QFN Wire Bond Dilemma - Stress Neck & Stress Heel

Abstract

In the competition of market share of quad flat no-lead (QFN) package, there were several improvement methods being utilized, such as leadframe design for higher density, cheaper tape, and more functionalities per chip (with smaller package size, larger die, and smaller pad size). Most of these improvement methods created extremely weak wire bonding structure, therefore causing further amplification of resonance effect, generating more stress in wires. Eventually, QFN production can not be avoided from having stress effect on wire neck and wire heel, and failing at customer applications. Normally, stringent automotive and hand-phone stress requirements were the two critical applications having failures. In the initial evaluations, QFN stress issues were found difficult to be resolved by executing design of experiment (DOE) and response surface methodology (RSM) on suspicious main factors. By utilizing statistical approach, a new evaluation methodology was developed, assuming severe complication in interaction effects with full resolution for factors more than five. In further analysis of interaction effects, there was ignorable effect of independent factor, and instead all significant factors were best to be presumed as combined interaction. Each suspicious factor need to pass through screening tests respectively either by DOE or comparative method, in order to measure its significant level. Once all significant factors were identified with recommended optimum process window, they were combined back into an interaction configuration. The established interaction configuration was then tested on its robustness in wire bonding responses including stress heel and neck. With the approach of bonding to fail, actual performance margin could be obtained, and hence production margin can be derived or designed properly. Throughout evaluations, several knowledge tools were created, and enabled discovery of some breakthrough findings. From high speed and high resolution camera, combination of machine hardware and leadframe design revealed different level of mechanical bouncing control during wire bonding process. The bonding rigidity has direct impact to resonance effect, with easily bounced structure caused more unnecessary energy transfer or stress in wire. From mapping process, peripheral units are highly affected area, followed by wire bonding dimensions such as loop height and length. Utilization of bonding sequence was found useful in altering resonance effect on sensitive units or wires, hence reducing stress buildup. Furthermore, a lot of finite element analysis (FEA) analysis helped in determines rigid leadframe design against resonance effect and also new wire bonding mechanism to divert stress to proper locations. Despite that, material properties selection and control were found very important in keeping stability in production, especially QFN tape, wire type, and leadframe tolerance. Finally when the combined optimum interaction was put to test, it could withstand up to four times larger ultrasonic power than normal configuration. Despite the package robustness, improvement actions have not sacrificed package features for market competitiveness. With extensive verification with scanning electron microscope (SEM) and high power scope, optimum interaction was proven to be robust & stable in production mode.