DIY Oculus Rift - schematics, instructions and build pics!

[Rothchild]

I thought the Hydra base tracked the controllers relative to its own position - if they are all tied together, how is that going to work?

I think they put the hand units on the helmet for storage, not actual use.

[rfurlan]

I thought the Hydra base tracked the controllers relative to its own position - if they are all tied together, how is that going to work?

I think they put the hand units on the helmet for storage, not actual use.

That would make a lot of more sense!

[brantlew]

Thanks for the quick answers everyone. That Move contraption looks a tad top heavy, but I'm sure it works like a charm.

Is there a guide for using the iPhone as a tracker? Also, I haven't seen any Android based trackers, are the iPhone's sensors superior?

The iPhone has nice tracking. Android quality is going to be totally dependent on the phone manufacturer. Here is one way to get tracking via iPhone.

http://www.mtbs3d.com/phpBB/viewtopic.php?f=139&t=15484

[3dvison]

So is there only a PS-Move & PS-Eye in that picture or is there also a hydra ?
If yes to the Hydra, does that mean they are using all three together at the same time ?

[rfurlan]

The Hydra is the best tracking solution have I tried so far, by a wide margin. I have also tried: Gyration remote, Wiimote, iPhone, Ardupilot and infrared beacons + camera.

[Fredz]

So is there only a PS-Move & PS-Eye in that picture or is there also a hydra ?
If yes to the Hydra, does that mean they are using all three together at the same time ?

Yes, it seems they are using the 3 at the same time. But the Hydra is only used for hands tracking, for head tracking it's the PS Move and the PS Eye.

John Carmack said in his 3h speech that the Razer Hydra was not suitable for head tracking, its accuracy degrades sensibly when the distance to the base augments.

[Yoder808]

The Hydra is the best tracking solution have I tried so far, by a wide margin.

...the Razer Hydra was not suitable for head tracking, its accuracy degrades sensibly when the distance to the base augments.

Oh Mylanta I'm confused.

[coadyj]

Hi rfurlan, firstly can I just say great job, I am a long time lurker, I also followed another thread you started about making a 3d printed case for the HMZ-T1. I too was thinking of making a home made oculus and I came across the leep on the cheap website. The lenses that palmer suggested are only one lens, but the leep optics in the leep on the cheap website uses two, but palmer talks about using leep optics in the quake con interview. How come we don't need the two lenses per eye in this set up?

You also posted a link to a display with higher resolution, do you think that the diy oculus could end up being better than they dev kit version? Also, would we not be able to upgrade the dev kit version with a better quality screen. Palmer said no in an interview on Reddit, but I am starting to think that was a business answer and not an answer to hackers. I mean the set up is so simple, how hard can it be. Does the focal lengths change in higher resolution?

Anyway, I have ordered a pair of the lenses from ebay and I am going to give it a go.

[Dycus]

Welcome to the forums, Coady!

I work for Oculus. I'm not sure where Palmer said we were using leep optics... We're only using one lens per eye as that's all we need. The distortion is handled in software, and is actually good because it puts more pixels in the center of your vision.

I'm not sure on the final case design for the dev kit, but Palmer may be right about not being able to swap the screen. Besides having to find one of the exact same size (unless you wanna be gluing stuff in, but good luck getting the focus right!), you'll have to take it apart, of course. I don't know how easy that will be, though, or if it will be possible.

The focal length is dependent on the lenses, not the screen, so no, focal length will not change with a higher-res screen.

[2EyeGuy]

For a normal HMD, you have information like this:

"equivalent to a 63 inch screen at 9 feet"

That's for a VR920 (if I remember correctly). And it really is equivalent to 63 inch screen at 9 feet. Not just in terms of FOV, but if you show the same image on both screens, the parallax will be equivalent to objects at 9 feet (I checked). The optical focus seems to be similar to 9 feet too. And of course each eye can see the whole "screen" at that distance (but not at any other distance! think about it, there's no such thing as 100% overlap except at one depth). That kind of setup only makes sense for low FOV HMDs with low distortion (the VR920 does have some minor distortion though). For a Rift there's no need for it to be like a virtual screen at a certain depth.

