JP3168988B2 - Fault location estimation method for sequential circuit, candidate extraction in failure location estimation and weighting method thereof, and apparatus therefor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for estimating a fault location of a sequential circuit, a method for extracting candidates in the fault location estimation and a method for weighting the candidates, and a device therefor. The present invention also relates to a method for extracting a candidate for estimating a fault location and a method for weighting the candidate, and a device therefor.

[0002]

2. Description of the Related Art Conventional sequential circuit failure diagnosis methods include:
There are a method of creating a failure dictionary in advance and collating it with an actual fail result to narrow down a failure location, and a method of estimating a failure location by setting / reading out the state of a flip-flop by a scan path.

A method using a failure dictionary is disclosed in
As disclosed in Japanese Patent No. 20485, etc., a simulation is performed in which a fault is inserted into an LSI using an actual test vector, and a data file as a fault dictionary in which an assumed fault location is associated with information on an output pin that fails at that time. Was created in advance, and from the failure state of the output pin that actually failed, the failure dictionary was indexed and the candidate points of the failure location were obtained, and the candidate points obtained from a plurality of candidates were obtained from the fail output of all vectors. The failure location is estimated by assigning priorities in order from the one that is most likely to use the failure estimation location.

A method using a scan path is disclosed in
As shown in JP-A-194416 and the like, it is necessary to prepare in advance a check circuit capable of reading / writing the state in the circuit, and use the test circuit for this check to check the state of the flip-flop. After the setting is performed, the state of the flip-flop and the like is read out using the test circuit similarly after the circuit operation based on the state. This is a method of comparing the read state with an expected value to determine whether or not propagation has occurred from before, and narrowing down a failure portion sequentially.

[0005] Further, the fault location estimating method of the sequential circuit proposed by the applicant of the present invention in JP-A-8-146093 or JP-A-8-94714 assumes that the propagation of a fault is similar to the present invention. This is a technique of sequentially extracting assumed combinational circuits, assuming a fault propagation in the combinational circuit, and obtaining an input state value that satisfies the output state, thereby performing only the combinational circuit input terminal state estimation.

[0006]

The above-described conventional fault diagnosis method using a fault dictionary is a method of searching an event matching an actual fault output from the fault dictionary and narrowing down a fault location. The problem is that it is necessary to prepare a fault dictionary for the fault output by fault simulation,
This means that the operation time for creating the failure dictionary increases. The reason is that a failure assumption is made for all expected nodes, so that the number of failure simulations increases, and it takes time to create a search dictionary of the failure assumption part and the failure output terminal. The increase in the calculation time required for creating the dictionary becomes remarkable as the size of the LSI increases.

A second problem is that a single stuck-at fault is generally used as a fault model handled in a fault simulation, and thus a multiple fault such as a short fault may not coincide with an actual fault. The reason is that if the fault simulation is extended to multiple faults, it is necessary to make fault assumptions in multiple combinations. Even if a double fault is assumed, the number of combinations increases explosively, and the processing time of the fault simulation increases. Because it is not realistic.

As for the method using a scan path, the first problem with this method is that it is necessary to incorporate a flip-flop having a check circuit for enabling setting / reading of an internal state, that is, a scan path at the time of circuit design. This is not applicable to an LSI that does not use a campus.

A second problem is that, in the case of partial scan, a fault estimating method as a sequential circuit is required instead of a fault estimating method as a combinational circuit.
The reason is that in the case of the partial scan, the circuit between the scan flip-flops is not always a combination circuit. Further, if the failure estimation method used at that time is the failure dictionary method, there is the aforementioned problem.

The problem with the method of estimating the fault location of a sequential circuit disclosed in Japanese Patent Application Laid-Open Nos. 8-146093 and 8-94714 is that weighting is performed on the frequency of occurrence assuming a single stuck-at fault. Therefore, the estimation accuracy may be reduced. The reason is that the weight of the node near the output terminal increases even when the state at the time of an open (release) fault or a short (short) fault changes with time or in the case of a fixed fault.

Further, since all the vectors obtained by the estimation are registered in the estimated value table for the fault propagation state estimated value of the combination circuit input terminal and the combination circuit is sequentially traced, the extracted combination circuit is extremely difficult. LSI to estimate
There is a disadvantage that the calculation time becomes very large as the scale becomes large.

[0012] It is an object of the present invention to provide a method for determining a failure in an LSI by using actual tester pass / failure information, circuit connection information, and LSI
An object of the present invention is to provide a fault estimating method for a sequential circuit which can perform an estimation process using expected value information for all test vectors of all internal flip-flops.

Another object of the present invention is to provide a method and an apparatus for improving the accuracy of a fault estimation candidate for a single stuck-at fault in LSI fault location estimation.

Still another object of the present invention is to provide a method and an apparatus for improving the accuracy of an open fault fault estimation candidate in LSI fault location estimation.

It is another object of the present invention to provide a method and an apparatus for improving the accuracy of a fault estimation candidate for a double fault such as a short fault in estimating a fault location of an LSI.

[0016]

