Research Article  Open Access
Abhilasha Rani Goel, Mohd Wajid, "Optimal Method for Catastrophic Faults Diagnosis in RC Ladder Network", Advances in Power Electronics, vol. 2016, Article ID 7176981, 15 pages, 2016. https://doi.org/10.1155/2016/7176981
Optimal Method for Catastrophic Faults Diagnosis in RC Ladder Network
Abstract
The RC ladder network has been analyzed for various catastrophic fault detection using minimal number of measurements. Generally, electronic circuit testing procedure is very exhaustive and includes higher cost; the presented approach will save fault diagnosis time. It is not possible to analyze the big RC ladder network to give the good fault coverage, so the ladder network has been broken into segments of different sizes. However, if segment size is small, it will cause more area overhead compared to bigger step size in terms of the interconnections and pins on the integrated circuit. A systematic and detailed analysis for onestep, twostep, threestep, and fourstep RC ladder networks has been carried out for various faults and optimal step size is proposed. It has been investigated that three measurements are optimal to localize different catastrophic faults in a RC ladder network.
1. Introduction
The resistive ladder network has been analyzed for detecting various catastrophic faults associated with it, where only resistive components are considered [1]. However, there are many circuits which use capacitor also, like RC ladder network. The ladder networks with resistor and capacitors are widely used in filters, phase shifter circuits, oscillators, and so forth [2, 3]. As per the knowledge of authors, the related literature on the fault diagnosis in RC network is limited. In this and next paragraph, we have presented available existing methods/algorithms associated with fault identification in the circuits. Huang et al. have presented an approach based on an assemblage of learning machine that is trained to guide through diagnosis decision. It diagnoses the hard/soft fault by using defect filter. The hard faults have been diagnosed using the multiclass classifier and the soft fault is diagnosed using the inverse regression functions. The disadvantage of this method is in resolving the ambiguity; this method uses some auxiliary circuit specific fault diagnosis rules [4]. Kyzioł et al. [5] have given an algorithm to diagnose the catastrophic faults in analog circuits. This algorithm uses the Particle Swarm Optimization (PSO); it uses more than one dimension like load resistance and reactance, generator resistance, and reactance and generator frequency to diagnose the faults. They have shown that increasing the numbers of dimensions of search space influences the identification of states of circuit under test (CUT). The disadvantage of this method is that it can diagnose only single catastrophic fault of CUT.
Starzyk et al. [6] have proposed an algorithm based on entropy index of available test points, where twodimensional integer coded dictionary is created whose entries are measurements associated with faults and test points. Though this algorithm can be used for medium and large networks also, it is very costly. Huang et al. [7] have given a method to diagnose the local spot defects in analog circuit. This method is based on the combination of multiclass classifiers that are trained using data from fault simulation. The problem has been viewed as pattern recognition task. The method starts by inductive fault analysis (IFA) which results in a list of probable defects; then the defects are ranked based on their probability of occurrence. After that, multiclass classifiers are used to detect the fault. It is a probability based method, not so much accurate.
In the present work, we have analyzed the RC ladder network (Figure 1) for fault diagnosis. Two catastrophic faults, open and short, are considered in resistors and one catastrophic fault is considered in capacitors; that is, capacitor impedance is equal to zero. In integrated circuit, the separation between two wires cannot be very large, so infinite impedance of capacitor is not considered. Different size segments of RC ladder networks, that is, 1step, 2step, 3step, and 4step, are assessed. In this paper, the impedance is varied as a function of as given in (1), where and other parameters , , and are kept constant ( KΩ, pF, and rad/sec):We have assumed that the network has only one fault at a time. is an equivalent impedance measured between th and th nodes (Figure 1). Notations gives the list of notations used.
To calculate the fault coverage, two characteristics are considered, that is, distinguishability and ambiguity. Distinguishability means that the impedance plot for any faulty case can be distinguished from the impedance plot of the faultless case, but it will not cover the overlapping of plots of any two distinct faults. Because distinguishability does not cover the overlapping of two plots for any two faulty cases, ambiguity is defined separately. Ambiguity means that the plots for any of the two faulty cases overlap each other and it does not cover the faultless case. A deviation (see (2)) of 2% is considered as acceptable value in the given impedance plots as taken in [1]. If the plot for one fault lies within 2% of another plot for the given value of , then both of those plots are considered as nondistinguishable or ambiguous:
2. Analysis
In this section, a large RC ladder network is broken into small size RC ladder networks, namely, onestep, twostep, threestep, and fourstep RC network. These different step size networks are analyzed for fault detection for different values of , . In this paper, all graphs are shown only for to , because for to 4 some of the plots have very high values as compared to others, so that they cannot be plotted. Also, it has been assumed that . All the analysis is done mathematically and all the graphs are plotted in the MATLAB software.
2.1. SingleStep RC Ladder Network
The singlestep RC ladder network is shown in Figure 2. In this network, only three faults exist, that is, short circuit, open circuit, and which is short; that is, :
It can be concluded from Table 1 that all the three values of for different fault types are distinct and faults can be detected with no ambiguity, so the fault coverage is 100%. But computational complexity and area overhead are very large, because number of measurements will be equal to number of components and many numbers of singlestep RC ladder network have to be tested to identify the faults of a big network [8]. Also, large numbers of observable nodes are required, so this cannot be considered from integrated circuit point of view. When we are breaking the big RC network into small RC network, this will also use additional circuits for the switches which lead to increase in area overhead [1, 9].

2.2. TwoStep RC Ladder Network
In this section, the detailed analysis of twostep RC ladder network as shown in Figure 3 is conducted.
Table 2 gives the , , and expressions for considering each and every fault as well as for faultless case.

Figures 4, 5, and 6 show the values of magnitude of , , and , respectively, plotted against “.”
Table 3 shows the conclusion derived from Figures 4, 5, and 6 about faults distinguishability and required measuring parameters for the same.

Tables 4, 5, and 6 show the ambiguity for each fault which is not distinguishable from the measurement of , , and , respectively.



Table 7 gives the conclusion derived from Tables 4, 5, and 6.

Figure 7 shows the fault coverage for twostep RC ladder network. Fault coverage is 100% except for . All three measurements, that is, , , and , are required to calculate the distinguishability and ambiguity.
2.3. ThreeStep RC Ladder Network
In this section, detailed analysis of threestep RC ladder network (Figure 8) is done.
For faultless network, the expression for is given by
Table 8 gives the expression for considering each and every fault. Similar analysis is done for and and the results are shown below.

For a faultless network, the expression for and is given by
Table 9 gives the expressions for considering each and every fault and Table 10 gives the expressions for each and every fault.

