Complex fractionated signals in intracardiac EGM recordings contain information about the underlying pathophysiology and are thus promising to guide clinical ablation procedures. For instance higher fractionation complexity is associated with fibrotic tissue. Yet, the diversity of mechanisms that can cause EGM fractionation renders the inference of physiological information highly nontrivial. The details to extract from these types of signals are particularly sensitive to being obscured by noise. Lowpass filters are commonly used to smooth EGM recordings. However, it is known that interesting physiological information can have frequency components in spectral ranges that overlap with the high frequency noise. Wavelet denoising allows to make a distinction that does not rely on a constant separation in frequency space. Unipolar EGM recordings of 2.5s registered during a contact mapping guided ablation procedure were exported from the Carto3® (Biosense Webster) mapping system. Different wavelet denoising algorithms are applied. The success in reducing noise while retaining the physiologically relevant signal morphology of each algorithm is assessed. For this purpose recordings with a prominent signal are selected such that an area of interest with dominant signal and an isoelectric background window with dominant noise can be identified. The mean, range and standard deviation of the deviation from baseline in the isoelectric window are calculated to quantify noise. Quantities of interest such as number of peaks and maximal downslope in the area of interest are determined to quantify the stability of physiologically relevant morphology.