468 karma · joined April 10, 2016
We use the peak PPD for the Aero.
I need to confirm how sharply it drops after that, but in any case it should be above 20 PPD +-(15 to 30) degrees and above 12 PPD for +-(30 to 50) degrees.
We push about 3.4 GByte/s (or 27.5Gbit/s) to our displays. Even a 26Gbit/s DP1.4a HBR3 link needs display stream compression for that. So a wireless link will absolutely need strong compression.
H.265 can push that to maybe (ballpark) 200Mbit/s, which is fairly feasible. But that will need to be evaluated and optimized for text quality and latency. Some googling suggests 100ms end-to-end latency as good. That's pretty high for VR, IMO.
[Edit: fixed the DP1.4a bandwidth)
* Preliminary number, might change as we make tradeoffs of battery life vs thermal solution vs weight.
I put a link to the paper [1], but it's paywalled.
[1] https://www.spiedigitallibrary.org/conference-proceedings-of...
So the resolution is basically there, but what about latency? Output-to-photon is as direct as possible (NUC Displayport -> DP-to-MIPI -> Displays), so the only question is how low we can get the motion-to-output latency.
Unfortunately that's not a question I can answer yet. Tracking is hard, especially when you're designing for a portable headset. But I'm pretty sure we'll be able to get it to sub-frame latency if my mmWave idea works out.
The compute pack will be at the back of your head with the airflow going away from you, so it shouldn't be too noticeable. Thermal design is definitely a priority for us.
Putting it on your head reduces the amount of cables on your person, so you don't accidentally tangle yourself or something. Also, it improves the balance so it might actually feel better than a front-heavy headset (compare a Vive with an Index for example).
No depth mapping yet, so it'll only be in the background, but first things first.
* Square displays are what's readily available in the VR form factor (2-3" diagonal)
* Rectangular displays will murder your minimum IPD, which kills accessibility for a lot of users
* Reducing the size/increasing pixel density makes the optics exponentially more complex. We're already pushing the limits of what's possible without pancake lenses.
* Pancake lenses are a nightmare of complexity when you want high-quality images. Ghost images, low transmission, etc. all increase development cost by a lot. And the unit cost is significantly higher, but that's not an issue here (the NREs are the main cost driver)
The drawback is that the optical train is long and there's 3 complex lenses as opposed to 1 glued assembly (not sure if it's 1 or 2 lenses). Adds per-unit cost and assembly labor, and at a $300 ($800 realistically) price point that's a lot.
In our experience, a balanced headset is more important than a light one.
Compared to a Quest/Immersed combo:
* we offer significantly higher PPD (and better optics with no ghost images, Fresnel rings, etc.; but that's hard to quantify) than any other headset on the market except super-high end ones. You would not be able to read text on a Quest like you can on a normal 27" monitor; you can with our headset.
* Immersed does not do window management. We support unlimited, actual windows. Immersed only does a handful (up to 10 at reduced resolution? need to check) virtual displays, which is pretty annoying from a UX perspective.
* Immersed needs to be tethered, which is reduced quality and higher latency due to the WiFi connection.
Front-facing cameras, hmm. I'm not sure I like it from a usability POV, but it's something to consider. I think going the extra mile and including internal cameras for a seamless UX is better, though.