Taking a Photonic Device from Prototype to Manufacturing:
Ten Common Pitfalls and How to Avoid Them

EDWARD J. PALEN, Ph.D., P.E.
P. O. Box 3192
Redwood City, CA 94064

phone: 415-850-8166
email: epalen@earthlink.net
website: http://www.palensolutions.com

Proceedings from IEEE's Photonic Devices & Systems Packaging Symposium (PhoPack)
held at Stanford University, California, July15th-16th, 2002

Abstract
Packaging of photonic devices has been the largest cost factor in their delivery to the telecom market. Taking a photonic device from prototype to manufacturing involves both packaging design and assembly processing. This paper will share experiences learned in taking photonic devices from product concept, through to proof-of-concept, package design, design for manufacturability, prototype sample builds, product Telcordia qualification and volume manufacturing. Ten common pitfalls observed in this process will be described. Key areas of optical and thermo-mechanical modeling, materials selection, and available assembly alignment and attachment techniques are covered for their influence upon product and assembly process reliability and production yield. These pitfalls apply to a wide range of photonic devices which require optical coupling of optoelectronic elements to optical fiber. Special attention is addressed to active photonic devices.

The ten common pitfalls to be avoided in taking a photonic device from prototype to manufacturing are:

1) Not Understanding Positional Tolerances of the Product's Optical Design. Understanding a device's optical design allows for informed decision making on the device package design and assembly processing. A lack of understanding can lead to poor attachment choices. For example, choosing lower tolerance attachment methods for components requiring high positional tolerance requirements leads to device performance and assembly yield problems. Likewise, specifying high tolerance attachment methods for components with lower tolerance requirements leads to increased materials and processing costs and to assembly process constraints. The use of optical modeling software early in the device design for manufacturing is a good investment that defines the device performance and its sensitivity to the positional placement of its components. These placement tolerances, ranging from submicrometer to several micrometers, can lead to different decisions in device packaging and assembly process choices. Smaller allowable placement tolerances require assembly methods with higher alignment and attachment repeatability. Optical modeling allows for understanding which axes the position tolerances are most sensitive: in x, y, or z axes and for some products in rotation, pitch, or yaw. Modeling of optical coupling from waveguides to single mode fiber (SMF) should include both ray tracing and mode matching in order to deliver true positional sensitivity analyses. The foundation to design of the device packaging, submounts and assembly sequencing is laid by understanding a device's component positional tolerance requirements. A dimensional stack-up analysis of all components and submounts is an early checkpoint milestone in the packaging design. The use of optical modeling allows for the matching of device waveguides and component optical parameters to each other in order to achieve superior device performance and lower sensitivity to component placement. For example, if a device's semiconductor waveguide dimensions are already fixed, the focal lengths of coupling aspeheric lenses may be chosen using the model to optimize optical coupling and minimize placement tolerance sensitivity. In another example, the radius-of-curvature of fiber tip cone lenses or wedge lenses may be defined by optimal modeling for optimized optical coupling to laser facets. Thermo-mechanical finite element modeling (FEM) can be used at the front end of the packaging design to assess the mechanical design positional stability and component temperature gradients over package temperature exposures and device thermal loading. Both affect device optical performance. Stability of submounts and attached components is necessary to maintain optical alignment and device performance under thermal loads of elevated package temperature environment, TEC operation, and laser diode operation. Design solutions to these thermo-mechanical packaging issues include: choice of submount thickness for resistance to thermal gradient warpage; proper TEC sizing; material selection of submounts for thermal conduction; and submount ability to provide temperature uniformity across its surfaces.