According to the present invention, there is provided a method for diagnosing faults in a sequential circuit, comprising: all expected value information for all vectors of all flip-flops in an LSI prepared in advance; pass / fail output information for an actual tester; Using the connection information of the circuit,
Based on the fact that a fault is estimated by dividing the circuit into a flip-flop and a combinational circuit, an actual fail output pin or a flip-flop estimated as a failure is determined as a sequence for obtaining a failure propagation estimation value at an input boundary of the combinational circuit for each fail vector. The procedure for extracting the combinational circuit including the input line and the estimation of the fault propagation value at the input terminal of the combinational circuit assume that a fault exists in the combinational circuit. The estimated state value or estimated fault propagation path of the location and the combinational circuit input terminal is determined.If it is assumed that no fault exists, only the estimated state value or the estimated failure propagation path of the combinational circuit input terminal is determined. If a fault must be present, the procedure to find only the estimated fault location and the estimated A procedure for determining the number of failure points that can be assumed for each type of failure among the failure points, and reducing the estimated state obtained based on the number of failure points, and the estimated state value of the combined circuit input terminal obtained by the reduction After extracting the fault propagation terminal from the above, several path selection procedures for simplifying the Boolean algebra and effectively summarizing the fault propagation paths, and repeating the above procedure, estimating with all the vectors, the above obtained Using all estimated fault locations and estimated fault propagation paths, classifying each time and determining locational coincidence at all times, or taking the frequency of nodes obtained as estimated fault locations and estimated fault propagation paths, and taking priority A procedure for performing the ranking, and for each of the obtained fault-predicted places with priorities, examining an individual estimation result and classifying each failure type,
In order to further improve the estimation accuracy, the method further comprises a step of performing a failure simulation again on all the vectors based on the obtained estimated fault location, its type, and the state value, and performing a coincidence determination.

In addition, when applying the fault estimation in an actual LSI, in addition to the above procedure, the expected value of the clock line of the flip-flop is checked for each vector, and it is determined whether or not the clock has entered at that time and the data has been updated. If updated, the estimated state value is sequentially propagated to the output terminal of the preceding stage according to the above procedure, and if data is not updated, the data update is held until the last updated time before that, It is determined whether or not the clock line and the scan path line include a fault. If the fault is not included, only the normal data line of the flip-flop is extracted and estimated as a combinational circuit. A procedure for estimating the scan path line and estimating the entire failure location together with the estimation of the data line, LSI input terminal, ROM not only the flip-flop output, also assumes the output terminal of RAM, ROM, in the case of RAM write bit fault propagation, make assumptions to determine the supply address fault propagation, respectively ROM,
A procedure of going back to the previous stage from the input terminal of the RAM, and creating a final failure estimation list.

The method for extracting candidates and estimating the weight in the method for estimating the failure location of a sequential circuit according to the present invention is described below. A procedure for dynamically extracting a combinational circuit sequentially from each fault output terminal of each vector obtained as a test result using the connection information of all the circuits, and an output state value or estimated value of the extracted combinational circuit. Estimating the state value of each flip-flop in the LSI by estimating the input state value of the combinational circuit from, and extracting the terminal, element and its state value that are estimated to have propagated a fault from the difference from the expected value And a test vector from the fault propagation estimation terminal obtained in this way,
The procedure of repeating the above operation sequentially, such as extracting a combinational circuit in the input direction, and the procedure of performing the same operation on all test vectors to obtain a fault propagation estimation element and its state value for each extracted combinational circuit Candidate extraction and its weighting method in fault location estimation including:
A procedure for retrieving circuit connection information and reconstructing a connection relation as a fault propagation path to an output terminal using the fault propagation estimation elements and state values in each of the combinational circuits; Check the state value of the point, and (1) always select a node having the same state value as an estimated failure candidate point; (2) select a node whose estimated state value changes with time as an estimated failure candidate point; ( 3) Simultaneously select nodes that require two estimated fault locations as estimated fault candidate points; include a procedure and count the number of fault output terminals of the LSI that arrive at each of the selected estimated fault candidate points as a frequency And a procedure of rearranging the order of appearance of the failure candidate points in descending order of the counted frequency number and performing weighting.

Further, the candidate extraction and weighting apparatus for estimating a fault location in a sequential circuit according to the present invention stores a netlist of an LSI to be estimated and registers the netlist divided into partial circuits. Possible,
A netlist management unit that processes search and deletion requests and performs unified management of a netlist; an input / output terminal / flip-flop expected value management unit that manages expected values of input / output terminals and flip-flops; An estimated value / actual value management unit storing actual measured values, a failure candidate management unit storing estimated results such as a failure candidate point, an estimated failure state value, and a failure mode obtained from each failure location estimation function; A combination circuit extraction unit for dynamically extracting a combinational circuit from an output terminal or a failure estimation output terminal, a combinational circuit state estimation unit for estimating a state value inside the combinational circuit, and an estimated failure propagation path from a state estimation result inside the combinational circuit. A fault propagation path extracting unit to be extracted, and a path selection for selecting a fault propagation path using the obtained fault propagation terminal and state value of the combinational circuit input terminal And a fault propagation path reconfiguration unit for reconstructing the obtained fault propagation path and associating a fault location with a fault output terminal, and examining the obtained estimated state of each fault candidate for all vectors. Estimated state value determination unit to determine whether the state value is constant or changing,
A failure mode classification unit that performs failure mode classification in consideration of the number of failure points based on the results obtained in the state value determination, and counts the number of failure output terminals or the number of failure propagation paths for each failure candidate point. A frequency weighting unit that weights each failure mode, and a sequence control unit that controls an overall estimation sequence such as control of the processing order of the failure estimation functions and queuing control between the failure estimation functions. And

The operation of the present invention will be described. The above fault estimation procedure estimates a fault estimation location and a fault propagation value by assuming the inside for each combinational circuit divided for each detected fail vector, and further estimates the failure propagation estimate with the output of the preceding combinational circuit. The fault propagation of the assumed input part is estimated. In this way, the combinational circuits are sequentially extracted from the output side of the LSI, the estimation is performed independently for each combinational circuit, and the fault propagation value at the input boundary of the combinational circuit is estimated for all the failures. Even if a state propagates through a plurality of combinational circuits and is fed back to a combinational circuit having a propagation failure point, state estimation can be performed without causing an estimation error.

Further, since the fault propagation path can be efficiently selected after estimating the input state of the combinational circuit input terminal, the estimation is not repeated for every fault propagation estimation state as in the prior art.
There is an effect that the amount of calculation for failure estimation can be greatly reduced, and the time can be significantly reduced. At the same time, real-time processing can be performed on a large-scale circuit. Further, multiple faults such as short faults and timing faults are assumed for each type of fault, and estimation is performed.

