Using charged device model testing to eliminate real-world ESD failures

Abstract

Of 30 bipolar, BiCMOS, and CMOS monolithic, integrated circuit products that were ESD classified to the socketed Charged Device Model (CDM), 27 had ≥500 V withstand voltages and experienced no real-world CDM failures. Two of the three focus products with <500 V withstand voltages initially had numerous manufacturing-induced CDM failures. Analysis of these two products showed that both socketed and non-socketed CDM testing induced damage at the same failure sites as identified on real-world CDM failures. However, only non-socketed CDM testing consistently reproduced the subtle damage observed on the real-world failures. On one of the focus products, the more severe damage induced by socketed CDM testing resulted in an open circuit rather than the resistive short that occurred on both the non-socketed and real-world CDM failures. Once the physics of CDM failure on the three focus products were fully understood, the ESD redesigns were relatively straightforward. On all three products, diffused series resistors and/or clamping devices with fast response times were added to the pins with inadequate CDM robustness. For each product, these redesigns boosted the socketed CDM withstand voltages for the previously susceptible pins to ≥1500 V and eliminated real-world CDM failures. Based on this work, a combined socketed and non-socketed CDM test approach is proposed for classifying/evaluating products and driving CDM robustness improvements. Guidelines for CDM testing and CDM improvement programs are also provided.