DIY Oculus Rift - schematics, instructions and build pics!

Post by **Fredz** » Wed Aug 22, 2012 4:41 pm

PalmerTech wrote:Not quite. Those lenses are nice, but the field of view is quite a bit lower than the optics in the Rift. I used those UltraOptix lenses for some prototypes, but we are having custom lenses made.

Are you talking about the 7X (UltraOptix) or the 5X you linked to in your last post ?

Post by **rfurlan** » Wed Aug 22, 2012 6:48 pm

PalmerTech wrote:Not quite. Those lenses are nice, but the field of view is quite a bit lower than the optics in the Rift. I used those UltraOptix lenses for some prototypes, but we are having custom lenses made. Cheaper, and even more geometric distortion = high pixel density in the center!

Exciting!!!

PalmerTech wrote:Another thing to note: Because you are using a pure SBS signal, the center of each image is only 60.5mm apart, not the 65mm or so that is an average IPD. Because of this, you will have to move the lenses closer together than they should be to shift the image outwards, which means the exit pupil is not perfectly aligned. It works for testing, but does a lot better when you can shift the center of each image a few millimeters outward.

Thanks Palmer, it is really great to be able to tap into your expertise! I can't even imagine how many hours it took to figure out all the details for the Rift. What I am learning is that reality is a harsh mistress - at least in comparison to software.

Post by **rfurlan** » Wed Aug 22, 2012 7:25 pm

brantlew wrote:Seems like you'll be getting your Rift 3 months ahead of time. I'm jealous

With some luck because building physical things is a completely different game from building software, which is what I am used to

brantlew wrote:Have you tested your tracker yet? It comes with mouse emulation right out of the box so you should be able to get a feel for it even just holding the assembly together with your hands.

I did, but with the HMZ-T1 instead. Without proper application support the experience is interesting, but not great - at least at first. The good news is that our brains are really eager to adapt to new modes of interaction and after a while I was feeling quite immersed even though I ended up holding the Hydra on my hands. Mapping your viewport to the position of your hand, while still unnatural, still feels much more natural than using an analog stick.

Once the developer's Rift is out, we will learn a lot more about the optimal modes of interaction for a domestic VR setup. I believe head-tracking is a necessity for immersion, but it will not be sufficient for locomotion (since most users will be sitting down, with a limited range for rotation). Mouse and keyboard might turn out to be impractical (because users will be rotating their shoulders) - so what are we going to use?

Some options:
(1) A Xbox 360 controller, the obvious choice, not terribly "natural", maybe it should be bundled with the consumer version of the Rift.
(2) A Razer Hydra, a not so obvious choice but interesting since you could give the user "virtual arms" - how cool would that be?
(3) Your idea goes here

The Sixsense tracking system used by the Razer Hydra supports up to 4 independent endpoints - which is great because an integrated VR setup would need at least 3 - one as the head tracker and one for each hand. Integrating the Sixsense platform is something Palmer could pursue for the consumer version of the Rift since they will probably need to provide a "plug-and-play" solution for the masses (HMD and joypad/controller, fully integrated, as a kit).

Post by **brantlew** » Wed Aug 22, 2012 7:39 pm

rfurlan wrote:The Sixsense tracking system used by the Razer Hydra supports up to 4 independent endpoints - which is great because an integrated VR setup would need at least 3

I thought the Hydra had only two controllers?

Post by **rfurlan** » Wed Aug 22, 2012 8:06 pm

brantlew wrote:
rfurlan wrote:The Sixsense tracking system used by the Razer Hydra supports up to 4 independent endpoints - which is great because an integrated VR setup would need at least 3
I thought the Hydra had only two controllers?

That is correct, the Hydra has only two controllers, however the Sixsense platform can handle up to 4 if someone were to license and integrate it on their own devices.

Post by **Brandon9271** » Wed Aug 22, 2012 8:38 pm

I was very impressed with this video. I've never used one but it seems more responsive than the demo videos I've seen of Doom 3 bfg and the hillcrest tracker.
[youtube]http://www.youtube.com/watch?v=vVrflj30LGY[/youtube]

Post by **rfurlan** » Wed Aug 22, 2012 8:48 pm

I just updated the parts list (on page 1) to include a link to the latest AutoCAD file for the foldable case I have been working on

Post by **rfurlan** » Sat Aug 25, 2012 1:10 pm

I uploaded a revised AutoCAD file (link on the parts list, page 1). I also added a line connecting the center of both lenses to make IPD adjustments easier for everyone

Post by **rfurlan** » Sat Aug 25, 2012 8:29 pm

Update:

Almost there!

