Fresnel lens stack for "supernatural" FoV

This is for discussion and development of non-commercial open source VR/AR projects (e.g. Kickstarter applicable, etc). Contact MTBS admins at customerservice@mtbs3d.com if you are unsure if your efforts qualify.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Information for this topic is scattered in multiple related threads, but it is time for its own thread.

I will pull in quotes from my other various posts on this topic, so you do not need to jump all over the place to follow the previous scattered "thread". Please forgive any duplicate post contents here. I really should have started a new thread to begin with.

UPDATE: Be sure to check out the new "Fresnel Lens Stack Theory of Operation" post (with lens position diagram) here:
http://www.mtbs3d.com/phpBB/viewtopic.p ... 865#p98869

geekmaster wrote:Although I have the recommended 5x lenses, I just tried an experiment with a cheap dollar store "Page Magnifier Bookmark" plastic fresnel lens with fine pitched ridges. It is about 2 inches tall and 6 inches wide. I cut it into 3 pieces each two inches wide and overlapped them (all with ridges toward the eye, no rotation). Then I cut out the corners to fit my nose and inner eye brow. When in focus on my Nexius 7 screen, it completely fills my horizontal FoV, including where my nose should be and as far as I can stretch my gaze away from my nose. That is more FoV than my natural FoV. Although it causes chromatic aberation on text (mostly blue), that is rarely noticeable in a game environment.

I played one of cybereality's Rift-adjusted videos from the Vireo Perception thread at fullscreen on my Nexus 7 (7-inch LCD), and it was great. If you normally wear glasses, but blurriness at the corners is less than from real glasses (even large aviator-style lenses). In fact, after having such a super-normal FoV, it makes me much more aware of my eyeglass frames (and my nose) which are obstructing FoV that I saw inside the game environment. MUCH better than I had expected, and extremely lightweight lenses.

I got the idea from an old "Virtual Reality Construction" book that came with fresnel lenses and a cardboard fold-up frame, which had a string to tape to a CRT monitor so it could hang down against the screen and to use like binoculars on a displayed stereo image. These cheap dollar store lenses are much higher quality (finer pitch) than used in that device.

By overlapping the fresnel outer edges, their extreme offset simulate looking through the edges of the 5x lenses causing more geometric distortion (desirable in this case, pushing the image beyond the nose boundary).

Any perceived distortion is quickly lost in the game environment in my experience, but it could be compensated in software (including shifting blue for chromatic aberration adjustments).

And these lenses are cheap too (only one dollar per eye from my local dollar store).

Here is the lens I used (UPC 731015162413):
http://www.dollardays.com/i789846-whole ... ml?print=1

This is a popular lens available from many stores, but most places sell it in multiples of 24:
http://www.antarespro.com/5977131-item- ... 62413.aspx

Image


I may mount my cut-down fresnel lenses very close to my eyes in modified swim goggles (also available from my local dollar store):
http://www.fishingdiscountdirect.com/pr ... direct.com

Image


And I may try attaching my Nexus 7 to dollar store saftey goggles like these (but they may need a better strap):
http://www.dollartree.com/Tool-Bench-Sa ... /index.pro
Image

Image
geekmaster wrote:
Mel wrote:Interesting stuff, geekmaster. Can you clarify the steps for how you modified the lens? A photo is worth a thousand words :-)
...
I guess it's that 'stacking' thing I'm asking about. You said you cut it into three pieces, so what's the stacking arrangement such that three pieces covers both eyes?
My camera battery is dead so words will have to suffice for now.

I cut the fresnel portion of the lens into three pieces about 2 inches wide. Then I stacked them with the ruler marks at top and bottom and ridges toward my face. I don't think the stacking order matters, but I stacked them left edge piece, then right edge piece, then center piece (starting from the side closest to the eye). Then I trimmed the edge toward my nose until the lens stack fit comfortably under my eye brow and overlapping the bridge of my nose. Be sure to keep the ridged (non-smooth) side of all 3 lens elements towards your eye. Then do the same with another lens stack for the other eye.
geekmaster wrote:
cybereality wrote:Sounds good, geekmaster. I always thought that fresnels would work well.
I am surprised at how well it worked with these newer cheap fine-pitched fresnel lenses. By stacking them, I am taking advantage of the extreme off-center distortion from the edges of these 6-inch diamemeter lenses. Due to my stacking arrangement it does not stretch vertically near the edges as much as horizontally, but still it looks amazing, and the correction you used in your "rift-adjusted" video actually shows doorways as rectangular in these lens stacks. There is more chromatic abberation near the edges (mostly blue), but that could be adjusted in software. Even when I rotate my eyes in their sockets (painfully) to their extreme positions, all I can see is video (even where my nose should be), giving me a "supernatural" FoV.
:woot
For low-power devices (no GPU) I plan to use a software displacement map to do the geometric and chromatic correction, like this:

Image

I used displacement mapping like this for animated "magnifying glass" effects back in the early 90's, when 360x480 256-color VGA (Michael Abrash's "Mode X") and 386 CPUs were state-of-the-art. It should work well on low-power devices to correct for lens distortion.

BTW, the fresnel lenses that came with the cardboard "HMD" in the "Virtual Reality Creations" book were much coarser pitch than these dollar store fresnel lenses, and consequently did not work as well. Here is that book:
http://books.google.com/books/about/Vir ... AIAQAAMAAJ

And these are listed at Amazon as "new", so may even come with the fresnel goggles:
http://www.amazon.com/gp/offer-listing/ ... dition=new

I have been into 3D for about 55 years now. It was always a big thing in my family. I have my grandmother's stereopticon viewer and a large collection of 3D cards for it:
Image

I was thinking about trying to stick my Nexus 7 into my stereopticon in place of a 3D card. A classic steampunk HMD, for sure... ;)
geekmaster wrote:I need to get another fresnel magnifier. I only have a fresnel lens stack for one eye. I just tried the Fov2Go demo on my Nexus 7 (held in my left hand), while holding the lens stack (with taped top and bottom edges) to my right eye (close enough to brush eyelashes on it). With my left eye closed, I turned slowly around while standing, which navigated me around in the VR space of the Fov2Go demo. When turning at a casual pace, such as when exploring the scenery, it was very immersive without noticable latency. However, when turning quickly, the latency caused an immediate dizzy feeling that persists several minutes later (along with a feeling that I was punched in the stomach). It does not help that I just ate a meal a few minutes before trying this demo and I still have a full stomach.

Now, turning my head OUTSIDE the VR environment increases my vertigo. I am normally not prone to motion sickness, so perhaps the "supernatural" FoV increased my proximity to the "uncanny valley" (which is known to cause queasiness)...

I will need to turn slowly when using the Nexus 7 internal IMU (or at least the Fov2Go implementation of it). Now, I am anxious to get another fresnel magnifier, and to construct a cardboard and tape mount and shroud to hold the lenses in place at the correct position in front of my Nexus 7 display. I will try using dollar store safety goggle first, before purchasing more expensive sky goggles.

The goal here is to create a template for an extremely inexpensive add-on for an Android tablet, using locally available cheap components. Just for "fun", you know? :P

Then later I will use my 5x loupe aspheric lenses (if I can remember where I put them "away").
geekmaster wrote:
MSat wrote:Very interesting, geekmaster! I'm trying to wrap my head around the premise of the stacked fresnels, and why it should work, but I might just have to stop by the dollar store tomorrow and see what they got to try it out for myself. What is your thought on image quality? Does it remain clear overall? Also, if they're giving a super wide FOV on a 7", would they still work well on smaller displays (I suppose this really depends on the fresnel design, correct)?
How much it stretches the image depends on how close it is to your eye. Mine focuses with the triple stack at less than two inches, but I am near-sighted (myopic). YMMV. For a smaller display you may need another lens element for greater magnification.

At the "sweet spot", the image is surprisingly good with sharp focus. With different fresnels, it may not be as good a focus. I have seen many fresnel magnifiers with much coarser pitch (sginificantly fewer ridges per inch). I recommend the ones I showed in my first post on this, but at least get the finest pitch fresnel magnifier that you can find.

The angle of the lens stack is important too -- it should be roughly parallel to the screen surface (but adjust it for best image). And very close to your eye, which required trimming to clear your nose and inner eyebrow ridge. And ridges inward toward the eye for all lenses too...

The lens stack using the outer edges of the 6-inch lens scretches the image horizontal edges much more than vertical. However, and my optimum focus, the vertical (just barely) covers my full FoV. I actually lose a small amount of the horizontal image, but nothing that detracts from immersion.

I will take some pictures of the lenses and through the lenses when I get around to charging my camera battery. :mrgreen:
geekmaster wrote:
MaterialDefender wrote:@Geekmaster: how bad is the chromatic aberration with fresnels? I already wondered whether fresnel lenses might make sense some time ago, but came to the conclusion that CA would most likely be much worse than with good aspheric lenses.
The chromatic aberration is mostly at the edges of the total FoV, with blue stretched farther than other colors. This causes objects to have a bluish inner edge and yellowish outer edge, mostly visible on text or other sharp high contrast content. Not very visible on "realistic" scenery such as in games. And it does not detract significantly from immersion, which is the main goal here. This is not a desktop environment. And if required, the chromatic aberration could be compensated in software.

The centers are sharp and devoid of chromatic aberration. And with head tracking, the edges are just peripheral vision and not a big problem. Just having moving VR content there (even if distorted) is vastly superior than having that portion shrouded from view. You really need to experience it. After having a "supernatural" FoV, going back to normal vision makes you very aware of the visibility of your nose, eyebrows, and cheeks as you roll your eyes around in their sockets. All those natural occlusions contain VR video with my fresnel lens stacks close to my eyes, and although those boundary pixels are somewhat distorted, they ARE part of the peripheral vison and contribute greatly to immersion.

The key to this is the very fine pitch (ridges per inch) of the fresnels I used (and recommended above). These work MUCH better than attempts in the past that used much coarser pitch (fewer ridges per inch) than these. I believe that the focus is better and the chromatic aberration is less annoying with these newer cheaper fresnel magnifiers.
geekmaster wrote:
MaterialDefender wrote:You make me really curious... Since I'll disassemble my build anyway to put in new lenses, I guess I will try to find some good fresnels too for testing.
To get maximum FoV, the fresnels are trimmed to fit under my eyebrow ridges, and against the side of my nose. A pair of lenses actually touch together on the upper nose bridge (but below the center of the eyebrow ridge).

