I am currently playing a lot with those. Brandano I followed your link too, as well as Gene. I read a lot and find some implementations used in robotics.
I will publish my hall design and my findings when I finish so others can benefit too. What you can expect to see is 1x magnet...probably 20x5x5 mm with polarisation thru 5 mm.
Magnet is fixed in a shaft in a way that it's passing thru center but not centered..it has an offset from center. Sensor is also offset from center.
In general it+'s a combination of magnet pass by and rotation. Magnet is still rotating over the sensor but not thru center of sensor.
So far I achieved this : Sensor reading thru full 0-5 volts ( using full resolution) on cca 65 degree of roatation. In this concept sensor's reading degree of rotation can be fine adjusted by using more or less magnet offset from center, but with magnets I have smallest angle at which the sensor read full resolution is cca 65 degree.
Also, noone mentioned this anywhere, but it is obvious that full resolution is obtained with magnet very very near the sensor, less than 1mm. Further the magnet is away less resolution you get.
I didn't use flux concentrator at all as I can accomodate bigger magnet in my design.
