BY GAIL FLOWER, Editor-in-Chief
In the March 2007 cover story, we asked the industry about the demands that new packages placed on sockets, lead-free issues, standards, cost control, fine pitch, and the overall consensus on the Burn-in & Test Sockets Workshop (BiTS) where users talk to suppliers. Part II covers sockets as an industry, thermal issues affecting sockets, measuring burn-in socket life, and perplexing technology issues surrounding sockets.
Socket Industry Forecasts
How is the socket industry doing overall?
The industry is tending to merge a bit. According to Jamie Andes of Synergetix, there are, and will continue to be, mergers and a reduction of players in the burn-in and production test socket area. Valts Treibergs of Everett Charles Technologies (ECT) added that mergers, acquisitions, name changes, and growth will continue for sockets. The only certainty is change.
At Antares Advanced Test Technologies - formed by acquisitions of WELLS-CTI in 2003, DB Design and ACI in 2005, and merging with Antares conTech (a spin-off of Kulicke & Soffa) - change has been quite significant. Mark Murdza, director of marketing, Antares, adds that the industry is strong and changes will continue, especially consolidation in the burn-in and socket realm. Reasons for consolidation abound. Anyone can make a spring-probe test socket, but suppliers with value-added variables that customers need will survive. The need for alternative solutions to test next-generation devices with higher speed and power is another reason for innovative interconnect technologies. Many companies cannot afford the R&D expenses needed to generate these solutions; it takes horsepower in global markets. A socket supplier must be able to service and support the business anywhere in the world where a customer’s production requires it. Thus mergers and acquisitions make sense in today’s economy.
Though some see change, others reply that the burn-in market, after year 2000, has remained more or less constant. Franz Rosenberger of Yamaichi Electronics explained that significant continuous growth cannot be expected because products have built-in self test (BIST) more often, and the functionality is guaranteed by process control.
Figure 1. A liquid-cooled spring-loaded heatsink. (Photo courtesy of Gryphics, Inc.)
Bob Fenton of Gold Technologies, Inc., mentioned that his company has a significant OEM business outside of the semiconductor arena that, beginning in 2005, helped them out in terms of revenue. The semiconductor test area has been profitable with incremental growth for Gold in 2005 and 2006. Fenton sees some indications of a slowdown in their Tier I semiconductor customers, though orders have remained the same as in the fall 2006, and are improving incrementally.
Markets adjust over time and many socket suppliers sell products outside the “normal” range. Therefore, making a blanket statement is not as useful as knowing the capabilities and strengths of each company within the industry.
How do sockets handle rising thermal issues?
Thermal management of a BGA socket involves material selection, design, analysis, optimization, and verification of a cooling system for producing reliable sockets for testing high-power devices according to an expert at Ironwood Electronics, Inc. These include thermal interface material (grease, silicon pad), heatsinks made of finned aluminum on the IC’s copper spreader, and other elements to control rising temperatures.
Steven Bozsi of Bozsi Engineering notes that there seems to be a trend toward higher current capability while still requiring ever-higher pin counts. Some manufacturers use a variety of different types of pins in the same socket based on current and temperature requirements.
Fenton reports that Gold takes the same approach to rising thermal issues as to other challenges, combining new materials and engineering sets with testing and validation. Many of these issues are due to increased current density and testing methodologies. For example, the move to array or strip testing, and the use of conductive thermal conditioning in strip-test handlers, has been a thermal challenge their firm overcame just last year. For a device with many I/Os to surface area, such as a BGA or LGA, if you do nothing but build a standard socket, the handler will not keep the DUT at temperature without an unacceptably long soaking time. They solved the problem by adding supplemental heating or cooling as needed to the socket.
Figures 2a & b. Crown-tipped probes working with oxidized material must be inspected and cleaned for optimized effectiveness. (Photo courtesy Nu Signal LLC)
Most burn-in suppliers report some form of heat control. Rachel Lufkin of Aries Electronics explains that their burn-in socket material is plastic with high thermal characteristics, and that heatsinks are also added if required. Yamaichi has active- and passive-cooled socket systems, some with each device individually sourced and controlled. And, of course, heatsinks are included where needed. Antares also reports that customers can individually control device temperatures during burn-in using their existing burn-in equipment. The company offers high-power device thermal control products in test applications from benchtop through production. Heatsinks are used more with the advent of finer architectures, increased current leakage, and higher power densities. Therefore customers may require more control than is offered by standard passive heatsinks.
Knowing what each application requires dictates the approach a socket supplier takes to handling increased heat. ECT uses an advanced modeling and thermal simulation tool as well as lab testing to measure socket performance at temperature and high-current extremes. Working closely with the customer to specify and provide passive and active thermal manual actuators for sockets, as well as provide thermal flow channels with sockets, helps stabilize temperature.
Jim Rathburn of Gryphics, Inc., says that his company has a selection of heatsink styles: CAM-actuated spring-loaded passives, some with fans, and many other configurations (Figure 1). In the user arena, Frank Navarret of National Semi Corp. uses heatsinks for high-rel devices.
Materials, such as Honeywell’s burn-in thermal interface material (TIM), are used to maintain thermal performance for more than 1,000 cycles. More often than ever, materials, heatsinks, and new designs take the heat.
Measuring Burn-in Socket Life
How do you measure burn-in socket life?