The images are stretched far enough the even with full FoV there is no visibility of the other half of the screen (but the outer corners are outside the FoV). When inside a game, it really is an amazing experience. Like I said, this arrangement provides a "supernatural" (beyond natural) FoV. Cutting down aspheric lenses may not provide this experience. I think it is the stacking of the outer 1/3 of 6-inch lenses with the center 1/3 that contributes to the extraordinary horizontal FoV.

Using aspherics may provide a better quality image, but at a reduced FoV. Immersion is more FoV dependent than image quality dependent, in my experience.

I have yet to try a pair of lens stacks for a stereoscopic test, but I do not see why it should be any problem.

Beware that the lens stacks must be in the "sweet spot" (try adjusting their position) or the result will be unsatisfactory. My fresnel lens stack may be optimized for my near-sighted vision, so YMMV.

If others try this, please report back...
geekmaster wrote:For an understanding of why these stacked fresnel elements work like this, look up information on "eccentric fresnel". I am using one outer 1/3 of a 6-in lens to stretch the inner side of the image behind my nose. The other outer 1/3 stretches the outer edge of the FoV so all you see is pixels even when painfully rolling your eyes to the outer limits. And the inner 1/3 adds to that, and also stretches the vertical FoV above my eyebrows, and below my cheeks. It is like slicing smaller lenses from the outer edges of a large 6-inch lens, resulting in a large sidewayes stretching of the FoV.
geekmaster wrote:
PalmerTech wrote:Some pictures would be great! I have been messing around with things like this, but I want to test your actual setup. I have the fresnel lenses you used on order, should be here soon.
Cool! I need to find my camera battery charger now...

EDIT: Found it. Charging...
geekmaster wrote:
3dvison wrote:That would be great, because I am not sure where the cut edges go ? Are the three stacked one on top of the other or are they butted edge to edge or even if they are shifted a little left and right, can't you then see the edges of the center fresnel ?
As I mentioned earlier, I cut the 6-inch wide lens into 3 2-inch wide lenses then stacked them face-to-face with the ridges all toward the eye. I trimmed the edge against my nose and eyebrow so it all touches the skin of my nose, under my eyebrow, and my cheek.

I just looked into a mirror while holding my lens stack to my face and I see that it touchs the corner of my nose where it joins the cheek above my lip.
geekmaster wrote:
Fredz wrote:
geekmaster wrote:I got the idea from an old "Virtual Reality Construction" book that came with fresnel lenses and a cardboard fold-up frame
This one I guess ? :
Image
Yes, that is the one that gave me the idea to try this. I still have it, but the lenses fell out of it. They are around somewhere. That tab on top has a red string tied to it (included in the kit), which gets taped on top of the CRT so that you can hang it down over the screen and use it without holding it, or place it on top of the CRT when not using it. It worked fairly well, but you had to lean into it to use it, and you had to size and center your 3D image pair to fit into it.

Those were very coarse pitched fresnel lenses, and the ones I used for this experiment are very fine pitched (so much better image). Also, to get the maximum area of the screen in focus, I am holding them roughly parallel to the screen surface.

I just bought 16 more lenses to play with. I just cut one up and got a couple of photos before my battery was dead again. I lost my original charger while travelling, and this replacement charger takes forever and does not seem to give a full charge... :-(

I have a photo of it held up to my left eye, taken while looking in a mirror, so you can see how it touches the underside of my eyebrow ridge, and the side of my nose down to my cheek, and part of my cheek... I need to stack them together and trim their edge to match the curve now cut into one of them, then take more photos.

Now, I need to charge my camera battery overnight, and I need to find my microSDHC card reader so I can post the photos I have thus far... :-(
geekmaster wrote:
3dvison wrote:OK, I my be slow on the up take, but this is what seems to work for me, and it may well be what has been said here all along but I just did not understand it.

Found what seems to be very close to the same lenses at Office Depot.
For me using the center portion is what gives the best image. So if you wanted a 2inch dia lens you would start at the 3inch mark which is the center of the 6inch lens and go 1 inch left of the 3inch mark and cut and 1 inch right of the 3inch mark and cut.

So I needed 3 full 6inch lenses to make 1 lens. The lens is made from 3 identical lenes all cut from the center portion of the larger 6inch fresnel..
I will have to try that. I started with one fresnel magnifier. It is cut at the 2 and 4 inch marks, just like yours. It works well enough, but perhaps your stack of 3 centers would be even better...

Using the eccentric (off-center) edges requires keeping the ridges toward the eye. Otherwise the image distortion is not correct to compensate for the Rift-adjusted images.

With eccentric lense, the image can stretch beyond normal facial occlusions such as the nose. It that the same for only using lens centers?
geekmaster wrote:
3dvison wrote:Ah Ha,
My stack has the ridges facing away from the eye.
I was testing it with pre-warped Rift photos.

The advantage of this stack over a regular round lens, is the rectangular shape, you dont feel like you are looking through a port hole.
Try the eccentric (edge) portions in your stack, with the ridges facing toward your eye. Be sure to cut notches for your nose and eyebrow ridge in the inner edges (but not too much or you lose too much of the outer edge). I was actually holding mine below the eyebrow inward toward the eye. The key to stretching half of a 7-inch screen to full FoV is getting it very close to the eye, which requires trimming a little to fit the nose and inner eyebrow ridge...

Because the lens portion of my magnifiers is a bit larger than 6-inches, my elements were about 1/8-inch wider than 2-inches.

I still think the eccentric portions from the outside of the 6-inch lens are important to get the full FoV that I am describing. One side stretches the inner image beyond the nose with none of the other image visible (and not even the nose visible). The other side stretches the outer image to the edge of the lens stack without showing the left border of the video image. The center part gets rid of the top and bottom margins...

I have not tried 3 centers yet. I bought 16 more magnifiers so I will try various arrangements to see if I can improve on it.
geekmaster wrote:
Zoide wrote:I can't wait for Palmer to buy into geekmaster's Fresnel lens idea so we can have a consumer Rift with superhuman FoV :woot

Palmer: Any news regarding your experiments?
My camera battery is charged now, so I should get the photos that Palmer requested posted some time this weekend (I hope). Demands on my time have been just nuts lately...

Hopefully those photos will help Palmer (and others) to duplicate my results. I tried cutting another lens stack to fit my nose more snuggly, but that reduced lens coverage of my outer peripheral vision. Getting (almost) full coverage at the outer edges requires the lens stacks to be positioned just right, so they may not provide total FoV for everybody (depending on facial features).

What I would really like to try is a custom acrylic lens that can duplicate my fresnel lens stack, but with superior image quality (necessary when the Rift gets a higher resolution display).
geekmaster wrote:
Diorama wrote:I've seen others mention that the bottom of the screen being invisible is really important for immersion; in fact in some of the official oculus dev kit photos it looks like they may have also placed the screen slightly offset down:

Image
That makes sense. I cannot look up past about 45-degrees before I see my eyebrows, but I have to look down at a very steep angle (about 80-degrees) to see my mustache. The Rift photo looks about right for those vertical viewing angles. That is also how I held my Nexus 7 while testing my fresnel lens stacks, so that I could fill that FoV (and somewhat beyond).
geekmaster wrote:I did an experiment with fresnel lenses, where I stacked 3 elements such that the left and right edges of the images are stretched, then I trimmed the inner edge so it fit onto the upper bridge of my nose but under the inner eyebrow ridge. That places them very close to the eye, so the eyelash brushes them. The side of the fresnel lenses with ridges must be toward the eye on all 3 lenses in the stack. It gives me a view of my 7-inch Nexus 7 screen (showing a rift-compensated video), which extends beyond my normal visual borders. I can see video where my nose should be, and even at the extreme limits of rotation of my eyes. In fact, there are pixels everywhere I look with no obstructions (well beyond the limits of my eyeglasses for my normal vision).

The inner edge is cut somewhat similar to what PasticheDonkey showed:
Image
but the other edges are more rectangular so they cover all of my field of view.

If you can get the lenses close enough to your eyes (using a flat or concave inner surface), you can have a FoV beyond what is natural. Supernatural FoV. ;)
geekmaster wrote:
PasticheDonkey wrote:...
would maybe have to render a nose and brow etc. i think chromatic abrasion could be sorted in software as well.
The perceived chromatic aberation is mainly blue (and mostly visible on text displays). In a 3D game it is rarely noticeable except when you look for it. And yes, you could adjust for it in software (as mention in the other thread).

Unlike your lens images, mine have the tops trimmed flat so the fit below my eyebrows so I can get them very close to my eyes. They must be very close to fill the entire FoV even with extreme eye rotation. In my case, I am near sighted so it is all in focus even with complete FoV coverage. Pulling the lenses away from the eye even a small amount reduces the FoV coverage quite a bit. If my focus does not work for you, you may need to add another lens element to the stack.

Stangely, after using this for larger than normal FoV, taking this off makes me very aware of natural vision occlusions such as my nose, eyebrows and cheeks (but especially my eyeglass frames and unfocused FoV outside them). Normal vision feels restriced and limited after such a "supernatural" FoV experience.
geekmaster wrote:
PasticheDonkey wrote:in that case moving the screen rather than lenses could adjust focus.
Moving the screen away brings the screen edges into your FoV. Better to increase magnification and lose a few pixels around the edges than to bring unnatural occlusions into the picture.

As it is, the edges of the FoV are not in good focus, but neither are they in real life when wearing eyeglass (like I do). That does not detract from the experience, because when you see motion in the corner of your eye you quickly rotate your head to bring the movement into the center of your FoV where you have overlapped stereoscopic convergence. In the "heat of battle", you do not notice any lack of focus around the edges...
geekmaster wrote:
PasticheDonkey wrote:...
would maybe have to render a nose and brow etc. i think chromatic abrasion could be sorted in software as well.
The perceived chromatic aberation is mainly blue (and mostly visible on text displays). In a 3D game it is rarely noticeable except when you look for it. And yes, you could adjust for it in software (as mention in the other thread).

Unlike your lens images, mine have the tops trimmed flat so the fit below my eyebrows so I can get them very close to my eyes. They must be very close to fill the entire FoV even with extreme eye rotation. In my case, I am near sighted so it is all in focus even with complete FoV coverage. Pulling the lenses away from the eye even a small amount reduces the FoV coverage quite a bit. If my focus does not work for you, you may need to add another lens element to the stack.

Stangely, after using this for larger than normal FoV, taking this off makes me very aware of natural vision occlusions such as my nose, eyebrows and cheeks (but especially my eyeglass frames and unfocused FoV outside them). Normal vision feels restriced and limited after such a "supernatural" FoV experience.
geekmaster wrote:
PasticheDonkey wrote:in that case moving the screen rather than lenses could adjust focus.
Moving the screen away brings the screen edges into your FoV. Better to increase magnification and lose a few pixels around the edges than to bring unnatural occlusions into the picture.

