Abstract: A quantizaiton method is presented to aim at the fractional saturation Synthetic Aperture Radar (SAR) raw data. Norm distribution is the basis of Block Adaptive Quantization (BAQ) algorithm. But in the application of BAQ, some trouble may encounter, one of these is fractional saturation of raw data. A compression algorithm based on the traditional BAQ algorithm is presented, which can select different ways to quantify the raw data. This algorithm reduces the error and increases the Signal to Quantization Noise Ratio (SQNR) without influence the compression ratio. Finally, a block of real SAR raw data are processed by the two algorithms and the result is compared with each other.