If SLAM is integrated into Dream Air, does it mean only the SLAM version get to enjoy hand tracking? And if you have the Lighthouse version, you wouldn’t be able to use it on for example on an airplane, right? Why do you use the word “integrated” if it wasn’t integrated into the Lighthouse version? Please clarify!
Also, if you implement COBB into Dream Air, how do you control apps in COBB? Do you require SLAM to use COBB or does COBB come with its own input device similar to what they did with Rokid Station?
I’ve been using a PCL for about a year on the same Ryzen 5900X based machine. I often have intermittent out-of-sync tracking where I turn my head but the PCL doesn’t show that I’m looking in the new direction until a moment later. This is obviously uncomfortable. Is my CPU too slow? I can understand tracking updates being slow under heavy CPU load, like any game having to run at a lower frame rate, but do you consider tracking being out-of-sync a bug that will one day be fixed?
Credit to Pimax support though! They’ve worked with me on this several times. With their help and with some recent software updates, the frequency and duration of the out-of-sync tracking events have improved.
Are you considering doing SLAM tracking on the HMD in future revisions of the Crystal Light and/or Super? Considering that a PC can be quite busy doing the work of driving these demanding HMDs, I can imagine it’s tempting to move something as critical as SLAM tracking to dedicated hardware where it wouldn’t have to compete for CPU time.
Why do SLAM tracking cameras use relatively low resolution sensors? Would SLAM tracking work any better with higher resolution cameras? I have a PCL and a Meta Quest 2 and both appear to use cameras of roughly the same quality, at least based on the pass-through view.
Is it possible to make headset support both SLAM and base station tracking?
To put both tracking hardware and allow users to toggle which one is used.
There are multiple posts and comments asking for lighthouse faceplate for Crystal Super.
Moreover on Dream Air you won’t be able to add base station tracking if you chosen SLAM version. Since SLAM tracking for new Pimax headsets require months to mature (Super tracking still has issues today judging by user comments) it will put Dream Air SLAM version users into difficult situation and will result in frustration, more RMA and support cases and devaluation of Pimax’s image.
Why such heavy investment into SLAM?
From what I understand SLAM tracking is difficult to implement because it’s a software solution and tracking on newly released Pimax headsets requires a lot of time to mature and become stable which makes negative user experience for early adopters.
Maybe it would be better to offer headset and optionally base stations tracked controllers (without base stations). To keep to cost low. It’s more difficult to find controllers today rather than base stations (as they are produced by Vive). There are few options since Valve stopped producing Index knucles. And some of them are non-standard (Etee or Shiftall Flip). People even reported using heavily outdated Vive Wands due to lack of options.
Base stations tracked controllers production cost is not higher than inside out tracked controller by much.
If you’re going to design your controllers, please don’t invent the wheel. Shiftall did a smart move by replicating Index knuckles. While Play for Dream replicated Quest 3 controllers.
I would also like to summarize issues that users shared on Reddit and Crystal Super discord.
I can isolate following issues related to Pimax’s SLAM tracking:
Tracking generally not as good as base station tracking in sense of reliability and accuracy. E.g. users reported being thrown away from their cabin during the flight. Users reported height (headset position) not being accurately calculated. Headset and controllers jitter. Controllers loosing tracking and flying away. Tracking quality (accuracy) overall is not as good as base station tracking.
Tracking for new headsets require months or years to mature and become stable.
Not possible to use with motion rigs due to not having fixed point of reference inside the rig. Motion compensation (if it exists) doesn’t work well enough. It makes Pimax headsets unusable until base station tracking option appears.
CPU bottleneck heavily impacts tracking to the point of not possible to be used. Not an issue for OG Crystal as tracking happens onboard.
Small issues unrelated directly to SLAM but related to the tracking. Like necessity to redo room setup each session.
Hello, I’m running pi_server with dual 9684x CPUs, which has 96 physical cores per socket.
I have noticed that SLAM could consume a lot of cores but each core is only 20% time running, would you plan to optimize SLAM to consume more cores(over 64 cores on windows) and improve per core utilization usage to provide better track stability to the controllers, reduce random screen shake when moving my head?
And would you provide more setting options for SLAM so advanced users could adjust for a better behavior, also this could be helpful for those running on limited cores users, as pi_server consumes too much core that the game they are playing could not get enough cores and FPS decreases a lot.
I’ve run into a serious issue with the SLAM tracking on the Crystal Super that seems tied to room lighting. When I start the headset in a dim or low-light environment, even with IR lights that work flawlessly for other VR headsets, including in complete darkness, the tracking goes completely haywire. In moderate lighting, it causes my PC (i9-14900K, RTX 5090) to freeze entirely. It’s bad enough that when I’m running audio apps like Voicemeeter Banana with buffers over 2048 KB, the freeze crashes the audio drivers too. It seems one of your services is heavily taxing the CPU during this time.
The workaround from support is to delete the SLAM folder in the Pimax directory and restart the services, which gets things running again. However, this only works if I significantly increase the room’s lighting, for me, that’s four 75W recessed lights at full power, especially at night.
Are you aware of this issue, and is a fix in the works? I suspect some users might think their CPU isn’t up to par when it’s actually just a lighting sensitivity problem. Thanks!
Will it help if headset has more tracking cameras?
Play for Dream has 10 cameras (2 of them I think are color passthrough) and their tracking is pretty good. Not without issues (due to controller form factor and size LEDs can easily be obstructed) but pretty good.
I’ll pass on the 10 dollar coupon, but to bring up a few questions-
1: Discussing it with Pimax before, I’ve heard ultrawide monitors are not ideal and can effect tracking Just curious about the technical reason for this. I assume it’s related to the size but is it more so being a large reflective surface, typically pretty bright for a standard room light, or both?
2: What is it about occlusion in particular that causes SLAM wo many issues, even when the controllers are noticeably held apart, in cases like aiming with a gun? Most forms of SLAM tracking I’ve tried, Pimax included, tend to work quite well with one handed and rapid moment, actions such as swinging a melee weapon, but camera tracking tends to have issues there. Technically speaking, what is it about keeping the reference points separate that poses such an issue? Is it more that the camera system loses the exact position of the controller and tries to find it again, causing the jitter, or does it start confusing the positions of the left and right hand from the close proximity?
3: On a hardware level and assuming it were improved to keep up with a higher resolution, do you think that increasing the capabilities of the SLAM cameras, such as higher resolution and ability to see color instead of just IR to make out those finer details with greater accuracy, or adding more lower resolution cameras to view more angles would be better in increasing tracking accuracy.
4: Related to #3, what would be more computationally expensive, significantly increasing the resolution the existing SLAM cameras record in(With appropriate hardware) or adding more lower resolution cameras?
Why aren’t there more cameras on the current headsets to eliminate deadzones and improve accuracy?
Why aren’t external usb slam cameras used to track/calibrate the headset and eliminate deadzones (aka hands behind back don’t float into the air) and maybe accurately set height. (maybe as an optional addon?)
Will the super ever have hand/mouth tracking, as an addon or with current cameras and if yes when?
If you make a hand tracking module or with built-in cameras, will it support gloves (for medical reasons) and haptic gloves?