As it is, the edges of the FoV are not in good focus, but neither are they in real life when wearing eyeglass (like I do). That does not detract from the experience, because when you see motion in the corner of your eye you quickly rotate your head to bring the movement into the center of your FoV where you have overlapped stereoscopic convergence. In the "heat of battle", you do not notice any lack of focus around the edges...
geekmaster wrote:
PasticheDonkey wrote:well if the screen size was set so people with whatever vision required it further away could see the whole screen, then others could get focus moving it closer but would lose a bit of res.

how blurry are we talking at the centre of vision with an eye turned to full extreme? anyway this is a problem with all versions thus far.
My screen center is not blurry at all. I can see individual pixels. It is the screen edges that blur, but even so it is much better to have blurred VR content at the visual boundaries than just a blackened shroud.
geekmaster wrote:
KBK wrote:The contrast and resolution junkie in me says that single lens solutions will always be better looking than any set of stacked Fresnels.

That in stack of Fresnels, the air must be there, in order to produce the desired optical effects, and this stacking alone can ruin the CR and definition. The best case scenario is a around an overall what, 5% reduction in both CR and definition, which would look worse than the raw numbers would suggest.

A custom cut single lens is still the best bet, in my offhand analysis.

The single lens is a big advantage, when it comes to a custom cut, for the purposes of enabling eye movement. Regular optics, such as camera/projection and the like, make it so the cut has to be a specific type, but the singular interface of the HMD display and the lens make it so the DUT or lens/HMD display combo, in conjunction with the eye, can make a second cut work. and the second cut, is the one that allows for eye motion.

Glasses for human use are an example of a single cut. that cut staked with that of an aspherical as a secondary or paired concern, may be possible. But I admit I am weak on this particular subject. I suspect that the more expensive HMD's use something similar. However, eye/lens relational positioning then becomes critical. Messy subject, at best.
Yes, glass or acrylic aspherics would be better, but for a really inexpensive experiment, fresnels are good enough to give a feeling of immersion. And the pixels in the center area of of the screen were in sharp focus for me. Although the outer edges were stretched and blurred somewhat, they still added greatly to a sense of immersion, and when adjust "just right" I could read the part of the text visible at the right edge of the "Rift-warped" L4D YouTube video.

My idea was to make a simple cardboard frame with cheap dollar store lenses, similar to the one that came with my old VR Creations book, which could be attached to a tablet computer such as my Nexus 7. I published this because my experiment yielded results far better than past experiments using coarser fresnel lenses, and with lenses trimmed to fit snugly against the side of my nose and under my eyebrow ridge, the FoV exceeded what I see naturally.

Agreeing that non-fresnel lenses would be superior, I would like to duplicate this experiment with solid custom lenses that include outer offsets of 6-inch lenses, just like these fresnel magnifiers. These fresnel stacks may not be good for watching movies, but they are adequate for VR. But then my eyes are not so fussy as I get older. I used to clean my eyeglasses many times per day, but now I just adjust my head to look around the grime. I think that started when I got some scratches in my glasses and never bothered to replace them. Perhaps the blur at the edges that I do not even notice may annoy others who still clean their glasses often...

As you get older, available time becomes more precious, and expedient (quick and dirty) solutions are often more than acceptable. When younger, I was a perfectionist and needed everything scrupulously spotless. But that is not important any more. Just getting stuff done while I still can is most important to me.
:roll:
I am interested to see the opinions of others who try fresnel lens stacks...
geekmaster wrote:
Zoide wrote:... I think your contributions here have been very valuable, and I hope to see your superhuman FoV ideas in the consumer Rift! ;)
A future Rift with "supernatural FoV" would need custom lenses. The fresnel stack is just a prototype. I need to build an HMD with them so I do not need to hold my tablet PC and lenses with my hands. The real key is for the lens edges to actually touch the side of the nose, and the cheek below, and the eyebrow ridge above. That lets you see beyond your facial obstructions. And the offset portion from the edges of a 6-inch lens bends the 7-inch display to fill your full FoV.
geekmaster wrote:With my fresnel lens stack, I have think I am getting 150 degrees horizontal FoV (including seeing video where the nose should be. I do not think it necessary to render a virtual nose occlusion. In fact, my nose distracts from immersion in the real world after taking OFF my VR viewer. Small critical details near the edges would not be good, but even blurred motion there is a huge help for immersion. After this experience, I think that including an "ambilight" for peripheral vision beyond the FoV in the Rift would aid immersion. Unless you are emulating an HMD/HUD, there should be no text or indicators fixed time the screen. They should be on a virtual vehicle dashboard, or HUD on the windshield, or on your avatar wristwatch, IMHO. :)

To maintain the best immersion, we must remember to emulate reality even for our status displays.

EDIT: I just did a little experiment that made me change my opinion about fixed text or status indicators on the display. On my Nexus 7 (with IPS display), I scrolled some text (web page here) by holding my finger still and sliding my tablet, to simulate head tracking. The text got blurry while it was fixed on my FoV but the pixels were moving under it. So small text would require getting your avatar close to it to make it readable. Until we get 4K displays and faster pixel switching time, we may still need status info locked to the screen, but I think that should be avoided if possible. Fixed text is fine for config menus though...
geekmaster wrote:
TheLookingGlass wrote:I can't wait to try your experiment with the fresnels when I get some time this week. I had read some great things about using fresnel lens for Monocular HMDs but haven't seen anyone stack them for a greater FOV. Have you taking any pictures yet of your setup? Or you are holding off until you have experimented more?
I took some pictures, but I really want to do more experimentation. After cutting another lens stack to fit closer against my nose and eyebrow ridge above, it sacrificed too much peripheral vision at the edge.

I also need to make a pair of them, and mount them in some sort of a frame. Because of their symmetry (both left and right offset edges), they should work fine for stereoscopic vision, but it needs testing...

I do not wish to post my CRUDE experimental photos just yet. They proved the concept to me, but after less than satisfactory results after cutting the second lens stack, I want to take more careful measurements so that my experiments are easily reproduceable.

After cutting the lenses into thirds, you just need to take the corners off to fit it closer against the nose and inner eyebrow ridge. As close to the eye as possible gives the widest possible FoV.
geekmaster wrote:
budda wrote:Hi,

Interesting work geekmaster on using fresnel lenses to achieve a super wide field of view.

My ideas above refer to a basic concept (which others have probably thought of too), but you have moved on with a practical realisation.

Post away, but I would have thought a more general thread for your work, such as "Fresnel lens stack for superior field of view" would cover both the fresnel lens approach and its other possibilities.

Thanks.
Done! I just created a new thread:
Fresnel lens stack for "supernatural" FoV
http://www.mtbs3d.com/phpBB/viewtopic.php?f=140&t=16373

It contains large quotes taken from my posts on this topic, scattered through four different threads, plus some PMs.

It makes more sense to follow this information in a single thread.
Last edited by geekmaster on Tue Apr 25, 2017 10:09 pm, edited 8 times in total.
TheLookingGlass
Two Eyed Hopeful
Posts: 73
Joined: Fri Aug 31, 2012 6:22 am

Re: Fresnel lens stack for "supernatural" FoV

Post by TheLookingGlass »

Wow! Thanks for pulling all of this into a new thread! I had stumbled across of few of your different posts across multiple threads so it'll be nice to come back here to check on your progress. :)
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Fresnel Lens Stack Theory of Operation

Here is a crude drawing (from overhead) showing how the fresnel lens stacks should work with a 7-inch display (Nexus 7 in my case):
fresnel_stack.jpg
In this drawing, the fresnel lens stacks are red.

The solid blue lines (bent inward by the lenses) are the light rays from the edges of the left SBS image to the eye. The dashed blue lines are the VIRTUAL FoV for the left eye.

The solid green lines (bent inward by the lenses) are the light rays from the edges of the right SBS image to the eye. The dashed green lines are the VIRTUAL FoV for the right eye.

Notice that the Virtual FoV images extend horizontally inward behind the nose. A similar effect lets me see vertically beyond my eyebrow ridge above, and beyond my mustache below.

As mentioned above, this not only does NOT reduce immersion, but in fact, REMOVING the "headset" makes Real Life "less immersive" due to the distracting sudden awareness of natural facial obstructions. With the lenses, you get a "supernatural" (beyond natural) FoV.

I would estimate that the total FoV (from the outer dotted blue line to the outer dotted green line) is about 150-degrees horizontally. Likewise, vertically I think it is about 130-degrees (about 10-degrees beyond my 120-degree "natural" vertical FoV). The key to achieving this "supernatural" FoV is getting the lenses extremely close to the eye, so that it touches the side of the nose, the underside of the eyebrow ridge, and the cheek. You get used to the eyelashes touching them, but mascara-extended eyelashes could get messy.
:o
You do not have the required permissions to view the files attached to this post.
Last edited by geekmaster on Sun Feb 10, 2013 12:34 am, edited 5 times in total.
German
Binocular Vision CONFIRMED!
Posts: 209
Joined: Tue Nov 23, 2010 5:18 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by German »

I thought you were going to post pictures of what you made? Is that still happening?
User avatar
twofoe
Cross Eyed!
Posts: 164
Joined: Mon Dec 17, 2012 7:16 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by twofoe »

I do not wish to post my CRUDE experimental photos just yet. They proved the concept to me, but after less than satisfactory results after cutting the second lens stack, I want to take more careful measurements so that my experiments are easily reproduceable.
Have you PMed Palmer with the pics, at least, since he asked for them?
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

German wrote:I thought you were going to post pictures of what you made? Is that still happening?
twofoe wrote:Have you PMed Palmer with the pics, at least, since he asked for them?
I took photos while making another lens stack, but when trimming them to better hug the edge of my nose, I moved them inward, which reduced the outer edges and sacrificed some of the peripheral vision.

I need to find my original lens stack and photograph that...

Also, I need to make a matching mirror-image lens stack for my other eye, and mount them in some sort of frame.

I have been extremely busy, and what little free time I had was spent READING these forums instead of continuing my fresnel lens stack experiment.

I reorganized these posts into a new thread, and added a "theory of operation" diagram in my previous post above. Sorry for any delays, but I am EXTREMELY busy lately, but I will try to post photos RSN. You "should" be able to duplicate my results with information previously provide in this thread. Just trim the inner corners from the lenses to fit you face as described...
STRZ
Certif-Eyed!
Posts: 559
Joined: Mon Dec 05, 2011 3:02 am
Location: Geekenhausen

Re: Fresnel lens stack for "supernatural" FoV

Post by STRZ »

Image
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

I was not able to find my trimmed lens stacks. I remember only where the WERE before organizing my office. :( However, I did find my new unopened lenses, so I will experiment with those now...

