Noise Behavioral Model of Testing Systems

Quality of parameter identification (diagnosis) is strongly dependent on the equal conditions of testing the actual circuit (circuit under test CUT) and simulation of the nominal circuit (circuit under simulation CUS) [1]. We assume that topologies of CUS and CUT are identical and known. The classical approach uses the same excitation J to simulation of the nominal and testing of the actual circuits. Diagnosis equations are constructed using integral sensitivity W of the nominal circuit and difference between vectors: response of CUT - VM, and nodal voltages of CUS - V, and are presented in [2,3]. This approach gives good results when the testing system is noiseless or level of noise is insignificant. But in the actual case excitation J (applied to CUS) not equal JM (applied to CUT), and V not equal VM, even when CUT is fault free. Classical hardware and software tools are ineffective because CUT is a nonlinear object.

We propose an approach to conditioning the process of simulation and testing, which includes noise behavioral model (NBM) of testing system (TS) to simulation algorithms. Time delay of channel, quantization and jitter effects, additive noise of amplifier and attenuator are considered in the NBM of TS. By optimizing NBM of the function generator we achieve J is equal JM, and optimizing NBM of the response recorder unit we obtain V is equal VM. Incorporating NBM of the testing system into simulation of CUS allows us to improve convergence of the procedure for solving diagnosis equations and reduce influence of measurement noise, that is improving the quality of parameter identification of CUT.