By adding the above procedure to a method for estimating a clock line, a scan path line, and a memory block of a flip-flop, a fault can be estimated for an actual large-scale LSI and can be performed efficiently.

In any case, a single fault, an open fault, and a short fault have a procedure for reconstructing a connection relation as a fault propagation path from a fault propagation estimating element in each combinational circuit to an output terminal. A failure candidate point can be selected. This is because an actual fault is always on the estimated propagation path since the fault propagation path is sequentially estimated from the failure output terminal of the LSI for each combinational circuit and goes back toward the input direction.

Conversely, a failure cannot be present in an element that is not extracted as a failure propagation path. This is a single failure,
Not only for open failures, but also for short failures
Since the path from the ort faulted portion to the LSI fault output terminal is assumed to be fault-transmitted through multiple paths at the same time, a true fault propagation path is included in the estimated fault propagation paths. Because it is included.

There are two types of estimated propagation paths, one that reaches from the output terminal of the combinational circuit to the input terminal of the combinational circuit and the other that does not reach and stops in the combinational circuit. . The number of fault candidates in the method of using nodes on the propagation path as fault candidates depends on the logical depth of the LSI itself, that is, the number of propagation elements from the LSI output terminal to the input terminal, and the number of fault candidates directly depends on the scale of the entire LSI. However, there is an advantage that the explosion of operations can be suppressed even when the scale is increased.

In addition, single failure, open failure, short failure
Since the number of paths existing between each estimated fault candidate point and the failure output terminal of the LSI or the number of failure output terminals of the arriving LSI is counted as a frequency regardless of the failure, there is a procedure for weighting according to the frequency. In addition, the accuracy of weighting the failure candidate points is improved as compared with the related art. Because the number of paths or the failure output terminal of the arriving LSI is counted, LS
Even if the logical depth from the I output terminal is deep, a node that satisfies more failure propagation paths or failure output terminals is extracted as a node with high frequency, and may be higher than a node on a propagation path close to the LSI output terminal. Is large.

[0027]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a flowchart showing one embodiment of the present invention. The fault diagnosis method for a sequential circuit according to the present invention uses all expected value information for all vectors of all flip-flops in an LSI prepared in advance, pass / fail output information of an actual tester, connection information of all circuits, and It is based on the assumption that a fault is estimated by dividing the fault into a loop and a combinational circuit.

Here, considering the circuit configuration of the entire LSI, it is a sequential circuit including a circuit having a latch function of temporarily storing data, such as a flip-flop or a memory.
A circuit portion that does not include an element for storing data, such as a flip-flop or a memory, is usually defined as a combinational circuit.
The sequential circuit includes a flip-flop and the like, and the input state given is the same, but the output state differs depending on how the clock is given. Since the combinational circuit does not include a flip-flop or the like, the output state is uniquely determined when the input state to be given is determined.

First, in procedure 1 for extracting a combinational circuit, an LSI
A circuit trace (or circuit extraction) from the fault output of or from the output terminal that has been presumed to be faulty to the input direction
When the signal reaches the input terminal of the LSI or the output terminal of the flip-flop, the circuit trace (or circuit extraction) ends. Next, a circuit trace (or circuit extraction) is performed in the output direction using the input terminal of the LSI or the output terminal of the flip-flop obtained at that time, and the output terminal of the LSI or the input terminal of the flip-flop is obtained.
Further, circuit extraction is performed in the input direction from the obtained output terminal or the input terminal of the flip-flop until reaching the input terminal of the LSI or the output terminal of the flip-flop to obtain a combinational circuit.

Here, the circuit trace refers only to tracing the connection relationship without extracting a circuit, and the circuit extraction refers to extracting a traced portion while performing a circuit trace to form a partial circuit. Is represented. Therefore, the circuit trace and the circuit extraction are traced (or extracted) until reaching the input / output terminal of the LSI or the input / output terminal of the flip-flop.

The state estimation of the combinational circuit obtained in this way is based on the combinational circuit input terminal state estimations # 1 to # 3 (procedures 2 to 3).
4), and the required conditions are different. Combinational circuit input terminal state estimation # 1 (procedure 2) is based on the assumption that no failure exists in combinational circuit 100 and a failure signal is propagated from the preceding combinational circuit as shown in FIG. This is an input terminal estimation procedure. Note that 101 is an input latch, and 102 is an output latch.

Combination circuit input terminal state estimation # 2, # 3
(Procedures 3 and 4) are based on the assumption that a fault exists in the combinational circuit, and are further divided into two according to the estimated number of faults. One is a combinational circuit input terminal state estimation # 2, which is a procedure for obtaining an estimated failure location and a combined circuit input state estimated value when it is assumed that one failure exists in the combinational circuit as shown in FIG. It is. The other is the combination circuit input terminal state estimation # 3 (procedure 4), which assumes the presence of two failures in the combination circuit and the combination circuit input terminal as shown in FIG. 2C. This is a procedure for obtaining a state estimation value.

Steps 2, 3, and 4 are performed using search trees or BDDs.
(Binary Dicision Diagram), Transitive Closure Alg
This is a procedure for obtaining an input state value that satisfies the combination circuit output state such as orithm. Taking a search tree as an example, an input state value is obtained from an output state value by repeating assumption and implication operations. In other words, starting from the fault output terminal, assuming a state that satisfies the fault state value for each input / output terminal of the gate,
The operation of determining the state values of the related gates and paths based on the assumption is repeated, and the state values are sequentially determined in the input direction so as not to cause inconsistency as a whole, thereby obtaining the combination circuit input state.

FIGS. 2A to 2C show the differences between the conditions obtained in steps 2, 3 and 4. In the procedure 2, in the stage where the assumption and the implication operation are repeated, only when all the paths can reach the combinational circuit input terminal without any inconsistency can be extracted and the fault propagation path in the combinational circuit is obtained, and even one path causes inconsistency. Therefore, the state value that cannot be reached to the input terminal because it cannot be assumed is discarded.

