Title :
Discrete time analog polynomial type CNN with digital state
Author :
Laiho, Mikn ; Paasio, Ari ; Kananen, Asko ; Halonen, Kari
Author_Institution :
Electron. Circuit Design Lab., Helsinki Univ. of Technol., Espoo, Finland
Abstract :
In this paper a CNN with digital state and analog multiplication is presented. Better control of delays of different order feedback terms can be achieved with the mixed mode approach. Also, errors generated at the beginning of transients can be reduced and storage of the initial states is straightforward in the digital domain. Euler-Cauchy, Heun´s and Runge-Kutta integration methods are compared in terms of convergence time and aspects of cell level implementation. Heun´s method is found to be the best compromise in hardware realizations where network dynamics are critical. Also, a current squarer and four quadrant multiplier with low quiescent current are presented
Keywords :
Runge-Kutta methods; cellular neural nets; circuit feedback; delays; discrete time systems; mixed analogue-digital integrated circuits; multiplying circuits; neural chips; Euler-Cauchy integration methods; Heun´s integration methods; Runge-Kutta integration methods; cell level implementation; convergence time; current squarer; delays; digital state; discrete time analog polynomial type CNN; feedback terms; four quadrant multiplier; hardware realizations; mixed mode approach; network dynamics; quiescent current; Algorithm design and analysis; Cellular neural networks; Delay effects; Electronic circuits; Epilepsy; Feedback; Laboratories; Neural network hardware; Polynomials; Voltage fluctuations;
Conference_Titel :
Circuits and Systems, 2001. ISCAS 2001. The 2001 IEEE International Symposium on
Conference_Location :
Sydney, NSW
Print_ISBN :
0-7803-6685-9
DOI :
10.1109/ISCAS.2001.921356