Measurement of the mutual inductance is one of the key techniques in the joule balance to determine the Planck constant h, where a standard-square-wave compensation method was proposed to accurately measure the dc value of the mutual inductance. With this method, analog switches are used to compose an analog-switch signal generator to synthesize the excitation and compensation voltages. However, the accuracy of the compensation voltage is influenced by the non-ideal behaviors of analog-switches. In this paper, the effect from these non-ideal switches is analyzed in detail and evaluated with the equivalent circuits. A programmable Josephson voltage standard (PJVS) is used to generate a reference compensation voltage to measure the time integration of the voltage waveform generated by the analog-switch signal generator. Moreover, the effect is also evaluated experimentally by comparing the difference between the mutual inductance measured with the analog-switch signal generator and the value determined by the PJVS-analog-switch generator alternately in the same mutual inductance measurement system. The result shows that the impact of analog switches is 1.97 × 10 −7 with an uncertainty of 1.83 × 10 −7 (k = 1) and confirms that the analog switch method can be used regularly instead of the PJVS in the mutual inductance measurement for the joule balance experiment.