模拟电路统一软故障诊断的研究

在模拟电路故障诊断的诸多方法中,模拟电路统一软故障诊断方法成为新的研究热点。从模拟电路统一软故障诊断方法的提出,到模拟电路统一软故障诊断方法的改进,将其与基于多种方法融合的模拟电路软故障诊断方法做了比较。最后重点分析了模拟电路统一软故障诊断的研究现状和存在的不足,基于实用化的角度,从统一的模拟电路故障诊断的功能模块划分原则和方法、通用的模拟电路故障可诊断性分析方法、通用的模拟电路模块化分级诊断软故障诊断方法、统一的诊断效果评估指标4个方面指出了模拟电路统一软故障诊断的发展趋势。     

Extension of the variational equation to analog/digital circuits: numerical and experimental validation

Conventional shooting methods cannot be used to determine the steady‐state solution of circuits whose model is characterised by a vector field exhibiting zero order discontinuities. In the analysis of circuits working on a stable limit cycle, this limitation prevents the use of methods that exploit Floquet theory to compute the variational model and thus the properties of the fundamental matrix and, for example, phase noise in oscillators. In this paper, we use an improved shooting method that solves this drawback by resorting to saltation matrices and show how this method makes possible the correct computation of the first left eigenfunction (known as ppv ) of the fundamental matrix. ppv s are a key aspect in determining phase noise. The results obtained through numerical simulations are compared with measurements on a relaxation oscillator serving as a simple but significant comparison vehicle. Copyright © 2012 John Wiley & Sons, Ltd.     

基于深度学习的模拟电路软故障诊断

为降低模拟电路软故障特征提取与分类的人工成本,提高软故障诊断的通用性,提出一种基于深度学习的软故障特征提取方法.利用通道注意力机制对深度学习中的卷积神经网络进行改进,将时域电压波形数据输入至改进的卷积神经网络中进行卷积池化,实现数据降维和故障特征提取,并利用注意力机制对所得的故障特征进行深度选择,突出通道内关键的故障特征,抑制不重要的特征,最终使用Softmax分类器对故障特征分类.针对四运放双二次高通滤波器进行故障诊断,故障诊断平均准确率为98％.实验结果表明改进的卷积神经网络模型可以实现对模拟电路的故障诊断,并避免了传统故障识别耗费大量人工的故障提取和选择.     

基于时间卷积网络的模拟电路故障诊断方法

对模拟电路故障诊断中的故障特征提取进行了研究,引入了卷积神经网络模型,提出了一种基于时间卷积神经网络的模拟电路故障诊断方法,在四阶Butterworth低通滤波器电路上分别对不同深度的时间卷积网络进行分类对比实验,实验结果证实了深度时间卷积网络在故障特征提取中的有效性。同时设计实验对比了时间卷积神经网络、SAE-SOFTMAX、深度信念网络和长短期记忆网络的特征提取能力,结果表明,时间卷积神经网络模型在模拟电路故障诊断中能够提取出更能反映数据本质的特征,取得更好的诊断准确率。     

模拟电路的马氏距离故障诊断方法研究

针对欧氏距离在模拟电路故障诊断中的不足,提出模拟电路的马氏距离故障诊断方法。利用MULTISIM对电路的软故障和硬故障进行仿真获取故障特征样本,在MATLAB中计算样本与故障类之间的马氏距离,并选择距离最小的故障类作为诊断结果。仿真结果表明:在20次不同特征维数的试验中,欧氏距离,正规化欧氏距离和马氏距离故障诊断方法的平均诊断准确率分别为:86.7%,87.1%和92.9%。马氏距离的诊断方法不受特征相关性的影响,能够正确诊断出欧氏距离和正规化欧氏距离诊断方法的误诊样本,在模拟电路的故障诊断中的准确率更高。     

Bridging the gap between digital circuits and microprocessors

Most electrical and computer engineering students understand digital circuits and microprocessors, but fail to appreciate that a microprocessor is just a complex finite state machine. The authors present a three experiment sequence which takes the students from the design of a simple EEPROM-based finite state machine through a two-chip microsequencer to a 4-bit central processing unit (CPU)     

数字电路封装的串扰测试方法研究

梳理了数字电路封装串扰测试方法的发展历史,总结了串扰的影响因素。基于对业界串扰测试方法的分析,比较了各种方法间的区别和优缺点,分析了美军标MIL-STD-883K中串扰测试方法很少修订却长期存在的原因。明确了美军标及其他串扰测试方法的适用范围,提出了我国相应标准制修订方向的建议。测试结果表明,标准中的现有方法在评估串扰对整体电路影响方面手段有欠缺,需要结合电磁仿真对串扰影响做出整体评估,并利用实际测试对仿真结果进行校准,从而提高模拟的准确度,此方法可以在我国相应标准的制修订中加以采用。     

