Hey guys,

Here is the InfinitEye open source hardware HMD project release v1.0, so people (including Palmer Luckey himself ) can start to make their own build and modifications. Special thanks to MSat for the name InfinitEye.

I don't know much about licences (especially for hardware projects) but I think something suitable for a project like this, is Creative Commons CC BY-SA (and to be credited by the "BY" part, my real name is Lionel Anton (I'm the author of F-ZERO for TI68k if anyone tries to google my name))

Specifications:

- Horizontal FOV = 180° (calculated)
- Horizontal stereoscopic FOV = 63° (calculated)
- Vertical FOV = (to be determined)
- Weight without tracker and head mount = 412g (measured)
- Weight with head mount = 512g (measured)
- Resolution per eye = 1280x800 spread over 121° horizontally (calculated)
- Bounding box of the headset (without mount) = 310x220x140 mm

The components I used :

- 2 x N070ICG-LD1 + controller board from ebay : 260$
- 4 x Fresnel lenses FRL021 from optolife (size 165x105mm, focal length 120mm, thickness 2mm, groove 0.3mm) : 20$ + shipping cost
- 1 sheet of expanded PVC thickness 2mm from castorama (french hardware store, but I'm sure similar light weight 2mm thick and flexible material can be found elsewhere) 500x750x2 mm : 8.45 eur ~= 11$
- neoprene glue
- headmount from faceshield from castorama : 17 eur ~= 23$

cost ~= 320$

Additional components :

- Zotac mini DisplayPort to dual HDMI from amazon : 69$
- YEI 3 space sensor embedded for head tracking (website): 99$

total cost ~= 500$
Now the plans (Version 1.0) :



(I have no website to host the file so I'm just leaving it attached to this thread for now)

I made the following drawing to compare horizontal FOV between this HMD, the Rift and the other HMD I own: ST1080


Note that in the drawing, I used the 7" display to scale the Rift but with the 90° HFOV 100% overlap which is less likely (I think it will be something like 115-120° HFOV with 80% overlap but we'll see in march )

build pics of the prototype :

















(I can't attach other pictures, 15 files is the maximum)
Next steps :
- paint inside top and bottom parts in black
- glue the small parts to the box (green ones to prepare the LCD rails and then the yellow ones to maintain the box)
- glue the box nose part
- glue the lenses (top and bottom)
- insert the left LCD from the left
- insert the right LCD from the right
- insert the pink parts to maintain the LCDs in place
- glue the box left and right part to close it
- drill the holes to attach the head mount of your choice

Steps that I haven't done yet :
- extend LVDS cables (or tape the controller board to the headset, simpler but heavier)
- add head tracker
- modify VireIO perception drivers to make them compatible with this design
- say goodbye to reality and enjoy VR

Feel free to ask any question and post pictures of your builds in the thread

Awesome work man! Can you take some pictures through the lenses so we can get an idea of the quality?

I really think Fresnels are the best way to get an immersive experience. Every time I have used them, at least with monitors, it has been really fun and a different experience than just 3D or HMDs. I have my own design of an HMD that is very similar to yours, except using pico-projectors instead of LCDs. But overall very similar.

So what is the experience like? Do the Vireio drivers work for you?

Bravo !

Very interesting design indeed, it brings some questions :
- how do you intend to synchronize the two displays ?
- how is the viewing quality compared to your previous design which used spherical glass lenses ?
- isn't the binocular overlap a bit limited for a good stereoscopic perception (63° from the ~120° available) ?
- why didn't you go for higher angles toward the nose (31° instead of ~60°) and lower angles temporally ? Does the design makes it impossible ?
- won't the pixels be no more square with such a FOV (121x115° for 1280x800 pixels) ?
- won't the white color be problematic with the reflection of light ?

Thanks for the french ref. about PVC, may be better than foamcore.

EDIT : found this reference for black PVC : http://www.polydis.fr/epages/8d7c81fe-5 ... s/PVCE3N04
- 3mm - 50x40 cm : 2.56 € HT
- 3mm - 50x120 cm : 7.67 € HT

Wow! I am impressed by the specs.

