Network centrality based analog / mixed signal circuit measurement point set compression method

Abstract

The application discloses a kind of based on network centrality simulation / mixed signal circuit measuring point set compression method, and the node dictionary set and component dictionary set are extracted from the netlist file of analog / mixed signal circuit, the connection between node and the current direction relationship of extraction according to component dictionary set and node dictionary set establish the directed network graph of node, the centrality index of each measuring point is calculated, based on weighted Borda number method is comprehensively sorted to measuring point, and from measuring point sequence, the measuring point of early sorting is selected as alternative measuring point, and finally the measuring point set is further preferred based on criticality to alternative measuring point and obtained.The centrality of the application is combined with the sensitivity of defect to carry out measuring point screening, so that the measuring point set screened is more accurate and effective.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of analog / mixed signal circuit, more particularly, to a network centrality-based analog / mixed signal circuit test point set compression method. BACKGROUND

[0002] Integrated analog / mixed signal circuits have developed to a stage of higher and higher integration of functional blocks in the era of SoC (System on Chip), and the testing difficulty is also increased, which has become a prominent problem. Since the internal test points of the circuit cannot be observed after packaging, and the number of external accessible test points is limited, it is almost impossible to achieve complete defect coverage. Therefore, a certain number of excellent test points need to be found at the circuit design stage so as to carry out DFT (design for test) targetedly, so that the selected internal test points can still be observed after packaging. However, the introduction of additional DFT will cause area overhead, so the number of test points cannot be increased unlimitedly, but the combination of relatively excellent test points in the circuit should be found to meet the defect coverage while controlling the number of test points. Therefore, how to select test points is an important research direction in the field of testability design. At present, the most widely used method is full simulation, but due to the large amount of data and long simulation time, it is more suitable for small circuits. For large-scale analog / mixed signal circuits, the complexity of using the full simulation method is too high, and it does not meet the actual application requirements. SUMMARY

[0003] The present application aims to overcome the deficiencies of the prior art, and provides a network centrality-based analog / mixed signal circuit test point set compression method, which combines the centrality and sensitivity of test points to screen test points, so that the test point set obtained by screening is more accurate and effective.

[0004] In order to achieve the above-mentioned application purpose, the network centrality-based analog / mixed signal circuit test point set compression method of the present application comprises the following steps:

[0005] S1: Obtain the netlist file of the analog / mixed signal circuit, and extract the component dictionary set and the node dictionary set therefrom, wherein the key of the component dictionary set is the component name, and the value is an array for storing the node names connected to the component; the key of the node dictionary set is the node name, and the value is an array for storing the component names connected to the node;

[0006] According to actual needs, set M defects to be detected for the analog / mixed signal circuit, and then select a plurality of nodes from all nodes as candidate test points for detecting the M defects, to obtain a test point set T, and record the number of test points as N;

[0007] S2: Extract the connection relationship and current direction relationship between nodes according to the component dictionary set and the node dictionary set, and generate an adjacency matrix A with a size of QxQ, Q represents the number of nodes in the analog / mixed signal circuit, and the adjacency matrix generation method is: when the current direction between two adjacent nodes a and b is from node a to node b, the element corresponding to node a to node b in the adjacency matrix is 1, otherwise it is 0; then a directed network graph of the nodes is established according to the adjacency matrix;

[0008] S3: Calculate the centrality index of each measurement point i, i=1, 2, …, N, including eigenvector centrality, degree centrality, betweenness centrality, and closeness centrality, wherein:

[0009] The calculation method of the eigenvector centrality of measurement point i is to find the largest eigenvalue of the adjacency matrix A, and the D-dimensional eigenvector corresponding to the eigenvalue, the dth element in the eigenvector is the eigenvector centrality of node d, and the eigenvector centrality of measurement point i is denoted as C e (i).

[0010] The calculation formula of the degree centrality C d (i) of measurement point i is as follows:

[0011]

[0012] Wherein, degree(i) represents the sum of the in-degree and out-degree of measurement point i.

