The basic IGBT structure was reported about 40 years ago and has become more complex to achieve lower losses, higher robustness, higher reliability, and cost-effectiveness. Today's power devices are also required to improve noise characteristics corresponding to its increasing switching speed. In noise analysis based on switching waveforms, Fourier transform (FT) is often applied to estimate the causes as device structure has frequently been improved. However, the time information is implicitly deleted according to FT, and it is difficult to consider the causes by extraction of specific parts on waveforms because every waveform's time axis includes important information for device structures. This article presents analysis methodology to identify the key parts without deleting time information by using a wavelet transform and elucidates the effect of transition of dominant radiated noise generated at turn-on operation in case of different carrier stored layer density in CSTBTs.