Implementing Planar Waveguide Technology.

The market for optical networking components is growing at a staggering rate. Some 4,000 miles of new glass fiber are deployed every day. Telecom operators, frantic in their need to increase network capacity, are clamoring to ensure their access to the core building blocks of fiber infrastructure. This situation has industry analysts predicting continued dramatic growth for the sector, with even conservative estimates placing yearly increases in the use of fiber optics at more than 35 percent. By 2004, the worldwide market for optical components is expected to hit $24 billion (See Fig).

But there is an important caveat to this rosy picture. The manufacturing processes that have taken the optical components market this far are not up to the task of bringing it to the next level. Current processes simply aren't scalable or robust enough to meet dramatically increasing demand. Fortunately, there are new processes--or rather new applications of established manufacturing technologies--that can meet the demand, while at the same time reducing costs and time-to-market, improving yields, and giving the industry more flexibility.

Optical components expand the bandwidth capacity of fiber networks by breaking the laser light that carries data through the network into different colors, or wavelengths. Each resulting wavelength is capable of carrying a discrete data channel. Optical components therefore enable fiber networks to transmit vastly more information.

Companies like Corning and JDS Uniphase, which specialize in the design and manufacture of optical components, have enjoyed incredible growth in the recent past, spurred by the needs of system providers to accommodate the delivery of more and more data through their systems. To date, these companies have benefited from circumstances supporting near monopolistic conditions because the barriers to entry into the components business have been so high. The optical component business is characterized by long design cycles, short product cycles, and prohibitive capital requirements. Typical component manufacturing processes demand extensive manpower and vast and specialized manufacturing facilities.

But given the vast size of the emerging market for optical components, even these market leaders aren't ideally positioned. Despite ongoing efforts to ramp-up production capacity, many of the industry's giants are facing six-month order lead-times in order to accommodate the delivery of even basic components. In a market characterized by such a rapid pace of technological advancement and the high costs of addressing these advances, these companies are running full speed just to stay in the same place.

As the market progresses along a curve of exponential growth, it becomes increasingly evident that the large thin-film filter process manufacturers will not dominate the market based on their size alone. At the end of the day, optical component makers not only have to be able to produce products in high volume, but they have to be able to stay very close to technology's leading edge. Ironically, it appears that the same "high barrier" conditions that once made life ideal for the industry's big players are now setting the stage for a dramatic industry shake-up that might leave the 800-pound gorillas behind.

As the demand for more, and increasingly specialized, components intensifies, a number of smaller companies are emerging in order to meet the challenge. Equally at issue for these new players is the ability to ramp-up to a significant production capacity. As always, speed is the determining factor. The key to a breakthrough has relatively little to do with an impressive collection of registered patents, and everything to do with the more decisive ability to deliver needed product on a mass scale.

The crux of the issue is neither between established and emerging powers, nor even between companies and their competing technologies. The real issue is how quickly and easily new technology can be embraced and promoted by automated, large-scale manufacturing.

Thin-Film Processes

The prevailing approach to designing and producing passive optical components is based on a technological process known industry-wide as thin-film filter processing. Essentially, this process applies multiple layers of film to a single piece of glass in order to carry photons between component channels.

A defining characteristic of the manufacturing processes implicit to thin-film filter applications is the amount of manual labor required. As an example, consider a 16 channel DWDM, a standard optical networking component that splits a single signal into 16 signal streams. With the standard thin-film approach, armies of technicians are dedicated to the highly specialized and laborious task of joining two different strands of material together with a special adhesive. Each setting takes approximately 20 minutes to complete. Add to this the fact that each component of this type will require 16 similar connections, and that these will typically require a 50 percent rework rate in order to achieve the desired specifications.

The above scenario addresses but one stage of the basic manufacturing process inherent to the thin-film production. Subsequent to this, the component must be tested, and then very carefully packaged; a highly delicate process given the importance of limiting signal loss.

