Sensor and actuator bus networks for better OEE
06/01/2000
Paul Grosshart, MKS Instruments, Methuen, Massachusetts
James Moyne, University of Michigan, Ann Arbor, Michigan
overview
The evolution of sensor and actuator bus networks in the semiconductor industry and their emerging adoption into 300mm process tools will bring a new level of overall equipment effectiveness in semiconductor manufacturing.
Average overall equipment effectiveness (OEE) for 0.25µm process tools is approximately 40%. Future industry goals include evolution to <0.15µm technology, a transition to 300mm wafers, and >60% average OEE, the latter a 50% improvement. Standardized sensor and actuator bus networks—attractive alternatives to conventional point-to-point intra-tool device communications and associated wire harnesses—could play a significant role in improving OEE to this level.
 Figure 1. The position of a standardized sensor bus in the hierarchy of semiconductor manufacturing automation. (Source: MiTeX Solutions) |
A sensor and actuator bus is a networking solution for distributed input and outputs that connects tool components such as mass flow controllers (MFCs), vacuum pumps, RF generators, capacitance manometers, pressure control valves, etc., to process module controllers (Fig. 1). The many advantages of this technology that help to improve OEE include reduced wiring and wire harness costs, improved system diagnostics and mean time to repair, support for intelligent devices, and support for distributed control.
For example, a sensor and actuator bus solution can provide information about whether or not a system component is correctly calibrated and performing within predefined operational parameters. Instrument-initiated alerts and alarms can be set to notify operators so they can correct a problem before it affects process results.
Numerous component-specific cases provide examples of the benefits of intelligent sensors and vacuum components. Consider, for example, the ability to monitor current in a MFC. If the device begins to draw current levels outside a user-defined band, it could be because the MFC is wearing out or clogging. An alarm can be programmed to notify the operator that the MFC should be inspected.
In another example, precision capacitance manometers used for direct pressure measurements are temperature-controlled to maximize accuracy and to minimize condensable-effluent deposition within the sensor cavity. However, tool users do not always wait until the necessary temperature is reached, which can lead to a major problem unrelated to calibration and measurement. Insufficient heating can result in buildup of condensates—by-products of CVD or etch processes. Over a bus network, sensor temperature can be measured with interlocks and error bands established so the system does not run the process until the proper operating temperature for the sensor is reached.
The standards
Both Semi and International Sematech have already recognized the semiconductor industry's need for an interoperable sensor and actuator network standard:
- International Sematech has provided recommendations to Semi, educated the industry about alternatives, fostered competition among alternatives, and provided a canonical device description for the industry (i.e., a taxonomy and methodology for definition and characterization of sensors and actuators in bus systems). Sematech conducted an in-depth study in 1992/93 that included a requirement analysis, requirements derivation, candidate solution evaluation, and final specification developments. Results of this study indicated that a Controller Area Network (CAN) protocol solution— a low-level protocol used extensively in industrial control networking—would be one of the network standards best suited for semiconductor manufacturing [1].
- Semi developed the comprehensive Semi Sensor-Actuator Network (SAN) standard framework [2] for sensor actuator networks to ensure interoperability of sensor bus systems and interchangeability of devices in semiconductor manufacturing applications (Fig. 2). This standard defines three levels of sensor system standardization:
- Specific device model (SDM) standards identify device models for specific devices, such as MFCs. SDMs define baseline structure and behavior that all compliant devices of that type must support.
- Common device model (CDM) standards define the minimum structure and behavior required for any device on a SAN-compliant network. The parameters and behavior of CDMs (as well as SDMs) are described in terms of software objects, attributes and behaviors required in order that a device be compliant.
- Network communication standards (NCSs) provide standardized communications over a sensor and actuator bus.
The Semi SAN framework supports a multiple-choice standard at this level, allowing a user to choose a standardized SAN communication protocol that best suits system requirements. Currently, Semi NCSs exist for DeviceNet, Smart Distributed Systems (SDS), LonWorks, Profibus, Seriplex, and Modbus/TCP over TCP/IP Ethernet.
