Development of a Burn-in and Test Process for Producing Known-Good-Die

Abstract

Multichip modules(MCMs) are a major growth market in the IC industry. However, a problem hindering the widespread use of MCM technology is the inability of the MCM manufacturer to obtain bare die with the same level of reliability as those die in packaged devices. If the MCM is determined to be faulty after burn-in and test, it is difficult and expensive to repair. In fact, it can be difficult to isolate the individual chips during bum-in at the module level, making the task of determining the failing chip more laborious. One contributing factor to these failures is the lack of burn-in of the die before the MCM is assembled. If every chip on an 8-chip module has a 4% probability of failing(common for state-of-the-art devices), the module has a 28% chance of failing. This calculation is derived from the equation Pp = (l-P(f))" where; P = probability, p = passing, f = failing, and n = number of die per module. The goal is to deliver to the user known-good-die (KGD) with the same level of electrical test and burn-in reliability as currently available at the package level.