A systematic examination of measurement error in acoustic Doppler current profiler (ADCP) velocity estimates, in the limit of large signal-to-noise ratio, is made using a system model and sonar signal simulations coupled into an ADCP. The model is extremely successful in predicting ADCP performance. The signal simulations provide model validation. Three main sources of error are examined: frequency tracking, measurement variance (inherent variance of pulse-to-pulse incoherent volume reverberation), and measurement bias. A theoretical lower bound on measurement variance is directly tested by coupling simulated signals into an ADCP. The observed measurement variance is approximately twice the theoretical value and varies as the inverse of the product of the pulse and averaging period (bin). Model predictions of velocity errors for back-to-back beam pairs measuring a sequence of increasing velocity-shear profiles in a medium of randomly distributed scatterers are in excellent agreement with errors measured from simulated signals coupled into an ADCP. Trade-offs between velocity error, vertical and temporal resolution, and maximum range are discussed, with specific focus on optimizing parameters available to users of commercial instruments. For reasonable parameter choices in low velocity-shear ocean conditions, the predicted error in horizontal velocity from effects considered in this study is 1–2 cm s−1. In large-shear conditions, the predicted error using the same parameters as in low shear is much worse, approximately 10 cm s−1. Optimal parameter choices, however, can reduce the error in large-shear conditions to 1–4 cm s−1.