2) Poor Selection of Attachment Methods and Materials. The selection of attachment methods determines the materials requirements for submounts and optical components. Poor selection of materials is a key pitfall in taking a photonic device from prototype to manufacturing. This pitfall can occur frequently because prototype development teams are not constrained by manufacturing issues while demonstrating the device proof-of-concept. Device prototype assembly configurations, which then form the basis of the design to be manufactured, are baselined from the proof-of-concept build experience. The device assembly configuration at this stage is often not reviewed with regard to component placement tolerance data, gathered either from optical modeling or from experimentation, nor with regard to manufacturing constraints. Also the selection of attachment methods may be determined by the chosen outsource manufacturer's capabilities without assessment of the component placement tolerances from the device's optical design. Available component attachment methods include:

thermally cured epoxy

ultra-violet (UV) cured epoxy

solder attachment

laser soldering

laser welding

Material Outgassing Considerations Materials outgassing can affect device performance reliability and its ability to pass Telecordia qualification. Material outgassing is particularly a concern for laser transmitters and laser pump modules. Outgassing affects include degradation of optical coupling by condensation on optical surfaces and degradation of reflection coatings at high optical densities on laser facets. Material outgassing data that is relevant to photonic device reliability is not commonly available beyond mass loss from thermal graviometric analysis (TGA). While NASA databases for material compatibility to spacecraft applications are available, the ASTM-E-595 test measurement for total mass loss (TML) and collected volatile condensable material (CVCM) at 125oC in vacuum does not directly relate to photonic device reliability. Pitfalls in selecting epoxies for photonic device application can be avoided by selecting materials that are used in the disk-drive industry and by collecting relevant outgassing data for the materials under consideration. While outgassing data on materials used in the disk-drive industry is typically not available from their manufacturers, their use in the disk-drive industry provides a basis of consideration for photonic device application based upon similar requirements.

Epoxy Attachment Epoxy attachment is generally considered the lowest cost entry attachment process for building photonic devices. It is commonly used in photonic device proof-of-concept demonstration builds and early prototype builds because of its ease of use, low cost, availability and the absence of reliability requirements. While epoxy is an excellent choice in these early prototypes, its use in photonic devices should be assessed by experienced process and reliability engineers for each photonic device application. Issues that should be reviewed include its:

outgassing properties

cure profiles (UV snap cure and thermal cure)

shrinkage upon cure

thermal expansion (CTE)

mechanical stability

adhesion strength

refractive index and optical transmission in some applications

ability to act as a reversible gas getter within the device package

ability to maintain device optical coupling over Telecordia requirements

processing cost relative to solder and laser weld attachment in a production environment

Thermal cycling can lead to positional shifts in epoxy attachments that can degrade optical coupling in the device. This can be due to material expansion, loss of adhesion, and material property change due to environmental exposures such as water vapor. The amount of optical component alignment shift can be process dependent and is a function of epoxy volume, cure profile, and shape after cure. The use of UV snap cured or thermally cured epoxy for precision alignment attachment of components often requires the development of cure profiles that minimize both component alignment shift and epoxy outgassing potential. UV snap cured epoxies are typically cured at room temperature and are not as sensitive to large epoxy shrinkage from elevated cure temperatures that thermally cured epoxies can experience. The use of epoxies that can be both UV snap cured and thermally cured allows for minimized alignment shift in the initial component attachment by UV snap cure. This is followed by adhesion strength increase by thermal cure without alignment shift due to the thermal cure. The outgassing and stability of UV cured epoxies should be reviewed in detail to assess their compatibility with the photonic device in question.

Solder Attachment Solder attachment of submounts is a common method used in building photonic devices. A distinction should be made between the use of solder attachment for mechanical structures such as submounts and the use of solder for attachment of pre-aligned optical components. Reflow separation or repositioning of solder after attachment is possible if assembly rework or optical realignment is necessary. The use of flux in the soldering process requires special attention in the manufacture of photonic devices due to reliability concerns from flux contamination and cleaning process restrictions of many optoelectronic components. Solder attachment of components can be performed with flux, as a no-clean process, and as a fluxless process. In order to use solder attachment, the components require a compatible metal interface. Submounts may be plated with nickel/gold to assure a good repeatable solder joint. Ceramic substrates with electrical circuitry fabricated by thick film or low-temperature-cofired ceramic (LTCC) processes have a metal surface layer compatible with solder attachment. Optical fibers, thin film filters, and lenses require an expensive metallization process, either by thin film metal evaporation or by electroless plating to adhere metal onto the glass. Solder attachment may be successfully implemented to photonic package assembly by using the following processing tools and techniques:

choice of a sequence of solder temperatures so that earlier solder joints do not reflow at subsequent assembly attachments. This is aided by elevation of the subsequent solder reflow temperature by gold plating uptake into the solder during the initial solder reflow.