To do:
(1) A hole for my nose

(2) Head mount
(3) Head tracker

Post by **brantlew** » Sat Aug 25, 2012 9:34 pm

Very cool.

Post by **rfurlan** » Sun Aug 26, 2012 12:26 am

Here is a PDF version of the plans for the foldable case:
http://bit.ly/OCJS1g

Instructions:
(1) Make sure to print at 100% scale, in poster mode and with cut-marks enabled using Adobe Reader.
(2) Tape all 4 pages to a 4mm tick sheet of foam-core using transparent masking tape.
(3) Using an x-acto knife, cut out the lens openings completely first, then mark all other lines (but don't cut through).
(4) Cut through all outside lines to separate the shell from the sheet of foam-core.
(5) All the internal lines must be v-cut, either using this tool (http://bit.ly/OCKkgb) or manually (not as hard as you would imagine).
(6) Fold and enjoy

Post by **Chriky** » Sun Aug 26, 2012 2:32 am

Are you going to add physical blinkers so that you can't see the edge of the screen?

Post by **torc** » Sun Aug 26, 2012 4:20 am

rfurlan nice work! i'd like to have your skillz!

I have one question: so are you using lenses that Palmer suggest to you?

How is the FOV? can you give an estimate?

Last question: if you put your eyes near the lenses do you increase the FOV or you have more distortion?

Thanks a lot...Keep it up the Good work!

PS: excuse my bad english...i am Italian.

Post by **cybereality** » Sun Aug 26, 2012 11:41 am

Wow! Nice work.

Post by **rfurlan** » Sun Aug 26, 2012 11:43 am

Chriky wrote:Are you going to add physical blinkers so that you can't see the edge of the screen?

Yes, I that is the idea

Post by **rfurlan** » Sun Aug 26, 2012 11:53 am

torc wrote:rfurlan nice work! i'd like to have your skillz!
I have one question: so are you using lenses that Palmer suggest to you?
How is the FOV? can you give an estimate?
Last question: if you put your eyes near the lenses do you increase the FOV or you have more distortion?

Yes, I am using the lenses suggested by Palmer (http://bit.ly/MALA4X). The FOV is very large, if I had to guess I would say it is about ~100deg vertical and ~90deg horizontal. It covers enough of my field of view that I don't feel like I am looking at a screen anymore. Putting my eyes closer to the lenses increase the FOV and Crysis 2 in SBS mode looks great even without any correction for the lens distortion.

Post by **rfurlan** » Sun Aug 26, 2012 11:53 am

cybereality wrote:Wow! Nice work.

Thanks Cyber!

Post by **Chriky** » Sun Aug 26, 2012 12:11 pm

So I've decided to jump on this bandwagon as well. I'm just going to make a low res one for a bit cheaper, I'm very likely to screw this up and I don't want to waste too much money. Plus, it's really just to tide me over until my Rift arrives.

I ordered this screen/driver board combo - http://www.ebay.com/itm/300641746468 - and various budget lenses to try out. It's only 400x480 per eye, but pretty cheap ($100 delivered) and hey I used to play Unreal on 320x240 so I'm sure it'll be fun.

I will help me learn to code for the Rift; with head-tracking, stereo rendering and pre-warping.

Post by **rfurlan** » Sun Aug 26, 2012 1:18 pm

Chriky wrote:So I've decided to jump on this bandwagon as well. I'm just going to make a low res one for a bit cheaper, I'm very likely to screw this up and I don't want to waste too much money. Plus, it's really just to tide me over until my Rift arrives.

I ordered this screen/driver board combo - http://www.ebay.com/itm/300641746468 - and various budget lenses to try out. It's only 400x480 per eye, but pretty cheap ($100 delivered) and hey I used to play Unreal on 320x240 so I'm sure it'll be fun.

I will help me learn to code for the Rift; with head-tracking, stereo rendering and pre-warping.