But how does it work for a real Rift and for these DIY rifts?

Firstly, what is the optical focus depth? What depth do the lenses in your (real) eyes have to focus on? That may affect what virtual distances 3D objects are comfortable to look at and what depths hurt your eyes or give you a headache. Which in turns affects what depths are best for drawing 2D things (or 3D things). I don't know if this affects "doubling" in 3D images.

Secondly, how much parallax is there between the centre of each eye? Are they at infinite depth or some other depth? This is important for how a 3D driver will work.

Thirdly, what's the aspect ratio? Are we dealing with basically square pixels? And does that mean this is a tall-screen display in full-SBS and not half-SBS? Is it 8:9? When they were saying 110 degree diagonal field of view I was picturing super-widescreen (sort of like their illustration) but if it is actually taller than it is wide, that would be awesome.

I want to have a go at making some software for this thing.

[Fredz]

Firstly, what is the optical focus depth?

I think Palmer said it was at 12 feet, which was equivalent to infinity according to him IIRC. But I guess it may be different now since they are using different optics.

Secondly, how much parallax is there between the centre of each eye? Are they at infinite depth or some other depth? This is important for how a 3D driver will work.

I'm not sure I understand your question, but the lenses are supposed to be around 6.5cm apart from what I've read if that's what you ask.

Thirdly, what's the aspect ratio? Are we dealing with basically square pixels?

The FOV is 90x110° and the resolution is 640x800 per eye, so the pixels are almost square (110/90 ≈ 1.22, 800/640 = 1.25)

And does that mean this is a tall-screen display in full-SBS and not half-SBS?

I don't think full or half SBS makes sense for such a display, if you feed it the right content it could be considered full SBS I suppose.

I want to have a go at making some software for this thing.

Have a look in the forums, there are already 6 or 7 projects (stereo drivers and test demos) fiddling with that, with at least one being open source.

[2EyeGuy]

On another note, I've noticed people throwing around terms like 9DOF. I'd like to remind people that there's no such thing, unless you are tracking more than one object. For example, if you fully tracked a pair of scissors, that would be 7DOF, with the opening and shutting of the scissors being another DOF. (The Razer Hydra has 16DOF if you include the joysticks). Degrees Of Freedom are by definition independent (or as close as possible given the nature of 3D rotation).

A 6DOF head-tracker, which is what the Oculus Rift is advertised as, is by definition capable of tracking both head orientation and head position. I don't know whether the 6DOF spec is true or a lie. In all the demonstration videos the testers are all too stupid to test more than 2DOF, so I don't know what it can actually do. If I was testing one, I'd be leaning to the side, rolling my head, and moving around.

You can't just put accelerometers and gyros in something and claim it's 6DOF (unless you're doing some very fancy calculus, and even then it's slightly dodgy).

[Fredz]

The Hydra is the best tracking solution have I tried so far, by a wide margin.

...the Razer Hydra was not suitable for head tracking, its accuracy degrades sensibly when the distance to the base augments.

Oh Mylanta I'm confused.

Have a look at this post, the problem is pretty well explained :
http://www.mtbs3d.com/phpbb/viewtopic.p ... 942#p81942
The Hydra is only accurate when you're close to the base, when your quite far it's pretty bad. As John Carmack said it's not really a problem with hand tracking (since the precision seems to be consistently good) but it's a no-go with head tracking (since the accuracy is quite variable).

[Fredz]

On another note, I've noticed people throwing around terms like 9DOF. I'd like to remind people that there's no such thing, unless you are tracking more than one object.

Yes, it's an abuse of language but it's commonly used and well understood. It's only because these devices (MARGs) use 3 types of sensors (accelerometer, gyroscope, magnetometer), hence 3x3 = 9dof. But we all know that we can only get 6dof from them, the advantage over IMUs is that the magnetometer can correct the drift in rotation due to the gyroscope. But anyway they're unable to provide true 6dof because of the translation drift of the accelerometer that can't be corrected (or with an external sensor like an optical tracker (cam) or an EMS (electromagnetic tracking system))

A 6DOF head-tracker, which is what the Oculus Rift is advertised as, is by definition capable of tracking both head orientation and head position. I don't know whether the 6DOF spec is true or a lie.

