Lower pixel density on 7" (DevKit) versus 5.6" (prototype)

[Tirregius]

So has some sort of magic panel size been come up with that would be best of all worlds for this design? Was the 5.6 just right or would a 6" or whatever be the "best" size? Just curious, not trying to start fights!

PPI (pixels per inch) is the main spec ... then there is a minimum horizontal dimension which might be around 5" ... After that its all about weight/heft.

[geekmaster]

There's no real problem, and as far as I can see what you say is exactly true. It's just interesting the different tradeoffs (and why) are being made.

All along, Palmer has made it exquisitely clear that this design is specifically for immersive GAMING, where a huge wide FoV is more important than pixel density. If you want to use primarily desktop apps in your HMD, you need a different HMD with less (intentional) optical distortion, and more resolution.

For its intended purpose (immersive gaming), the Oculus Rift Dev Kit has no affordable readily available competitive product that even comes close to what Oculus is providing, and not even at a much higher cost.

It is what it is, and for its intended applications, the Oculus Rift Dev Kit is the best that is available that most of us can afford at this time. Things will only get better for VR in the future...

[jaybug]

So has some sort of magic panel size been come up with that would be best of all worlds for this design? Was the 5.6 just right or would a 6" or whatever be the "best" size? Just curious, not trying to start fights!

a 6 inch 16:9 screen would probably be around the sweet spot. Maybe a little smaller.

The ultimate screen would be one that curves around your face. But that's a little further in the future..

[Tirregius]

So has some sort of magic panel size been come up with that would be best of all worlds for this design? Was the 5.6 just right or would a 6" or whatever be the "best" size? Just curious, not trying to start fights!

PPI (pixels per inch) is the main spec ... then there is a minimum horizontal dimension which might be around 5" ... More screen is better to a point. After that its all about weight/heft.

[blazespinnaker]

Thx Tmek, going through some of your old posts. So the lenses are co-planar on the Rift? Very interesting stuff.

Reading all this, I have to say, I think unless the gear has a very rigid structure it might easily fall out of alignment and create a risk for visual divergence in people who use it too much.

Or if it's not properly aligned for that individual.

Ugh. The liability could be awful. Sigh.

[tmek]

So has some sort of magic panel size been come up with that would be best of all worlds for this design? Was the 5.6 just right or would a 6" or whatever be the "best" size?

That is an interesting question and one I've thought about a bit.

But it's also hard to answer in a way everyone agrees upon because of all the little details involved and possible ways of doing things.

Here is my way of thinking about it.

Let's assume:
a. We're talking about a Rift like solution: a single physical display with each half dedicated to an eye's view using simple optics to magnify and focus.
b. Each eyes field of view can be described as an extremly wide cone. (ignoring things like your nose, eyebrows and other features of the face that can oclude part of the cone).
c. We want to support IPDs between 60mm and 70mm.

You can intersect that cone with a plane at any point along the forward looking vector and you would have a disc. The closer to the eye the smaller that disc can be.

There are a few things that make you want to restrict the disc size.
#1 Since we're building a HMD you want it small and light for comfort and freedom of movement.
#2 The look forward vector from each eye needs to line up in the center of it's disc.
#3 The the discs can't overlap.

Magnification helps with keeping the disc size down and helps with the overlap issue but ultimately the minimum IPD is what determines the maximum size each disc can be (without wasting pixels around the edges);

I've attached a diagram for what I think the ideal physical dimensions would be given the assumptions above. Note that it has a uncommon aspect ratio (effectively 2 to 1, but slightly more than that to allow an adjustable IPD range). Unless you had the clout to go to a manufacturer and get this type of exacting specifications you would actually probably just find screen that was close to these physical dimensions, and within that you want as much pixel density as you can possibly get.

[KBK]

the next thing that happens, is that the current manufacturing always specifies known numbers,

The Sheet sizes for the glass during LCD manufacturing is standardized

The cutting hardware for smaller sheets is standardized

The sheet sizing is standardized for those breakdowns in sizing and scrap minimization.

Even more aspects (considerations beyond the basics mentioned) of the panels are standardized. No-one, from Sharp to Samsung to Panasonic or whomever... can suddenly pull customized panels out of their butt. Not happening.

This ends up saying, in the end, that tablet panel sizes are pretty well set in stone, unless you want to buy 200k-300k up front, and then they'd probably still not do it. bigger numbers and bigger commitments, as they can use the space and the hardware to make a lot more money than you can provide. Servicing your demands could be a loss, not a gain.

