We present a new methodology to discriminate random telegraph noise (RTN) and flicker (1/f) noise components from set-up noise. We illustrate it for a strong RTN case (∆ID/ID ≈ 30 %) measured on a 26 nm gate length nMOS transistor. The approach is based on high-accuracy time-domain measurements. An iterative Schmitt trigger-like algorithm was developed to properly identify and model the RTN and 1/f noise. Statistical analysis allows to assess the accuracy of the extraction. The power spectral densities (PSD) of the different noise models accurately match the frequency measurements.