The mathematical differentiation of experimental signals is often used as a resolution enhancement technique, to facilitate the detection and location of partially overlapped gaussian peaks in a multi-component signal. In this paper the analysis of the peak shapes of real ion mobility spectrometry (IMS) signals indicated that a gaussian approximation leads to less than 0.035% of mean square error (using a window of ± 3σ around the peak) while the analysis of the noise signal (performed both in the time and frequency domains) showed that it could be modelled as filtered white gaussian noise. The experimental signal data were provided through the IMS sampling of the chemicals diazepam, lorazepam and bromazepan. Given these findings, a model capable of simulating IMS peaks with any desired time separation, relative amplitude, or additive noise level was developed. It then became possible to quantitatively evaluate the performance and limitations of the second derivative peak detection algorithm. Peak resolution down to 0.27 ms and signal-to-noise ratios down to 18 d⊙B were achieved. Some preliminary results obtained with higher order derivative algorithms are also presented.