DIY Oculus Rift - schematics, instructions and build pics!

[Silversurferx]

Hi,
to all here in this very amazing forum.(there are some geniuses underway)
I´m observing this thread for a while, searching for good idea´s to get
the rift up to HD resolution (I play most of all flightsim´s) so I found Jotschi´s idea very good.

Are you able to diverge your eyes at such a large angle like that? Honestly don't think that's safe for long periods of time.

But I hope not ending like this guy

Queral_Marty_Feldman.jpg

Jotschi isn´t it possible to correct the lenses with a prism or use rectangular lenses ? (greater fov?)
http://www.lupenhandel.de/product_info. ... r-5x-.html
http://www.lupenhandel.de/product_info. ... nomy-.html

Panel Brand : ORTUSTECH
Panel Model : COM48T4Mxxxxx
Panel Size : 4.8 inch
Resolution : 1920x1080
Display Colors : 16.7M (8-bit)
Signal Interface : LVDS (2 ch, 8-bit)
Input Voltage : 5.0V (Typ.)

Trying to buy a couple, will keep you all posted

rfurlan had you any Luck or is this too soon (vaporware)?

Regards
Silversurferx
PS. Sorry for my bad english

[Jotschi]

Interesting design. I assume that you have worked out the angle and position of the lenses such that they are acting like prisms, hence no need to actually angle your eyes out? I'm trying to work out the optics of it, since I would have thought that convex lenses positioned like that would actually bend the light the other way, unless you have mirrors in the middle there.

No, i will handle that problem with a special shader. I hope that this will work. Each image will be rotated for about 20°.

[PatimPatam]

Hmm this design looks familiar!

http://www.mtbs3d.com/phpBB/viewtopic.p ... 4&start=72

I really hope you can make it work Jotschi!!

[Silversurferx]

Hmm this design looks familiar!

http://www.mtbs3d.com/phpBB/viewtopic.p ... 4&start=72

I really hope you can make it work Jotschi!!

+1
Thanks for the link PatimPatam this thread answers alot of questions to me
I´ve got to learn alot before my dream comes true
Regards
Silversurferx

[android78]

Interesting design. I assume that you have worked out the angle and position of the lenses such that they are acting like prisms, hence no need to actually angle your eyes out? I'm trying to work out the optics of it, since I would have thought that convex lenses positioned like that would actually bend the light the other way, unless you have mirrors in the middle there.

No, i will handle that problem with a special shader. I hope that this will work. Each image will be rotated for about 20°.

Won't that mean that you have very little overlap between the two images though?

[TheLostBrain]

What up all!

Couldn't help but notice some of the misconceptions regarding divergent partial overlap...although I really do dig these pics of the poor souls with diverging eyes lol.

So... this outward-angled 'divergent' approach has actually been one of the most common ways of implementing partial-overlap in HMDs for decades at this point. Hell, just take a look at the HMD in my avatar - The Datavisor 80 (originally built in the late 90's I believe) has a monocular FOV of 80 deg @ 100% overlap... but with the oculars rotated outward (as depicted in that image) we get 120 deg. @ a 50% overlap.

Jotschi is right on the money - In a divergent partial overlap configuration when we rotate the screens outward we don't (thank god) need to rotate our eyes outward to compensate. The compensation is handled by simply rotating the corresponding left and right viewports by a complementing degree.

As for the amount of overlap that will be available w/ these screens we're using in our DIY designs... well it just depends on how everything (screens, optics, imagery etc.) is configured. I believe Jotschi made mention of only using a portion (the inner portions - in front of the eyes) of each screen for actually rendering image data. In that configuration it's possible to have 100% overlap (as we're only rendering using the screen real-estate directly in front of the eyes)...and by rotating the screen outward we can achieve overlaps of 25%, 50% or anything in-between and beyond (50% is the usual) with a corresponding increase in FOV. Note: For an even wider FOV (and more resolution) you could use more of the physical horizontal screen real-estate... but at the cost of your maximum overlap and lesser, your max vertical fov.