In the procedure 3, the case where only one path that cannot reach the input terminal due to inconsistency at the stage of making the assumption and cannot reach the input terminal is extracted. (From x in FIG. 2 (b) to merge with another route). Discard otherwise. In procedure 4, the case where only two paths cannot reach the input terminal due to inconsistency due to inconsistency at the stage of making the assumption is extracted, and the fault propagation path in the combinational circuit and the two paths (the assumption is no longer possible) 2 (c) until it joins another route). Discard otherwise.

In each of the above procedures 2, 3 and 4, when no route is obtained as a result, the processing shifts to the extraction of all the failure candidate points in the procedure 7. Therefore, if written in a more understandable manner, the flow becomes as shown in FIG. In this case, new fault candidate point registration in step 6 is performed for all routes obtained in steps 2 to 4. If no route is obtained, new fault candidate point registration cannot be performed. The process shifts to all failure candidate point extraction determination.

If the path is found, the fault has propagated from the input side of the combinational circuit, so that it can be determined that the extraction of all candidate points has not been completed. Then, the combination circuit extraction of the preceding stage connected through the flip-flop is performed (st
art). If the path is not found, the path is traced. If the estimation has not been completed for all the failure outputs of the LSI, the combination circuit extraction is newly performed from the uncompleted LSI failure output terminal. Do.

As a special example in each of the procedures 2 to 4 of the combination circuit input terminal state estimation # 1 to # 3, an estimated value of the combination circuit input terminal state cannot be obtained, and only one or two failure estimation positions are obtained. May be In this case, there is no input state that satisfies the combination circuit output terminal state, indicating that no fault has propagated from the preceding combination circuit.

The Boolean algebra is simplified in the simplification procedure 5 using the input estimation vector obtained in the combinational circuit input terminal state estimation # 1 to reduce the number of estimated vectors. In this simplification procedure, as shown in FIG. 4, first, an input state estimation value that satisfies the combinational circuit output is checked to determine whether the terminal state value is 0 or 1, whichever value, that is, a fault at the combinational circuit output. And terminals and status values that are not directly related to are extracted.

In FIG. 4, each vector surrounded by a broken line has the same state value except for a terminal surrounded by a broken line. In this way, terminals having different values are extracted, and if all the four status values of 00/1, 01/11 and 00/01/11 are included for one terminal, those terminals are included. Replace the status value with "X". In FIG. 3, a portion surrounded by a broken line is a portion that becomes “X” by simplification, and indicates that four input estimation vectors are combined into one vector including one “X”.

More specifically, the right four vectors in FIG. 4 (0100101001, 0100100001, 0100101101, 0100100101)
Have the same state values except for the terminal surrounded by the broken line.
As described above, terminals having different values are extracted.
In the case of 1 and 2 terminals, if all four status values of 00/01/10/11 are included, those terminal status values are replaced with “X”. Therefore, in the case of the four vectors, simplification results in (010010XX01).

Similarly, the four left vectors (0100100001,0100
100000,0100100010,0100100011)
01001000XX. In FIG. 4, a portion surrounded by a broken line is a portion which becomes “X” by simplification, and indicates that four input estimation vectors are combined into a vector including one “X”.

As shown in FIG. 4, the simplification procedure is as follows. First, an input state estimated value that satisfies the combinational circuit output is examined to determine whether the terminal state value is either 0 or 1, that is, the combinational circuit output. The terminal and the state value which are not directly involved in the failure are extracted. The part replaced by "X" is that part, and the other common parts indicate that the state is determined to be 0 or 1. This simplification operation can greatly reduce the number of input estimation vectors. Also,
The estimated fault location obtained by the combinational circuit input terminal estimation # 2, # 3 is registered in the procedure 6 of the newly extracted fault candidate point registration. In step 7, it is determined whether or not the fault candidate point extraction has been performed for all the vectors, and as described in the special example, no more estimated fault propagation input terminals from the preceding stage can be obtained, and Steps 1 to 6, 8 for the test vector
If the estimation has been completed, the process proceeds to the failure candidate point location coincidence determination in step 9.

If the estimated fault propagation input terminal from the preceding stage cannot be obtained, but the estimation has not been completed for all the test vectors, the procedure from step 1 is performed for the test vector for which the estimation has not been performed. Repeat the operation.

If the estimated fault propagation terminal from the previous stage is obtained, the estimated input vector is reduced by the procedure of the combination circuit input estimated input vector simplification procedure of step 8, and the fault propagation terminal from the previous stage is extracted. Then, the procedure goes back to the preceding combination circuit in the combination circuit extraction procedure of procedure 1. When the extraction of the fault candidate points is completed in step 7, the location match determination of all the estimated fault candidates obtained in step 6 is performed in the failure candidate point location match determination in step 9. In this coincidence determination, an operation is performed to compare and detect, for each location (for each gate or each pass), whether or not it appears several times in the failure candidate point list registered in the procedure 6.

By the weighting and extraction in step 10, the frequency of appearance is determined for each candidate fault point, the weight is calculated, and the fault estimation list is created by extracting the weight in descending order of the weight. In step 10, the frequency of appearance of the same portion compared and detected in step 9 is obtained and sorted in descending order of appearance frequency. That is, among the estimated fault propagation paths and the estimated fault locations extracted in the estimation for each combinational circuit, the fact that the frequency of appearance of the true fault or the propagation path increases is used. Because a single stuck-at fault,
This is because, assuming an open fault and a short fault, a true fault will always be included as one of the results of estimation of any fault output.

Combinational circuit input terminal state estimation vectors # 1 to # 1
Means for estimating a fault estimated location and a fault propagation path in each combinational circuit in place of the procedure for estimating the state of the input terminal of the combinational circuit and finally extracting the location of the estimated failure in the procedures 2 to 4 of # 3 It is also possible to carry out according to each flow of FIGS.