At the current time it's only used as a 3dof tracking system, with rotation drift because it doesn't use a magnetometer. But I guess (and hope) that this may change before release.

[2EyeGuy]

Firstly, what is the optical focus depth?

I think Palmer said it was at 12 feet, which was equivalent to infinity according to him IIRC. But I guess it may be different now since they are using different optics.

That's probably a good choice for most games. Although I'd have to wear it over my glasses or I wouldn't be able to see anything at 12 feet except a big blur. I can't even count my toes without my glasses (although I am tall).

Secondly, how much parallax is there between the centre of each eye? Are they at infinite depth or some other depth? This is important for how a 3D driver will work.

I'm not sure I understand your question, but the lenses are supposed to be around 6.5cm apart from what I've read if that's what you ask.

Not what I was asking. I know real eyes are 6.4cm or 6.5cm apart (except rfurlan's). But if you actually render a stereoscopic image on a 3D TV, or on most HMDs, you can't just place the two virtual cameras 6.4cm apart and display them to each eye, because there's already some parallax between you and the TV that you need to compensate for. On a 3D TV it would look like everything at infinity was actually at screen level (no depth behind the screen) with anything closer than infinity popping out slightly.

EDIT: What I want to know is what depth a 2D image will appear to be if I show the same image on both screens, or what depth something will appear to be if I draw it in the centre of each screen.

Thirdly, what's the aspect ratio? Are we dealing with basically square pixels?

The FOV is 90x110° and the resolution is 640x800 per eye, so the pixels are almost square (110/90 ≈ 1.22, 800/640 = 1.25)

You can't actually do divisions like that with degrees, I think. I think you need to use tan or something like that. So it probably works out as perfectly square (as you would expect from circular lenses).

And does that mean this is a tall-screen display in full-SBS and not half-SBS?

I don't think full or half SBS makes sense for such a display, if you feed it the right content it could be considered full SBS I suppose.

Although you sort of need to feed it specially distorted images, I'd still call it full SBS. If you fed a straight full-SBS image with an 8:9 aspect ratio, it would probably look about right.

I want to have a go at making some software for this thing.

Have a look in the forums, there are already 6 or 7 projects (stereo drivers and test demos) fiddling with that, with at least one being open source.

Sounds great! I've made VR software before, but it will a lot cooler with this. I'm kind of embarrassed about my own skills though when I see what other people on here are doing.

[Fredz]

Not what I was asking. I know real eyes are 6.4cm or 6.5cm apart (except rfurlan's). But if you actually render a stereoscopic image on a 3D TV, or on most HMDs, you can't just place the two virtual cameras 6.4cm apart and display them to each eye, because there's already some parallax between you and the TV that you need to compensate for. On a 3D TV it would look like everything at infinity was actually at screen level (no depth behind the screen) with anything closer than infinity popping out slightly.

I still don't get what you want to know, but basically I think it's not a problem with the Rift. The centers of the screens should be spaced at around 6.5cm, as are the lenses and the eyes. So the calculus of the off-axis projection matrices looks quite straightforward to me, the cams are basically spaced 6.5cm from each other in "game space" (using whatever unit the game uses) and using 90x100° FOV each.

You can't actually do divisions like that with degrees, I think. I think you need to use tan or something like that. So it probably works out as perfectly square (as you would expect from circular lenses).

I don't see how using tangents would help in this case. Anyway pixels are not really square since there is distortion, but their height/width ratio is around 1 (1.22 ≈ 1.25).

Sounds great! I've made VR software before, but it will a lot cooler with this. I'm kind of embarrassed about my own skills though when I see what other people on here are doing.

We all have to start somewhere. I didn't knew anything about all this when I started to implement my stereo driver.

