MicroE Systems Natick, Mass.
Typical Scara arm configuration
The semiconductor industry is beginning to realize it has a good opportunity to drive down its equipment costs. The opportunity concerns the hundreds of Selective Compliant Assembly Robot Arms (Scaras) found in scores of atmospheric and vacuum wafer-processing applications. Cost savings come from new direct-drive motor designs which make it possible to replace motorgearboxencoder loops. The benefits are improved robot performance accompanied by lower cost and a robot actuator that is smaller and lighter than its predecessors.
However, direct-drive designs need positionfeedback devices that have special qualities. Most encoders used in ordinary motion feedback loops have drawbacks that become apparent in semicon fabs. But a new generation of absolute encoders have been built specifically for wafer-handling robot applications. These devices overcome many of the difficulties inherent in conventional position encoders.
It is useful to briefly review how older Scara robots were configured. Nearly all of them had gearboxes which multiplied motor torque.
But the gearbox is expensive, bulky, and introduces its own mechanical errors that affect position accuracy. Backlash and dynamic stiffness, for example, can both be problematic. In addition there are maintenance and reliability issues. All in all, there are a number of incentives to eliminate this component from next-generation Scara robot designs.
Elimination of the gearbox will improve robot performance by reducing the size and weight of the drive mechanism for each joint. This more-compact configuration has several benefits: Shorter driveshafts make motion assemblies potentially stiffer and lighter weight. Better position accuracy and higher acceleration is the result.
The downside, however, is that encoders for position feedback in direct drive need substantially higher resolution.
Many older Scara robots also used incremental encoders to provide position feedback. They gauge distance in terms of increments from a single reference point or index mark on the encoder wheel. But there is a problem with this scheme in the event of a power failure. In the absence of battery or power backup, the system loses count of the joint position. It has to go to a home position on powerup to get its bearings.
Such a procedure is impractical if the robot happens to be holding a wafer. In the worst case, a fully processed wafer could be part way in a cassette when the lights go out. The robot risks dropping the wafer on the floor of a load dock if it makes any sudden moves to find itself once power returns. With some wafers representing hundreds of thousands of dollars worth of in-process ICs, the perils are simply unacceptable.
The way around such a dilemma is through an absolute encoder. Absolute encoders, of course, let the robot know exactly where it is upon recovery from a power failure. Each point on the encoder scale has its own unique signature, so the system instantly can read encoder position at start-up.
There are a variety of ways to realize absolute encoding. One alternative is to use a battery-backed incremental encoder-with supplemental memory. With this strategy,-the encoder never forgets its position and never loses power.
The problem with this approach for Scara applications is that each individual joint needs its own battery and memory. This can be a maintenance nightmare for the typical fab which would entail monitoring and regularly replacing hundreds of batteries. For these reasons, SEMI has called for elimination of batteries from next-generation fabs.
Pseudo-absolute encoders are another way of solving the problem. They use multiple distancecoded reference marks that let the system resolve position when the robot reenergizes. The procedure is to jog each joint slightly to find the nearest reference mark. But this "bumping" approach is no longer acceptable because it can damage wafers if the robot happens to be handling one at the time.
The best way to address this issue is with absolute encoders capable of retaining joint position without having to find an index or reference point.
Such devices are indeed available, but until recently they have lacked the resolution needed for wafer-handling robots. They have also been bulky and relatively expensive.
These were among the reasons for the development of absolute encoders specifically targeting Scara wafer-handling-robots. Like other true absolute encoders, they have enough positioning information on any point on the code wheel to know exactly where they are on start-up. Single-turn resolution is 24 bits, yielding a rotational positioning accuracy of one part in 16 million. Thus they can resolve shaft position to better than ±2.1 arc-sec, more than sufficient to track joint movement in wafer-handling robots.
Equally important, these encoders are dramatically smaller than other devices providing similar specs thanks to a proprietary data-track design. The high density of information on the code wheel makes it possible to be physically small. The encoders are just 3 to 5.5-in. diameter, and contain a large throughhole as needed for Scara robot designs.
Installed height is also important because it affects the overall length of the joint. A thin encoder body helps minimize the length of the driveshaft and thus promotes a lighter, stiffer, and more accurate joint. New designs keep physical dimensions down through use of a kit approach. The code wheel and electronics mount individually. Moreover, electronic components are distributed around the circumference of the motor shaft in the space just above the code wheel. Such a layout takes advantage of free space that would otherwise be wasted inside the joint.
In contrast, the electronics in most ordinary absolute encoders stack vertically. The height of the device is often set by the space requirements of the circuitry. This is why conventional absolute encoders take up about 30 mm of space compared to under 15 mm for the new version.
The economics of the new generation of encoders is worth mentioning. They are more expensive than low-resolution, incremental encoders found in current Scara designs, but the total system cost of the robot is considerably less. Cost savings come from eliminating the gearbox as well as the physically smaller size of the overall mechanics. Estimates are that the overall system cost typically drops by one-third thanks to the direct-drive approach made possible by the new encoding technology.
MicroE Systems Div., GSI Lumonics,