FIG. 5 is a flowchart showing a procedure for estimating a fault propagation path and a state of an input terminal of a combination circuit when it is assumed that no failure exists in the combination circuit. Node state estimation procedure 1 using all output states of the combinational circuit (procedure 11)
2, the state is estimated for each gate. In step 11, state values are set for all output terminals of the combinational circuit. A failure state value is set for a fault output or a terminal estimated to have propagated a fault, and an expected value is set to a normal terminal or a terminal estimated to be propagated a normal state. In addition, it is unknown whether a fault state has propagated or a normal value has propagated.
"0" which does not directly contribute to fault propagation to the output terminal,
An undefined “X” is set for “1” terminal. This operation is the initial setting 11.

Then, in step 13, the state is compared with the expected value, and it is determined whether there is a difference from the expected value. Register a node. If the estimated state matches the expected value, another node is compared with the expected value, and a node different from the expected value is searched.

Further, going back to the node different from the expected value,
The propagation path from the combination circuit input terminal and the combination circuit fault propagation terminal are extracted. FIG. 6 does not calculate the operation for estimating the propagation path by comparing with the expected value for each node, but first estimates the state value of the input terminal of the combinational circuit (step 17).
I do. A logic simulation (step 18) is performed using the obtained estimated state value of the input terminal of the combinational circuit to obtain a state value of each node. Further, in step 19, the difference from the expected value is calculated for each node.
At 0, it is registered as an estimated propagation path.

FIGS. 7 and 8 show a procedure for estimating a fault location and a fault propagation path on the assumption that a fault exists inside the combinational circuit. FIG. 7 shows a method of estimating a fault location and estimating a fault propagation path at the same time as estimating a state in a combinational circuit, similarly to FIG. FIG. 8 shows a method of estimating the input state of the combinational circuit and then obtaining the result by logic simulation, similarly to FIG.

In FIG. 7, first, in step 21, the terminal whose state is estimated and the terminal whose state is determined at the output terminal of the combinational circuit are initialized, and the procedure is performed based on the state value set in this step 21. At 22, the estimation of each node state value is performed in the same manner as in the procedures 2 to 4 of FIG. In step 21 of FIG. 7, as in steps 11 and 16, the state values are initialized for the terminals whose states are estimated and determined at the output terminals of the combinational circuit. That is, regardless of the normal state / failure state, the state value is set where the state is determined, and the terminal whose state value is undefined is treated as undefined.

In step 23, this node state estimation procedure 2
In step 2, it is determined whether an estimated value has been obtained. If the estimated value has been obtained, step 25 is performed. If the estimated value has not been obtained, step 24 is executed. If the estimated value cannot be obtained, it means that it is not possible to go back to the previous node, and the input terminal of the gate whose input state is to be estimated has already been determined by estimation from another net. In some cases, a state satisfying the output cannot be obtained. In such a case, the failure propagation path cannot reach the combinational circuit input terminal, resulting in a failure inside the combinational circuit. FIG. 9A shows this state.

A node that can estimate a node state from an output failure terminal and can estimate the node state up to the input terminal of the combinational circuit is an estimated failure propagation path, which cannot reach the input terminal and stops estimation in the combinational circuit. Location. The estimated fault location branches off from the estimated fault propagation path in FIG.
It indicates the route to the point where estimation is impossible and it stops. If the node state can be estimated in step 23, the state is compared with the expected value in step 25, and if it is different from the expected value, the path is temporarily stored as a propagation path due to a failure in step 26. This is because at this time, it is not possible to determine whether the estimated path is a fault propagation path from the preceding combinational circuit or a failure point in the combinational circuit, and it depends on whether there is an estimated state value when the next node estimation is performed. Because it can be determined, it is necessary to temporarily store it.

If the node state cannot be estimated in step 23, it means that there is a fault in the combinational circuit. The number of faults already estimated in step 24 in the combinational circuit and the number of faults newly obtained in step 23 are obtained. By checking the estimated failure location, it is confirmed that the assumed failure number is 1 or 2. If the estimated number of failure points in the combinational circuit exceeds two, the path obtained in step 26 is deleted because the above obtained estimation does not actually match, and the obtained path is already obtained under the same conditions. The combined circuit input terminal state estimated value is also deleted for the same reason. If the estimated number of faults is 2 or less in step 24, the estimation can actually be performed, so the path accumulated in step 26 is registered as an estimated fault location (step 27). The state of each node in the combinational circuit is estimated, and it is determined whether the failure state of the output terminal reaches the input terminal of the combinational circuit and whether the estimated number of failures in the combinational circuit at that time is 2 or less, and the condition is satisfied. If the estimation is correct, the route accumulated in step 26 is registered as an estimated fault propagation route in step 30.

FIG. 8 shows a method using logic simulation. As shown in FIG. 9 (b), an input state value satisfying the state of the output terminal that becomes faulty is obtained, and a fault propagation path is obtained by logic simulation. An estimated fault location and an estimated fault propagation path are extracted by deleting a path through which a fault output propagates to an output terminal that does not become an error. In FIG. 9 (b), the hatched portion is the deleted path. As the actual processing, first, in step 31, only the terminal whose output terminal of the combinational circuit is in the estimated state and the terminal which is determined to be faulty among the determined terminals and the state value thereof are initially set. .

Further, in step 32, the state value of the combinational circuit input terminal is estimated based on the initially set state value, and in step 33, the combinational state is obtained using the state value of the combinational circuit input terminal obtained in step 32. Perform a logic simulation of the circuit.
In step 34, the obtained node state value in the combinational circuit is compared with the expected value, and in step 35, all the fault propagation paths propagating from the input terminal of the combinational circuit to the output terminal are extracted.

