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

[seanny]

Rewritten for April 2011

Don't let the thread title get your hopes up-- this is just a discussion thread that may not go anywhere. In short I'm mulling over designing a high FOV, hi-def (720p?) HMD that can play modern games in 3D (i.e. PS3 & PC DX11 without IZ3D-like hacks). As far as I know, such a product does not exist, and will not exist in the short term simply because there's no "enthusiast" market for this stuff. I'm going to lay out the two avenues that I'm aware of for creating a home-made enthusiast HMD, just to get the ball rolling:

1. A monocular, shutter-based HMD using nVidia's 3D Vision kit or something. The idea was mentioned in passing in this thread. I guess the trick is sourcing a tiny DVI (dual-DVI?) input panel that can run 120hz & works with 3D Vision, or a tiny HDMI 1.4 120hz panel w/ glasses kit... or a tiny 3D polarized panel. I assume none of those are easily obtainable, if they even exist at all.

2. A true stereoscopic HMD, which seems like the least sketchy (though still sketchy) route. This design presents the following challenges:

- 2A. Build/Optics/ergonomics

People have attempted twin-pico-projector rigs in the past, but bulkiness aside, there are various challenges involved. Someone experimented with a pico-projector-backlit design, but found that at the scale of HMDs, all the material you could rear-project on has too much grain/texture built in.

With larger displays (5.6" to 7"), they cannot be placed directly side by side. I've tried wedge prisms and half-penta prisms as solutions to redirect the optical path, but they introduce too much distortion. A mirror-based solution can be used, but using more than one is an FOV killer, and then you're stuck with the problem of flipping the image.

- 2B. Signal Standard / Controller Box
The most viable 3D signal standard I can think of is HDMI 1.4. The PS3 supports it, nVidia supports it and AMD supports it. The practical gaming spec calls for 720p 60fps 3D, which seems plenty fine until the day 1080p pico-projectors or micro-displays exist and become easily obtainable.

The problem is the controller box, assuming we have two displays to feed with a separate left/right-eye signal. There needs to be a way to demultiplex the HDMI 1.4 3D signal into two HDMI 1.3 (or whatever) signals. I can imagine doing this with a computer outfitted with a pricey HDMI 1.4 3D capture card (1, 2), but the only stand-alone solution that I'm aware of is buying two Optoma 3D-XL converter boxes for way too much money.

Edit 2011: Cypress CH-322 demultiplexeron the horizon at $300/pair

- 2C. Displays
If the HDMI 1.4 signal is demultiplexed via a computer, the computer's software & video card will effectively act as the scaler and signal converter for whatever kind of display you want to use, allowing for more esoteric displays (cell phone panels, camera viewfinders and what have you) provided you can find a way to control them. But assuming we're demultiplexing HDMI 1.4 via two Optoma 3D-XLs or a similar converter box, you would need head-mountable displays that can accept digital 720p signals, and ideally 1080p encrypted signals for movies, but that might be asking for too much.

Currently I'm looking at 5"-7" "field monitors" that accept HDMI, meant to be equipped onto cameras & camcorders. Some of them have image processing features like X/Y flipping, which would be critical in a one-mirror-per-eye design, and they are readily available and cheap ($160 for the cheapest China crap, $900 for 720p).

EDIT: Looking at the HV056WX1, a 1280x800 5.6-inch panel that can be eBayed dirt cheap.

- 2D. Head tracking / Audio
This part writes itself. There are decent head tracking solutions out there (TrackIR, FreeTrack). Headphones would then be necessary, as moving your head while still being able to localize your loudspeakers in your room would break the immersion. The headphones would also have to get out of the way of, or become an integral part of, the HMD's ergonomic design.

Sorry folks, that's all I have in the way of ideas.

[cybereality]

Hey man! Welcome to the forum.

I am very much interested in the possibility of a DIY HMD with high-res/high-FOV. Pretty much everything on the market right now (except for some ultra-expensive research units) are limited to around 35 degree FOV at SD resolutions. In 2011, that is basically unacceptable. They had stuff like that back in 1999 on the consumer market and even better technology if you want to talk about academic/military. 10 years later what do we have? Squat. I mean, PalmerTech let me try a monoscopic prototype he put together and it was pretty immersive. I think the resolution was 1280x800 and there was like a 90 degree FOV. So it seriously blew any headset I have tried out the water in terms of image size and quality, however it suffered from some usability issues and was not comfortable on my eyes. So, I don't think this is about technical limitations. There have been many significant developments in display technology over the past 20 years (since VR had its heyday) but these don't seem to have been applied to VR headsets. This is mostly an issue of economics. There just isn't a mass market for this stuff, and there probably never will be. The mainstream consumer is much more interested in un-encumbering transparent technology. So we will likely see more adoption of AR applications on mobile devices and, to a certain extent, this is already happening. With the recent interest in mobile auto-stereoscopic displays, this will likely be where the industry moves toward (meaning to AR and not VR). However, I still think there are professional applications where headsets make sense: medical, industrial, military, etc. But there are going to have to be see-through, AR-based displays. Think of Ghost Recon style HUDs for soldiers, doctors who can see 3D x-rays overlaid on patients in real-time, engineers or repairmen that can see schematics of what they are working on so they have both hands free. There are valid applications where AR makes more sense than VR. Now it would seem like they use similar technology, and they do, but the motivation is different. With AR, you don't necessarily need a super-wide FOV and it doesn't need to be stereoscopic (even a single eye-piece is enough to display a simple HUD). However they still use HMDs, so micro-display technology remains essential. But the point is that there are not as many consumer applications outside a niche enthusiast market. Of course, there are companies that cater to those niches. Nvidia heavily promotes its 3D Vision, AMD has their Eyefinity, there are companies that offer expensive racing motion-seats and things like that. So I don't think VR has to go completely mainstream for companies to make some profit on it. However the amount of effort, investment, and R&D into making a breakthrough in this field seems significantly higher than the commercial opportunities (ie potential ROI). I guess it will happen eventually regardless, its just a matter of if it happens in 5 years, 10 years, or longer. Preferably, I would like to see this happen with-in my life-time and I am fairly confident that it will (even if I have to make it happen myself). So I definitely feel you about taking some initiative, and not waiting around around growing gray while companies neglect this field. There are a number of people on this forum that feel the same way, and a few of them already have projects started. I think you will like it here.

