How the Wiimote Works

The New York Times' 'How it Works' series touches on a remote with a twist: the Nintendo Wiimote. The article describes the micron-sized machines that make it work, displays cut-away graphics of the little white marvel, and rounds out the discussion with a breakdown of where the tech came from. From the article: "The controller's most-talked-about feature is the capacity to track its own relative motion. This enables players to do things like steer a car by twisting the remote in the air or moving a game character by tilting the remote down or up. 'This represents a fabulous example of the consumerization of MEMS,' the tiny devices known as micro-electro-mechanical systems, said Benedetto Vigna, general manager of the MEMS unit at STMicroelectronics, a leading maker of the accelerometers embedded in the controllers. (Nintendo itself declined to talk about the controllers' inner workings.) He said the motion sensors, using the technology that activates vehicle air bags, can accurately sense three axes of acceleration: up and down, left to right, and forward and backward."

Analogies rule!!!

[jibster]

These accelerometers are so sensitive, Mr. Vigna said, because electrons -- those subatomic particles that whirl around the nucleus of atoms like a video game in the making -- can sense the subtle atomic-level movement of the silicon structures.

Even for /. that's a messed up analogy. How is an electron like a video game in the making? Or is it the whirling around the nucleus that resembles the video game creation process?

Re:Not informative

[Telvin_3d]

Why would you want to have to calibrate it for the TV? The way it is set up right now, it is all relative. It doesn't have to aim exactly where you are pointing, it just has to change its position based on your movements in a consistent manner. The way it is now, i can go from a 23 inch TV to a 9 foot wide projector image with absolutely no calibration. Plug it in and go. That's really special.

Re:Motion or angle?

[acidblood]

Having done some inertial engineering work, I can shed some light on this issue.

The Wii-mote has accelerometers on it. These sense forces applied to it (gravity included), and their combined output is a vector indicating the direction and modulo of the resultant acceleration. Assuming you're standing still, you'd only be subject to the force of gravity, and by definition it always points to the ground. So do a few vector computations and you know the orientation of the Wii-mote. There is a problem with a `blind axis' (rotation in the same axis as gravity can't be detected), but ignore that for now.

On the other hand, if you want to estimate position, here's what you have to do: given an initial position, read the accelerometers, subtract the effect of gravity from the acceleration vector (harder than it seems, since the Wii-mote could be pointing anywhere, really), which then gives the `real acceleration' of the system. Now integrate this once to obtain velocity, and again to obtain position. There's just so much room for error here, that I don't know where to start. Limited accelerometer resolution, poor A/D converters, temperature drift, numerical accuracy issues, you name it. Integrating measurements (not only that, but integrating twice!) is just a recipe for disaster. Then there's a fundamental limitation to accelerometer devices: rotations can't really be distinguished from translations. Just think about it -- a given resultant acceleration vector could be the result of pointing the Wii-mote in any given orientation, added to a specific acceleration in a specific direction. You just don't have enough information to distinguish between the two -- not with accelerometers alone, at least.

Hope that helps.