A circuit for implementation of nonlinear analog functions has been designed. It is based on the fuzzy-logic paradigm which has been mapped into a modular programmable architecture: a software optimization procedure is used to compute the programming values by minimizing the error between the simulated circuit response and the target I/O relationship specified by means of linguistic and/or numeric information. The layout of dedicated macrocells featuring a specific I/O characteristic is then automatically created. A chip where the programming configuration is stored in a digital RAM has been fabricated using a CMOS $0.7 \mu m$ n-well technology thus allowing the overall accuracy of the proposed system to be tested. For several examples the difference between the measured characteristics and the desired I/O relationships is in the range of a few $\%$ RMS. The spread between different chip characteristics is below $ 1\%$ RMS. The output voltage variation for temperatures ranging between $0-70^\circC$ and power supply $5V \pm 5 \%$ keeps lower than about $1\%$ of the circuit dynamic range. The applications of the proposed system to the synthesis of a non-linear control law for a DC-DC power converter and of the compensation of temperature-sensor characteristics are presented.
|Microelectronics Group Home Page||Staff|
|Research Activities||Teaching Activity|
|DEIS Home Page||University of Bologna Home Page|