Now let me try to address the specifics of the topic.

1. A single display solution has its benefits. Mainly in that a larger display can (and must) be used than with traditional HMDs. So it is a lot more likely to be able to source the displays (ie from consumer portable devices, smart phones, tablets, etc.). There devices already are approaching 720P resolutions, in a couple of years we will likely have small gadgets with 1080P. I doubt smart-phones are going to get much bigger (even recent phones like the DroidX or EVO4G are on the verge of being too big). So lets just assume 4" 1080P screens are the best we will be able to work with, within say the next 5 years. Even with really good optics, I am not sure you could get more than 60 degrees FOV out of that (maybe less). Even if you could go higher, it would likely distort the image too much to be acceptable. However, even 60 degrees would be a nice jump from current headsets, so I am not discounting that. Also, I am not sure if these screens are going to be 120Hz or 3D-capable, so that adds another wrench into the works. Even if we have auto-stereo displays, they will be difficult to get working at such close distances and the optics might ruin the effect completely. Using a micro-polarizing sheet could be a good solution (like the Zalman) however it might be difficult to impossible to source the material or DIY the sheet yourself. So our best hope is that mobile devices move to 120Hz, but I don't really see much benefit for them to do this unless they expect people want to use active shutter-glasses. I also don't see any reason for Nvidia (and its partners) to start making smaller 120Hz computer monitors. They are only going to get bigger up until a certain point. Once everyone has 30" 2560x1600 120Hz 3D displays, where will we go from there? I mean a bigger display is not practical (and I have used a 32" HDTV as a monitor, you really don't want anything bigger on your desk). I guess the resolution can get higher, but where it is now is more than adequate. They could jump to 240Hz for flicker-free 3D and better support for 24P 3D content (without pulldown). Maybe they will move toward bezel-less displays, that can be easily combined ala Eyefinity. Maybe they will move to a different aspect ratio, like 21:9, which could be cool for enthusiasts currently on multi-screen configurations. Curved screens could also be pretty cool and they are close to being able to do this. Maybe displays that are just a clear piece of glass (like in Avatar), that would be nice. I could imagine Apple being the fist to provide something like this. Now I am going off on a tangent here. The point being is that computer monitors are not going to get any smaller, and smart-phone displays are not going to get much bigger. So we will have to hope that what is in-between, namely tablets, will be 120Hz 3D capable in the future. Not going to hold my breath for this.

2. Yes, stereoscopic HMD is the more straight-forward solution, and would be higher quality (ie no ghosting, 100% light throughput, greater horizontal FOV potential, no sync issues inherent with active solutions).

2A. Optics actually seem to be one of the major issues right now (aside from finding the correct displays). I have read that LEEP on the cheap article, so I know there are methods that could be explored. But there are certain limitations you come into with this technology, and there is only so much you can magnify an image without distorting it. Some of the more innovative solutions for high-FOV optics require an intentional warping of the source image. This can work for a simple tech demo, but then we will need a way for commercial games to support this. I am not sure the iz3D guys will dedicate precious time to supporting some one-off custom headset, so this means actually developing our own stereo 3d driver. I have considered this myself (for use with other experimental 3D hardware) so if we did go this route, I could help with the driver development. However this is a serious project on its own right, and could complicate things. Ideally we would want a solution that can work with existing driver solutions. In terms of using projectors (especially laser) instead of LCDs, I think it has some merit. I have long been interested in creating a pico-projector based HMD. There is a thread discussing HMPDs, which you might find interesting: http://www.mtbs3d.com/phpBB/viewtopic.php?f=26&t=11982" onclick="window.open(this.href);return false; . I think something like this could work, although the kind of enclosure you would need might be too big and bulky to be practical. Maybe it could be used as a stationary unit you just place your face in (similar to the Virtual Boy). But, like you said, these problems are not insurmountable, but may require research and investment since it does not appear there are any off-the-shelf viable solutions.

