Integrated packaging and testing of optical MEMS
04/01/2002
 Illustration by Gregor Bernard |
Working toward fully integrated solutions
BY JASON M. EICHENHOL
There has been significant demand for higher bandwidth from existing fiber optic networks during the past few years. Optical networks operating at 2.5 Gb/s just a few years ago are now planned to be replaced with 10 and 40 Gb/s systems today. These systems continue to evolve and are constantly pushing the envelope of current technology.
Optical microelectromechanical systems (MEMS) are regarded as one of the key enabling technologies of the "all-optical network." MOEMS (micro-optical-electromechanical systems, another term for optical MEMS) are currently being used for optical switches, cross-connects, add-drops and variable optical attenuators. Future uses of MOEMS include devices that can actively adjust all-optical network performance by changing wavelengths at will, as well as adjusting the total system dispersion and polarization.
New Packaging and Test Solutions
There are numerous MOEMS companies working on low port count switches, and each one is confident that their intellectual prop erty, performance, costs and process know-how will make them the market leader in this new optical networking revolution. The large amount of funding available before 2001 led to a rapid pace of commercialization of MOEMS, which, in turn, has generated a need for production-oriented MEMS/MOEMS test equipment.
The demand for reduced cost of test and high throughput follows similar trends that have accompanied widespread deployment of other MEMS applications, such as accelerometers and pressure sensors. As companies develop new MOEMS devices, it is clear that there must be a pathway to effectively and efficiently test these devices at the wafer, die and final product (or packaged) levels. Reducing the cost of test and increasing product throughput are essential elements in the successful commercialization of MOEMS.
Test and packaging com prises a major portion of the production cost of a typical MEMS device. It has been estimated that the costs of packaging and test of optical MEMS can approach up to 60 to 75 percent of the cost of the device. Successful commercialization of MEMS requires a systematic approach to the design.
There are distinct differences in the test requirements for MEMS and MOEMS devices as they transition through their product development life cycle. The three basic phases of product development and the associated test requirements are:
- Product research and development phase: Prove device works, and that it can be manufactured.
- Pilot production phase: Prove manufacturability in volume with high yields, develop production equipment solution, as well as testing needs for volume.
- Volume production phase: Maximize throughput and reduce costs.
The requirements for manufacturing and testing must be considered at the design phase to enable efficient and cost-effective fabrication and testing. Given the embryonic nature of MEMS and MOEMS technology, testing needs to be performed at the wafer, die and packaged device level in most cases. An integrated testing program allows more problems to be solved at the wafer and die levels, minimizing the cost of packaging bad die.
Currently Available Technology
MOEMS testing is currently dominated by manual and semi-automated processes. There is a lack of test equipment supplier infrastructure, and device manufacturers have had to develop much of their test equipment in-house. This is not unlike the early days of the semiconductor industry. A few companies have recently introduced equipment for semi-automated assembly and packaging of MEMS, but these product offerings typically consist of taking their existing R&D or lab-based equipment, placing it in "black box" and putting it on the factory floor. Despite the fact that these companies have acquired some substantial industrial-based firms, they are performing mainly systems integration, as opposed to systems engineering that balances the various elements of a system for production. Simple "rack and stack" system integration may be adequate for introducing a product to market for today's low volume production needs, but it will not necessarily enable the needed cost structure for wide-scale deployment.
Given the rather specialized customer requirements that exist, each packaging and test system requires some customization. For a proper systems engineering approach to be effective, the engineering team must have the ability to run simulations that can compare and calculate parameters, such as throughput, test parameters and overall system performance. The end result is an optimal solution that meets a given customer's needs in the short term, and at the same time, provide expandability in throughput as volume increases.
The most efficient and cost-effective way to test, measure, assemble and package optical components is through the use of a set of flexible and scalable "building blocks" that were specifically designed for the manufacturing floor. The current state of packaging in the optics industry might be best characterized as random, with each manufacturer using different packaging schemes. There are no industry standard packages for MEMS and MOEMS at this time, and many packages require customized material handling. Customers are reluctant to discuss specifics of their device and especially packaging, so packaging and test have inadvertently become major factors in overall price competitiveness. Manufacturers that continue to neglect the critical aspects of testing their MEMS/MOEMS devices will find it tough to sustain a competitive advantage. To reduce the costs associated with packaging and test, a new approach to assembling and packaging devices must be used.
MEMS Testing Requirements
The inherent advantage of MEMS technology is that it uses semiconductor-like fabrication techniques to generate miniature integrated systems capable of performing optical, electronic and mechanical functions better than conventional components. It is exactly these advantages and likenesses to the semiconductor industry that MEMS designers and manufacturers must leverage. MEMS-based optical component test requirements present unique challenges, including non-standard packaging, customized material handling, and difficult measurement needs (such as wavelength-dependent, polarization-dependent loss).
A viable production solution will incorporate a known good die (KGD) solution that can be integrated with fiber alignment and packaging technologies in a flexible system configuration. At this stage in MOEMS development, it is safe to assume that 100 percent of die will require testing before packaging.
Optical stimulus and measurement instruments must be fully integrated into an automated test system. Optical parametric measurements (e.g., optical spectrum, insertion loss, etc.) may need to be carried out on wafers, die or packaged devices. The approach to wafer-level testing depends of the type of MEMS device. MEMS-based tunable vertical cavity surface emitting lasers (VCSELs), for example, allow for immediate testing of optical parametrics at the wafer level, while other devices, such as simple 2-D switches, require presentation of the fiber to the wafer, making optical parametric testing at the wafer level difficult, time consuming and expensive. The latter is a better candidate for vision-based testing systems.