I tried stretching the FoV using fresnel offset lens stacks, but the resolution at the edges is really stretched. Yours has twice the horizontal resolution and (probably) less lens distortion, but at the cost of increased weight. Is the weight load okay, or does it work better with rear counterbalancing weight?

Optolife looks like they offer AR coatings. Thus, those lenses ..could be coated. It is quite a bit more expensive, but..it would probably make quite the difference. Interference from the coatings? Hard to say. But definitely worth the experiment.

http://www.optolife.com/diy_projector_lens/FRL021.html

It is an exciting time for VR.

Very interested in how those angled displays go for you. Going to be tough to get the geometric correction right, and getting the focus correct across the entire field may be impossible.

I have the same problem with my fresnel lens stacks too. But in my case, having the lenses actually touch the skin (nose, eyebrow ridge, and cheek) makes the image visible in all directions. The outer edges only have full coverage when the lenses are angled as shown in this thread.

Regarding not being in focus, my experience while playing with my lens stacks is that FoV is more important (for me) than being in focus. The stuff outside you foveal view does not need such a focus. I planned to correct for chromatic aberration using separate displacement maps for each color plane.

As mentioned in my "lens stack" thread, my eventual goal is to use HYBRID lenses that use fresnel technology at the outer peripheral zones (with full skin contact alla around), but keep solid aspheric lens tech at the inner foveal zones. If you wear eyeglasses, imagine the solid aspheric portion as the foveal view inside your eyeglass rims, and the fresnel portion as the peripheral view outside the eyeglass rims. Just like with real glasses, you can move your central gaze beyond your eyeglass rims. At least that is what I want. Getting around to it is the price.

In the dual-monitor version shown in this thread, similar methods may also improve the FoV (but it already sounds awesome as it stands now). My point was that for me, focus is less important than FoV, and in fact, a slight loss of focus can actually help reduce screen-door effect to some degree.

The problem is not so much the lack of proper focus as it is the disparity between the focus of each eye. Suppose that when looking directly forward, the center of each image is focused at 10 meters. Because each lens/display is tilted in opposite directions, the focus is going to change as you move your view across the field. Hypothetically, for example, you may be looking 10 degrees to the right of center, with the right eye having to focus at 5 meters and the left eye having to focus at 15 meters.

In addition, when you move your eyes, their physical position relative to the lenses is changing, changing the distortion slightly as well. Your brain can accept that pretty well when both eyes see the same geometrical changes, especially since those changes are the same no matter which way you look. I have done tests of setups like this, and a big problem is that any eye movement will result in mismatched distortion and focus between eyes that can cause significant eyestrain.

If eyetracking technology were more advanced, you could solve as least some of these problems with realtime distortion adjustment.

Thank you guys !

while I'm trying to upload a video of the HMD to youtube, I'll try to answer some of your questions

@cyber:
TBH I don't know yet how is the experience, I need to finish it before I can tell but I bet it will be awesome. I'll try vireio driver at this time does it support YEI tracker ?

@Fredz:
- I use a Zotac mini DisplayPort to dual HDMI, it is recognized as a single 2560x720 monitor under windows and ubuntu, that's how I manage synchronization (but i don't know if it is really synchronized until I test the completed HMD)
- Image quality is somewhat degraded compared to real glass lenses of course, but it is still really acceptable (maybe you can get an idea with the upcoming video).
- The drawing with the HFOV comparison was made using my own MRI so I could have my exact IPD, nose bridge position etc.. and I calculated that my stereoscopic vision is about 90° (you can see that the 90° FOV of the Oculus Rift Dev kit is limited by my nose bridge).
- the design is limited by the FoV per eye which is calculated from focal length (3 stacked fresnel of 120mm focal -> 1/f = 1/f1+1/f2+1/f3 -> f = 40mm), lens witdh (I choose to place the optical center not too far away from the eye to minimize distortion and image degradation, my lens stack is 126mm wide) and display width (149mm) that gives 121° horizontal FoV per eye. I just wanted to reach 180° because I like to tell myself that I made a 180 HFOV HMD ^^
- Pixels are square, non square pixel would involve different magnification power on horizontal and vertical dimension of the fresnel lenses and it is not the case. This is one point I'm not completely satisfied with the current design and which I have to improve, it occludes some pixels..
- I thought about that but since I didn't find some "fond d'encadrement noir" at Castorama, I used white and it doesn't seem to be a problem (but I did not test the HMD for a long time for now)

I don't know the density of the white expanded PVC (I assume it is the same for any color) but be careful, the empty box folded is 100g with 2mm PVC, you may add 50g using 3mm thick PVC sheet.