2B. HDMI 1.4 is a mistake for a project like this. Now if you wanted to get a production run and actually sell the headsets, then I agree HDMI 1.4 is a necessity. But for a custom HMD, I think it is overkill. It just adds cost and complexity to the project without much benefit besides support for PS3 (and I am not really interested in spending $1000's on a DIY headset just to play a console game at sub-720P res @ 30fps). The first problem is the cost of the hardware. Unless we are building this module from scratch (unlikely) we will need to buy an HDMI capture card (costing maybe $1000 or more). Now we we need a separate PC, just to capture the video and convert it. This is another $1500-2000 right there. Then the time/effort in writing a program that will demultiplex the video. Seems like an awful lot of money and trouble, when we could just use the dual output mode on the iz3D driver with no additional hardware or cost and actually get a better quality image (ie 1080P+, 120Hz, etc.).

2C. Again, I think HDMI 1.4 is not worth it for something like this, as explained above. So there is no need to source HDMI/HDCP capable displays (which may be problematic if we are dealing with pico-projectors and stuff like that). Even straight VGA should be good enough for what we want. In this respect, I think pico-projectors have the most potential because of their cheap price, and variable parameters (like screen size) that make them ideal for experimentation. Especially laser projectors which can focus at any distance (even on curved surfaces) giving use incredible room for trial and error. They mostly all support common interfaces, like VGA, so they will work out of the box on a PC. The problem with re-purposing consumer devices (like a smart-phone) is that we would need to build the control panel custom, which may not be that easy or practical. So pico-projection seems a lot more desirable. In terms of resolutions, they are already on par with smart-phones, and 720P prototypes have been shown at trade shows. 1080P is almost a guarantee within maybe 3 years. Upcoming consumer devices will be integrating this technology, already there are cameras and phones that do this. So the technology is going to get small and cheaper (perfect for building into a headset). I am going to watch this space carefully. In fact, I plan on buying some picos and giving this a shot once I am finished with my DIY auto-stereo mod.

Now if we had some way to build (or commission) custom LCD panels of any size we wish at HD resolutions, then we might have some more options. We could play in the space of maybe 0.5" to 2.0" diagonal and see what we could do with that. However we don't have a facility to DIY an LCD panel, and the common suppliers either cannot or will not produce quality HD displays in those sizes (and even if they could, we couldn't afford it unless we were buying bulk). So going with LCD micro-displays seems like a dead-end at the moment. If this were viable then why can't a company like Vuzix produce a better headset? Even their newest model (announced a few weeks ago) is only 800x600 @ 35 degrees and it is for the military. Even Carl Zeiss, one of the most respected companies in the field of optics, could only reach 35 degrees with their upcoming Cinezimer OLED. And they have the facility to make custom optics, we don't. And I am not to believe that all these companies are just fools and are missing something obvious. If it were really that easy to make HD microdisplays and wide-FOV HMDs, why has no CEM done it? I know there are some professional quality HMDs that cost in the tens of thousands and do have higher specs than stuff on the consumer market. However if these consumer companies could make a high quality device and sell for a reasonable price, why wouldn't they? Its been like 20 years since the whole VR thing got popularized, surely we are not the only ones interested in having this technology. I mean, you see the kinds of research projects companies like Sony or Microsoft do in their spare time. Much of this stuff never gets released. So even if the ultimate VR headset were too expensive to be market viable, a prototype would at least have been shown at trade-shows, don't you think? I am not saying the traditional microdisplay HMD is impossible to improve upon, but it feels like there is some physical or technical limitation holding it back. And if all these multi-million dollar companies cannot find a solution over 20 years, I am not sure how much we are going to accomplish in our garage with a couple hundred bucks.

Another option is to use multiple displays, ala Eyefinity or Nvidia 3D Vision Surround. Lets say we can get a hold of 4" 120Hz 720P LCD displays and have a method to drive them. Now we could mount 3x of these displays in landscape orientation (angled in) and use 3D Vision glasses for stereo 3D. With 3 displays we would have 3840x720 resolution, which is more than enough. If the screens were mounted at a 6" distance or more, you might not even need any optics to focus the image at all (but, of course, they can help to increase the screen size if desired). Just by eyeballing it, I think that setup would give about 90 degrees horizontal FOV, with about 60 degrees stereo overlap. Vertical FOV would be smaller, but still better than most HMDs today. This is about the best we could hope for in a setup like this. And again, this is without using optics. If we could find some adequate lenses we might be able to get an even bigger image and/or more immersive experience by masking the fact that we are looking at a screen so close to our face). We could also try things like using LED ambilighting around the edges to "fake" wider peripheral vision. I think it is worth looking into. The same concept could also be used with triple pico-projectors and, using laser, we could even project on a curved screen for a seemless experience. Projection may also allow the use of polarizing filters so now we can use passive polarized glasses for stereo 3D. This would, however, now require 6 pico-projectors, so we are about past the limit of what can reasonably be housed in a headset. But if it were stationary, this might be ok. Although if we are now needing 6 projectors you have to wonder why even use a headset at all when we can build a CAVE...

