Photonics Assembly in an EMS Environment

Estimated reading time: 33 minutes

This article appeared in two parts in the January and February issues of SMT. It is presented here in its entirety.

By Glenn Woodhouse

The explosion of the Internet and Web-based multimedia applications has driven phenomenal growth in networking hardware requirements. This rapid growth trend has required original equipment manufacturers (OEMs) to turn to the electronics manufacturing services (EMS) industry as a means to quickly and efficiently satisfy their manufacturing needs. The most pronounced Internet trend of the new millennium is an insatiable thirst for bandwidth. Conventional copper-based networks, although improving in speed through evolving hardware solutions, cannot promise the quantum leap in bandwidth capacity that the fiber optic solutions demonstrate with currently available technology. Recent investments in the long haul fiber-based core have pushed bandwidth constraints closer to the enterprise and/or end user. Unfortunately, the hardware side of the fiber optic-based networking infrastructure outside of the core is in its infancy. As a result, the majority of the fiber optic networking hardware manufacturing has resided within captive OEM plants. The transition has begun for the migration of photonics assembly from OEM to EMS in order to meet the huge increased demand for low cost optical network bandwidth.

Optical network hardware employs all of the same SMT-intensive electronics as the traditional copper network hardware. However, fiber optics presents new challenges in the incorporation of on-board active components such as laser, switches, photo diodes, and passive optical components such as splitters, filters, multiplexers and demultiplexers (mux/demuxs), optical fiber, and fiber connectors. All of these photonic components require interconnection, mounting and testing processes outside of traditional EMS competencies. In the past these skill-intensive processes, residing in the labs and plants of the OEM, have not lent to mass production automation. Herein lies the challenge of the technology transfer from OEM to EMS and the scaling of manufacturing processes to meet the increasing volume demands while insuring high quality and reliability with ever decreasing manufacturing costs.

This paper introduces the reader to photonic components, their assembly and test processes, and the unique manufacturing challenges and paradigm shift to be faced by "SMT-centric" manufacturers.

BackgroundWe are amidst the "Internet Era." The most pronounced Internet trend of the new millennium is an insatiable thirst for bandwidth. This thirst is driven by a rapidly expanding Internet user community with expectations well beyond basic services such as e-mail, Internet chatting and Web searches for informational reference. Today's user expects expanded services such as e-commerce, Web-based interactive applications, integration of the Internet/Intranet for corporate enterprise business systems, and streaming and/or downloadable multimedia. The Internet has become a key means of doing business in both our professional and personal lives. This environment is responsible for the amount of data traffic on public networks growing at a rate of more than four times per year for the last three years, already surpassing voice traffic as the dominant traffic type on some networks. The economic downturn resulted in a pause for 2001/2002, but according to a recent report by Dataquest Inc., forecasted sales of $36 billion in the optical transmissions systems segment in 2003 should finally eclipse the $33.8 billion in global sales realized in the year 2000. EMS companies garnered 20 percent of this market in 2000, with an anticipated penetration rate increasing to 25 percent in 2003. Thus the outsourcing of optical network hardware manufacturing is anticipated to increase to an approximate $9 billion market opportunity for EMS players by 2003.

This explosion of users, applications and data traffic is placing considerable demands on the existing networking infrastructure and technology. At the forefront of the quest for expanding network bandwidth lies new optical networking hardware technology, driving rapid growth in demand for photonic components and modules, and integrated optoelectronic networking cards and systems. Dense wave division multiplexing (DWDM) technology, for example, is gaining tremendous popularity primarily because of its capacity to provide a significant increase in the capacity of fiber optic data transmission systems. DWDM is an optical technology that couples many wavelengths (>40) in the same fiber, thus effectively increasing the aggregate bandwidth per fiber to the sum of the bit rates of each wavelength. Thus, the carriers can get more out the existing fiber network infrastructure for less cost by adding technology such as DWDM hardware without the need to add more fiber (i.e., burying cable) to increase the capacity of the network. This is a compelling reason among others to expect the continued growth in optical networking hardware and the continued increase in speed of optical data transmission rates. This growth of photonics in the datacom space, combined with the significant trend toward manufacturing outsourcing over the last decade, has optical OEMs turning to the EMS industry as a means to scale.

