Power transformers are one from more important and the most expensive elements of the AC electric power systems. For this reason, their in-service operation and management are usually based on the Condition Based Maintenance (CBM) strategy, which uses diagnostic data obtained by various methods to assess the real condition of supervised objects. The method analyzing the transfer function (TF) of the transformer windings makes it possible to track changes in the electrical and mechanical parameters of the windings due to the occurrence of the various operating stresses (e.g. high current short-circuits). The paper presents analysis regarding the use of wideband, random signals (applied from a pseudo-white noise generator) for determining the transfer function of transformer windings. Particularly, the influence of the basic digital signal processing parameters (e.g. waveforms sampling frequency, data record length and selected window function) on the evaluated characteristics of the transfer function is shown.