The Wide 5 (reviewed in the link below by Palmer himself last year) appears to be implementing divergent partial overlap and is likely achieving most of the horizontal fov by simply employing the full field of both screens but with a relatively small overlap region. And those 'sexy' optics... they are likely nothing more than a pair of large aspheres w/ a chop taken out of them for nose relief..maybe cast from a high abbe, high-index material to keep weight down.
http://www.mtbs3d.com/phpBB/viewtopic.php?f=120&t=13741


I believe the Wide 5 to be an excellent example of what we (as a DIY community) could likely achieve and happens to fall pretty closely in line with the direction Jotschi is already taking with his design.

Also just thought that I'd mention... if you've got some ideas custom optics are not out of our reach as can be seen in an earlier post I made: http://www.mtbs3d.com/phpBB/viewtopic.p ... 270#p81276
Heck

Once I have a bit more time (and wrap up this last Rift clone design I'm working on) I believe I'm also going to join this DIY 'multi-screen partial-overlap' bandwagon Jotschi has started.

BTW: Here are a couple pretty good resource for learning more about Partial Overlap implementation in HMDs:
http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA392013
http://kirkmoffitt.com/hmd_image_configurations.pdf
http://www.usaarl.army.mil/TechReports/99-19.pdf
http://www.nvisinc.com/literature/M-Series_Manual.pdf (Page 20, 21 ...)

[PatimPatam]

Thanks a lot for the explanation TheLostBrain, so it seems like my idea was not that crazy in the end..

I think another advantage of having a 2-screen design is that you could use smaller displays than the required 6 to 7 inches of a single-screen HMD like the Rift. If we move to a 4 to 5 inch range it will probably be a lot easier to find a wide variety of 1080p displays for instance.

[mwarren]

Couldn't help but notice some of the misconceptions regarding divergent partial overlap...although I really do dig these pics of the poor souls with diverging eyes lol.

So... this outward-angled 'divergent' approach has actually been one of the most common ways of implementing partial-overlap in HMDs for decades at this point.

Great links! I'd like to add one more - LEEPVR's web site has a lot of good info and this article explains why only partial overlap is necessary once the display's FOV is wide enough (our eyes only have partial overlap).
http://www.leepvr.com/periphery.php

[zacherynuk]

Partial overlap.

Can anybody point me towards a 3d driver, demo or similar which actually processes partial overlap ?

I have been looking for years...

[TheLostBrain]

Partial overlap.

Can anybody point me towards a 3d driver, demo or similar which actually processes partial overlap ?

I have been looking for years...

At one point I accomplished a hacky (but usable approximation) w/ IZ3D by:
1) Decreasing convergence until each view frustum was pointing outward
2) Decreasing separation until items directly in front (in close proximity) were in back in view
3) .... Or something like that... lol

...Or... was it... decreasing separation until the left and right eyes swapped and THEN working w/ the convergence...and possibly also swapping video inputs as well....

Lol.. I can't remember... was literally years ago. I just know after a bunch of messing w/ it I was able to get something remotely usable to come up w/ the DV 80.

[Silversurferx]

Thank you TheLostBrain and mwarren very informative links.

If we move to a 4 to 5 inch range it will probably be a lot easier to find a wide variety of 1080p displays for instance.

Unfortunately not at the moment AFIAK the ORTUSTECH 4.8" COM48T4Mxxxxx is the only one
in this size but don't know where to buy and price?

There are several 720p but with MIPI interface
http://www.panelook.com/sizmodlist.php? ... e=30&sizes

and one with an adapter to lvds coming soon.
http://q-vio.com/AMOLED_displays.php

Regards
Silversurferx

[mwarren]

The iPad gen 3 screen (9.7", 4:3, 2048x1536) has a resolution of 264 dpi which is nearly as high as the Hydis HV056WX1-100 (5.6", 16:10, 1280x800, 270dpi). The iPad also includes an orientation sensor, magnetometer, gps, GPU, etc... Has anyone tried using the iPad 3 with a set of the 5x aspherical lenses suggest by rfurlan/PalmerTech? A 3d iPad demo running at native resolution with proper SBS, warped rendering, head tracking, etc... would be great, but just viewing some test images would probably give some insight more quickly. Thanks!

