基于稳健神经网络的CMOS电路开关级设计  

Design of CMOS circuits at switch level based on robust neural network

作　　者：王柏祥[1] 林弥[1] 吕伟锋[1] 

机构地区：[1]浙江大学信息与电子工程学系,浙江杭州310028

出　　处：《浙江大学学报（工学版）》2004年第7期835-838,共4页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(210200N198).

摘　　要：基于数字电路开关级设计的优越性、稳健布尔神经网络和开关级电路结构的相似性,提出了一种基于稳健布尔神经网络的互补金属-氧化物半导体(CMOS)晶体管电路的开关级设计方法.该方法通过神经元激活函数推导开关级电路的开关函数,同时引入神经网络中卡诺图化简和最小项限制的化简方法.所得电路可以和传统的开关级电路相吻合,电路结构清晰,并进一步减少了金属-氧化物半导体(MOS)晶体管的数目.Taking advantage of the superiority of switch level design and the similarity between the robust Boolean neural network structure and the circuit structure designed at switch level, a design method for the complementary metal-oxide semiconductor (CMOS) transistor circuit based on the robust neural network was proposed. This method takes the switch operations for CMOS circuits through the active functions of neurons, and takes the neural network's k-map and the minterm suppression into consideration. The obtained circuits can accord with the traditional switch level circuits. As a result, the circuit structure is clear and the number of MOS transistor can be reduced further.

关 键 词：稳健神经网络 数字逻辑 CMOS电路 开关级设计 

分 类 号：TN432[电子电信—微电子学与固体电子学] TP183[自动化与计算机技术—控制理论与控制工程]