By Terrence E. Thompson, Editor-at-Large
Once upon a time, Rube Goldberg's cartoons showed incredibly elaborate and over-engineered mechanisms designed to perform seemingly mundane-often simple-tasks. Ultimately, Rube's inventions became synonymous with any Herculean design effort destined to achieve dubious results. His remarkable work certainly contains its own exquisite logic, proving that the unnecessary can indeed be the mother of invention, often with hilarious results [www.rubegoldberg.com]. Just imagine what Rube might have done with CAD to design today's complex products. He, in fact, may unknowingly have been optoelectronic packaging's first role model, putting even Dilbert to shame. No Simple Answers What's to test? A very good question, but certainly one with no simple answers. At this point, some perspective is needed. Consider that the photonics community, a very talented physics-intensive group, has been inventing and producing everything from lasers, LEDs and detectors to exotic optical lenses and systems for decades, albeit in relatively low volumes. Much of the production has been custom and usually expensive. Some optical systems suggest a Goldbergian thought pattern when first viewed by those outside the optical community. Now, however, the intense focus of interest in "electronics" is shifting from the exotic, virtually hand-made photonic/ electronic interface products to higher-volume, mass-produced ICs packaged with photonic devices (a.k.a., optoelectronics). And the reason is simple: The need to extend long-haul fiberoptics bandwidth that final mile is becoming imperative for revenue generation. Many of the hot, new products and services expected to breathe new life the global economy will need more bandwidth, where people need and want it. Therein lies another challenge: after you build them, how will you test these complex critters? What If? Assume that your company (or a customer) wants to add some optoelectronics to a next-generation module. This module will make installing and interconnecting fiberoptics to electronics economical, whether it's in the last few meters to the office or in the home. This module, in the form of a hybrid package, might contain very high performance electronics. It might also provide the requisite photonic input/output (I/O) interconnections, not to mention requisite electrical connections for power/ground and I/Os.
We're no longer just talking about silicon, since its problematic compound semiconductor cousins are joining the party. It is a given that packaging these diverse components will be challenging for some time to come in many applications. How you test these components may not be what you want to hear. In all likelihood, you will have to build your own test system, a frightening prospect at first glance, conjuring up visions of Goldberg again. Optoelectronic designers are reaching for every trick in the function book. Why be limited to microelectronics when we can add mechanical and optical functionality too? Speaking of test challenges, test expert Masood Alavi hit the nail on the head when he wrote that today's IC devices are typically divided by functionality into digital, analog, mixed signal, RF/microwave and other functions. Each category is further segmented into devices, such as processors, logic arrays, volatile and non-volatile memory, DACs (digital to analog converters), gigabit data transceivers, audio receivers, etc.(1) So if testing the silicon ICs weren't challenging enough, adding photonics and microelectromechanical systems (MEMS) will make it more complex. Get used to MEMS, too, since this fast-growing component market is widely used for optical switching and other photonic and/ or RF applications. So What's to Test? There are many-typically small-vendors that supply very specific test and measurement equipment to make the testing job easier-or at least possible.
First, let's try to define what is actually needed and who tests what. Many optoelectronic components are being built in hundreds of categories, but how are they being tested? Of special interest are the packages that remind us of the military hybrids of yesteryear. How are they being tested? The Requirements, Please The odds are that you will be producing a module containing differing technology-base components (electronic with photonic and/or physical MEMS/MOEMS). You will often see terminology leftover from yesteryear: e.g., hybrids, multichip modules (MCMs) and multichip packages (MCPs). There's a sensible reason for this, of course: Many of today's mainstream opto-electronic packages were developed with the old AT&T Bell Labs "40+ years of operation or else" thinking. The packages were then often built by those with considerable military or aerospace expertise. So, do not be surprised when you learn that many packages are, in fact, hermetic-ceramic and/or metal packages (butterfly comes to mind). There is a larger question that begs to be answered: Are hermetic packages really needed? This issue will be addressed further and resolved as the industry matures. Many believe that much lower-cost plastic (termed "near-hermetic" by some) packages are just fine, certainly good enough for the benign environments in which they will function and good enough for short product life cycles. How many will still be using their present computing or communication devices in 40 years (or in four years, for that matter)? Certainly, the answer is obvious! The End of Copper-Based Telephony Steve Anderson, president of Silicon Bandwidth, Milpitas, Calif., [siliconband width.com] agrees. As he told me recently, "We are primarily looking at datacom, not voice/telecom, applications for new high-capacity landline and switching applications. For voice, many have already switched and now use cellular telephones as their primary home or business lines. More are joining them every day, which suggests that the end of copper-based telephony is in sight. Massive data transfer, or datacom, when we want it, is what we need to address." Anderson adds that there is considerable long-haul fiber with large switching centers already in place. Now this fiber needs to connect to offices, homes and factories, or anywhere else, economically. A thousand-dollar fiber-optic modem is just too expensive for most of us. A Contract Manufacturing Perspective When an OEM selects a contract manufacturer to build optoelectronic modules, the manufacturability is evaluated as well as how the modules will be tested. According to Philip M. (Phil) Yates, CTO and vice-president of Advanced Assembly Technology at contract manufacturer Nextek, Inc., Madison, Ala., "There are no standard off-the-shelf solutions for testing optoelectronic modules now. "Every project requires something different. If you need to test an OC-768 transceiver module running at 40 Gbits/ second, you might need two racks of equipment from several vendors-one for transmit and the other for receiver verification. What do you use to generate 40 Gbits of meaningful data and how do you measure it?"
