Synchronous dimming control for a cold- cathode fluorescent lamp driver - Electronics Letters

lation is created of randomly transformed copies of the original DFG. A fitness is allocated to each member of the population. This is determined by calculating the supply voltage and effective capacitance of the DFG. Area requirements are computed for the adders, delays and multipliers. The total required area is used to estimate the effective capacitance. The GA selects a member of the population for transformation using the probabilistic roulette wheel [6] method. A library of transforms is used, from which transforms are selected at predetermined rates. (Pipelining = 1 %I, retiming = 20%, unfolding = lo/, auto-pipelining = 1%). These rates were calculated using a combination of design heuristics and experimental data. Repeated application of the transformations will form the new population of designs. Genetic evolution terminates when the design with required power specification is produced, i.e. that design with the highest fitness over all generations. The C A may produce multiple designs that satisfy the power requirements, giving greater flexibility at the implementation stage.