DIY Oculus Rift - schematics, instructions and build pics!

[Naru]

Your build looks really great, Naru. I love the curved eye-pieces. Are they crafted specifically for your face? How close can you get your eyes to the lenses?

I originally crafted the eye-pieces to my face, but I had a few other people try it on and it fit their faces perfectly too. I even had my 5 year old nephew try it on and it fit nicely even with his smallish ipd still being very centered on each eye hole. I think most people have very curved faces on the sides and it's why ski goggles are like that. The only problem would be for people that have sharper curves in the forehead/eyebrow area.

I got the lenses pretty close, but not to where my eye lashes are hitting them. I'm thinking a quarter of a cm more and it would get to that point. If I had managed to build a custom housing for each lens instead of using the black rings, I would have to deliberately move them away to where they aren't touching my eyelids.

[cybereality]

@3DHMDGuy: That sounds very interesting. I can't wait to see more of your setup.

[2EyeGuy]

What I have developed myself now is a way to capture real video in 360x360 panoramic format also in true 3D (not fake).

That's incredible! Well done. We've been wondering how to do that and figured it was too hard.

When you say "true 3D", do you actually mean true 3D, where you can see behind objects a bit at the edges? Because I wouldn't have thought that was possible. I assume you are just measuring the depth and colour of each pixel, which isn't quite enough to replicate what our eyes truly see in terms of 3D. You'd have to have a pair of cameras that spin around their centre, sort of like our eyes do, to see pixels behind other pixels from the correct angles.

Nobody knows exactly how to warp for the Rift dev kit, except Dycus and PalmerTech, because they still haven't released the exact specs for the improvement's they've made. But we can take some educated guesses. The DIY Rifts will mostly be different though because people went with the smaller 5.6 inch screens before they announced the change to 7 inch screens. Here's what the Doom 3 BFG source code does (suitable for DIY rifts): https://github.com/id-Software/DOOM-3-B ... Warp.pixel and there's another thread on here somewhere where we collected the various open-source rift warping implementations, but I don't remember where. I wouldn't crop the top and bottom of the screens like Doom 3 BFG does though.

[cybereality]

@3DHMDGuy: Check the warping equation given in this post:
http://www.mtbs3d.com/phpBB/viewtopic.p ... =60#p74912

Thats the one I've been using and its been tested on the Rift and said to work OK.

[djdevin]

I am ashamed to admit that I discovered the reason for my being able to see so much of the top and bottom of the screen: The rez for the LCD was set too low, vertically. I had it at 1280x800 (which is what I thought the LCD's native resolution is...have to investigate further), but it turns out that I have to set it at 1280x1024 to get images to fill it vertically. There may be more to this than meets my eye, though (pun happily intended).

Can you try my monitor .inf file?

We have the same display and I think the EDID stored in the LCD makes it do this (it was really geared to handle 720P). And it doesn't seem to have a mode setting for 1280x800, and falls back to some other rez.

With this .inf the default recommended resolution should be 1280x800@60hz and it should also fill the screen. The OSD should also say 1280x800.

I've been using this since I got the display and games etc that use 1280x800 work fine now.

[Mel]

I am ashamed to admit that I discovered the reason for my being able to see so much of the top and bottom of the screen: The rez for the LCD was set too low, vertically. I had it at 1280x800 (which is what I thought the LCD's native resolution is...have to investigate further), but it turns out that I have to set it at 1280x1024 to get images to fill it vertically. There may be more to this than meets my eye, though (pun happily intended).

Can you try my monitor .inf file?

We have the same display and I think the EDID stored in the LCD makes it do this (it was really geared to handle 720P). And it doesn't seem to have a mode setting for 1280x800, and falls back to some other rez.

With this .inf the default recommended resolution should be 1280x800@60hz and it should also fill the screen. The OSD should also say 1280x800.

I've been using this since I got the display and games etc that use 1280x800 work fine now.

