Qualcomm was showing off three VR technologies at Siggraph plus one 3D scanner. SemiAccurate thinks two of the four have short-term useful possibilities while the other two are a bit farther off.
The first one was technically the most interesting, it was called six degrees of freedom (DoF), As you know with VR, you usually have three DoF, essentially rotating on any axis, but moving along those axis requires a controller of some sort for most consumer devices. Going to 6DoF with handheld controllers tends to be a clunky experience, you can’t easily dodge a bullet for example, or at least do so in an intuitive manner.
To solve this with their 6DoF tech, Qualcomm took a simple 820 dev board and added a low rez camera, 640*480 monochrome, with a fisheye lens. You can do the same, somewhat, with motion tracking accelerometers and the like, but this solution is both elegant and more importantly, cheap. It uses the camera to do edge detection on nearby objects and then tracks their position by horizontal movement or size change. The heavy lifting is done by the DSP in the 820 and the translation matrix is fed to the GPU for 3D work.
This allows the VR rig to quickly, easily, and relatively computationally inexpensively track the users movement along the three remaining axis. The normal accelerometers already present in a vanilla 820 board still track rotation, the other three DoF, and a simple camera does the rest. SemiAccurate tried it at Siggraph and it works as advertised. For consumer VR headsets in a generation or two, this could have a big impact on user experience but it is a bit farther off than the some of the other demos.
Pico isn’t actually very small
Next up is Pico, one of the short-term technologies that SemiAccurate thinks should be implemented by someone very quickly. It is a VR headset with a wired controller that looks like an 8-bit console gamepad. The hardware is nothing spectacular, in fact we didn’t even write down what it was so probably just a vanilla 820 dev board. The smart bit is that the heavy stuff, the board, battery, and all that, is in the controller in your hand. The headset itself is a screen and a plastic shell, plus maybe a sensor or two. Why is this smart? Light headsets mean less fatigue and hands are better at holding the heavier bits. Sure you could put it in a backpack ala HP but, umm, there are downsides there too. Someone needs to do a Pico-like headset soon.
The other further out technology is VR related but on the audio side. Qualcomm was showing a reference design, again 820 based, that did 3D positional audio on a VR helmet. This isn’t exactly a trick but doing so in realtime with decent result from phone based hardware is new to SemiAccurate. Every VR helmet should have this but quite a few don’t, showing it off on inexpensive hardware is a good portent for things to come in a couple of generations.
Last up is the closer to market non-VR based technology, but it has uses in VR as well. It is a Qualcomm demo system that uses two cameras, one RGB and one with structured light filters to effectively get RGB + depth information. Better yet it runs entirely on a Qualcomm 820 based tablet reference system with an added camera and filter. Again not expensive and not exotic tech, it looks like this.
Just a tablet from a brief glance
No the camera isn’t that big, it is just an external module glued to the back of the tablet, a production version could do this in an ultra-slim design with ease. The neat part is you scan the image by moving the camera around like the 72 other similar solutions, but this one does the 3D reconstruction on the tablet. There is no cloud needed, no data uploads, server-side crunching, and re-downloaded images. It all happens on a tablet with a single extra camera, the entire process takes 3-4 minutes most of which is the user moving around scanning the object from multiple angles. For the cost of a second camera module with color filters removed and polarized ones put in place, it is a very useful addon.
So in the end the demos from Qualcomm at Siggraph were not of the jaw-dropping far future tech type, they were much closer to production. This is a good thing because they could all be very useful in mainstream devices. Pico and the structured light 3D scanner could be done in very short order, the 6DoF VR headset and the audio side are a little farther out because hardware changes are needed along with software. None of the four are more than a few years out though, and I for one can see uses for them all today.S|A
Latest posts by Charlie Demerjian (see all)
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018