[discussion] home-made high FOV 3D HMD for modern games

[seanny]

well, I'm gradually amassing a collection of lenses, mirrors and prisms. It seems like a lens set (if more than one is even needed) would be easy to design. The problem of redirecting the optical path to accommodate two 5.6" screens remains however. Wedge prisms have too much distortion & chromatic aberration. I thought half-penta prisms would be a panacea, but they have FoV limitations related to the IOR trickery that allows them to do their thing... on top of that, they are impossible to source in large sizes. I'm gonna have a full penta prism to play with soon. Even though they are available in larger sizes, they have their own problems like cost and weight, and I wonder if the two-bounce internal optical path will be a FoV-killer.

Meanwhile I'm gonna play some more with a mirror-based setup, displacing my woes to the image-mirroring issue.

In our collective fantasy land, 3" high-def displays would be readily available so we wouldn't have to deal with all of the above.

[Synexious]

In our collective fantasy land, 3" high-def displays would be readily available so we wouldn't have to deal with all of the above.

Could the Cinemizer OLED microdisplays possibly be sourced?

[PalmerTech]

Even if we could, they will not be anywhere near 3". The highest I can imagine them being to fit in the shell they are using is 0.7", which would be an industry standard for higher resolution displays, as well. And designing optics for microdisplays is tough.

[seanny]

yeah, the OLED displays on the Cinemizer are going to be Kopin-sized and not high-FOV-ready.

Anyway a mirror-based design would work... which I guess is not a revelation. I remember seeing a teardown of an existing HMD that used such a design, with image processing to flip the video. Meanwhile I'm imagining something way more complicated involving a prism and concave & convex lenses, but I'm still waiting for that penta-prism to come in...

[AntiCatalyst]

I just dissected an old Sigma 70-200 telephoto lens i had for my old camera, thinking that the front lens would make a good magnifier. It works!

FOV is around 60°(calculated, not guessed) on my 4" phone, and focal length is about 10cm. Obviously there is a lot of pincushion distortion, but sharpness is very uniform and there's no CA at all. The front lenses from regular binoculars should work much like this one, and then you'd even have one for each eye.

P5020737.jpg

[typop]

I am hoping that someone here has the patience to help an extreme newbie …

First, apologies in advance for any incorrect terminology. Someday this will all click for me, but at the moment it is more than a bit confusing.

What I am interested in is hardly unique: the possibility of mainstream immersive gaming with an HMD. The difference between what I am looking for and what most of you are looking for is that I am NOT concerned with porting modern games into a VR setting. All I am concerned with is creating the illusion that objects are fixed in space. (Most likely I am describing something that has a name, but I don't know what the name is!) To use a specific example, I imagine a monochrome outline of a cube floating at approximate ground level. The user wearing the HMD could turn his head, approach the cube, whatever, and still have the feeling that cube was really there. I want to be clear that the user is (of course) aware that the cube isn't real, and that the HMD has various technological limitations. But he still is able to have the feeling that it is there.

I want to further assume that the motion-tracking aspect of the system has been solved, as well as the processing requirements.

Also, I want to define “immersive” as high FOV (120° or so).

I am guessing that such a system would require a grid of OLEDs and the optics (or serious math?) to “map” the pixels onto a sphere with very little distortion and/or blur (for the overlapping pixels). If this is close to correct, then I guess I am ready to pose my question to you, the expert: What do you think is the lowest resolution, smallest grid, and cheapest optics that could possibly achieve the illusion I am looking for?

All answers, including partial answers, much appreciated! This question has been gnawing at me for quite some time ...

[cybereality]

Sounds like what you are talking about is augmented reality. This could be done with a transparent HMD or, more practically, a standard HMD with two small cameras mounted on it. The cameras would film the surroundings and display it on the screen, so it would appear you were just looking though glasses. The cameras would also be used for computer vision and motion-tracking. The easiest thing would be to use a tracking pattern combined with software like ARToolKit. This will allow you to have your cube positioned in real space. Not sure if thats exactly what you are talking about, but it would require the same type of technology.

[youtube]http://www.youtube.com/watch?v=NTy_kJfElWU[/youtube]

[typop]

Thanks very much for the response; the tech you describe is interesting, and I've seen some examples here and elsewhere. But I realize now that my post was unclear. When I talk about the feeling of an object fixed in space, I do still mean a fully virtual object, rather than augmented reality. The reason I said that I don't care about porting modern games is that I was worried someone would think I was asking for the latest FPS extravaganza in 60fps and 1080p, when all I'm thinking of is BattleZone or Tempest! So they would say "impossible" and move on.

