mm-wave Microelectronics Manufacturing

mm-wave Microelectronics Manufacturing

Kevin Danehy and Mark Wolf
M/A-COM, a division of AMP Inc.
Lowell, MA

The question as to whether or not the telecommunications industry will fully utilize the promise of mm-wave technology is still highly debated. However, the development and interest in commercial mm-wave products have skyrocketed. The days of performance-specific, low volume, high cost mm-wave components and systems are fading. Today’s commercial interests focus on speed, quantity and cost of mm-wave technology.

Producing mm-wave microelectronics is a challenging venture. The success of commercial mm-wave manufacture depends on the ability to implement products rapidly, maintain low manufacturing costs through automation and maintain a stringent design-for-manufacture discipline.

Markets That Utilize Millimeter Frequencies

The huge diversity in the communications marketplace from a customer, application and system architecture standpoint makes the introduction of standard products exceptionally difficult. Therefore, creating an efficient production strategy based on manufacturing high volumes of the same or similar equipment is an unlikely approach. However, economies of scale must be reached to satisfy the low cost needs of the marketplace.

The local multipoint distribution service auctions that were completed recently by the Federal Communications Commission are a case in point. Spectrum winners will be looking for turnkey system solutions that include base stations and customer premise equipment (CPE). Potential volumes for CPEs are huge (greater than 500,000 units per year globally), yet it is unlikely that the market will adopt such a standard. Instead, system integrators will develop their own architectures, protocols and electrical specifications, thereby requiring distinct designs for subscriber units.

The point-to-point radio link market for telephony and data traffic is another good example. Globally, there is an increased density of cellular base stations in urban areas that has required service providers to utilize millimeter frequencies (18 to 50 GHz) to support the high capacity links between base stations and the telephone network. This market has expanded radically and today’s forecast for high frequency radios exceeds 100,000 units per year. Figure 1 shows estimated radio link growth derived from industry interviews and internal sources.

While the majority of the links deployed are waveguide Gunn oscillator (cavity)-based systems, 1997 saw the introduction of many solid-state radio links that utilize different architecture approaches and offer higher reliability with lower price. As the volumes increase, these new systems will need to be manufactured on automated production lines to meet the cost targets of the market.

One final example: Automotive radar and sensors will require the production of millions of high frequency products over the next 10 years. Although this market is in its infancy, many automotive manufacturers have expressed interest in adding additional electronic features to car and truck platforms such as collision-avoidance radars, intelligent cruise control systems, parking aids and alarm sensors, which will have extremely low cost targets. As with the two markets described previously, individual manufacturers will strive to differentiate themselves in the marketplace by offering special styles, attributes and system performance, once again eliminating the development of standards.

The prices and volumes required to support these and other consumer markets will exert extreme pressure on manufacturers of millimeter products. Only by integrating the design and manufacturing teams can high volume, low cost millimeter manufacturing be realized.

mm-wave Construction Techniques

The commonality of the mm-wave technology applications outlined is the fundamental requirement for generating designs that are manufacturable and easily automated. Generally, mm-wave designers select the design platform that they are most familiar with and/or that will produce optimal electrical performance. The new low cost commercial applications add a new dimension to the equation: how to produce at low cost. The commercial mm-wave designer does not have the luxury of sacrificing material or labor costs to achieve the ultimate system performance.

Waveguide vs. Microstrip
The predominant assembly approach for mm-wave products has been the use of waveguide technology. A waveguide approach has inherent performance advantages at high frequency, including less insertion loss, high power handling capability and better signal filtering. However, automating the manufacturing process is difficult. Waveguide assemblies are typically more labor intensive to assemble due to their size and odd shapes.

Microstrip technology is conducive to automation because the circuit topologies are planar and easily accessible. The microstrip approach can be implemented even if the customer imposes restrictive geometric outlines for housings. For example, a carrier-based subassembly utilizing microstrip could be prefabricated using a high degree of automation and subsequently dropped in if necessary.

Substrate Media
Substrates such as polytetrafluoroethylene (PTFE) soft substrate and alumina are two commonly used media for mm-wave circuits. Glass Microwave Integrated Circuit (GMIC™), a proprietary glass/silicon laminate, is used as a thin-film alumina equivalent. GMIC is a good choice for high frequency. It also has the added advantage of enabling thermal management of active devices through the use of silicon pedestals. The lumped-element construction can be utilized with both glass and alumina, reducing overall component count. Figure 2 shows a simplified cross section of passive glass/silicon technology. Other traditional substrate media include AlN and BeO, which are also considered during the design phase.

