Issue



Optimizing the Wire Bond Process


07/01/2007







Collaboration Is the Key

BY PAUL REID, K

Demands for continuously smaller IC packages with increased I/O capability are requiring tighter wire-bond loop tolerances. Die with multiple rows of bond pads, called tiers, are common in the current generation of package assembly (Figures 1a and 1b).


Figures 1a and 1b. Close-up views of wire loops in some of today’s packages.
Click here to enlarge image

Achieving this high level of precision in loops on a wire bonder takes more than simply programming the right loop shape. It requires a deliberate collaboration and choice of materials, equipment, and motion control software to create reliable and repeatable loop profiles, while still adhering to the universal demands of maintaining the highest throughput, lowest cost, and highest yield. The diversity of package and customer requirements is causing process engineers to find creative ways to mix-and-match elements of the process to achieve the desired wire-bond loop profile.

Looping Materials

The diameter of bonding wire is getting smaller for two reasons: to achieve finer pitches in smaller packages and - given the current price of gold - to lower costs. Obviously, the wire itself is an integral structural component of the end product, and there is a limit to simply reducing the size used in a package before it adversely impacts performance and reliability. Not so obvious is that optimal wire type for various packaging challenges is required to achieve the correct level of final product quality and reliability. It will be an enabler of the specific wire-bond looping challenges of each package.

For example, applications that have die-center bond pads and strict package height limitations typically require very long, low, in-board wire loops. After the first bond is placed, the wire loop must travel a long, flat span to the die edge, and then descend to the second bond very sharply as the wire goes over the edge of the die. To achieve an optimal loop profile, as shown in Figure 2, the wire type chosen must be of sufficiently high tensile strength to allow for diameter reductions without compromised mechanical stability. The choice of loop shape, along with the programmable motions and trajectory of the bonding tool to create the loop itself, must also be considered. During the travel from first to second bond, a combination of forward and reverse motions put small “kinks” in the wire that give it the intended final shape and added strength. Choosing an ultra-low loop (ULL) profile on the wire bonder achieves the optimal loop height, span length, and descent angle to second bond.


Figure 2. Example of an optimal loop profile.
Click here to enlarge image

In choosing wire from a strength perspective, there may be consideration to using copper wire instead of gold, which is inherently stronger and stiffer. Even though copper wire use is increasing in many applications, gold wire is still used in the overwhelming majority of the advanced packages under discussion. No longer can the old industry-standard 25-µm 4N wire be assumed to be optimal, particularly in these newer packages. The bottom line is that for each application, a review of the packaging and looping requirements such as pad pitch, desired bonded ball size, wire lengths, loop heights, desired pull strengths, etc. must be cross-referenced to the specific properties of gold wire such as size, tensile strength, elastic modulus, length of heated affected zone (HAZ), resistance to mold sweep, etc.

Capillary Properties

Similar to the choices in wire, more bonding tool choices are available to achieve an optimized process. The right capillary must be dimensionally specified to fit the diameter of the wire being used. However, even in this seemingly simply choice, there are some less obvious pitfalls to avoid. The internal diameter of the hole in the capillary is not exactly the same as the wire; there needs to be some room so the wire pays out smoothly through the cap as it is used. Generally, more room is better to reduce potential clogging from contaminants during production. However, too much space will affect placement accuracy. Many process engineers use the general guideline of a capillary internal hole diameter between 1.2× and 1.4× the diameter of the wire as a starting point.

After hole diameter, there are still many other decisions to make. Each capillary has more than 10 different parameters to specify (Figure 3). The capillary tip size, shape, and surface angles determine not only the gold ball size and shape; they also serve the added function of forming the second, or stitch, bond that cleanly severs the wire and allows the bonding process to start again for the next wire loop.


Figure 3. Each capillary has more than 10 different parameters to specify.
Click here to enlarge image

Process engineers should evaluate the attributes of capillaries that are designed to address specific issues of stacked die and multi-tier applications. For instance, one capillary choice* was designed for wire loop reliability and repeatability, and works the wire to minimize wire sagging. Also, for finer-pitch applications, it can be specified in a contained inner chamfer (CIC) configuration, which optimizes the bonded ball shape to provide the most pad and ball contact in the smallest space, therefore providing the strongest possible bond with a smaller ball.

Equipment Ramifications

The automatic wire bonder remains the interconnect method of choice for many reasons, such as the extent of the installed infrastructure and capacity, the bonder’s inherent flexibility for adapting to a wide variety of packages, and the low cost of the overall wire-bond process. The other key element is the continued advancement of wire bonders’ capabilities. Continuously reducing the pad pitch on die with one row of perimeter bond pads was once the mainstay of filling the demand for increased I/O in a package. This has shifted to stacked die and multi-tiered packages as they achieve the increased I/O and provide increased device functionality, while the package size continues to decrease. These new packages have shifted the focus of the primary bonder spec discussion from accuracy and pitch capability to looping control and portability. New, tighter looping tolerances are called out as percentages of the diameter of the wire itself. These requirements simply could not have been achieved in previous generations of wire bonders. The enabling technologies in the wire bonders are the servo control systems and positioning systems, and the resulting improved resolution in the X, Y, and Z axes. The motions are being made with accelerations and decelerations of more than 12 Gs and with accuracy approaching ±2 µm. The motion-control systems are, in turn, enabled by the computational power and speed of the latest microprocessors, which allow continuous real-time execution of complex algorithms that tell the bond head where to go to create the desired loop and execute the requested bonding process.

Other bonder subsystems have made equally impressive jumps in capability. The most prominent is the vision system, which locates alignment reference points on both the die and substrate, and enables the bonder to correct for alignment inaccuracies within the package. New search algorithms and increasing camera translation speeds have kept the vision system from being the throughput-critical path. Advanced vision systems can perform certain additional functions on an as-needed basis, such as looking for backup reference points if the first is not found. This could occur in the simplest cases of when a die is not present or the site has contamination covering the primary reference point. This added functionality can keep the bonder from stopping and waiting for an operator to provide assistance due to a material problem. Such capabilities directly increase the mean time between assists (MTBA), translating into increased bonder productivity, while still keeping the yield at the highest levels, ultimately lowering the overall cost of production.

Conclusion

Cell phone vendors shipped 1.019 billion phones in 2006, 22.5% more than the 832.8 million mobile phones sold in 2005. Each phone will typically use more than two stacked-die packages. Increasing wire bonders’ technology - particularly their complex looping capabilities - has enabled these packages to be wire bonded. As new packages come to the assembly facility, process engineers can choose an optimal bonding wire and tool and the appropriate loop shapes and, in a short amount of time, develop a production-worthy process. Combining the latest high-performance bonding wire with an arsenal of process-customized bonding tools enables wire bonding to continue meeting challenging packaging demands.

* Kulicke & Soffa’s ARCUS capillary

PAUL REID, product marketing director, equipment division, may be contacted at Kulicke & Soffa, 1005 Virginia Drive, Fort Washington, PA 19034; 215/784-6792; E-mail: preid@kns.com.