Photomultiplier tube (PMT) has been an essential instrument for low background experiments, and the accuracy of single photoelectron (SPE) calibration will directly affect the energy reconstruction of the detector. The SPE spectrum can be acquired from the charge distribution of dark-noise pulses or the low-intensity light signal triggered by LED. However, there are arguments on whether the two methods are consistent. In this paper, the Hamamatsu R5912-20mod PMT and the EMI-D642KNFL PMT have been tested adopting the two methods with signal waveform recording electronics. As comparison, traditional charge-to-digital conversion system is also used to test SPE spectrum. The results show that they are consistent when the PMTs are working at typical recommended high voltage (HV).