Does that make sense? (And is it still impossible?)

[cadcoke5]

I am still not totally clear on your questin. But, suspect that you are talking about is simply called "Virtiual Reality". This involves a display that can present the user with a view of an object that is generated by the computer.

Such a system typically involves either a head mounted display or a projection onto a screen (or multiple screens). Also, the viewpoint of the user is determined using a device called a "head tracker". The computer then calculates what the image on the display should look like by integrating the viewpoint and its knowledge of the geometry of the display.

Since you say a 120deg field of view, you are talking about technology that is not really available to users as a head mounted display. Although many are trying, it is proving to be technically difficult to achieve.. So, you will probably need to go to either multiple projections in a setting known as a "cave", or a projection onto a curved screen.

The outline of an object, as you describe it is called a "Wireframe display" as opposed to a shaded display. But,nowdays the technology behind it all is the same. It used to be in very old systems (like from the 70's) that some CRT displays were only capable of a wireframe display, but today that is no longer an issue.

Joe Dunfee

[typop]

Thanks, Joe. That's pretty close to what I was asking.

I guess there are probably a lot of reasons that a good HMD VR solution is difficult to achieve, but I am curious if you see a particular obstacle that is the hardest to get past. It seems to me that the prices of monochrome lo-res OLEDs (even with some grayscale for "shading") are coming down. The sensor and processing aspects can be done cheaply, or will be able to be done cheaply very soon. I'm guessing --- though I certainly don't know --- that the biggest issue is in what I was calling the "mapping" issue. That's probably a misnomer, but I don't really know how else to describe it. (The user needs to see the object on a virtual sphere, and the pixels are displayed on a grid of discrete non-spherical displays.)

Anyway, thanks for taking the time to respond. I know you have better things to do ...

[cadcoke5]

The challenge is that the optics necessary to obtain a 120 degree view are a problem. If you are willing to settle for a 30degree view, then it can be done.

Joe Dunfee

[typop]

Yeah, I was afraid of that. (Sort of hoping that you might be able to compensate for a cheap wide-angle lens with some complicated math ....) Anyway, thanks again!

[seanny]

If image processing is not a constraint, you can always hood up a high-FoV lens set and attach them to a couple of phones, or do something like this... and then use image processing to correct for distortion and chromatic aberration. That's how the commercial and once-patented "LEEP" system worked.

And as Cadcoke mentioned, you can also build a 3D CAVE:
[youtube]http://www.youtube.com/watch?v=MQeHPnMCKWY[/youtube]

The ready-to-rock, driver-supported PC gaming version would be a triple 3D projector setup, for panoramic 3D gaming. If desired, add a headtracker (trackIR, freetrack, whatever) and some headphones.

My particular quest is to make something that anyone can build with off-the-shelf parts and a minimum of electrical engineering that "just works" with modern games & game consoles in hi-def, so my particular hangup is designing an optics set that can accommodate a particular pair of 5.6" 800p panels. As far as I know, that's the smallest hi-def thing a hobbyist can get his hands on that'll accept a digital video signal. A more creative or resourceful person, or one who is unafraid of engineering an image processing board, may find a better way.


...while I'm at it, I received the "bed prisms spectacles" and realized they were just (coated) Littrow prisms, which are functionally similar to (coated) half-penta prisms. You can get a decent viewing angle out of them, like sitting 4ft away from an HDTV, but like the half-pentas, the IOR effect that they rely on prevents them from achieving immersive FOVs... in one axis anyway. At least they're easier to source. You could precious-bounty them off of the bed prism glasses if you wanted, but I think I'd need bigger ones...

[cybereality]

I liked those old wireframe games like Tempest. What's really sick is playing them on a vector-based monitor, like the original Asteroids arcade unit. Forget 1080P. It has infinite resolution! Sick!

[Synexious]

Infinite resolution?

[crim3]

Infinite resolution?

A monochrome vector display doesn't have pixels but a continous phosphor coating.

[Okta]

Thanks very much for the response; the tech you describe is interesting, and I've seen some examples here and elsewhere. But I realize now that my post was unclear. When I talk about the feeling of an object fixed in space, I do still mean a fully virtual object, rather than augmented reality. The reason I said that I don't care about porting modern games is that I was worried someone would think I was asking for the latest FPS extravaganza in 60fps and 1080p, when all I'm thinking of is BattleZone or Tempest! So they would say "impossible" and move on.

Does that make sense? (And is it still impossible?)