Collapse of MOSFET drain current after soft breakdown

Gate-oxide soft breakdown (SB) can have a severe impact on MOSFET performance even when not producing any large increase of the gate leakage current. The SB effect on the MOSFET characteristics strongly depends on the channel width W: drain saturation current and MOSFET transconductance dramatically drop in transistors with small W after SB. As W increases, the SB effect on the drain current fades. The drain saturation current and transconductance collapse is due to the formation of an oxide defective region around the SB spot, whose area is much larger than the SB conductive path. Similar degradation can be observed even in heavy ion irradiated MOSFETs where localized damaged oxide regions are generated by the impinging ions without producing any increase of gate leakage current.     

Elman神经网络在非线性模拟电路故障诊断中的应用

Elman神经网络把隐层输出反馈回输入层,是一种动态递归神经网络,改进的Elman网络能逼近任意动态非线性系统。本文研究了Elman网络的结构和学习算法,并把改进的Elman网络应用于非线性模拟电路故障诊断激励信号参数优化中,以提高故障正确诊断率。首先用改进的Elman神经网络建立故障电路和无故障电路系统模型,然后用遗传算法搜索电路故障诊断的最佳激励信号参数。     

Circuit-level modeling of soft errors in integrated circuits

This paper describes the steps necessary to develop a soft-error methodology that can be used at the circuit-simulation level for accurate nominal soft-error prediction. It addresses the role of device simulations, statistical simulation, analytical soft-error rate (SER) model development, and SER-model calibration. The resulting approach is easily automated and generic enough to be applied to any type of circuit for estimation of the nominal SER.     

基于乘法器的模拟电路参数测量系统的设计与实现

针对现有模拟电路参数测量方法复杂、测量结果精度低的缺点,构建了基于乘法器的模拟电路参数测量系统。系统中由信号源产生2路正交同频正弦模拟信号,取出一路信号通过待测模拟电路与原信号源产生的2路信号做乘法、滤波处理,产生两路直流信号,之后通过DAQ、LABVIEW采集,得到模拟电路相位差和幅度。经测量,幅度和相位的误差均小于3％,表明该系统具有电路简单、易于实现、误差小、运算速度快等优点,为模拟电路参数测量的研究提出一种可行性方案。     

基于小波神经网络的模拟电路故障诊断改进

在运用神经网络进行模拟电路故障诊断的过程中,代表着故障特征的网络输入至关重要,由于小波变换的时频局部化和多尺度分析等特性,将两者结合起来,通过小波变换对模拟电路的输出响应进行故障特征提取,同时解决PSPICE与MATLAB之间的数据通信问题,提出将蒙特卡罗分析产生的所有训练样本经过处理后输入到一个神经网络进行训练的方法,从而避免了训练多个神经网络。利用神经网络对各种故障模式进行分类,实现模拟电路的故障诊断,并进一步与传统的BP网络故障诊断法进行比较。仿真结果表明,该方法可以实现故障检测及定位,诊断的准确率显著提高,适用于模拟电路故障诊断。     

Analysis of soft breakdown failure with ESD on output buffernMOSFETs and its improvement

The leakage increase of the off-state MOSFETs after an ESD event has been studied for output transistors with the thin gate oxide and LDD structures. Leakage increase called “soft breakdown” has been found at relatively low ESD testing voltages (200-300 V). This soft breakdown is caused by the creation of interface traps due to the snap-back stressing during the ESD event. The creation of interface traps has enhanced the interface trap to band tunneling current at the drain side of the MOSFETs. The improvement of the ESD threshold has also been proposed with an additional arsenic implantation into the n^$ region. It has been confirmed that the arsenic implantation improved the HBM ESD threshold to more than 2000 V     

马氏距离在模拟电路硬故障检测中的应用研究

针对模拟电路硬故障检测问题,提出了一种基于马氏距离的快速检测算法,并利用pspice电路仿真软件对一个二阶滤波电路进行了电路建模及故障检测仿真.在pspice环境下,进行含容差待测电路无故障和各故障模式下的蒙特卡洛仿真,选取输出电压作为采样点,利用输出电压的小波分解系数求出无故障态和各故障模式下的时-频马氏距离,并以此作为电路故障判据.matlab下的计算结果显示,这种故障检测算法对模拟电路硬故障具有良好的检测能力,故障诊断正确率可达92%以上,且具有算法简单、计算量小的优点.     

