A novel compact architecture for a programmable full-range CNN in 0.5 /spl mu/m CMOS technology

Describes an analogue hardware implementation of a programmable full-range CNN. The used technology is the MIETEC 0.5 /spl mu/m CMOS process. The most important building blocks in each cell are its multipliers and an integrator with a hard-limited output. For the multipliers it is shown that the application of 2-quadrant types suffices, without loss of generality of the resulting network. As the number of multipliers per cell can be quite large, this means an important reduction of the circuit complexity. The integrator is implemented as a single capacitor. Hard-limiting is incorporated by a small clamper circuit. The resulting low-power and low-voltage circuit stands out for its low number of components and dense implementation. Its usefulness is illustrated with simulation results of this CNN used as a connected component detector.

[1]  Lin-Bao Yang,et al.  Cellular neural networks: theory , 1988 .

[2]  Kari Halonen,et al.  High speed 16 by 16 CNN chip , 1996, 1996 IEEE International Symposium on Circuits and Systems. Circuits and Systems Connecting the World. ISCAS 96.

[3]  F. Lobato-Lopez,et al.  A very fast CMOS artificial cellular neural network , 1996, 1996 IEEE International Symposium on Circuits and Systems. Circuits and Systems Connecting the World. ISCAS 96.

[4]  J.A.E.P. van Engelen Implementation of an analogue programmable Cellular Neural Network , 1995 .

[5]  Leon O. Chua,et al.  A current-mode DTCNN universal chip , 1994, Proceedings of IEEE International Symposium on Circuits and Systems - ISCAS '94.

[6]  Á. Rodríguez-Vázquez,et al.  Current-mode techniques for the implementation of continuous- and discrete-time cellular neural networks , 1993 .