| Wafer Handler End-point Position Analysis |
The Problem
A customer noticed that one of many robotic arms occasionally failed to correctly grasp a silicon wafer after a programmed move.
The Measurements
To identify and rule out any differences between what the servo system believed to be the correct position and the actual arm position, a laser distance sensor was employed to monitor the end of the arm. This particular sensor produced an analog voltage that was proportional to distance. Connecting both the encoder of the servomotor and the output of the laser to the SensorPlot™, the customer programmed the arm to repeatedly move horizontally back and forth 7.000 inches.
After a few runs the customer immediately saw that the robot arm was resonating with a peak value of 0.068" at the end of the move. As indicated by the graph above, it took the end of the arm nearly a full second to settle.
Comparing the position as measured by the encoder to the position as measured by the laser, indicated that the resonance was not a servo loop-tuning problem. Instead, the resonance was caused by the mechanics of the arm (couplings, mountings, belts...). Subsequently, the problem required altering the velocity profile and adding mechanical damping to solve the problem.
Comments
Simultaneous sampling of position and analog voltage ensured that the positions measured by the encoder and laser were time and position correlated
In this application, the customer chose to use a low-cost laser triangulation sensor to measure the distance at the end of the robotic arm. A more accurate and much more expensive non-contact displacement sensor (not needed in this case) is a Laser Interferometer. It provides a digital quadrature signal output similar to that of an encoder, and can be directly connected to the SensorPlot™.
The SensorPlot™ can tap onto the encoder feedback signals of a servo system without affecting the controller.
The SensorPlot™ identified and assisted in solving the customer's problem in less time than it would have taken using the blind "trial and error" method of modifying accelerations, wait times, and PID loop parameters.
Euclid Research · 2 North 1st Street, 6th Floor · San Jose, CA 95113-1201
Tel: (408) 283-9020 · Fax: (408) 283-9029