This paper describes an improved lumped circuit model of power bipolar junction transistors (BJTs) that can predict the turn-off fall time to a greater accuracy than currently available models. Though the existing models simulate the storage time and delay time to a good accuracy, the fall time performance is neglected. This is because the existing models do not account for the charge decay due to recombination. The model presented in this paper is based on the charge dynamics of the device. The charge dynamics is explained in detail using simulation results from an advanced two-dimensional (2-D) device and circuit simulator. Based on a physical understanding of the charge dynamics, this model is implemented to incorporate the charge decay due to recombination to account for the current tail during turn-off. The lumped-circuit model is implemented in PSPICE using the existing quasisaturation model along with controlled sources. To validate the model, the device was subjected to hard-as well as soft-switching conditions (zero current switching and zero voltage switching). The modeled results are observed to have a good match with measured results.