[2EyeGuy]

Yes, it's an abuse of language but it's commonly used and well understood. It's only because these devices (MARGs) use 3 types of sensors (accelerometer, gyroscope, magnetometer), hence 3x3 = 9dof. But we all know that we can only get 6dof from them,

But you can't get 6DOF from them, because they all measure either the angles or angular velocity. None of them measure position. So it is 3DOF. And there's a huge difference between the immersiveness of 3DOF and 6DOF tracking. I was really excited about the 6DOF part.

You could also use ultrasound for position tracking. I've seen some devices do that.

[Malfate]

I'm new here but why is the hydra not an ideal solution/supplement. Is that cause it induces it's own magnetic field?

[mahler]

Yes, it's an abuse of language but it's commonly used and well understood. It's only because these devices (MARGs) use 3 types of sensors (accelerometer, gyroscope, magnetometer), hence 3x3 = 9dof. But we all know that we can only get 6dof from them,

But you can't get 6DOF from them, because they all measure either the angles or angular velocity. None of them measure position. So it is 3DOF. And there's a huge difference between the immersiveness of 3DOF and 6DOF tracking. I was really excited about the 6DOF part.

Don't worry too much. Palmer has expressed that 6DoF for the consumer version is a very high priority.
Besides, with those three sensors (9-Axis) it is possible to get positional tracking up to a high degree of accuracy and with minimized drift (although with the level of immersion the Rift provides, it might still too much)

See this video for an explanation: Sensor Fusion on Android Devices: A Revolution in Motion Processing @ 23:20

These guys have made a very impressive 6DoF demonstration with their LPMS-CU: 9-Axis IMU using a 'novel' filtering algorithm:

• http://www.lp-research.com/introducing-lptracker/
• http://www.lp-research.com/lpms-head-tracking-demo/

Discussed with the one of the creators here: http://www.mtbs3d.com/phpBB/viewtopic.p ... est#p78786

[MSat]

Post malfunction

[brantlew]

I'm new here but why is the hydra not an ideal solution/supplement. Is that cause it induces it's own magnetic field?

The consensus is that it probably is the best device that we have right now for full tracking. There is some slight concern over the stability of the readings, but to me the biggest problem with the Hydra is that it's a wired device so it decreases mobility and prevents you from easily creating 360 degree setups.

[rfurlan]

Hi rfurlan, firstly can I just say great job, I am a long time lurker, I also followed another thread you started about making a 3d printed case for the HMZ-T1. I too was thinking of making a home made oculus and I came across the leep on the cheap website. The lenses that palmer suggested are only one lens, but the leep optics in the leep on the cheap website uses two, but palmer talks about using leep optics in the quake con interview. How come we don't need the two lenses per eye in this set up?

You also posted a link to a display with higher resolution, do you think that the diy oculus could end up being better than they dev kit version? Also, would we not be able to upgrade the dev kit version with a better quality screen. Palmer said no in an interview on Reddit, but I am starting to think that was a business answer and not an answer to hackers. I mean the set up is so simple, how hard can it be. Does the focal lengths change in higher resolution?

Anyway, I have ordered a pair of the lenses from ebay and I am going to give it a go.

Hi Coady, welcome and good luck building your HMD!

I tried the cheap leep optics from that article with the 5.6" screen used by the Rift prototype. The image output was really nice, but sadly the whole contraption was too heavy to be practical.

Regarding the DIY Oculus having a better screen than the official one, it is possible but not likely. I am sure that the Rift will ship with the best screen available for its price point.

Finally, regarding upgrading the official Rift, it is hard to say before it is out. Unless the higher-resolution screen is exactly the same size it would be hard to make everything fit back inside the headset. Another roadblock is that the higher-resolution screen might also require a different kind of controller.

[rfurlan]

The Hydra is the best tracking solution have I tried so far, by a wide margin.

...the Razer Hydra was not suitable for head tracking, its accuracy degrades sensibly when the distance to the base augments.

Oh Mylanta I'm confused.

Ok, I should make it clear that I have no intention to walk around wearing my HMD. My Hydra setup has me sitting less than a meter away from the base

While I love the idea of a 360deg setup, I wonder how practical it is since the Rift itself is wired - won't we find ourselves tethered no matter what?