Okay, I took some photos of my Nexus 7 showing cybereality's "DIY Stereo 3D Drive (Test 8)" youtube video (paused).

First, a pair of uncut fresnel magnifier lenses, stacked as shown in the following photo. Notice that the ridge (rough) side is up (toward the eye). The overlap is offset downward only for this photo, so you can see the ink markings on the reverse side of both of them:
fresnel_pair.jpg
Next, a photo with the camera a fixed distance from the right image of the SBS pair on my Nexus 7 display. To achieve reasonable focus with the same FoV as when using the lenses, I had to zoom in until the image filled the viewfinder in the vertical dimension:
no_lens.jpg
Now, a photo with the camera in the same position, but with the outer edges of the lenses stacked with 5 cm overlap (camera zoomed out to compensate for vertical magnification). Notice that the vertical FoV is the same as the previous photo with no lens stack. The horizontal FoV is stretched on the left and right edges to fill the camera FoV. For naked-eye viewing, I add another lens center portion to the stack to stretch the image in all directions for increased FoV of my eye beyond what my camera lens can focus. [For some reason, the editor here will not let me inline this third image, so no text can follow it]:
You do not have the required permissions to view the files attached to this post.
Last edited by geekmaster on Sat Feb 09, 2013 11:58 pm, edited 4 times in total.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

In the photos in the previous post, it is the eccentric (outer left and right) portions of the magnifier lenses that achieve more offset magnification in the horizontal direction. One lens stretches the left edge of the image, while the other lens stretches the right edge. These distortions can be compensated in software, similarly to how the Rift pre-warping is done. In fact, existing Rift pre-warp seems to work reasonably well here, as can be seen by how the text on the right edge is stretched back to "normal" by the lens distortion.

When viewing with the naked eye, another magnifier lens (center) is added to the the lens stack, to fill the complete FoV vertically (and all but the very outer edges horizontally). This was left out for the photos, because with it my camera would only focus with too much zoom to be useful for this demonstration.

Also, the inner edge of the lenses need their corners trimmed off to get the lenses positioned close enough (touching) the nose, while held very close to the eyes.

In the "heat of battle", the lack of focus and chromatic aberration (mostly blue) in the peripheral vision areas is hardly noticed (unless specifically looking for it). The chromatic aberration should be correctable in software pre-warping if desired.

Remember that the above images are intended to fill your entire FoV even when rolling your eyes in their sockets. In fact, it actually hurts to rotate my eye enough to see the text on the right side. That pushes the distortion out where it is not so noticeable as in the small FoV of the photos as viewed in a typical web browser.

Also, remember that using fresnel lens stacks is just an expedient experiment. Although useful for an extremely inexpensive "Fov2go"-style add-on for a 7-inch tablet computer, custom acrylic lenses with these same optical properties would (probably) supply a superior-quality image for a consumer-grade "supernatural FoV" HMD peripheral device.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

WiredEarp wrote:
My experiments show that stretching the image beyond natural facial occlusions (giving a "supernatural FoV"), makes returning to the real world LESS immersive. In other worlds, VR with invisible facial occlusions (depending on your head shape) can be MORE immersive than reality.

After viewing video for a few minutes with this lens stack (left 1/3, right 1/3, middle 1/3, all trimmed to fit against lower nose and inner eyebrow ridge), taking it off makes me suddenly and unusually aware of the visibility of my nose, eyebrows, and mustache which I can see back in the "real world", actually giving me LESS immersion for a minute or so in the real world.
Thats really quite interesting geekmaster. I wonder if one day (brain interfaces for example) we might not achieve 360 degree vision, making a return to 'reality' seem very unreal.
This new quote came from another forum thread discussing the same "supernatural" FoV, which had previous relevant posts already quoted above. Please post to this new dedicated thread instead of the others that still contain old (now quoted) information about fresnel lens stacks and "supernatural" FoV. Thanks.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Fresnel Lens Stack Provides Partial Anamorphic Correction for SBS-Half Video

It just occurred to me that because my fresnel lens stacks stretch the images more in the horizontal direction than they do in the vertical direction, they should do a better job of viewing SBS-Half (normal side-by-side half-width stereoscopic) video pairs. Although the vertical dimension of the video may still look a bit "too tall", it will still be an aspect-ratio improvement over fully-symmetrical aspheric lenses, and the increased vertical FoV over the "correct" aspect-ratio may help improve immersion (subject to viewer preferences).
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Custom hybrid lenses:

I just had a new idea to make a custom HYBRID lens, with a solid acrylic lens in the center zone (for foveal vision) and a fresnel outer zone (for lightweight peripheral vision).

This new lightweight hybrid lens design could give us a lens with full skin contact at the nose, eyebrow ridge, and cheek (for "beyond the skin" extended FoV), plus a high-quality central vision where the pre-warp distortion places the highest pixel density.

Here is a similar concept (like earlier fresnel lenses designed by Augustin-Jean Fresnel for lighthouses, except I would use fine-pitched fresnels as in previous posts above in the outer peripheral zone):
Image
I still think fresnels will have their place in future HMD designs, especially in the peripheral areas of the FoV. Now we just need to continue our experiments in this direction.

History of this idea:

My plans when I first started experiment with stacks of fresnel lenses were twofold:
1) Make a simple and inexpensive "Fov2go" type of accessory available from local "dollar store" parts.
2) Use fresnel lenses to MODEL a higher quality CUSTOM solid acrylic lens that covers the entire FoV of the user.

After some experiments and thought about custom lenses, I realized that a solid acrylic lens that covers the entire FoV would be large and THICK.

Perhaps would could instead design a custom HYBRID lens, that has a solid inner zone for increased central optical quality while keeping the overall size and weight down, with a fresnel outer zone (at the edge and beyond what the fovea can reach) giving huge peripheral vision while keeping size and weight low. It would be nice to keep the high-resolution center zone free of "fresnel distortion".

There are previous discussions here regarding using a CNC to make custom lenses, but hybrid lenses such as described above are a new consideration. Until them, I may try cutting holes in a fresnell lens stack and mounting solid aspheric lenses in those holes. It could make software pre-warp a bit more complicated, but perhaps worth it to get the benefits of both high-quality acrylic aspheric lenses AND the "supernatural FoV" of large fresnel lens stacks.
Leahy
Cross Eyed!
Posts: 181
Joined: Thu Jun 07, 2012 1:36 pm
Location: France

Re: Fresnel lens stack for "supernatural" FoV

Post by Leahy »

I am not so well versed in optics to be sure of my terminology so feel free to correct me. Seeing as you are using a stack of three what if three types of fresnels were used, the same pitch but with fewer facets. It's probably better to illustrate what I am trying to say. You might lose a bit of FOV but you should regain the loss by better detail in the center.
fresnel stack.jpg
You do not have the required permissions to view the files attached to this post.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Leahy wrote:I am not so well versed in optics to be sure of my terminology so feel free to correct me. Seeing as you are using a stack of three what if three types of fresnels were used, the same pitch but with fewer facets. It's probably better to illustrate what I am trying to say. You might lose a bit of FOV but you should regain the loss by better detail in the center.
fresnel stack.jpg
As I described in my previous post, you would gain FoV in the peripheral zone because my proposal places high resolution ecccentric (offset) fresnel lens material touching the skin at the side of the nose and the cheek below and the eyebrow ridge above. You cannot get a wider FoV than that (except at the side of the head).

An acrylic aspheric lens cannot give any FoV beyond its edges (which may partialially be occluded by the lens frame).

The point being that a fresnel lens stack gives extreme FoV (seeing VR through facial obstructions) and an acrylic aspheric lens gives high quality foveal (central) vision, so we should combine both to get the best of both.
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Fresnel lens stack for "supernatural" FoV

Post by PasticheDonkey »

i think eye position would have to be too exact in that case to be workable. you'd end up with the boundary between fresnel and aspheric creating huge problems where you see aspheric corrected through fresnel or visa versa.
German
Binocular Vision CONFIRMED!
Posts: 209
Joined: Tue Nov 23, 2010 5:18 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by German »

PasticheDonkey wrote:i think eye position would have to be too exact in that case to be workable. you'd end up with the boundary between fresnel and aspheric creating huge problems where you see aspheric corrected through fresnel or visa versa.
A dual lens setup would probably have this issue, yeah because there's a gap between the two types of lenses. However, an actual single fresnel/aspheric lens like geekmaster first showed shouldn't suffer from that. You wouldn't need stacking either because the properly designed lens would have the correct magnification to begin with.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

As you can see in my earlier post that showed a photo taken through a stack consisting of the left and right sides of 6-inch fresnel lenses, it functions primarily as an anamorphic lens, similar to what is used in CinemaScope movie projectors to expand 35mm film out to fit the wide projection screen. Here is what a circle looked like on the film before anamorphic expansion.
http://www.poc.com/media/10416/hmd_datasheet_r2.pdf
Image

Comparing the photos I took with and without that 2-lens stack shows a similar (but nonlinear) anamorphic expansion (along with some aspheric magnification). When I add a fresnel center section to the stack, it gives increased aspheric magnification to fill my FoV.

Image
Image

The problem with cutting lens stack elements from a flat magnifier is that we do not have complete total coverage at the outer peripheral edges, but if designing a custom lens, we could solve that by starting with a meniscus lens (similar to eyeglass lenses), giving it enough curvature to touch the skin beside the eyes even at the outer edges. To get this with the properties we need, it would have to be thin and lightweight, requiring that all but the center foveal zone be collapsed into a curved fresnel lens.

As a thought experiment, think of cutting a section of a (head sized) sphere down to fit around your eye, touching the skin on all sides. Then make the center foveal zone into an aspheric lens like we need, and make the rest of the spherical surface into an fresnel lens with properties of my experimental fresnel lens stack. Look at my previous "theory of operation" diagram for an example of the path the light rays need to follow, but make the lens stack spherical instead of flat. Here is the previous diagram, with the right lens changed as described in this post to give an even larger FoV (dashed green lines):
fresnel_curved.jpg
Notice the thick portion of the curved right lens, which in this diagram represents a solid (without fresnel grooves) portion of the hybrid lens.

