Baseline noise has a significant contribution to the repeatability of chromatographic measurements. Noise can be analyzed by calculating the power spectrum of the noise sequence, because the logarithmic plot of the power spectrum against the frequency reveals the major characteristics of the noise. There exist several different models for describing the baseline noise. More often the flicker noise is found to be a suitable model, but it can be shown that in many cases the noise characteristics are different in different frequency domains.

The heteroscedastity or homoscedasticity of the noise is another important point that influences the precision of analytical measurements. Usually it is assumed that noise level is proportional to signal intensity. This is however a rather rough estimation. We propose several approaches – including spectral analysis – to evaluate these noise features, too.