SELF-CORRECTING DEVICE

PURPOSE:To easily attain the self-correction of a signal conversion error even with a signal converter having a complicated input/output function by simulating the learning ability of a neural circuit. CONSTITUTION:When a test input signal is sent to a device 10 to be tested from a test signal generating circuit 1, an output signal is transmitted from the device 10. Thus output signal is converted by a signal converter 3, and an error deciding circuit 2 decides an error between a test ideal output signal and the actual output signal. When this error is larger than the set value, a correction signal is sent to the converter 3 to change the characteristic of the converter 3. At the same time, the output signal is outputted again from the circuit 1 and repetitively processed. Therefore the circuit 3 simulated to a neural circuit performs the conversion of signals so as to secure the coincidence between the output signal produced by being subjected to the characteristic of the circuit 3 and the test ideal output signal. Thus it is possible to easily apply the self-correction even to a device to be controlled having the input/output characteristic that contains complicated correlation.