Predictably, labor costs represent almost 50 percent of the costs of producing an optical component, as well as being a major determinant of the number of components a thin-film facility can produce. In short, the relationship between productive capacity and the ability to increase that capacity is linear. To make more product, you need more technicians, and more, or larger, manufacturing facilities. As you move further along this curve, it becomes increasingly evident that the ability of this labor-based manufacturing model to address the need for increased productive capacity and upwards scalability is somewhat limited.

Another complication limiting the ability of thin-film component manufacturers to scale-up production results from the fact that appropriate automated manufacturing equipment has not yet been invented. In general, the trend is for manufacturing equipment to follow the evolution of the product.

An additional obstacle facing the thin-film manufacturing paradigm can be seen in reference to the astonishingly regular increases in component complexity that appear to run parallel to the growth of the optics market. With each technological advance the ability of current processes to produce components in volume becomes further constrained. In the optics market, each successive performance hurdle requires higher performance components, and more of them.

For example, the current industry standard for one-to-many channel optical components is a 40 DWDM component that allows a single strand of fiber to be split into 40 distinct channels. Even now, components capable of delivering as many as 160 distinct channels are being sought by the carrier industry. It is safe to predict that this trend will persist.

When dealing with thin-film filter process architecture, as channel spacing is reduced to increase the number of channels in a single component, greater constraints are placed on the component filter. Notwithstanding the considerable challenges of applying filtering material to incrementally shrinking surfaces, decreasing the space between channels invariably results in a reduction of component yield. Over the long-term, the spacing requirements of thin-film based components are an insurmountable, limiting factor.

In tandem with all of this, perhaps the most damning indictment of thin-film's limitations has to do with the changing paradigm of the optical component industry. The next challenge for component manufacturers is to address the increasingly specialized needs of customers. Building customized components means in turn that component architecture and manufacturing processes must be capable of addressing a number of variants quickly. That's something the existing environment and current processes are loath to support. As always, the chief factor is time to market, and now, more than ever, a scalable and flexible manufacturing process is key.

In an ideal world, manufacturing is a primary element of the product conceptualization process. This assumes that from the outset there is a strategy in place, predicated on the notion that success of the venture depends on the ability to deliver product, and moreover that the process to deliver that product is scalable. To borrow a word from the new age lexicon, manufacturing must be addressed from a "holistic" perspective.

Simply stated, scalability has never been a prime concern for optical manufacturers. Typically, optical component companies spring from scientists or engineers who have incredible insight into the science of photonics, but who also have a tendency to think compartmentally. Optics companies tend to be focused on the complex science and physics at work in making a component technically more efficient and faster. Invariably, the primary focus of these companies is almost always geared towards the ability to deliver a working prototype, while registering a few patents along the way.

The result is that component design is rarely conceived of in such a way that lends itself to automated manufacturing. The working assumption is that the ability to produce 10 components is necessarily a testament to a company's ability to deliver 500 or 1000 components. When it comes to manufacturing capacity, clearly, this is not the case. It is a strange irony that in an industry where time to market is considered a decisive factor, the ability to scale up appropriately to ensure product delivery is a distant concern.

Effective manufacturing process engineering identifies hurdles to the product delivery cycle, and addresses issues such as supply scarcity and unnecessary cost. The guiding ethic argues the necessity of extensive data and metrics, core communications systems, and an elaborate and detailed manufacturing resources plan. In an industry as inclined toward rapid growth as the optical component business, it also demands that these systems be in place long before they are required, because instituting them after the fact is a major challenge.

Clearly, the ultimate goal for process engineers is to achieve a mature manufacturing process, characterized by the ability to measure a myriad number of factors dependably, thereby ensuring that everything is in place to keep the manufacturing line running smoothly. Contrast this with a thin-film filter process environment that uses immature processes and that will be increasingly compelled to embrace near-infant processes in an attempt to try and maintain market share.

Optical component manufacturers are starting to realize that a new paradigm is emerging, and that it is very unlikely that the thin-film filter process is in a position to address it. Faced with this new paradigm, the companies that will lead are those that have the appropriate manufacturing processes in place from the outset. In other words, those companies have decided to align themselves with mature manufacturing processes and proven ways of doing things. It stands to reason that if there is not a manufacturing paradigm that supports what you are looking to achieve, the smartest thing you can do is borrow one.