A considerable challenge facing the rapid scaling of optical component manufacturing today lies in two key areas: automation (or lack thereof) and knowledge transfer. Although there are some emerging semi-automated solutions, the size, delicacy, diversity and historical volumes of photonic components present a material-handling dilemma that impedes full scale automation and the development of turnkey in-line assembly equipment. Because of this, past assembly activity has primarily been performed manually by highly skilled workers in the labs and factories of the OEM. This skill set is very different than that required to operate an automated SMT environment.

Dissect an optical network switching system and you will find a synthesis of complex mixed electronic placement and soldering technology, photonic modules and optical fiber routing. This union of electrons and photons lends to the popular terms "optoelectronics" or "optronics." For the sake of clarity in this paper, a further definition of terms is warranted. As in the industry today, the terms "optics" and "photonics" are used interchangeably within this paper, pertaining to datacom devices which transmit light (photons), amplify, filter, concentrate/split, receive or otherwise influence the behavior or properties of light. For this paper, the term "devices" as used in "optical devices" or "photonic devices" is intended to incorporate the family of modules and components. A photonic/optical "component" is the lowest level of integration that can be found on an optoelectronic assembly, such as a laser or photodiode, or passive like a splitter or filter. An optical/photonic "module" is a substrate-based (typically laminate) assembly incorporating SMT components and photonic devices into packaging that can be incorporated into a higher-level optoelectronic board.

Optical Network Component OverviewHigh-end optical switching and routing systems typically require custom configuration of several complex optoelectronic boards, combined with passive photonic modules, power supplies, media drives and configured software to comprise the end unit. The manufacturer of such systems must possess configure-to-order capabilities both in their business and physical manufacturing systems. System assembly knowledge and resources for the integration of electronics and sheet metal are required. Full test capabilities to support functional, parametric, and statistical testing including high-speed data traffic testing capabilities, burn-in and/or environmental stress screening (ESS) must exist in-house.

Focusing on the mixed technology boards and passive photonic modules within the system takes you one step deeper in the dissection. The SMT content of the optical system boards is similar to that of high-speed copper-based networking systems. These boards consist of high layer count double-sided PCAs (16 layers typical) ranging from 6 x 6" (152 x 152 mm) to 18 x 20" (450 x 500 mm) in size. Organic substrates potentially exhibit "islands" of high frequency that emerge in the 10 to 40 Gbps applications that may require the of alternate high frequency laminate materials. SMT actives including fine pitch devices and BGAs on the secondary side are not unusual. Component counts ranging from 1,000 to 4,000, passives down to 0402 in size, BGAs (some exceeding 1,000 balls), (BGA (down to 0.75 mm/30 mil pitch), TSOPs, QFPs, fine pitch down to 16 mil (0.4 mm) are typical. High-density press fit and wave solder through hole header connectors (up to 1,200 pins) provide interconnection to the system backplane. A variety of wave soldered, paste-in-hole and manually soldered components such as electrolytic capacitors, oscillators, power converters, unique connectors, sockets, flex circuits, photonic devices, and optical fiber are incorporated.

Where the optoelectronic networking board departs in manufacturing technology from the copper-based network board is in the assembly of the various on-board active and passive photonic devices. Traditional electronics relies on copper traces and solder joint interconnections to channel electrical data I/O through its various components, whereas the photonic device employs light through optical fiber. These devices include active photonics such as lasers, photodiodes and GBIC (gigabit interface converter) modules. These active components incorporate analog and/or digital I/O (including power and ground) as well as photonic I/O, meaning that they require both soldering and optical fiber interconnection. Passive photonics include filters, splitters, concentrators, fiber termination connectors and on-board fiber management components such as fiber reels, fiber tracks and fiber anchors. Although these passive devices do not require soldering, they, along with active photonic devices, do require special care in handling, cleanliness, and a unique set of processes and equipment for interconnection and test.

The optical datacom networking application described above typically employs 250 or 900 micron fiber, in either single mode (SM) or multi-mode (MM) format for on-board photonic device interconnection and routing.

