Example of the ability and importance of the PRESTO rednoise filter to reduce the EMIR locally generated signals that interfere with the search algorithm. The plots show a blind detection of the Galactic centre magnetar PSR J1745−2900 in one epoch in which the pulsar was relatively weak (2017 May 18, mJy). Left panel: candidate plot from prepfold when the raw data are folded and rednoise is not applied. Right panel: same candidate when the data are cleaned by rednoise prior to folding. The filter does subtract part of the pulsar signal, as apparent by the narrow profile and the dip in the profile below the mean noise level, but the improvement compared to a non-filtered version is so substantial that the pulsar may easily have been missed if the filtering had not been applied. The main panels in each plot show (a) the folded profile twice, (b) the candidate and observation details including name, telescope, epoch, time sampling, statistics, source coordinates, period and period derivatives (in topocentric and barycentric references), and binary parameters if applicable, (c) waterfall plot of folded signal intensity vs. time, (d) accumulated reduced χ² of the integrated profile vs. a model of noise alone, and (e) two-dimensional waterfall plot of the reduced χ² of the candidate profile vs. noise as a function of folding period and period derivative. The white gaps in the phase vs. time panel are intervals in which the telescope did not observe the source, e.g., during calibration, pointing, or focus, or for some epochs, when observations were taking place at different frequency bands that were not used for the pulsar search (see Sect. 2).