I think the best bet actually would be for me to just take a camera and do a through-the-lenses shot through both. Our current optical train needs a second iteration for the mold, but once it meets our specs we'll do a blog post about it.
468 karma · joined April 10, 2016
I think the best bet actually would be for me to just take a camera and do a through-the-lenses shot through both. Our current optical train needs a second iteration for the mold, but once it meets our specs we'll do a blog post about it.
uOLEDs with waveguides (for AR) or pancake lenses (for VR) are... really complicated. Both systems currently do not offer the visual quality that a traditional refractive system can offer due to low light transmission, ghost images, and so on.
The pixel size on a high-res uOLED is so small that getting the same PPD requires pancake optics to fit in the headset form factor. Otherwise your spot size (i.e. the minimum resolution of the optics) is too large and you lose the benefit of a high-res display.
Also, bandwidth is a big concern. Our double 2448x2448@90Hz displays require DSC to run even with a DisplayPort 4-lane HBR3 link. I don't think phones can drive that.
In tethered mode, I see no reason why it should interfere with the host OS.
* Main heat generation is on the back of the head and pointed away from you
* If it's still required, active airflow for the user in the front of the headset
* Weight will be balanced
* IPD is auto-adjusted, so there's no chance of having the wrong setting and hurting your eyes.
* Better optical quality leads to less eye strain
* Adjustable padding
That being said, I'm not a Quest 2 user. I don't know what things it does wrong, so it's possible I missed a feature that we have and didn't list.
In tethered mode, totally possible.
Case 2, you build a window manager. Each window can be positioned individually and you can give specific windows custom handling, or even integrate 3D scenes. Problem is, now you're limited to Linux. Linux headset support is very limited. People don't have Linux installed, even.
How do you solve this? You build a Linux compatible headset. You minimize the friction. You allow people to put on the headset and get to work. Nobody cares that their Android phone is Linux based, after all.
That being said, right now this absolutely is an enthusiast, early adopter product. We'd like to get the costs down, but going into a costs battle when you're doing a new product type is insanity.
Our software won't phone home. The cameras will have a kill switch. If we include a microphone, so will the mic.
We have two wide-angle RGB cameras mechanically locked to the users' PD. The cameras itself are around 2k x 2k at 90 fps. For the standalone mode, the data is pushed over a PCIe gen 3x4 link so we can hit <12ms latency. Tethered is a bit more tricky as we can't count on the bandwidth being available, but we'll either compress it or do a reduced resolution.
The actual sensor is still TBD. I'd like to do this properly with a global shutter sensor (IMX547), but they're a bit too expensive. If we don't have the margin for it, something like a IMX715 is the fallback.
The compute pack will be on the back of your head, so a bit easier to cool and not as severely annoying. Also helps balance the headset.
We're not doing this lightly. Hardware design is expensive, and risky, and low margin. But we want a cultural victory, not necessarily a financial one. If that means doing it ourselves, then so we will.
Will we reach Varjo's level of polish? No way. But I'm confident we can get 80% of the way there at least.
Depends. We're planning a diopter adjustment (either as baseline or as a stretch goal), which might be sufficient. Alternately, there's enough space in the headset for glasses and it'll also be possible to get prescription lens inserts.
> I dot lots of video conferencing. How would I appear in a zoom or msteams call with this on?
You would appear like you're wearing a headset. Not much we can do about that, unfortunately.
Centroid separation for lateral color is roughly 0mm for 0-30 degrees then sharply decreases at about -0.1mm/5 degrees
When eye is rotated 25 degrees horizontally, pupil swim is within 10-20 arcmin for (+-40 V, 0-40 H) field angles. This is already fairly low, but we also plan to correct it with eye tracking if it's noticeable.
Allowable eye shift is +-2mm, eye box accounting for eye rotation is around 18mm. Full 20um RMS spot size in that region for the foveal region (+-25 deg).
Display is a LCD panel. 600:1 contrast, 150 nits brightness, 83% NTSC.
We actually considered pancake lenses, but the low transmission (15% in a pancake design vs ~85% in ours) and the ghost image issues made us step away from them.
XR2 does all the hard work, but Qualcomm pretty much ignored us when we messaged them. Besides, it'd pretty much be dead weight since we want an x86 processor for our primary use case as a laptop replacement.
And yeah we're planning to integrate the Ultraleap module if they get their Linux software sorted. If not, there's fallbacks.
(I think we also messaged Goertek and got nothing, so... a big part of why we're doing this, and will likely open source most of it, is that all the usual vendors didn't talk to us)
All the software is or will be open source, so much more than any other headset on the market. Ideally the FPGA code would be open source to, but I don't know if that's possible with licensed IPs
> Does it come with controllers?
No controllers.
> Can I hook in controllers from another system (vive, index, oculus, etc.)?
Any controller that's supported by OpenXR is supported, and we'll likely have SteamVR support as well.
> Can I play video games with this?
In tethered mode, absolutely. In standalone mode, only very light games, but might be doable.
> Can I stream a video feed from a real world camera (like with a raspberry pi) to the interior displays?
Yes. In fact, we'll have AR passthrough by default, but nothing's stopping you from using different cameras.
If you are not interested in the standalone capabilities, the headset without compute pack (i.e. tethered only) will be ~2k.
The power consumption is inline with any other ultra-low-power x86 (~15W TDP for the CPU). That's pretty comparable to a Snapdragon XR2 with ~10W TDP and worse performance.
Users would want to use proprietary apps that are only compiled for x86. Emulation for those on ARM is not particularly feasible for any acceptable performance. If you're buying a laptop now, you would not buy an ARM device.
That being said, I'm not married to x86. But I don't think ARM is the future, either.
We're considering offering a belt clip option, but I'm not sure it's worth it overall.
We're thinking about a separable battery for the V2. V1, too much complexity for now.
You'll be able to power it off the charger or any USB-PD source while charging the battery (provided it has enough power, but that should be doable).