The cycle-counting function of LDV processors provides a means of measuring the transit time of a seed particle in the probe volume in order to correct velocity bias and velocity gradient errors by weighting individual velocity measurements by the transit time. The main objectives of the investigation are to study the dependence of the number of counted cycles on signal level, and to check that the cycle-counting process is independent of the Doppler frequency. Tests of TSI 1990A and 1980 counters are performed with different kinds of simulated signals generated by function generators. Measurements show that the number of counted cycles depends on the ratio of signal to threshold detector levels in a predictable way. The cycle-counting process is not frequency dependent, so that no bias will be introduced on weighted velocity measurements indirect accuracy limitations will unavoidably arise from the dispersion introduced by the effect of signal level. Major difficulties have been encountered because the signals used for test purposes were different from actual LDV signals, causing wrong cycle-counts: occurrence of end-of-burst failure due to improper signal shape, and interference of the dead time of the processor with cycle-counting. Other difficulties are related with signal shape imperfections. In actual measurements errors of the same kind are likely to occur owing to the lack of validation of cycle-counts. Requirements for a suitable cycle-counting processor, as well as the conditions of proper signal generation for testing Doppler processors are defined.