Fiber optic cable consists of a glass "core" surrounded by an optical material called "cladding." Differences in refractive properties of the core and cladding form a cylindrical "mirror" at the interface of the two materials. It is this mirror effect that serves to channel light down the tiny core for very long distances with minimal light loss. The cladding is coated with an acrylate covering called the "buffer coating" that protects it from moisture and other damage. The diameter of SM core is typically 8.6-9.5 microns (0.00034"-0.00037"), cladding 125 microns (0.0049"), and buffer coating 250 microns (0.0098"). More protection is provided by an outer plastic covering called a "jacket," bringing it up to a final diameter of 900 microns (0.035"). Optical fibers used in "off-board" applications such as patch cords and cables will include additional protection in the form of a woven Kevlar fabric layer covered by an outer protective plastic casing resulting in a final diameter in the 3 mm range (0.118"). Because of the glass component of optical fiber, special care must be used in handling, assembling and routing of fiber cable both for the safety of the fiber and the human handler.

Traditional Manufacturing Models And Current TrendsComponentsPhotonic component manufacturing traditionally has resided in specialized OEM manufacturing facilities for both passive and active components. This model continues today with low level outsourcing occurring within a handful of niche service providers. An EMS industry trend of note for 2000/2001 was the flurry of acquisition activity by Tier One EMS companies among the photonic assembly niche market. Although current economic conditions have tempered this activity, acquisition will be the most prevalent model for an EMS entry into the photonic component (versus module) contract assembly market. This is primarily due to the unique and disparate technologies employed in photonic component assembly as compared to the board and system level assembly core competencies possessed by the EMS provider today. These technologies include IC fabrication processes, silicon to fiber bonding processes, fiber-to-fiber bonding/fusing processes, optical polarization and "micro-alignment" processes, hybrid electronic assembly processes, and plastic and hermetic component packaging processes.

The manufacture of complex active components such as lasers and photodiodes modulated with OC-48 and OC-192 (2.5 Gigabit/second and 10 Gigabit/second, respectively) typically require specialized equipment, a clean room environment, manually intensive high skill assembly processes, tuning, and test with relatively low yields in comparison to the monolithic IC and board level assembly industries.

Many passive photonic components such as standard fiber connectors, patch cords, filters and splitters are being manufactured in volume by offshore OEM and contract manufacturing facilities. Equipment and processes are still unique in comparison to the EMS "tool kit" but the volumes and labor content requirements warrant going offshore into dedicated facilities. Custom passive component assembly predominately remains in domestic OEM plants and with a few niche contractors who typically don't possess any module/board level SMT assembly capabilities.

ModulesOptical/photonic modules integrate supporting digital and power circuitry with active photonic components to produce standard and application-specific functional photonic receivers, transmitters and transceivers. There is considerable effort in module miniaturization taking place to facilitate the "on-board" incorporation of modules into optoelectronic board designs and as a means to lower costs for what has traditionally been a high-cost device. Although small, typically ranging from 0.50 x 1.0" to 6 x 6" (12.7 x 25.4 mm to 152.4 x 152.4 mm), photonic modules are far from simple, employing the latest in SMT advances. These substrate based (typically 16 layer, fine line/trace laminate) assemblies incorporate SMT devices including passives down to 0201 in size, (BGA (ceramic substrate (BGA very typical), TSOPs, TQFPs, fine pitch SOICs down to 16 mil (0.4 mm), and high density area array SMT connectors. As mentioned previously, specialty high frequency laminate materials may be required. If the module is designed as a dual in-line package (DIP), press fit or soldered-in pins are required. Packaging can include sheet metal enclosures, potting/plastic encapsulation or metal cans with sealed lids. For the state of the EMS industry today, with regards to manufacturing capability (versus capabilities for the acquisition of photonic assembly companies), this synthesis of SMT and photonic components is where the most logical opportunity for market entry lies. The EMS provider can augment their SMT core competencies with new capabilities in the areas of photonic device handling, assembly, and test for the manufacture of photonic modules. Both the photonic component and optical network system OEM would be potential customers to an EMS provider offering photonic module assembly and test services.

This augmentation of the EMS provider's core competencies cannot be taken lightly. The knowledge and equipment set employed by an electronics manufacturer for component placement and soldering provide minimal, if any, contribution in the handling, interconnection and routing of optical fiber and photonic devices. As a result, new tools have to be acquired and/or developed by the EMS provider to support these activities. In some cases, these tools may not be readily available and have to be developed by the EMS provider. For example, Plexus Corp. has developed tool set and method for managing the routing and splicing of high fiber count modules. This tool set, called the Fiber Ramp(tm), provides an efficient means soldering optical components and subsequently managing the routing, splicing and termination of their pigtails. Looking forward, additional unique manufacturing tools will have to be developed as new optical module designs are created.

