Systematically testing optical MEMS speeds production

Thomas A. Cellucci

A system approach to optical-MEMS testing can ensure data integrity from R&D through high-volume production, while meeting time-to-volume demands. Automated test solutions currently exist at the wafer/die level, with final test systems on the horizon.

Microelectromechanical system (MEMS) technology uses semiconductor-like fabrication techniques to generate integrated systems on a submicron scale that are capable of performing electronic, mechanical, and optical functions better than conventional components. Optical MEMS are deployed in telecommunications components and subsystems such as all-optical switches, optical crossconnects, and variable optical attenuators. The compound annual growth rate (CAGR) of the optical-MEMS market from 1996 through 2003 is conservatively estimated at 50% or more.¹

As companies engage in the development of optical MEMS, clearly there must be a pathway to effectively and efficiently test these devices at the wafer, die, and final product (or packaged) levels. As much as 60% to 75% of the unit cost of a MEMS device is represented by its packaging and testing. Reducing the cost of testing and increasing product throughput, therefore, are essential elements in successful commercialization. Companies that fully understand their test and packaging requirements and continually reduce associated costs will be able to meet time-to-volume demands and to enhance profitability.

SYSTEMS APPROACH TO TESTING
Both conventional wisdoM² and experience gained from successes and failures in the inertial and pressure MEMS market segments point to five major attributes of a successful MEMS company:

Understand customer's needs and the impact of those needs on the MEMS device and the design and deployment of testing solutions.
Use a system approach, in that the MEMS actuator or sensing structure, electronics, package, assembly technology, and all aspects of testing are developed as a single system—an approach that optimizes overall performance characteristics including cost.
Follow a strategy of solving the maximum possible number of test problems related to packaging and assembly at the wafer and die levels.
Implement a higher level of automation for assembly and test.
Design and deploy an integrated test solution that offers a flexible and scalable platform with seamless continuity of data integrity from R&D, to pilot, to the high volume production phases of a product's manufacturing life cycle.

The concept of a systems approach to optical-MEMS testing is crucial because test is frequently an afterthought considered once a device design has been approved. The result is a higher cost-of-test, and lower throughput and yield. Indeed, "design-for-test," while it may require a more rigorous, systematic approach to optical-MEMS product development, usually results in lower cost-of-test and higher throughput—the two primary parameters in the successful launch and sustained profitable revenue of a MEMS product.

As the young optical-MEMS industry matures, downward pricing pressure will eventually occur, creating a commodity-like market. This trend has been observed in the more mature inertial and pressure MEMS market segments. In fact, price erosion could occur faster in the optica-MEMS market because of the large number of optical-MEMS designs and startup firms. The majority of these companies will not be competitive in a price-sensitive environment.

Part of developing a systematic approach to testing requires the development of tasks and procedures that will support a product from initial R&D through to volume production. Table 1 outlines the key test requirements as a function of the three phases contained in a typical product life cycle. During the R&D phase, testing must be interactive, flexible, and provide the basis for a pilot test system. During the pilot phase, testing must validate the product and the production process and predict production test costs. Finally, during production the testing system should be a scaled-up version of the pilot system that provides automated material handling with the capability of testing multiple devices simultaneously.

FIGURE 2. During testing of an optical-MEMS device, the in-plane frequency response plot (top) and the out-of-plane response plot (bottom) allows quality control personnel to determine whether a device is within specifications.

Systems assembled in a piecemeal fashion and run with pieced-together software are neither scalable or flexible, and may cost a firm competitive advantage. Interesting parallels can be seen in the early days of the semiconductor industry when manufacturers opted to design and build their own test and measurement equipment. The current success of the semiconductor equipment industry points to the utility of off-the-shelf tools that enable higher-volume testing with flexible and scalable integrated test solutions. These tools are capable of being fully automated, with testing capabilities at the R&D, pilot, and production levels (see Fig. 1).

