The implementation of real-time blood velocity estimators using time-domain cross-correlation is investigated. The basic algorithm for stationary echo canceling, cross-correlation estimation and subsequent velocity estimation is presented. Sampled data acquired at rates of approximately 20 MHz are used in the algorithm, imposing a heavy burden on the signal processing hardware. The algorithm is analyzed with regard to the high sampling frequency, and a method for performing real-time high-speed-movement and cross-correlation is suggested. Implementation schemes based on using the sign of the data as well as the full precision are proposed. From an analysis of the process it is concluded that the sign data implementation can attain real-time processing. This can also be obtained for the full precision data, but at the expense of using a number of dedicated signal processing chips. Both implementations suggested can handle the estimation of velocities for A-lines acquired from multiple directions.