July - August 1999
eMail the Editor
Still at the Head of the Class
Although fine-pitch lead bonding and flip-chip bonding are beginning to carve out market share, wire bonding will continue as the main interconnect method well into the next century.By Leon Oboler
 Figure 1. The K&S bonding head on this Model |
Every advance in packaging technology has placed new demands on interconnect processes, equipment and the infrastructure. In the past, high leadcounts were the technology drivers, but now
the price/performance/size demands created by smaller packages are leading the way.
This trend has been accelerated by the development of chip-scale packaging, with its rigorous demands on wire looping and the need to develop special fine-pitch lead bonding systems.
Drivers
While package size is important, it is no longer the exclusive reason manufacturers are looking at CSP and NearCSP (NCSP) packaging. Performance is now another concern. RAMBUS redefined the requirements from a packaging standpoint, with SDRAM and linear logic, increasing the pressure.
Increased device clock speeds require low-inductance packaging, which, in turn, points to chip-scale packaging solutions. For other applications there are cost/performance trade-offs.
The cost/performance alternatives breakout as follows:
- For primarily cost-driven applications, the choice is obviously the most affordable option. This means using as much existing capacity as possible and using existing solutions to limit investment costs.
- Where the application requires high performance and the lowest possible cost, the packaging technology choice is not always clear. Low-inductance- driven packages such as DRAMs and RAMBUS require high performance. But since the DRAM is a commodity, whoever has both lowest cost and best performance will provide the winning solution.
- For some applications, only performance counts-assemblers of packages with a great number of I/Os may be ready to pay a premium for precision fine-pitch bonds-in the form of an advanced single point TAB package or a wirebonded CSP or NCSP.
Where does this leave the equipment leaders like Kulicke & Soffa and ESEC (Figures 1 and 2)? Is all this new CSP and NCSP demand going to mean vast numbers of new bonders? According to industry leaders, they expect an approximately 50/50 mix of new and retrofitted older bonders to meet increased production requirements.
A brief survey of major equipment manufacturers shows a high level of confidence in the future of wire bonding as the interconnect technology that will continue to support CSPs and other small packages into the first decade of the 21st century. The assumption is, however, that other technologies, such as fine-pitch lead bonding and flip chip, will constitute important parts of the mix.
Figure 2. The ESEC 3008 wire bonder offers extremely
low wire looping.
The key elements for interconnecting such parts are:
- The wire bonding part cannot use up too much space on the substrate.
- Wires must stay as close as possible on the die.
- Wire bonds must have an extremely high level of repeatability and very fine pitch.
 Figure 3. Proprietary algorithms combine with advanced |
Unique Algorithms
Some types of CSPs, notably the Tessera µBGA and TI's MicroStar BGA, also use conventional wire-bond technology, but are bonded using special looping software and bottle-necked capillaries (figure 3).
These CSP and NCSP needs have resulted in the development of a variety of unique looping algorithms that move the wire straight down the edge of the die as close as possible to the die itself.
Availability of these extremely short loops can allow package designers to reduce the size of packages.
Package Solutions
The second stream of small package solutions is exemplified by the even smaller ratio "classic" lead-bonded CSPs (Figure 4) represented by the Tessera µBGA package. They require recent-generation wire bonders that have been modified with special software and hardware to enable single point TAB (sometimes called µTAB).
Many recent-generation wire bonders, already in the field, can be upgraded to perform single point TAB, although earlier-generation equipment, such as the K&S 1484, cannot. The current crop of products specifically marketed as lead bonders, including the ESEC 3018 and the K&S 8070, are repackaged modifications of wire bonders, specifically the ESEC 3008 and the K&S 8020.
A third emerging CSP or NCSP packaging stream is wafer-level chip-scale packaging (WLCSP), which places additional demands on the wire bonder, including the need to use wafer mapping to identify which die should or should not be bonded.
An advantage of wafer-level CSPs is the throughput benefit that comes from reduced package indexing. A large bonding area WLCSP bonder, such as the Palomar CST6000, needs to index only once for every 500 or so die (depending on the number of die per wafer).
One difficulty with wafer-level bonding is the instability of a laminated polyimide surface, which requires that touch and bonding forces be carefully controlled to avoid damaging to the wafer.
Currently, many CSPs are lead bonded, but most major equipment manufacturers see wire bonding taking over as much as 50% of the small package market, with lead bonding dropping to 40% or less in the next five years. The remainder of die in small packages is expected to be handled with flip chip. Estimates for flip chip range from 10% to as much as 50% of the market, up from the insignificant market share held by flip chip today.
 Figure 4. This illustration shows the typical single point TAB bonded leads on a CSP device. The leads are broken from the matrix, positioned and ultrasonically welded, using a specially designed tool. (Source: ASM Pacific) |
Production Levels
This is the year that the CSP is beginning to reach production quantities, according to most reports.
"We're a year into a CSP production level of 300-400 million units per year-and we expect that number to double within the next few months. Within five years, the CSP market looks to be as large as 4-5 billion units per year-which represents a compound annual growth rate (CAGR) of 80%," according to Christian Rheault, Kulicke & Soffa's strategic marketing manager for assembly equipment, Willow Grove, Pa.
What does this mean for the manufacturers of wire and lead bonders that support this growth?
Does it mean new interconnect technologies on the horizon and portend the often predicted end for wire bonding? Probably not.
The wide variance in estimates reflects the many different paths available and the uncertainty of what may be hidden in labs throughout the world.
As Peter Buehlmann, ESEC business unit manager for wire bonding, noted, "I'm sure there are new things in the labs. Whether they will survive no one knows. However, the future is in substrates with die and conventional wirebonding-with very short loops."
Conclusion
As device manufacturers review their processes, and balance their price/performance needs while looking to the future, they will be comforted to learn that current platforms for ball bonding are the same for CSPs and regular bonding.
While there are some new software features needed for CSPs, such as low loop special trajectories, as well as some simple hardware changes, advanced bonders designed for CSPs can also be used for other advanced packages. This equipment can even be retrofitted to perform single point TAB.
Mr. Oboler spent more than a decade working for Kulicke & Soffa in Willow Grove, Pa. Readers may contact him at Oboler Communications, leonobol@voicenet.com or by phone at 610.635.8800.
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
