The suppliers of test and burn-in sockets surveyed for this article predict a strong market for the year ahead, fueled by a growing demand for memory, telecom and ICs for portable electronics. As a subset of this growth, chip-scale packages will continue to be a key socket industry driver-but now cost of ownership is beginning to play a larger part in the overall socket-buying equation. Technology, too, will push the industry toward finer pitch sockets for finer pitch devices. (Figures 1-6 show representative sockets.) Although we counted more than 24 suppliers a year ago, we "lost" four in that period to acquisitions: 3M bought Tronix; DCI absorbed Liberty Research; ECT acquired PrimeYield and OzTek became part of Cerprobe Inc. Although there may now be slightly fewer players, it's likely that vendors are going to face a more competitive (read "price sensitive") environment in the year ahead. While uniformly optimistic, socket vendors broke ranks on how much the market for their products will grow over the near-term. Responses varied from "considerably" to a specific figure of 35%, forecast by Aries Electronics, Frenchtown, N.J. Nearly all responses, however, cited-to one degree or another-the "explosion" in handheld devices. Consumer Demand Bruce DeChillo, Aries' product manager, says growth will lean toward smaller packages, such as MLFs and CSPs. "This growth will be driven by the enormous consumer demand for mobile communications. Leaded device test sockets are still in strong demand," DeChillo adds. "These will soon plateau, however, as leadless devices-with smaller footprints and lower price tags-take center stage."
"Moreover," Langston adds, "I believe the systems houses-especially those driven by net-centric computing applications, are squeezing more integration and capability into new packages, mostly CSPs." Second, says Langston, the consumer market is driving the convergence of wireless and network applications, such as stock quotes and global positioning, "so the need for more miniaturization is compelling. This translates into more CSPs and more sockets." David Pfaff, vice president of the Plastronics Socket Co., Irving, Texas, estimates a "short-term increase of 20-30% in burn-in socket demand, based on a higher number of wafer starts this year." Brian D. Abrams, 3M Sockets Business Unit manager, Austin, Texas, looks for more customization "resulting from specialized needs, such as thermal management." Abrams believes the focus must now move from the cost of socket acquisition to the cost of ownership. "During the socket selection process, engineers will look to products that include robust designs, high performance, accurate data recovery, low maintenance and long operational life to maximize test floor productivity." High-Volume Manufacturing As competitor Ed Craig, sales and marketing manager for Texas Instruments Interconnection Business, Mansfield, Mass., observes, "As a supplier of burn-in sockets, the ability to transition seamlessly from product development to high-volume manufacturing is a must. If you miss the initial ramp, you won't see a return on your investment." There was little agreement by socket suppliers on technology direction. Most concurred, however, that finer pitch and the increased emphasis on functional test at burn-in will play a role in their business plans. "More and more customers are asking about doing functional test during burn-in," notes Joe Bunch, product specialist at Synergetix, Kansas City, Kansas. "This will drive the need for burn-in socket makers to reconsider their contact methodology," says Bunch. "Low cost, short-signal-path contacts will have to be developed to fit the need." Scott Novak, worldwide director of sales for ECT's Semiconductor Test Group, agrees that the increased emphasis on test at burn-in will reduce the market for older, low-performance contact technologies. "Short contact, high-bandwidth probe technology will be needed to maintain the signal integrity required by the increasing functional integration and built-in self test demands at the device level. "Manufacturers will seek longer cycle lives and generally lower costs of ownership for these types of ATE interfaces," Novak adds. "With the competition for wireless technology devices driving prices down, alternative, less expensive test and burn-in practices will be explored," he says. 6-10 GHz Range As several socket vendors were quick to point out, including Jim Brandes of PrimeYield Systems, St. Paul, Minn., "Contactor manufacturers will focus on supplying sockets for the last generation of communications devices in which signal frequencies are extending into the 6-10 GHz range." Karen F. Lynch, Cerprobe's marketing vice president, Gilbert, Ariz., agrees that >10 GHz will be the socketing industry's goal for communications ICs. Additionally, Lynch says, "CSP pitches are moving down to 0.4 mm for communications and memory applications. The need to provide the complete interconnect solution for controlling the electrical signal from the IC to the ATE is becoming a stronger requirement, too."
Plastronics' Pfaff sees burn-in focusing more at the die and wafer level. While this technology has been developed over nearly a decade, major obstacles still exist, he says. "Contacting the small die pads, the number of contacts needed in a small area and the exact method of locating the die or wafer are major socketing issues." Process issues, he adds, "include the excessive amounts of heat produced in a small area, as well as the difficulty in monitoring all the signals on a single wafer with existing burn-in systems." Thermal Management "The industry will see a tremendous advance in thermal management for burn-in and a scramble for a reliable, low-cost, fine-pitch interconnect solution," according to Mark Murdza, product marketing manager for Wells-CTI, Phoenix, Ariz. Murdza contends that the "next year will bring 0.5 mm pitch as commonplace, and we will begin to see experimentation with 0.4 mm and even 0.3-mm pitch. The trend towards sub-0.5-mm pitch will require an advanced interconnect technology beyond the standard stamp-and-form contact, or even spring and probe."
Emerging smaller devices used in commercial applications, specifically the handheld market, says Knudsen, are driving speed and size. "Contactors need to interface to these small, high-frequency packages, including several new LCC-type packages with and without body grounds." Although package footprints continue to shrink, package thickness will increase due to component stacking, he says. Ariane Loranger, sales and marketing manager for Loranger International. Warren, Pa., notes that advances in microprocessor speed and I/O requirements are "continuing to press the engineering creativity and capacity of the socket designer. Pincounts of 1000 I/Os or more are coming out faster with smaller pitches." Implementing Rambus Now that Rambus' RDRAM technology has been officially implemented by such giants as Intel Corp., the reaction from socket vendors is mixed, with some seeing an increase in Rambus RDRAM socket orders. Others, however, say RDRAM interest has remained, as characterized by Ariane Loranger, "at the status quo." TI's Craig says orders are beginning to increase, driven by RDRAM implementation in PCs, such as those made by Dell, and, in turn, Dell gaining support from such major chip makers as NEC, Samsung and Toshiba.
Wells-CTI's Murdza sees activity surrounding RDRAM sockets increasing slightly. "Intel has announced its intention to use the RDRAM architecture with its Willamette processor. This will be the first step toward a wide market acceptance for RDRAM technology." However, the potential influence of lower-cost alternatives, such as Quad Data Rate and Double Date Rate memory architectures may work against RDRAM's wide acceptance, he adds. Competitor Aries has seen a decrease in demand for RDRAM-type sockets, compared to an increase during Q4/98-Q1/99. "We believe the reduction in demand is due to the extremely volatile memory market," says Aries' DeChillo. For that reason, he adds, "Customers do not want to make a firm commitment to use RDRAM devices in their products." Rambus' recent lawsuit against Hitachi may have made DRAM suppliers leery about licensing new products from Rambus, says Synergetix' Bunch. Additionally, a recent alliance of big names like Hyundai, Infineon, Intel, Micron, NEC and Samsung, formed to develop a new DRAM architecture for PCs, has also slowed Rambus RDRAM acceptance, Bunch believes. Conclusion
Contact the editor at chipscale@cs.com. |
