Modeling the dynamic behavior of the I2L inverter

Abstract

r VlODE ,LS WHICH characterize 12L inverters have been presented, as well as methods to determine the model parameters’ >’. Since these models rely on the charge control principles, a quasi-static charge distribution is assumed during transient analysis. This reflectsessentially in a one-pole approximation for the small signal input impedance of the device. The deviation of this approximation from reality is a valid measure for the applicability of the charge control model, even though the latter will be used in switching problem. Measurements of the small signal input impedance of an 12L inverter a t medium and high power levels, result in curves which strongly deviate from the -20 dB/decade rolloff of the one pole approximation; Figure 1. This indicates that the chargecontrol principles are not a priori valid. It has been reported that the charge distribution in the Y direction (perpendicular to the surface) is in principle non quasi-static in the epitaxial layer3. However, straightforward analysis proves that for realistic values of the epitaxial width W and of the NN’ interface recombination velocity s, such that the product sW is much smaller than the diffusion constant of holes in the epitaxial region, the one pole model remains a good approximation in the measured frequency range4. The influence of the distributed nature of the base resistance in the X direction constitutes a second source of non quasi-static behavior; Figure 2. The general distributed network (Figure 3 insert) can be analysed by introducing the following assumptions. ( I ) The overall impedance RB and the overall small signal admittance r, and C, are uniform over the length L of the distributed network. This implies that lateral variations are ruled out by adopting mean values for RB, r, and C,. (2) The voltage dependence of r and C, along the base is neglected for the ac analysis. Computer simulations have shown that this condition limits the applicability of the results t o moderate dc biasing, such that r, < RB. ?