The integration of optical components into an optoelectronic module is not an exercise in placing smaller components faster and more accurately with existing or new turnkey equipment, or enhancing existing processes to solve soldering challenges presented by unique package requirements. Photonics assembly is an entry into a strange new territory that affords the EMS provider little luxury in drawing upon existing process technology knowledge. A process engineer who has mastered the art of convection reflow profiling will be at a loss if asked to develop or troubleshoot an optical fiber fusion splicing profile. This goes beyond equipment familiarity. None of the rules of soldering apply to fiber splicing. A new "mindset" is required that comes with training, practice and experience. Prior to any photonic component integration, the photonic module will go through SMT assembly and in-circuit test (ICT). Active and passive optical components are then mounted and interconnected via fusion splicing and optical fiber connectors are installed. Any remaining optical fiber pigtails for interface with the optoelectronic datacom board are terminated at this time as well.

Without special training, an SMT or test operator, technician or engineer is not prepared to handle optical fiber, fiber terminations/connectors and photonic modules which incorporate optical fiber pigtails and terminated cables. Commercial training services can be used as a means to seed your production or training team, and qualify for military or government contracts.

Once trained, your photonics assembly and integration team is a unique resource that is not easily replaced. Paying special attention to the selection and retention of a good production team can not be over stressed in its importance.

Examples of new rules for the optoelectronic manufacturing environment:

• Safety -- Disconnected optical connectors may emit radiation if the far end is coupled with a working laser or light-emitting diode (LED). Do not view the fiber end of a cable or plug with an optical instrument until absolute verification is established that the fiber is disconnected from any laser or LED source. Short segments or "shards" of raw optical fiber can enter the skin, eyes and mouth. The utmost in care and caution must be employed in the management and disposal of fiber debris including the wearing of protective equipment such as gloves and safety glasses.
• Handling -- Fiber should be handled carefully without pulling, twisting or bending.
• Bend Radius -- The recommended minimum bend radius for SM fiber routing and handling is 1".
• Cleaning -- Fiber connector performance is degraded by the presence of contaminants, particles and dust. Compressed air and isopropyl alcohol cleaning of fiber connectors should be performed as a final step in the assembly operation. Dust caps should remain on all unused connections and should be replaced after any removal or test process.

As simple as these rules sound, it is extremely easy to violate the minimum bend radius rule. For example, fiber can be overstressed during transportation of a photonic component or module from its bubble pack to a workstation by not paying attention to how the fiber coil is supported. In addition, it is generally not possible to detect if the fiber has been damaged until the photonic device is mounted to the optoelectronic board later in the process and a downstream functional test is performed. Replacement or repair of a module fiber pigtail is often not an option due to fiber length and/or customer specification constraints.

Although use of a clean room is not a requirement, at a minimum, an order of magnitude in manufacturing process cleanliness and awareness is required, not just at the operator level but for anyone who comes in contact with the product, including test debug technicians and engineers. Particles and oils present on the faces of fiber connectors can be passed to the faces of mating sockets. Although connector faces can be readily cleaned, lack of adherence to cleaning procedures can introduce scratches and additional contaminants to the fiber surface. Sockets, once contaminated, are very difficult if not impossible to effectively clean, and may require costly removal and replacement.

The manufacture of photonic modules, and subsequent assembly of active and passive photonic devices onto an optoelectronic board will require optical fiber splicing and termination processes foreign to a traditional EMS provider. Methods of joining optical fiber include fusion and mechanical splicing, as well as alignment of two fiber connectors with a socket.

The joining method that provides the least signal loss, the most compact junction, and the greatest assembly flexibility is fusion splicing. Fusion splicing involves taking two fibers, heating the fibers to their melting point and fusing the two fibers into one continuous fiber, while maintaining near perfect alignment of the mating microscopic cores.