2D. Well in terms of audio, that has already been mastered. You can obtain a 5.1 headset for around $100 and it is already supported on many games with no additional software patches. It should be pretty easy to modify an audio headset to work with the HMD housing and could even be used as a basis for the frame. Head-tracking should also not be a problem. There are 3rd party solutions available that should be adequate, and its possible to DIY your own solution using cheap hardware like the Nintendo Wiimote, Playstation Move/6-Axis, Microsoft Kinect or any combination of the 3. For any quality head-tracking, you would need to be able to move your view in the game independent of which way you are moving/aiming. This type of control is available on certain racing sim titles and games like ArmA 2. However the vast majority of popular titles do not support this. Mouse-emulation is a crappy solution and is not really what VR is about. So for any meaningful head-tracking, we will need to mod these games. This goes back to the need for a custom 3D driver, which adds development time and complexity. I know the TriDef driver devs added a beta head-tracking support on their 3D driver (using a web-cam) but it was absolutely horrible and I could not get it to work at all. In theory, it would be very cool though and add a significant amount to the immersive. For example, with the VR920 I have played UT2004 with this sort of setup (in 3D with independent headtracking and mouse movement) and its pretty cool. Its still somewhat limiting, though, if you remain seated facing forward. So for this to reach its full potential, you need to be standing up and moving freely. This adds its own can of worms in terms of cable management. Unless you can somehow make the whole setup mobile (either with a gaming laptop in a backpack or some wireless solution) then you are constantly encumbered by cabling. Even if you got 100 foot VGA cable, you could still get easily tangled up. If you primarily turn to only one side in the game, you would only get around maybe 15 full 360 degree revolutions before there was no cable left. In a fast-paced twitch FPS, it is not unreasonable to think you could do this in like the first 5-10 minutes playing. I guess you could build some sort of ceiling-mounted gyro, which would certainly help, but I am not sure even then you would have 100% full mobility. The laptop in the backback seems a lot more practical (although this limits the power of the system, especially for crazy stuff like triple-screen 3D). And if we are now standing up and turning freely, we now need something besides and mouse and keyboard. At the most basic level, an Xbox360 controller would work (although for all the time and money spent on this advanced getup, I would hope for something better than a console gamepad). So maybe something like a Nintendo Wiimote in a gun shell would be nice for FPS games. That ForceTek Xio controller might also be perfect for this (if it ever really comes out). There are also air-mice that could be used for this. You would still need some way to move around in the game, so a second controller would be necessary. The Playstation Move Navigation controller looks perfect for this. So assuming you had a nice sized room, and developed a driver so you could control games like this, then I think it would work.

Now, practically, this setup seems like it would be very expensive to build and might be too cumbersome to be comfortable to use. You could also potentially spend $1000 or more to find out it doesn't even work at all. Lots of risk here. Especially when you can obtain 720P 3D Vision projectors for $550, and potentially run them in Nvidia 3D Vision Surround mode. For under $2000, you could practically have a 300" ultra-widescreen 3D display in your living. At a certain distance, this would completely fill your horizontal FOV and I imagine it would be pretty immersive. It is also an off-the-shelf solution, we wouldn't need to be writing custom 3d or input drivers for games to work. We would lose the more advanced VR elements like head-tracking and free-movement, but even this could be added with TrackIR/FreeTrack (although spinning around 180 degrees obviously wouldn't work). But we could use alternative peripherals that require a stationary desk, like force-feedback steering wheels, HOTAS controllers, the Novint Falcon haptics force-feedback controller, maybe the Razor Sixense (if it ever comes out), plus the other stuff I mentioned like a gamepad, wiimote, air-mouse, etc. Of course, going with the pre-packaged commercial solution is not as fun as paving new road with a DIY research project. So I certainly see the appeal of building a custom HMD. But at some point you have to look at whats out on the market and wonder if existing commercial solutions may just be better than whatever we think we can cobble together. I don't know, something to think about.

OMG! This has turned into a friggin' thesis paper!!! I really ought to just write a book one of these days...

[seanny]

To be totally honest, I'm in a wait-and-see mode about a lot of this. With the supposed popularization of 3D movies and 3D gaming, and (finally) a universal standard (HDMI 1.4), it wouldn't be hard to imagine a small enough panel with HDMI 1.4 3D support that can be slapped on a home-made single-display HMD. It also wouldn't be hard to imagine Chinese clones of the 3D-XL converter box or some other future 3D demultiplexing solution that makes the HDMI 1.4 route affordable for a dual-display HMD. Pico-projectors seem like they're on the cusp of 720p micro-HDMI designs, closing the door on those analog 640x480 VGA things. It feels like a lot of enabling products are right around the corner (a feeling that might be perilous, given the stagnant history of enthusiast VR, but still).

I still peg HDMI 1.4 as the best standard looking forward. It's the only conceivable way I can think of to get 3D out of nVidia's "3D Vision" for dual-display stereoscopy... Otherwise 3D Vision is limited to nVidia-approved 120hz shutter monitors & projectors, as far as I know. There is probably going to be a flurry of devices supporting HDMI 1.4 in the near future, so unlike past proprietary standards that have come and gone like the wind, HDMI 1.4 will be here to stay. IZ3D & TriDef support dual output stereoscopy, but are non-DX11 and otherwise iffy. TriDef, being an active and officially sanctioned (by AMD) effort, may improve in the future though.

There's no doubt that building an HMD is an impractical hobby/nerd project. You could definitely make a triple-head 3D projector system, or a near-CAVE-like rig with less fuss and with official driver support for gaming from AMD or nVidia. I'm well aware of that. For a completely dedicated gaming setup, a triple-head stereoscopic system and an array of SLI'ed graphics cards to drive all the displays would be the best thing.

