In the above-described overlap phase shift, an intended overlap period t may be set arbitrarily. For example, a constant (t=5, or the like) may be set. A period required to raise the measured output voltage VH to the target value VH* may be set for the overlap period t. With this configuration, a delay (separation) of the rising edge of the PWM signal PWM2 from the falling edge of the PWM signal PWM1 is avoided over the entire period in which the measured output voltage VH is raised to the target value VH*, so an additional auxiliary pulse is avoided.

In FIG. 9, for the sake of simple illustration, the rate of the PWM signal PWM1 is set to 1.0 (constant); however, the rate is not limited to this configuration. In short, in the case where the edge-alignment phase shift needs to be executed multiple times, there can be an advantage of executing the overlap phase shift instead of the edge-alignment phase shift. That is, when the on period of the PWM signal is shortened and the rising edge of one of the PWM signals delays from the falling edge of the other even with the edge-alignment phase shift once, there is an advantage of executing the overlap phase shift.

The above-described case can occur when the total period of the on-duty periods of the PWM signals PWM1, PWM2 based on fluctuation predictions become short once every period, that is, when the average rate of the PWM signals PWM1, PWM2 becomes smaller than 1.0. For this reason, when both the rates of the PWM signals PWM1, PWM2 are taken into consideration by extending the mathematical expression (2), the mathematical expression (3) is derived as a mathematical expression for obtaining the shift amount θsft_t of the overlap phase shift.