M. Andrejevic

Analogue electronic circuit diagnosis based on ANNs

Feed-forward artificial neural networks (ANNs) have been applied to the diagnosis of nonlinear dynamic analogue electronic circuits. Using the simulation-before-test (SBT) approach, a fault dictionary was first created containing responses observed at all inputs and outputs of the circuit. The ANN was considered as an approximation algorithm to capture mapping enclosed within the fault dictionary and, in addition, as an algorithm for searching the fault dictionary in the diagnostic phase. In the example given DC and small signal frequency domain measurements were taken as these data are usually given in device’s data-sheets. A reduced set of data per fault (DC output values, the nominal gain and the 3 dB cut-off frequency, measured at one output terminal) was recorded. Soft (parametric) and catastrophic (shorts and opens) defects were introduced and diagnosed simultaneously and successfully. Large representative set of faults was considered, i.e., all possible catastrophic transistor faults and qualified representatives of soft transistor faults were diagnosed in an integrated circuit. The generalization property of the ANNs was exploited to handle noisy measurement signals.

Fault Diagnosis in Digital Part of Sigma-Delta Converter

In this paper the artificial neural network (ANN) is applied to diagnosis of defects in the digital part of a nonlinear mixed-mode circuit. Both catastrophic and delay defects are considered. The approach is demonstrated on the example of a relatively complex sigma-delta modulator. Delay defects in this example are delays of rising and falling edge of digital signals and catastrophic defects are considered as stuck switches. Fault dictionary is created, by simulation, using the response of the circuit to an input ramp signal. It is represented in a form of a look-up table. Artificial neural network is then trained for modeling (memorizing) the look-up table. The diagnosis is performed so that the ANN is excited by faulty responses in order to present the fault codes at its output. There were no errors in identifying the faults during diagnosis

Behavioural modelling, simulation, test and diagnosis of MEMS using ANNs

The design of micro-electrical-mechanical systems requires that the entire system can be modelled and simulated. Additionally, behaviour under fault conditions must be simulated to determine test and diagnosis strategies. While the electrical parts of a system can be modelled at transistor, gate or behavioural levels, the mechanical parts are conventionally modelled in terms of partial differential equations (PDEs). Mixed-signal electrical simulations are possible, using e.g. VHDL-AMS, but simulations that include PDEs are prohibitively expensive. Here, we show that complex PDEs can be replaced by black-box functional models and, importantly, such models can be characterized automatically and rapidly using artificial neural networks (ANNs). We demonstrate a significant increase in simulation speed and show that test and diagnosis strategies can be derived using such models.

Fault Diagnosis in Digital Part of Mixed-Mode Circuit

In this paper artificial neural networks (ANNs) are applied to diagnosis of catastrophic defects in the digital part of a nonlinear mixed-mode circuit. The approach is demonstrated on the example of a relatively complex sigma-delta modulator. A set of faults is selected first. Then, fault dictionary is created, by simulation, using the response of the circuit to an input ramp signal. It is represented in a form of a look-up table. Artificial neural network is then trained for modeling (memorizing) the look-up table. The diagnosis is performed so that the ANN is excited by faulty responses in order to present the fault codes at its output. There were no errors in identifying the faults during diagnosis

ANN application in electronic diagnosis-preliminary results

In this paper artificial neural networks (ANNs) are applied to diagnosis of catastrophic defects in a linear analog circuit. In fact, today the technical diagnosis is great challenge for design engineers because the diagnostic problem is generally underdeterminate. It is also a deductive process with one set of data creating, in general, unlimited number of hypotheses among which one should try to find the solution. So, the diagnosis methods are mostly based on proprietary knowledge and personal experience, although they were built into integrated diagnostic equipment. ANN approach is proposed here as an alternative to existing solutions, based on the fact that ANNs are expected to encompass all phases of the diagnostic process: symptom detection, hypothesis generation, and hypothesis discrimination. The approach is demonstrated on the example of a simple resistive electrical circuit, and the generalization property is shown by supplying noisy data to ANNs inputs during diagnosis.

ANN Application to Modelling of the D/A and A/D Interface for Mixed-mode Behavioural Simulation

Artificial neural networks are applied for modeling the input and output circuits of the digital part of the digital–analog and analog–digital interface, respectively, in CMOS mixed-mode circuits. The generalization property of the neural networks is exploited to apply the models in a set of previously unknown situations, the most important being loading the model generated from the unloaded circuit. The models developed are applicable in mixed-signal behavioral simulations.

Electronic modelling using ANNs for analogue and mixed-mode behavioural simulation

We describe the state of the art and some preliminary results obtained by application of artificial neural networks (ANN) to modelling of dynamic non-linear electronic circuits. ANNs are used for application of the black-box modelling concept in the time domain. The ANNs topology, the testing signal used for excitation, together with the complexity of the ANN is considered. Examples of Pion-linear dynamic modelling are given encompassing a wide variety of modelling problems. Verification of the concept is performed by verifying the ability of the model to generalize i.e. to create acceptable responses to excitations not used during training. Implementation of these models within a behavioural simulator is exemplified.

Chaos Based Analog-to-digital Conversion of Small Signals

In this paper a typical one-dimensional chaotic system, a tent map, can be introduced. This tent map implements A/D conversion function as well as small signal amplifying function. We can show how this system can be applied to analog-to-digital conversion of different types of signals. In order to get full characterization, we made three experiments with the circuit. We performed A/D conversion of 1) a small DC signal, 2) a sine wave, and finally 3) a ramp signal. A correction scheme, taken from literature, was applied in order to get better accuracy. In addition, simulations were performed both with ideal and realistic models of operational amplifiers so getting better information on the circuit behavior

Modeling the D/A interface for mixed-mode behavioral simulation

Artificial neural networks (ANNs) are applied for modeling the output circuits of the digital part of the analog/digital interface in CMOS mixed-mode circuits. The generalization property of the neural networks is exploited to apply the models in a set of previously unknown situations, the most important being loading the model generated from unloaded circuit. The models developed are applicable in a mixed-signal behavioral simulation environment.