That is awesome Chriky, welcome aboard!

Post by **3dvison** » Sun Aug 26, 2012 3:59 pm

rfurlan, Great Work.

Quick question,
I know it is much larger, but if it is all I can find or buy, can I use the screen from vitrolight with that 7 in 1 controler board that they use ? Even though the controler will be larger than yours, will it still work just the same as your LCD & control board combo ? Still be able to use 3D drivers and games set-up just like yours ?

Post by **rfurlan** » Sun Aug 26, 2012 4:21 pm

3dvison wrote:I know it is much larger, but if it is all I can find or buy, can I use the screen from vitrolight with that 7 in 1 controler board that they use?

Absolutely! Palmer has been using that same 7-in-1 board for his prototypes. Before he was cutting it to make it smaller but the latest known prototype uses an uncut board inside a separate control box:

3dvison wrote:Even though the controler will be larger than yours, will it still work just the same as your LCD & control board combo? Still be able to use 3D drivers and games set-up just like yours?

You shouldn't have any problems, I am simply feeding a standard 720p signal with the left/right frames rendered side by side. While Crysis 2 supports this mode natively, other games would need a 3rd-party tool to generate the SBS frame.

Post by **3dvison** » Sun Aug 26, 2012 9:16 pm

Thanks rfurlan,

Post by **Fredz** » Mon Aug 27, 2012 7:48 pm

I just received my UltraOptix 7X magnifiers and tested them with the 5.6" display of my old Fujitsu UMPC and the result is great. I've tested it with this video from Cybereality :
http://www.youtube.com/watch?v=Ovf5TLiIfZ8

More details here : http://www.mtbs3d.com/phpBB/viewtopic.p ... 329#p79329

The result was not really convincing with the images from Cybereality and Emerson I found on the forum, not bad with the one from John Carmack but it doesn't stand the comparison with the video.

I was holding the screen and the magnifiers in my hands while playing the video, but now I need to build a box so it's more practical.

I'll probably receive the 5X magnifiers PalmerTech talked about in some days, but now I'm not really sure the result will be better. The only problem with the 7X was some color fringing (and the lack of warping in the video), I'll see if that can be corrected in the rendering.

I'll most probably try to experiment with head tracking next, using my PS3 eye or buying a Hillcrest or Sparkfun IMU. I'll also try to do some software experiments too, by writing a small demo or by adapting the old anaglyph stereo driver I developed for Linux.

Post by **rfurlan** » Mon Aug 27, 2012 11:03 pm

Hi Fredz, I posted this a while ago regarding the 7x lenses:

rfurlan wrote: The diameter of the (7x) lenses limit the effective FOV. After mounting the lenses on a faceplate, the FOV is significantly narrowed because you can't get close enough to see the whole screen anymore (plus lots of pixels are wasted outside the visible area).

The 50mm suggested by Palmer are actually *much* better

Post by **sellars** » Tue Aug 28, 2012 4:14 am

rfurlan wrote:
The 50mm suggested by Palmer are actually *much* better

Giving this a go myself too - already have the screen and board from Ebay, and a Hillcrest (which I've got working correctly in a simple OpenGL demo viewing a pyramid with correct head movement, thanks to Brantlew's code for grabbing the hillcrest signal. Still awaiting the 5x lenses - have you already worked out the correct distance from the screen the lenses need to be to focus correctly?

Post by **Fredz** » Tue Aug 28, 2012 4:14 pm

rfurlan wrote:The diameter of the (7x) lenses limit the effective FOV. After mounting the lenses on a faceplate, the FOV is significantly narrowed because you can't get close enough to see the whole screen anymore (plus lots of pixels are wasted outside the visible area).

The 50mm suggested by Palmer are actually *much* better

Yes I've read your previous posts, but using the 7X magnifiers I can have a clear vision and see the whole half screen with each eye. I need to make a box to be able to measure the FOV, I hope it won't be reduced because of the box.

By the way the 5.6" screen I'm using is 12.3 x 7.2 cm, but since it's 1024x600 (1.7 aspect ratio, supposing square pixels) it may different from yours (1280x800 and 1.6 ar if square pixels). Could you measure it to compare ? Thanks.