Do you want 3d or 2d display? Sounds like all you want is a high FOV 2d HMD with good head tracking and a controller for movement (or do you want the setup to map your actual movement to the virtual world?)

[pierreye]

Guys, looking at the spec, it is possible to use this lens to provide wide FOV for HMD on 5" screen?

http://www.loreo.com/pages/products/loreo_lubot.html" onclick="window.open(this.href);return false;

[martinh]

Are we going to cannibalising our older HMD for the LEEP optics and fitting new displays?

I have a 17 year old rig, the same as the one I used in 1994 here, from Division. It has a dVisor HMD, which as far as I can tell is very close to being a VR Flight Helmet, but grey and restyled, with LEEP optics and a Polhemus Fastrak. After 17 years it seem to be state of the art in everything but resolution.

I'm planning to attach is to a modern computer shortly, but have no time for a few months.

Also, I was blown away by VR in 1994, but I've only ever used LEEP HMDs, is there much less of a feeling of presence with the lower hfov ones?

Just found the forum, so hello everyone...

Cheers,

Martin.

[ERP]

Yes the difference between the older HMDs with 90 degree FOV's and modern HMD's with ~35 is dramatic to say the least.

[bobv5]

After 17 years it seem to be state of the art in everything but resolution.

Sad, isn't it!

[cybereality]

@martinh: Hey man! Welcome to the forum.

I've been into VR since the 90's as well, and it is a bit sad technology hasn't gone further. But I think things will get better, we will just have to build it ourselves, though. No sense waiting for some big companies to throw us a bone.

[Johnny-Mnemonic]

Are we going to cannibalising our older HMD for the LEEP optics and fitting new displays?
...
Just found the forum, so hello everyone...
Cheers,
Martin.

Hi Martin!
I would say to keep your HMD as it is, it's very rare stuff you know, not to cannibalise it in sake of optics.
Few talented guys in the community already developing their own HMD's, and not bad optics can be built from stock lenses, so please don't hurry up with breaking such museum-grade stuff

[martinh]

Thanks everyone.

No, I shan't break it up - it's the full set, an IPU (Integrated peripherals units - with the Fastrak embedded, and the serial passthrough for mouseclicks), a Division 6D mouse, and the HMD. It came with a pair of SGI Indigo2's one of which had the Impact Channel Option to drive the HMD, plus cables and a converter box to take VGA to NTSC for the HMD.

Didn't come with any software though, it *should* run with dVS/dVISE for IRIX, but the boxes had been wiped. I did get one eye working with Maverik on the SGIs and jittery tracking, but I think the best bet is to connect it to a new (Windows) machine and run it from there. I only have Macs at the moment and no dual displays - my plan is to get a Windows PC with dual displays and something like Vizard. None of the software today seems as user friendly as dVS/dVISE was, but I can script OK (Unix background).

I was a student at a pharma company in 1993/4 and work on this rig, but with a ProVision 100 VPX for VR and dVS/dVISE for IRIX for making models.

What I really want to do is show other people and my kids the sensation of full presence in a VR, this really blew me away at the time. I still look for dead pixels in the sky after all these years....

[martinh]

@martinh: Hey man! Welcome to the forum.

I've been into VR since the 90's as well, and it is a bit sad technology hasn't gone further. But I think things will get better, we will just have to build it ourselves, though. No sense waiting for some big companies to throw us a bone.

Thanks!

I'm hoping the games and AR people will generate tech that can be reused in VR. That seems to be accelerating, so I'm hopeful, especially around trackers and vision.

I'm not against AR to be honest as an adjunct. One of the downsides of immersive VR is that you're helpless and vulnerable in the real world. An AR overlay like a wireframe real life that gets thicker as your near obstacles would be a cool way to make VR safer. Something like a Kinect or Roomba sensor could do the input, and the AR overlay algorithms will be done by the AR people. That sort of thing.

So VR becomes a special case of AR, and AR is getting a lot of interest. That said, AR might end up being the next VR!

Will our kids live in two or more realities simultaneously?

[cybereality]

Yes, AR is the new VR. But eventually AR will be so immersive it will become VR again.

[martinh]

Exactly what I'm hoping... and with the added benefit you can wireframe the real world over the VR world, so as not to walk into the wall...

[martinh]

Looks like LEEP patent expires in 2013? Wonder if we'll see more wide FoV HMDs turn up after that happens?

[cybereality]

Looks like LEEP patent expires in 2013? Wonder if we'll see more wide FoV HMDs turn up after that happens?

Interesting. I sure hope so.