The original version of the most successful approach for power-line (PL) interference removal from ECG, called subtraction procedure, is based on linear segment detection in the signal and hardware synchronised analogue-to-digital conversion to cope with the PL frequency variations. However, this is not feasible for battery supplied devices and some computer-aided ECG systems. Recent improvements of the procedure apply software measurement of the frequency variations that allow a re-sampling of the contaminated signal with the rated PL frequency followed by interference removal and back re-sampling for restoration of the original time intervals. This study deals with a more accurate software frequency measurement and introduces a notch filtration as alternative to the procedure when no linear segments are encountered for long time, e.g. in cases of ventricular fibrillation or tachycardia. The result obtained with large PL frequency variations demonstrate very small errors, usually in the range of ± 20 μV for the subtraction procedure and ± 60 μV for the notch filtration, the last values strongly depending on the frequency contents of the QRS complexes.