The Research And Design Of Intelligent Pressure Transducer Based On MSP430F448

Piezoresistive sensors produced using MEMS technology on a silicon chip, is widely used in all kinds of pressure measurement, such as tire pressure, deformation pressure and transient pressure. However, there are temperature drift, consistency and nonlinearity in the actual application. In order to improve the accuracy of measurement, it is necessary to compensate the output signal of sensor using transducer which is consist of calibrating circuit, diver circuit and transmission interface. Based on the research of nonlinearity and existing compensating algorithm, the intelligent pressure transmitter of piezoresistive sensor is designed with MSP430F448 as core, and function- link artificial neural network and LM algorithm are used as main arithmetic.The research of the designing scheme of transmitter, intelligent compensating algorithm, analog signal processing circurt, hardware and software designing and the program of compensation are introduced in paper.In intelligent pressure transducer, the instrumental microprocessor MSP430F448 is the core of system. The ADC in processor and DS18b20 is used to sample pressure and temperature data. Using the strong transmission ability of MSP430F448, the results of compensating is transmitted through SPI, UART interface and displayed on LCD. As the clock management function, the transducer can be set to low power mode, which can save the power consumption. In the power supply system, the reference voltage of sensor drive circuit and ADC is using the same source, which could offset the interference from 2.5V source.Function-link aritificial neural network and Levenberg-Marquardt algorithm are used as compensating arithmetic in system, and the column-row matrix-multiplying method which can reduce the memory space of Jacobian matrix and sample data compared with traditional row-column way is used during the computation of correction. To improve the training speed and compensation accuracy, the decay of variation is recommend into LM algorithm. The 1-wire protocol has to be programmed in microprocessor, because the DS18b20 communicates with MSP430F448 through 1-wire bus. And the network training results could be send to computer through serial interface and detected. All software of the system is programmed with C code.The system is achieved by hardware and software debugging. Operating stabilization and credibility are approved by spot test. The analysis of a large amount of experimental datas proves that the system is feasible.