The odd size required for HMD use...is not ever going to erupt in the world of tablets, which tend to hit more standard numbers like 4:3, and 16:9.

16:9 is the closest standard aspect ratio...and this end up giving you a 6.5 inch diagonal screen required, at a minimum.

7.0" diagonal ends up being what you get your hands on, with a 16:9 aspect ratio.

And..right now, in the world of 7" diagonal 16:9 screens..at the time the choices had to be made, there was maybe 2 that fit the bill. That's it. Only two that would do the right job.

But...that is changing.

[tmek]

And..right now, in the world of 7" diagonal 16:9 screens..at the time the choices had to be made, there was maybe 2 that fit the bill. That's it. Only two that would do the right job.

But...that is changing.

Yes the 7" is clearly the best for the Dev Kits but looking forward for possibilities of a consumer version Rift look how nicely this 6.1" IGZO Sharp Prototype matches up with the Rift's needs.

It's just a matter of pricing and availability but this display's dimension and PPI would make for an amazing consumer version of the Rift. (providing it's strong enough in all the other required areas refresh rate, color, contrast, pixel switching, pixel fill. But these things typically all get better with future generations of display technology, not worse)

The purple area is the 6.1" display area, you can consider the purple and gray areas as giving "bonus FOV" which actually works out great because you want that extra fov in the periphery of your vision.

Edit: my dimensions were off, it's a 16:10 display not 16:9.. I've fixed and updated the image below

[jack612]

I don't think most people are going to be willing to pay for the kind of hardware needed to run games at 60fps in 1600p. That kind of resolution is probably overkill.

[druidsbane]

I don't think most people are going to be willing to pay for the kind of hardware needed to run games at 60fps in 1600p. That kind of resolution is probably overkill.

That's the wrong attitude though. If a cell phone can drive this display and it doesn't cost much we should definitely put the best in there. The same way games on PC's run on anything from the lowliest PC with integrated graphics and are quite ugly to the best ones on eyefinity setups with 3 or 4 30" monitors we should keep the Rift as the best hardware device out and let the market support it with the right cards when they are ready. Even if you need simpler geometry to get that kind of resolution that would still be preferable to a blurry pixelated experience of a smaller display.

[twofoe]

I don't think most people are going to be willing to pay for the kind of hardware needed to run games at 60fps in 1600p. That kind of resolution is probably overkill.

That's the wrong attitude though. If a cell phone can drive this display and it doesn't cost much we should definitely put the best in there. The same way games on PC's run on anything from the lowliest PC with integrated graphics and are quite ugly to the best ones on eyefinity setups with 3 or 4 30" monitors we should keep the Rift as the best hardware device out and let the market support it with the right cards when they are ready. Even if you need simpler geometry to get that kind of resolution that would still be preferable to a blurry pixelated experience of a smaller display.

And if you have a game you can't run at that resolution, then just run it at a lower resolution. This could be great for even those with less powerful PCs - think of being able to use a virtual reality desktop and actually being able to read the text. If you're running it at a lower res, it should also cut down on screen door effect.

[tmek]

I don't think most people are going to be willing to pay for the kind of hardware needed to run games at 60fps in 1600p. That kind of resolution is probably overkill.

Well the broad point is technology gets cheaper and cheaper over time, eventually this type of display would be available cheap enough to put in a $300 HMD. This display is intended to be mass produced for broad consumer adoption in tablets.

Consider not every gamer is willing to pay for a $1000 for an Nvidia Titan, yet they exists as consumer products. VR gamers will hunger for (not need, but hunger for) higher resolution, potentially more than your typical desktop gaming experience because it is spread out over such a wide field of view. Just like there are multiple tiers of graphics cards there is no reason not to offer consumers multiple choices for headset specs, for example "Rift" and "Rift-HD" at a higher resolution. Both completely compatible with the same line of games one just offers an improved experience at a higher price Just like a GTX660TI versus a TITAN, or your various quality tiers of Oasis haptic interfaces.

[Lark]

Hardly overkill. 2560x1600p is merely 1280x800 per eye and thats taking up almost your entire view. 1280x800 is pretty low resolution these days, even when its taking up a small portion of my view.

[geekmaster]

Hardly overkill. 2560x1600p is merely 1280x800 per eye and thats taking up almost your entire view. 1280x800 is pretty low resolution these days, even when its taking up a small portion of my view.

