CSP and flip chip underfill
06/01/2001
Optimizing production throughput by leveraging dual-lane dispensing
BY S.J. ADAMSON
The widely expanding use of solder bumped area array packages, such as flip chip and chip scale packaging (CSP), has driven change across many of the fundamental processes used in electronics production. The dispensing of underfill encapsulant between the die and the substrate is one of the critical arenas that recently has received attention and has undergone significant process development. Over a period of time, underfill dispensing techniques have continuously matured and adapted to meet overall production requirements, steadily increasing in accuracy, repeatability and speed.
In a complex production environment, the maturation of each individual operation cannot occur in isolation. As the speed, efficiency and throughput rates improve for any particular point in the production line, it is important to continually reevaluate how to adjust other parameters to maintain optimal overall production flow rates. For example, the dramatic increases in underfill dispensing speed during the past few years have now shifted the bottleneck in the process to the time that it takes for the encapsulant to flow out beneath the die.
While some industry observers maintain that underfill dispensing is an inherently slow process, it is important to understand the difference between dispense and flow-out time. Dispensing is the time taken to put the fluid down and flow out refers to the time for the fluid to be drawn under the die by capillary action and flow out at the opposite end of the die. New advances in the workflow process are now being used to dramatically boost sustained dispensing speeds. By systematically supplying more parts to the work envelope and programming a system for efficient multi-pass dispensing for parts simultaneously traveling on two parallel lanes, it is possible to achieve higher dispensing speeds.
The Growing Importance of Underfill Processes
To achieve higher input/output (I/O) densities, smaller package sizes and lower cost for high-volume production, solder bump arrays have rapidly transitioned from relatively exotic processes just a few years ago into widely used mainstream technologies today. A significant percentage of high-performance microprocessor designs now use flip chip technology to meet increasing I/O requirements. Similarly, the whole spectrum of today's miniaturized and portable electronics products, such as cellular phones, personal digital assistants (PDAs), video cameras, global positioning satellites and more, are all dependent upon flip chip and CSP technologies to achieve ultra-small size and weight specifications.
In addition to providing higher circuit densities and packing more functionality into very compact packages, flip chip and CSP techniques can also deliver consistently higher reliability and longer product life if the production processes are properly defined and controlled. As solder bump designs and production techniques have matured, underfill encapsulation has become one of the standard processes for ensuring high levels of product reliability.
A well-controlled underfill process can significantly improve product reliability by adhering equally to all surfaces under the die and effectively distributing the stresses created by the coefficient of thermal expansion (CTE) mismatch between the die and the substrate. This minimizes the stress on individual solder bump joints, which are inherently the weakest points in a non-underfilled structure, thereby increasing thermal fatigue life. The underfill also protects the solder joints from transient mechanical shock. In addition, the presence of a uniform voidless underfill between the die and substrate has been shown to provide an improved level of environmental protection.1
The underfill dispensing process fits well into the overall production flow after the reflow soldering step, thereby enabling all upstream and downstream processes to proceed without unusual delays. Because the dispensing step comes after reflow, some manufacturers choose to insert an electrical testing step before underfill encapsulation in case they need to rework a component.
When evaluating an underfill process, the process appears to be very slow unless dispense time and flow time are considered separately. The time to dispense is very short (less than one second in most cases), but flow out takes much longer. Modern dispensing platforms handle this by allowing flow out on a post dispense, flow-out station. Some board manufacturers factor in flow out taking place at the start of the oven cure. In either case, the throughput of the line is not impacted by the flow-out time.
However, for large die or any other dispensing application where multiple dispense cycles are required, the whole process can be governed by the speed of flow out. No-flow materials have been proposed as a faster underfill solution. However, if the flow-out time is removed from the equation, it is very difficult to find other methods of fluid application that are as fast as standard dispensing. No-flow underfills do have an advantage of not requiring a second oven for underfill cure, but there can be trade offs in thermal expansion and process latitude to use these types of underfill materials.
High-speed Dispensing
The raw speeds of underfill dispensing systems have increased greatly during the past few years. High-speed single action (easy to clean) linear positive displacement (LPD) pumps and high-speed dispensing systems that can achieve 2G accelerations for high-speed short movements enables advanced dispensing equipment to deliver precisely controlled amounts of material in dispensing passes that take a fraction of a second. From a total throughput perspective, today's dispensing systems can consistently run much faster than the flow-out process. For example, with a relatively straightforward 12 mm by 12 mm die, an L-shaped dispensing pass can be accomplished in approximately 0.6 seconds with motion systems capable of achieving 2G acceleration. The time required for complete flow out under the die, however, is about 20 seconds. Flow-out times can become even more lengthy in designs with larger die, such as microprocessors with large numbers of bumps or fine pitch devices with very little space between the solder bumps. The amount of encapsulant that can be put down at one time is also restricted by the need to avoid dispensing underfill onto the back of a die or to surrounding components.