choice of a processing sequence that allows for thorough solvent cleaning of soldered subassemblies prior to the integration of the sensitive optoelectronic components that can have their performance degraded by solvent cleaning.

use of cover gas belt furnace or vacuum autoclave ovens. Both require the fabrication of fixtures for the process.

pre-tinning attachment surfaces, which can be solvent cleaned afterwards if desired, to improve the surface wetting capability of the solder and to reduce dependence on flux to ensure solder surface wetting.

use of no-clean fluxes, which do not affect the product's ability to pass Telcordia requirements, that aid in solder surface wetting.

Solder is an excellent method for mechanical attachment of submounts and components, especially for designs that require high thermal conduction through the submounts. Solder attachment of thermo-electric coolers (TECs) in photonic devices requires special attention in order to avoid entrapment of voids in the solder. Voids reduce the effective thermal conduction ability and limit the TEC cooling ability of the device in the package. Solder attachment of TECs in a production environment requires a stable and repeatable process with regard to void entrapment. The methods described above can be used to do this with the aid of non-destructive inspection methods such as ultrasonic imaging to confirm the process result. TECs have a temperature exposure limit, which if surpassed during TEC solder attachment, will lead to degradation in the cooling ability of the TEC. Therefore, the maximum solder temperature and duration should be monitored and controlled during TEC solder attachment process set-up. An alternative method for TEC attachment in photonic devices is to use thermally conductive epoxy (epoxy filled with small solids). However, epoxy outgassing issues makes this alternative a less desired choice because photonic devices that require active cooling usually have laser facets whose stability is sensitive to material outgassing. Precision attachment of optical components and fiber pigtails using solder requires special care in process set-up in order to achieve attachment position repeatability. Process controls that enable solder attachment at high positional tolerance and repeatability include:

solder volume control

placement of solder volume, often resolved by the use of solder preforms

conditioning the solder preforms for optimal solder surface wetting by etching oxides off of preforms prior to use

solder solidification shape

application of a repeatable and controllable solder reflow temperature profile. The use of pulsed laser energy, resistance heating, and inductive heating are methods that can deliver such repeatable temperature profiles for solder attachment.

Laser soldering has an advantage for solder attachment within photonic packages because it is a noncontact heating method that can easily access the small components to be soldered that are recessed within small device packages. These recessed components are difficult to access for resistance heating contacts and inductive heating coils.

Laser Weld Attachment Laser welding is a process that can deliver highly repeatable submicrometer attachments in a production environment if the weld-attachment processes are well understood and if the device is designed for laser-weld manufacturability (reference 1). A pitfall in taking a photonic device from prototype to manufacturing using laser weld attachment can occur in the understanding of the weld attachment process by both the device designers and by outsource manufacturers. If the outsource manufacturer does not have prior experience in a laser weld attachment, particularly in a production environment, a learning process will be encountered. The associated costs, schedule and assembly yield risks would likely fall upon the prototype device being taken to manufacturing. The key areas of laser weld process knowledge to be aware of are; materials selection, joint tolerances, laser weld set-up, post-weld-shift (PWS) repeatability and PWS correction. An outsource manufacturer may have a preferred configuration for laser weld attachment, such as bendable weld clips or coaxial weld configurations. If a configuration is proposed based upon outsource manufacturer configuration preference, the maturity of their processing should be assessed. A less mature process will consume more devices in developing acceptable process results, leading to increased cost and schedule risk in taking the device from prototype to manufacturing.