In step 36, an illegal output terminal to which a fault has propagated other than the fault output terminal set in step 31 is extracted from the obtained estimated fault propagation paths. In step 37, not only the propagation path from the combinational circuit input terminal to the illegal output terminal but also all the propagation paths in the input direction from the node where the propagation path to the faulty output terminal and the propagation path to the illegal output terminal intersect are extracted. Then, in step 38, this route is deleted. This is because if a fault condition appears at that node, including the crossing node, the node whose output is assumed to be normal always fails, that is, there is an activation path from that node to the output terminal. This is because the output terminal cannot be satisfied.

In step 39, the propagation path obtained in step 38 is registered as the estimated failure propagation path reaching the input terminal of the combinational circuit, and the path that cannot reach the input terminal is registered as the estimated failure point. The fault estimation of the entire sequential circuit is performed by using the fault estimation location and the fault propagation path obtained by the processing of FIG. 5, 7 or 6 or 8.

As a second embodiment, a flowchart of the fault estimating method shown in FIG. 10 will be described. Step 40-
43 is a procedure similar to procedures 1 to 4 in FIG. 1, in which a fault is estimated based on whether or not a fault exists in the combination circuit after extracting the combination circuit from the fault output terminal. Combination circuit input terminal state estimation # 1 to # 3 (procedures 41 to 4)
In 3), the fault propagation path in the combinational circuit and the fault location are estimated using the method shown in FIGS.

In step 44, the combinational circuit input estimation vector is simplified, and in step 45, path selection is performed. This route selection procedure 45 is means capable of greatly reducing the state value obtained for each combination circuit input terminal state estimation.

This procedure 45 is performed after the simplification of the procedure 44 to further reduce the estimated tracking path. In the simplification procedure, since the Boolean algebra is simplified, the estimation vectors that cannot be simplified remain as they are, and as a result, the number of estimation vectors to be tracked may increase. However, failures can be broken down into single stuck-at faults, open faults, short
If a fault is assumed, a true fault should be reached no matter which fault propagation path is traced. Therefore, by tracing together a vector having a common fault propagation terminal (and state), the path is further tracked. It is possible to reduce the number of estimated vectors to be performed (because it is necessary to track each estimated vector).

Also, this procedure 45 uses the fact that the fault propagation path is always connected to the true fault location, and that even if any fault propagation path is traced, the true fault location is reached, and the propagation path is traced back. There are four selection methods as shown in FIGS. In the path estimation method shown in FIG. 11, a terminal of “fail” is extracted for each vector from an input state estimated value that is an estimation result at an input terminal of a combinational circuit.
The estimated input vector obtained by simplification is classified for each terminal including l, and 0/1 / Y / Q /
This is represented by the state of D.

"0" indicates a state where the expected value is "0", the estimated value is "0", and the same value is "0" in all of the classified estimated vectors. Similarly, "1" indicates the expected value "1". , The estimated value “1”
And the same value “1” is shown in all the estimated vectors that have been classified. “Y” indicates the expected value “0”, the estimated value “1”, and the same value “1” in all of the classified estimation vectors, and “Q” indicates the expected value “1” and the estimated value “ 0 "and the same value" 0 "in all of the classified estimation vectors. “D” indicates that both the expected values “0” and “1” are good, but the estimated values are not common to the estimation vectors, and both “0” and “1” are present depending on the vectors.

In FIG. 11, Fail1 to Fail6 classify the estimated input vector for each fail terminal surrounded by a broken line, and summarize the estimated values from the classified vector by the above procedure.
In the route selection method shown in FIG. 11, in order to cover all the fault propagation routes, a route selection and a collective operation are performed for all the failed terminals, and the same vector is included in some vectors after the route selection. Will be. According to this route selection method, the estimated vector obtained by the simplification can be reduced to the number of input terminals of the combinational circuit at the maximum.

FIG. 12 shows, similarly to FIG. 11, a terminal of fail for each vector extracted from the input state estimated value as an estimation result at the input terminal of the combinational circuit, and obtained for each terminal including fail in a simplified manner. The estimated input vector is classified into 0/1 / Y / Q /
This is a method of collecting the estimated vectors in the state of D and selecting a route. In this case, instead of classifying all fail terminals,
In order to sequentially classify all the simplified estimated vectors, the classification is repeated until all the vectors are covered.

FIG. 12 shows that classification is first performed at the terminal of Fail1, and then classification is performed at the terminal of Fail2. When the classification of Fail 2 is performed, the input estimation vectors are all covered, and the failure location can be reached by tracing the selected route, so that no further classification is performed. At this time, the vectors classified as Fail1 are Fail3, Fail
Includes Fail3, Fail5, Fail6 and Fail2
6 vectors. The order of the Fail terminals to be collected is not particularly defined, and is arbitrary. In this method, since the number of estimated vectors is always equal to or less than the number of input terminals of the combinational circuit, the number of estimated input vectors can be significantly reduced.

The route selection method of FIG. 13 is the same as that of FIG. 14, and extracts the terminal of “fail” for each input state estimation vector.
Are classified and the estimated input vectors obtained by simplification are classified for each terminal, and the estimated vectors are collectively routed in the state of 0/1 / Y / Q / D. In FIG. Was optional. In this route selection method, each Fail
The frequency of occurrence of Fail is obtained through all the estimated input vectors for each terminal, the vectors are classified in order from the Fail terminal having the highest frequency of occurrence, and the classification is performed until all the estimated input vectors are covered.

In FIG. 13, the appearance frequencies of the Fail terminals of Fail 1 to Fail 6 are 8, 3, 3, 2, 2, and 5, respectively, and the order of classification is Fail 1, Fail 6, Fail 2, or Fail 3, Fa.
It becomes il4 or Fail5. In the example of this figure, Fail
No further classification is performed to cover all the input estimation vectors at the stage where the classification is performed by the terminal of No. 1 and then the classification is performed by the terminal of Fail 6. At this time, the vectors classified by Fail1 include Fail2, Fail3, Fail4, and Fail5, and Fail6 includes the vectors of Fail2, Fail3, and Fail5. In this route selection method, the output frequency is calculated for each Fail terminal, and the classification is performed in order from the Fail terminal with the highest output frequency. Is more likely to be reduced.

