Abstract: Collecting accurate data of stability parameters is a key to monitoring and fault diagnosis of hydropower units, particularly the core stability parameters: peak-to-peak amplitudes of vibrations and pressure pulsations. First, this paper examines different methods for calculating peak-to-peak values specified in different standards and discusses the concept of confidence interval from the two viewpoints of statistics and engineering application. Then, considering signal collection in a complicated field environment, we verify each method using both numerically simulated signals with normally distributed noises and field measurements. Results show that for low frequency dominated signals, the ±1.96σ method produces large errors and thus is not applicable, while for signals with high frequency components, different methods often result in great differences. This study recommends the method of 95% confidence interval for calculation of peak-to-peak values of the signals, which is conservative but robust and insensitive to noises. These results would help the revision of GB/T 17189, IEC 60994 and other related ISO specifications.