By using the Force plate, a sticky-foot and the ForceSensingAnkle in parallel, a calibration matrix can be determined that transforms strain gauge signals to forces through the load cell. Assume that the ball-joint in the foot will prevent moments from being read at the load cell, we can assume that forces through the force plate are the same as the forces through the ball-joint.
The ForceSensingAnkle and sticky-foot are mounted at the end of a 4-ft 5/16" diameter wooden dowel, to maintain the orientation of the ankle. The flexiblity of the dowel ought to allow small translations while preventing any rotations. A piece of printer paper is adhered to the force plate to provide a clean surface for the sticky foot to contact with.
LabVIEW was used to sample data from both the force plate and ForceSensingAnkle. The data rate was 1 kHz over a 45 second trial. The strain gauge data was filtered using a Gaussian window filter with a diameter of 100 msec. Second, creep is accounted for by producing an offset signal from a piece-wise linear signal, set to zero the signal when no forces are applied. (The no-force points are chosen manually.)
8.4732 | -1.7714 | 1.2175 |
-5.8559 | -6.6459 | -9.9062 |
-23.7120 | 7.1519 | -36.3181 |
Using this matrix, we can transform the ankle data:
The drift was measured by measuring the strains across a cold (recently powered) ankle for slightly less than a half-hour. The assumption is that current across the strain gauges will warm the ankle, causing the readings to drift. These are plots taken from the trial (the data is filtered over 1 second intervals.)