A study of spaceborne SAR raw data compression error based on a statistical model of quantization interval transfer probability

SAR raw data signal to noise ratio (SNR) after compression is of great importance since the choice of compression ratio is dependent on it during SAR system design and application analysis. The signal to quantization noise ratio (SQNR) generally used may not precisely indicate the relationship between signal and noise. Considering the thermal noise, a statistical model of quantization interval transfer probability is proposed in this paper. SNR mapping between SAR raw data before analog to digital converter (ADC) and after block adaptive quantization (BAQ) over the whole set of saturation degree is obtained using this model. When the power of echo is small with low SNR, after 1, 2, 3 or 4 bits BAQ compression, SNR has tiny difference among the four compression levels. When the power of the echo is medium with higher SNR, the SNR degradation after BAQ is about 5 dB with each bit decreasing from 4 bits. If voltage of the echo is higher than the clipping point of ADC, SNR after ADC and BAQ degrades stepwise. The higher the saturation degree of SAR raw data, the worse the SNR is. Simulated Gaussian data and real SAR raw data are used to verify the theoretical results, which are useful in the choice of BAQ compression ratio and further application analysis.