Combined Device and System Simulation for Automotive Application Using SABER

For Automotive applications, device optimization requires both system and device level simulation in order to properly predict its performance and thermal response, especially in the case of adverse dynamic conditions. In this paper, the SABER model of IR’s power MOSFET IRFP2907 is validated in a half-bridge circuit using a low frequency two-pulse test. This model is then applied to simulate the performance of an inverter system. Simulation results accurately match the measurement. This work can be extended to evaluate power MOSFETs in other automotive applications such as ISA, EPAS and DC/DC conversion as well as IGBT and anti-parallel diodes in inverter switching applications Index Terms – MOSFET, Modeling, Simulation, SABER I: Introduction: All automotive applications are cost sensitive. It is nearly always desirable to optimize the die size to reduce the cost. However, because of reliability requirements, power devices have to survive system transients and very harsh electrical environments in today’s car. Therefore, accurate performance prediction of the power device is essential for the designer to satisfy the trade-off between performance and cost. Although analytic methods are widely used to estimate the device and system steady state performance, some dynamic characteristics such as device switching loss, stray inductance effect, device avalanche and machine transient are not easily handled by this simple approach. On the other hand, numerical simulation can effectively combine system and device level characteristics to accurately calculate the loss and can predict the device junction temperature, which is the key parameter for power device selection. So with good prediction of the transient and the device thermal response, a device with optimized performance and low cost in the dynamic system environment can be selected. From the design standpoint, this numerical approach can save not only design time but also the cost of building prototypes. SABER [1] provides a good platform for device performance prediction in a system environment. It has the capability to deal with both circuit and device level models. Moreover, its mixed signal simulation can handle electrical analog, digital, mechanical, thermal and magnetic blocks in their native domain, with optimized algorithms implemented to balance the accuracy and speed. So in this study, SABER is used as the simulation tool. Also, device SABER models are now readily available for most new MOSFET products, which provide broad resources and is convenient for benchmark comparison. II: Simulation and Results: For device modelling, different levels of model have been developed to support analysis of different level of behavioural complexity [2]. The essential difference is the accuracy of prediction of switching characteristics, especially in automotive applications from medium (12K to 20KHZ in ISA and EPS) to high bandwidth (75K-120KHZ in DC/DC converter) where the switching losses account for a large portion of the total losses and are a nonlinear function of current and device parasitics. By comparison, conduction loss is easy to calculate even considering its temperature dependency. In this study, to validate the SABER model, the switching characteristics are investigated in a half-bridge circuit, using a low frequency two-pulse test. The simulation results are then compared to measurement. After that, stray inductance effect is examined to clarify the loss distribution between turn-off and turn-on. Following this, a system with a PWM driven three-phase inverter and passive load is simulated to predict the thermal dynamic. Results are compared to the measurements from a real system. International Rectifier’s IRFP2907 is used as an example in this study. Its SABER model was downloaded from IR’s website at www.irf.com. Section 1: Half-Bridge, two-pulse test Figure 1 shows the circuit diagram. Figure 2 compares simulated and measured Vds and Id during turnoff, while the waveform in Figure 3 is the corresponding gate to source voltage Vgs. Data fitting is achieved by adding stray inductances to the drain and source leads such that Ld_stray=16nH, Ls_stray=14nH. Figure 4 and 5 illustrate the waveforms during the turn-on process with the same parameters given above. Figure 1: Schematic of the two-pulse test circuit Vds, Id During Turn-off -20 0 20 40 60 80 100 2.61E-05 2.63E-05 2.65E-05 2.67E-05 2.69E-05 2.71E-05 2.73E-05 2.75E-05 Time(s) Vo lta ge (V ), C ur re nt (A ) Id(A)-Measured Vds(v)-Measured Id(A)-simulated Vds(v)-simulated Figure 2: Comparison of Vds and Id during turn_off