I think that if we can accomplish this, having EVERYTHING THE EYE CAN SEE being all in VR space would be as good as you can get (even better than natural FoV).
You do not have the required permissions to view the files attached to this post.
Last edited by geekmaster on Thu Aug 15, 2013 9:51 am, edited 5 times in total.
Mystify
Certif-Eyed!
Posts: 645
Joined: Fri Jan 11, 2013 5:10 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by Mystify »

I like the hybrid lens concept. That seems like it has potential.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

I see that the fpvlabs guys have discovered this thread:
http://fpvlab.com/forums/showthread.php ... post208934

And people are now using these "dollar store" lenses in other FPV HMD builds too:
http://fpvlab.com/forums/showthread.php ... post212826

Some of my ideas that I posted here seem to be gaining traction in the FPV community. It is good to see it evolving into something useful, even if not here...

I wish I had more time available to experiment with this. No time soon that I can see though...
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

For all these new "software pre-warp" HMD designs, we must give credit where due, to Palmer Luckey, for the idea that it is time to replace complex and expensive "post-warp" lenses with "pre-warp" software. And we must give credit to his Kickstarter campaign for getting all these Dev Kits into our hands (soon?), and we must credit John Carmack for endorsing the Rift so that it could become so popular.

That said, I think it is a good idea for a variety of experimental enhancements to this software pre-warp concept to take hold, allowing an evolution of design concepts to flourish so that the CONSUMER Rift can become a much more awesome device. To allow this to happen, we must remember to license our experimental versions with an open license that can be used in a commercial product. If our ideas end up getting incorporated into the final Consumer Rift, it would be nice to get some recognition and/or financial compensation (but with the Creative Commons license I normally use, such recognition and compensation or optional, but nice to have anyway).

I have been thinking about my "Supernatural FoV" HMD design more recently, and I just had some fresh ideas for it. Unlike the recent awesome DIY "InfinitEye" dual-display design that also uses fresnel lens stacks (http://www.mtbs3d.com/phpBB/viewtopic.php?f=120&t=16434), I am using a SINGLE 7-inch display to keep the costs down. Additionally, my first iteration will be designed to attach to a 7-inch tablet computer such as my Nexus 7, making it an evolutionary step beyond Palmer Luckey's Fov2Go kit for SmartPhones (http://www.mtbs3d.com/phpBB/viewtopic.php?f=120&t=14658).

An additional benefit of using a tablet computer instead of a bare LCD panel is that my tablet has a CAMERA! It may be possible do eye-tracking (for one eye). Software should then be able to extrapolate the gaze position of the other eye using a depth map for the stereo image pair.

The KEY to getting huge FoV from a single 7-inch display is to get it as close to the eyes as possible so it covers as much FoV as possible, and then to use a semi-custom fresnel lens stack to expand the image outward to the peripheral limits of visibility as much as possible. With the lens stacks not using frames around them, and actually touching the skin at the side of the nose, cheeck, and eyebrow ridge, the viewed images are actually bent "around" these facial occlusions to give a FoV greater than natural Real Life FoV.

One problem that showed up during testing was that my Nexus 7 touchscreen resizes the image (software problem) when my nose touches the bottom of the screen (in landscape mode), but the display actually needs to be that close for my eyes to focus on it and have maximum FoV. I just realized hours ago that I do NOT need to use the same pre-warp distortion as the Rift, so I can get the display EVEN CLOSER to my eyes for even LARGE FoV (especially at the outer sides of my peripheral FoV), in which case I need to adjust focus by sliding the bottom of my display diagonally up and down along the bridge of my nose instead of in and out perpendicular to my eyes. This will require that my pre-warp software shift the image horizon up and down to match the vertical position of the display, as determined by the shape of the bridge of my nose.

Because I will be able to define my own pre-warp parameters, I can stretch the image downward with fresnel optics to compensate for the display bottom rising vertically to get it closer to my eyes. I just need to change the third lens in the stack to use an outer portion of the 6-inch fresnel instead of the center portion. The reason why this will work is that the outer portions of the fresnel magnifier lenses stretch and shift the viewed virtual image outward away from the lens center proportional to their distance from lens center. While the lens center expands equally in all directions, a lens trimmed from outer portions of the 6-inch lens stretches much more in a direction corresponding to their offset distance of that lens portion relative to its position on the original 6-inch lens. I can use one outer lens portion to stretch the virtual image inward so my eye sees only its own half of the stereoscopic image pair even when looking "beyond" the nose. I can use another to stretch the outer edge of the image far into my outer peripheral vision (way beyond the side of the 7-inch display), and for a display shifted vertically to sit on the bridge of the nose, I may be able to use a third outer fresnel lens portion rotated 90-degrees to stretch the virtual image downward to (or beyond) my cheek (or the visible part of my mustache in my case).

Then I will need to define displacement maps that contain my pre-warp compensation. I can use separate displacement maps for each color plane (R,G,B) to compensate for the chromatic aberration that is greater on the outer portions of these fresnel lenses.

For now, this new vertical offset idea to get the display even closer to my eyes (without cutting a "nose notch" out of my tablet computer) is untested. But it will be tested soon. I also have some ideas for how to create a housing to hold the display and lenses to my face while also blocking residual light from leaking in, while providing for focus adjustment similar to how the new 7-inch Rift Dev Kits do it (telescoping inner and outer HMD shell portions). I plan to start by cutting out a pair of telescoping HMD shell parts from a cardboard box, and then trimming them to fit. I will use adhesive-backed foam weather stripping now available at my local Dollar Tree store, to pad the part of the display that touches my face. The lenses will not use padding or frames because that would obstruct part of the view. Better that they actually touch the skin around their trimmed edges.

These will be interesting and fun experiments, ultimately creating an interesting DIY "Supernatural FoV" HMD device on the cheap. I will use open source CC-BY-SA (Creative Commons) license, of course (http://en.wikipedia.org/wiki/CC-BY-SA).

EDIT: Corrected numerous typographical errors. Nobody warned me that typing accuracy decreases with age. Grr...
Last edited by geekmaster on Mon Feb 25, 2013 2:22 pm, edited 9 times in total.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

One downside to my design is that stretching a 640x800 image to cover a lot more FoV than the Rift uses is that there may not be enough central pixel density for some games (especially flying or driving simulators). Even FPS games may be a problem with distant objects being very limited in available pixels. How can you tell friend from foe in multi-player mode? Perhaps multi-player games will need some sort of HUD assistance (such as a "team color" halo surrounding onscreen players).

Until we get MUCH higher resolution displays, we will certainly need content that is actually designed to take advantage of the increased FoV, while compensating for the reduction in available pixels for the central viewing area.

It may be that the lesser FoV of the Oculus Rift Dev Kits is a better compromise, giving a larger quantity of smaller pixels in the central viewing area at the expense of some FoV. But my design may be better for just strolling around and enjoying the virtual scenery, where missing fine detail is not a matter of virtual life or death...

Different designs for different purposes. After all, there is a reason we still have both acoustic and electric guitars (and hybrids containing features of both). We can do the same with our Rift-like HMD designs.
:D
EDIT: More typo fixes. It takes way too many edit passes before I stop finding more annoying little errors. I hope you guys can understand me when I first post these. I really cannot see these errors until much later, no matter how many times I proofread them. Sorry about that. The younger me was a much more fast and accurate typist. Oh well, such is life. I try not to let it slow me down... Please PM me with obvious typo fixes if you find them in my posts. I want them to be correct where possible. Thanks.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

I played with moving the image closer than my nose touching the bottom of the display, by having it touch the bridge of my nose. Not good. Even with a side of the 6-inch fresnel lens pulling the bottom of the image down, it was not enough. Nose touching the display is as close as it gets unless you go with a smaller display. The fresnel lens stacks DO focus at that distance for me (I am near sighted, focusing between about 10 and 15 inches away).

At least that test simplified how I need to construct the focusing mechanism for my display. I will first construct and test fit it from cardboard, then purchase some foam core sheets from the dollar store for the final prototype.

The final HMD device will provide an estimated 150-degree Horizontal FoV. I really need to measure the FoV somehow after I construct it.

I have some ideas about software rendering that may be a novel approach. The latency problem with head movement until pixels hitting the eye can be bypassed with some perceptual tricks using the power of my Nexus 7 as the display. If the host PC sends an image with MORE horizontal and vertical FoV than you plan to display, quick head movements can do local pan/tilt/zoom on the current image, which will be replaced (or morphed) with a newly rendered image as soon as it becomes available. I believe that substituting an appropriate PTZ portion of the most recent image will substitute nicely for a missing NEW image, to avoid motion sickness from latency issues. Of course, the rendered image must be larger than the displayed image, to provide sufficient peripheral margin to pan into while waiting for the next rendered image. Depending on the application, this might even work okay for high-latency network connections.

The idea for this came from videos I have seen where 4 FPS video had intermediate frames inserted that using morphing instead of motion interpolation. It was a interesting "dream-like" effect, which I think would be better than motion lag.

I will find out more how effective this will be when I get around to testing it. Even though I just "invented" this idea recently, for all I know somebody else may have invented it before me. It has happened before and it will happen again. I just hope some patent troll will not prevent me from using my idea...

In summary, I think that simple head-tracked pan/tilt/zoom can compensate for high-latency rendering to prevent motion sickness, and I want to test this with my "Fresnel Lens Stack Fov2Go Clone for 7-inch Tablets"...
budda
Sharp Eyed Eagle!
Posts: 388
Joined: Thu May 22, 2008 9:13 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by budda »

Hi,

I have been wondering if a 'field flattener' may be useful for a wide angle head mounted display.

This 'field flattener' is a negative power lens which is used to lower the Petval Sum of the lens system and so reduce the field curvature of an image.

So what you say?

Well, if the optics of a head mounted display were operated backwards as a landscape camera, the landscape being far away, would be imaged on a curved surface - the Petzval Surface - if astigmatism was totally absent.

Now, the display screen in a head mounted display is flat and does not follow the Petzval Surface. So the screen will project different depths of field across the field of view. This may be noticeable to the eye and not desired.

So what can be done about it?

The idea of the field flattener is to make the primary lens 'see' the display screen as being suitably curved for projecting a uniformly deep virtual image.

In camera optics, the field flattener takes the form of a plano-concave lens which lies flat near the sensor plane - or next to the display screen in the case of a classical head mounted display system.

But since space, weight, cost and practicality are foremost for makers, there may be another way to achieve a solution.

A fresnel lens of suitable negative power can be placed on the display screen to act as the field flattener.

This negative power field flattener at the screen can be combined with the primary positive lens or fresnel stack, as the case may be.

The beauty of the field flattener is that is doesn't reduce the optical power of the primary lens, but it can reduce the Petzval Sum as a design variable.

