P. O. Box 3192
Redwood City, CA 94064

phone: 415-850-8166

Optoelectronic Packaging Consulting

Proceedings from Fiberoptic Automation EXPO
San Jose, California, Dec.4-6, 2002, Session FA10AB "Markets, Trends and New Technologies"

The Holy Grail was lost and Camelot lay in ruin. King Arthur was morbid, his castle's candles unlit, and the gallant Knights of the Round Table were dispersed and some were already deceased.
Will fiber optic assembly wield the sword of Excaliber in automation to return the Kingdom of Photonics to prosperity? Or will it take a combination of Merlin's process assembly knowledge to bring back the light? The present market situation for telecommunication photonic components is certainly as dark as any Medieval winter's night. Market pull is perceived to be non-existent. Technology push ventures have been slain by both VC financiers and established OEMs. Cash is being conserved.
How long can the market no-pull continue? To be realistic we are still sliding leaving skid marks of closed companies, laid-off telecom employees, closed factories, auctioned equipment and frail nerves along the roadway. Market conditions determine success in all industries.
This paper will address the trends in photonic device packaging and automated assembly that may result from the present environment. Challenges and opportunities will be shared.
A short recap of the events that led to the present market no-pull condition parallels King Arthur's Camelot:

The Holy Grail: "Commodity Priced Fiber Optic Bandwidth Availability Enabled by Low Cost Photonic Devices"
Lower cost photonic packaging certainly seems like a Holy Grail for the realization of commodity bandwidth delivery. The search so far has been just as elusive.

The Dragon: "No Downside Financing"
Visions of unlimited data bandwidth demand launched the bubble dragon of no downside financing. The resulting abundance of venture capital allowed a splurge in capital expenditures for optoelectronic device development, equipment for fiber alignment, test, and instrumentation and for network build-outs.

The Call of the Dragon: "Excess Bandwidth Growth Prediction"
Network applications that would drive bandwidth demand beyond supply called the Dragon to the bubble. Fueled by network use forecasts and investment capital, the Dragon's fire burned in every sector.

The Sword Excaliber: "Automated Assembly Solutions"
Automated assembly process solutions hold an ability to lower the cost of photonics devices1. Early on the number of available automated photonic assembly solutions was limited just as there was only one Excaliber. Sword makers duplicated new versions and created a bazaar of automation options. Solutions began to integrate together yet the Knights, like the OEMs, seemed to have the better swords of automated assembly in their grand estates. And then the darkness fell on the land.

The Wizard Merlin: "How to Use the Power"
The magician's power comes from knowledge. Equipment capability alone will not determine success in optoelectronic automation assembly2. Knowledge of materials and assembly processing is necessary in order to design photonic products and automated solutions that can achieve repeatable yields. However, process knowledge wizards are not necessarily sword makers of automation equipment.

The Sword in A Stone: "Over-Inventory of Components"
Drastic and dramatic changes in the market pull of carrier expenditure plunged photonic component suppliers into a situation of few if no new orders. Compounding this has been the large downward price pressure on existing components due to over-inventory at suppliers and network providers. The Photonics Kingdom was plunged into a seemingly irreversible position much like Camelot when Excaliber was driven back into the stone.

The Funeral Pyre: "Close Down, Exit & Auction"
The photonic components world then collapsed. Equipment auctions, consolidation of major OEMs or their exit from the optoelectronic component business, and mass closings of venture financed new component suppliers and network suppliers lit the funeral pyre.

Technology push ventures rarely succeed. The recent period of huge IPO valuations and start-up company buy-outs by large OEMs during the telecom bubble resulted from a distortion of what the market pull was in the face of embracing the believed hype in technology push. Current loss write downs testify to a harsh re-evaluation of the market.