You may have four eyes, but I only have two eyes, so I will get 1280x1600 per eye from a 2560x1600 display.

[marbas]

Hardly overkill. 2560x1600p is merely 1280x800 per eye and thats taking up almost your entire view. 1280x800 is pretty low resolution these days, even when its taking up a small portion of my view.

You may have four eyes, but I only have two eyes, so I will get 1280x1600 per eye from a 2560x1600 display.

I almost wish we had monovision. Everything would be so much easier then

VR Day 0. 1280x800 to.eye

[Lark]

You may have four eyes, but I only have two eyes, so I will get 1280x1600 per eye from a 2560x1600 display.

Hah woops. My judgement still stands though, pretty far from overkill. 4K displays actually sound useful in a VR headset.

[KBK]

I don't think most people are going to be willing to pay for the kind of hardware needed to run games at 60fps in 1600p. That kind of resolution is probably overkill.

That's the wrong attitude though. If a cell phone can drive this display and it doesn't cost much we should definitely put the best in there. The same way games on PC's run on anything from the lowliest PC with integrated graphics and are quite ugly to the best ones on eyefinity setups with 3 or 4 30" monitors we should keep the Rift as the best hardware device out and let the market support it with the right cards when they are ready. Even if you need simpler geometry to get that kind of resolution that would still be preferable to a blurry pixelated experience of a smaller display.

And if you have a game you can't run at that resolution, then just run it at a lower resolution. This could be great for even those with less powerful PCs - think of being able to use a virtual reality desktop and actually being able to read the text. If you're running it at a lower res, it should also cut down on screen door effect.

Your next problem is that you need an on board (On the HMD LVDS) scaling engine and that adds lag. So you end up wanting a fast direct throughput, with no scaling in between, to cause lag. You want pure display, no scaling, direct throughput to the panel. This means it all has to be done at the PC, if it is fast enough. The PC GPU then needs to be fast enough. so you are back to square one, driving a quite decent high pixel density HD panel at a true v-synced 60hz.

There is the small speed advantage, possibly...of not having to compute the render, but just interpolate the rendered frame (pros would know the answer to that one, I don't, offhand). When and if the horsepower of the PC and GPU finally get high enough, then you can go to a full v-synced 60hz render.

[geekmaster]

My judgement still stands though, pretty far from overkill. 4K displays actually sound useful in a VR headset.

Why stop at a measly 4K? With a sufficiently large exoskeleton, you could mount a pair of these 8K displays in the "HMD". That should be adequate until we get our direct neural implants:



Just sit between a pair of 8K displays in your exosuit/mech cockpit with a mirror arrangement like this:



Attempting to turn your head would actively turn your body (seated or standing) instead to keep your head facing into the mirrors.

[Lark]

haha thats pretty whacky!

But something like this http://www.theverge.com/2012/10/25/3552 ... 4k-display while not yet practical for VR, shows that 4K devices are shrinking.

[DaveNagy]

Tmek, thank you for your contributions to this thread. They have been a big help to me.

It does seem like a 6" diagonal 16:9 may be the sweet spot going forward, a compromise between the "perfect" size, and relative commonness. Palmer must have been bummed when the Kickstarter outgrew those 5.6" panels he had found!

I wonder how big an order you would need to justify a custom 2:1 ratio screen. Huge, I'd imagine.

[KBK]

I'm at 7.1cm on my IPD, which means that the 5.6 was a bit small for me. I'm glad it is a 7' panel. (Lots of other things I'm not getting into here, I'm talking panel size alone)

When pixel density goes up, which it is..then IPD vs pixel density issues go away. The range of potentials goes up, you might say, when building a commercial release unit.

In that case of higher PPI, you can select a wider panel and end up being able to have a truly universal panel for a wide range of IPD's.

[KBK]

Tmek, thank you for your contributions to this thread. They have been a big help to me.

It does seem like a 6" diagonal 16:9 may be the sweet spot going forward, a compromise between the "perfect" size, and relative commonness. Palmer must have been bummed when the Kickstarter outgrew those 5.6" panels he had found!

I wonder how big an order you would need to justify a custom 2:1 ratio screen. Huge, I'd imagine.

Qualcomm had to throw $120M US at a starving and drowning sharp corporation in order to get that 6.1" IGZO HD panel moving toward production. Only when Sharp was desperate, did the funds/action happen so cheaply and easily. Imagine what it takes financially...to approach a healthy going concern - with a proposal. The Apple iPad stands at 100 million units sold, to date.