[0013] The calculation formula of the betweenness centrality C b (i) of measurement point i is as follows:

[0014]

[0015] Wherein, represents the number of shortest paths passing through measurement point i in the shortest path between nodes s and t, g st represents the number of shortest paths connecting nodes s and t.

[0016] The calculation formula of the closeness centrality C c (i) of measurement point i is as follows:

[0017]

[0018] Wherein, represents the average distance from measurement point i to each of the other nodes:

[0019]

[0020] Wherein, d ij represents the shortest path length from measurement point i to each of the other nodes j.

[0021] S4: Based on the weighted Borda number method, the N measuring points are comprehensively sorted according to the eigenvector centrality, degree centrality, betweenness centrality and closeness centrality of the N measuring points, and a measuring point sequence is obtained; the first K measuring points in the measuring point sequence are taken as candidate measuring points to form a candidate measuring point set T' of the analog / hybrid signal circuit T';

[0022] S5: The normal state and M defect states of the analog / hybrid signal circuit are simulated, and the response eigenvalue P k of each candidate measuring point k in the candidate measuring point set under the normal state and the response eigenvalue P k (m) of each candidate measuring point k in the candidate measuring point set under the defect m state are obtained, k = 1, 2, …, K, m = 1, 2, …, M;

[0023] The detection threshold ε and the sensitivity threshold β are set in advance according to the actual situation, ε < β, and the sensitivity α k (m) of the candidate measuring point k to the defect m is calculated by using the following formula:

[0024]

[0025] Then, the criticality C k of the candidate measuring point k is calculated by using the following formula:

[0026]

[0027] The K measuring points in the candidate measuring point set T' are arranged in descending order of criticality, and the first R candidate measuring points are selected to form a final preferred measuring point set, wherein the value of R is set according to actual needs.

[0028] The analog / hybrid signal circuit measuring point set compression method based on network centrality extracts a component dictionary set and a node dictionary set from a netlist file of an analog / hybrid signal circuit, extracts the connection relationship and current direction relationship between nodes to establish a directed network graph of nodes, calculates the centrality index of each measuring point, comprehensively sorts the measuring points based on the weighted Borda number method, and then selects the measuring points with high ranking from the measuring point sequence as candidate measuring points, and further optimizes the candidate measuring points based on criticality to obtain the final measuring point set.

[0029] The present application has the following advantages:

[0030] 1) The centrality of circuit nodes is used to preliminarily screen the measuring points, reduce the size of the candidate measuring point set, reduce the amount of calculation data, and reduce the complexity of measuring point optimization;

[0031] 2) The sensitivity of the measuring points to defects is used to further optimize the measuring points, which can make the screened measuring point set more accurate and effective. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a specific embodiment flow chart of the network centrality-based analog / mixed signal circuit measurement point set compression method of the present application;

[0033] Figure 2 is a circuit netlist file example diagram;

[0034] Figure 3 is a directed network model diagram of the circuit shown in Figure 2

[0035] Figure 4 is a circuit diagram of the Benchmark benchmark circuit in the present embodiment;

[0036] Figure 5 is a directed network diagram of the Benchmark benchmark circuit in the present embodiment;

[0037] Figure 6 is a node centrality ranking diagram in the present embodiment. DETAILED DESCRIPTION

[0038] The specific embodiments of the present application will be described below in conjunction with the accompanying drawings, so that those skilled in the art can better understand the present application. It should be particularly noted that in the following description, when detailed descriptions of known functions and designs may obscure the main content of the present application, these descriptions will be omitted here.

[0039] In order to better illustrate the technical solutions of the present application, the basic theory of the present application will be briefly described first.