@geekmaster:
I read your posts about fresnel lens stack Actually I read a lot of post here on MTBS3D even if I'm not participating a lot (mainly because my english is not so good and I don't know of to tell things sometimes), and I was quite happy to see that I was not alone to think about using stacked fresnel lenses, it gave me motivation to finish the construction of the HMD.
The weight is not too much thanks to the light weight LCDs, simple design and the use of fresnel (I can't imagine the weight of equivalent glass lenses)

@KBK:
I didn't know PMMA fresnel could be coated (worth the experiment indeed ). (FRL021 is the type I used)

@Palmer:
I think a drawing must be better at explaining why the focus is maintained at infinity across the entire FOV than me and my english ^^, here you are

Also I don't think that the distortion changes when I rotate my eyes (it would be the case if I had the ability to translate them^^) so designing the warp shader will not be so hard (maybe something tunable depending on IPD, astigmatism, etc.. that I can set once at the beginning, saving a profile file for each user).

@everyone: Note that when I talk about FOV I only take into account rays that could theorically hit the center of retina (if I could turn my eyes enough to look 90° to the sides ^^). In reality the eyes get rays from a wider angle because the lens of the eye is not in the center of rotation (you can test it easily: look straight forward with left eye closed and place a finger of your left hand at arm length and at the limit where you can see before it get occluded by the nose. If you turn your right eye in the direction of the finger, it disappears (I don't know if I explained that correctly :s)

...
It looks like it took me a long time to write all these answers, the video is ready (sorry about the quality I only have my iPhone able to take the video close enough to the lenses, it doesn't give what it's like to see with your eyes)

foisi, your device looks great! Thanks for the explanations. The use of fresnels is very exciting. Any thoughts on how they could be improved - such as application-specific geometry and/or groove density?

By the way, your English is very good. Please don't let the language barrier prevent you posting more when there's something you want to share or comment on

I was worried about that. So far I only tested a single lens stack. I was hoping their symmetry from using stacked left and right offset edges would compensate. I still plan to try it though, hopefully soon. You have been doing these experiments much longer than I, but I hope I can still find something useful like the hybrid lens idea I posted elsewhere. If you can keep the blur restricted to the peripheral vision, I think it would still add to the FoV by surrounding the aspheric lens with a fresnel lens.

This project looks great foisi, love the improved resolution and FOV over the Rift. If it avoids the focus issue mentioned by PalmerTech, this will rock. The more FOV the better, IMHO!

Nice job man. Can't wait to hear how this turns out.

Awesome job with this!
You'll have to let us know how it goes in practice once you find a driver that can handle it.

Hi,

I have been wondering whether the principles of tilt/shift photography can be applied to the lens and screen arrangement of head mounted displays - to remove or minimise any distortion effects caused by the relative placement of the screens, lenses and eyes.

Tilt/shift photography seeks to correct perspective effects in a scene by applying a geometrical optical correction which effectively 'squares up' the image. It sounds a bit like the perspective correction process used for interactive whiteboards.

For a head mounted display, there may be an optimum tilting or shifting of the lens which balances the perspective of a tilted or offset screen and corrects the change in magnification across the field of view as it appears to the eye.

If anyone can reveal a similar viable principle for virtual display optics, I would like to know.

Thanks.

scheimpflug principle :

http://en.wikipedia.org/wiki/Scheimpflug_principle

Cool! I was not aware of that. The "hinge rule" shown there looks interesting. Nice and simple, making it easy to remember and to apply. I like simple...

Brilliant job foisi!