Fusion splicing equipment is available in "hand-load" or clamp- (fiber holder) based equipment. Hand-load based equipment is generally less expensive and faster but requires high operator skill level, produces additional opportunity for fiber damage and contamination through excessive handling, and introduces additional process variation due to a greater "human factor" in the process. Hand load equipment is popular among smaller niche contract shops and OEM facilities that possess an experienced and stable skill base where one operator or technician performs the entire range of assembly steps, building up an entire unit themselves, without handing off to other workers for subsequent steps. With clamp-based equipment the fiber is located in a clamp at the onset of the process, and remains in the clamp throughout its stripping, cleaning, cleaving, fusing and splice protection process steps. The clamp ensures accurate registration of the fiber for cleave length and splicer alignment, thus reducing worker skill level requirements and the opportunity for handling based fiber damage and contamination. Once in a clamp, it is safer and easier to pass fiber from step-to-step in an assembly line fashion. This facilitates palletization, work flow balancing, optimized utilization of critical high dollar equipment, and the reduction in skill sets required to perform any particular function in the assembly line flow. This methodology is more familiar and conducive to an EMS model where rapid work force and production line scalability, predictable throughput and aggressive return on equity is critical to protect the low assembly profit margins typical in the industry.

The basic equipment set for fusion splicing is described below in order of process steps:

Jacket/Buffer Stripper -- A "pliers"-style hand tool or a more sophisticated clamp-based tool for the cutting and removal of the 900-micron jacket, and in some applications a portion of the acrylate buffer. The outer jacket for 3 mm cable can also be removed with these tools.

Buffer Stripper -- Similar to the stripper described above, a hand tool or clamp-based tool is employed to remove the 250 micron acrylate buffer coating down to the 125 micron glass cladding. The more sophisticated tool is recommended, incorporating a heated blade set, as a means to reduce bending and axial stresses imparted on the glass components of the fiber during the stripping process. For high strength splicing applications, chemical and ablation stripping methods are employed.

Fiber Cleaner -- Hand wiping with lint free wipes and high purity isopropyl alcohol (99+ percent) or the use of an ultrasonic cleaner with high purity isopropyl alcohol is applied for cleaning acrylate residues, foreign particles and organic contaminates from the stripped fiber. The ultrasonic cleaning approach requires the fiber to be secured in a clamping mechanism. This process has more opportunity for process repeatability and control and minimizes the opportunity for handling based bending stresses to be imparted on the fiber. It also removes the opportunity for a "pennywise" member of the production team to reuse lint-free wipes, thus creating a contamination opportunity.

Fiber Cleaver -- The most critical fiber preparation step is cleaving. A cleaving tool operates in a manner similar to cutting window glass but at a microscopic level. The fiber is clamped in a special holder, scoring the glass with a circular blade a fixed distance from the end of the holder, and then breaking the fiber by imparting a concentrated bend precisely at the score line.

Fusion Splicer -- Fusion splicing equipment serves to align two fibers, heating the fibers to their melting point, and "fusing" the two fibers into one continuous fiber by mechanically merging the molten glass ends in a precise fashion. Although SM-to-SM is the most prevalent splicing activity in the datacom optoelectronic environment, various splicers are available that have the ability to join dissimilar fiber types, dissimilar fiber diameters, and perform radial alignment of polarization maintaining (PM) fibers such as "Tiger" and "Panda." Depending on the splicing system, fibers are aligned either through optical core alignment technology (recommended for high accuracy alignment of the 8.6 to 9.5 micron cores in single mode fiber) or less expensive v-groove systems that align off of the fiber's outer protective buffer.

Step 1 -- The splicer will inspect for excessive cleave angle, inclusions and contamination on the fiber that is detectable by the cameras. The user can set acceptance thresholds for these parameters.

Step 2 -- Because a fiber's cladding has different optical refractive properties than its core, a camera system with appropriate lighting can detect the cores of two opposing fibers. With two cameras located 90 degrees apart and pointing perpendicular to the axis of the fibers, an alignment of the axis of the two cores can be performed. Very high-resolution cameras and stepper motors are required to achieve alignment of the microscopic cores. The alignment step can fail due to misclamped fiber or fiber curling. The user can set acceptance thresholds for these parameters.

Step 3 -- The splicer will "fire off" a cleaning arc of set energy over a defined time interval. A cleaning arc does not contain sufficient energy to melt the glass fiber but it can burn off small particles and contaminants such as oils or cleaning solution residue. Driving a current between two electrodes produces the arc.

Step 4 -- The splicer will then initiate the fusing arc at a set energy over a defined time interval. This arc will result in the simultaneous melting of the ends of both fibers, forming molten glass globules. At a key point in the arcing process the fibers are "stuffed" together by a defined distance. The application of the fusing arc energy is continued, to evoke a wetting action between the two globules. With the fusing arc energy still applied, the two fibers can be pulled apart by a slight distance as a means to provide fiber diameter adjustment at the splice junction.