My original post was more like a list of assumptions I had for people to critique or suggest alternatives, so I really appreciate the monster of a reply that you wrote. If all the challenges of building an enthusiast HMD can be overcome with enough effort, then the challenge becomes getting the cost and complexity down as much as possible. Making a TriDef/IZ3D-only display out of some cheap VGA-connector 640x480 pico projectors would probably be the best thing if one were to start building today, but I'm hoping there will be some things on the horizon that will change this calculation

[Tone]

So lets just assume 4" 1080P screens are the best we will be able to work with, within say the next 5 years. Even with really good optics, I am not sure you could get more than 60 degrees FOV out of that (maybe less).

You can get much wider FOV from 4" screens. Remember that the Virtual Research VR-4/V6/V8 achieves 60+ deg. with 1.3" displays and the distortion is acceptable (to me, anyway!) The original Virtual Research Flight Helmet used real LEEP optics and 2.5" displays and you got 90+ deg. at 100% overlap. However, with today's sub 1" displays, getting wide FOV isn't easy (or practical.)

Another barrier to entry is the optics. "Leep on the Cheap" showed a proof-of-concept with off-the-shelf glass optics, but at the expense of severe distortion and chromatic aberration. The original Leep optics were plastic (to bring down the weight) and used quite large custom molds for each of the two elements. Cost to design and fabricate is on the order of $100,000 per mold. If you're building 100s (or even a few thousand) consumer HMDs, you can't afford large diameter custom optics.

Virtuality's Visette 2 (and beyond) used custom plastic lenses with the same 1.3" displays as Virtual Research's VR-4/V6/V8. Can't imagine they ever amortized the cost of the molds. On the other hand, Virtual Research figured out how to do it with off-the-shelf optics and saved all that tooling expense.

For these reasons, DIY projects are pretty much limited to off-the-shelf optics.

[cybereality]

So lets just assume 4" 1080P screens are the best we will be able to work with, within say the next 5 years. Even with really good optics, I am not sure you could get more than 60 degrees FOV out of that (maybe less).

You can get much wider FOV from 4" screens. Remember that the Virtual Research VR-4/V6/V8 achieves 60+ deg. with 1.3" displays and the distortion is acceptable (to me, anyway!) The original Virtual Research Flight Helmet used real LEEP optics and 2.5" displays and you got 90+ deg. at 100% overlap. However, with today's sub 1" displays, getting wide FOV isn't easy (or practical.)

Ah yes, you are right. I knew about those headsets in passing, but I never had the chance to use them and didn't remember the specs off-hand.

[UncleJoe]

Guys, it's clear for a DIY project to succeed it will have to be a combined effort, but I still think it's quite worth it.

I'm also very interested in steroscopic HMD, and have done a simulation of some professor's idea & implementation based on his publication. See here http://www.mtbs3d.com/phpBB/viewtopic.php?f=26&t=10181 Although the idea seemed very good - projecting a small image onto a large FOV with only a *single* display, unfortunately it seems to be a dead end because of small exit pupil. As the eyes look towards the periphery, the rays completely miss the small display. You can see this effect quite clearly in my video. Small movements (not including rotation) of the camera eye result in the LCD no longer being seen.

Regardless, I think simulation will be a very important tool for building a HMD. If anyone has a good HMD idea and want to simulate it, I'd be happy to do so. I can 1st prototype in Matlab, then implement in the PBRT ray tracer, which is quite powerful & extendible.

For the problem of interfacing the display with the computer, I think HDMI or any digital signal should be doable with a FPGA board and some video break out board and allow almost any signal & image processing desired. I don't have much experience with this, but I have a Spartan 6 SP601 board http://www.xilinx.com/products/devkits/ ... P601-G.htm that I can experiment with.

Anyone want to join?

[cybereality]

Yeah, if any of us want to finish a DIY HMD its likely going to be a community effort. As much as I'd like to help, I am pretty busy at the moment. I am finishing up on my DIY auto-stereo display and I just started working on a 3D image gallery for MTBS3D. I also have some other stuff in the pipeline I don't want to announce yet but which I have been eager to do. So I'll still be on this forum and try to contribute in any way I can, but I don't really have time for any hardcore development right now. However I would still like to have this ultimate HMD we all want, so if I can help make it happen I will try when I get a chance.

[Okta]

I'll put in my 5c.

From my testing i think the simplest method will require this.

2x 3 or 2.5 inch lcd's with driver boards. These are the largest you want side by side. You can almost manage 3.5 inch but you will be using at least 2 offset lenses per eye to get the size but most importantly the natural center of vision. And with 3.5 you will have much distortion from looking through the edges of the lenses you need to see inwards at a natural angle.

At the moment there doesn't seem to be any consumer screens that size available of a useful resolution. Vitrolight on ebay had teh 3.5" vga's with driver but they all seem gone now.

So unless someone can source some displays we are left with playing around with all sorts of monstrosity's of pico's and such.

Just made a sketch example of what i had to do to view 2 3.5 inch size screens.

3dhmd.JPG

[Fredz]

Why not use iPhone 4 replacement screens ? They are also 3.5" in diagonal and have a much better resolution than the Vitrolight ones (960x640 vs 640x480).

[PalmerTech]

Because they are even wider than a 4:3 ratio screen, and we have no control circuitry for them. No way to feed them a video signal.

[Fredz]

The Vitrolight are 70.08 mm wide and the iPhone screens are between 74 and 75 mm wide according to my measurements. Would the 4-5 mm wider size really render these screens unusable ?

And by extracting the needed parts from an iPhone couldn't you drive them with a video signal ?

