We don’t recall having heard the term ‘collimated display’ before, but we’ve seem them in action. These are mirrored projection display that give the viewer a true peripheral vision experience thanks to well-designed optics. Here is a project that [Rob] and [Wayne] have put a ton of time into. It’s their own version of a DIY collimated display that uses a shop vac and Arduino to form the screen shape.
The frame above is the structure that will support the screen. A sheet of mylar was later attached to the edges of that frame. That is pulled into place by the suction of the vacuum. But it needs to be stretched just the right amount or the projected image will be distorted. They’ve got something of a PID controller to manage this. A valve box was built to vary the amount of vacuum suction inside the screen’s frame. A switch positioned behind the mylar sheet gives feedback to the Arduino when the screen reaches the appropriate position and a servo closes off the suction box. If you lost us somewhere in there the description in the clip after the jump will help to clear things up.
Here’s an unrelated project that implements the same concept on a smaller scale.
[youtube=http://www.youtube.com/watch?v=QC7H61palXs&w=470]
[Thanks Keith]
http://www.youtube.com/watch?v=4R2bcZ-3eX8&list=UUpSdEDGUppf-I8SUMXjPBZQ&index=2&feature=plcp
the team from your youtube example built a large one too and it works really well.
You’re probably wondering what the hell does this thing do? Yeah I know, me too.
Turns out Collimated displays are used in flight simulators to provide more realistic out-the-window views. It more accurately simulates the focal length of distant objects.
That “PID” control algorithm is defenedly over-compensating!
@martinmunk: You’re correct. The PID controller hasn’t been tuned propery yet. Wayne is also looking at using the new PID library for the Arduino to replace our custom PID with that one to see if it’s easier to tune.
@Sam: The image pre-warp is being handled by a program called Sol7 – it’s the “commercial” edition of their “consumer” product, NThusim+.
If you want to see all the pictures & write up I did while the project was under construction, you can head over to htto://www.diy-cockpits.org/coll
The funny thing is the “non related” project that is linked is also mine & Waynes. See youtube.com/f15sim for video on us cooking hot dogs with it after we decided it’s “productive” lifetime was over. 🙂
g.
I’m the ‘Wayne’ mentioned above. I can tell you that a PID control is -not- overcompensating. Without active control, it’s next to impossible to keep the mirror in the right spot.
Having worked on those displays and full flight simulators, I can say from experience that aligning the overlap areas of multiple projectors on curved, collimated displays is a real bitch. Almost as tedious as calibrating the color curve and brightness of each projector so that they match. That last part is made worse by the fact that no two projector lamp ages in quite the same way.
Wouldn’t you just need to align and blend the images on the intermediate rear-projection screen? It seems as far as blending the image, the mirror is pretty passive.
Indeed. Most multi-projector systems I see now do some form of automated calibration to deal with overlap. It’s kind of necessary if you want to do it easily.
normally you have a projector array, whos light bounce off the mirror and onto a back projection screen.
this kinda actually looks too small for its job, as you will still need a screen in there, and a place for the cockpit mockup. ptojectors are usually mounted on top of the mockup. reguardless this is still pretty cool and should work for a small cockpit, like a cessna.
Every time I see these things I think how the vacuum ads to the sound of a jet engine 🙂
I can’t wait until my four month old son is old enough to help me build my cockpit… my wife would rather I not, though.
So what is the advantage to this instead of something like structured projection? I’m genuinely curious.
Couldn’t you just have a projector show an image distorted in software that appears corrected when displayed on a curved surface?
Theoretically you could, but how would you keep that in focus? Projectors focus onto a plane.
What about those laser pico projectors which claim infinite focus, apparently their light source is close enough to a point light. They’re too weak for this yet, but with time, maybe?
I’m also in the flight simulation industry and this is almost exactly how we build the optics. I’m really impressed with their mirror, as installing the mylar is the hardest thing to get right. It helps that theirs is pretty small, but I’m sure it’s still a real pain.
As for getting the geometry right, we used to do that in the projector hardware (and it was almost an art form in itself) but nowadays we do all the corrections in software. It’ll be interesting to see how they tackle that.
