For Make: Magazine volume 52 (available late July), I got to go to a place I had dreamt of for many years; Valve Software. If you’re not a video gamer, that name might not mean much. If you are, then you’re probably cursing at your screen right now out of jealousy.
Valve is a unique and wonderful creature. They consistently put out amazing games and push the boundaries of how gamers interact. Not only do people play the games, but there are entire ecosystems in place where gamers can create, sell, and buy digital goods around those games. For some, these ecosystems are more lucrative than their day jobs.
While I was marginally let down by the fact that Valve does truly exist as an office building, and not some other imaginary science fiction based facility, I was quickly reinvigorated by glimpses at top secret prototypes.
The magazine articles always feel criminally short since there’s only so much space on the paper. I felt like we needed a ridiculously long gallery to satiate the curiosity of our readers.
Before I get started, I want to say a special thanks to the folks at Valve who took time to talk with us. Valve doesn’t use titles, so just know these folks make awesome stuff.
- Alan Yates
- Christen Coomer
- Monty Goodson
- Scott Dalton
- Jeremy Selan
- Phil King
- Yasser Malaika
There were a few more folks who stopped by from time to time and shared some anecdotes, but I didn’t capture all their names.
I also want to thank two of our internal crew who helped make this happen:
- Tyler Winegarner, who made this beautiful video
- Hep Svadja, who took these fantastic pictures
Facility
Yes, Valve has normal offices. They have normal walls, normal fluorescent lights, normal windows, normal carpet, etc. I know, it would be cool if it were some kind of lair like a villain from James Bond might have, but it just isn’t.
As you’ll see below, there are plenty of cool things to look at anyway.
This was probably my favorite part of the facilities. These markers were what were used in the initial VR prototypes for motion tracking. They’re called fiducial markers and they’re more or less randomly generated before being printed and taped to the wall. A camera on your headset (or whatever needs to be tracked) sees these on the walls and uses them to determine its own position. It is how Valve did their first VR rooms that allowed you to walk around.
As employees looked at these, they began to see shapes in them, and labelled them as such. These touches of human interaction are so wonderful and I couldn’t help but smile ear to ear as I read them. My favorite was “ghost + trashcan fire”
The funniest part is, these little names will probably ensure that the fiducial markers stay on the walls even though the systems have moved beyond that technology.
The famous test room, visible in some of valve’s demos
The vive headset pictured with some old fiducial markers
This robot arm helps assess the accuracy of the tracking systems
One lab area had multiple small scale manufacturing tools such as CNC mills.
Rigs for testing the steam controller. A 3D printer sitting right next to this happened to be spitting out prototypes we weren’t allowed to see. That excites me for the future!
I have no idea.
See those lenses? Those are for the HTC vive.
A peek into another lab area for hardware design.
More tools. Yes, there’s a tiny mill, band saw, and belt sander at Valve Software.
A station for testing some optics.
This room was used for lots of early VR testing.
These 3d printers are in use nearly constantly.
The molds for the steam controller.
Steam controller prototypes on display.
Random Hardware Hacking
Before Valve settled into the product path of the HTC Vive, they were exploring other random facets of AR and VR. Here are some examples of random prototypes as well as some custom tools that were created along the way.
My favorite prototype was this gnarly contraption called “The Susan”. That is a gaming monitor that has been disassembled and mounted to a lazy susan so that you can grab a handle with your teeth and rotate the entire assembly. Pieces were missing from this, but Alan Yates graciously agreed to model it for us.
Another view of the crazy mouth controlled prototype.
Galvonic vestibular stimulation device. It uses pulses of electricity to make you “feel” motion. They called this the “zombie device” because you can steer someone who is wearing it.
An early AR prototype. There’s a pico laser projector there, as well as some fancy optics. A camera on the back tracks feducial markers allowing you to have a fulling tracked VR monoscope.
A pocket oscilloscope, hacked to test the sensors and output data about the signal
Alan’s signal tester. It buzzes when it detects oscillation or RF interference.
A display for showing what the lighthouse sensors were seeing. constructed by Yates early in lighthouse development, he still uses it!
A camera used for optical tracking. Those LEDs emit IR light for reflecting off of tracking dots
Those plastic dots help with the angle of incidence, there are sensors beneath them.
Early prototype of a tracking system that uses lasers. Similar to lighthouse, but different.
Testing sensor layout called “UFO”. At that point, 5 points were the minimum for tracking, so this was the test item.
The back of the tracking tester. An FPGA and a bunch of lighthouse sensors.
Alan’s motor tester
Breakout box prototype
Driving cellphone screens at much higher frequency than previously thought possible.
Vive Headset
This is the prototype headset that used the fiducial trackers you saw above. Note the little camera mounted on it. You can also clearly see that it utilizes two screens in portrait mode.
Testing lighthouse
Prototype with integrated sensors.
HTC Vive prototype
Final product
In the initial shift away from fiducial markers, the headset was tracked with a camera.
An early prototype of the lighthouse design
Another view of a prototype with lighthouse sensors attached.
The optical cups of the htc vive
Inside the HTC Vive’s front cover
Vive Controllers
This controller design looked really cool. It turns out that these sensors were all initially placed by hand by Yasser Malaika, guessing where they should go. Later they created an algorithm for optimal placement and found that this prototype was surprisingly well placed for being a guess. This design, while neat looking, was dropped for the current style.
Update: Yasser clarified this point for me a bit.
At the time I did have software made by Gordon Stoll that simulated and evaluated my CAD designs as I iterated, so it wasn’t completely manual. Would have taken forever otherwise… It didn’t take occlusion into account, but it gave use a sense of which design directions might work better than others.
It was only much later, as we were doing the final design with HTC, that we started using the automated sensor placement utility that Ben Jackson created for that purpose.
Another view of the crazy looking design.
Another view of the optically tracked steam controller.
An early experiment with using the steam controller for VR or AR.
Just be careful not to squeeze the clamp. An early tracked controller experiment.
Early controllers were tracked with cameras.
note the 3d printed prototype body
With the face plate off.
The “hat”
The final design
Lighthouse Base Stations
The lighthouse base station sends a fanned-out laser beam across your room over and over. The best way to do this is with a rotating system. This prototype uses a pair of hard drives sawed in half. Hard drive motors are very reliable for long term spinning, and a convenient item to prototype with.
A handful of hard drive motors.
Another view of the hard drive based system.
Evolution of the lighthouse base station.
The insides from the front.
The insides from the rear.
Early prototype rear.
Early prototype internals from the front.
Prototype with synchronizing LEDs.
Another prototype with the LEDs.
Alan’s first attempt at sweeping a laser. This prototype uses an Xbox controller motor and a laser and would go on to become lighthouse!
though the placement of the diodes looks like it is a lighthouse sensor prototype, this was explained by Yates as being a test of IR syncing.
Initially the base stations synchronized via cable. This is the prototype of the board that allows them to sync optically.
early tests in syncing wirelessly between base stations
I may look deep in thought, and I sort of am. I was wondering if some Valve employee has to periodically grease the bearing in the prop to keep it spinning smoothly.
ADVERTISEMENT