I could use a TH2GO with X-plane and stretch the horizontal FOV out to 180. I've tried this with an aux monitor on my iMac i7 and set it to 120, and it worked fine. As long as the computer can keep up I would think it would work without having to go to a multiple pc config (and the ridiculous commercial X-plane license). I'm probably going to build a PC to run the sim/visuals, so I'm not totally wedded to X-plane, but that's where I'm starting.

S

[wledzian]

X-plane doesn't do 180°. It's not a matter of stretching.

Like just about every other graphics engine out there, the visuals in X-Plane are generated with a perspective frustum - picture a pyramid drawn as lines from your eyes to an imaginary rectangle in front of you. To go wider you make the rectangle bigger and/or move it closer to the eyepoint. No matter how big/close it gets, you cannot achieve 180°. Also, as you approach this limit, the detail becomes very compressed towards the center and the edges get very stretched. Anything beyond ~120° becomes nearly unusable.

We get around this limit in FSX by using multiple windows. Gene and I are using three windows of ~80° each, to allow for overlap and field loss due to the warping. As X-plane does not support the use of multiple windows, you're limited to what is reasonable to produce in a single window

While the X-Plane manual implies that extreme wide views are possible (cylindrical / spherical projection modes), these are post-render effects which simply warp the perspective frustum after the fact. Due to the distortion required, the edges of the image in these modes are fairly sharp, but the center gets very blurry, especially at wide zooms, and is still not capable of achieving or exceeding 180°.

[Salire]

Ah, ok. The pyramid illustration made it clear. Thanks. This would explain why trying ~120° on 2 monitors seemed pretty good as a test. Would you say that ~80° is close to the limit you'd go for each?

[wledzian]

For our display, that's about as wide as we'd go. When you're dealing with front-projected screens, there comes a point where the curvature of the screen starts to limit the width you can use. If you're using a rear-projected screen, however, you don't have such limitations. If you can project to it in focus, you're good to go.

[Salire]

Did you build basically to the limit of the available mylar width to still give you 40° vertical?
In thinking of the variables one has to consider, I was wondering what the relationships are between distances between viewers, screen, and mirror and how they affect, for lack of a proper term, an acceptable viewing zone (perhaps also an oblate spheroid). From your ray drawings, I understand that the screen/mirror distance is determined by ray convergence and the specific shape of the oblate spheroid of the screen are driven by the viewer offset. I suppose then that the acceptable viewing zone would be a function of the mirror radius and the degree of viewer offset from the axis?

BTW, what software did you use for the ray diagrams?

[wledzian]

you are largely correct in your understanding of screen geometry generation. our design basically assumed a single viewer at the display axis, offset a bit because your eyes are in front of your neck. Due to the way the ideal screen shape changes with offset, a real multi-viewer cross-cockpit display cannot be fully collimated at all points. The screen shape will be determined by a surface which will appear distant to both viewers, but will not be behind the ideal surface for either viewer. So long as the image forms at a sufficient distance to negate stereopsis, it will work.

Regarding our design size, it was a grand confluence of coincidence. I had a design worked out for 15° up, 25° down with a 48" radius, but Gene had already begun cutting parts for +/- 20°, also at 48" radius. I did the raytrace and determined that it would work, albeit with a bit of crop at the top from the bottom screen edge. We decided that this would be acceptable, as it would allow the additional immersion of leaning forward and looking up at otherwise-hidden sky. 40° vertical was chosen as the largest vertical FOV that we could get with a centrally-located pilot, mainly due to the ~2% maximum stretch of mylar at the peak of the stress/strain curve. Pull more than that, and the mirror deforms uncontrollably and pops like an overinflated bubblegum bubble. 210° horizontal was just going to 180° by design, with about 15° usable in each 'ear'. It also happened to coincidentally be exactly the limit of what we could get out of the 56" mylar that was available.

The software for the ray diagrams is a spreadsheet in Excel, entirely of my own creation. At some point, I may convert it to a stand-alone app coded in VB.Net, but I'm not planning to release the Excel version.

[Salire]

Having done this, what would you do differently in the next implementation? What would you say is the optimal vertical FOV centerpoint? I can see the benefit of having more sky visible for effect (when leaning forward), but is it worth it to restrict the view of the ground when looking out the side window?

How large would you say is the zone where you get an acceptable image? Are there disorienting effects by moving your head? Is it more sensitive to distortion in one axis? I'm guessing it would be most sensitive to vertical head movement.

Thanks,

Sal

[Salire]

I put together a spreadsheet and played with various configurations. It seems that for a couple of reasons, the ±20º FOV worked well for you. I had a lot of trouble arranging a cockpit to work with the 15/25 and 20/20 is more in line with FAA cockpit design guidelines (see FAA AC25.773-1), at least for transport category airplanes.

I understand your mylar stretch constraint, but this appears, at least in the 48" radius design, to have limited your effective vertical viewing area of the mylar to be about 36" of the width (at a 40º FOV). It's a shame that you can't use more. Don't you have about 48" or so available after accounting for a margin to tape the mylar to your 'goggles'? I haven't worked out the geometry yet, but perhaps you can use more of the mylar as the radius increases?

Another issue I had and thinking about the 48" radius was that the proximity of the mirror comes into to conflict with placement of an instrument panel/glare shield if you use a typical eye to instrument panel distance of ~30".

S

Collimated 20-20.001.jpg

[wledzian]

You are correct - the bottom edge of the mirror does start to intrude on the panel structure if you want the panel that far out. The entire design becomes an exercise in compromise.

When the mylar is fitted to the frame, it is initially in a conic shape. This conic shape unrolled requires more mylar width than the mirror chord length might indicate. You can go bigger if you limit your horizontal FOV, but for the configuration we've used, the design requires a full 56" width.