Over the years I’ve had to design several joint torque sensors. When developing these sensing systems you have to consider a few key aspects of the system. The number one thing you should always remember is “what is this system going to do?”. Is this an actuator that is going to deal with impacts that you can’t predict like a leg that will impact the ground at an unknown time? Is this something that is going to need to apply extremely high bandwidth, controlled forces at another object like a voice coil trying to vibrate a speaker? When you answer this initial question you can usually better design a system that is capable of some incredible things.
Stiff Hips.
Pictured above are two different load cells developed to provide high resolution force (left) and torque (below) feedback in a robotic actuation system. The requirements for the sensors were that they be drop in replacements for already existing components, provide feedback for a force control application, and reduce part count when possible. The linear force sensor on the left was designed to serve as the coupler from a ball screw nut to a set of linkages that at times applied large off axis loads which were taken by a linear rail that ran parallel to the ball-screw. Due to runout in the ball-screw shaft thin flexure sections were implemented that allowed for 0.25mm deflection in the vertical and lateral directions. To provide enough length for the two full bridge axial load cell segments I had to get crafty with the linear rail mount placement and geometry. The decision to go with two full bridge axial load cell strain gauge arrangements came about due to the loads seen during runout conditions of the load cell. When two full bridges are implemented their respective signals compliment one another during any off axis deflection under load which allowed for a linear force measurement that was immune to any force ripple that occurred during the rotation of the ball screw. To amplify this load cells analog signal we developed a small amplifier board that cantilevered off the carriage as to not take any loads during flex. This board provided gain and offset trim pots that could be used to calibrate the sensor then be potted to ensure no more adjustments were made. For a sense of scale the ball screw shaft is about 8mm in diameter.
Stiff Knees
This torque sensor similarly needed to be a drop in replacement component. As it had a longer moment arm to work with I decided that a bending beam load cell with full bridge foil strain gauge setup would work nicely. The joint pivots about the pin joint near top center and is actuated via a link seen on the right. In certain joint angles the axial loads into this segment are actually quite high so the full bridge arrangement was crucial to get just the bending beam torque acting on the joint. Again, a small signal conditioning board was implemented in an adjacent piece. This joint, while overall simpler in structure and function, was rather difficult to design to handle the loads needed while providing clean torque signals with the right resolution. In the end the Ti-4Al-6V component worked great. The main challenge during manufacturing, as always seems to be with Ti was the tight tolerance bearing bores.