[brantlew]

From what I gather, Carmack was thinking in terms of 6DOF of freedom within an area comparable to the Kinect. He often talks about bending down to look at the floor textures. While admittedly a desirable goal, I can understand how the Hydra would fail to scale to this size. But for the more limited purpose of unconscious lateral head motion, neck bending, and potentially even lateral dodging all reports seem to indicate that the Hydra is sufficiently accurate and stable.

While I love the idea of a 360deg setup, I wonder how practical it is since the Rift itself is wired - won't we find ourselves tethered no matter what?

Speaking for myself - I have no intention of being tethered. Backtops and/or HDMI transmitters will be integral to my setups.

[WiredEarp]

From what I gather, Carmack was thinking in terms of 6DOF of freedom within an area comparable to the Kinect. He often talks about bending down to look at the floor textures. While admittedly a desirable goal, I can understand how the Hydra would fail to scale to this size. But for the more limited purpose of unconscious lateral head motion, neck bending, and potentially even lateral dodging all reports seem to indicate that the Hydra is sufficiently accurate and stable.

Actually, the Hydra works fine for all of that. I think you can bend down and look at floor textures etc no problems (certainly it picks up the gun held at under waist height). Its basically similar tech to what was used in the Virtuality system, and there were no problems with that. It just won't scale to walking around - but for a stand up system, its the best bet currently IMHO. Personally, I dont see being able to move around while wearing a HMD as a particularly desirable thing at this point for home VR, unless you have an empty room and software to prevent you hitting the walls.

[rfurlan]

Speaking for myself - I have no intention of being tethered. Backtops and/or HDMI transmitters will be integral to my setups.

Unless the simulation is aware of the shape and size of your "safe area" and is able to give you some sort of feedback to keep you within it, it simply isn't practical to walk around your home or office blindfolded without a spotter.

It is not even very safe to stand while wearing an immersive HMD without a spotter because the simulation can sometimes trigger your reflexes. For example, if you fall from a cliff in the virtual world, you might automatically bend your knees to lessen the impact, since you are not really falling, you end up losing your balance and falling to the floor in the real world. This happens in a split second and isn't something you should try to suppress (reflexes are useful after all).

I experienced this first hand a few times, it is really confusing/fun and I was glad I had someone spotting me

Also, even with very good tracking, walking around can be a bit disorienting. Bottom line: if you are not sitting, you will need someone to spot you,

[jaybug]

Man, TheLostBrain's model really got me interested in jumpin aboard on this DIY thing

I'm just afraid I'll spend a couple hundred and then inevitably break the screen or fry the boards somehow :p

[rfurlan]

Man, TheLostBrain's model really got me interested in jumpin aboard on this DIY thing
I'm just afraid I'll spend a couple hundred and then inevitably break the screen or fry the boards somehow :p

TheLostBrain is very talented, that is for sure!

I cracked a screen few weeks ago but haven't fried a board yet, I am sure it is just a matter of time though. Either way, it was worth it

[zacherynuk]

I'm just afraid I'll spend a couple hundred and then inevitably break the screen or fry the boards somehow :p

Hehe, I have gone through both a board and a screen. And 3 projectors. I have also gone through 3 packs of elastoplast. Luckily no full fingers yet.
I'm not very careful.

[Jotschi]

What is the best solution for mounting the lenses?
I saw some pictures in which the lenses were stuck down using some tape.
Is the tape being used to clip the FOV? ( I guess you can otherwise see the edge of the screen )

Has anyone tried to use the 2" 5x pocket loupe mount?

[zacherynuk]

What is the best solution for mounting the lenses?
I saw some pictures in which the lenses were stuck down using some tape.
Is the tape being used to clip the FOV? ( I guess you can otherwise see the edge of the screen )

Has anyone tried to use the 2" 5x pocket loupe mount?

IMHO Hot Glue Gun is your friend. Just 3 or 4 small dots will hold them soundly enough.

