Fixture-Based Industrial Robot Calibration for Silicon-Wafer Handling

Mike Tao Zhang and Ken Goldberg 
<August, 2001 ~ April, 2002> 
ALPHA Lab, University of California, Berkeley. 
 


Abstract:

Semiconductor manufacturing industry requires highly accurate robot operation with short downtime. We develop a fast, low cost and easy-to-operate calibration system for wafer-handling robots. The system is defined by a fixture and a simple compensation algorithm. Given robot repeatability, endeffector uncertainties, and the tolerance requirements of wafer placement points, we derive fixture design and placement specifications based on worst-case and statistical tolerance models.  We verify our resultant design by physical experiments in a factory-floor environment. 

 

Introduction:

During front-end semiconductor manufacturing, silicon wafers are transported in cassettes via Automated Guided Vehicles (AGVs) or Over Hoist Vehicles (OHVs). The wafer-handling robot picks up the wafer from the cassette, and then hands it to a pre-aligner. After aligned, the wafer is transferred to IC manufacturing equipments by the robot.

Wafer-handling robots generally have high repeatability, but their absolute position accuracy is rather low due to relative position/orientation errors. Currently, the wafer-handling robot is most often programmed by playback method using teach pendants. This procedure requires human operators to teach the robot all the wafer placement points. If we have to change the endeffector due to damage or other reasons, this manual teach has to be performed once again completely. Therefore, this procedure is costly, inaccurate, and time-consuming.  To avoid the manual teach, robot calibration is often applied to increases the robot accuracy via proper software compensation. It has considerable impact on the reliability, throughput, and total cost of the overall manufacturing system.

A market survey reveals that the leading [robot calibration] performers are characteristically easy to set-up, operate and, most important, more economical. This is especially true for semiconductor manufacturing industry. In such a dynamic changing environment, to reduce setup time for changeover makes significant contribution to overall cost saving.

We propose tooling concepts to eliminate manual re-teach during robot/component field replacement. As illustrated in Figure 1, a standard fixture is attached to the base on which the robot sits. Initially, the operator manually teaches the robot all the wafer placement points. The fixture will help to accurately locate one critical placement point. If the endeffector changes, the operator does not have to re-teach all the placement points again. He only needs to re-teach the critical placement point defined by this fixture. Then, the robot automatically generates the other placement points based upon the offset between the initial teach and the re-teach data of the critical placement point. The robot is insured to reach these placement points within the tolerance requirements. In this paper, we describe the mathematical and conceptual design and placement rules of the fixture to guarantee the tolerance requirements are satisfied.
 

Publications:

  • Mike Tao Zhang, Dimitri Kambouridis, Rodney Lum, Tom Wahl, Paul Larskulsint, Greg Hirth, Brian Carlisle, and Ken Goldberg. "Fixture-based industrial robot calibration for silicon wafer handling," in IEEE International Conference on Robotics and Automation, Taipei, Taiwan, China, 2003 (in preparation). [PDF]

Acknowledge:

Collaboration between ALPHA Lab and Adept Technology, Inc. makes this work successful. I would like to thank Brian Carlisle (Chairman and CEO) and Greg Hirth (Director, Semiconductor Equipment Division) for enlightening discussions. I also would like to thank engineers Dimitri Kambouridis, Rodney Lum, and Tom Wahl for help in modeling and implementation.