The classical PGA algorithm assumes that the azimuth phase error is redundant in the range direction. Therefore the range gates that contain strong point targets have to be selected. The isolated, strong targets are selected, so that the phase error can be accurately estimated. There are different target selection strategies, including Energy Maximum [5], SNR Maximum [6], Contrast Ratio Maximum [7] and so on. The combination of Energy and Contrast Ratio Maximum strategy shows a promising performance in the target selection. Define the contrast ratio C_{n} of the nth range gate as

where 1П denotes the amplitude of the nth range gate. By setting up the threshold of C_{n}, the range gate can be selected.

In the selected range gates, the strongest point targets must be circular shifted into the center of the scene to remove the impact of the first-order phase error.

(b) Adding Windows

Adding windows to the strong point target is to contain the phase error and eliminate interference from the phase of other targets and noises. The incorrect length of the windows will induces estimation errors, thus the window must be added accordingly.

In practical PGA algorithms, different window length selection strategies are chosen according to the scene. In the case of isolated strong target, the window is 1.5 multiplies the length of 10 dB [4]. In the case that no isolated strong target exists in the scene, the window is calculated by using the iterations.

An effective window length calculation method is presented in [6]. Suppose the power of an area far from the peak is P_{c}, the width from the peak to P_{c} is W_{b}, the length of fast-descending area is W_{a}, then the window length can be calculated as