Thanks.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

A field flattener may be important when we get much higher resolution displays. But with the low resolution of these 1280x800 displays, after they are pre-warped to a huge FoV, the outer edges are distorted enough that increasing the focus may not be important (at this time). And in fact, a little blur actually improves the image by reducing the screendoor effect. In fact, older books about DIY VR recommended using a diffusion layer (such as waxed paper) over the display surface in an HMD in order to blur the image to reduce both screendoor and blocky objects (allowing the eye to better interpret what it is seeing). We are better at understanding blurry content than we are at blocky content.

When the pixels get too small to discern, then all the focus we can get will be important, and a field flattener may be a useful tool to achieve that. Thanks for the suggestion.
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Fresnel lens stack for "supernatural" FoV

Post by PasticheDonkey »

geekmaster wrote:One downside to my design is that stretching a 640x800 image to cover a lot more FoV than the Rift uses is that there may not be enough central pixel density for some games (especially flying or driving simulators). Even FPS games may be a problem with distant objects being very limited in available pixels. How can you tell friend from foe in multi-player mode? Perhaps multi-player games will need some sort of HUD assistance (such as a "team color" halo surrounding onscreen players).
that's only a worry in the now tho. so get wide FoV working as best you can to be ready for the future.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

To summarize the Theory of Operation of the HMD proposed for this thread:

1) Use the left 1/3 of a 6-inch fresnel magnifier to stretch the left edge of the image outward to the left.

2) Use the right 1/3 of a 6-inch fresnel magnifier to stretch the right edge of the image outward to the right.

3) Stack these two fresnel lenses and trim the top and bottom corners to fit along the side of the nose and slightly under inner eyebrow ridge. This stretches the image horizontally, proportional to the distance from the lens centers.

4) Add additional "aspheric" magnification to stretch the top and bottom of the image to the upper and lower limits of FoV.

The result of these steps (when the lens overlap is adjusted optimally) is that the onscreen image fills the inner FoV with none of the opposing image visible to the wrong eye. It is also stretched outward into the peripheral vision well beyond the outer edge of the display surface. And it is fills the vertical FoV as well.

For step #4 above, I just tried substituting a 3x "aviator-style" pair of reading glasses to provide magnification required for full vertical FoV. I taped the trimmed left and right fresnel stack to the INSIDE of the eyeglass frames, between the eyeglass lens and eye. It worked quite well. The high quality reading glass (polycarbonate) lenses gave better image quality than a third fresnel layer, as expected.


I hope to get away from fresnels if I can. By having the user just put on a pair of ordinary (strong) reading glasses, the display can be mounted to just "swing down" in front of the glasses, and the glasses will be attached to the head instead of the display, allowing the display to be moved in or out to adjust the focus. To change the FoV, just put on a different pair of reading glasses. A "ready to assemble" kit could be shipped with a set of reading glasses for multiple magnifications, so that users may use whichever work best for them.

Now, for my next experiment, I just ordered a couple of pairs of "full-lens" 5x and 6x reading glasses from here:
http://stores.ebay.com/CAROBALL/STRONG- ... ub=3449935
I already know that using just the 3x "full-lens" glasses from the dollar store work well, when my Nexus 7 tablet is held about 5-inches from my eyes. These new higher magnification reading glasses will let me hold it much closer (perhaps even too close). I am anxious to try them.

Using ONLY glasses (no fresnel lenses) may require a blinder (divider) between the pair of images though, unless the asymetric stretching using the offset (outer) portions of the large fresnell lenses. I suspect that with sufficient magnification, no blinder will be needed (at the expense of more unused pixels on the display). Unused pixels are not a problem if you have plenty to spare, such as on future higher resolution displays.

Because offset lenses essentially have a built-in prism effect, the help reduce the need for a divider between images by shifting the view outward (and/or outward). It would be ideal to get avaitor-style glasses with anamorphic lenses that preferentially stretch the image half in the horizontal direction, similar to what I am doing with the offset fresnel lens stacks. Even better if the anamorphic lenses stretch the inner and outer portions the most, preserving more pixels in the central foveal view (just like my fresnel lens stacks).

I really like the idea of using ordinary eyeglass frames to hold the lenses close against the face, rather than attaching them to the display surface. Much simpler mounting mechanics (and easy to obtain off-the-shelf components) are always a good thing.

Anyway, my experiments have proved to me that this can be a very simple and lightweight HMD with extremely high FoV, using a single 7-inch (or smaller with higher magnification) display. This is currently at the expense of pixel density, but that will correct itself as higher pixel density displays become affordable for a consumer device (or hobbiest kit).

If I were to order a boatload of reading glassses and pallets of foam board, I suppose I could start a Kickstarter campaign to fund mass-production of these kits. And supplying a display device (instead of using your own 7-inch tablet) could be an option too.

I described elsewhere how I think we can get away with lower framerate VR rendering for tablet computers (or Raspberry Pi?). To keep the head-tracking latency low for small shaking or jittering motions (or even small head turns), we should be able to use PTZ (pan/tilt/zoom) to move around in the static image while waiting for the next "animated" image to arrive. I strongly believe that viewing low-latency head-tracked stereoscopic 3D (perhaps even a full spherical projection) will be effective at providing useful immersion even when the scenery update speed is limited. If necessary, we can use a faster IMU than what the tablet PC uses, perhaps with bluetooth or near-field communications.

Any thoughts on these ideas, guys?
Last edited by geekmaster on Tue Feb 26, 2013 1:33 pm, edited 1 time in total.
PalmerTech
Golden Eyed Wiseman! (or woman!)
Posts: 1644
Joined: Fri Aug 21, 2009 9:06 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by PalmerTech »

geekmaster wrote:A field flattener may be important when we get much higher resolution displays. But with the low resolution of these 1280x800 displays, after they are pre-warped to a huge FoV, the outer edges are distorted enough that increasing the focus may not be important (at this time). And in fact, a little blur actually improves the image by reducing the screendoor effect. In fact, older books about DIY VR recommended using a diffusion layer (such as waxed paper) over the display surface in an HMD in order to blur the image to reduce both screendoor and blocky objects (allowing the eye to better interpret what it is seeing). We are better at understanding blurry content than we are at blocky content.

When the pixels get too small to discern, then all the focus we can get will be important, and a field flattener may be a useful tool to achieve that. Thanks for the suggestion.
This is true, but only to a certain degree. Using a diffuser to blur the entire screen was indeed common, and so was slightly defocusing the entire image. That is different than having a curved field.

If you don't have a flat field, it is pretty easy to cause eyestrain. Your eyes will constantly be shifting focus as they look at different parts of the image, something that is more apparent in certain content, closed indoor spaces for example.

Not a dealbreaker, but something to be aware of!
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

PalmerTech wrote:
geekmaster wrote:A field flattener may be important when we get much higher resolution displays. But with the low resolution of these 1280x800 displays, after they are pre-warped to a huge FoV, the outer edges are distorted enough that increasing the focus may not be important (at this time). And in fact, a little blur actually improves the image by reducing the screendoor effect. In fact, older books about DIY VR recommended using a diffusion layer (such as waxed paper) over the display surface in an HMD in order to blur the image to reduce both screendoor and blocky objects (allowing the eye to better interpret what it is seeing). We are better at understanding blurry content than we are at blocky content.

When the pixels get too small to discern, then all the focus we can get will be important, and a field flattener may be a useful tool to achieve that. Thanks for the suggestion.
This is true, but only to a certain degree. Using a diffuser to blur the entire screen was indeed common, and so was slightly defocusing the entire image. That is different than having a curved field.

If you don't have a flat field, it is pretty easy to cause eyestrain. Your eyes will constantly be shifting focus as they look at different parts of the image, something that is more apparent in certain content, closed indoor spaces for example.

Not a dealbreaker, but something to be aware of!
In my experiments described in this thread, my personal preference is for extended FoV at the expense of sharp focus at the outer edges. Even blurry content out in the periphery (especially where it is painful to turn your eyes that far to get those pixels into the foveal view) strongly increases immersion (at least for me).

That is why having multiple designs to play with is important. We can experiment with all of them and figure out which optical parameters are subjectively important for a larger intended audience. Then we can try to incorporate the best features of all them into an evolutionary new device. It is important to improve not only FoV, resolution, and low-latency, but also simplicity, reliability, and low manufacturing cost. All of the ideas we are playing with here are a good thing for the evolution of our VR devices and algorithms.

In my case, to keep costs down for now, I want to base my design as a tablet accessory, in the tradition of the Fov2Go. We can add a dedicated LCD or OLED panel later. :D

And regarding a "flat field", my experiments show (to me) that as long as the fresnel lenses are parallel to the display surface, the entire surface remains in focus even when the lenses close enough to be brushed by my eyelashes. It may be mostly my own personal subjective judgement, but the focus is good enough for me, and the FoV is superior to the Rift Dev Kit design (not having had the pleasure to actually EXPERIENCE a Rift demo yet).
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

I now have two sets of fresnel lens stacks (one pair of left and right magnifier ends for my left eye, and another pair for my right eye). Instead of using the center section from the two magnifiers, I instead used a pair of "aviator-style" (full-lens) 3x magnified reading eyeglasses.

I first tried taping the fresnel lens stacks to the inside of the eyeglasses lenses. But because the eyeglasses curved around the face and hugged it closely, the fresnel lenses were not parallel to the LCD screen. This caused the prewarp and focus to be uneven (as Palmer Luckey suggested could be a problem).

Next, I pulled the fresnel stacks loose from the eyeglass lenses, leaving them attached only at the outer edges. Then I taped the inside corners of the trimmed fresnel lenses together with a folded-over bridge of tape. This made the fresnel lens stacks parallel to the LCD screen. When putting this arrangement on my face, the tape bridge of the fresnel lenses touched the top of my nose between my eyebrows, and the bridge of the eyeglasses stayed about 1/2-inch from the fresnel lens inner edges, but the fresnel tape bridge rested comfortably on my nose. Although the front of the eyeglasses were being held away from my face a bit, they fit securely enough and the fresnel lens stacks stayed parallel to the LCD screen while I was wearing this "eyeglass and fresnel lens stacks and tape" contraption.

While wearing these, I tried viewing the wide-eye view of a scene from Quake, posted as an attachment in the "Quake for Rift-alike HMDs" thread (first page), using my Nexus 7 tablet. When viewed full-screen through these glasses, the image separation was too much for me to converge the stereo pair, even though I can converge them with my naked eyes.

