A novel FIR filter implementation using truncated MCM technique

Multiplication is performed frequently when implementing finite impulse response (FIR) filters. In most cases, rounded products are necessary to circumvent the growth of word length in order to reduce the area requirement and power dissipation and to speed up the circuit. However, due to rounding or truncating, errors are caused. In addition, in the FIR filter implementation, because of internal truncation, the coefficient multiplication varies based on the input. This makes directly analyzing its frequency response impossible. In this paper, FIR filter implementation using truncated multiple constant multiplication (MCM) is proposed, while developing a method to estimate the error due to truncation. Simulation indicates that our technique can reduce the number of full adders by nearly 30 percent, compared to traditional MCM algorithms, while only causing small errors in the frequency response.