Often the customer will monitor your packaging techniques and then do all of the testing, notes Yates. "You need laser sources to test photonic detectors, digital oscilloscopes and bit error rate testers (BERTs) from sources such as Hewlett-Packard and Anritsu. "It's expensive," Yates added, "Two racks of equipment, each costing between $400,000 and $750,000 is not unusual for an optoelectronic transceiver."
Opto Solutions "There are some opto test solution companies out there such as Instrumentation Engineering Inc.'s Optical Testing Division, Merrimack, N.H., [http://www.ie-ate.com] that have supplied complete optical test solutions for years," Yates notes. IE provides component tests for couplers/ filters, VOAs, laser diodes including PINs and VCSELs, optical switches including MEMS and optical amplifiers for EDFA, as well as Raman and module test platforms for transceivers from 2.5 GHz to 40 GHz, MUX/DEMUX, optical add-drop MUX (OADM), optical amplifiers, gigabit Ethernet, optical switch matrices and routers. Heard of an optical bench? You will, since the base substrate in some optoelectronics is referred to as an optical bench (see sidebar on page 55.) This so-called optical bench is the miniature equivalent of a full-size workbench that has a physical arrangement of lasers, detectors, optical lenses and/or other specific functional components, some made by the user. Major test companies from the IC side are looking at this emerging test market, but relatively little exists at the moment. Technology Roadmap You also might want to take a close look at the ITRS-2001 (International Technology Roadmap for Semiconductors) and look at optoelectronic test challenges and goals [http://public.itrs.net]. The entire document and supporting information are available on the web gratis, including many specifics that you will want to review before embarking on an optoelectronic module project. Optical/Photonic Test Functions Some specific test functions, encountered when building optoelectronic modules, include mirror testing; optical testing; glass including configuration, lens sets as well as environmental testing, laser damage testing, MIL-SPEC testing, photometric (the measurement of light intensity); spectrometric (analysis of spectra, generally, the electromagnetic spectrum within the wavelength region extending from 40 nm to 1 mm). Will you need to test for all of these functions? Probably not but some of these tests will be needed.
Conclusions Conclusions on test strategies and specific "must have" equipment decisions are a bit premature. Suffice it to say that many companies around the world are in R&D, prototype or limited volume production with modules containing a mix of electronics with photonic and/or physical MEMS/MOEMS. No real package standards exist, and there are not enough guidelines on device performance (usually for proprietary reasons). Are photonic devices sophisticated from a design perspective? Certainly, but packaging and testing them need not be rocket science. It is back to basics. Identify critical functions and choose appropriate equipment to validate performance. Many companies are using contract manufacturers or IC packaging foundries to tweak module designs and manufacturability with the intent of establishing de facto standards when they introduce optoelectronic-based devices, modules and products in volume. Some day, semi-monolithic integrated optoelectronic devices, housed in simple non-hermetic plastic packages, may make some test issues of today irrelevant. Until then, pull out your test equipment catalogs, because it will still be mix and match for a quite a while. Moreover, while the evolution continues, there will be those few designs that will surely evoke images of Rube Goldberg's early drawings.
References 1. M. Alavi, "Package Final Test," in press for July 2002 issue of Chip Scale Review. 2. ITRS-2001 (International Technology Roadmap for Semiconductors), International SEMATECH, Austin, Texas.
|
Copyright © 2002