Introduction

Linear motor three-axis gantry used for precision optical inspection— Y0, Y1, and X. The linear motors in this gantry application are using glass-scale sinusoidal encoders that allow for a very high level of accuracy and speed in point-to-point moves. In conjunction with an MEI XMP-SynqNet^® motion controller, typical performance consists of move and settle at micrometer resolutions and settling times of less than 100 µsec with well tuned motors. Quick settling times in point-to-point moves helps increase throughput and efficiency.

Problem

The load attached to the linear motor (X axis) was oscillating during fast moves. The load in this case was the camera and associated equipment that impacted the optimal operation of the machine and subsequently overall throughput. It's important to understand the system was well tuned, but accelerometer readings at the camera indicated adverse vibration. Motor feedback indicated optimal settling time.

Due to the extreme machine motion requirements, further mechanical modification of the interface between camera and mount yielded no better performance due to the natural frequency response of the entire machine under normal operation.

Solution

To compensate for the unwanted vibration on the X-axis, the control schemes were modified, along with the H-bridge gantry control, Y0 and Y1. The entire custom MIMO (multiple-input multiple-output) model with the addition of state-observers and proprietary control schemes was developed with MechaWare™ to obtain optimal vibration control to achieve the required settling times and position accuracy.

This image shows a basic gantry positioning with two Y axes.

No mechanical changes were necessary; the vibration control was performed in software only. In essence, the control algorithms to the gantry were modified to greatly minimize the vibration of the camera and achieve the necessary mechanical requirements without further mechanical cost in modification and/or retooling. The solution required the motion of the linear X axis to behave in a non-ideal way to allow the load to behave predictably. The diagram shows controlled motor oscillations in the move to achieve optimal vibration control of the camera. The accelerometer indicated optimal settling, while the motor feedback indicated a dampening effect.

The usual challenge in this type of application is developing an accurate and workable control model, but more so avoiding a performance decrease based on additional computational overhead each servo cycle. The XMP-SynqNet motion control family employs a dedicated DSP onboard so the additional algorithms could be implemented while maintaining an 8 kHz servo update rate in this case.

MechaWare is a plug-in package that works within the Matlab^® /Simulink^® environment and consists of predefined control "blocks" that can be used to quickly develop sophisticated control schemes ranging from PID with feed forwards to complex vibration cancellation methods like the one explained here. The subsequent models can be used in simulation mode or used in realtime with existing mechanics.

Top: Motor settling response vs. camera settling without MechaWare. Bottom: Camera settling with MechaWare.