| The history of digital design is one of continuous movement towards ever-higher been abandoned for other than pedagogical purposes, today's design methodologies of hand coded hardware description language (HDL) will eventually prove to require impractical levels of human effort. Advances in semiconductor technology permit design complexity to grow exponentially, yet engineering manpower has no such increase. Removing the tedious and error prone tasks of manual design will enable engineers to work at the system level, relegating HDL development and component verification to automated design tools, transparent to the user.; This dissertation proves the practicality of moving to a higher level demonstrate the superiority of automating tasks that are now considered to require a high degree of engineering expertise.; Finite Impulse Response (FIR) filters are the class of digital designs selected for this development of design automation technology. Based on the principle that architectural tradeoffs can be quantified, predicted and thus automated, the synthesizer presented here can perform analysis and repetitive calculations to find an optimal design that a human designer would find overly burdensome and time consuming.; Because filter design with this synthesizer is so temporally efficient, it was possible to use it to generate vast numbers of filters and gather empirical data regarding their performance. Analysis of these data shows that conventional wisdom regarding optimization strategy is frequently misguided. These data were then used to refine the synthesizer, which is now capable of producing designs to rival and surpass hand crafted ones, yet requiring minimal operator intervention or knowledge of circuit design. |