Step 5 -- The splice is now complete. The splicer will again optically inspect the finished product for bubbles and inclusions and will reject the splice based on user defined acceptance criteria. It will also relocate the center axis of each core through an optical process similar to the original axis alignment in step 1, calculate the degree of physical misalignment, and output the result into an estimated theoretical dB splice loss. The splicer will reject the splice based on a user defined maximum acceptable estimated splice loss. Because two active and/or passive photonic devices are typically being joined with this method, the fiber ends are not accessible for insertion into an optical power source or light meter and actual splice light loss measurement is not possible. This fact makes robust splice loss estimation an important machine feature. Finally, every splice undergoes a nondestructive proof test pull by the machine. With each splice, the possibility exists that some optically non-detectable interface defect could cause a weak splice even when the splice has low estimated loss. The proof test does not break the fiber, but it does verify that the splice is not weak.

Although cleanliness is important in the fusion splicing process it does not necessarily require a clean room assembly environment. At a minimum an isolated clean environment is recommended that undergoes frequent wipedowns. A walled off room separate from the large SMT production area will suffice. Non-particle-generating paint can be applied inexpensively to walls, ceilings and furniture. Suspended ceilings should be taped off from above to prevent overhead dust from settling through the joints in the ceiling. Reduce opportunities for particulate contamination by keeping cardboard, wood and other paper products to a minimum in the splicing area. Shoe covers and lint-free smocks will reduce traffic-related contamination. Wearing of makeup and hair gels/oils by production personnel should be discouraged. Use of a laminar flow bench with HEPA filtration at the splicing operation is an effective and affordable supplemental contamination control.

Upon completion of the fusion splicing process, it is typical for some form of splice protection and reinforcement to be applied at the splice location. The two most common methodologies are recoating and the installation of splice protectors or "splints."

Recoating is the process of over-molding and curing liquid jacketing compound around the splice area and back to the 900-micron jacketing material (for our SM fiber example). This technique has been in existence for many years and is popular for high-strength applications such as suspended, underground and marine optical cable bundles because it does not add bulk to the outline of the fiber which is critical when grouping dozens of fibers into a single large cable. Because of its prevalent past use and familiarity, recoating has carried over into the datacom optoelectronic space. In this application its advantages are again no increase in fiber outline as well as a degree of flexibility in the splice area. These attributes make fiber routing around an optoelectronic board a bit easier in the event the splice falls on a radius such as a fiber reel or fiber track corner. The drawback to recoating is mainly centered around manufacturing process and flow. Recoating requires a variety of molds to match different fiber and splice geometries. These molds require continual cleaning while in use. Recoating compound must be stored and handled correctly, and the opportunities for contamination of the splicing work space with compound can jeopardize product yields and reliability.

Splice protectors are a hollow cylindrical sleeve consisting of heat shrink tubing, a hot melt adhesive and an inflexible stiffener.

Splice protectors are a younger technology than recoating and historically have been large and bulky. A typical splice protector package is 30 to 40 mm in length (1.18" to 1.57") and 1.2 to 2.5 mm (0.047" to 0.098") in diameter depending on whether it is fitting over a 900-micron jacket or 250-micron buffer. The splice protector is a "low profile" 30 mm sleeve for unjacketed 250 micron fiber. The splice protector is installed by slipping it over one of the fiber ends to be spliced and sliding it back out of the way. Upon completion of the splicing process, the splice protector is centered on the splice junction, overlapping the jacketing or unstripped buffer of each fiber, and placed in a tube heater. In 30 to 60 seconds the hot melt adhesive will flow, the heat shrink will contract squeezing all air out of the splice protector cavity, and the adhesive will securely bond to the stripped fiber and jacketing material. The splice interface is now well protected from tensile and bending stresses. This process is clean, requires minimal additional equipment and tooling (tube heaters are often integrated into fusion splicers and stand alone units cost less than $700), and no special skills are required beyond good fiber handling practices. The drawbacks of splice protectors are the cost of the actual splice protector and the bulk and rigidity of the resulting package. The raw material cost of a splice protector can contribute as much as 25 percent to the total cost of the splicing process. New small package splice protectors have recently addressed the issue of bulk.