Soft Substrate Handling
Rigid-backed soft substrate is preferred for assembly automation. A low dielectric constant must be utilized to maintain relatively large feature size. Glass weave is sometimes added to PTFE to improve rigidity and temperature stability, but the insertion loss of the material increases. Thin dielectric material is flimsy, making the delicate dielectric difficult to handle. PTFE circuit handling is more robust, using a rigid backing material such as aluminum.

Soft Substrate vs. Glass/Ceramic
The decision to use soft substrate or glass/ceramic depends on a number of factors. Critical parameters such as lead time, material cost, electrical specification margin and assembly yield must be considered. For example, soft substrates have the advantage of a fast material lead time and traditionally lower material cost. Soft substrate cost drivers include circuit size (material and process costs), line resolution (minimum gap and process yield), postmachining (number of vias and subsequent machining) and selective plating requirements (that is, does the substrate use wire bondable gold and/or plated thru holes).

GMIC circuits are processed on wafer and, like most foundry processes, have associated wafer yields and longer lead times. GMIC cost drivers include wafer processing costs that are dependent on circuit size and complexity (yield/number of elements), and wafer dicing, poking and handling of discrete substrates.

Soft substrate is inherently more difficult to wire bond. Thermosonic wire bonding requires heat and ultrasonic energy. Applying heat to a thermoplastic such as PTFE changes the physical properties by softening the material. This spongy characteristic makes wire bonding difficult. Poor first-pass bonding yield may negate the lower material cost that was appealing initially.

Other Soft Substrate Considerations
Typically, Cu/Ni/Au plating is used on PTFE. Undercutting of base copper and circuit trace adhesion strength have been notorious problems with soft substrate metallization. Variations in the etching processes often limit the ability to maintain the fine gap spacing that is desirable at mm-wave frequencies. The metallurgy on glass is dramatically different. Line resolution and circuit trace adhesion are less of an issue with glass or other thin-film equivalents.

GaAs MMIC vs. Discrete Device Technology

The decision to implement GaAs MMICs or discrete devices in the mm-wave design is another factor affecting manufacturability. Considerations include the availability of the product and reliability of the supplier’s process. Specifically, with higher frequency MMICs, inherent wafer fabrication process limitations, process repeatability and lack of suppliers make implementation of MMICs difficult. Higher levels of integration at the IC level tend to drive overall wafer yields down, thereby increasing MMIC unit costs.

MMICs have the significant advantage of lowering the component and wire bond count. Typically, MMICs have larger bond pads (roughly 100 mm square) as opposed to the small-geometry discrete FET bonding pads. Larger pads make the wire bond process easier. Automatic component pick and place is also generally less difficult with MMICs as opposed to small, unpackaged discrete FETs.

Product Integration

Millimeter-wave microcircuits can be fabricated in a myriad of ways. The cost of integration of the RF circuitry is linked closely to the RF design approach. Figure 3 shows the relationship of the various integration techniques to cost. For example, a popular application at mm-wave frequency is soft board (PTFE) laminate. To reach the RF ground plane in a soft board laminate chip-on-board (COB) design, the active device (such as a MMIC) must be mounted into a cavity (through the dielectric). The added high frequency requirement to reduce parasitics by keeping interconnect wires short results in placement of the MMIC in a very tight cavity. The high accuracy placement often results in slower speed, thereby driving manufacturing costs up. An alternative COB approach where a coined ground mesa is used reduces the need for die placement accuracy.

Procuring Packaged Components and the RF Transition Dilemma

Although great efforts have been expended in improving MMIC and discrete performance, similar technology improvements have not been realized in packaging these respective RF components. Paying more initially for the packaged components and using an automated, high speed surface-mount technology (SMT) assembly approach may result in lower assembly costs.

Elements Necessary for Volume Production of mm-wave Microelectronics

Design for Manufacture
Building microcircuits at mm-wave frequencies is challenging. Issues at high frequency must be considered as well as the fundamental elements for efficient manufacture. Design for manufacture is a critical element for success. Understanding the inherent manufacturing process variations and limitations is essential for the mm-wave design engineer.

Formal Design Guidelines
Documenting critical processes and capabilities helps provide a structure for design implementation. Developing the working document highlights the strengths and weaknesses of the assembly process. Establishing critical feedback from lessons learned greatly improves the effectiveness of the design guidelines.