Funnily enough, i asked if this type of design could work in one of my very first posts on MTBS last July (with crappy MS Paint diagram included and all):
http://www.mtbs3d.com/phpBB/viewtopic.php?f=140&t=14777&start=648

Now i understand it's not so simple, but i hope you can prove that it's possible in fact! If it is, this could solve 3 of the big remaining issues for consumer VR at once: even wider FOV, higher resolution, and a nicer HMD profile.

Best of luck, and looking forward to your progress!

For now I can say that there is no such effect with this design. I didn't read all the stuff on the wikipedia article yet but I can say that it doesn't look like this



The lenses are parallel to the screens and focus is set to infinity so in theory every ray of light cast by each pixel of the screen come out of the lenses in parallel
I repost the drawing I made for palmer to illustrate that



I did some testing to see why some pixels of the screen are hidden because of the high magnification the 3 stacked lenses (resulting in a 40mm focal length system) and while in theory it should be a perfect match, it is not. I tried with 2 stacked fresnel placed at 60mm from my nexus 7 and it seems to be OK.
I will have to make another box with the screens 20mm away from where they are now and using 2 fresnel lenses per eye (the weight should be in the same range as now). It will improve the image quality because there will be more pixels for the same horizontal FOV per eye and also because light will have to go through 2 lenses instead of 3

I'll update as soon as I can, don't worry I'll prove that this design is possible

edit : Fredz, you were right about reflection of light on the white material btw, I'll have to do something about that.

Very impressive work

I hope that it will be compatible with the game developments for the Rift, it would be cool to have a highend VR alternative for different situations where the weight of 2 screens isn't a problem.

Too bad for the reflections. I've ordered some sheets of black PVC, I'll see if weight is a problem compared to foamcore.

Looking at your last drawing, I wondered if you knew what was the position of the center of the entrance pupil of the eye ? I can't find any definitive answer about this after reading a ton of papers on this subject.

So what do you guys think about Foisi's explanation of the way the eye focuses on his HMD? If we could overcome the obstacles that Palmer mentioned this would seriously rock That said, Palmer's concern about each eye having to focus at a different distance does make a lot of intuitive sense...

Lockheed Martin. Look at the file dates. Probably filed due to others finally approaching their 'private' work.

http://www.google.com/patents/US2012012 ... &q&f=false

There is also a nice illustration of what Palmer was speaking of. The illustration shows that the distance from eye is kept even in all viewing directions.

It does give a nice illustration showing the issues that need be overcome, which appears to be a layer of separated Fresnel, separated according to the optical considerations in the given 'area'.

This is what I expected when it came to the Fresnel lens and wide FOV. Layered acts of collimating Scheimpflug into an exit pupil, not a focal plane -done in stages, individually, per Fresnel. You need the positioning of each, and minimal warping per component to stay away from distortions that cannot be easily corrected. In this way the corrective coatings and the Fresnel can be at their most effective. That one might be able to press a Fresnel into the correct shape. Flat will not do.

figure 16 is particularly interesting.

one side of the Fresnel is flat. the issue is damaging the ribbed surface.

Has anyone here ever done any acrylic shaping?

well, you can pressure or vacuum form, with a heat gun as the assist to get the flow required.

since the eye is looking for a spherical curvature, this works in your favor.

You can build a mold that is like a rectangle, but is a cut from a sphere's skin. this forms one side of your mold. which ever way you want to go, pressure, or vacuum. Not too many PSI will be needed.

You make it larger than the required lens, so you don't damage you lens area you need. Just a hair larger. place this in sealed cavity, with the open area inside the rectangle in open air. The lens needs to be placed so it sucks itself onto the surface of the rectangle, at the same time the pressure forms the spherical shape, at the same time the low heat allows the lens to reform. Via air pressure alone, so the Fresnel len's primary functionality is not disturbed.

something like that is doable.

I received the black 3mm PVC and it's awfully heavy. Also it's glossy, so I'm not sure it would be a good candidate for DIY HMDs. I'll stick to foamcore for now...

couldn't you tape black sheets of paper to the inside?

