Ultrasonic sensor performance improvement using DSP-FPGA based architectures

Traditional ultrasonic systems are based on isolated transducers, which provide reduced information with a considerable acquisition time. To improve this performance, transducers can be associated in more complex sensors, taking advantage of some geometrical questions. Also, multi-mode techniques (simultaneous emission and reception) allow increase of the scanning frequency. These systems are able to provide more information, even for carrying out high level tasks. But they require more computational capacity to implement the new processing algorithms. First, it is necessary to measure accurately the ultrasonic times of flight, using correlation techniques. Then, geometrical and probabilistic approaches are performed to interpret correctly the measured times of flight in high level tasks, as reflector classification or mapping. This work develops a sonar sensor, and it proposes a DSP-FPGA based architecture to process the ultrasonic signal. Used algorithms are explained, and an operation mode is developed to show how the different tasks are performed by the DSP or the FPGA, according to their intrinsic characteristics.