3) Not Understanding the Sensitivity of Device Design to High Bandwidth Performance. High bandwidth, OC-192 and OC-768, photonic devices require additional packaging design attention. An investment in the following device characteristics is required in order to realize high bandwidth performance of photonic devices in manufacturing:

component performance characterization by s-parameter measurements

component interactions by RF circuit modeling

modeling of circuit trace layout of substrates, bond pads and package feedthrough interfaces for impedance matching

The understanding gained allows for a design that can reproduce device performance in production. These characterizations also enable efficient troubleshooting of packaging issues involving circuit frequency resonance, cavity resonance, and signal reflections that affect device performance. Pitfalls that affect the achievement of device bandwidth performance are commonly experienced if these component and device performance characterizations are not made.

4) Not Collecting Empirical Optical Coupling Tolerance and Stability Data on Proof-Of-Concept Build and Early Prototypes. Optical modeling data alone can not deliver a working photonic device. At some point within the venture of understanding a device by optical modeling a decision should be made to start collecting experimental data by assembling device components. Experimental data of achievable optical coupling, component placement sensitivities, device performance and device stability may confirm or challenge modeling results. Common pitfalls in taking a device from prototype to manufacturing are to either not engage in optical modeling at all or to become enveloped in the modeling at the expense of not moving the program forward. The empirical data can substantiate the optical modeling data, reveal inappropriate modeling assumptions, and introduce real component characteristics that the model could not know. Collecting empirical data by performing component optical alignments allows for a reality check in the device design, and for intimate learning of the device assembly constraints. This early empirical data allows for evaluation of device design limitations, for device design changes that allow for larger component placement tolerances, and for measurement of device performance variation resulting from different attachment processes.

5) Poor Component Specification and Assuming Vendor Quality Control. Working with vendors may be required for components that are critical to a device design to ensure their ability to supply the component to the specifications required. Management of the component supply chain can be essential for the success of many new photonic device designs. In order to assure the manufacturing viability of a new prototype device there must be a commitment from the developer of the new device to invest in engineering interaction with the component supplier for component processing improvements and quality control. A common pitfall in taking a photonic device from prototype to manufacturing is in not recognizing critical component availability and its deliverable quality control. Photonic devices that advance the industry performance capabilities, such as in the case of widely tunable lasers, are more sensitive to critical component availability and performance issues. Specifying component performance is part of the challenge. Component performance requirements can be defined by using optical modeling, circuit modeling and by using assembly process engineering experience for packaging configuration constraints. Examples of components where performance specification and vendor quality control are important include:

Reflection coatings optical performance and stability

Semiconductor waveguide beam divergence variation and its affect upon device optical coupling and device performance

Optical fiber and IR window metallization hermetic seal ability

Implementation of appropriate acceptance screening tests of components aids in minimizing the impact of component lot-to-lot variations. Acceptance screening criteria required for the component function in the device may well not be the same as the screening criteria performed by the component supplier. Problems in component quality variation can be addressed by fabricating proxy samples in large enough quantity to perform component derisk activities. An example of such a derisking activity is thermal cycling of soldered metallized fiber while testing leak rates for hermetic seal. Early detection of vendor supply problems allows for corrective actions to be implemented or, if necessary, selection and qualification of another vendor.

6) Poor Outsource Manufacturer Selection. Utilization of outsource manufacturing can leverage capital equipment, processing knowledge, and manpower for both start-up companies and established OEMs as detailed in reference 2. However, the selection of an appropriate and capable outsource manufacturer may not be a clear decision. This is because outsource manufacturers may not have all the desired capabilities for taking a given device from prototype to manufacturing. Table 1 provides a general guide for assessment of outsource manufacturers. A near fatal pitfall in taking a photonic device from prototype to manufacturing is to select an outsource manufacturer that can not meet the device assembly needs and the device market needs. An outsource manufacturer without the necessary capabilities, either in available equipment, processing knowledge or appropriate development staff and production staff will not be able to perform to the needed expectations. Development costs in such a situation could become too large to bear in addition to the associated delay in device manufacturing readiness.

Table 1
Evaluation Guide for Selection of Outsource Manufacturers

 

Competency

Questions

Optical Modeling

Configurations modeled?

Thermo-Mechanical Modeling

FEM capability in-house?

Die Attach

& Wire Bonding

Automated equipment?

