A programmable g/sub m/-C CNN implementation

An implementation of a programmable cellular neural network is reported. It overcomes some of the limiting characteristics and restrictions inherent in CMOS VLSI technologies, and allows an arbitrarily large continuous-time analog CNN to be built up by modularly connecting CNN chips with a modest number of cells. The template values are implemented as sets of unit and half-unit OTAs and are digitally step-wise programmable. The design incorporates an offset compensation and initialization circuit. All external input, output and control signals are electrical and digital. The design was carried out in a 0.8 /spl mu/ CMOS technology. Each cell occupies 0.78 mm/sup 2/, including all support circuitry. Matching accuracy was measured and operation was verified on numerous uncoupled and propagation-type templates.

[1]  George S. Moschytz,et al.  Stochastic and hybrid approaches toward robust templates , 1998, 1998 Fifth IEEE International Workshop on Cellular Neural Networks and their Applications. Proceedings (Cat. No.98TH8359).

[2]  G. Moschytz,et al.  The Influence of the Boundary Conditions on the Robustness of a CNN , 1998 .

[3]  Ákos Zarándy,et al.  CNN template design strategies and fault tolerant CNN template design - a survey , 1997 .

[4]  D. Lim,et al.  A programmable, modular CNN cell , 1994, Proceedings of the Third IEEE International Workshop on Cellular Neural Networks and their Applications (CNNA-94).

[5]  Ángel Rodríguez-Vázquez,et al.  A CNN UNIVERSAL CHIP IN CMOS TECHNOLOGY , 1996 .

[6]  B. Mirzai,et al.  Applications of CNN processing by template decomposition , 1998, 1998 Fifth IEEE International Workshop on Cellular Neural Networks and their Applications. Proceedings (Cat. No.98TH8359).

[7]  A. Dawidziuk,et al.  Minimum size 0.5 micron CMOS programmable 48 by 48 CNN test chip , 1997 .

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