多特征故障字典及其在模拟电路可测性分析中应用

提出一种新的模拟电路可测性分析模型——多特征故障字典模型。该模型首先获得输出端的阶跃响应输出波形,提取出波形的多个特征量,将所有特征量形成一特征向量,不同故障类别对应的特征向量组成一特征矩阵。然后由PSPICE软件中的Monte-Carlo分析的结果计算出阈值,在该阈值下将特征矩阵转换成整数编码表,从而计算可测性分析指标-检测率与隔离率。仿真与实测数据表明:与传统模拟电路可测性分析方法相比,提出的方法优点有:1)不需要传递函数,可以适用于复杂电路;2)仅需要一个输出测点,容易实现;3)对于容差模拟电路,同样可以取得较好的检测率与隔离率。     

基于可编程电源的数模动模混合仿真系统研究

为满足现代电力系统实时仿真的发展需求,提出一种基于可编程电源的数模动模混合仿真系统。利用动模实验室搭建动模仿真系统,并在RTDS中搭建39节点系统作为数字仿真系统模型,结合以理想变压器模型算法作为功率接口算法的接口系统,形成一个800 V/345 kV的动模数模混合实时仿真系统实验平台。在研究混合仿真接口两端系统延时对系统稳定性和准确性影响基础上,提出用二阶相位超前环节对下发通道和上传通道的接口延时进行相位补偿,以提高系统稳定性和准确性。通过对所搭建的数模动模混合仿真系统的实验,并与纯数字仿真进行对比分析,进一步验证了混合仿真系统的有效性。     

Multiple catastrophic fault diagnosis of analog circuits considering the component tolerances

This paper deals with the diagnosis of multiple catastrophic faults, being cuts (open‐circuits) of some connecting paths and/or short‐circuits of some pairs of points in analog circuits. A method enabling us to detect and identify the faults, taking into account the deviations of the circuit parameters within their tolerance ranges, is developed. The method exploits an appropriate fault dictionary. The fault dictionary is used only for preliminary identification of the faults, because it is based on the analysis of the circuits with nominal parameters. The crucial point of the method is a verification procedure, proposed in this paper, based on the linear programming approach. It leads to the results considering the component variations within their tolerance ranges. In addition, a procedure for selecting appropriate test points, employing some evolutionary techniques, is developed. Although the approach presented in this paper is described in detail for linear circuits, it can be directly generalized to nonlinear circuits. Three numerical examples, including two linear and one nonlinear circuits, illustrate the proposed method and show its efficiency. Copyright © 2011 John Wiley & Sons, Ltd.     

Operation of analog MOS circuits in the weak or moderate inversion region

In analog metal-oxide-semiconductor (MOS) circuit design, the ratio of channel width to channel length (W/L) can become very high, often exceeding a value of 1000. This high ratio may result either from wide channel devices, designed to achieve high transconductance, or from short-channel-length devices used in high-frequency circuits. Since drain current is generally proportional to the W/L ratio, the effective gate-to-source voltage to achieve a desired drain current is typically quite low. In many cases, this set of circumstances leads to operation in the weak or moderate inversion region accompanied by electrical behavior that differs significantly from that in the strong inversion region. In designing analog MOS circuits, an understanding of operation below the strong inversion region is necessary to relate design predictions to simulation results. In addition, biasing simple amplifier stages near the weak or moderate inversion region can be used to advantage. Such a bias can lead to maximum voltage gain, low device dissipation, and minimum total harmonic distortion.     

基于FPGA的多路A/D转换控制接口设计

在嵌入式系统设计中,常涉及到ADC的处理问题。文中介绍一种基于FPGA(FieldProgrammableGateArray)的多路ADC(AnalogtoDigitalConvertor)控制转换器,它能够独立地控制多路A/D转换,并能实时地将各路ADC的结果存储到相应的寄存器中。利用该接口控制器能极大地降低ADC对MCU的资源占用,提高嵌入式系统的效率。     

数字—物理模型互联方法及混合仿真系统稳定性研究

现代电力系统通常包含高压直流输电(HVDC)或灵活交流输电(FACTS)等由开关器件构成的高频元件,单一实时数字模型因其离散性而受仿真截止频率限制,无法准确模拟这类元件的高频响应过程,目前最有效的仿真措施是建立包含这类高频元件物理模型的数字-物理互联混合模型.根据电路替代定理提出节点电压重叠互联方法,即在数字模型中将接口点电压变量通过D/A转换输出通道形成物理模型中的电压源替代支路,又将该物理支路的电流变量通过A/D转换输入数字模型中形成接口支路的替代电流源,从而实现数字一物理模型的互联;从系统稳定性的角度对基于该方法互联的混合仿真模型进行了理论分析,证明混合仿真模型为有条件数值稳定系统且推导出模型的稳定条件判别式;最后,通过混合仿真实验验证了理论分析结果的正确性.