 Figure 2. The suite of Semi sensor bus standards. (Source: MiTeX Solutions) |
DeviceNet [3], for example, is based on the CAN protocol that was developed by Bosch Corp. in the early 1980s for the European automotive market. (SDS is also CAN-based.) CAN was used to replace expensive wire harnesses on automobiles with low-cost network cable. The CAN protocol has proven fast and reliable in demanding applications, such as anti-lock brake and air bag systems. The yearly demand for CAN ICs is 10 million and should increase to 50 million by the end of 1999, which will ensure competitive prices. Because of the large and increasing investment in CAN protocol across many industries, cost leverage of CAN products will continue to increase.
The CAN protocol is ISO Open System Interconnect (OSI)-based, but only specifies a portion (see Fig. 3). CAN uses a unique, nondestructive bit-wise arbitration mechanism (i.e., a sending prioritization scheme whereby no information is lost); this CAN-specific feature allows resolution of collisions (i.e., situations where multiple devices try to send data simultaneously) without loss of throughput or re-sending of data by higher-priority nodes. This means that it is ideal for systems that require a high-speed deterministic response for some data transmissions, but also allows for lower-priority, lower-speed, nondeterministic transmissions on the same bus.
For example, an MFC and a sensor may both want to send messages over the network at the same time. With nondestructive bit-wise arbitration, the device with the higher priority transmits first and the other device immediately follows, with no risk of lost data packets.
Practical application
A sensor and actuator bus provides communication between devices as well as important device-level diagnostics not easily accessible or available through hardwired devices. It is an interface specification that defines a common set of behaviors and digital communications protocol to which smart and dumb sensors and actuators must conform if they are to remain and function within a well-defined distributed control environment. In other words, such a bus is an intelligent cable that can replace hundreds or thousands of individual wires conventionally used on a process tool. Because it replaces traditional analog wiring with a digital cable, each device is able to share valuable information with other devices and controllers on the system.
Recognition and early adoption of sensor and actuator bus networks has paralleled the emergence of 300mm wafer process tools. Many semiconductor tool suppliers observed how these buses were being used in other industries and started the process of customizing them to meet semiconductor manufacturing requirements (See Fig. 4 on p. 221). Initially, this involved more than 90 semiconductor manufacturers around the world spending hundreds of hours in committees and in the lab to design profiles for system components. Profiles are guidelines on certain parameters that devices must include to be interoperable.
As a result of this work, semiconductor manufacturers introduced beta 300mm wafer tools equipped with DeviceNet in 1998.
Five key groups have had a role in defining the open, sensor and actuator bus standard: the Open DeviceNet Vendor Association (ODVA), wafer manufacturers (AMD, Zilog, etc.), process tool suppliers (Applied Materials, Watkins-Johnson. etc.), component suppliers (Rockwell Automation/Allen-Bradley and MKS Instruments), and Semi.
 Figure 4. Building a Semi sensor bus system. (Source: MiTeX Solutions) |
A key requirement of all sensor and actuator network-enabling technologies is that they be open to users, integrators, and technology providers. The resulting market opportunities and competition foster technology improvements and lower costs, and avoid pitfalls such as single sourcing. DeviceNet technology, for example, has been transferred to ODVA, an open association which manages the DeviceNet specification and supports worldwide growth of DeviceNet. ODVA currently has more than 300 member companies (see www.odva.org), with more than 50% having semiconductor manufacturing interests. The rationale for giving proprietary technology to the public domain is that it needs to be supported by the entire industry to be successful. ODVA consists of 33 special interest groups (SIGs) that work on committees and in the lab on specific initiatives.
While there is no doubt that the migration to 300mm wafers will involve a relatively large capital expenditure on new equipment, leading tool suppliers predict that sensor and actuator bus networks will help reduce the initial cost by providing enhanced diagnostics, predictive maintenance, and in situ calibration.
A typical wafer fabrication line is down an estimated 10% for preventative maintenance and process yield issues. Due to the high cost of raw 300mm wafers—$1500/starting wafer and up to $250,000/processed wafer—the industry is targeting 99.995% equipment process yields. Reduction in the consumption of test wafers is being pursued as an additional cost saving. Digital communication on transducers will help wafer manufacturers reach these targets by increasing awareness of equipment readiness without the need for test runs and sample wafers.