[PatimPatam]

@Silversurferx

Yes when i said "it will probably be" i meant in the near future! (i hope!)


@zacherynuk

It's funny how sometimes it's hard to communicate properly between hardware and software people! when we discussed my post in the other thread i was taking as a given that you would need to render each view with a different view-plane normal (vpn) to match the screen/lenses angles.. if i understood correctly that's all you need to "process" partial overlap.

It shouldn't be too difficult to do a little demo with opengl (or unity maybe).. I'm quite busy at the moment with another project but maybe you could ask Owen to modify his "armored ops" prototype for you? As it happened with Palmer and Carmack, you need the right combination of HW and SW to archive something awesome!

[TheRealistWord]

The iPad gen 3 screen (9.7", 4:3, 2048x1536) has a resolution of 264 dpi which is nearly as high as the Hydis HV056WX1-100 (5.6", 16:10, 1280x800, 270dpi). The iPad also includes an orientation sensor, magnetometer, gps, GPU, etc... Has anyone tried using the iPad 3 with a set of the 5x aspherical lenses suggest by rfurlan/PalmerTech? A 3d iPad demo running at native resolution with proper SBS, warped rendering, head tracking, etc... would be great, but just viewing some test images would probably give some insight more quickly. Thanks!

As soon as I received my 5x aspherical lenses, the first thing I attempted was using them with my iPad(3rd gen)'s display. I tested out the Tales from the Minus Lab app formatted in SBS for the FOV2GO, but because the app was created specifically for the iPhone's display, it's scaled up for the iPad (which means, of course, it wasn't in the iPad 3's native crisp, high resolution ) Still, the experience was pretty immersive! I didn't build any enclosure or anything to hold the iPad and the lens - I just managed to hold the lenses and the iPad in my hand at the correct distance to focus, and since the app makes use of the iPad's gyro, I'm able to rotate my head in all directions (and wow, I haven't had much experience with other head tracking devices, but the iPad's gyro/accelerometer seem extremely accurate with little lag as far as I can tell). But from what I tried, it was an impressive experience! It makes me even more excited for the Rift!

This weekend, I'm going to attempt to rig something up to hold my iPad and the aspheric lens, and I'm thinking that perhaps I'll stream my desktop's image to my iPad so I can use any video stream, not just restricted to apps purely on the iPad (I'm afraid of the lag though). I know next to nothing though about if it's possible for the iPad's gyroscope/accelerometer to be read directly from the PC without have an apple dev account (or jailbreaking or any other means), I'll have to look into that.

[cybereality]

@zacherynuk: If I'm not mistaken, I think you can do this with my driver using the convergence controls. Basically all it does it control the sheer of the viewport, by adjusting it you just shift the image left or right without changing the perspective at all. I believe this is all you need to achieve partial overlap, but if I'm wrong please explain and I will see if I can help.

[Owen]

Partial overlap on the software side should have the cameras turn outward at the same angle as the optics, then apply barrel distortion as normal. Shearing wouldn't give you correct perspective. Think of it this way, when the lens and screen are turned outward, its no different from looking through the rift with your eye turned inward.

[zacherynuk]

OK - this really isn't the place for this; but Palmer is too busy for me to ask him to split the thread! Can anybody else ?

Partial Overlap Example picture

Anyway, I have lost all the assets I had when I was testing partial overlap - I had it working superb with 2 screens in portrait with still images a good few years ago, but I could not find any software which would do the cropping. CROPPING IS ALL IT IS!

So I have just thrown together this image, for me it works perfect with the 5X loupes and my head size. The rest of you should be able to get alignment without too much issue. If this is your first time partial overlap; you will see why we say it is the only true way to get immersion. It's freakin awsome IMHO.

