METHOD OF PROCESSING SERIES OF COMPUTATIONS ACCORDING TO LINEAR FORECAST ENCODING METHOD AND DIGITAL LATTICE FILTER FOR EMPLEMENTING THE SAME

摘要

A lattice filter for processing lattice equations includes a fast adder (78) for adding partial products to partially perform a multiplication step. A full adder (44) is provided for completing the multiplication and then adding the product with a previously calculated and stored value. The input to the full adder (44) is multiplexed with a multiplexer (74) for selecting the sum output of the fast adder (78) and a multiplexer (76) for selecting the carry output of the fast adder (78). The multiplexer (74) also selects prestored values for addition with the summed output of the full adder (44). This summed output is selected by the multiplexer (76). The fast adder (78) sums partial products simultaneous with addition operations of the full adder (44). In this manner, the full adder (44) operates at a slower rate than the fast adder (78). Storage registers (58), (62), (70) are utilized to delay results output by the full adder (44) for later selection and operation thereon. These values are utilized as both the multiplicand the addend in subsequent operations. The multiplier is stored in a K-stack (90) and selected for the appropriate operations. A bit correction circuit (190) provides corrections for truncated bits in the form of a carry input to the full adder (44).