The appearance frequency for each Fail is classified into a case where the appearance frequency obtained first is used for each classification, and a case where the appearance frequency of the Fail terminal which has not been classified for each classification is obtained again and used. Can be The former is a simple procedure,
The latter has the advantage that it is easier to combine with a smaller number of vectors.

FIG. 14 does not fix the number of Fail terminal selections to 1 as in the route selection method of FIGS. 11 to 13, but calculates the appearance frequency, and then extracts and summarizes a plurality of Fail terminals in descending order of frequency. It is a procedure to go. FIG. 14 shows a case where two input terminals are extracted and collected until all the estimated input vectors are covered.

In this method, since common estimation input vectors are collected for a plurality of terminals, as shown in FIG.
6, Fail1 & 4, Fail1 & 5, Fail2 & 6, Fail1 & 3
, Resulting in the estimated number of vectors shown in FIG.
Although the number is larger than that of each of the thirteen methods, the estimated vector compiled for each Fail terminal has fewer terminals “D” including a fault, and has more terminals whose values are determined.

As another method of the route selection method shown in FIG. 14, a method of extracting and classifying Fail terminals based on a threshold determined in advance from the frequency of appearance, a method of extracting a plurality of Fail terminals having an appearance frequency equal to or higher than a predetermined threshold is used. Extracting and classifying the Fail terminals with the determined number of terminals when the classification is below the threshold by this classification, and classifying the state estimation vectors of the combinational circuit input terminals collectively, the route selection method etc. is there.

In the above-described route selection method shown in FIGS. 11 to 14, the description has been made using the state values of 0/1 / Y / Q / D. An equivalent function can be realized by expressing the terminal state using a state value (for example, M / N) that is classified in more detail.

In this case, “M” is the expected value “0”, and the estimated value is not common to the estimation vectors.
It is assumed that both “1” are present, “N” is the expected value “1”, the estimated value is not common to the estimation vectors, and the vector indicates the state where both “0” and “1” are present. I do. Compared to the case where the state value “D” is used, the use of M / N has an advantage that the expected value 0/1 can be determined. However, by examining the expected value, the failure state can be determined even for D, so that the same effect is obtained.

The same applies to the case where the route is selected using the result including don't care “X” for simplicity. However, the state value of the combinational circuit input terminal may be “0” or “1”. If the state and another 0/1 / Y / Q / D state are present at the same terminal, they are replaced with "D". This indicates a state in which “X” can be either 0/1, while the other states indicate either 0/1. Therefore, it is necessary to set “D” as a result. It is.

In step 46, the estimated fault location for each combinational circuit obtained in combinational circuit input terminal state estimation # 1 to # 3 including the route selection method shown in FIGS. 11 to 14 is checked (see FIG. 10).
Determine whether the sum of the estimated number of fault locations under the conditions up to that point when the estimated state of the output terminal of the combinational circuit is obtained and the newly obtained estimated fault location exceeds 2, and if the sum exceeds 2, In step 47, the new fault candidate points obtained by the estimation of the combinational circuit in step 47 and the estimated fault propagation path under the same conditions are reduced.

In this operation, as shown in FIG. 15, when it is assumed that one fault exists, the fault types include 0/1 degeneration, open, and short. The estimation results up to immediately before the combinational circuit include (1) when only the propagation path is estimated, (2) when the propagation path + one estimated failure point is obtained, and (3) when the propagation path + estimated failure is obtained. There are three types when two locations are obtained.

Further, when the estimation of the combinational circuit is performed based on the estimation result, similarly, the propagation paths and the estimated failure locations of the above (1) to (3) are obtained. However, assuming a single fault, the physical location is 0/1 stuck-at fault, op
Since there can be only one location for an en failure and two locations for a short failure, the number of estimated failure locations must be two or less. Therefore, when the estimation result up to immediately before is only the propagation path, in the estimation of the combinational circuit, (1) only the propagation path, (2) one propagation path + one estimated failure point, and (3) propagation path + one estimated failure point There is no problem in any of the locations, and the type of failure can be classified according to each.

There are cases where the estimated failure location is short at both of the two locations. However, this causes an error in one of the two nodes at the same time due to the short, and as a result, only one estimated failure at that time This is the case where the location cannot be obtained. For short with two estimated failure locations, short
Due to t, both nodes may make an error at the same time, and two failure locations may be seen. When there is an actual conflict, the state value of one of the nodes is strong, so there are many single-point failures. In this case, a special short example is also considered. Represents.

In both cases where the estimation result up to immediately before is the propagation route + estimated fault location and the propagation route + estimated failure location at two locations, the criterion is the same as in the case where the estimation result immediately before the previous period is only the propagation route. However, when the number of estimated failure locations becomes three, location coincidence determination is performed. Even if there are three or more estimated failure locations, if there are two or less physically estimated failure locations, they are left as failure candidate points. The portion described in parentheses in FIG. 15 is that case. In that case, in step 48, it is registered as a newly extracted fault candidate point. At this time, if two or more estimated fault locations have already been obtained, only the locations where the physical positions match among the estimated fault locations determined by the respective combination circuits as failure candidate points as shown in FIG. Treat and register. FIG. 17 shows one of the two locations, and the bold line in the figure is an estimated fault location in each combinational circuit, and the thick line portion that matches when superimposed is the estimated fault location to be registered.

Next, combination circuit input terminal state estimation # 1
Simplification is performed in step 49 using the estimated vector of the combinational circuit input terminal obtained in # 3, and the path is selected in step 50 based on the simplified result. Step 51 is a procedure for determining whether or not the failure estimation has been completed by checking whether or not there is a route selection result for which estimation has not been performed, similarly to the procedure 7 of the first embodiment (FIG. 1). If not, the preceding combination circuit extraction estimation is repeated. When going back to the preceding combinational circuit, it is checked whether or not a clock has been applied to the flip-flop and the state has been updated from the expected value of the clock line of the flip-flop which is presumed to have propagated the fault. If the clock line is activated, it goes back to the previous time, and if it is not activated, it goes back to the activated time and repeats the circuit extraction and estimation of the previous stage.

