Virtual Qualification of IC Sockets Using Probabilistic Engineering Methods

Some high performance computing systems interconnect the microprocessor and the printed circuit board (PCB) using an integrated circuit (IC) socket. In these systems the socket provides a low resistance electrical connection when its contacts are compressed between the IC and the PCB pads. Obtaining and maintaining an acceptable contact resistance is essential for the seamless operation of these interconnects. The electrical resistance of an IC socket contact is a function of applied load, contact compression, and contact surface quality. To apply load to a socket (and compress the contact) a mechanical system is utilized which may consist of springs, screws, bolster plate, heat sink, and other components. The analysis of contact compression and contact resistance in an IC socket assembly is not a simple task. It requires extensive material characterization and testing. Due to rapid design, vendor selection cycles, sample size requirements, and cost, the physical testing to estimate the variability of contact resistance in socket assemblies is not feasible. Probabilistic methods such as finite element analysis (FEA) are used to overcome this obstacle. FEA provides the means to model the response of an assembly to compressive loads and variation in the mechanical load system components. An issue with this approach is that each finite element analysis run takes hours to execute, making it an impractical approach for evaluations that consider numerous parameters and parameter values. This paper describes how the response surface methodology (RSM) can be utilized to reduce the number of FEA runs that are required for the analysis of contact compression and contact resistance. Statistical distributions of the parameters of interest are constructed and used with RSM and Monte Carlo simulations to generate a transfer function. This function is evaluated thousands of times by means of Monte Carlo simulations to quantify the probability that the contact compression (and contact resistance) requirements may not be met, while considering manufacturing variability. The approach allows the virtual qualification of socket assemblies without resorting to expensive testing and prohibitive FEA computation cycles