Neuro-space mapping modeling approach for trapping and self-heating effects on GaAs and GaN devices

In recent years, neural networks have been successfully applied for modeling the nonlinear microwave devices as GaAs and GaN MESFETs/HEMTs. Many modeling approaches have been developed for small and large signal applications. In this contribution, a neuro-space mapping approach is proposed for modeling the trapping and the self-heating effects on GaAs and GaN devices. The Angelov empirical model is used as the coarse model, which can be adjusted using DC and Pulsed I/V measurements at different static bias points. The proposed approach is tested for the MGF1923 GaAs MESFET and for an 8x75m AlGaN/GaN HEMT. DC and transient simulation results are compared to DC and Pulsed I/V measurements. Good results are obtained for the DC and dynamics I/V characteristics at different static bias points.