[PalmerTech]

It is not that it would render them unusable, just that it would render them even more unusable than even the Vitrolight ones are. The problem is still fixable, but every millimeter counts.

And unfortunately, it is not that simple. Driving LCD screens takes a special digital signal, and special chips are needed to convert standard video inputs into something that the LCD can take. The iPhone outputs those digital signals directly, and is not equipped to take any kind of standard video input. I suppose you could use the iPhone as a network video streamer from a PC, but the latency would make it useless for gaming. Could be great for videos, though.

I have been messing around with a ShowWX for a few weeks now, and I have come to the conclusion that it is really not much better than a standard pico for our uses, though the focus free capability is very nice. It is just really a shame that miniature projectors are all going DLP or LCOS, I wish we still had some LCD projectors being made.

[Tone]

@okta - To use two large (i.e. 3.5") screens side-by-side, just add 3M Press-On fresnel prisms. http://www.west-op.com/pressonprism.html. Solves the problem of the screen centers being wider apart than your eyes.

[PalmerTech]

What kind of effect does that have on the quality, Tome?

[Tone]

3M Press-on fresnels are used by ophthalmologists and optometrists on patients with strabismus http://en.wikipedia.org/wiki/Strabismus (misaligned eyes, lazy eye, etc...) LEEP, VPL Research, and Virtual Research all used them on their early HMDs to achieve 100% overlap with large LCD panels. In practice, they slightly reduce contrast, but otherwise simply act as prisms.

[Okta]

Do they add distortion and spreading of colors like prisms?

[ERP]

I used the 3M press on prisms on a set of reading glasses between prescriptions once.
Honestly they're fine, they are a fresnel prism, but on the glasses you certainly didn't notice them.
They are just a prism, so you'll see the same color shifting, but given the axial correction you'll be making, it'll probably unnoticeable compared to the distortion already introduced by the lens pair.

[Fredz]

It is not that it would render them unusable, just that it would render them even more unusable than even the Vitrolight ones are. The problem is still fixable, but every millimeter counts.

It seems the solution proposed by Tone and ERP with 3M Press-On fresnel prisms could solve this problem, nice to know that such solutions do exist.

And unfortunately, it is not that simple. Driving LCD screens takes a special digital signal, and special chips are needed to convert standard video inputs into something that the LCD can take. The iPhone outputs those digital signals directly, and is not equipped to take any kind of standard video input.

I'm not at all versed in this kind of technologies, but since the iPhone 4 uses a PowerVR SGX535 GPU that is also used in many other mobile devices and even in graphic cards for laptops like the Intel GMA 500, I guess the digital signal must be somewhat compatible with some known standards.

I can understand that it should still be quite complicated to drive this type of LCD screen, but I guess someone knowledgeable in this field could find and use the appropriate chip to drive them.

[seanny]

Since it's been a few months, my interest in a 3D HMD picked up again and I placed an order for some 40-diopter press-on prism lenses just to see what they're like (thanks for the tip, Tone). I was thinking of a design that had two large (7-inch?) displays side by side... at least I can sort-of prototype that with some pieces of paper, the prisms, and by taking off my eyeglasses-- for once, myopia works to my advantage.

Another design that came to mind, or maybe I saw it on the forums somewhere, was one that involved two mirrors per eye, and had the screens placed on either side of the head, as to not make the headset so front-heavy and to avoid image processing (mirroring the display). It would be crazy-wide though. Either of these designs might end up very heavy and too unwieldy for head-tracking.

There's another coming-soon HDMI 1.4 3D demultiplexer, the Cypress CH-322, at $300 for a necessary pair. That's a lot less than a dual Optoma 3D-XL setup that some people use for a dual-projector polarized 3D home theater. Maybe some months later there will be even cheaper China crap for this purpose.

I was also trying to find small displays that accept digital inputs, and all I came up with are 7-inch "field monitors" meant to be equipped on DSLRs and camcorders. 800x480 ones run $200-$400/ea depending on how Chinese you want it. Marshall produces a 5" one for $500. SmallHD makes a 720p one for $900. It's all priced for pros, I guess.

So I'll probably put my 3D HMD prospects to rest until some cheaper alternatives arrive, but things are certainly looking up...

[seanny]

I tried out a 40 diopter (which people say is like 30 degrees?) press-on prism fresnel lens. It works, but man is it blurry-- chromatic aberration up the ass, and a general blurriness that I can't attribute to anything. Is it the fact that it's fresnel, a floppy piece of plastic and not a real piece of glass, or both? It also has a viewing angle of sorts, so if placed closed to the eye, the thing has to be curved (like a cylinder, on the axis the prism does its thing). But even if it was ultra sharp, the chromatic aberration is a beast, so I'm thinking a mirror-based design is the way to go. I don't know enough about optics to make a lens set with all the corrections involved.

Apparently some field monitors incl. the cheap Lilliput ones can flip the image internally, which would make a 1-mirror (per eye) HMD possible without requiring a separate video processing stage to flip the video.

[seanny]

I messed around in a raytracer for a minute to virtually prototype the design. It kinda works, but all the sizes are guesstimated so who knows how this'll pan out IRL. I need to grab some small mirror plates from somewhere to experiment some more.

