Abstract
In many Multimedia and DSP applications, the fixed-width multipliers are used to avoid infinite growth in the word size. Fixed-width multiplier produces an [Formula: see text]-bit product with two [Formula: see text]-bit inputs. This paper presents probabilistic estimation technique applied for the fixed-width radix-8 Booth multiplier for the generation of the compensation bias circuit. The probabilistic estimation circuit for the fixed-width radix-8 Booth multiplier is derived systematically from theoretical computation in preference to time-consuming exhaustive simulations. Results show that the radix-8 direct truncated multiplier reduces the maximum absolute error by 33%, the average error by 22% and the mean square error by 39% for a 12-bit multiplier compared with the radix-4 direct truncated multiplier. Results also demonstrate that, with the probabilistic estimation technique applied to the fixed-width radix-8 Booth multiplier, there is a reduction of 25% in the maximum absolute error, 13.4% reduction in the average error, and 25.13% reduction in the mean square error have been realized compared with the existing fixed-width radix-4 Booth multiplier with probabilistic estimation technique. Standard EDA design tools are used for simulations.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have