Joe,
Here's another article about flip-chip development posted today on CMP Inc., which notes that companies are now finding technlogies to expand flip-chip beyond ceramic substrates into plastic and other forms (i.e., answers part of the rhetorical question in the last line of the article above which you posted Joe) That is important as then these new types of substrates can be glued to a circuit board-i.e., expands the use of flip chips to the printed circuit board and surface mount ciruit board realms...
Perhaps Smartflex was just a little ahead of the technology curve in the flip chip arena...
Sincerely,
Doug F.
prime time
By Terry Costlow
EAU CLAIRE, Wis. -- Flip-chip technology seems finally set to come to market, as companies ready major product launches and as the infrastructure for manufacturing with flip-chip methods falls into place.
W.L. Gore & Associates Inc. this month revealed a development that could help dismantle a remaining hurdle to wide-scale implementation: the availability of high-density substrates for flip-chip assembly.
Gore executives told EE Times that the company is set to unveil a laminate technology for very-high-lead-count flip-chip packages. The approach allows flip chips to be put on ball-grid-array substrates for easy attachment to a circuit board.
The company will ship the laminate next year and has signed licenses with two large Far East providers of package substrates, company executives said. The laminate itself is a variation of Gore's low-dielectric-constant Speedboard material.
Flip-chip assemblies provide high density without increasing total product size beyond the perimeter of the chip, and the shorter leads of flip-chip configurations increase product speed. But most of the flip-chip approaches available today employ ceramic substrates, which can be costly and difficult to attach to laminate circuit boards. If flip-chips are to see their predicted success, industry observers say it must be housed on organic materials, such as laminates.
"One of the things that is still missing in the infrastructure is high-density substrates," said Luu Nguyen, senior engineering manager at National Semiconductor Corp.'s packaging lab (Santa Clara, Calif.), which is readying flip-chip approaches for a number of internal and federally supported programs. "There are some becoming available from overseas companies, but there's nothing yet from domestic suppliers."
Gore is not the only vendor readying organic substrates, however. Nguyen noted that Sheldahl Corp. (Northfield, Minn.) has developed substrates that are being used by many in studies by the Defense Advanced Research Projects Agency (Darpa). Sheldahl has built a factory that will begin producing its highly publicized flexible substrate in volume sometime next year.
While few chip companies have gone public with their flip-chip plans, rumors of pending launches are making the rounds.
"We think Intel's P7--the Merced--will start out as a flip-chip package," said Jan Vardaman, president of TechSearch International Inc. (Austin, Texas). "Because of the high I/O the chip has, it will probably be in an area-array flip-chip format."
Nguyen was more forthcoming, reporting that National is "involved in several programs internally and with consortia supported by Darpa. We have several programs, and many of them will ship fairly soon."
National is focusing on analog devices with fairly low lead counts. But flip-chip's largest impact is likely to come from its use in putting high-lead-count devices in ball-grid-array (BGA) packages. Other IC vendors are moving forward quickly with such high-lead-count programs.
"We're building two parts now that have over 1,000 leads, and we're doing prototypes on several parts that have up to 1,800 leads," said Linda Matthew, product marketing manager at LSI Logic Corp. "The primary driver when we got into this was speed, but cost comes right on the heels of that. In the workstation and supercomputer market, the driver tends to be performance. In telecom, they have a lot of signals, but performance isn't as high, so their driver is cost."
Approaches such as Gore's pending introduction let system manufacturers work with larger spacings that may be handled with common circuit boards, without sacrificing flip-chip's density and speed advantages.
Gore's introduction marks something of a coming out for the Wisconsin part of the company. In 1993, the packaging supplier bought the remains of SSI, a supercomputer company that was spun out from Cray Research and had hefty funding and technical support from IBM. Many of the designers who created a 78-layer board for SSI in the early 1990s also had a hand in developing the high-density, low-dielectric-constant material that Gore will introduce later this year.
Even some designers who worked on the supercomputer project are surprised at how quickly such technologies as high-density flip-chip are moving from the fringes to the mainstream of electronics design.
"We're looking at anything from 600 up to 2,800 leads," said Bill Petefish, technology leader at Gore. "We're seeing interest in the high end of that this year. This surprised us; we didn't think people would get to that level this fast."
Vardaman said a broad movement from wire bonding to flip-chip attachment will take place over the next few years as laminates which support higher densities come to market. "About 3 percent of all flip-chip devices will be mounted in packages this year, but that share is projected to increase to more than 27 percent by 2004," Vardaman said. "Computers will grow from 7 percent of all flip-chip usage to 31 percent."
Chip makers are virtually compelled to adopt flip-chip to combine density and high performance. But while chip companies are working closely with substrate providers to get flip-chip into higher volumes, system designers are staying away from it.
Despite much discussion, a movement to attach flip-chip directly to circuit boards is still in its infancy. Getting the density needed to route signal lines to flip-chip solder bumps is still beyond the capabilities of most board producers.
"Flip-chip is being used more and more, but a lot of people hate it," said Jack Balde, senior consultant at Interconnection Decision Consulting (Flemington, N.J.). "The driving force is essentially that no one can put flip-chip directly on big circuit boards. Lead counts are going up, so they have to do [flip-chip], but the system houses want an alternative to flip-chip that they can put on the boards they're making today."
Gore's forthcoming material is capable of 40-micron lines and spaces with blind vias as small as 25 microns. Those figures are much finer than the 300-micron feature sizes provided by the buildup materials that are now being used to achieve high density with laminates.
Gore produces only about 50,000 IC substrates per month, shipping them to 10 companies. The company will go into full production near the end of next year, about the same time as its licensees start shipping substrates, Petefish said.
Gore's material will be priced at about 2.5 to 5 cents per pin, depending on the size of the package. While that's higher than some alternatives, the end cost will be below those of ceramics.
"Those prices are similar to ceramics, but in some cases ceramic is four times as much," said Daniel Johnson, business leader at Gore's Elkton, Md., facility. "Even when we're at the same price, ceramic can't be soldered to the board like our substrate, so you need a socket."
The technology is expected to find use primarily in high-end products. "The first applications for flip-chip in package will probably be in workstations and high-end telecom products," Vardaman said. "A lot of these applications are not as much cost-sensitive as performance-sensitive. People are talking way above 1,000 leads, up to 2,000 leads. The volumes at that density aren't very high, but the dollar value of those parts and packages is pretty impressive." |