I was using a couple of old IBM trackpoint (mouse) sensors to use as a 2 axis force sensor.

I built a structure to house the sensor and attempted to calibrate the sensor.

While calibration provided data that showed the sensor was extremely consistent and linear, I also discovered that when I applied a force directly along the one of the axis, I would always get a change in voltage in the other direction as well. I thought perhaps that I was creating a moment around the sensor, that was causing changes in the voltage across both axis, but even when I applied a force directly on the sensor, the voltage across the other axis would change significantly as well.

If you look at the sensor, the stick portion is not aligned with the base of the sensor. When I took the sensor out of the laptop computers, the base was parallel to the keyboard and the stick portion appeared to be crooked in the keyboard. I assumed that the axis in which the sensor was aligned to, was aligned to the stick itself, but now I think it may be aligned to the base. I tried to apply forces that were aliged with the base, and noticed that there was less change in voltage in the axis I was not applying the force, so I think I need to rebuild another structure that will allow me to test this.

• trackpoint.jpg:
  

Here's a chart of the calibration I did, in one of the directions, which is the axis in the same plane of the skin contact area. You can see there is very little hysteresis and it's very consistent and linear. This graph shows 3 sets of data, where I increased the applied load from 0 to 600 and then decreased the load from 600 to 0. (the y-axis on the graph is voltage (V))

But another thing I noticed was that the sensor (as was with the previous sensor) was very sensitive to where the load was applied. If I changed the point at which the load was applied during calibration by even just 1 mm, I would get different outputs.

• graph.jpg: