In this article the problem of errors arising from a specific design o
f a real-time multichannel scaler is presented. The dead-times error,
double-pulse error, and the error due to the gate width modulation are
discussed. The absolute values of the systematic errors of the gate w
idths of a single channel are of the order of hundreds of picoseconds,
making for a severe problem when the gate is small. The problem is ov
ercome by a floating allocation of the gates. This method leads to a d
ecrease in the differences between the variances of the number of coun
ts for all time channels and as a consequence improves the precision o
f the measurements. (C) 1998 American Institute of Physics.