Fiber Alignment
Once KGD are delivered to packaging, one of the greatest challenges is the alignment and attachment of an optical fiber or an array of fibers to the device. This is a critical step in the packaging process. Several companies make motion stages for nano-positioning of fibers; these stages must have motion accuracy and resolution on the order of 50 nm or better. Many automation vendors directly incorporate either these commercially available stages or their own proprietary stages into their optical alignment solution. However, these are often stages that were designed for laboratory use, rather than production manufacturing applications. There is a need for fiber alignment stages that have built-in force sensors to "feel" when the fiber or MEMS device is grasped, as well as when the fiber has been placed into the proper position. The forward-looking MEMS designer will most likely incorporate V-grooves or alignment rails into a MEMS design, and also incorporate passive structures to grip and hold the fibers in place after the alignment is completed. None of the automation manufacturers have reliable solutions available that actively use and take advantage of these advances.
Most major manufacturers use an assembly floor of technicians who manually adjust the fiber into the die, making many unconscious, and therefore, non-repeatable steps while assembling the product. Each time a technician must handle the device, the fiber (or fragile MEMS die) can be damaged. A simple process, like vacuum pickup of a device, is not sufficient for MEMS packaging operations. Each MEMS device is different and the fixturing of the device is a major area of complexity that is often overlooked.
There is a multitude of potential, unwanted scenarios obtained even with KGD and perfect manual alignment because of the process being operator-dependent, severely affecting yield and throughput. Typically, the operator is adjusting manual positioners throughout the alignment process while looking at the transmitted power through the device to minimize insertion loss. The operator's performance and evaluation may be dependent on a given throughput of parts per shift. The operator is, therefore, encouraged to push through as many parts as possible that meet minimum specifications but might not perform at optimum performance. The use of a computer-controlled automated test and alignment system can be used to maximize the performance of each component assembled.
Systems for Testing MOEMS Devices
MEMS systems engineering equipment suppliers are now introducing a new generation of MOEMS assembly, testing and characterization systems. These systems are designed and deployed as part of an integrated test solution that offers a flexible and scalable platform with continuity of data integrity throughout all phases of a given MOEMS' product life cycle. Solutions for MOEMS testing have already been designed and implemented that can assist in the packaging and test of devices at all three levels of product maturity: R&D, pilot and high-volume production. There are systems that use stroboscopic illumination to measure in-plane and out-of-plane dynamic motion in the X, Y and Z directions, as well as angular displacement (six degrees of freedom). This is important because it allows a user to measure the motion characteristics of a device dynamically. While static measurements are not without merit, it is important to be able to measure the following characteristics under realistic operating parameters: device sensitivity or gain (measured in degrees/volt, microns/volt, or amps), frequency response, mechanical resonance and bandwidth, transient response, settling time, overshoot and ringing, motion repeatability, motion trajectory, relative motion and device-to-device cross-talk. Results from these tests are often compared directly to expected results predicted by the CAD tools used to design the MEMS.
Other systems offer the same MEMS motion capabilities described above, but can be designed for either pilot runs (semi-automated testing) or full volume production (fully automated testing). These systems offer solutions to MEMS manufacturers who wish to be able to deliver KGD to their packaging processes.
Integrated and Automated Packaging and Test
While typical automated test and packaging processes might perform optical power measurements to guide fiber alignment and minimize the insertion loss, these simple optical parametric tests do not go far enough. In a standard automated alignment system, after the insertion loss is adjusted to less than the minimum acceptable level, the fiber alignment is completed and the fibers in the device under test are attached using either epoxy or laser welding. The device is now hermetically sealed and sent to final test before shipping. What if the device passes the insertion loss test, but other parameters, such as return loss or crosstalk, are at an unacceptable level? A better solution is to detect these problems earlier in the process so they can be fixed.
The next generation of automated packaging and test systems will allow for a full suite of optical parametric tests prior to making the ultimate decision to attach the fibers. These systems will measure the devices under test for parameters such as optical power, output spectrum, spectral linewidth, insertion loss, adjacent and non-adjacent channel cross talk, channel accuracy, dispersion and polarization-dependent loss. While each device might not need all of the tests discussed, it is important that these new systems are capable of making "go/no-go" decisions as soon as possible in the packaging process. It is inexpensive to repeat the fiber alignment (and even replace the fiber if necessary) before the final attachment of the fibers but impossible afterward. Comprehensive testing before final packaging will significantly increase the yield of the packaging process, and if imple mented, via a complete systems engineering methodology, will have little direct effect on overall throughput. These complete systems will also include smart functionality, which will automatically or manually regulate tests as needed. If yields are consistently high, the engineers are able to further increase the system throughput. Conversely, when the inevitable yield issues appear, the system allows the process to slow down to get an "all tests on" picture of where the problem is arising in the system. This can save a quality control or process engineer's time by eliminating the need to perform offline tests to determine where the cause of a problem is.
Conclusion
The optical MEMS market needs to become more mature as it moves out of development and into deployment. Because packaging and test are the largest contributors to overall device cost, new methods to lower these costs need to be cost-effectively implemented for MEMS to be cost-competitive with other technologies. To minimize the cost-of-test and increase throughput and yield, it is crucial to adopt an interdisciplinary systems engineering approach to MEMS/MOEMS packaging and testing. Commercially available automated test solutions currently exist at the wafer and die levels, delivering KGD to be packaged, with final test solutions on the near horizon. Lower overall systems costs (with increased per-part profits) will only be realized by merging the typically discrete processes of packaging and test into fully automated and integrated packaging and test systems.
AP
Acknowledgements
The author would like to acknowledge the contributions of Mr. Arthur Holzknecht and Dr. Thomas Cellucci for their support and assistance in the writing of this article.
For more information, please contact Henry Klim, Etec Inc., 83 Pine Street, Peabody, MA 01960; 978-535-7683; Fax: 978-535-7003; [email protected].
null