Market pull for photonic devices is governed by:
Deployment of new equipment beyond stock-piled inventory
CAPEX spending return of carriers which are driven by:
Availability of investment capital
Bandwidth demand outpacing supply
Fundamental change in carrier revenue generation from voice to data
Deployment change over to more efficient networks requiring new enabling hardware

The perceived market pull of the telecom bubble launched an armada of technology push activities in photonic components by both venture financed start-ups and established OEMs. There are underlining technology changes that have been embraced and implemented by telecom networking, such as optical amplification and DWDM. However, the number of technology push options overwhelmed the readiness of the market to use them. Technology push activities include:

Me-Too Device Suppliers
New component vendors offering similar devices assembled by the same packaging methods
Need For More Functionality
Transmitter wavelength tunability, in narrow and wide tunable ranges
MUW/DMUX functionality integrated with active device functionality
Higher Bandwidth
10GBps over 2.5GBps
Higher DWDM channel count deployment
Tighter DWDM channel spacing
More Device Performance
More pump power
More transmitter power
Less dispersion, PDL, etc.
More Integration
MUX/DMUX integration with active functionality such as receivers, power tap monitors, VOAs, and optical switches
Integrated wavelength lockers
Integrated EAM modulators on wafer
Integrated SOAs
Arrayed transmitters and receivers
Change of Status-Quo Networks to Something Better
CWDM networks applications
SONET to MESH network reconfiguration for better traffic management efficiency. This requires large port count OXCs and network reconfiguration ownership
Dynamic DWDM wavelength provisioning. This requires wide tunable lasers and OXC switches
Niche Applications
Free space network communicationv

Development of new photonic devices has led to some substantial advances in photonic packaging solutions and include:
Planar waveguides: Thermal control solutions for AWGs and optical coupling processing for arrayed single mode fibers.
External cavity tunable lasers: Unique solutions to each design have met the significant packaging challenges.
OXC switches: Packaging solutions that address switch core insertion loss variation, stability and qualification challenges are being implemented for both 3D and 2D optical MEMS switches. Challenges in packaging large (1000) port count switches together with lack of network implementation readiness has limited deployment more towards medium (40-126) port count switches.

Standardization is a natural market progression that can enable lower automated assembly cost. Lower costs can be realized by: equipment solutions being able to handle a wider range of products; reducing the amount of customerization efforts for the assembly processes and associated tooling fixtures; and aiding interchangability of automated equipment from different vendors. While vendor Multiple Source Agreements (MSA) have worked to standardize outside package interfaces and are a natural result of the down market conditions, no standardization efforts exist for inside the package. It is the configurations inside a photonic package that affect the assembly processes and their automation.

The need for lower cost packaging designs and processing methods is widely recognized as a necessary element for network growth. Substantial challenges in achieving this will require new solutions in the device packaging. Each one of the following trends has seen some degree of implementation. They are listed as trends not because of the amount of their acceptance to date, but because of their capabilities to deliver the market demand for lower cost photonic devices by lower packaging and assembly cost.

1. Low Cost Hermetic Fiber Feedthroughs
Hermetic packaging of active photonic devices is considered necessary to pass Telcordia environmental requirements. Hermetically sealing the fiber to the package has required a costly process of metallizing the optical fiber. This metallization process is also not compatible for integration with automated assembly. A solution based on Impact Mount Technology (IMT) eliminates the need to metallize the fiber and is inherently compatible with integration into automated assembly processing3. This IMT solution has already demonstrated Telcordia qualification and will likely become the perferred choice for low cost single fiber hermetic feedthroughs.

2. Low Cost Hermetic Packages
Hermetic packages such as 14 pin butterfly or mini-DIL configurations represent a significant ratio of photonic device bill-of-materials (BOM) cost. Price pressure on the devices will require lower cost options, which are constrained by the present low volume demand of the market. At high volumes metal injection molded (MIM) packages offer lower cost options than traditional package fabrication methods. MIM packages also have the ability to add cavity features without incurring machining costs. This package cost dilemma can not be resolved by proven low cost TO-can configurations for applications with data rates above 1-2GBps.

3. Automated Fiber Pigtail Preparation
Fiber pigtail preparation is a sequence of processes for which automated solutions have not yet been prevalent. The processing includes fiber stripping, cleaving, cleaning, optical end face preparation by polishing or other means, and may also entail fiber metallization. Fiber pigtails are needed for transmitter and receiver modules with specifications ranging from low to high end tolerances. Lower cost fiber pigtails will be demanded at both ends of the specification range. This is an area of photonic packaging in which automated solutions will become dominant, much like automated fiber jumper connectorization processing, when market pull returns.