[0040] EMBODIMENT

[0041] Figure 1 is a specific embodiment flow chart of the network centrality-based analog / mixed signal circuit measurement point set compression method of the present application. As shown in Figure 1 the specific steps of the network centrality-based analog / mixed signal circuit measurement point set compression method of the present application include:

[0042] S101: Obtain circuit information:

[0043] Because the wiring in the schematic diagram of a large analog / mixed signal circuit is very numerous, and many of them are marked by node names, it is very difficult to confirm and find the connection relationship of components in the circuit diagram by the naked eye, the present application needs to use the netlist file to extract the circuit connection information. The circuit netlist is usually in Spice format or Spectre format, and the netlist file contains all the component information and related node information. Figure 2 is a circuit netlist file example diagram. As shown in Figure 2 ​The circuit shown is a class B push-pull amplifier circuit schematic diagram, in the corresponding netlist file, the r at the beginning of R3 represents a resistor, and the node names at both ends of the resistor R3 are output and 0, and the signal flow direction is from output to 0. These are the information needed for subsequent construction of a directed network model.

[0044] Therefore, in the present application, first, the netlist file of the analog / mixed signal circuit is obtained, and the component dictionary set and the node dictionary set are extracted therefrom, wherein the key of the component dictionary set is the component name, and the value is an array for storing the node name connected to the component. The key of the node dictionary set is the node name, and the value is an array for storing the component name connected to the node.

[0045] According to the actual needs, M defects to be detected are set for the analog / mixed signal circuit, and then a plurality of nodes are selected from all nodes as candidate test points for detecting the M defects, to obtain a test point set T, and the number of test points is N.

[0046] S102: Establish a directed network model:

[0047] According to the component dictionary set and the node dictionary set, the connection relationship and the current direction relationship between the nodes are extracted, and a QxQ adjacency matrix A is generated, Q represents the number of nodes in the analog / mixed signal circuit, and the adjacency matrix generation method is: when the current direction between two adjacent nodes a and b is from node a to node b, the element corresponding to node a to node b in the adjacency matrix is 1, otherwise it is 0. Then a directed network graph of the nodes is established according to the adjacency matrix.

[0048] Figure 3 is Figure 2 The directed network model diagram of the circuit shown. As Figure 3 shown, Figure 2 There are a total of 8 nodes in the circuit shown, and the directed edges between the nodes are determined by the current direction.

[0049] S103: Calculate the centrality of the test point:

[0050] In the network model, centrality is used to measure the importance of each node in the network. For circuit defect detection, excellent test points are also important nodes. Highly central nodes may be excellent test points because they have greater influence in the network and can better cover other parts of the network. Therefore, the centrality of the nodes in the analog / mixed signal circuit is used to screen the optimal test points in the present application. The centrality indicators used in the present application include eigenvector centrality, degree centrality, betweenness centrality, and closeness centrality, so the centrality indicators of each test point i, i = 1, 2, …, N, are calculated, and the calculation methods are as follows:

[0051] • Eigenvector centrality

[0052] Eigenvector centrality is a measure of node centrality based on the node's connection pattern in the network and the centrality of its surrounding nodes. Eigenvector centrality assumes that the centrality of a node is influenced by the centrality of the nodes connected to it, regardless of the degree of the node. The calculation of eigenvector centrality relies on the concept of matrix eigenvalues and eigenvectors. The calculation method of eigenvector centrality of measurement point i in the invention is as follows:

[0053] Find the largest eigenvalue of the adjacency matrix A and the D-dimensional eigenvector corresponding to the eigenvalue. The d-th element of the eigenvector is the eigenvector centrality of node d. Denote the eigenvector centrality of measurement point i as C e (i).

[0054] • Degree centrality

[0055] The definition of degree centrality is: the proportion of the degree of a node to the total degree of all nodes. In an undirected graph, the degree of a node represents the number of edges directly connected to other nodes. In a directed graph, the degree of a node is divided into two indicators: in-degree and out-degree, representing the number of edges pointing to the node and the number of edges from the node, respectively. The calculation formula of the degree centrality C d (i) of measurement point i in the invention is as follows:

[0056]

[0057] Where degree(i) represents the sum of the in-degree and out-degree of measurement point i.

[0058] • Betweenness centrality

[0059] The definition of betweenness centrality is: in all shortest paths between pairs of nodes in the network, the proportion of the number of shortest paths passing through the node to the total number of shortest paths. The calculation formula of the betweenness centrality C b (i) of measurement point i in the invention is as follows:

[0060]

[0061] Where, represents the number of shortest paths passing through measurement point i between nodes s and t, g st represents the number of shortest paths connecting nodes s and t.

