As a key component of power system, DC/DC converter needs to operate normally. Related DC/DC converter fault diagnosis technology has been widely proposed, among which the output voltage is often used to indicate the health status of the converter system. This paper mainly studies the influence of various sampling frequencies of output voltage on soft fault diagnosis accuracy of DC/DC converters. A buck converter is used as an example, and its output voltage signal is collected at different frequencies. Based on the obtained output voltage samples, various machine learning models are applied to diagnose converter soft faults. The diagnosis accuracy of these models under different output voltage sampling frequencies is studied. The appropriate sampling frequency of the models is investigated. This research provides a reference for the selection of signal sampling frequency for soft fault diagnosis of DC/DC converters in practical engineering applications, because the implementation cost of high-frequency sampling circuits may be high.