 Figure 1. An L-shaped dispense pattern, followed by a seal pass, can achieve reliable underfill at high throughput. |
For large or fine pitch die, it is often necessary to dispense a small amount of encapsulant in an initial pass and then to return later for subsequent passes. For example, when dispensing to a large die using a typical L-shaped pattern, the dispense head would need to lay an initial bead of encapsulant along the two adjacent edges. Care needs to be taken to prevent overdispensing, which risks getting epoxy material on the back surface of a die, like a microprocessor, where a heat sink would be mounted.
The initial fillet acts essentially as a reservoir that feeds the capillary flow-out process beneath the chip. After the first pass of encapsulant has begun flowing under the chip, the dispense head needs to come back and add more material to the fillets along both edges that feed the continuous flow out. In addition, upon completion of the dispensing/flow-out process, many manufacturing situations also call for a final seal pass to finish encapsulating the die. Multiple dispense applications are also used when underfilling CSPs under radio frequency (RF) shields, in which case it is often necessary to perform multiple dispenses through a small hole in the shield (Figure 1).
The Key Challenge
Increasing throughput: A major issue of concern is getting more parts into the work envelope at one time to increase the utilization of high-speed dispensing capabilities. For instance, using the previous example of 0.6-second dispense time per die plus a 1.0-second move time to the next die and assuming a typical five-up parts boat, then the entire process time for dispensing L-shaped passes to an entire boat would take 7.0 seconds, not including time for transport and fiducial location. Even with a single-pass dispense cycle, the need to wait for 20 seconds of flow-out before doing the sealing pass can require the system to spend more time waiting than dispensing.
 Figure 2. Twin lane dispensing allows for dispensing, underfill fluid flow-out and part transport to occur simultaneously. |
In contrast, if efficient methods can be devised for getting more parts into the work envelope on a continuous flow basis, then the currently available performance headroom from raw dispensing speed can be exploited to deliver real sustained throughput increases.
Bigger Boats or Multiple Lanes
The most obvious alternative for getting more parts in the dispensing system work envelope is to either get more parts into each carrier boat or to put more boats in the system at the same time. A manufacturing process engineer has to take into account any potential impacts on the rest of the production line. Shifting to bigger boats could have significant implications for other equipment on the line.
In contrast, running multiple lanes of boats through the dispensing machine has shown to be effective in boosting total throughput of the underfill process while having the advantage of not disrupting any other production line operations. By splitting the incoming flow of boats into two lanes for transit through the dispensing work envelope, these new system designs can almost double the throughput of the underfill process, depending on the application (Figure 2).
Implementing Twin-lane Dispensing
Simply sending more boatloads of parts through the system is not enough by itself to yield higher throughputs. The entire system must be carefully designed to maximize actual dispensing time to avoid unnecessary movements and to minimize or eliminate wait states. A high degree of machine programmability and multi-tasking flexibility is also required to tailor the sequence of tasks to deliver optimal utilization and throughput rates for specific product configurations.
For example, when receiving boats into the dispensing work envelope, the system needs automatically to find all of the fiducials and height-sensing data for every part in both boats at the start of the process. A system where boats in the two lanes arrive sequentially, rather than in parallel, and the vision and height sensing system has no waiting time lays the foundation for all subsequent operations to be carried out in a single seamless process flow.
After finding all of the parts in both boats and registering their positions, a system can then step through a pre-programmed dispensing process with an efficient interweaving of dispense operations across both lanes of boats. For instance, in a program set up for two dispense passes and a seal pass, the system could go through the sequence shown in Figure 3.
Some system design attempts have added a second lane of boats and process two lanes in parallel and dispense on both boats before moving two more boats into place. However, without the built-in programmability and flexibility for sequencing the multi-tasking operations throughout the work envelope and interleaving steps between different boats, having two lanes of boats within the system actually amounts to little more than bringing the buffer queue inside the machine.
Putting It All Together
Another key issue is the need to split the incoming flow of boats into two lanes and to feed them into the dispensing system. Leading-edge dispensing system providers need to offer customers complete production-cell processing solutions, which include the functionality for dividing the flow on the fly, efficiently handling all of the twin-lane dispensing operations, and reunifying the flow for downstream operations. The objective is to provide a complete multi-tasking dispensing solution that fits seamlessly into the existing process flow, while simultaneously reducing or eliminating the time currently spent waiting for flow-out of underfill encapsulant. A two-lane solution can provide maximum throughput for demanding applications that require multiple dispense passes.
AP
Reference
- Martin Bartholomew, "An Engineer's Handbook of Encapsulation and Underfill Technology," 1999, p. 2., Electrochemical Publications, Ltd., Port Erin, Isle of Man, British Isles
Steven J. Adamson, product manager of semiconductor assembly and packaging, can be contacted at Asymtek, 2762 Loker Avenue West, Carlsbad, CA 92009; 760-431-1919; Fax: 760-431-2678; E-mail: [email protected].