In this paper the authors deal with the implementation of the neural network on an Xilinx based field programmable gate array for a non - linear system. Here back propagation neural network is used. The implemented hardware is then used to efficiently calibrate the U-tube manometer, which is the relation between the level of mercury and the capacitance developed across the copper plates of the manometer. The relationship is found to be highly non-linear. The results shows neural network effectively measures the levels of mercury in U-tube by training frequency and acts as efficient tool for non linear systems.