In this paper, we diagnose the fault in the CA analyzer by setting up the initial value such that the final test signature is a constant regardless of the circuit being tested. This method makes the CA test procedure short and clear. In addition, we detect the fault of the faulty input sequence by using the inverse matrix of the CA state transition matrix.