This paper proposes a method to separate distortions of the input signal, amplifier, and analog-to-digital converter (ADC), which are cascaded blocks in the testing of an analog front-end system. The proposed technique extends the stimulus error identification and removal (SEIR) method to separate the low-quality input test signal, the nonlinearity of amplifier and ADC. The system model was built by using MATLAB/Simulink, and the nonlinearity of the input signal with distortion, amplifier, and ADC were successfully separated. With a 12-bit ADC, the estimation error of each block is less than 1 LSB in the simulation, and the proposed method was verified by measurement. The relations between the estimation error, calculation cost and each parameter in the proposed method are systematically studied.