[0062] Betweenness centrality measures the importance of a node in the network by calculating the shortest paths between each pair of nodes. The higher the betweenness centrality, the more frequently the node appears in the paths connecting other nodes in the network, indicating that the node has a higher intermediary role in the network.

[0063] · closeness centrality

[0064] Closeness centrality represents the inverse of the average shortest path length from a node to all other nodes. In a network, the shortest path length from a starting node to a target node refers to the minimum number of edges required to pass from the starting node to the target node. If there is no path from the starting node to the target node, it is said that the starting node and the target node are not connected.

[0065] Closeness centrality of the measurement point i in the present application c The calculation formula of (i) is as follows:

[0066]

[0067] wherein, represents the average distance from measurement point i to each of the remaining nodes:

[0068]

[0069] wherein, d ij represents the shortest path length from measurement point i to each of the other nodes j.

[0070] A node with higher closeness centrality generally indicates that it is closer to other nodes in the network and can quickly transmit information to other nodes, thus having an important influence in the network. It should be noted that the calculation of closeness centrality depends on the connectivity of the network, and if the network is not connected, the closeness centrality of some nodes may not be calculated. In practical applications, closeness centrality is usually used together with other indicators to evaluate the importance of nodes in the network.

[0071] S104: Screening to obtain a candidate measurement point set:

[0072] According to the eigenvector centrality, degree centrality, betweenness centrality and closeness centrality of the N measurement points, the N measurement points are comprehensively sorted based on the weighted Borda number method to obtain a measurement point sequence. The first K measurement points in the measurement point sequence are taken as candidate measurement points to form a candidate measurement point set T' of the analog / mixed signal circuit.

[0073] It should be noted that too many candidate measurement points will result in poor compression effect of the measurement point set, and too few candidate measurement points will result in low accuracy of measurement point screening. A reasonable trade-off is needed, and the optimal value range of the number K of candidate measurement points is [0.2N, 0.4N].

[0074] S105: Measurement point re-selection based on sensitivity:

[0075] In evaluating the importance of the test points, the number of defects that can be detected is usually used as an indicator. However, in calculating the number of defects that can be detected by a test point, a detection threshold needs to be set. That is, only when the difference between the defect response and the non-defect response exceeds a certain threshold, the defect can be considered to be detectable. This threshold is usually set artificially according to different circuit scenarios, which is highly subjective, and thus the number of defects that can be detected lacks persuasiveness. To solve this problem, the present application calculates the number of defects that can be detected based on a sensitivity factor, thereby evaluating the criticality of the test points, and reselecting the candidate test points, the specific method being as follows:

[0076] simulate the normal state and M defect states of the analog / mixed signal circuit, and obtain the response eigenvalue P k of each candidate test point k in the normal state and the response eigenvalue P k (m) of each candidate test point k in the defect m state, k = 1, 2, …, K, m = 1, 2, …, M.

[0077] According to the actual situation, set the detection threshold ε and the sensitivity threshold β in advance, ε < β, when the response of a defect at a candidate test point deviates from the response in the normal state by more than or equal to the set detection threshold ε, the defect is considered to be detectable at this test point, and the greater the deviation within the sensitivity threshold β, the higher the sensitivity of the test point. When the deviation is exactly equal to the set detection threshold ε, the sensitivity is the lowest, but it is still considered to be detectable, and when the deviation is less than the detection threshold ε, the defect is considered to be undetectable at this test point, and there is no sensitivity. Based on the above analysis, the calculation formula of the sensitivity α k (m) of the candidate test point k to the defect m in the present application is as follows:

[0078]

[0079] It can be seen that the existence of the sensitivity threshold β ensures that when the deviation value is large enough, further increase in the deviation will not increase the sensitivity of the test point to the defect. The size of the sensitivity α k (m) can reflect the credibility and ease of detection of the candidate test point k to the defect m, and the greater the α k (m) represents the stronger the feature detection capability of the candidate test point to the defect m. The existence of the sensitivity allows the test points in the circuit to be rearranged according to their criticality size values, thereby weakening the influence of the subjectivity and blindness of the threshold setting.