Planar Waveguide 

While by no means a recent development in the optical component industry, an alternative technological process for designing and assembling optical components called planar waveguide shows considerable promise in facing the challenges inherent to component manufacturing and upward scalability.

The planar waveguide process involves placing glass fibres on silicon chips in order to provide pathways capable of routing a light signal between fiber optic strands. The core manufacturing process behind the technology addresses the objective of building a silica pathway across a silicon wafer, and relies on a variant of the same process employed in manufacturing traditional electronic silicon chips.

Planar waveguide processes borrow directly from the semiconductor industry, relying on standard and proven production machinery like Plasma Enhanced Chemical Vapour Deposition (PECVD) equipment, or high-pressure oxide equipment. Better still, planar waveguide borrows from the semiconductor world without assuming its massive infrastructure costs. While a semiconductor line might represent a $1 billion capital investment, an optical component manufacturing line will cost a comparatively scant $20 million.

Besides the fact that the process lends itself to automation, another compelling argument in favor of planar waveguide is that the component architecture also lends itself to batch processing. What this means in a manufacturing context is that a number of dies can be mounted on a 6- or 8-inch wafer and processed simultaneously. Moreover, 70 or more wafers can be processed together through the same process cycle.

Planar waveguide supports a manufacturing process that addresses the rework of multiple component connections in less time than it takes existing thin-film processes to address a single connection. In short, human labor is largely banished from the manufacturing equation.

In addition, the architecture that allows for the mass manufacture of planar waveguide components also supports the ability of this technology to address the customization needs of the new paradigm. As stated earlier, thin-film process manufacturing encounters considerable difficulty in addressing the junctions between component connections, largely because the process is labor-heavy, slow, and costly. Furthermore, each circuit-to-fiber junction represents a significant source of signal or light loss.

Planar waveguide reduces the number of these interfaces by mounting these connections on the wafer itself. Circuit-to-circuit connections are subject to significantly less loss of light, and are therefore more efficient. These "on-chip" interfaces also introduce considerable flexibility to the design process, creating significant "head room" for the technology to address future requirements and increased capacity needs of next-generation components.

Finally, planar waveguide processes also benefit from the semiconductor world's experience with component level testing. It stands to reason that testing throughout a complex manufacturing process is an important control in terms of limiting cost and increasing throughput, especially in a context of where a throughput of less than 30 percent is the norm. The objective is to take the loss early and eliminate faulty components from the process chain before they drain additional resources. By eliminating these components at each processing stage, line capacity can be increased by a factor of two or even three.

The prevailing mentality and approach to optical component testing adheres to the notion that it takes roughly 15 minutes to test a component at the wafer level. Thin film manufacturers, used to dealing in discrete components, test each component separately and then re-test them following assembly.

Conversely, semiconductor processes allow for the testing of arguably more complex components in less than 15 seconds. Notwithstanding the challenge of developing an appropriate test cradle, and material properties aside (namely the tendency of photons to be less cooperative than electrons), there are considerable lessons to be learned here.

The next challenge for optical component manufacturers is to realign their current manufacturing processes so as to address the emerging market for customized components and solutions. As an overview, this article has argued that planar waveguide-based architecture engenders a far greater technical potential than its thin-film counterpart. More importantly, it has also demonstrated that planar waveguide leverages the considerable benefit of scalable, automated manufacturing.

Given the growing market for optical components, the advantages of a technology that lends itself to automated manufacturing cannot be underestimated. The history of widespread technological adoption is also the history of manufacturing advantage. Just as pick and place automated tools for printed circuit boards lost the war to silicon integrated circuits, planar waveguide's manufacturing advantage makes a compelling argument for the end of thin-film component manufacturing. The only question that remains is who can leverage this advantage most quickly. Chances are, it won't be the 800-pound gorillas.

Terry Caves is the vice president of engineering and manufacturing at Zenastra Photonics, Inc. (Ottawa, Ontario, Canada).

[Graph omitted]