[play2lose]

What about the rumored 5.9" 1080p AMOLED for the Galaxy Note 3? I would love that screen in the rift.

Seems like it would be a good size/resolution combination, plus having TRUE blacks would probably help a lot in VR. On LCD all our blacks will look dark grey and throw the illusion off a bit. Also with the new 5" 1080p AMOLED in the Galaxy S4 they have shown you can calibrate it to natural colors so the "amoled is too over saturated" argument is null.

The only thing that may be a deal breaker is response time, as I have no idea how these screens perform in this area.

[Kazioo]

Your next problem is that you need an on board (On the HMD LVDS) scaling engine and that adds lag. So you end up wanting a fast direct throughput, with no scaling in between, to cause lag. You want pure display, no scaling, direct throughput to the panel. This means it all has to be done at the PC, if it is fast enough. The PC GPU then needs to be fast enough. so you are back to square one, driving a quite decent high pixel density HD panel at a true v-synced 60hz.

There is the small speed advantage, possibly...of not having to compute the render, but just interpolate the rendered frame (pros would know the answer to that one, I don't, offhand). When and if the horsepower of the PC and GPU finally get high enough, then you can go to a full v-synced 60hz render.

Can't the warping shader do the scaling without significant additional latency?

[druidsbane]

What about the rumored 5.9" 1080p AMOLED for the Galaxy Note 3? I would love that screen in the rift.

Seems like it would be a good size/resolution combination, plus having TRUE blacks would probably help a lot in VR. On LCD all our blacks will look dark grey and throw the illusion off a bit. Also with the new 5" 1080p AMOLED in the Galaxy S4 they have shown you can calibrate it to natural colors so the "amoled is too over saturated" argument is null.

The only thing that may be a deal breaker is response time, as I have no idea how these screens perform in this area.

I think the only remaining problem is the pentile layout of the pixels. They aren't true pixels and I believe that will be more annoying when the pixels are as large as they are though a headset.

[Endothermic]

Well most of the rimors for the Note 3 are of an IGZO panel not an AMOLED one.

The Note 2 didn't have a pentile matrix (though it wasn't a std rgb matrix either, but atleast it was rgb not pentile), so you if they did stick with AMOLED for the Note 2 that doesn't mean it's going to be pentile.

Won't know anything for sure untill there is some official word on it.

[KBK]

Your next problem is that you need an on board (On the HMD LVDS) scaling engine and that adds lag. So you end up wanting a fast direct throughput, with no scaling in between, to cause lag. You want pure display, no scaling, direct throughput to the panel. This means it all has to be done at the PC, if it is fast enough. The PC GPU then needs to be fast enough. so you are back to square one, driving a quite decent high pixel density HD panel at a true v-synced 60hz.

There is the small speed advantage, possibly...of not having to compute the render, but just interpolate the rendered frame (pros would know the answer to that one, I don't, offhand). When and if the horsepower of the PC and GPU finally get high enough, then you can go to a full v-synced 60hz render.

Can't the warping shader do the scaling without significant additional latency?

My experiences with playing with pure resolution in highly modified CRT projection systems tell me that this would create a secondary concern of unwanted blur in motional frame to frame in-situ eyeball analysis and use. Ie, how the eye sees it. the eye would see the warping shader and the scaling interact, for a lowered quality aspect in the form of rings or strands/spots. Places where the interpolation and the warping interact in an unwanted and noticeable manner. The scaling/interpolation is designed for flat screens and specific aspects that have nothing to do with warping algorithms. Throwing extra conversions on top can damage or modify their basic assumptions and mathematical implementations.

When you stack algorithmic functions like that, in a non linear manner, you get patterns showing up, that can lower perceived resolution and can literally be seen. You have to do the basic analysis of potentials, then train the eye to see it. To consider a potential quality aspect, and then look for them. Conversely, to see any quality aspects of note, in any direction, then reason out the causes behind them.

This is what I expect to see, but my experience is in dealing with stacked systems of conversions and interpolations on flat screens In CRT projection...then switching to highly modified digital projection (all types) systems and doing the same 'visual' experimentation work with them. The warp could make 'untoward' patterns, I'm thinking. Experience predicts this. Reality may be different. The level of perceptibility, how much it changes the image fidelity - is to be seen. Some may notice it more and some may notice it less. I try to see the issues and then work at describing them, so the source can be ascertained, and a solution then worked on.