Fredz, I think it's not expanded PVC like the one from castorama because they are absolutely not heavy (check the weight on the website. I think they changed their provider though, because I use less than half the area of the sheet and the box weights ~100g)
Taping black paper or painting is a good and cheap solution (even with a black permanent marker it should be fine, you only need to blacken the inside up and down where the light might be reflected)

I made some changes to the plans and printed them, hopefully I will be able to build another box at lunch time today and test the 2 fresnel per eye design this evening.

About scratching the plastic lenses problem, I thought about using 7inch tablet screen protections (the flat side of the fresnel lenses is directed towards the eyes so there is no risk of scratching the ribbed surface)

Nice job Foisi, your HMD looks awesome.

airons1972

Hi,

I just did a quickie experiment with an A4 fresnel sheet magnifier and my 20 inch CRT monitor.

Just some observations:

1: The fresnel lens works much better in one direction than the other - in this case with the ridges on the screen side.

2: There was significant curvature of field of the image, but the image was fairly clear from edge to edge and I did not see any ridging.

3: There was noticeable chromatic fringing outward from the central position.

4: It seems maximum magnification appears when the lens is midway between the screen and the eye.

5. When bending the sheet about the vertical axis, the horizontal image curvature was lessened and the vertical image curvature was increased. This 'toroidal' effect might be quite useful. It looked good on the 4:3 aspect ratio monitor.

The various geometric and chromatic distortions could be corrected for in software, to give a convincing virtual reality display. I am not too sure if I would use just a single fresnel lens sheet for viewing with both eyes though.

Overall, this simple experiment shows me there is great promise for the fresnel lens in head mounted displays.

Thanks.

Exactely, but it can be resolve with flexible fresnels with the focus center in front. The lenses could be curved until the edges are placed 180 deg from each eye. This way, the spherical distortions on the left and right eye are the same. If not, there coud be a severe stereographic error. There are some products on the market that uses also tilted optics similar to this HMD but the optics are rather aspherical onces.NVIS nVisor SX111 http://www.nvisinc.com/product.php?id=48 and SEOS http://cgsd.com/SEOSHMD/
Damn, I forgot to mention the screens.

I'm not sure to understand why people tend to believe that focusing would change depending on the position of the eye... for what I know the focus is set by the distance between the lens and the screen. This distance is the same for the whole FOV because the lens is parallel to the screen. There is chromatic aberration (that can be corrected in software) and spherical abberation (because the fresnels are not aspheric lenses) but it doesn't have a significant impact because the eye is near the optical center

New version :







note that there are black pixels at the bottom and at the top of the panoramic image I used to make something displayable with the HMD so vertical FoV is not as it should be in the video.

I didn't paint the inside in black (or tape black paper) for the moment, I just rebuilt the box in a few hours today at lunch time and this evening just to try to prove people that the design is possible.
I took about 20 pictures when I was building the new box so I will be able to make some sort of tutorial if anyone is interested in making a similar HMD (I think it deserves a name, any suggestion ?) and I will release the printables very soon.
My mom who happened to be visiting me today tried the HMD when I just finished the new build and she said something that could be translated by "It's f***ing awesome !" ^^

The new weight is 512g (~50g less than the previous version, the PMMA lenses are heavier than the expanded PVC)

Congrats foisi!!

I think this is a BIG step forward for VR displays!

EDIT: It does look like it from the videos, but did you notice a big improvement in image quality in comparison to your first prototype?

This is awesome, simply awesome! I would die to have Oculus hire you and give us a 180 degree FoV, 1920x1200 consumer Rift.

A good name would be the "180", or "Pano" (as in "panoramic"). Pair it up with a company name and it would sound good.

For example: "Virtua 180", "Visualize 180", "VisualWorks Pano", "VisualTech Pano"

This is great stuff! Have you been able to determine if the displays are in fact synced?

As for a name, the first thing that popped into my mind was "InfinitEye", or some derivation such as "Infinite-i"

Looks pretty good and the weight is not that bad. I wonder if it can be lowered a bit more, do you think using foamcore instead of PVC would help ? Also are the controller boards mounted on the HMD or in a separate box ?