Next, I viewed the smaller thumbnail version embedded in the post (not full-screen). I did the tablet pinch-and-stretch to expand the size to the maximum that I could comfortably converge into a 3D view. It worked very well, but it was clear that I need to use a non-overlapped pre-warp (like the 7-inch Rift Dev Kits need) so that I can converge these image pairs without shrinking them to get their centers closer together, while still filling the FoV with image pixels.

One good thing that this experiment proved to me is that I could see a slight screendoor effect on ALL visible pixels, as long as my fresnel lenses were parallel to the screen. This proved that they were properly focused over the full FoV.

Because my fresnel lens stacks are functioning primarily as anamorphic lenses, it should be possible to get special reading glasses with magnifying anamorphic built into them, duplicating my experimental setup without reduction of image quality from fresnels. However, because I can see the screendoor effect (when I look for it) all over the screen, the fresnel lenses may actually work well enough by themselves. It is the anamorphic property of my lens stacks that gives a wider FoV while eliminating the need to mask the opposing image with a divider or blinder.

One thing left to fix is the problem with the eyeglass lenses being curved around the face instead of perpendicular to the face, which causes them to also curve away from the fresnel lens stacks, making a cumbersome arrangement. What is really needed here are eyeglasses that have flat lenses set into a flat frame, similar to 3D glasses used at movie theaters, and for passive 3D TVs. I still want the "oversized" aviator-style lenses though, for maximum FoV.

Using this setup still leaves some area at the outer limits of peripheral FoV not used for VR content, but other than that outer edge, I can see pixels everywhere else, including where my nose should be.

In VR, I have an invisible nose. :lol:

Of course, software COULD draw in my missing nose, but why? I get used to it quickly, and leaving VR makes my REAL nose annoying for awhile as it restricts my inner FoV.

This experience was cool, but it will be MUCH better with head tracking. In fact, I want to try the Fov2Go demo (which does support head tracking) with these glasses, except I will not be able to converge the image pairs unless I can shrink the images (simulating a 5-inch display) or make them partially non-overlapping, so they work on my 7-inch Nexus 7 tablet display.

Conclusion: Successful experiment, but more work to do (software pre-warp for 7-inch display, and some what to mechanically hold it all together and attach it to my head).

FYI, "anamorphic" means "asymetrical magnification", which in this case magnifies more in the horizontal direction than it does in the vertical direction.

If you think of the shape of an eyeglass lens, it is wider than it is tall (i.e. landscape mode). Half of an SBS-Half (side-by-side half-width) stereo image pair on a 1280x800 screen is 640 wide by 800 tall, which is taller than it is wide (i.e. portrait mode). What the anamorphic lenses do is stretch the "portrait mode" image into a "landscape mode" image, more closely fitting the shape of the eyeglass lens (and corresponding FoV for one eye). This means that to effectively map a widescreen LCD panel to a pair of eyes REQUIRES the use of anamorphic lenses, which my experiments have proven here. This all seems so obvious in retrospect, but it could not have been completely obvious, or it would have become the defacto standard for HMDs long before now. This may change soon...

Is there "prior art" for anamorphic HMDs? A quick search found that there is (http://www.poc.com/media/10416/hmd_datasheet_r2.pdf) but it is only 39.5-degree diagonal. At least that validates my research here.

Perhaps the future consumer Rift needs anamorphic lenses too, to increase the FoV width and eliminate the need for a center divider or restricted optics.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Here are some photos of the lens stacks taped to my 3x reading glasses. While on the face, the fresnel lens stacks rest against my nose and cheeks, and sit parallel to the LCD screen that I hold in front of my face, just far enough away to fill the FoV covered by the lenses (all but the extreme outer edges). The eyeglass front sits out in front of my nose (by the distance shown between the center of the eyeglasses and the center of the fresnel lens stacks), but is well supported by the tape holding the lenses together, which rests on the bridge of my nose (just between my eyebrows). It worked well enough to perform this experiment, but is a VERY temporary arrangement. When the diplay is in focus, it is close enough to almost touch the tip of my nose.
ls3.jpg
ls1.jpg
ls2.jpg
These eyeglasses with attached inner fresnel lens stacks would fit together much better if the eyeglasses had a flat front instead of a curved (face hugging) shape, as shown in the above photos.

To get more FoV than this, we would need wider lenses. To get even wider FoV (such as full 180-degree FoV), we would need two diplays, as shown in another DIY HMD thread.

FYI, "FoV" mean "Field of View" (how much viewing angle the eyes see, measured in degrees).
You do not have the required permissions to view the files attached to this post.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

To restate my important independent discovery here (confirmed as viable by experimental evidence, and by a prior-art search):
geekmaster wrote:A Rift-style HMD *MUST* use anamorphic lenses to properly map the SBS-Half (side-by-side half-width) portrait-mode images onto the landscape-mode shape of "aviator-style" (large-lens) eyeglass lenses, which give the maximum available FoV (Field-of-View). This is the ONLY way to effectively and efficiently map pixels from a widescreen display onto the full natural FoV of human eyes, or even the somewhat horizontally-restricted FoV of large-lens eyeglasses. Simple aspheric lense are simply not good enough for this application.
You can quote geekmaster on that.

For more information about this significant discovery and how this conclusion was reached, you can read this entire thread, starting from the first post:
http://www.mtbs3d.com/phpBB/viewtopic.php?f=140&t=16373
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

3D TVs do anamorphic stretching electronically, to make one image from a 3D SBS-Half image pair twice as wide, to fit the entire widescreen display (for either active or passive viewing).

Rift-style (fully immersive) HMDs also need to stretch one image from a 3D SBS-Half image pair to MORE than twice as wide, to fit as much of the horizontal FoV as it can. However, an HMD must use optical lenses to do this anamorphic stretching, preferably in a fashion that dedicates more pixel density to the central viewing area. This can be done with "aspheric anamorphic" lenses.

You can read more about the history of anamorphic lenses here:
http://en.wikipedia.org/wiki/Anamorphic_format
The word "anamorphic" and its derivatives stem from the Greek words meaning formed again. ... The process of anamorphosing optics was developed by Henri Chrétien during World War I to provide a wide angle viewer for military tanks. The optical process was called Hypergonar by Chrétien and was capable of showing a field of view of 180 degrees. After the war, the technology was first used in a cinematic context in the short film Pour Construire un Feu (To Build a Fire) in 1927 by Claude Autant-Lara. ... In the 1920s, phonograph and motion picture pioneer Leon F. Douglass also created special effects and anamorphic widescreen motion picture cameras. ... Anamorphic widescreen was not used again for cinematography until 1952 when Twentieth Century-Fox bought the rights to the technique to create its CinemaScope widescreen technique.[1] CinemaScope was one of many widescreen formats developed in the 1950s to compete with the popularity of television and bring audiences back to the cinemas. ... The development of anamorphic widescreen arose due to a desire for wider aspect ratios. The modern anamorphic widescreen format has an aspect ratio of 2.40:1, meaning the picture width is 2.40 times its height. ... It may seem that it would be easier to simply use a wider film for recording movies; however, 35 mm film was already in widespread use, and it was more economically feasible for film producers and exhibitors to simply attach a special lens to the camera or projector, rather than investing in a new film format, along with the attendant cameras, projectors, editing equipment and so forth.
That last statement compares using the anamorphic process to using a wider display. For our HMD application, this is equivalent to using a software pre-warp filter (or fisheye camera lens AND anamorhic lens) that horizontally COMPRESSES the image more than it does vertically (non-linearly to preserve central pixel density), then using OPTICS (aspheric AND anamorphic lenses) to expand the (portrait mode) image from HALF of the screen to (wider than landscape mode) full vertical FoV AND almost full horizontal FoV.

Returning to the analogy between motion picture formats and HMD content:
Cinerama was an earlier attempt to solve the problem of high-quality widescreen imaging, but anamorphic widescreen eventually proved to be more practicable. Cinerama preceded anamorphic films, but consisted of three projected images side-by-side on the same screen: the images never blended together perfectly at the edges, and it required three projectors; ... Anamorphic widescreen was attractive to studios because of its similar high aspect ratio (Cinerama was 2.59), without the disadvantages of Cinerama's added complexities and costs.
What this shows us is that an attempt was made to provide a higher-quality content with the same wide (or wider) aspect ratio, but this was abandoned because of its increased cost and compexity. This is the exact equivalent to HMDs that use multiple displays to get higher quality with the same (or wider) aspect ratio than anamorphic.

I think that for a unit intended for mass distribution to the general public, a multiple-display HMD will not achieve full market mass because of its comparitively expensive use of dual displays, when compared to anamorphic HMDs.

The anamorphic process is also used for home distribution of motion picture content on DVDs distributed in widescreen (or letterbox) format. A 16:9 image (1.78 aspect ratio) is compressed horizontally to 4:3 (1.33 aspect ratio). To achieve a full 2.40 theater aspect ration, additional letter-boxing (black bars at top and bottom) is applied.

You can read more about DVD anamorphic compression:
http://en.wikipedia.org/wiki/Anamorphic_widescreen
Anamorphic widescreen, when applied to DVD manufacture, is a video process that horizontally squeezes a widescreen image so that it can be stored in a standard 4:3 aspect ratio DVD image frame. Compatible playback equipment can then re-expand the horizontal dimension to show the original widescreen image. ... A DVD labeled as "Widescreen Anamorphic" contains video that has the same frame size in pixels as traditional fullscreen video, but uses wider pixels. The shape of the pixels is called pixel aspect ratio and is encoded in the video stream for a DVD player to correctly identify the proportions of the video. If an anamorphic DVD video is played on standard 4:3 television without adjustment, the image will look horizontally squeezed. ... It doesn't matter whether the filming was done using the anamorphic lens technique: as long as the source footage is intended to be widescreen, the digital anamorphic encoding procedure is appropriate for the DVD release. As a sidenote, if a purely non-widescreen version of the analog-anamorphic Star Wars were to be released on DVD, the only options would be pan-and-scan or hardcoded 4:3 letterboxing (with the black letterboxes actually encoded as part of the DVD data).
Anamorphic compression is also used for digital television broadcasting, to save bandwidth.

Why this affects how Rift-style HMDs SHOULD be designed:

What the current version of the Rift Dev Kit is doing, is essentially displaying the equivalent of (what SHOULD be) movie theater 2.40 aspect ratio (or WIDER) content at NARROWER than even 4:3 (1.33) aspect ratio. Assuming that its current optics are simple aspheric lenses similar to the recommended 5x aspheric loops for DIY Rift clones (but I can only speculate about Rift lenses at this time), then the image seen by the eye is optically stretched non-anamorphically (proportionally the same in all directions), maintaining its aspect ratio of 0.80 (640/800), when for OPTIMUM pixel use it SHOULD be using an aspect ratio of AT LEAST 2.40 as used in movie theaters. This is essentially what my "aspheric anamorphic-like" offset fresnel lens stacks are doing to map an SBS-Half image pair to the (amost) full FoV of both of my eyes.

