Proprietary systems have traditionally dominated the control of three and four-axis wafer handling robots for in-vacuum and in-air wafer applications. As a result, these systems can be limited in their capability, difficult to support, and dedicated to the specific robot being used.
Certainly, the control challenge is significant in these applications. Current robots from Yaskawa, Genmark, Hirata, and Brooks Automation are designed to perform the task of picking and placing 300 mm wafers repeatedly in a sequential fashion. The robot axes are taught positions in which to pick and place wafers based on the joint location of R-Theta and Sacra style robots. A typical sequence of moves found in the ion-implanting process from a robot perspective would be:
Extend to wafer pick location
Retract to safe rotate position
Rotate wafer to place angle
Extend robot arm to place location
Retract and Repeat
Each move is generally separated by a brief dwell to ensure that the robot has reached the position it has been commanded to go to. As a result, wafer-handling robots perform their tasks in a series of moves involving quick accelerations and decelerations. To increase machine throughput, robot manufacturers have focused on decreasing the time required for each individual step to meet production needs. This means that expensive wafers are being exposed to high accelerations and velocities.
To solve this problem, robot manufactures are open to creating custom “macros” to blend several steps together. However, it generally takes several months for a “custom macro” to be created and implemented within the framework of a proprietary control system.As wafer-handling machines continue to emphasize machine throughput to meet the demand for faster chip production, this process can become a bottleneck in the long wafer production process.
Developing an Alternative
Along with its customer, Axcelis Technologies, a Massachusetts-based semiconductor equipment supplier, Axis New England embarked on a project to address these challenges. We turned to a standard, commercially available control system, Delta Tau’s Turbo PMACII Ultra light controller. Configured with a fiber optic MACRO ring, the controller provides a 32-axis control capability.
Built-in forward and inverse kinematics made this controller an ideal choice to handle the various mechatronic designs of the robots axes. In less than six months of development time, the design team was able to create and prove out the kinematic equations for four different robots.
Once these equations were implemented in the controls, programming became very straightforward. All motion is programmed from a center of wafer perspective in an X-Y coordinate system. This approach lets you create without stopping smooth single-motion move trajectories, such as pick-to-place.
With Delta Tau’s automatic inverse kinematic calculation capabilities, we were able to achieve constant velocity at the center of the wafer by varying the independent mechatronic joints of the robots. Linear and circular move types are assembled and blended to create seamless move profiles that eliminate the need for multiple accelerations and decelerations. The “macros” are limitless and can be developed in hours and days instead of months.
This system demonstrated its ability to increase throughput while at the same time reducing the velocity and acceleration rates that the wafer was subjected to. Further, different robots can be interchanged simply by loading in a new set of kinematics equations, while preserving the core software.The present architecture can control 14 axes of kinematics robots (two 4-axis robots, and two 3-axis robots), while also handling up to an additional 10 axes of motion.
What’s more, the fiber optic network allows Axcelis to use distributed hardware to reduce wiring, but maintains centralized control for tightly coupled motion. This configuration also eliminated the need to manage five separate controllers (four for the robots, and one for the remaining motion) with their associated cost and communication latencies.With all these benefits, the semiconductor industry appears poised to take advantage of this technical breakthrough.
Greg Ellrodt is a sensor systems engineer with Axis New England in Danvers, MA.