[brantlew]

Also, even with very good tracking, walking around can be a bit disorienting. Bottom line: if you are not sitting, you will need someone to spot you,

Out of curiosity. Have you experienced nausea from any motions? Swaying, looking down, etc. What about walking? Has that brought about any nausea?

[PatimPatam]

The Hydra is the best tracking solution have I tried so far, by a wide margin.

Sorry for staying a bit out of topic but, while using the hydra for headtracking, have you noticed the low accuracy (roll off by 10 degrees for instance) that Carmack mentions in his keynote?
John Carmack on the Razer Hydra for head tracking

If you haven't noticed, could this innacuracy be unit-dependant? meaning basically you had "good luck" with your hydra unit and carmack was not so lucky?

[Owen]

It is pretty inaccurate. I have used two hydra units and they were indistinguishable, but the distortion varies significantly from one environment to another. Correcting for it would probably require a very lengthy calibration routine checking several known points and orientations in your play-area. Electronics and large metal objects in particular will skew the results. The first time I used it in my office the movements were very sloppy and at one point the Z translation axis actually inverted, though it always worked great in my apartment.

You compensate naturally when its tracking your hand, especially if you are using it to simulate an object held in your hand since then you would need to compensate for the shape of the object in real life too, but it would definitely be nauseating for your head. I would only use it to correct for drift, not as the primary head tracker.

[rfurlan]

What is the best solution for mounting the lenses?
I saw some pictures in which the lenses were stuck down using some tape.
Is the tape being used to clip the FOV? ( I guess you can otherwise see the edge of the screen )
Has anyone tried to use the 2" 5x pocket loupe mount?

I have been using tape so far, it is not as bad as it sounds. I recommend black tape and you can position it so you can't see the edges of the screen for better immersion. I am using the 2" 5x Palmer recommended, it works great!

[brantlew]

I think a sensor fusion of an IMU device for orientation and a Hydra for position would be the a good way to compensate for quirks in the Hydra readings. So far though, I don't know anyone who has done that yet.

[Jotschi]

What is the best solution for mounting the lenses?
I saw some pictures in which the lenses were stuck down using some tape.
Is the tape being used to clip the FOV? ( I guess you can otherwise see the edge of the screen )
Has anyone tried to use the 2" 5x pocket loupe mount?

I have been using tape so far, it is not as bad as it sounds. I recommend black tape and you can position it so you can't see the edges of the screen for better immersion. I am using the 2" 5x Palmer recommended, it works great!

Ok, thanks for the feedback. I think i won't have a problem with the edges of the screen because i'm currently aiming for a dual screen solution. I think it won't be a rift clone - more likely a rift mutation

[rfurlan]

Also, even with very good tracking, walking around can be a bit disorienting. Bottom line: if you are not sitting, you will need someone to spot you,

Out of curiosity. Have you experienced nausea from any motions? Swaying, looking down, etc. What about walking? Has that brought about any nausea?

No nausea at all, I am lucky that I am not prone to motion or simulator sickness. The odd feeling of disorientation I am talking about is caused mostly by tracking lag, your brain notices something is "off" and as you walk around the motion can feel somewhat unnatural even on a high-end setup. The closest metaphor I have for it is that you feel like you are very mildly drunk. Does this make sense?

[rfurlan]

Ok, thanks for the feedback. I think i won't have a problem with the edges of the screen because i'm currently aiming for a dual screen solution. I think it won't be a rift clone - more likely a rift mutation

Props to you, diversity rules!

[rfurlan]

The Hydra is the best tracking solution have I tried so far, by a wide margin.

Sorry for staying a bit out of topic but, while using the hydra for headtracking, have you noticed the low accuracy (roll off by 10 degrees for instance) that Carmack mentions in his keynote?
John Carmack on the Razer Hydra for head tracking
If you haven't noticed, could this innacuracy be unit-dependant? meaning basically you had "good luck" with your hydra unit and carmack was not so lucky?

It has been very accurate for me, however my setup has me seated right in front of the Hydra base (< 0.5 meters) - maybe that is why it works well for me?

EDIT: Also, I am not using mouse emulation, I am using a test program I wrote that reads the tracking info using the SDK.