Evolution is driven to find circuits that perform larger numbers of logic functions.
A modification in a method of automated evolutionary synthesis of voltage controlled multifunctional logic circuits makes it possible to synthesize more circuits in less time. Prior to the modification, the computations for synthesizing a four-function logic circuit by this method took about 10 hours. Using the method as modified, it is possible to synthesize a six function circuit in less than half an hour.
The concepts of automated evolutionary synthesis and voltage-controlled multifunctional logic circuits were described in a number of prior NASA Tech Briefs articles. To recapitulate: A circuit is designed to perform one of several different logic functions, depending on the value of an applied control voltage. The circuit design is synthesized following an automated evolutionary approach that is so named because it is modeled partly after the repetitive trial-and-error process of biological evolution. In this process, random populations of integer strings that encode electronic circuits play a role analogous to that of chromosomes. An evolved circuit is tested by computational simulation (prior to testing in real hardware to verify a final design). Then, in a fitness-evaluation step, responses of the circuit are compared with specifications of target responses and circuits are ranked according to how close they come to satisfying specifications. The results of the evaluation provide guidance for refining designs through further iteration.
- Typical circuitry can be less complex and can occupy smaller areas; because only a single analog control line is needed to select different functions.
- If voltage-controlled multifunctional gates were used in the place of the configurable logic blocks of present commercial FPGAs, it would be possible to change the functions of the resulting digital systems in much shorter times;
- Relative to conventional circuits designed to perform single functions, multifunctional circuits can be synthesized to be more tolerant of radiationinduced faults.
In the unmodified method of automated evolutionary synthesis, the target responses of a multifunctional logic circuit are fixed: that is, the user specifies in advance which logic function the circuit is to perform at each of several discrete values of control voltage (for example, AND at 0 V, NOR at 0.9 V, and NAND at 1.8 V). In the modified method, the user no longer specifies which logic function occurs at which control voltage: Instead, the evolutionary algorithm is allowed to find the control-voltage levels at which various logic functions appear, and the fitness-evaluation function is modified to assign a higher fitness score to a circuit that exhibits a greater number of logic functions over the full range of the control voltage. Thus, evolution is driven to find circuits that perform a larger number of logic functions.
In order to be able to score fitness in this way, one must ensure that circuit output is a digital waveform at every value of the control voltage, so that the output can be classified as a particular logic function. Nevertheless, it has been observed that the circuits generated during evolutionary search typically generate analog outputs, taking values between zero volts and the power-supply voltage. In order to solve this problem, the output of an evolving circuit is digitized by use of a buffer, as illustrated in the figure. Whereas the direct output of the evolving circuit is evaluated in the unmodified method, the buffered output is evaluated in the modified method. In effect, for the purpose of evaluation, the buffer becomes part of any such evolved circuit.
This work was done by Adrian Stoica and Ricardo Zebulum of Caltech for NASA’s Jet Propulsion Laboratory. For further information, access the Technical Support Package (TSP) free on-line at www.techbriefs.com/tsp under the Semiconductors & ICs category. NPO-40934
This Brief includes a Technical Support Package (TSP).
Faster Evolution of More Multifunctional Logic Circuits; (reference NPO-40934) is currently available for download from the TSP library.
Please Login at the top of the page to download.