[0080] Then, the criticality C k of the candidate test point k is calculated by the following formula:

[0081]

[0082] The K measurement points in the set of alternative measurement points T' are arranged in descending order of criticality, and the first R alternative measurement points form the final set of preferred measurement points, wherein the value of R is set according to actual needs.

[0083] To better illustrate the technical effects of the present application, a specific example is used for experimental verification. Benchmark benchmark circuit refers to a set of standard circuits specially designed for evaluating the performance of circuit design tools and the effect of circuit optimization algorithms. Benchmark circuits are widely used in the field of circuit design and testing to evaluate the accuracy, speed and reliability of circuit design tools or algorithms. The advantage of Benchmark circuit is that it can provide a standardized testing method for circuit design tools and algorithms, allowing different tools and algorithms to be compared and competed with each other, thereby promoting the development of circuit design technology. At the same time, Benchmark circuit also provides a standard test platform for circuit design engineers, which can be used to evaluate the performance of different circuit design tools and algorithms in actual design, so as to select the optimal tool and algorithm. In this paper, the Bandgap Voltage Reference Circuit in the Benchmark benchmark circuit will be used as the verification circuit. Figure 4 is the circuit diagram of the Benchmark benchmark circuit in this embodiment.

[0084] The directed network graph of the nodes is generated according to the component dictionary set and the node dictionary set of the Benchmark benchmark circuit. Figure 5 is the directed network graph of the Benchmark benchmark circuit in this embodiment. As shown in Figure 5 , the Benchmark benchmark circuit in this embodiment includes 77 nodes. Assuming that all these 77 nodes are used as measurement points, the degree centrality, betweenness centrality, closeness centrality and eigenvector centrality of each measurement point are calculated based on the directed network graph. Table 1 is a centrality index data table of some measurement points in this embodiment.

[0085]

[0086]

[0087] Table 1

[0088] The calculation and sorting of the four centrality indexes are performed for the 77 nodes, and the 30% critical point value is marked in each index. Figure 6 is the node centrality sorting graph in this embodiment. As shown in Figure 6 , the straight line is the 30% critical point value. As can be seen, not all alternative measurement points have four centrality indexes in the top 30%, so it is not accurate to select measurement points according to a single index.

[0089] In the present application, the weighted Borda number method is used to comprehensively rank 77 nodes, and the weights of eigenvector centrality, degree centrality, betweenness centrality and closeness centrality are 0.1, 0.2, 0.3 and 0.4 respectively. Eigenvector centrality is a centrality index based on the global structure of nodes in the graph. In contrast, degree centrality and betweenness centrality pay more attention to the local connection of nodes, and measure the degree of nodes in the network and the degree of intermediation of nodes to other nodes in the network respectively. Closeness centrality focuses on the distance from a node to other nodes, and measures the accessibility of a node in the network. Therefore, eigenvector centrality is more suitable for evaluating the importance of nodes in the global structure, while degree centrality, betweenness centrality and closeness centrality are more suitable for evaluating the importance of nodes in the local structure. In the comprehensive ranking, although the importance of global and local structure needs to be considered, the local centrality index is more suitable for emphasis and highlighting, so the weight value is higher. Among them, the weight of degree centrality is set to 0.2, which is the lowest of the three, because the degree distribution of nodes in the network model of Bandgap circuit is relatively uniform, and degree centrality cannot highlight the key nodes. In this embodiment, the top 30% of measurement points are selected as candidate measurement points. Table 2 is a data table of the top 30% of measurement points in the centrality index ranking in this embodiment.

[0090]

[0091]

[0092] Table 2

[0093] The present application finds that the relationship between the centrality of the measurement point and the criticality is not completely consistent, that is, the nodes with high centrality are not necessarily key nodes, but the nodes with high centrality have a higher probability of becoming key nodes. Therefore, the size of the possible measurement point set can be reduced by the centrality of the nodes, thereby reducing the workload of simulation and data calculation in testability design.