For the name I guess you could use your nickname to personalize it a bit, after all it's your own creation. Something like the Foisi 180 for example.

@PatimPatam: yes I forgot to say that, image quality is way better than before (mainly due to the fact that there are only two fresnel lenses per eye compared to 3 in the previous prototype)
Even without applying any counter-distortion effect and chromatic corrections, the result is perfectly clear across the entire 180° horizontal FOV.

@Zoide: Maybe Dycus could make a prototype when I release the build instructions to have Palmer admit that it works ^^
even without the prints, the important things of the design are :
- 2 stacked fresnel lens 165x105mm cut with a dremel to make room for the nose and resized to 126.5mm wide (optical center is 44mm away from the cut side) 120mm focal. The flat surface need to be set toward the eyes (both fresnels for each eye)
- the distance between the surface of the screen and the flat surface of the fresnel lens that is the nearest to the eye is 66mm.
- the angle between the screens is 134°.
- the two panels are set like in the second drawing to gain a few mm (because bezel are not the same at the left and right of the panel) :


For the name I kinda like Pano-something =) I do not intend to make a company and sell the HMD as a product. I think this design will be released as an opensource HMD so DIYers can make their own for free (except the cost of components) and improve the design etc..
If the Oculus SDK is flexible enough, the only things to change in Oculus ready games will be the angle of the virtual camera (23° divergence for each from the central axis), FOV of each camera (set to ~121°) and distortion correction shader.

@MSat: I still don't know if the displays are correctly synchronized, it only depends on the Zotac device.. If a company like Oculus were to produce a dual screen HMD like this one, I hope they design a dual LVDS, single input controller board to solve this problem.
InfinitEye sounds good, but doesn't it sound like AMD eyefinity ?

@Fredz: to answer your previous question, I don't know where is the eye exact iris position.. sorry
The weight of 512g take into account the headmount and you're right it's not bad at all
The controller boards are separate (but not in a box yet, I still have to extend LVDS cables and make the box). About foamcore (ca veut dire carton-plume c'est bien ça?) can you tell me the density g/cm^3 so I can compare to expanded PVC ? Don't you think foamcore is a little weak to resist from tearing at the mounting points ?
Oh and Foisi is a terrible nickname (I will not explain why I chose it), it sounds a lot like "moisi" (french for moldy) and I don't think people want to wear something moldy on their face (even if it has 180° HFOV)

Not as much as Sony's new EyePad sounds a lot like Apple's iPad.

But then "EyePhone/iPhone" was used in the film "Johnny Mnemonic", so who knows where Apple got their inspiration for that product name:
viewtopic.php?f=120&t=16464&p=101310#p101310
viewtopic.php?f=120&t=16464&p=101310#p101452

The people who do the most copying are the most protective of the stuff they copied. Take for example the Mac GUI copied from the Xerox Star, then they sued others for copying the same stuff:
http://en.wikipedia.org/wiki/Xerox_Star

I like "InfinitEye" (especially for a non-profit Open Hardware project). I think if Sony could be so close with commercial "EyePad", you should be safe (not legal advice, see a professional for that).

I LOVE your design and the results you claim for them. After my experiments with stacked fresnel lenses, I know that the immersion provided far outweighs any sacrifice in focus or image quality caused by fresnels. I am looking forward to trying one of these (probably requiring that I build one). I still plan to build my 7-inch "Fov2Go" clone inspired by the "HMD kit" included with my "Virtual Reality Creations" book. It would be very similar to yours, but would use a single display (my Nexus 7).

Virtual Reality Creations "HMD" (circa 1993):
http://www.mitchwaite.com/writing7.htm


Notice how the fresnel lenses wrap around the face, just like the HMD featured in this thread. It worked surprisingly when when attached over a stereoscopic image pair on a CRT. Extending this concept with newer better fresnel lenses it certainly bound to succeed. Your nose cutout allows for taller lenses, giving a much larger vertical FoV than the VR Creations model.

Thanks for keeping us posted on your progress!

Epic stuff! This is a 100 times more exciting than all the boring PS4 news from Sony!

Btw should this thread not be moved to the "VR/AR Research & Development" section?