The sensor shaft on my rudder pedals snapped a few days ago (it was a nylon shaft from a potentiometer, and I had carved a groove in it for a small clip retainer... bad idea) so I made a replacement for it using another pot with an aluminium shaft. I tried out a new concept to see if I could get the same sort of performance using a single magnet... well the results are mixed. Here is a comparison of the new and the old sensors side by side, in a blurry cellphone picture:

The old sensor is on the right, and it's the classic "sensor between magnets" setup, with a small metallic strip acting as a flux concentrator to increase the magnetic field between the magnets and provide some magnetic shielding.
The new sensor is on the left. On both sensors the potentiometer tracks were filed down to break the electric contact, and neither respects the original potentiometer pinout, so be careful if you try to replicate this. The actual hall sensor is buried in a slot between two metal flux concentrators, cut out from three washers stacked and glued with CA glue. The magnet is in the center. The problem with this approach, or at least with this iteration is that the full range is over 180 degrees of swing rather than over 90 degrees, that would suite me better for my direct drive setup. The advantage is that it does use a single magnet and is slightly more compact, at least in thickness. I bet that the sensing angle can be made smaller by using shorter arcs for the flux rings, and that the linearity can be adjusted by careful shaping of the flux concentrators. Does anyone know if there is software available to run simulations of magnetic flux?

[edit] To be more precise, "new" means it's the first time I try this, not the first time this has been done. A quick search on Google patents shows that the concept is not new in any way.