4. Low Cost Transceivers
Low cost transceivers are one type of photonic product seeing some market pull through applications in datacom and 10G Ethernet. Leaders in this field have made signficant advances for low cost small form factor (SFF) automated assembled transceiver modules. Packaging advances by these vendors will likely continue and lead their configuration trends.

5. Packaging Design Utilizing Silicon Micro-Optical Benches
Silicon micro-optical benches offer low cost precision submounts that can reduce photonic packaging costs and enable automated assembly solutions. Silicon micro-optical benches are fabricated using silicon wafer batch processing of: anisotropic wet etching; photolithography; and thin film metallization. Their designs incorporate alignment planes, precision placement planes for optical fiber and optoelectronic chips, and provides circuit trace on the bench at very low cost. They can also be designed to leverage automated pick'n'place equipment capability by use of their in-plane attachment approach. Silicon micro-optical benches have been incorporated into photonic packaging by several vendors. However, their value in lowering assembly costs will be recognized more in devices manufactured in high volume using pick'n'place automation equipment.

6. Design of Components, Subassemblies and Packages for Compatibility with Pick'n'Place Automation
Photonic packaging typically has not been designed from the ground-up for automated assembly. Pick'n'place automation equipment that is widely available and used in electronic assembly manufacturing has constraints which are often not compatible with photonic assembly processing. Precision component placement at 5mm resolution, available with flip-chip and other pick'n'place equipment, is frequently required in photonic assembly and is difficult to achieve with manual assembly. Hence there is a compelling need to utilize these automated pick'n'place capabilities for photonic device assembly. However, the photonic packaging design and assembly processes need to be constrained within the automated equipment limitations. Also design features should be incorporated that help leverage the automated equipment capabilities. These approaches can have a large impact on processing cost and achievable yield. Packaging design for compatibility with pick'n'place automation includes the use of common reference planes for passive alignments, reduced degrees of freedom for component location and placement, and the integration of fiducials onto components and submounts to aid machine vision guidance.

7. Integration of Optical MUX/DMUX Functionality with Active Component Functionalities in One Package
The trend of DWDM applications seeking more functionality per module is driving the integration of optical MUX/DMUX with active device functionality. Integration of these presents many packaging challenges where the solutions will be specific to the incorporated technologies. Solutions that require large hermetic packages will be prohibitively expensive. Module packages will tend towards using internally packaged components that comply to Telcordia requirements inside a non-hermetic larger package.

8. Implementation of Low Cost YAG Laser Weld Attachment Production Lines
Component attachment by laser welding has the unique ability to deliver repeatable submicron precision4. This combined with the elimination of active component reliability issues caused by incorporation of organics from epoxy attachment or flux from solder attachment will make laser welding a preferred trend for automated assembly. Low cost fiber pigtails prepared by automated assembly using IMT ferrulization will support implementation of laser weld to meet low cost, high yield assembly demands. Solutions that lower capital costs by multiplexing YAG laser weld sources over several production lines at one time using beamsplitter boxes will become a trend. Laser weld attachment for both high end and low end photonic devices will be implemented in fully automated assembly production lines and as weld stations using manual component loading.

Successful benefits of automated assembly include:

I. Improved Product Consistency
This is probably the most impelling reason to implement automated assembly for photonic devices. Successful implementation reduces product preformance variation and increases production yield. Photonic device yield loss can easily affect manufacturing profitability due to the high BOM cost for telecom application devices.

II. Decreased Number of Assembly Staff
Automation reduces operation expenses as the large population of staff for manual assembly processes is not needed. Maintaining quality of production on volume upswing requires fewer resources and the pain of laying-off staff on volume downswing is also reduced.

III. Improved Ability to Modulate Production Volume
Production throughput may be easily modulated if the processes and equipment are robust. Throughput may be ramped up by extending shift operations with fewer staff. Automated processes may be readily duplicated at relatively low risk with additional assembly stations at the same or a different site. However, depreciation of capital expenditure for automated equipment is a serious consideration for low utilization. Low utilization may be due to extended market down-turns or perhaps due to difficulty in achieving assembly yield because of issues in the product design for manufacturability and automation.

