Speed Acquisition Methods for High-Bandwidth Servo Drives

• Main Author: Bähr, Alexander
• Format: Others
• Language: German; en
• Published: 2005
• Online Access: http://tuprints.ulb.tu-darmstadt.de/553/1/baehr_alexander_2.pdf; Bähr, Alexander <http://tuprints.ulb.tu-darmstadt.de/view/person/B=E4hr=3AAlexander=3A=3A.html> : Speed Acquisition Methods for High-Bandwidth Servo Drives. [Online-Edition] Technische Universität, Darmstadt [Ph.D. Thesis], (2005)

A servo control needs the actual values of speed and position.Usually, the latter is computed from the signals of a position encoder; its 1st derivative is smoothed by a low-pass filter and used as actual speed signal. A number of enhanced and alternative methods is experimentally investigated in this thesis. Based on an equal steady-state behavior, the controlled servo's dynamic stiffness is used as the performance measure. The used setup has a special feature: because of its rather high resonant frequencies (870 and 1280Hz), the encoder's oscillation against the drive can no longer be neglected. The mechanical resonance can be met by using notch filters to damp the resonant frequencies out of the controller spectrum, leading to major improvements. By identifying and modeling the mechanical setup at different levels of precision, observers were designed to provide an alternative actual speed signal, leading to a further improvement; however, active damping was not possible due to the configuration of the resonant system. The use of a state controller allowed active damping, but at the expense of reducing control gain and thus dynamic stiffness. The signals of an optical position encoder show characteristic errors. Using measures to correct those errors, it was tried to improve steady-state speed quality and allow a higher control gain. Two table-based and one on-line adaptive method were investigated. As stated in previous works, the correction of signal records resulted in a considerable error reduction with all methods. However, the improvement due to correction used in the control loop is small, because the loop gain is quite low at the error signals' high frequencies. The use of an acceleration sensor for speed acquisition has the advantage that the signal is integrated instead of derived, reducing noise instead of amplifying it. The improvement in the experiments was only low, because oscillation and not noise is the problem limiting control gain. Another advantage of the acceleration sensor is a much easier fixing compared to the position encoder. By mounting the acceleration sensor at an optimal location concerning oscillation, it is possible to damp the oscillation considerably without any knowledge about the resonant frequencies. The thesis is completed by theoretical investigations of speed quality and dynamic stiffness, investigations of drive-side and load-side behavior and necessary computation power for the investigated algorithms.