The effects of inserting a video filter of finite width on the observability of pulsed signals in random noise are examined. It is found that at match (when the pulse and IF filter are each other's conjugate Fourier transforms) no improvement is gained from such a video: the infinitely wide response yields the optimum results. This is strictly true when the second detector is a quadratic rectifier; (a slight improvement on narrowing the video is noted, however, for strong signals when a half‐wave linear rectifier is used). Away from match a video filter does give noticeable improvement when the pulses are overlong, i.e., narrower spectrally than the IF (&lgr;>1); more noise than signal is then removed by the narrower video filter. The greatest gain over the infinite video is observed for final filters slightly wider than the original pulse. On the other hand, for pulses that are too short (&lgr;<1) the performance is worsened. The mean maximum signal level is decreased more rapidly than the interfering noise background. In all instances the familiar phenomenon of modulation suppression arises. Two different pulse shapes are considered: (1) Gaussian and (2) rectangular, but for wide videos pulse shape is not a critical factor. Curves showing the output signal‐to‐noise ratios as a function of the input ratio are included for conditions of match (&lgr;=1) and mismatch (&lgr;≠1).