The torque measurement setup is pretty straightforward. Instead of fixing the motor mounting points to the fixed world, the mounting points are fixed to a body that can rotate around the motor axis. That body is then connected by a torsion spring to the fixed world. What looks like copper is actually a steel spring, coated with copper to prevent corrosion (custom spring made from 3mm steel TIG welding wire). When the motor exerts a torque the springs twists. I’ve added a small magnet to the back of the motor, an absolute hall position encoder then reads out the magnet rotation/spring twist. The value it gives lies between 0-5V or can be read out over an SPI bus, so no additional sensing electronics are required.
The main reason I chose this configuration, is that I was afraid that if strain gauges were used, the sensitive signals would be disturbed by the hard EMI radiated by the wire, which would then lead to all sorts of shielding issues. I also wanted to minimize the number of wire guiding wheels, to make wire handling as easy as possible. Especially during experimentation the wire will break frequently.