A robust maintenance and calibration routine is paramount to the success of an ongoing fusion splicing process. Paying special attention to routine splicer arc calibration, electrode replacement cycles, sharpness and alignment of the cleaver blades, condition of stripper blades, and equipment cleanliness will insure that opportunities for process variation and failed splices are minimized.

Another key process integral to photonic module and optoelectronic board assembly is fiber termination. Custom fiber termination is a highly manual operation. Some semi-automated equipment is available to aid in the performance and repeatability of the termination process steps. A summary of this process is provided below:

A strain relief boot is slid over the fiber jacket. For 3 mm cable, the cable jacket is stripped back to the appropriate length and the woven protective Kevlar yarn is cut away. The 900 mm jacket and underlying buffer coating is stripped back to the appropriate length. Acrylate remnants are removed using an isopropyl alcohol-soaked lint-free wipe. Epoxy is injected into the connector package until a small bead forms on the tip of the ferrule.The fiber is gently threaded into the connector, flaring the Kevlar back as it enters, until the buffer is fully seated. The connector sleeve is crimped and a drop of epoxy is applied where the jacket meets the crimp. The strain relief boot is seated onto the back of the connector. The epoxy is cured. The excess fiber is cleaved 1.0 mm or less above the cured bead of epoxy by running a cleave blade against the fiber to produce a score on the edge of the fiber. The tip of the fiber is deflected causing the fiber to cleanly break off or "cleave." The end face is polished by performing a series of polishing steps using progressively finer grits of polishing film. Finally, the termination is cleaned and inspected with a fiber microscope.

With the SMT, photonic component assembly, fiber splicing and fiber termination steps complete the module is ready for testing. The purpose of this test is to ensure the proper function of active photonic components such as lasers and photodiodes and validate that all fiber splices and terminations were successful and do not exceed light loss limits. Optical module testing requires a set of test equipment and software that is typically unique to the module.

These test sets include a precision power source, digital multimeter (DMM) and optical power meter. Thermal management of the laser is also necessary. This can be done by using the transceiver card to interface to the test set or using a parameter analyzer. This type of test would measure raw (no modulation) power. It could also be used to measure the linearity of the device relative to the supply or temperature. These test sets consist of readily available and affordable components. Engineering time is required for the integration of the test set and the development of test software.

Optoelectronic Board Level IntegrationThe nature of SMT content and complexity at the optoelectronic board level for datacom systems is described earlier in the paper. The manufacturing technology employed up to this point is representative of a capable datacom EMS provider. Presently the outsourcing of true optoelectronic products is in its infancy with either the OEM installing the photonic components themselves onto outsourced PCAs, or shipping the board to a third party photonics assembly house for photonics integration. The goal of the datacom EMS provider is to offer "one stop shopping" for an optoelectronic datacom OEM.

Once the optoelectronic board is complete through the SMT/mixed technology assembly process, the board will undergo ICT test and then be forwarded as a subassembly to photonic integration. At this step active photonic components such as laser, detector and/or GBIC modules will be soldered or socketed onto the board, passive components such as concentrators, splitters, filters and connector duplex adapters are mounted, any additional fusion splices performed, and fiber wound, routed and secured throughout the board.

Many of the photonic modules/components require hand placement and soldering due to unique form factors, packaging limitations (i.e. provided in bubble pack), fiber handling challenges, and temperature and/or cleaning sensitivities. This presents a number of obstacles in the automation of the photonic integration step. Fiber must not violate minimum bend radius specifications (typically 1") and must be securely anchored to the board. Fiber terminations should be cleaned with isopropyl alcohol and a lint-free wipe and connector ports should be blown out with inert dusting gas before insertion of fiber connectors or dust caps.

The testing of the optoelectronic board presents some new challenges to the EMS provider versed in the testing of conventional networking boards. Tests include parametric tests and bit error ratio testing (BERT). To have a statistically significant sample, BERT testing may take several minutes. This time may be reduced by stressing the system's power or temperature. It can also be stressed by changing the input pattern. Bit error ratio (BER) is measured statistically. Therefore a confidence level is associated to a BER measurement.

The parametric tests consist of "eye diagram" optimization mask testing through the testing of extinction ratio, rise time, average power and jitter.