Many tool suppliers are working together with component suppliers to address key requirements of the semiconductor industry. More than 80 process tool and device component suppliers are members of ODVA's semiconductor SIG. This SIG has been unique since inception; while most SIGs are product-specific—with the common thread being a motion controller, pneumatic valve, etc.—the semiconductor SIG is industry-specific. Issues being addressed include:
- increasing tool complexity and set-up time,
- addressing the fact that 40% of faults can be attributed to control system software,
- documenting that instrumentation is often inaccurately perceived at fault,
- improving OEE needs to exceed 60%, and
- interchanging different manufacturers' components without impacting system controllers.
The semiconductor SIG began by organizing vendor subgroups to develop common device profiles for industry-specific products, such as MFCs to manage process gas delivery. These profiles are based on SDM and CDM model structures in the Semi SAN standard. But their first-of-a-kind work did not stop with conventional device profiles. SIG members drafted an interface guideline, a SIG compliance test procedure, and have made arrangements with the University of Michigan to conduct additional SIG testing beyond ODVA's compliance test suite.
- Interface guidelines: The semiconductor industry DeviceNet interface guidelines are application notes that choose preferred options from the many selections made possible by DeviceNet specifications. They were written to convey a "semiconductor capital equipment manufacturing" point-of-view. The narrowing of choices is aimed at reducing options and taking commonality to a new level. For example, DeviceNet specifications define three ways for setting node addresses: software selectable over the network, dip switches, or rotary switches. Semiconductor Interface Guidelines require rotary switches. It is based on actual lessons learned during large-scale integration of DeviceNet products from multiple suppliers into 300mm tools.
- Test procedures: A companion document to the interface guidelines is the interface guidelines conformance test procedures. The purpose of this document is to provide a comprehensive test procedure to verify conformance to the interface guidelines, requiring a DeviceNet certificate of conformance from an independent test laboratory.
Conclusion
Semi continues to emphasize development and promotion of open standard solutions for the semiconductor industry. The Semi SAN standard represents an important success story for this organization. Through unselfish cooperation among device suppliers, network solution providers, and users within the Semi organization, a group of industry volunteers has created a framework of standards that provide interoperable and interchangeable open sensor network solutions. This framework has helped foster further interoperability specification development in sensor bus technology organizations such as ODVA, resulting in open and standardized solutions for semiconductor tools that not only address today's OEE improvement needs, but provide a continuous improvement path to sensor bus solutions in future tools.
As capital equipment is delivered based on DeviceNet, end users will use this interface to improve their manufacturing processes. Employing a sensor and actuator bus enables new technologies to be explored effectively. For example, end users in semiconductor manufacturing will have an avenue to integrate new metrology tools with process tools to improve performance.
References
- J. Moyne, et al., "Analysis of Sensor/Actuator Bus Interoperability Standard Alternative for Semiconductor Manufacturing," Sensors Expo '94, Cleveland, Sept. 1994.
- Book of Semi Standards: Equipment Automation and Software, Vols. 1 and 2, Semi E54-0997.
- DeviceNet was designed by Allen-Bradley (now part of Rockwell Automation) in 1994 to serve industrial applications in food processing, materials handling, oil and gas, and automotive industries.
Paul Grosshart received his BSEE from the University of Connecticut. He has worked for Perkin-Elmer and Lam Research, and is currently engineering manager of the Process Controls and Instrumentation Group at MKS Instruments, 651 Lowell St., Methuen, MA; ph 978/682-4567, fax 978/682-4956, e-mail [email protected].
James Moyne received his BSEE, BSE-math, MSEE and PhD from the University of Michigan. He is president of MiTeX Solutions Inc., a supplier of control solutions for semiconductor manufacturing. Moyne is also an assistant research scientist and head of the DeviceNet and ControlNet conformance laboratories at the University of Michigan, 2360 Bonnisteel Blvd, Ann Arbor, MI; ph 734/936-3645, fax 734/936-0347, e-mail [email protected].