Placement accuracy?

Substrate Fabrication

In-house thick film or LTCC production?

Layout capability?

Submount Design

& Fabrication

In-house fabrication?

Silicon micro-optical bench or v-groove assembly capability?

Hermetic Seal

& Test

Seam seal equipment?

Fine & gross leak testers?

Fiber Metallization

In-house or outsourced?

Quality control methods?

Fiber Pigtail Preparation

In-house or outsourced?

Arrayed fiber?

RF Design,

Manufacture & Test

Engineering staff depth?

Test equipment in-house?

Circuit modeling capability?

Lightwave Test Measurement

Capital equipment on hand? 

Sufficient test engineering experience?

Fiber handling experience?

DWDM test capability?

Component & Device Burn-In

Ovens available?

Tooling & test equipment?

Precision Alignment Equipment

Commercially available equipment only? 

Ability to deliver closed loop controlled nanostage alignment stations in-house?

Epoxy Attach

UV cure experience?

Cure ovens?

No-Clean & Fluxless Solder Attach

Reflow ovens, cover gas or vacuum ovens?

Demonstrated precision attach & yield data?

Submount assembly soldering history?

Laser Weld Attach

Production experience?

Weld alignment stations in-house?

Weld configurations used?

Process maturity & understanding?

Telecom Production Experience

Telecom volume production history?

Related electronic assembly experience?

QA competency?

Telecom Product Qualification Experience

In-house qualification capability?

Engineering derisk capability?

Ability to Provide Custom Processing Solutions

Depth of engineering team?

Process engineering, mechanical tooling & test equipment?

Cost

Overhead rate?

Development support cost or production cost?

Management Responsiveness

Is your business important to the outsource manufacturer?

 

7) Not Designing the Device for Manufacturability. Photonic device assembly is limited to a number of equipment vendor solutions and proprietary home-grown solutions. Assembly equipment and tooling used may well be different at various outsource manufacturers. The lack of packaging standards in the photonic industry exasperates the situation for new device developers as there are no standard interfaces to which to design devices for compatibility with manufacturing equipment and processes. Wherever the device is assembled, the device should be designed for compatibility with the alignment and attachment equipment. Tooling refitting of the alignment and attachment equipment can be substantial, timely and require processing changes particularly with alignment software programs. It is more cost effective to design the device for compatibility with available assembly equipment and processes than to develop or modify assembly processes and equipment to meet unique device requirements. Design of devices for manufacturability extends beyond the more complex processes of precision alignment and attachment to the assembly processing of submounts and subassemblies. Designing devices for manufacturability requires defining the manufacturing process, equipment and tooling constraints. Assembly process sequencing and attachment materials compatibility are part of these constraints. Not designing a device for manufacturability is a clear pitfall leading to poor assembly yields and higher manufacturing costs.

8) Not Performing Derisk of Device Design, Components and Assembly Processes Prior to the Start of Qualification Build. The perceived pot of gold awaiting at the end of a device qualification rainbow can lead to pitfalls by rushing a photonic device into qualification testing without addressing derisk of the device design, components and assembly processing. Derisk activities prior to qualification are a fraction of the cost of any problems encountered during qualification testing. The potential impact upon device availability to the customers, market perception, and investor confidence far outweighs the prudence of investing in such derisk activities prior to the qualification build. Examples of derisk activities include:

thermal cycling stressing of device and subassemblies for mechanical fatigue, alignment shift, loss of optical coupling and device performance parameters

development of weld gap tolerance control and PWS correction methods for laser weld attachment

repeatability assessment for precision solder attachment

verification of package hermeticity over Telecordia environmental exposures

The use of proxy samples, such as device submounts and packages, allows for effective derisk activities at lower materials cost. A lack of device samples for assembly process development and for process tooling is a pitfall that is very common in taking photonic devices from prototype to manufacturing. A sufficient number of prototype devices should be budgeted for assembly process, tooling and device design derisking; and if possible adequate schedule time should be allotted.

