A feedback proportional control repeats the following processing of starting control of an optical modulator by a control signal that corresponds to a reference value set in the preliminary search or in the previous cycle, calculating a new reference value based on an error signal obtained from the optical modulator, and controlling the optical modulator by the control signal that corresponds to the new reference value, thereby acquiring the control signal in which the error signal is minimized as an optimal control signal, setting the acquired control signal as a signal for controlling the optical modulator, and storing the optimal control signal as the reference value to be used at a time of starting the feedback proportional control in the next cycle.