Implementation and Investment in Automation
A distinct competitive advantage for any mm-wave manufacturer is investment in manufacturing automation. A key to lowering manufacturing cost is the selection of the component pick-and-place platform.

Traditional microelectronic placement automation equipment vendors have concentrated on two distinct markets: high speed SMT and high accuracy bare die placement (hybrid circuits). Original equipment manufacturers (OEM) have endeavored to optimize either speed or accuracy. Unfortunately, a race toward speed and accuracy together has not taken place. (Frankly, the number of customers asking for both has been small.) The recent migration to mixed SMT/COB technology at high frequency is changing how OEMs address equipment offerings and the market.

The best available option for the mm-wave producer is to capitalize on existing automation by directing projects to the appropriate work cells or by working with an OEM to develop equipment that maximizes both machine accuracy and machine speed. Table 1 lists typical characteristics of available pick-and-place equipment.

A High Speed SMT and COB Manufacturing Line
Equipment in a typical high speed SMT and COB line should include an in-line screen printer, high speed component pick-and-place machine(s) and SMT furnace. Automatic gold wire/ribbon bonding equipment should be available to support interconnects. Automatic epoxy or SMT adhesive dispensing via Cartesian robot is also desirable. Figure 4 shows a high volume SMT/COB manufacturing line.

Flexibility, speed and accuracy are key to automation success. Strategic advantages can be gained by improving component placement speeds. Adjustable in-line conveyors maximize productivity by allowing large panels of circuits to be processed in one pass. The higher the number of components that can be presented on tape-and-reel or other formats, the more component handling time is reduced. An automation line also capable of soft solder flip-chip and ball-grid array (BGA) processing allows improved flexibility for future mm-wave development.

High Accuracy, Small Component Placement Manufacturing
High accuracy pick and place certainly plays a role in mm-wave microelectronic manufacturing. A design using a high number of unpackaged discrete components is an example. Placement accuracy of ±0.0005" is ideal for this application. Of course, the trade-off is speed.

A wafer poking station(s) as an alternate to tape and reel or waffle pack provides improved capacity. It is beneficial for wafers with a high die count such as discrete FETs to be left on wafer and poked during the placement operation.

Fixturing for Automation
The often overlooked element for successful automation is fixturing. As mentioned previously, substrate rigidity is important to automation because it allows units to be processed in panel form or discretely in a carrier boat. Most microelectronic OEMs can support in-line (where machines are adjacent to one another and material is fed on a belt continuously) or magazine-to-magazine (where discrete substrates are handled in carrier boats and transported between machines via a magazine holder) configurations. Both material feed systems are totally automated, greatly reducing unnecessary touch labor.

Conclusion

The challenges facing mm-wave product manufacturers are common to all commercial microelectronic producers. To succeed in commercial markets, manufacturers must be able to produce in high volume at low manufacturing cost. The added challenges unique to mm-wave frequencies such as sensitivity to placement tolerances or control of wire bond length make understanding the inherent manufacturing variations imperative. Yields can be improved by characterizing process variability and educating the design community about the manufacturing constraints. An essential objective for a mm-wave design engineer is to select a design platform that maximizes performance while simultaneously minimizing total manufacturing costs. If this equation can be solved the probability of a successful product is high.

Acknowledgment

The authors wish to thank the members of the M/A-COM mm-wave design and manufacturing community for their respective technical assistance; Charles Howell for his marketing data; and J.P. Lanteri, Dick Anderson and Rick Perko for their graphical inputs.

Kevin Danehy received his BSIE and MBA from Northeastern University in 1987 and 1997, respectively. He has over 11 years of experience in microelectronic manufacturing. Danehy has been employed by M/A-COM, a division of AMP Inc., for seven years. Currently, he is a principal manufacturing engineer in the company’s Millimeter and Source Product Group. Prior to joining M/A-COM, Danehy worked as a microelectronic process engineer for GTE Government Systems. He is a member of the International Microelectronics and Packaging Society.

Mark Wolf received his BS and MS in materials science from MIT in 1982 and 1983, respectively. He has more than 15 years of experience in the microwave and RF manufacturing industry and has been with M/A-COM, a division of AMP Inc., for nearly 10 years. Wolf is responsible for M/A-COM’s high frequency communication products (LMDS, point-to-point and point-to-multipoint) and automotive radars and sensors. Prior to joining the company, he was part of a start-up company (Tachonics Corp.) and worked in engineering and sales.