Forum Discussion
Fiber optic lens
So I have been thinking of how to correct the pincushion/barrel distortion optically without adding too many lenses. Directly on top of the OLED panel a plastic with an optic channel over each pixel (either hollow or filled in with a plastic of a different refractive index) channels the light from each pixel to the correct position on the other side. This not only corrects the lenses pincushion but solves the problem of the edge of the image being farther away from the lens, this solution can bring all the pixels into the curved focal plane of a single lens. Hopefully despite its crudeness this concept image I made in paintshop can help you understand what I am talking about.
fiber optic len.png
Now the question is how to manufacture this. I don't expect we can create an injection mold with 2 million different needles that need to be pulled out at different angles, and we don't have the time to drill all the holes individually but perhaps a photo lithographic process with a photo active acid can get create channels along light beams focused on the correct convergence.
GPU based barrel distortion has worked quite well so far but there are distinct advantages to an optical solution. My understanding of time warp is that it is just shuffling the pixels in the desired direction before applying the barrel shader so if you can skip the barrel shader it reduces the latency on time warp, the could even be done in the displays own electronics. And to work well with the barrel shader it is recommended you super sample the image, working with the displays native resolution would obviously save GPU use, although you would still need a slightly wider virtual FOV for time warp.
fiber optic len.png
Now the question is how to manufacture this. I don't expect we can create an injection mold with 2 million different needles that need to be pulled out at different angles, and we don't have the time to drill all the holes individually but perhaps a photo lithographic process with a photo active acid can get create channels along light beams focused on the correct convergence.
GPU based barrel distortion has worked quite well so far but there are distinct advantages to an optical solution. My understanding of time warp is that it is just shuffling the pixels in the desired direction before applying the barrel shader so if you can skip the barrel shader it reduces the latency on time warp, the could even be done in the displays own electronics. And to work well with the barrel shader it is recommended you super sample the image, working with the displays native resolution would obviously save GPU use, although you would still need a slightly wider virtual FOV for time warp.
7 Replies
- No collimation, no IPD adjusting, no anything. And low resolution.
"raidho36" wrote:
No collimation, no IPD adjusting, no anything. And low resolution.
You should be able to adjust the lens focus a bit before too much distortion creeps in, or you could get prescription lenses. You can still adjust the IPD if there are two movable screens instead of one, Post Facebook and Samsung deals they will be getting custom screens. And the resolution in the image is not supposed to be to scale lithography can make optical channels down to a micron in width.- We have a "swear jar" for every time someone in the office says "fiber optic faceplate". They sound appealing, and they have certainly been tried in VR a few times. I'm not an optics guy, so I don't really know the details of why, but we're certainly aware of them, they're just not a good solution to what we're trying to do.
"tomf" wrote:
We have a "swear jar" for every time someone in the office says "fiber optic faceplate". They sound appealing, and they have certainly been tried in VR a few times. I'm not an optics guy, so I don't really know the details of why, but we're certainly aware of them, they're just not a good solution to what we're trying to do.
You guys should have a "neural interface now" jar instead imo.- what about micro lenses? aint those a good cheap solution? also for shrinking the form factor.
"tomf" wrote:
We have a "swear jar" for every time someone in the office says "fiber optic faceplate". They sound appealing, and they have certainly been tried in VR a few times. I'm not an optics guy, so I don't really know the details of why, but we're certainly aware of them, they're just not a good solution to what we're trying to do.
It does seem counter to the spirit of innovation to financially punish new staff members for thinking a thought perhaps instead you should give them an informative document on the problems with them, who know maybe someone will solve them.
I am now curious as to those reasons are. A google image search for "fiber optic face plates" reveals some to have a muddy color and I dont seem to be able to find a price I guess they are hard to manufacture so getting a precise barrel distortion could be hard. The width of optic fibers is around .1mm which would only give a ppi of 250. These are the reasons I was speculating on novel ways to manufacture them in my first post instead of fused optic fibers, etching optic channels in plastic or aluminum with photo lithography and lasers but I suppose that would take quite a while to get through something so thick.