The return of market pull will encourage the following assembly automation trends to be implemented. Some aspects of these trends have been addressed prior to the bubble burst as requirements for successful automation equipment solutions:

1. Pallets for Loading and Transfer During Assembly Processing
The use of pallets allows for efficient loading and transfer of components and subsassemblies at and between islands of automation. Pallets may contain multiple devices for assembly processing prior to fiber pigtail integration. One device per pallet may be the preferred solution for optoelectronic test considerations once the device enters the fiber pigtail integration process. Two features in the design of pallets for photonic device automated assembly are: the interface with the automated equipment for pallet transfer and component location within the pallet; and allowance for holding, locating referencing and assemblying components and subassemblies on the pallet during the assembly process sequences. Pallet designs today are customized and configured specifically to a particular product or a vendor's automation equipment. There is a need to establish a set of standards for pallets used for photonic assembly automation.

2. Flexibility in Automated Production Line to Manufacture Multiple Product Types with Fast Reconfiguration
The ability to assemble different products on the same assembly line increases equipment utilization and will be a differentiating cost factor for photonic manufacturers. The use of pallets aids in the reconfiguration ability of automated assembly. Elements that enable flexibility include: design of assembly process and equipment with degrees of freedom, placement resolution, and machine vision guidance to allow multiple products to be assembled on the same production station. This flexibility may be necessary in order to realize the acceptable return-on-investment (ROI) for the automation equipment.

3. Pick'n'Place Automated Assembly of Device Subassemblies
There is and will remain a divide between automation equipment that performs pick'n'place of components and submounts and equipment that performs the optical alignment and attachment for photonic devices. This is due to the different machine abilities and process requirements of these two groups of assembly processes. While these two station process types will likely remain as two separate automation islands, they should share the same component and device pallets. If the fiber pigtailing station uses a different pallet to accomodate the pigtail and optoelectronic testing needs, then the subassembled device should be removed from the pick'n'place process pallet and placed in the pigtailing pallet by the pick'n'place station. Pick'n'place automation station processes should include: vision guidance component location, orientation and placement, component solder or epoxy attachment, and wire bonding. These stations may integrate many of these processes into one or configure the processes sequentially as process islands with linear transfer of pallets.

4. Automated Optical Alignment Assembly Stations
Optical alignment requirements can vary substantially depending on product type. Submicron component alignment tolerances are common for many edge emitting semiconductor products coupled to single mode fiber, while larger placement tolerances are allowable for other products such as VCSELs coupled to multimode fiber. The packaging design and assembly processes are likely to remain specific to the product type. There is a need for automated assembly equipment that is flexible enough to perform the various optical alignment processes of different product types. However, unless the products and the assembly processes have been designed with this flexibility parameter in mind, which is currently not the case, then this processing flexibility translates into higher capital expense for the automated system. Solutions that minimize capital cost for automated optical alignment and maintain processing flexibility will become a growing trend. One such solution is to configure the process such that each alignment step or group of steps is separated onto individually hard tooled alignment stations with the fewest degrees of flexibility and motion possible. This way each alignment station has the least number of motion controllers and the lowest capital cost. These individual assembly steps are then joined in a linear fashion either in a manual transfer mode or fully automated using transfer pallets. The hard tooling provides for more robust processing. Flexibility in processing different products is achieved at lower cost by reconfiguring the assembly order and by modifying the hard tooling. The optical alignment sequence is best automated using active alignment optical beam profiling for high yield production. Certain process steps that are easy for human operators but difficult for machine automation need not be automated. Such steps include the loading and handling of fiber pigtails and treading a fiber through a package snout.

5. Optoelectronic Test Stations
Automation of optoelectronic testing is necessary for data collection, analysis and documentation. Indeed automated data information storage, archival, retrieval and SPC analysis will be mandatory due to the volume of data collected. Loading of the assembled devices at the test station, either individually or in pallets or trays, may remain manual for modest volume production but in high volume require automated loading solutions. The pallet used to hold the device during fiber pigtailing should be designed to be compatible with the test station. The large capital cost for optoelectronic test equipment has been temporarily minimized due the availability of lower cost test equipment via factory auctions.