[0094] Next, the normal state and 120 defect states of the Benchmark benchmark circuit are simulated, the sensitivity of each candidate measurement point to different defects is calculated according to the response eigenvalue (the detection threshold ε is set to 0, and the sensitivity threshold β is set to 0.0105), then the sensitivity is summed to obtain the criticality, and then the 23 candidate measurement points are ranked according to the criticality from large to small, and the top 10 measurement points are selected to form the final measurement point set. Table 3 is a criticality data table of the top 10 criticality in this embodiment.

[0095]

[0096] Table 3

[0097] While the foregoing specific embodiments of the application have been described in some detail to provide a clear understanding thereof, it will be apparent to those of ordinary skill in the art that numerous modifications can be made to the specific embodiments described without departing from the spirit and scope of the application defined by the appended claims.

Claims

1. A method for compressing measurement point sets in analog / mixed-signal circuits based on network centrality, characterized in that, Includes the following steps: S1: Obtain the netlist file of the analog / mixed signal circuit, and extract the component dictionary set and node dictionary set from it. The key of the component dictionary set is the component name, and the value is an array used to store the node names connected to the component. The key of the node dictionary set is the node name, and the value is an array used to store the component names connected to the node. Configure analog / mixed signal circuits according to actual needs. There is a defect to be detected, and then several nodes are selected from all nodes as those used for detection. The set of test points is obtained by selecting test points for each defect. The number of measurement points is recorded as follows: ; S2: Extract the connection relationships and current direction relationships between nodes based on the component dictionary and node dictionary, and generate a dataset of size [size missing]. adjacency matrix , The adjacency matrix represents the number of nodes in an analog / mixed-signal circuit. The method for generating the adjacency matrix is ​​as follows: when two adjacent nodes... , The direction of the current between nodes is from node Flow to node Then the nodes in the adjacency matrix To the node The corresponding element is 1, otherwise it is 0; then a directed network graph of nodes is built based on the adjacency matrix; S3: Calculate for each measuring point The central indicator, , These include eigenvector centrality, degree centrality, betweenness centrality, and proximity centrality, among which: measuring point The method for calculating the eigenvector centrality is as follows: obtain the adjacency matrix. The largest eigenvalue, and the corresponding eigenvalue. 3D eigenvector, the eigenvector of the eigenvector is the 3rd eigenvector. Each element is a node. Eigenvector centrality, denoted by measurement point eigenvector centrality is ; measuring point Degree centrality The calculation formula is as follows: ， in, Indicates the measuring point The sum of in-degree and out-degree; measuring point betweenness centrality The calculation formula is as follows: ， in, Represents a node and The shortest path between them passes through the measurement point The number of shortest paths, Indicates the connection node and The number of shortest paths; measuring point proximity centrality The calculation formula is as follows: ， in, Indicates the measuring point Average distance to all other nodes: ， in, Indicates the measuring point To other nodes The shortest path length; S4: According to The eigenvector centrality, degree centrality, betweenness centrality, and proximity centrality of each measurement point are analyzed based on the weighted Borda number method. Sort the measurement points in descending order of their Borda numbers to obtain the measurement point sequence; then, in the measurement point sequence, sort the first... These measurement points serve as candidate measurement points, forming a candidate measurement point set for analog / mixed-signal circuits. Number of alternative measuring points The range of values ​​is ; S5: Normal state of analog / mixed signal circuits and Simulations were performed on each defect state to obtain a set of candidate test points. Each of the alternative measurement points Response characteristics under normal conditions and in defects response eigenvalues ​​under state , , ; Pre-set the detection threshold according to the actual situation. and sensitivity threshold , The candidate measuring points are calculated using the following formula. defects sensitivity : ， Then, the candidate measurement points are calculated using the following formula. criticality : ， Set of candidate measurement points middle The measurement points are arranged from most critical to least critical, and the first ones are selected. The candidate measurement points constitute the final set of preferred measurement points, among which The value should be set according to actual needs.

Images