But here is a partial overlap picture, print out 100% and have a play, it is 21cm accross.



All you are doing is restricting what each eye sees, like in real life - allowing you to use the rest of the screen real-estate for higher non 3D FOV.

I have created a quick post here for reference: http://www.wastedspace.co.uk/cms/2012/0 ... l-overlap/

Whilst I am here - a few quick other points (These may be wrong, but it is my understanding): You do not want to change the angle of each eye camera! You do not look at the world like that, our eyes are separated only, and until we get real time focussing in HMD's nothing else need change when presenting the image.

Professional HMD's which support partial overlap have to have each eyepiece moved into place when this mode is utilised; this is not because the act of diverging our vision increases the FOV, it's because the act of using partial overlap effectively increases the virtual image size - so we need to angle the eyepieces slightly to widen the IPD and make it feel more natural; since the focus point extends.

[TheLostBrain]

CROPPING IS ALL IT IS!

You do not want to change the angle of each eye camera! You do not look at the world like that, our eyes are separated only

That is 100% correct... when the two physical displays are simply positioned side by side in a collinear fashion (essentially resting their backs on the same virtual plane). In this scenario all we need to do is a bit of linear displacement as you have shown.

However, when the screens are angled (convergent / divergent) then the view frustum / eye camera needs to be needs to be angled in a complementing manner. Our eyes still look straight ahead but the scene imagery is rotated to compensate for the screen itself being angled.

Each approach (convergent, collinear, divergent) are all usable and each has it's own set of advantages and disadvantages. Ex: Some reports show target acquisition to better w/ the convergent approach vs the divergent approach... However, I believe the divergent approach inherently covers more of your peripheral vision and thus offers a larger potential horizontal FOV...While the collinear approach is technically achievable via a single screen of proper dimensions.

They all have their place and application. However, I believe I'd like to experiment w/ 'divergent' myself for the inherent peripheral vision coverage.... Test it yourself: Looking straight ahead hold a piece of paper in front of you right eye... now hold the left edge in place and begin pulling the right edge back (essentially turning the sheet clockwise if looking from above).. that same sheet angled in this manner covers much more of your peripheral vision. if this were a real screen you'd just need to rotate the scene data to compensate.

[zacherynuk]

Good stuff, thanks!

However, I believe the divergent approach inherently covers more of your peripheral vision and thus offers a larger potential horizontal FOV...While the collinear approach is technically achievable via a single screen of proper dimensions.

Agreed, it is the best and easiest, but in my testing I was hindered by the angles; really 5-10 was the absolute max I could make comfortably work without ruining exit pupil.

What we need is a bloomin' 3D drivers that will allow us to play with it

[Owen]

Yes, cropping works if you have coplanar displays and magnifying optics. When they are coplanar, the displays would need to be off center with respect to the optics in order to have partial overlap, so an off axis perspective projection is needed in at case as you said.

[TheLostBrain]

Yes, cropping works if you have coplanar displays and magnifying optics. When they are coplanar, the displays would need to be off center with respect to the optics in order to have partial overlap, so an off axis perspective projection is needed in at case as you said.

Coplanar... that is so much more consise than my 'colinear w/ their backs resting on the same virtual plane'... lol. And you'd think I would have used that term myself as I see it (and use it) all the time in SolidWorks lol.

[PatimPatam]

What we need is a bloomin' 3D drivers that will allow us to play with it

Just to clarify, for the divergent approach I doubt this could be done at a driver level, as Owen already mentioned. For the coplanar approach, I'm not sure..

For games to work with angled displays they would need to support this feature at a graphics-engine level. All you need to test this is a simple demo app where the 2 view frustums have divergent angles. I'm sure some people from this forum can help you with that!

[TheLostBrain]

What we need is a bloomin' 3D drivers that will allow us to play with it

Just to clarify, for the divergent approach I doubt this could be done at a driver level, as Owen already mentioned. For the coplanar approach, I'm not sure..

