Odds of a 1440p display?

[Jossos]

So dk2 is out, which has pretty impressive specs. and apparently since consumer version will be a step above that, what are the odds for a 1440p display? I'm sure people are willing to drop the graphic settings for clearer vision. I got a 780 so it doesn't bother me. also whats the max hz for a 1440 screen? I hear it's like 83.33 or something. I use a dvi coord so idk if that makes a difference.

[MrMonkeybat]

"owenwp" wrote:
The optical solution is well known and bound by physical laws. And it contains at least half a dozen lenses stacked on top of each other. Abrash showed a diagram at one of his presentations. If they went this route, it would be like an HMD from the 90s in form factor, because thats how they used to do it before we had pixel shaders.

If your willing to take a bit of chromatic aberation you can use a few less lenses than in that slide.

[bz1]

"xtrasmiley" wrote:
My first post, and it's to ask a question in this thread!

Since screen rez and size are so different across different platforms, how come we don't discus the OVR screens in terms of PPI? I mean, wouldn't that make more sense than talking about direct rez, expecially since a 5" vs 6" screen would result in a differnt picture seen by the user in the end product.

Anyway, I'm in no way an expert and am just curious.

I don't own the DK1, but have been looking hard at the DK2. I'll wait until it starts shipping before deciding to jump on board or not, but if I don't it will be only b/c I'll be waiting for the final version!

The size of the screen doesn't really come into it though. I mean, whatever field of view they aim for, it is the number of pixels that are stretched across that field of view that will determine the perceived resolution.

Example: Two 1080p screens of different sizes. The smaller one has a much higher PPI. Some boffin tweaks the optics to reach the target field of view. End result? An HMD with 960 pixels spread across the field of view in both cases; i.e., no perceived difference.

Pixel density is a more enlightening way to describe the display in terms of the perceived resolution, but you would need to express it as pixels per degree, or pixels per steradian. The problem with this is that it's not an inherent property of the display, but a result of the combined properties of the display and optics. So it's kind of meaningless to talk about the angular pixel density without it being in the context of a target field of view anyway.

[mrjazz]

"mrmonkeybat" wrote:

"mrjazz" wrote:

And the lenses' property of concentrating the pixels in the middle where they are needed is not too bad, is it?

The pin cushion lens distortion concentrates physical pixels in the middle, but the barrel distortion shader does the opposite spreading them out at the middle so if you want to make use of those physical pixels you have to render an image much larger than the actual screen size wasting allot of rendered pixels squished together at the edges. Also your eyes do tend to dance about a bit tracking objects, you wont always be staring strait forward.

I guess building displays with unevenly distributed pixels is not feasible. So thinkable options are:
-Using distortion-free optics (probably not feasible): Then we have a higher pixel-density (pixels/degree) at the borders than in the center (alpha_out = arctan(radius/focal_length))
-Using a single plano-convex lens per eye (the current setup of the Rift): Then the pixel-density is even over the whole field-of-view (alpha_out = pi/2 * radius/radius_max)
-Using multiple bi-convex lenses in a row (as proposed in this kickstarter-project): Then we have a higher pixel-density in the center than at the borders (probably something like alpha_out = pi/2 * (radius/radius_max)^n).

In my opinion the last option would be the best choice, because you are looking most of the time at the center. Yes, with the latter two cases you have to render more pixels than the display has, which are then more or less discarded (better: low-pass filter/averaging) by the distortion-shader. But on the other hand you are utilizing the pixels of the display more efficiently, regarding the way people typically use their eyes. And just using higer-res displays (with distortion-free optics) would increase the needed render-resolution, too. Hence, if you want a higher pixel-density, the GPU-load will increase in every case independently of the type of optics.

[Cgpnz]

What!!! 1K x 1K per eye!!! Thats what we have with FHD 1920x1080 approx.
Even 1280x800 offers 600x800 per eye.
We want 2K x 2K per eye.

[Cgpnz]

And thus we need GPU power to do 4k at 90 Hz.
Facebook needs to purchase NVidia and free everyone from
the stranglehold of microsoft and its dealings.

[MrMonkeybat]

If they managed get a screen with 2kx2k or more per eye then they would have to add eye tracking and foveated rendering to utilize it. A 0.5kx0.5k foveal resolution on a 0.5kx0.5k peripheral image should do it the foveal area is about 10 degrees so the same render should work on a 4kx4k view at 100 degrees. Including both eyes that would amount to a total of 1megapixel to render and pipe to the Rift so a modern card with display port would be capable of very high frame rates.

"bz1" wrote:

"mrmonkeybat" wrote:
A cheap but effective solution to screen door would be to mold a transparent plastic sheet with bumps the same size as the pixels. The little prisms will blend the light of the sub pixels, with the gaps between the bumps lined up with the gaps between the pixels it shouldn't reduce sharpness too much. Not too hard to do but you can see reasons why no one would bother doing that on a cell phone or TV screen but it would be perfect for an HMD.

I don't see how this would provide any advantages over a uniform diffusing film that has a point spread function with a radius roughly half the width of the space between pixels. You will only lose sharpness if the point spread function is too large.

It would seem that only a simple stamp mold would be needed to dimple the difusor film over each pixel allowing a thorough mix of each sub pixel without reducing sharpness, although reducing sharpness would reduce the amount of antialiasing required a uniform difusor could introduce slight chromatic effects on zebra patterns, although some chromatic effect are a given witout.

I tried aprouching my 1080p monitor and and plasma screen until it filled 90 degrees vertical. Then i tried the same thing on a friends 720p dlp rear projector. Eliminating sub pixels and screen door had a big impact on perceived quality. Still dreaming of a stacked barrel distorted Oled.

[Cgpnz]

It would be really interesting of Oculus can now ask for screens with v high density pixels in the middle,
and less on the outer. Perhaps OLED with 3 subpixels per pixel in the middle, and the penTile arrangement (2 subpixels per pixel) outer.

[ElectricMucus]

"Cgpnz" wrote:
It would be really interesting of Oculus can now ask for screens with v high density pixels in the middle,
and less on the outer. Perhaps OLED with 3 subpixels per pixel in the middle, and the penTile arrangement (2 subpixels per pixel) outer.

Only if HMDs become the primary market for display manufacturers. And the way pixels are addressed it's doubtful if it can be done.

[Guspaz]

Palmer and Oculus have repeatedly said that the CV1 will be higher resolution than 1080p. Where exactly that leaves them is uncertain. If they're doing a custom display (which remember will still be based on commodity OLED manufacturing technology), they don't necessarily need to stick to specific resolutions. If they need to shrink the display slightly, they could just cut a slightly smaller display and lose the extra pixels.

However, I think one thing that could probably offer relatively easy short-term gains is non-rectangular displays. The visible portion of the Rift's display isn't rectangular, so there's no reason the display needs to be either. Something as simple as a roughly octogonal display (trimming the corners off) could make significant improvements to the form-factor of the chassis. You could potentially even achieve that with a flexible panel, by folding the corners. Trimming the corners may seem simple, but it frees you from having the rectangular-box-on-face form factor that DK1/DK2 have had.

[MrMonkeybat]

Sharp and SEL showcase 13.3-inch 8K OLED display at SID-2014
http://news.oled-display.net/why-igzo-oled/

Cut down to Rift size that should still be 3k x 1.5k or 1.5k square for each eye. 1440, 1536 whichever fits best.