Now that I understand what it does…that’s pretty cool.
I love this project! Can anyone help me doing the same in Russia?
I am sure I should not have to point this out, but you should also mention that Gene Buckle is one of the authors of this setup, and I believe the assembly was cut on his shopbot. Incidentally, you might want to check out his rebuild of an actual F15 cockpit as a flight sim cockpit too (at http://www.f15sim.com/).
Very neat but when I first saw the picture I wasn’t thinking projectors.
I was thinking multiscreen and lenses to curve and combine the screens seamlessly. Maybe enlarge them a little depending on the brightness/contrast your willing to sacrifice.
Anyone know if the modern ATI and nVidia drivers support curved display/viewport mapping/processing?
i would love to see something like this for an FPS 🙂
I’m the ‘Wayne’ mentioned in the article. I did the optics design and some of the structural design. A couple corrections:
While Rob has been a big help, the credit should go to Gene Buckle. Without his skills, shop, machinery, experience and enthusiasm, this project never would have come together.
The “Unrelated” project is not unrelated. It’s the smaller 60° prototype Gene, Rob and I built to test the concept before we went through the effort of building the full wraparound version.
It occurred to me a week or so ago that you might not need to go the mylar route, as Alsa Corp makes two amazing chrome paints, Mirrachrome and Killer Chrome. (Check out how close they come to real chrome.)
That stuff looks nice on a set of wheels, but it’s too diffuse to make a proper mirror, and even the slightest bit of ‘orange peel’ would ruin the effect. Even if the paint were perfect, it would require a perfectly formed structure in order to work.
Please don’t think I’m insulting you (I hold everyone, including myself up to the same standards), but
1. Do you have a credible source as to it being too diffuse? The manufacturer states that “MirraChrome can achieve a 95-98% reflectivity”, as this picture seems to prove. (I don’t have the information as to whether or not it would work even as well as mylar, but if the difference in reflectivity is only around 5% would save a lot of effort and hassle compared to a vacuum system, I think that it would be worth the effort to give it a try by building a small test rig.)
2. ‘Orange peel’ is the result of incorrect paint application. Some education and practice can prevent this, or, preferably, a local paint shop could be used, as they would have the necessary equipment and skills to do the best job.
3. As (I assume) the shape of the mirror is a perfect toroid section, it lends itself to being shaped using circular tools. For example, a disc with sandpaper in a band along the edge would be able to remove the perfect amount of plaster/etc. Each section of the mirror would be mounted on a jig that let it swing along the arc of the torus. Yes, the disc would be large, but cutting a perfect circle, even one three meters across, is easily done.
Another method that occurred to me would be to use positive pressure to form the mylar to the proper shape, then place a support frame covered with hardware cloth almost touching the mylar, and finally spray resin/etc. to bind the two together. As long as the binding agent doesn’t expand/contract (and sufficient care is taken), the perfect shape should be maintained.
Hmmm… A bit of further research reveals that the best mirror shape is parabolic, rather than circular. That, in turn suggests using a smaller material removal tool, traveling along a parabolic path.
1: ‘Reflectivity’ is not the same as ‘specularity’. I haven’t done research on that particular brand of chrome spray, but from what I’ve seen in general, it’s less than ideal.
2: ‘Orange peel’ can be reduced via proper application, but it cannot be completely eliminated. For such a large optical surface, any orange peel is unacceptable.
3: The intended mirror surface is spherical. In fact, my original construction plans involved building a thick shell and carving the surface with a tool mounted on a swingarm pivoted at the mirror center. For the actual optical surface, however, it needs to be optically smooth. This approach has been tried unsuccessfully by several people before us, and the result has been disappointing. -ANY- imperfections (even a tiny piece of dust!) on the optical surface becomes blatantly obvious when mirrored. There is a company out there producing acrylic spherical mirrors, so Mylar is not the -only- way to do it. It’s simply the easiest way to do it without precision machinery and garage-size forming ovens.
Research is underway in the direction of positive-pressure forming and construction of a solid surface.
A parabolic mirror is optimal, but from a single point only. A spherical mirror is slightly less optimal, but is adequate from a much larger range of eye position and is much easier to construct.