In this paper, we present a quantitative study to find the best-suited mother wavelet and the optimal number of decomposition levels to analyze Narrow Bipolar Events (NBE) generated by Compact Intracloud Discharges (CID). The selection was made using one criterion namely, sample entropy. Eleven distinctive NBE pulses were used and three different signal-to-noise ratio were evaluated (-10 dB, 0 dB, 10 dB). A total of 58 mother wavelets were used to perform the denoising process using a soft/hard thresholding and four different approaches to defining the threshold to filtering the coefficients. The outcomes show that is not possible to find a unique solution and the final mother wavelet selection depends on the expected results of the analysis. In this study, the aim is to detect the existence of NBE in a noisy signal, thus sample entropy criterion proves to be a useful tool for this purpose. Therefore, according to our findings, mother wavelet rbio3.1 with four level decomposition, soft thresholding and rigrsure threshold selection is the best choice for detecting NBE pulses up to SNR of -13dB.