4 ms·
>Right now the closest tech to do this would be something like WiFi Is that true? 802.11ad (wigig) should be able to handle it. https://www.networkworld.com/
by basch 5y ago
>Right now the closest tech to do this would be something like WiFi
Is that true? 802.11ad (wigig) should be able to handle it. https://www.networkworld.com/article/2172394/understanding-where-802-11ad-wigig-fits-into-the-gigabit-wi-fi-picture.html https://www.networkworld.com/article/2172394/understanding-w...
Have you ever used a wigig dock? You plug everything into your dock (hdmi, usb, ethernet) and it all gets beamed to the computer. ~10 microseconds latency.
edit: I see another commentor mentioned its successor 802.11ay. Same thing applies, except that 802.11ad is a delivery from the past, not a future promise. we are already there.
- kanetw 5y agoUnfortunately still not high enough for low-compressed (DSC 3:1 ~= 9.2Gbit/s) or uncompressed video (~= 27.5Gbit/s) at least at our resolution. But might be viable for lower-res displays.
- basch 5y agoWhat about splitting up the load? Pre-processing happens on the floor cube, with final work done on the headset. The headset doesnt have to be a video screen only.
- kanetw 5y agoCould be doable, and might be something we investigate in future iterations. I don't think it's worth it for this one, though.
- freemint 5y agoThe intermediary stages take more information then the final picture. So by moving part of the processing to the device you increase bandwidth requirements
- basch 5y agoThe thought crossed my mind, but here is an example that I am not sure actually ends up being helpful but illustrates a point. Imagine some kind of game with rain where water is collecting on your lenses. You could render and compress the frame, send it over, and distort it further You could even render the first source at 60fps, and the rain at 120fps. Or the first image comes over at a lower resolution, is upscaled, and then the rain effect is rendered at full resolution. The same could apply to synthetic film grain. Compressing a more pristine image and then adding film grain later should allow for significant additional compressibility? Decoupling the rendering into two layers could possibly allow for more resolution and framerate tricks like this? Or even color space upscaling upon display? Would it be possible to send half the color depth in even frames, half in odd, and have a nn up color both frames to their original?