A Framework For Scaling Arbitrary Fixed Point Digital Filter Structures

The precise modeling of a digital filter structure through expanded state equations is used as a basis for scaling the filter states and summers. The scaling process is structured to allow each state and summer full utilization of its dynamic range. The scaling process is user defined and includes, but is not limited to, 1 norms, peak magnitude filter operation, and ran8om noise. The measured dynamic range is then used to "normalize" the summer and state dynamic ranges by changing or adding filter multipliers. This scaling process does not change the basic filter structure. Ihe scaling gives an improved signal-to-noise ratio for the same filter dynamic range. The scaling process has been computer automated for an arbitrary filter.