Just like the (failed) Cinerama approach that used multiple projectors, it *IS* possible to get a higher quality image at the expense of using multiple displays, but only at the added expense of cost and (hardware and software) complexity. The increased cost alone will make in non-competitive to an anamorphic HMD display.

To keep the design of an aspheric anamorphic lens simple and inexpensive, we can use a single-lens design similar to the lenses used by the Rift Dev Kits, but with a specially-shaped lens that also stretches the image much more horizontally than it does vertically while still preserving a higher central pixel density. We can then continue to use software pre-warp filters to compensate for any lens distortion.

My proposed FIX for the "non-anamorphic Rift optics problem":

A version of my proposal can easily be adapted to the current design of the Rift Dev Kits by simple replacing the lenses in the lens cups. But this will be limited by the dimensions of the lens cups and the edges of their frames. It will STILL be an improvement over the current anamorphic lenses, because it will waste less vertical FoV (vertical pixels that are out of view), and it will reduce (or eliminate) the black band on the inside of the horizontal FoV making a divider (or blinder) not needed, and it will also reduce the outside black band. There will still be a somewhat reduced FoV when wearing glasses, but this cannot be helped by simply replacing Rift Dev Kit aspherical optics with aspheric anamorphic optics. The software pre-warp filter may also need to be adapted to anamorphic compression.

The reason that this is so important is that without anamorphic compression, increasing the horizontal FoV will also increase the vertical FoV well beyond what is useful, wasting pixels that could otherwise contribute to important details in the vertical dimension. With anamorphic compression we are doing the equivalent of not just increasing horizontal FoV, but also reducing vertical FoV to only what is needed, diverting otherwise wasted pixels at the top and and bottom to an increased pixel density in the central portion of the display. Why waste pixels when the resolution is already too low for some applications?

Of course, I do not yet have a Rift Dev Kit to study, so my "Rift-FIX" proposal will only be useful if the Rift does not already ship with aspheric anamorphic lenses.

If the Rift Dev Kits still ship with simple aspheric lenses, we can fix that with an "aspheric anamorphic lens" upgrade kit that contains a set of lens cups that contain aspheric anamorphic lenses, along with a software pre-warp filter that performs anamorphic pre-warp compression.

I like the removable lens cups in the Rift, when that is exactly what allows the end-user to easily install a simple anamorphic lens upgrade kit to increase the horizontal FoV and/or to increase the central pixel density by reducing unused vertical FoV (depending on display position and whether eyeglasses are worn with the rift).
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Here is a little experiment to demonstrate the aspect ratio of a single eye:

1) Hold the back of your left hand against your nose.
2) Stretch your thumb down and your little (pinky) finger up, so that they are near the limits of your vertical FoV (Field of View).
3) Now hold your right hand the same way, in front of your right eye.
4) Move your right hand to the right until it is at the right-most limit of what you can see when you rotate your eye to the right.
5) Move your left hand a little to the left so it too is at the left-most edge of the limit of vision of your right eye.
6) Now move both of your hands outward (away from your face), and notice that the corners of the rectangle defined by them approximately matches the movie theater landscape aspect ratio.

That landscape aspect ratio is for one eye. This is why anamorphic lenses are needed, to stretch the right SBS-Half (portrait mode) image to fit the maximum useful FoV of the right eye.

Alternative to using anamorphic compression and anamorphic lenses, we could display our stereoscopic image pair in "over/under" format, which for a 1280x800 display would use a pair of 1280x400 images (a 3.2 aspect ratio), which is twice as wide as a 16:9 widescreen display, and even wider than the 2.4 aspect ratio used by modern motion pictures. It would be ALMOST PERFECT for duplicating the effect of the older 3-projector Cinerama motion pictures though (such as the musical movie "Oklahoma!"). And I think it would be great for an HMD, other than the complicated optics needed to map the top and bottom images to the left and right eyes.

For now, anamorphic optics are the most economically effective way to do what we need for both motion pictures and for Rift-style HMDs. HMDs also benefit from the addition of non-linear aspheric optics, to increase the central pixel density (where it is needed most).

EDIT: Please forgive any typos in my recent posts. Or better yet, please PM me with corrections for them. Thanks! And I would LOVE some feedback about these ideas/experiments/conclusions.
Last edited by geekmaster on Wed Feb 27, 2013 1:06 pm, edited 1 time in total.
User avatar
Diorama
Binocular Vision CONFIRMED!
Posts: 273
Joined: Mon Jan 28, 2013 10:37 am
Location: Brighton, UK (Sometimes London)

Re: Fresnel lens stack for "supernatural" FoV

Post by Diorama »

Interestingly, I think Namielus' "virtual cinema" project would be a fantastic way to experience Cinerama films.

Because of the way they are filmed, you get strange issues with projecting Cinerama onto a flat screen (characters standing on either side of the screen having a conversation do not appear to be facing each other, etc.).

The current workaround solution for this is to simulate the original curved screen using 'Smilebox' presentation:
Image

With the virtual cinema, I could finally watch these films on a real (i.e. virtual!) Cinerama screen!
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

Diorama wrote:... With the virtual cinema, I could finally watch these films on a real (i.e. virtual!) Cinerama screen!
Oklahoma! Here I come!

Oh, and "It's a Mad, Mad, Mad, Mad World" too:
http://en.wikipedia.org/wiki/It's_a_Mad ... _Mad_World

An interesting thing about "Mad Mad World" was that although it was designed to be projected onto a curved Cinerama screen, it was actually filmed with a special extreme anamorphic lens (Ultra Panavision 70) and used only a SINGLE camera.

EDIT: Oh, wait, "Oklahoma!" was not on the list of Cinerama feature films:
http://en.wikipedia.org/wiki/Cinerama#Cinerama_features
But it WAS a "super-anamorphic" format filmed with a single camera, just like "Mad Mad World" which WAS released to Cinerama theaters. Still, "Oklahoma!" would be great to watch in the Virtual Cine(ra)ma app, when it becomes available (and if it allows selectable screen layouts).


The big problem with Cinerama was that the filmstock had color variations, so you could see abrupt color changes at the multple projected image seams. This was bad for immersion, especially when following a character across the screen.

I am waiting (impatiently) to try the Virtual Cinema (or Virtual Cinerama!) VR application too.

@Namielus: When do you plan to add a virtual extra wide and curved Cinerama screen, for watching movies filmed in that format? Now that those theaters are gone, we will need your program to "time travel" back to experience that experience "in person".

Experience things that are gone could be advertised as virtual time travel. Virtual Adventures in Time and Space!
Last edited by geekmaster on Wed Feb 27, 2013 1:15 pm, edited 2 times in total.
User avatar
PatimPatam
Binocular Vision CONFIRMED!
Posts: 214
Joined: Thu Jun 28, 2012 1:31 pm
Location: Barcelona

Re: Fresnel lens stack for "supernatural" FoV

Post by PatimPatam »

WOW mind-blowing stuff geekmaster. You're doing an amazing job, not only with your own experiments and projects but by helping others as well.

Sorry i can't contribute much to this topic, my knowledge about optics is barely enough to keep up.. I just hope that if all this is confirmed by Palmer that they're still in time to provide a set of anamorphic lenses with the Dev Kit, or that at least that they can provide down the line an upgrade kit + different pre-warp options, as you mentioned.

And don't worry, i don't consider myself old but i ALWAYS find mistakes AFTER i posted something, if we find enough empirical evidence maybe it could be postulated as a universal law.
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Fresnel lens stack for "supernatural" FoV

Post by PasticheDonkey »

i think it's a personality thing rather than an age thing. luckily for the rest of you i have a spell check plug in for my browser, or things would be even worse. i'm not good at remembering arbitrary things; but better than most at remembering pertinent things. of course arbitrary things still need to be remembered, so it's good to have people around who can remember them or have them written down.

i could do with a grammar check plug in
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

For those new to this thread, here is a quick summary:

Because your eyes see in landcsape mode (wider than their height, like eyeglass lenses), while a SBS-Half (side-by-side half-width) stereoscopic image pair are each in portrait mode (taller than their width), anamorphic lenses are needed to stretch the images from portrait mode to landscape mode for maximum FoV (field-of-view).

With ultra-wide display screens (such as a PAIR of LCD panels) anamorphic lenses are not necessarily required because each eye sees a dedicated widescreen (landscape mode) display. But to keep size and weight and costs down by using a single display, anamorphic lenses are certainly less expensive (and lighter weight) than a second LCD panel (especially when fresnel and/or custom anamorphic lenses are used).

As I suggested before, hybrid lens that act like a pair of eyeglasses, except with anamorphic (or offset) fresnel lenses extending BEYOND the eyeglass frame area, would give the best of both types of lenses, using solid acrylics in the center, and thin lightweight fresnel in the outer peripheral zones.

Remember, to test HMD anamorphism, just cut the outer 1/3 from each side of a 6-inch magnifier, and stack them. The result is horizontal stretching more than vertical stretching, proportional to distance from original lens center (i.e. aspheric anamorphism). Add another 3x magnification layer (fresnell of reading glasses) to fill the FoV. Then view an SBS-half image through the lens stack, positioned such that all you see through the lens stack is pixels intended for that eye. You will see that there is almost no wasted pixels at the vertical or horizontal margins of the display.
Last edited by geekmaster on Wed Feb 27, 2013 1:31 pm, edited 1 time in total.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Fresnel lens stack for "supernatural" FoV

Post by geekmaster »

PasticheDonkey wrote:i think it's a personality thing rather than an age thing. luckily for the rest of you i have a spell check plug in for my browser, or things would be even worse. i'm not good at remembering arbitrary things; but better than most at remembering pertinent things. of course arbitrary things still need to be remembered, so it's good to have people around who can remember them or have them written down.

i could do with a grammar check plug in
I used to be a very fast and accurate typist. That has changed for the worse in the past couple of years, at an increasing rate. I have not noticed any accompanying changes in personality, nor has that been reported to me by others. The only thing changing here is my age, so I blame that...
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Fresnel lens stack for "supernatural" FoV

Post by PasticheDonkey »

you're avoiding being a grumpy old man then?
Post Reply

Return to “VR/AR Research & Development”