Acoustic Doppler instrument measurements suffer from random spikes and Doppler noise. Using a four-receiver ADV (acoustic Doppler velocimeter; Vectrino manufactured by Nortek) that allows recording beam velocities, we combine a spike-removal procedure on the beam velocities with a noise-reduction method on the flow velocities to improve turbulence measurements. We compare the results with those obtained from ADVP (acoustic Doppler velocity profiler) measurements under the same conditions, i.e., in turbulent open-channel flow over a coarse-grained bed. It is shown that spikes are best removed from ADV beam velocity data before calculating flow velocities, thereby correcting all three flow velocity components at the source. Spikes in beam velocities do not correlate with low correlation values. ADVP data generally have few spikes and do not need spike removal treatment, showing that spikes are instrument related. The noise reduction method is based on the decorrelation of the Doppler noise terms contained in two vertical velocities redundantly sampled in the same volume. The combined ADV data treatment is sufficient to significantly extend the resolved frequency range in the velocity spectra. It reduces RMS values by up to a factor of 2, and the corrected values agree with ADVP results and theoretical predictions, indicating that both treatments are needed. Owing to spatial averaging effects over the ADV sample volume, a sampling frequency limit of close to 50 Hz is determined by the deviation of the spectra from the -5/3 slope.