As you can see, one of the problems is the user's face being reflected back to the user on the extreme edge there. I don't know how much that'll occur IRL, but at least in the virtual prototype it limits the FOV you can achieve--

If you angle the mirror toward the user's face for more FOV, it reflects more of the face. If you angle the mirror away from the user's face to avoid this issue, you have to use a longer mirror and therefore the monitor must be placed further away, decreasing FOV and potentially messing up convergence (ugh). If corrective lenses are thrown into the rig, will those get reflected back too?

But even with these constraints, it still seems like you can achieve a decent FOV... ~66-ish in my retarded non-scale prototype. Maybe less IRL, but 66 is a good starting point even for an overly optimistic estimate.

[AntiCatalyst]

the canon 500d has a 3 inch 4:3 screen at 920000 pixels which is sold as a replacement part.. any chance of getting a pair of those working with a PC?

edit: meh, turns out it's just a VGA res screen, they just counted subpixels instead of real pixels.

[seanny]

I'm looking at that HV056WX1 1280x800 5.6-inch panel that's been used in some other single-screen DIY projects in this forum. For now though, I'm waiting for those mirror plates I ordered to come in so I can prototype a design (some printouts can stand in for the panels, heh).

[PalmerTech]

Just a side note: I have noticed that a lot of cheaper HMDs do not use proper front surface mirrors. Instead, they use what appears to be aluminized plastic sheets, very thin and slightly flexible. They are actually very high quality, even in comparison to real FS mirrors, and definitely beat normal mirrors. I am trying to track down a good supplier, if anyone has a good place, post it here.

Here is what I am working on:

[ERP]

Just a side note: I have noticed that a lot of cheaper HMDs do not use proper front surface mirrors. Instead, they use what appears to be aluminized plastic sheets, very thin and slightly flexible. They are actually very high quality, even in comparison to real FS mirrors, and definitely beat normal mirrors. I am trying to track down a good supplier, if anyone has a good place, post it here.

Here is what I am working on:

Are you making any attempt to resolve the chromatic aberration any sufficiently strong lens pair will introduce?
You're going to need something with a combined focal length in the 35mm range.
I've been messing around using variations on leep eye piece design, and the reason they add a 3rd positive miniscus lens next to the screen is to remove the chromatic aberration introduced by rest of the eyepiece design.
So far I've found lens that will counter some of it, but not completely removed it.
I'm just interested in if/how you plan to resolve it.

EDIT - I just realized that those aren't actually lens in your diagram. I'm still interested in the optics design.

[3dvison]

Why can't the 3D viewers be used in a DIY HMD with 2 LCD screens like the 5.6 1280*800 ?

Here is a link to a nice over under viewer.
http://www.rmm3d.com/viewers/vm.econ.html" onclick="window.open(this.href);return false;

[cybereality]

Why can't the 3D viewers be used in a DIY HMD with 2 LCD screens like the 5.6 1280*800 ?

Here is a link to a nice over under viewer.
http://www.rmm3d.com/viewers/vm.econ.html" onclick="window.open(this.href);return false;

Thats a great idea. I hadn't thought of that before.

[3dvison]

Yea, I think the over under viewer would be more compact.
Below is a link to a viewer where the LCD's would be placed side by side and would need a bit more distance in front of the viewer, so I think it would be more front heavy.
http://www.berezin.com/3d/wheatstone.htm" onclick="window.open(this.href);return false;

[cadcoke5]

Be sure to check out all the products at that site. They have quite a variety, at a variety of prices and features. They don't give a "field of vision", in their specs, but do give image sizes for the photographs. That should make choosing one to fit a display fairly straight forward. They may be interested to talk to you experimenters, and may have some good advice for you.

Joe Dunfee

[3dvison]

Be sure to check out all the products at that site. They have quite a variety, at a variety of prices and features.Joe Dunfee

Hi Joe Dunfee,
Do you see any 3D viewer that looks good to you, to try in a DIY HMD with two LCD panels ?

[mAchiNE]

Heres a side by side viewer from the same site as the over under viewer: http://www.rmm3d.com/viewers/vm.6x6.html" onclick="window.open(this.href);return false;
I think that if used with the 5.9" Vitrolight displays the displays can be positioned closer to the viewer because the viewer is designed for looking at prints larger than these displays, could solve the problem of side by side for larger displays but no mention of FOV. By the looks of it I'm guessing each side of the viewer has 2 mirrors to seperate the image and possibly some sort of lens or filter in the middle? would be interesting to see what they look like inside to see how they actually work.

[cybereality]

What about something a little more lightweight, like this:
http://www.berezin.com/3d/kmq_viewer.htm" onclick="window.open(this.href);return false;

[mAchiNE]

or this for side by side http://www.berezin.com/3d/3dprism.htm" onclick="window.open(this.href);return false;

[Okta]

Heres a side by side viewer from the same site as the over under viewer: http://www.rmm3d.com/viewers/vm.6x6.html" onclick="window.open(this.href);return false;
I think that if used with the 5.9" Vitrolight displays the displays can be positioned closer to the viewer because the viewer is designed for looking at prints larger than these displays, could solve the problem of side by side for larger displays but no mention of FOV. By the looks of it I'm guessing each side of the viewer has 2 mirrors to seperate the image and possibly some sort of lens or filter in the middle? would be interesting to see what they look like inside to see how they actually work.

Looked into these options long ago. They could work but you will have a tiny field of view.

