Skip to Main Content
Junction temperature is an important issue for a semiconductor package, influencing the package's thermal, mechanical, and reliability performance. An accurate prediction of junction temperature provides informative guidance in design, development and operation of the package. A compact thermal resistor-capacitor (RC) network approach is presented in this paper to accurately predict transient junction temperatures. The thermal RC network in this approach is a nongrounded Foster network. This approach consists of extraction of the thermal Foster network and prediction of the transient junction temperature response to a given power input using the extracted network. The network extraction part is based on Kirchhoff's current law and Laplace transformation technique, and uses the Foster network to facilitate changes of the RC network structure. The temperature prediction part is a direct substitution-and-calculation process, and therefore is fast to carry out. Since Laplace transforms are directly or indirectly available for most power inputs, their transient temperatures may be predicted by the proposed approach. Superposition is employed in cases where the Laplace transform of a given power input is not directly found in Laplace tables, or where the junction temperature is affected by multiple heat sources. The proposed approach is demonstrated with a power amplifier (PA) module; predicted junction temperatures are accurate in both single and multiple heat source cases.