For games to work with angled displays they would need to support this feature at a graphics-engine level. All you need to test this is a simple demo app where the 2 view frustums have divergent angles. I'm sure some people from this forum can help you with that!

In case it was missed...(a few posts back)... For 'divergent' I was able to accomplish a working approximation w/ iZ3d a few years ago.
I *believe* it was simply convergence rotated outwards...and maybe a little adjustment to the separation.
However, I also have some weird hazy memory of doing some funky stuff where I increased convergence and decreased the separation until the views flipped and then swapped the left/right video cables.... Maybe it was just some drunken haze? lol

[zacherynuk]

What we need is a bloomin' 3D drivers that will allow us to play with it

Just to clarify, for the divergent approach I doubt this could be done at a driver level, as Owen already mentioned. For the coplanar approach, I'm not sure..

For games to work with angled displays they would need to support this feature at a graphics-engine level. All you need to test this is a simple demo app where the 2 view frustums have divergent angles. I'm sure some people from this forum can help you with that!

In case it was missed...(a few posts back)... For 'divergent' I was able to accomplish a working approximation w/ iZ3d a few years ago.
I *believe* it was simply convergence rotated outwards...and maybe a little adjustment to the separation.
However, I also have some weird hazy memory of doing some funky stuff where I increased convergence and decreased the separation until the views flipped and then swapped the left/right video cables.... Maybe it was just some drunken haze? lol

But surely for a basic side by side Coplanar setup, all that needs to happen is that the scene from each eye is rendered wider than it has to be and then cropped for each eye. Seems simple enough in my tiny brain! Now I have so many screens and importantly longer lvds cables (and now a smaller board) I can have a go at these things properly; perhaps I'll see if I can create a partial overlap 3D clip so at least I have something animated to play with.

[Owen]

Actually you would use convergence from iz3d for coplanar displays, because they use off axis frustums for convergence. A 3D monitor is basically two coplanar displays that happen to occupy the same physical space after all.

A driver could handle the rotation for divergent optics though, just apply a rotation transform to the projection. Clipping would be a nightmare though.

[MaterialDefender]

Would it make sense to do something like this in a 2d only solution with a 90% FOV like the Rift has or is this only useful with greater FOVs? The lens distortion injection shader I released here two weeks ago crops 25% of the image on both sides when used in aspect ratio 'correction' mode. Doing the cropping like shown in the images above instead should be extremely easy.

If I find the time I will release an updated version with this functionality in the next few days. Would be nice to get some feedback from DIY Rift owners, unfortunately I don't have one myself. Last time the feedback was a bit underwhelming, and sharing is much more fun, when you receive something in return.

[zacherynuk]

Would it make sense to do something like this in a 2d only solution with a 90% FOV like the Rift has or is this only useful with greater FOVs? The lens distortion injection shader I released here two weeks ago crops 25% of the image on both sides when used in aspect ratio 'correction' mode. Doing the cropping like shown in the images above instead should be extremely easy.

If I find the time I will release an updated version with this functionality in the next few days. Would be nice to get some feedback from DIY Rift owners, unfortunately I don't have one myself. Last time the feedback was a bit underwhelming, and sharing is much more fun, when you receive something in return.

Needs to be 3D really... LInk me to your current injector please

[MaterialDefender]

It's just a few threads down. http://www.mtbs3d.com/phpBB/viewtopic.php?f=140&t=15473

[Bretspot]

Hi everyone,
New here, but love the idea of throwing together a cheap "3d Viewer" using some lenses, some hot-glue and a kindle fire.
It would be for viewing 3d images and maybe movies on the web, but not for gaming yet.

Can someone take a look at these lenses and let me know if they think they would work?

http://tinyurl.com/FrysLoup

Thanks!

[zacherynuk]

Hi everyone,
New here, but love the idea of throwing together a cheap "3d Viewer" using some lenses, some hot-glue and a kindle fire.
It would be for viewing 3d images and maybe movies on the web, but not for gaming yet.

Can someone take a look at these lenses and let me know if they think they would work?

