基于神经网络的模数转换电路动态误差源识别系统设计  

Design of neural network based dynamic error source identification system for analog-to-digital conversion circuit

作　　者：张晋博 丁传红 ZHANG Jinbo;DING Chuanhong(Beijing Institute of Computer Technology and Applications,Beijing 100854,China)

机构地区：[1]北京计算机技术及应用研究所

出　　处：《现代电子技术》2019年第21期53-57,共5页Modern Electronics Technique

摘　　要：为提升信号识别电路的电量采集精度,实现理想状态下的电力误差校准,设计基于神经网络的模数转换电路动态误差源识别系统。以CNN神经网络作为模数转换电路的物理依赖环境,通过合理选取动态识别元件的方式,实现误差源识别系统的硬件运行环境搭建。在此基础上,将模拟电流转化成数字信号,再将其完整存储于系统数据库中,利用既定数学运算公式对已存储的数字信号进行识别精度提纯处理,实现误差源识别系统的软件运行环境搭建,联合相关硬件执行设备,完成基于神经网络的模数转换电路动态误差源识别系统设计。实际应用结果表明,在加压环境下,新型误差源识别系统的电量采集精度达到90%,单位时间内的信号识别量超过7.5×10^9TB,理想状态下信号识别电路的电力误差校准能力得到有效保障。In order to improve the acquisition accuracy of signal recognition circuit and realize the power error calibration in ideal state,a neural network based dynamic error source identification system of analog-to-digital conversion circuit is designed.The hardware operation link of the error source identification system is constructed by reasonably selecting the dynamic identification elements and taking CNN neural network as the physical dependent environment of the analog-to-digital converter circuit.On this basis,the analog current is converted into digital signals,and then they are stored completely in the system database.The purifying processing for the stored digital signal is conducted to make identification precision improved by means of the established mathematical formula.The software running link construction of the error source recognition system is realized.The neural network based dynamic error source recognition system for the analog-to-digital conversion circuit is completed by combining the relevant hardware equipments.The practical application results show that,in the pressurized environment,the power acquisition accuracy of the new error source identification system reaches 90%,the signal recognition quantity per unit time exceeds 7.5×10^9 TB,and the power error calibration ability of the signal recognition circuit under ideal conditions is effectively guaranteed.

关 键 词：神经网络 模数转换 电路误差源 动态识别 数字信号存储 系统设计 

分 类 号：TN792[电子电信—电路与系统] 34[自动化与计算机技术—计算机应用技术] TP391[自动化与计算机技术—计算机科学与技术]