The high-speed nature of optical datacom products requires very sophisticated test equipment. Just the hardware alone in this example costs $650,000 to $700,000, exclusive of automated board handling equipment and test engineering software development time and fixture development/debug. Throughput of the system is still constrained by the time-intensive BERT testing, coupled with the parametric test time. For even a medium-volume manufacturing environment, multiple test sets would likely be required to keep pace with upstream production.

Optical Datacom Network System IntegrationHigh end optical switching and routing systems typically require custom configuration of several complex optoelectronic boards, combined with passive photonic modules and assemblies such as light rings and optical add drop multiplexer modules (OADMs), power supplies, media drives and configured software to comprise the end unit. The business system capabilities and manufacturing competencies required by the EMS provider to support the optical datacom OEM were addressed in the background section of this paper. There is a large EMS installed base possessing the majority of these capabilities today, as a result of the rapid growth and outsourcing of the copper based networking market over the last five years. Additional expensive high-speed data traffic testing hardware may be necessary to support OC-48 and faster products.

Business ConsiderationsCare must be taken when developing pricing models for optoelectronic products. Photonic module assembly is substantially more labor intensive than SMT, requiring higher skilled labor and more technical support. Expected yields will be lower due to the delicacy of the process, lack of automation, and often times no opportunity for rework (i.e. replacement not allowed due to fiber length constraints). Considering the "youth" of the technology and the blazing data transmission rates associated with optical networking, product design and component marginalities become a yield driver as well. There is substantial financial risk associated with yield fallout of high-speed photonic components such as lasers and photodiodes. Active components can run in the thousands of dollars. In-process fallout can result from component infant mortality, design marginalities, module level manufacturing-induced stresses, and module level manufacturing defects and mishandling. The development of material liability and "bone pile" agreements between OEM, EMS provider and photonic component manufacturer is recommended as a means to remove any gray areas in liability and equitably share the financial risks associated with in-process fallout. The very nature of these high-speed, light-based components make failure analysis and fault assignment a difficult and costly task.

Due to a current lack of packaging standards in the photonics environment, a high degree of product-specific tooling is typically required to handle, align and transport material in process. Test equipment is specialized and expensive. The technology is progressing at such a rate that the risk of obsolescence of product line, not to mention high dollar test equipment and tooling, must be considered. For these reasons, non-recurring expenses (NRE) that are to be distributed into product unit pricing should be front loaded to insure payback. Customer consignment of test equipment should be considered.

ConclusionsInternet usage and network traffic will continue to grow, driving increased integration of optoelectronics in the datacom space. This growth combined with the significant trend toward manufacturing outsourcing over the last decade has presented new opportunities and many new challenges to the EMS industry.

Photonic component manufacturing remains "a beast unto itself." Due to broadly disparate manufacturing technology, EMS acquisition of photonic component manufacturing companies is more suitable than developing processes in-house as a means for entry into the photonic component market. Examples of this can be seen in recent Tier One EMS acquisition announcements. Photonic module manufacturing and optoelectronic integration presents the most logical "organic" market entry point, and greatest opportunity for the EMS provider, by combining core competencies in complex SMT board level assembly and test with in-house development of new photonic module assembly capabilities.

Photonics assembly requires a paradigm shift for the SMT-intensive EMS provider. Soldering and component placement knowledge is not transferable. Fiber optic fusion splicing is a totally different material science than traditional lead-based soldering. The level of delicacy of optical fiber requires new procedures and a new mindset for care in manufacturing cleanliness and handling within the production team. The optoelectronic datacom networking hardware market has little standardization at this time, driving unique designs from component level up through module, board and system level. As a result there is little available automation to support the manufacturing and test environment, so a highly skilled work force is required to support the delicate and labor-intensive processes. Additional testing knowledge and expensive specialized equipment is required to perform parametric and statistical testing of these high-speed devices.

The EMS provider will need to be very careful with the business models surrounding capital investment, quoting, NRE expenses and turnkey material purchasing agreements when entering into the photonics assembly space. It is important that the unique challenges associated with material management, technology obsolescence, and human resource requirements are taken into account.

With all of these risks and challenges comes the potential for attractive rewards in an exploding market, an exciting technology, and an expanded service offering. EMS capabilities ranging from photonic module manufacturing through optoelectronic board and system assembly services provide the optical datacom OEM a very desirable "one-stop-shop" outsourcing solution. A solution that promises to provide the scalability required to meet the needs of the "Internet era."