In such a case, the shift amount θsft_t is obtained by setting the overlap period t to the period in which the sum of the on periods of the PWM signals PWM1. PWM2 is the shortest. With this configuration, a delay of the rising edge of PWM2 from the falling edge of PWM1 is avoided in all the periods including the overlap period t.

In the above-described embodiment, only PWM2 is subjected to the phase shift; however, the invention is not limited to this mode. For example, PWM1 may be subjected to the phase shift instead of PWM2 or the shift amount θsft may be distributed between PWM1 and PWM2. For example, both PWM1 and PWM2 may be subjected to the phase shift where the shift amount of PWM1 is set to ?θsft/2 and the shift amount of PWM2 is set to θsft/2.

Comparison between Overlap Phase Shift and Edge-Alignment Phase Shift

FIG. 11 illustrates the waveforms of the PWM signals PWM1, PWM2 at the time when the edge-alignment phase shift is executed and the waveforms of the PWM signals PWM1, PWM2 at the time when the overlap phase shift is executed. In this example, where the overlap period t is 4, the waveforms of both the phase shifts until the intended period are shown, in this example, the rate of PWM1 is set to 1.0, and the rate of PWM2 is set to 0.9.