The effect of baseline restoration on signal‐to‐noise ratio in pulse‐amplitude measurements is determined theoretically and experimentally. It is found that the signal‐to‐noise ratio with baseline restoration normalized to that without restoration is &eegr;r/&eegr;=[1+&agr;− 2&agr;½kx(&tgr;)]−½, wherekx(&tgr;) is the noise correlation factor at the output of the filter (amplifier), &tgr; is the measurement interval, and &agr; is the attenuation of noise power due to integration by the restorer capacitance and switch resistance (&agr;=1 for fast restorer). Methods of calculation of correlation functions are outlined. A method for measurement of correlation functions by oscilloscope is described. The results are presented for the three commonly used filters: CR‐RC filter, quasi‐Gaussian filter, and delay‐line filter. Baseline restoration is found to degrade the signal‐to‐noise ratio. The amount of this degradation can be reduced to negligible proportions in filters with long memory (such as CR‐RC) by a small degree of integration in the restorer. Peak‐to‐peak amplitude measurement of double‐differentiated signals performs, in effect, baseline restoration and, in addition, improves signal‐to‐noise ratio compared to peak‐amplitude measurement.