Post by **rfurlan** » Wed Aug 29, 2012 1:31 pm

sellars wrote:Giving this a go myself too - already have the screen and board from Ebay, and a Hillcrest (which I've got working correctly in a simple OpenGL demo viewing a pyramid with correct head movement, thanks to Brantlew's code for grabbing the hillcrest signal. Still awaiting the 5x lenses - have you already worked out the correct distance from the screen the lenses need to be to focus correctly?

Hi Sellars, welcome to the pack

Yes, I worked out all the distances already, everything should be perfectly aligned if you follow these schematics to build a shell using a 4mm foam-core sheet:

Download latest AutoCAD file here -> http://bit.ly/SVM7fZ (still work in progress!)
Or use the PDF version if you would like -> http://bit.ly/OCJS1g
(check the first post of this thread for step-by-step build instructions)

Post by **brantlew** » Wed Aug 29, 2012 1:32 pm

Just a tip.. I would pass on the Hillcrest and go straight to a Sparkfun for tracking. If you're going to do it, then do it right.

Post by **cybereality** » Wed Aug 29, 2012 1:45 pm

Is the Hillcrest really *that* bad. Seems OK from my limited testing. Keep in mind I have not worn it with an HMD yet, since the USB cord I have is only like 3 feet. I am waiting on a 10 foot cable so I can test the thing with my HMZ-T1. Will report back once I try it on proper, but seems OK so far.

Post by **Fredz** » Wed Aug 29, 2012 1:51 pm

I've inserted the magnifiers in a small cardboard mount but as was expected a lot of the FOV is lost unfortunately. The problem is mainly due to the nose which increases the minimum distance the lenses can be from the eyes, and the forehead didn't help either.

Here are some shots, but nothing really impressive :

IMG_0004.JPG

IMG_0007.JPG

IMG_0008.JPG

I'll wait for the 5X magnifiers and try again with a similar mount, if that doesn't work out I'll try to make a round eyeglass frame to support the lenses. That should take care of the nose problem.

Post by **brantlew** » Wed Aug 29, 2012 2:01 pm

cybereality wrote:Is the Hillcrest really *that* bad. Seems OK from my limited testing. Keep in mind I have not worn it with an HMD yet, since the USB cord I have is only like 3 feet. I am waiting on a 10 foot cable so I can test the thing with my HMZ-T1. Will report back once I try it on proper, but seems OK so far.

I think it's good matched with the right sensor fusion software. The responsiveness is good and when I tried it with the Rift it worked wonderfully and of course everyone else seems happy with it in the Doom demos. But that's Carmack's software. The firmware that comes preloaded seems to have problems. They're not as evident when you're just testing it out and watching your screen, but when you view it through a head-mount the jumpiness and discontinuities are really evident. I ordered a Sparkfun which should be arriving today. I've heard nothing but good things about it, so hopefully that will make a better solution.

PS. It's funny. I hooked up my Vuzix tracker this week to test on a 9DOF before my Sparkfun arrives. I remember thinking it was a really great tracker when I got it, but Holy Cow! - it is laggy as hell! I never realized it before, but now I'm accustomed to these low latency trackers and the Vuzix seems sooo slow to me now.

Post by **rfurlan** » Wed Aug 29, 2012 2:42 pm

Really cool Fredz!!!

Post by **rfurlan** » Wed Aug 29, 2012 2:50 pm

brantlew wrote:
cybereality wrote:Is the Hillcrest really *that* bad. Seems OK from my limited testing. Keep in mind I have not worn it with an HMD yet, since the USB cord I have is only like 3 feet. I am waiting on a 10 foot cable so I can test the thing with my HMZ-T1. Will report back once I try it on proper, but seems OK so far.
I think it's good matched with the right sensor fusion software. The responsiveness is good and when I tried it with the Rift it worked wonderfully and of course everyone else seems happy with it in the Doom demos. But that's Carmack's software. The firmware that comes preloaded seems to have problems. They're not as evident when you're just testing it out and watching your screen, but when you view it through a head-mount the jumpiness and discontinuities are really evident. I ordered a Sparkfun which should be arriving today. I've heard nothing but good things about it, so hopefully that will make a better solution.