WAFER AND DIE TESTING REQUIREMENTS
Fiber alignment and assembly procedures for optical MEMS are more complex, time-consuming, and expensive than inertial or pressure MEMS. This fact dictates the need to closely examine the testing that can be accomplished at the wafer/die level to avoid unnecessary and expensive alignment, assembly, packaging, and final test of unsuitable or suspect die. Therefore, go/no go tests are critical. This testing should incorporate optical parametric tests in addition to the mechanical and electrical tests required of optical MEMS. For example, a high-volume wafer/die production test equipment could have the following characteristics:

Automated wafer prober and wafer-handling robot to provide unattended cassette-to-cassette operation with an integrated wafer handler or film frame handler
Robust, optical-head assembly, configurable for a variety of specific test applications with the ability to provide motion measurements (such as real-time dynamic response to stimuli), three-dimensional (3-D) metrology, profilometry, and reflectivity (such as radiometry)
Precision electrical stimulus and measurement intrumentation with high-voltage drive capability
A user friendly, open-architecture software environment and operator interface
The ability to perform optical parametric testing (for example, polarization-dependent loss, optical power) on a routine basis with an easy-to-use beam delivery system and high-speed optical parametric test suite

Even in the case of an in-plane or two-dimensional (2-D) optical-MEMS device, 3-D metrology is recommended because many 2-D optical-MEMS devices exhibit out-of-plane motion. Whether the motion is intended or not, detection or observance of such phenomena can be important in determining device reliability and performance. Table 2 provides a list of suggested testing capabilities for an optical-MEMS device at the wafer or die level, while Figure 2 demonstrates typical output of an in-plane and out-of-plane frequency response for a optical-MEMS device.

FINAL PRODUCT TESTING
Unlike final package testing for inertial and pressure MEMS devices—which are characterized as low-product-mix/high-volume testing—optical-MEMS packaged products must be assembled, aligned, and tested to ensure they meet their performance specifications. In addition, optical-MEMS production is characterized by a higher product mix/ lower volume than what is observed in the inertial or pressure MEMS market segments.

At this time, the telecommunications industry is not using standard packages, nor are there many commercially available production tools that can align, assemble, package, and test optical MEMS. Optical-MEMS production will be cost-effective only when it is treated as an extension of field-proven production tools used at the wafer/die level to minimize the development of fully integrated and automated fiber alignment and package test system.

The heart of such a system begins with flexible and scalable automated test equipment and calibration/trim algorithms. The system engineering of the alignment assembly and packaging needs to occur in the proper sequence based on a fundamental systematic approach to volume testing. In other words, it is easier, from a systems engineering prospective, to add the optical subsystems to existing integrated and automated MEMS test systems than to develop the complex electrical and mechanical subsystems for a test system that is specific to optical MEMS (see Fig. 3).

MAKE VS. BUY
Many firms involved in producing optical MEMS will be faced with the dilemma of whether to make or buy a high-volume optical-MEMS test system. Rigorous models exist to calculate the cost-of-ownership (COO) of test equipment: ³

COO = CEO + CYL

where CEO is the cost of equipment and CYL is the cost of yield loss.

While developed primarily for semiconductor-wafer processing equipment, there are many concepts gained from such models in conducting cost calculations for optical MEMS test systems that are either "home-grown" or outsourced. It is especially important to make "apples to apples" comparisons when calculating fully burdened internal costs versus a purchase. In addition, consideration must be given to many other factors, including the impact on schedule, the documentation and training required, in-house experience in developing tools, scalability or potential reuse of the equipment, and maintenance and calibration costs.

REFERENCES

See, for example, Deutsche Bank Alex Brown, Optical Switching (April 2001), MEMS Update (January 2001), and MEMS (August 2000).
V. Voganov, Packaging and Testing Issues in MEMS Commercialization, SEMI Commercialization of Microsystems Conferences (1998).
See, for example, SEMI E 35.1-95, Guide for Cost of Equipment Ownership Comparison Metric, 1995, 1996 and SEMI E35-0299, Cost of Ownership for Semiconducter Manufacturing Equipment Metrics, 1995, 1996.

Thomas A. Cellucci is president and CEO of Etec, 83 Pine Street, Peabody, MA 01960. He can be reached at 978-535-7683 or [email protected].