Error Compensation via Command Signal Deconvolution

A signal processing technique has been developed to dramatically increase both the usable bandwidth and tracking accuracy of an actuator. All actuators exhibit varying amplitude and phase response to multiple frequency input signals. This makes it difficult to utilize their full small-signal bandwidth for high precision motion. For example, even a small phase lag in an axis of a turning machine may result in significant form errors due to the high relative velocity of the spinning workpiece and the cutting tool. The goal of this research is to use an understanding of the dynamics of an actuator to create an input command compensation technique to correct for anticipated motion errors. Deconvolution is a mathematical procedure that can produce a modified input to reverse the effects of attenuation and phase if the impulse response of the dynamic system is known. Since the modified signal is not calculated from the position feedback information, there is no delay in the response. The algorithm differs from feedback control in that it is implemented off-line to make use of the entire command signal to produce a modified actuator input signal. When the actuator is driven with this modified signal its motion will exhibit attenuation and phase lag that convolves with its dynamic response to produce the original desired position trajectory. In this way the individual motions of the axes can be synchronized to produce the correct machined surface. This technique critically depends on knowledge of the actuator dynamics an