Discrete sampling is widely used in broadband partial discharge(PD) measurement system, and the sampling parameters of PD pulse sequences have significant influences on the calculation results of eigenvalues. Cluster diagram is a tool developed with the development of broadband PD measurement system, which can analyze PD pulses in real time and apply cluster technique to the separation of PD source. In order to study the influence of the stationary noise from the vertical resolution of oscilloscope on the cluster separation, this paper quantitatively analyzes the stationary noise caused by the sampling frequency F_s, the sampling time T_s and the sampling bits in the broadband PD measurement system by simulation and measurement verification. The results show that different PD sources can be separated significantly by using the equivalent time-frequency classification method with the parameters of equivalent period T_eq and equivalent bandwidth F_eq, while the same PD source can not be separated in this method. Cluster results are greatly affected by the sampling bits, sampling rate and sampling time, and the sampling bits have the most influence on the results. Thus, an appropriate sampling bits can decrease the influence on classification figure.
 

clustering separation, equivalent bandwidth, equivalent period, partial discharge, sampling bits