You can view 8 inch wide prints from a distance of 1-2 feet. Larger prints are viewed from a greater distance. For example, 11 inch wide prints are viewed from a distance of 2-3 feet and 16 inch wide prints are viewed from a distance of 5-6 feet.

[cadcoke5]

I suspect that the idea is to get one which accepts photos that are the same width as the width of your two screens side-by-side. While several are in fixed locations, and the dimensions are given, some of them are adjustable. But, the field of vision is, to me, the biggest issue even though that is not something that is documented. Here are three that might be worth looking at;

3Dscope [$31 USD] "is a parallel image viewer that can view left and right pairs and has a knob which adjusts the pitch of the mirrors. It will work with pairs from around 5" to very large. "

The KMQ Viewer [$10] is just a piece of plastic that you put in front of your eyes. "for viewing over/under stereo pairs (def). This works well for panoramic pairs which are too wide to be viewed in side-by side format"

Hyper-View Large Format Stereo Print Viewer [$165]"The Hyper-ViewThis Large Format Stereo Viewer was designed and manufactured in order to provide the highest quality stereo image currently available. It does this by eliminating lenses from the system and allowing the viewer to see a large unobstructed image (up to 11 inches wide, 22 inches for the pair)."

Note that there is also an article from the creator of this above system that is worth reading. He says, "The difference between the Hyper-View Large Format Stereo Viewer... is that the others cannot cover a very large field of view. Our viewer covers a 10" square image from an eye position of 15""

I think this 15" distance is not the actual distance from the eye, but only of the device to the image. The eye-to-image distance is increased by the mirrors. But, it shows the field of vision is definitely less than 36 deg.

On the glossary for the web site, I noted a drawing which gives dimensions of the various stereoprint photos. I imagine any stereo viewer capable of showing something as large as your two screens, that would work for them. It is the field of vision that is an unknown. http://www.berezin.com/3d/Glossary.htm" onclick="window.open(this.href);return false;

Talking to them is the best thing to do. They are stereoscopic enthusiasts, so I would think they would greatly enjoy talking to an experimenter. Though, it is also possible they have to do a lot of explanation to novices, and have to guard their time. You are a person who is trying something new, and should set you apart from the crowd.

Joe Dunfee

[seanny]

also something to notice on that site... you can buy stereoscope lenses straight up-- http://www.berezin.com/3d/vieweracces.htm#Lenses" onclick="window.open(this.href);return false;

though 20mm diameter for the achromatic lenses seems pretty small, and the prismatic ones are meant to be used in viewers like these, which leads me to believe they aren't meant to be used in a high-FOV system, but I dunno as I've never actually owned an HMD or a stereoscope beyond a Viewmaster (toy).

[seanny]

I've been playing around with a 5" circular 2nd-surface mirror and a 7" printout of an Avatar screenshot, trying to imitate the setup illustrated here for one eyeball, just by holding stuff up with my hands (I'd need another mirror and rig to test both eyes). I can get a pretty damn decent viewing angle with enough effort. I tend to sit 35" away from my 42" HDTV on my desk (I use it as a computer monitor), and I can get something like it perceptually from my goofy prototype.

Apparently firstsurfacemirrors.com will cut you rectangular and circular mirrors. A 3x5" @ 1.1mm might be good for the 5.6" panel, but I might get 4x6s to have some margin of error. (Or you can cut and strip FS mirrors yourself, apparently)

But even with my myopia & my eyeglasses off, I find it hard to focus at such a close-up object as where the screen should ideally be. I'll have to look into optics that can change the focal length (I think?), which is a whole new bag of worms because I know very little about optics and where to buy them. Ideally it would be large (so as not to impede viewing angle) and somehow adjustable so that people with eyeglasses don't have to use them with the HMD. Or maybe with large optics that can magnify, you can get into immersive FOV territory and use smaller mirrors.

[Okta]

I've been playing around with a 5" circular 2nd-surface mirror and a 7" printout of an Avatar screenshot, trying to imitate the setup illustrated here for one eyeball, just by holding stuff up with my hands (I'd need another mirror and rig to test both eyes). I can get a pretty damn decent viewing angle with enough effort. I tend to sit 35" away from my 42" HDTV on my desk (I use it as a computer monitor), and I can get something like it perceptually from my goofy prototype.

Apparently firstsurfacemirrors.com will cut you rectangular and circular mirrors. A 3x5" @ 1.1mm might be good for the 5.6" panel, but I might get 4x6s to have some margin of error. (Or you can cut and strip FS mirrors yourself, apparently)

But even with my myopia & my eyeglasses off, I find it hard to focus at such a close-up object as where the screen should ideally be. I'll have to look into optics that can change the focal length (I think?), which is a whole new bag of worms because I know very little about optics and where to buy them. Ideally it would be large (so as not to impede viewing angle) and somehow adjustable so that people with eyeglasses don't have to use them with the HMD. Or maybe with large optics that can magnify, you can get into immersive FOV territory and use smaller mirrors.

If you work out a way to reverse the lcd display so it shows correctly let me know.

[PalmerTech]

The new IZ3D drivers have both horizontal and vertical flip, so you can do that in games.

[seanny]

If you work out a way to reverse the lcd display so it shows correctly let me know.

A large half-penta prism (or "45-degree prism") is what I'm currently trying to track down as an alternative to a mirror.