CTM on multiprocessor: solution for bottleneck of SAR

The corner turn phase, which is the real-time matrix transpose, forms the crucial memory bottleneck in the image formation process of the SAR. For the limited memory of the DSP, we cluster multiple DSPs to form a distributed memory multiprocessor system to satisfy the memory requirement in adaptive radar. In our solution, we reduce the total memory requirements by using a new structure of the corner turner matrix at first. With a Latin square as a template, the system performance is maximized. It resolves the memory bottleneck problem of the SAR image formation system with minimum additional cost.