For example, in case the 1/f noises are to be taken out in anultra-low frequency range of one several hundredth of thousandth, the d.c. coupling is generally convenient but it superposes the aforementioned drift due to the temperature change upon the 1/f noises, thereby causing a large error.As another means for avoiding the aforementioned temperature drift, on the other hand, there is a method in