Thanks, djd. My DIY Rift is out on loan, so I will pass the info on to the persons using it (Tom, this means you, if you're reading this )

[jaybug]

I bought the power supply from chinatobby and he packed it in the same box. 12VDC 4A MAX

Shucks, should have done the same. I just got the screen.
I tried the one adapter that fit (9V 450mA) , and it actually worked for a little while, but now the screen restarts constantly. Now, is that because the voltage is too low, or is it the ampere? And how much Ampere do I need?

[Chriky]

I think its the ampere. Interestingly I can run mine off USB with HDMI but not USB with VGA. I guess the HDMI cable carries some power as well.

[jaybug]

Found a 12V 2A adapter. Works like a charm

[jaybug]

Hooked up a quick test rig. The IPD was too wide, the lenses were centered too low, a tad too close to the screen, and the skimask eventually fell off. But otherwise pretty cool :p

[cybereality]

Looking good.

[LaserEdge]

Here is my first HMD build. The goal of this build was to prove that a LCD panel in the 10" class could be used with Rift style optics. The panel itself is 1920x1200 with 16.7 million colors. If the educated guess was correct the technology to build a consumer level version of a Rift display already exists today.

So far I have tested the build with Warp Tune Alpha, StereoScopic player and Doom 3 BFG. Right now the selection of software to test with is limited because this panel requires negative convergence (sometimes called parallax). So far the subject comments I have gotten from friends and family I have showed it too is "It feels like you are there." My own personal experience is I have found myself wanting to reach out an touch things with the 1080p Half SBS clips I have played.

I estimate the FOV to be in the neighborhood of 150 degrees. If I want to look to the edge of the screen I really have to try hard to get my eyes to do so.

So far two negatives with the build. There is some keystoning, but only noticeable at the far horizontal edges of the screen. The weight is around 60 grams more than 7" class. Perhaps some comfort issues over a 7" display. I can't say. I don't have a 7" inch build to test against. Negative convergence is required, but will be required for the majority of users of a 7" panel like the dev kit. Only the 5.6" panels can get by with a convergence of 0.

Busy day today for me. Post questions/comments and I will respond as I have time.

[Fredz]

Very nice HMD indeed !

Did you measure the weight of the whole ? I see that you have only one strap attached to the head, don't you feel the need of an additional over the top strap ? Did you try your HMD for long sessions ?

Also the arrangement of the lenses is not clear with your photos, did you do like most other DIY HMDs with lenses inserted on a plate, or are they secured at the top of cones so that the nose and the forehead don't prevent the eyes from being at the closest possible distance from the lenses ?

Are you using the same 5x magnifying lenses than rfurlan ?

[MSat]

Super cool, LaserEdge!

It doesn't look anywhere as bulky as I imagined. Could you go into some more detail about the panel you used, and the interface box?

[Dycus]

Ten inches!? Holy cow, that's huge! Looks pretty good, though, if I do say so myself. Is the 150 degrees an estimate, or did you measure it somehow?

[LaserEdge]

Very nice HMD indeed !

Did you measure the weight of the whole ? I see that you have only one strap attached to the head, don't you feel the need of an additional over the top strap ? Did you try your HMD for long sessions ?

Also the arrangement of the lenses is not clear with your photos, did you do like most other DIY HMDs with lenses inserted on a plate, or are they secured at the top of cones so that the nose and the forehead don't prevent the eyes from being at the closest possible distance from the lenses ?

Are you using the same 5x magnifying lenses than rfurlan ?

I will measure the weight when I get back to the office. So far my sessions have only gone for an hour. It does get a bit uncomfortable on the nose after an hour. A top strap would help.

The lens are inserted into the plate so I am not able to get my eyes right up next to the lens. This is something I would like to redo.

The lens I am using are different. I am using 60mm aspheric lens made by Donegan optical. They claim 5x magnification, but it is more like 4.3 based on the focal length I measured.

[LaserEdge]

Super cool, LaserEdge!

It doesn't look anywhere as bulky as I imagined. Could you go into some more detail about the panel you used, and the interface box?

I am using the AUO B101UAN02.1 panel. You can get it for $60 on eBay. The controller board is the RTD2662 board made by njytouch. This board can run with 24 volts so it pairs well with required backlight voltage for the panel. I had reprogram the firmware to get it working with the panel.

