In many separation techniques, such as field flow fractionation, liquid chromatography, and electrophoresis, chemical species form bands that migrate at distinct velocities. If these bands are to be quantified on-line using a shot-noise-limited detection system, then attention must be given to the data-digitization rate and to the removal rate of molecules from the analyte pool as a result of the detection process. A theory is developed for calculating the signal-to-noise ratio under such conditions, and it is specialized to the case of fluorescence detection in capillary electrophoresis. Using standard detection procedures in which the data-digitization rate and excitation intensity remain constant for the duration of a separation, detection sensitivity can vary by more than a factor of five for bands that arrive at the detection zone between migration times tau fast and 10 tau fast, where tau fast is the time after the start of the separation that the fastest migrating band arrives at the detection zone. To compensate for different band velocities, both the data-digitization rate and the excitation intensity must be decreased as separation time (tau) increases by the factor tau fast/tau. Only when these corrections are made can uniform sensitivity with the highest possible signal-to-noise ratio be achieved for each peak. These predictions are experimentally tested and compare favorably to observations.