Here it is:
It's based around a universal joint, part of a prop shaft from a car, $2 hall effect sensors (Allegro A1321EUA-T bought online from Farnell), and some 8mm cubic neodymium magnets for less than $1 off ebay. (And of course Leo Bodnar's BU0836 USB interface, which is mounted in the grey box)
The automotive uni joint is suited to use in a stick - it's tough, is a pre-made gimbal, uses roller bearings with unnoticeable slack, has a square limit of movement, has a mounting flange, and pole to put a grip on, and it was reasonably easy to attach the sensors. I don't think this will ever wear out!!
I measured the output from this hall effect sensor/magnet relationship, as fitted to this stick, and as you can see it's damn close to linear:
Details regarding the use of the universal joint, and the hall effect sensor arrangement which achieved the above output, are written up in this PDF:
But these pics summarise the essence of the angular sensing method. (hall effect sensor shown in blue, magnet is stationary for this axis, with one pole on the upper right face, and the other on the lower left)
And the rudder pedals:
And the rudder pedal's sensor:
This is same arrangement as used for the stick. The pedal bar pivots on the large bolt and bearing, and the magnet is glued to it's top. The sensor is mounted on the pedal bar, and rotates around the magnet's vertical axis. One pole of the magnet is on it's front right face, the other is on it's rear left face. So in this view, the sensor is parallel to the magnetic field lines flowing around the magnet, and reads centered.
This sensor/magnet arrangement is not new, it is an adaption of one I saw in a google e-book snippet, which used two magnets, with the sensor positioned between them. It turns out that there's no need for the second magnet, you can just use the field flowing around one of them, and mount it and the sensor on the axis of rotation.
Edit - changed file hosting