[LaserEdge]

Ten inches!? Holy cow, that's huge! Looks pretty good, though, if I do say so myself. Is the 150 degrees an estimate, or did you measure it somehow?

Don't have any professional device to measure FOV. What I did was a prebuild test with Legos and some test patterns I made with Adobe illustrator. It let me work out much of the required measurements before I designed the HMD in Autocad. Using the test pattern length and width plus distance of lens center from test pattern and estimate of eye distance from lens I can calculate the FOV using trigonometry. Atleast it gets me a decent ball park figure.

[cybereality]

Sounds amazing LaserEdge! Nice job.

[Kajos]

That screen is very popular in the DIY beamer market as well, atleast I think it's the same one.

[MSat]

There are a bunch of 9" screens with that resolution. I wonder if it would make a significant difference in overall size/weight though.

Regardless, LaserEdge now has the highest resolution Rift clone. Congrats!

[LaserEdge]

There are a bunch of 9" screens with that resolution. I wonder if it would make a significant difference in overall size/weight though.

Regardless, LaserEdge now has the highest resolution Rift clone. Congrats!

The real issue as I see it is software. Going to bigger sized panels requires software support for negative convergence. The display is unusable without it. On top of it you still have aspect ratio and warpping issues. Keystoning should be easy to fix in software. Need to try Cybereality's driver when he has it ready. Hoping he has a way to set to negative convergence.

[ChrisP]

wondering if I could get a quick favour. could someone give me the dimensions of the viewable area on the 7" N070ICG-LD1 screen?
(ex: ??mm x ??mm)

@LaserEdge : nice work!

[Matoz]

wondering if I could get a quick favour. could someone give me the dimensions of the viewable area on the 7" N070ICG-LD1 screen?
(ex: ??mm x ??mm)

@LaserEdge : nice work!

http://www.vslcd.com/Specification/N070ICG-LD1.pdf

[ChrisP]

thanks for the quick reply.

I have that document. It only gives the overall screen size, physical corner to corner. If I base my lens placement on that it will be off by however much of that measurement is unused viewable area.

[Matoz]

thanks for the quick reply.

I have that document. It only gives the overall screen size, physical corner to corner. If I base my lens placement on that it will be off by however much of that measurement is unused viewable area.

In the Appendix of the document you will find all the measurements in the drawings. Those match my display.

[ChrisP]

(smacks forehead with palm)... Thanks Matoz.

[3dvison]

Hey Cybereality,
Will your driver support the negative convergence needed for the 7" 9" and 10" inch panels being talked about here ?
I know you want to focus your time on RIFT support, but would a negative convergence feature be somthing that would be cross compatible between the RIFT and the DIY builds, and not take up anymore of your time ?

[LaserEdge]

I measured the weight of my AUO B101UAN02.1 Rift build. It is 360g (12.85oz). Here are two more pictures so you can get a better idea of its size and shape.

[LaserEdge]

Here is a picture the 5x Pocket Loupe used in the original rfurlan build and the Donegan Optical 60mm 5x Aspheric Lens. The pocket loupe is on the left and the Donegan lens is on the right. Using a larger lens increases FOV while at the same time reducing distortion. Even though magnification is reduced with going to the bigger lens it isn't reduced less than a proportionate increase in the display size over the original rfurlan build.

[cybereality]

@3dvison: I have controls in my driver that allows to shift the image either left or right without changing the separation or angle. This is what I am calling convergence, but it is not exactly the same as convergence in the Nvidia or other drivers. I believe this should allow for non-100% overlapping displays like these DIY Rifts, but I have not been able to verify this myself. I am planning on getting a version of my driver out very soon, so the wait will not be too long.

[LaserEdge]

From my perspective doing initial testing of the optics is vital to having a successful Rift build. Since Rift builds usually only have one lens pair it is fairly easy to test.

The most important feature to test of the lens is focal length. The center of the lens must be placed focal length distance away from the panel to achieve the desired pixel at infinite focus.

We also need to test the lens's ability to uniformaly focus the entire viewing area as well as chromatic aberration if any. After that geometry distortion needs to be tested. Finally range of usable IPD placement should be determined. Finally calculations of approximate FOV can be done.

