In this paper we present experimental results of a method for determining the path of a synthetic aperture sonar (SAS), with sub-centimeter accuracy, by autofocussing images of an active beacon. The beacon sits on the seabed, listens for sonar pings using a matched filter, and retransmits in a different frequency band after a fixed time delay. The sonar forms a separate image containing just this point-target, blurred by sonar motion, with a high signal to noise ratio (SNR) and no contamination from close range objects. A simple narrowbeam autofocus algorithm is applied to estimate the sonar path, providing motion compensation (MOCOMP) data to correct the SAS images of the other objects in the scene. Independent one-way and twoway approaches to measuring the time of flight (TOF) were investigated, with the two-way approach producing better results. Significant improvement was achieved for two runs at 30kHz with a small improvement achieved in one run at 100kHz. Fitting a hyperbola, to the separate beacon raw data produced a velocity estimate for the sonar that was within ±0.025 m/s of the tow boat velocity, for both runs, when compared with the global positioning system (GPS) velocity measurements averaged over the run length.