PS. It's funny. I hooked up my Vuzix tracker this week to test on a 9DOF before my Sparkfun arrives. I remember thinking it was a really great tracker when I got it, but Holy Cow! - it is laggy as hell! I never realized it before, but now I'm accustomed to these low latency trackers and the Vuzix seems sooo slow to me now.

That is great to know brantlew!

Has anyone tried using a Razer Hydra for tracking? I did some very limited testing and it felt pretty good, next I want to move the magnetic coils from one of the controllers to the HMD, leaving the other handheld tracker, the dual analogs and all buttons still functional.

Any thoughts on this idea?

Post by **sellars** » Wed Aug 29, 2012 4:38 pm

One... very rough box later...

and yes, I don't have a thin HDMI cable yet. Nor, apparently a supplied power adapter with more than 2 foot of cable!

Post by **rfurlan** » Wed Aug 29, 2012 4:46 pm

sellars wrote:One... very rough box later...
and yes, I don't have a thin HDMI cable yet. Nor, apparently a supplied power adapter with more than 2 foot of cable!

Looks awesome to me!

Post by **sellars** » Wed Aug 29, 2012 4:56 pm

I see what you mean about the noise in the cables though - twisted them, and I'm getting a horizontal wave of noise. Might be good to see if there are replacement cables around.

I'm now looking for a good screengrab with the actual barrel warp used in the Rift too - Cyber's got some great WIP on manipulating existing games, but either my IOD is wrong (very possible!) or the warping isn't high enough yet.

Also, I'm probably going to work on getting the centre point of the camera away from the centre of each half of the screen - PalmerTech did mention this a ways back I think, and that might be causing issues.

Now to spray it black and surgically attach it to a spare pair of ski goggles, somehow.

Post by **WiredEarp** » Wed Aug 29, 2012 6:36 pm

@ rfurlan re the Hydra:

I've tried a couple of ways of disassembling the Hydra. Theres the one where you cut off the coils, connect fine wires, etc. You need a shielded cable to pull this one off. Then theres the other way of cutting stuff off the Hydra till you are left with 2 circuit boards. This way is not as good in terms of making it compact, but it seems to generate less interference than the coil removal method. If you go that path however, you have to be careful you dont break your microcable that connects both circuitboards.

I got mine going acceptably with the coil removal method, but it still has a bit of a 'jump' when I turn to a certain point (I think this is interference caused by where I attached the cable to the coils). Also, i'm having issues with my microcable again, so I can't really fix it until I've fixed that up.

I'm planning on attempting another microcable repair over the next week, and resoldering my coil so that there is the minimum of metal/wire/solder I can possible achieve. Hopefully this will resolve the 'jumping' issue.

Post by **Jotschi** » Thu Aug 30, 2012 12:36 am

It is fun to see how many people are into the diy rift:
http://offer.ebay.de/ws/eBayISAPI.dll?V ... 0319537252

My current setup looks like this:

I used cheap ski goggles and removed the spacer that normally is used to keep the googles away from your face.

The 5x lenses that Palmer suggested are great. The FOV is enormous. I can't really say how my setup will work at the end since i'm still waiting for my displays to arrive.

The sparkfun razor 9dof IMU is a neat little device but the default firmware is just for testing purposes. The update rate is low and there is no onboard filtering being executed. I plan to connect the IMU together with a Pixart sensor that i extracted from a Wiimote to my Raspberry PI / Pandaboard by using SPI/I2C. The board itself will then output a small opengl demo.

Post by **brantlew** » Thu Aug 30, 2012 7:36 am

@Jotschi: Almost everyone flashes their Sparkfun with the FreeIMU AHRS firmware which is supposed to be good.

https://dev.qu.tu-berlin.de/projects/sf-razor-9dof-ahrs

Post by **Fredz** » Thu Aug 30, 2012 1:02 pm

Jotschi wrote:The 5x lenses that Palmer suggested are great. The FOV is enormous. I can't really say how my setup will work at the end since i'm still waiting for my displays to arrive.

Heh, nice setup ! You could still print some of the warped images found on this forum on a 120.96 x 75.6 mm sheet of paper if you want to test it before you receive your display.

I should do that also to be prepared to what it'll look like with higher resolution screens.

Dimensions taken from there :
http://www.ebay.com/itm/5-6inch-TFT-LCD ... 0734588827