High quality focused SAR imaging dictates that the relative phase error over an aperture must be kept below a fraction of a wavelength. On most deployed SAR systems the internal measurement systems ability to measure position uncertainty is not sufficient to achieve this required precision. This necessitates an additional post-processing step of data-driven phase error mitigation known as autofocus. We present results comparing the performance of a variety of autofocus techniques including image metric optimization based techniques and several variants of phase gradient autofocus (PGA). The degree of focusing is evaluated with an image focus metric, specific to SAR images, that is not biased toward any particular autofocus algorithm. This evaluation is performed on a variety of scene types using injected (known) phase errors. We show that PGA autofocus outperforms the image metric optimization techniques tested (based on minimizing image entropy) in low contrast SAR scenes.