Measuring the life of a burn-in socket depends on the customer’s specific usage and requirements. Rosenberger says that Yamaichi’s sockets are designed to be rigid mechanically and that end-life can be predicted by electrical behavior, which is influenced by temperature, contact alloy, and usage.
The real issue with burn-in is not the socket life but the cost of burn-in, says Roger Weiss, Ph.D., president and CEO of Paricon Technologies, Corp. Several device manufacturers are transitioning to strip burn-in, which reduces cost on several fronts: the number of components per test board can grow to as many as 800. This reduces cost of board, energy, and unit socket. Net/net, the strip burn-in may reduce the cost of burn-in by a factor of as much as 10×.
At CVInc., Terence Collier uses a tool that the company developed to supply electrical load and heat to simulate what an individual pin will encounter. From this data he estimates the life of a given design. CVinc. also takes customers’ sockets and runs simulation to estimate the life or time between preventive maintenance (PM). That data can be generated on a single pin or an entire socket. CVInc. runs failure analysis (F/A) on a socket to determine if the socket is the cause of poor yield. The socket is placed in the test fixture to toggle target pins and from that generate a profile of good versus bad pins. One also needs to know the impact of sockets and probe contacts on the DUT, adds Collier. This information can be modeled, but it can also be gathered in real time to provide hard data.
Users keep track of burn-in life. At Micron they monitor yield per site and make decisions based on pre-selected criteria. At National Semi an inventory database keeps track of the amount of hours that burn-in uses. This is tracked once the boards are released and 100% verified prior to use.
ECT builds burn-in sockets if a customer specifically requests it. However, for test contactors, they have a regimented lab test qualification plan for new interconnect designs, using robotic flying probers, pneumatic cycling machines, high-current pulsing apparatus, and a host of RF measurement tools to qualify socket signal integrity.
Cycle times vary per manufacturer. Fenton says that, though his company doesn’t do burn-in, they specify burn-in sockets at 20,000 insertions or more and have never had an issue with the specification.
Figure 3. Strip testing for multiple devices increases efficiency. (Photo courtesy Paricon Technologies Corp.)
Olek Cymbalski of OPC Technologies says that many companies supply his firm with high-end test sockets claiming more than 500,000 insertions. The sockets work well, but cost as much as 50× the average socket. They require extensive maintenance to remain at peak performance. Key elements of the technology are crown-shaped probe tips that make contact with the sides of the solder ball (Figure 2). Heavy spring force is used to displace oxidized surface layers and make good electrical contact. With regular preventive maintenance and cleaning, these sockets deliver a long service life.
Antares’ burn-in sockets go through a validation testing sequence to emulate the end-use application, and that is what they use to measure burn-in socket life. The objective is to validate the socket integrity through multiple burn-in cycles. Additionally, all burn-in sockets must withstand 10,000 mechanical cycles to be considered a reliable product, adds Murzda. Withstanding 10,000 cycles has become a de-facto industry standard.
Challenges in Technology
What are the most perplexing issues?
In this area, most participants in our survey tended to agree. The challenges today include: cost, high-densities, fine pitch, high power, lead-free, and thermal concerns. Advances in system performance, needs for high-speed memory, RF, wireless, and other applications drive demand for low-cost socket technologies that can deal with lead-free plating and low insertion loss, says Gryphics’ Rathburn. Tight tolerances and chips that don’t follow JEDEC standards add challenge, notes Aries’ Lufkin. Coming up with a design that eliminates the trade-offs between higher current-carrying capacity, low resistance, high frequency, greater contact travel, finer pitch, low inductance, low cost, and mass production is important, adds David Pfaff of Plastronics Socket Co.
Specifically, for engineers in the field, problems revolve around designing a reliable socket that can perform for an extended time at elevated temperatures. Trends seem to show these stress temperatures have been increasing to as much as 200°C and above, says Bozsi.
Weiss noted that the worlds of test and burn-in are coming together, with test being performed during burn-in. The speed of devices and contact pitch have moved beyond what has historically been economically feasible for conventional spring pin and formed contacts. This is compounded by the growing need to test components at actual performance. This must all be done while providing cost-effective solutions. These industry needs are the drivers for new technology.
The socket, board, and IC device must be optimized as an integrated assembly says Treibergs. To build a better socket that meets these needs, the designer and test engineer must come up with optimum spring loading for a given probe-tip geometry to determine the sweet spot for stable electrical contact. They must evaluate contact resistance (CRES), intermetallic formation at the probe tip impact, as well as contamination on the DUT pads, says Collier. Life testing is a must. Incoming inspection of assembled sockets must be done. And the impact of temperature on performance and components must be considered.
Sockets are vital to meeting the needs of future electronics. We thank the users and suppliers of burn-in test and production sockets for their participation in this survey. The better we understand what future technologies require, the more we can prepare to meet the needs of tomorrow.
The author would like to thank the following companies for their contributions to this article:
Advanced Interconnections Corp.
West Warwick, RI
Antares Advanced Test Technologies
Aries Electronics Inc.
Everett Charles Technologies
St. Paul, MN
Gold Technologies, Inc.
San Jose, CA
Kansas City, KS
Micron Technologies Inc.
Santa Clara, CA
Grass Valley, CA
Fall River, MA
San Jose, CA