I can't do any analysis until I have one in my hands. My 'job' for nearly 10 years, was to find the edge of image quality, and pick away at it. To find ways to improve it.

maximum fidelity in motional video that uses digital display systems is complex enough. Adding warping and then scaling can create some odd situations.

What I suspect, is that there will be a exact H x V feed for the panel that will give the best image fidelity compromises..... and that is it. Not one pixel higher, and not one pixel lower in HxV. Anything else might vary the perceived image fidelity by very noticeable amounts. That is what experience teaches me.

This, is an analog statement (driving CRT projection systems) in a digital world (digital panels and digital projectors), is the problem.

What this means is that digital does indeed create it's own problems. Specific resolution parameters that digital LVDSs' require in order to function, they may be far from optimal when all the conversion is added in. But, you are stuck with them, like it or not.

Then you have to go to PC/GPU based 'supersampling', and then reduction to the acceptable panel LVDS feed parameters. It is your only way past this limit.

And that is where the exact pixel ratio scenario will return as a thing to be pursued. In the form of specific aspects of H x V, and so on.. in the supersampling parameters. Fidelity will be found there.

Most of this will go way as a concern, when the Rift gets higher PPI and higher resolution panels. So it is almost a wasted bit of writing. But not quite. It will still play out for the people who pursue maximum visual fidelity.

[Visual Knight]

Pixels per inch, resolution, screen door affect, latency issues.

This device is still in prototype stages... a beta

My friend who is former military told me that currently low resolution displays for VR and AR costs upwards in the $10k range even nowadays. He also explained to me when your eyes get 2 screens mixed with a mesh of another 2 screens it interlaces where you will not see things pixely all the time. Our brains will see the overlap of the 2 screens and interlace them together- even though the 7" screen will give us a smaller FOV than the 5" model.

Another thing....
Anything beyond 1080p will be tough for the consumer market to play- they'd have to use the newest GPU's, spend quite a bit of money. But Oculus VR Inc is smart- but by lowering the specifications of the Oculus device- They force the market to make sure that the newer games come out work with this platform. I really like the idea of them giving us lower resolution panels to work with- it forces the game industry not to put a gamers wallet out of reach towards this technology- keeping it affordable. Much like the way Nintendo does business with innovation and not just specifications. The last thing Oculus VR wants is for there product to become a dud.

[Endothermic]

- even though the 7" screen will give us a smaller FOV than the 5" model.

The FOV on the 7" isn't smaller it's slightly larger.

[Parallaxis]

Resolution of the Rift is one of the most discussed issues on this board, and that is understandable and OK. But I feel that people are comparing apple and oranges when it comes to 3D HMD with low latency head tracking and standard monitors.

I have a few inputs about resolution from my experience.

Regarding 60fps per eye stereoscopic vision
Yes, the Rift is 1280x800 and that is divided between two eyes. That is 640x800. But the Rift is also stereoscopic and each eyes sees a different view with different information. So that actually means the resolution is somewhere between 640x800 and 1280x800. Higher for objects closer to you and lower for object father away.

Head tracking vs. resolution
When your move a video camera you will actually get more information than the resolution of the movie you are shooting. This is also true for head tracking. If you move your view just a few pixels, you will receive more information than the resolution of a still.

Another thing about head tracking is that your view is actually moving. So if you turn 360 degrees with a 90 degree FOV, your horizontal resolution is not 640 pixels, but 2560 pixels. Because of the stereoscopic vision it's actually between 2560 and 5120 pixels horizontally.

[Laserschwert]

So if you turn 360 degrees with a 90 degree FOV, your horizontal resolution is not 640 pixels, but 2560 pixels. Because of the stereoscopic vision it's actually between 2560 and 5120 pixels horizontally.

That's some wacky math there... in other words, it's wrong. It WOULD be right, if you could only turn your head in steps of single pixels (so, to make this simple, let's assume the 1280 pixels of the screen equal 90 degree, so one pixel equals 0.07 degrees) - if you could turn your head only in 0.07 degree steps.

But as you stated yourself, the "higher" perceived resolution comes from being able to turn your head on a subpixel level (no steps of 0.07 degrees, but anything inbetween as well). So technically, the "resolution" of the image around you is unlimited, you're just looking at it through a resolution-limited mosaic window

[Randomoneh]

So if you turn 360 degrees with a 90 degree FOV, your horizontal resolution is not 640 pixels, but 2560 pixels. Because of the stereoscopic vision it's actually between 2560 and 5120 pixels horizontally.

Nobody ever defined resolution in that way.