To test these things I create my own test patterns with Adobe Illustrator. Because Illustrator uses vector graphics I can make very precise test patterns which I print out on the printer. By doing this we can evaluate a potential panel by creating a test patterns that matches the panels active dimensions without actually buying it.

I used Legos to create a stand so I can control lens placement. Legos gives a 3mm increment control in focal length. To get finer control I used small cuts of paper as needed. See the pictures below of my optical test builds.

I take the measurements from these optical tests and use them in the Autocad design for the housing. In this way an error free one pass Rift build is possible.

[LaserEdge]

@3dvison: I have controls in my driver that allows to shift the image either left or right without changing the separation or angle. This is what I am calling convergence, but it is not exactly the same as convergence in the Nvidia or other drivers. I believe this should allow for non-100% overlapping displays like these DIY Rifts, but I have not been able to verify this myself. I am planning on getting a version of my driver out very soon, so the wait will not be too long.

Glad to hear cyberreality. I view your work as vital to the success of this technology. I will test and give you feedback with my build as soon as you are ready. Thanks for your efforts.

[bobjwatts]

Here's my build so far.

It's working really well and I have played a heap of games with it, it's a bit weird getting used to aiming with your head. I got MS flight sim working with mouselook and that was pretty amazing, until I started getting nauseous when the plan rolled. I also played the racing sim GTR with mouselook which is really fun, unfortunately the majority of racing sim mouselook implementations only allow you to look around a small amount.

The 3D effect looks amazing with this screen.

After 3 failed attempts to make a cable from the controller to the screen I ended up mounting it to the top, which adds plenty of weight.

I used the strap from the goggles for a while but went back to my welding mask brace, it enables me more ability to choose where the weight falls.

It has fully adjustable IPD which is pretty critical for a DIY given the fault tolerance of glue and board.

Here's the pics!

[PalmerTech]

That looks great!

I have used those Donegan lenses before, a suggestion: You can get the lenses centered on and closer to the eyes a lot more easily if you sand off the edge of the lens by the nose area. I just ran my lenses along a belt sander, but using normal sandpaper or even a file would work fine.

[cybereality]

Nice work, bobjwatts!

[LaserEdge]

Here are some Doom 3 BFG screenshoots running on my build. The settings are
stereoRender_convergence -6
stereoRender_warp 1
stereoRender_warpTargetFraction 0.9
stereoRender_swapEyes 1
stereoRender_interOccularCentimeters 12.5
g_fov 110

As far as I can tell setting stereoRender_convergence to any negative value requires stereoRender_swapEyes to be set to 1. Also the reason for stereoRender_interOccularCentimeters set to 12.5 is that I wanted to experiment with shift the lens outward versus my IPD to see what the tradeoffs were in my first build. Doing so increases FOV, but creates some undesirable geometry distortion that is noticed when looking left or right of center. On my second housing build I will move the lens closer together, but that will create some left/right overlap and decrease FOV. Ideally it would be nice to build a housing so the lens can be adjusted inward and outward. I imagine each user will have their preference with this tradeoff.

[LaserEdge]

That looks great!

I have used those Donegan lenses before, a suggestion: You can get the lenses centered on and closer to the eyes a lot more easily if you sand off the edge of the lens by the nose area. I just ran my lenses along a belt sander, but using normal sandpaper or even a file would work fine.

Thanks for the suggestion. I can imagine how that can help. I originally designed it this way wanting to increase FOV, but for my personal tastes the increased geometric distortion when looking off center isn't desirable. Also it would help to add more viewable image when looking near the nose too. Even despite the negatives I am pleased with the results of my first build. Now tweaking begins.

[LaserEdge]

Your build is coming together well bobjwatts. How is the comfort level so far? How long can you use your build before you feel like you need a break?

[xef6]

That looks great!

I have used those Donegan lenses before, a suggestion: You can get the lenses centered on and closer to the eyes a lot more easily if you sand off the edge of the lens by the nose area. I just ran my lenses along a belt sander, but using normal sandpaper or even a file would work fine.

What sort of precautions did you take to deal with the dust? I'd like to do this, but glass dust sounds very unfriendly.