In a reference signal comparison AD conversion scheme, a reference signal SLP_ADC and each of P and D phases of a pixel signal voltage Vx are compared. A count clock CKcnt1 is counted based on the comparison result. The counting result data is converted into signal data Dsig, i.e., the difference between the P and D phases, which is also subjected to CDS. At this time, the n-bit AD conversion is performed on each of the P and D phases of the pixel signal voltage Vx, followed by summation for digital integration. This prevents any possible detrimental effects that may be caused by summation in the analog domain. Although the signal data becomes W times greater, noise will likely become √W times greater. This alleviates the problem of random noise resulting from AD conversion such as quantizing noise and circuit noise that do not exist in the analog domain, thus reducing the noise.