6. Burn-In Stations
Solutions for burn-in of components and devices arrayed in pallets are readily available. The design of these pallets for compatibiliy with the pick'n'place assembly pallets and pigtailing process pallets may be necessary to realize full automation of high volume production lines.


I. Outsource Manufacturing
Many benefits can be leveraged by using outsource manufacturing of photonic devices to lower assembly costs4. This leverage potential is now more compelling given the ongoing consolidation of established photonics OEMs and the break-up of their vertically integrated supply chain. While lower business volume for optoelectronic outsource manufacturers has resulted in their own staff reductions, nothing better could have happened for their market capture opportunity in the long term than the bubble burst2. The cost of capital equipment to build photonic production capability has never before been so low with the abundance of factory auctions which seem to be never ending. These factors present improved opportunities for outsource manufacturers, particularly large ones, to capture manufacturing volume upon return of market demand. However, the outsource manufacturers may not have enough photonic assembly business to sustain their interest to invest in photonic manufacturing. Challenges in achieving high assembly yields remain for outsource manufacturers. Those who have mainly electronic assembly backgrounds will encounter significant hurdles in addressing photonic process yield issues. Reduced staffing and a market no-pull environment may delay the realization of these process yield challenges to the detriment of the outsource manufacturer's competitiveness.

II. Photonic Manufacturing Build-Up in China
Build-up of photonic component manufacturing capabilities in the Far East, especially in China and Taiwan, is a trend that will continue and see further investment6,7. The domestic market in China is one of the few market pulls that presently exist for global photonic communications products. Trends in assembly processes implemented in China emphasize utilization of low cost labor, manual assembly processing and epoxy attachment methods. These methods are well suited to the old ways of manufacturing photonic devices, especially passive devices, which dominated the industry. Products with low and medium level alignment tolerances are well suited for this blend of manufacturing solutions. The current trend of photonic manufacturing capacity build-up in China may be described more as a business solution and as a global opportunity, rather than a process or automation solution.

III. Yield Control
Control of assembly yield will dominate the justification of process automation. Because of the need for yield control, automated pick'n'place and alignment stations will be used for both high end products and high volume, low end products. The high volume low end products will probably be manufactured by larger outsource manufacturers at offshore facilities, principally in the Far East, as the infrastructure of these large outsource manufacturers are suited to the product cost demands. However, deploying the process assembly knowledge to achieve repeatable yields by robust processing will remain one of their greatest challenges beyond the present lack of volume demand.

The following three strategies when combined will empower automation of photonic assembly that can successfully meet processing requirements and market demands:
Design for compatibility with assembly process and processing equipment.
Design product and assembly processes for manufacturability.
Design automated assembly equipment for flexibility in reconfiguring for multiple product types.
What will become of Camelot and the Kingdom of Photonics? This land without a king will be led out of darkness by the return of the spirit of Merlin. His delivery of photonic assembly process knowledge will guide the new swords of automated assembly over the paths of peril to prosperity.

1. Dixon, R. "Low-Cost Manufacturing: The Future of Photonics" Compound Semiconductor, August 2002.

2. Palen, E. "Change is the Arena of Opportunity: A Look at Photonic Devices Following the Telecommunications Bubble Burst" Editorial in MEPTEC Report, July/August 2002.

3. Palen, E. "Impact Mount Technology Analysis & Strategies Roadmap" reproduced with permission on website

4. Palen, E. "Why Outsource Optoelectronics Manufacturing?" MEPTEC Report, July/August 2001.

5. Shannon G, and Palen E., "Laser-Weld Attachment Enables Repeatable Submicron Precision," Optical Manufacturing, May 2002.

6. Vardaman, E. and Urekew ,T. "Low-Cost Optoelectronic Manufacturing: The China Solution", Proceedings from IEEE' Photonic Devices and Systems Packaging Symposium (PhoPack), July 2002.

7. Kincade, K. "China Emerges as a Hotbed of Photonics Outsourcing Providers" Optical Manufacturing, September, 2002.

Edward Palen, Ph.D.

phone: 415-850-8166


--return to home webpage--

Copyright © 2002, 2004 E. J. Palen
all rights reserved