9) Not Considering Assembly Yield as a Parameter for Packaging Design and Processing Decisions. Assembly yield is often the largest cost driver in manufacturing photonic devices. While device bill-of-materials (BOM) are closely evaluated for new device designs, anticipated assembly yields are at best assumed. A common pitfall in taking a photonic device from prototype to manufacturing is to not consider assembly yield as a parameter to influence the device design and assembly process choices. Including assembly yield in the device design requires an in depth knowledge of assembly equipment, assembly processes, assembly materials and their influence upon the device performance. Device designers simply do not know these assembly process trades due to lack of assembly process experience. Integration of assembly process engineering early in the device design solves many assembly yield problems. Devices with high placement tolerances are more susceptible to assembly yield problems related to the packaging design. An example of one assembly process where yields are often the largest cost driver is in the repeatability of optical coupling of edge emitting lasers to SMF.

10) Inappropriate Time Sequencing of Tasks. A natural progression of technical and marketing tasks in taking a photonic device from prototype to manufacturing is given in Table 2. The time sequencing flows from the top to the bottom. Change in the order of tasks, overlap of tasks or elimination of tasks due to market pressure, investor pressure or management desires can lead to pitfalls that expand development cost and increase schedule delay. This is because necessary accomplishments and milestones that lay the foundation to proceed to the next task become either delayed, only partially completed, or missing entirely. Identification of the necessary exit criteria for each of these tasks during the planning stages and adhering to meeting these exit criteria helps avoid the pitfall expenses of inappropriate time sequencing of tasks.

Table 2
Roadmap for Taking a Photonic Device from Prototype to Manufacturing

 

Roadmap Sequence

Goals

Proof-of-Concept Build

Seed funding.

Demonstration of potential design performance.

Early Prototypes

Collection of customer requirements.

Definition of market needs.

Marketing plan.

Initial investor confidence.

Extensive Testing of Prototype Performance

Design evaluation; redesign as necessary.

Definition of Component Specifications

Device performance specifications.

Selection of Component Suppliers

Strategic decision for critical components.

Selection of Development Outsource Manufacturers

Strategic decision affecting delivery, development cost and schedule risks.

Leverage investment budget.

Derisk Components

Performance variation measurements.

Feedback to suppliers to control quality.

Design for Manufacturability with Outsource Manufacturers

Design subassemblies and attachment sequencing for compatibility with available equipment and processes.

Component & Device Drawings Complete

First design freeze.

Prototype First Article Builds

First customer samples.

Customer interest.

Investor risk assessments.

Derisk Activities on

Assembly Processes

& Product Design

Eliminate sources of device performance limitations, assembly process variation, yield loss, and qualification failures.

Process Documentation Complete

Enables processing traceability.

Prototype Builds

Customer samples.

Investor confidence.

Product Qualification

Customer confidence.

Selection of Production Outsource Manufacturers

Readiness to deliver.

Assessment of cost versus ability to deliver.

Production Build

Customer orders in hand.

Ability to deliver product.

 

 

Conclusions These common 10 pitfalls described in taking a photonic device from prototype to manufacturing can be avoided with awareness of the issues highlighted. Key elements in avoiding these pitfalls include:

understanding component placement tolerance requirements to meet the device performance for its optical design

using these tolerance requirements to determine component attachment methods

designing the device packaging for assembly process manufacturability

performing derisk activities on components and assembly processes

Knowledge of these pitfalls can enable start-up companies developing new photonic devices to avoid cost explosion and schedule delay in their product developments. These shared experiences also apply to delivery capability at outsource manufacturers and to cost competitiveness at established OEMs.

References
1. Shannon G, and Palen E., "Laser-Weld Attachment Enables Repeatable Submicron Precision,"
Optical Manufacturing, May 2002.
2. Palen, E. "Why Outsource Optoelectronics Manufacturing?" MEPTEC Report, July/August 2001.

EDWARD PALEN is a consultant specializing in optoelectronics-packaging design and manufacturing.

Edward Palen, Ph.D.

phone: 415-850-8166

email: epalen@earthlink.net
/font>

--return to home webpage--

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