I think the fire screen would be too big for 5X and I think those loupes are too small judging by the coins only...

[Bretspot]

Thanks for the feedback, yeah I'll have to see them in person first, but I figured, I can always scale the image down a bit, but yeah, it would likely be annoying.
If I put anything together, I'll post a pic.

[Guygasm]

Not sure if this is old news, but if anyone is still interested in experimenting with the Razer Hydra I just noticed a student discount on their website. $79.99 (without Portal 2) but you do need access to a .edu email address.

[Bretspot]

Hey 'yall,
So I picked up some 3x 2" magnifiers last night, I tried them with some images on my computer monitor scaled down and I think possibly the magnification isn't high enough, What is the relationship between magnification and screen size? Why is 5.6" usually used with 5x? Would a 4x with a 7" screen work possibly?
http://tinyurl.com/frys4x
I can get those at frys for 5$, but I don't want to waste money again. I need to find that program that lets you adjust the location and warping of an image, I guess back to the forums for more searching.

[troffmo5]

Good news:
http://sharp-world.com/corporate/news/121001.html

[2EyeGuy]

Display size: 5 inches
Number of pixels: 1,080 x 1,920 (full HD)
Pixel density: 443 ppi
Application: Smartphones

"Application: Smartphones" they have no imagination. We could do much cooler things with it.

[jaybug]

Maybe one of the "exciting surprises" is that Oculus have procured a batch of these screens for the dev-kits.

I can dream, can't I?

[MrGreen]

5 inches is too small no?

The current prototype screen is what, 5.6"? And Dykus mentioned the ideal size would be something around 7" if I recall correctly.

[Yoder808]

Hey DIY Rift Gurus! I need some help!

I got my LCD in today, and I hooked it all up, but I'm having some serious issues.



See the image links below for how it's hooked up.

The first thing I noticed when I unboxed it and assembled it, is that the plug for the "button board" didn't fit the socket on the LCD controller board. It was 1 pin too wide on the connector, and this 1 ping was unoccupied. So, I cut of the unused socket on the connector and it fit fine.

I got my universal power supply, set it to 5V, and made sure the polarity was correct according the bottom of the PCB. I plugged it and, and the green LED on the button board comes on, the screen lines up, (like a smashed GameGear), and does nothing. Eventually, the LED on the button board changes to red, but the screen looks the same.

There is no apparent damage of the screen in any way, I think my problem is in a connection perhaps. Could someone who owns this board look at my images? I'm going to try to contact the eBay seller for his advise.

http://s14.postimage.org/g5zcaum41/2012 ... _46_35.jpg

http://s14.postimage.org/kt5ecm9gx/2012 ... _46_50.jpg

http://s14.postimage.org/lwpioku41/2012 ... _47_10.jpg

http://s14.postimage.org/npsfcwfap/2012 ... _47_17.jpg

http://s14.postimage.org/q8iprbqep/2012 ... _47_28.jpg

http://s14.postimage.org/hvd0wnr75/2012 ... _47_35.jpg <--- THIS is the plug I cut the unused empty socket from, so it fits.

http://s14.postimage.org/h0v2od8q9/2012 ... _47_53.jpg

http://s14.postimage.org/fn3ft29gx/2012 ... _48_19.jpg

Thanks!

Yoder

[space123321]

Yikes - that does not look good.

As an FYI - I run my DIY Rift without my 'button board' plugged in... simply plug the unit in and it is up and running.

[Yoder808]

Yikes - that does not look good.

As an FYI - I run my DIY Rift without my 'button board' plugged in... simply plug the unit in and it is up and running.

That's actually what I tried first. It did that, so I plugged in the button board.

The stripes aren't always in the same place. You can unplug it, and plug it back in, and the position of the striped will have changed.

When I plug in the HDMI cable, it makes no difference on the screen, but the PC detects the display, and correctly sets the resolution.

I'm thinking maybe one of the LVDS pairs is wrong? I dunno - it doesn't seem possible to plug the LVDS connector in backwards either.

I'm at a loss.