As shown schematically in FIG. 18, the procedure 52 uses all estimated fault candidate points obtained at all times obtained in the new fault candidate point registration procedure 48 in FIG. This is a procedure for calculating the appearance frequency for each. The height of the bar in FIG. 18 represents the appearance frequency, and the higher the height, the higher the possibility of failure.

As described above, this procedure 52 uses not only the estimated candidate points obtained in the new fault candidate point registration procedure 48 but also the estimated candidate points obtained in the new fault candidate point registration procedure 48. There is a procedure using the fault propagation path obtained by the estimation of the combinational circuit. The latter case is advantageous in frequency calculation when a fault is present on a path obtained as a fault propagation path by estimation in a combinational circuit.

The procedure 53 assigns priorities in descending order of frequency using the obtained appearance frequencies of the respective failure candidate points, and calculates the physical position of the estimated failure location and the estimated failure state value as shown in FIG. The type of the failure is determined by the determination. FIG.
The judgment criteria of 6 are as follows: (1) When the physical position and the estimated fault state value are always fixed, there is a high possibility that a 0/1 stuck-at fault has occurred. (2) The estimated fault location is one and If the physical location is the same and the estimated state is different at that time, it is highly likely that an open failure or short failure has occurred.
(3) Two presumed failure locations and the same physical location,
Furthermore, if the estimated state differs at that time, there is a high possibility that a short failure has occurred.

Further, by adding a procedure for performing a timing failure determination as shown in FIG. 20 to the estimated fault location and the failure type obtained in the procedure 53, a timing failure can be handled for the failure type. In the timing failure determination flow shown in FIG. 20, first, in step 54, it is checked whether or not the estimated failure location includes a flip-flop. This is determined based on whether or not the extracted fault candidate includes the flip-flop terminal as shown by the bold line in FIG. In the case of FIG. 19, it is possible to assume that a fault exists on the path of the thick line, or that a timing failure occurs in the flip-flop in the hatched portion and that the output of the flip-flop is erroneous. Therefore, in the procedure 55, all paths in the combinational circuit reaching the flip-flop which is presumed to be the fault location are extracted.

In step 56, the path delay is calculated for all the paths obtained in step 55, and in step 57, the critical path is obtained based on the obtained delay for each path. Further, the critical path is compared with the expected value of the combinational circuit input terminal, it is confirmed that the state of each node on the critical path has changed, and it is determined according to a procedure for determining whether or not the timing is defective. If the estimated failure location does not include a flip-flop in step 54, timing failure detection is not performed.

Further, when a scanning flip-flop in a circuit is included as shown in FIG. 21, a scan path indicated by a bold line in the drawing is also extracted when a combinational circuit is extracted, and path tracing and fault estimation are performed. Do. At this time, as a means for determining whether or not the scanning flip-flop is faulty, it is determined in advance whether or not the scanning flip-flop is included in all the circuits, and further the estimation of the extracted combinational circuit is performed. In, a combinational circuit not including the scan path indicated by the bold line in the figure is extracted, and the path is tracked and a fault is estimated.

The result of the failure estimation including the scan path is compared with the result of the failure estimation not including the scan path to determine whether or not a failure exists in the scan path itself including the scan flip-flop. As a result, a scan path failure is detected.

When the terminal of the memory cell is reached at the time of extracting the combinational circuit, the procedure for tracing and estimating the path of the data line, address line, and control line of the memory cell is changed according to the conditions as shown in FIG. First, when the terminal reaches the terminal of the memory cell, it is recognized whether it is an address line, a data line, or a control line of the memory cell. Extract and estimate the deraline or address line.

In the case of the address line, the data line is tracked by using the internal storage data. In the case of the data line, based on the estimation result of the data line, as shown in FIG.
In step 59, data collation inside the memory cell is performed. After determining in step 60 whether or not another address contains the same data as a result of data collation in the memory cell, if the address is included, it is highly likely that the address holding the same data has been read. Traverses an address line or a control line based on the address data.

If the same data does not exist in the memory cell, it is possible that the data at the time of writing is wrong. Go back the data line. The normal LS can be obtained by estimating the fault location according to the above procedure.
The fault diagnosis of I becomes possible. Note that the flowchart of FIG. 10 can also be rewritten as shown in FIG. 24, just as the flow of FIG. 1 is rewritten as FIG.

Next, another embodiment of the present invention will be described in detail with reference to the drawings. FIG. 25 is a schematic flowchart of this embodiment. The procedure of the present embodiment includes a combinational circuit extraction 71, a combinational circuit input terminal state estimation 72, an estimated fault propagation path registration 73, a circuit backward determination control 74, a path branch,
It has the procedures of joint recognition 75, failure mode classification 76, and failure candidate frequency weight comprehensive judgment 77.

Step 71 in FIG. 25 is a combination circuit extraction, and step 72 is a procedure for obtaining an input terminal state from an output terminal state for each combination circuit obtained in the step 71. Also,
Step 73 registers the estimated fault propagation path when obtaining the input terminal state of the combinational circuit, and goes back to steps 71 to 73 until the estimated state cannot be obtained.
In step 4, it is controlled whether or not to continue the estimation for each combinational circuit.

When the fault estimation for each combinational circuit is completed for all the vectors, in step 75, based on the estimated fault propagation path for each combinational circuit obtained in step 73, path branching and joint recognition are performed, and the fault estimation is performed. The fault propagation path from the location to the fault output terminal is reconfigured. By this operation, it is possible to classify a single fault point, a multiple fault point, or a fault propagation path instead of a fault candidate with respect to the fault candidate points on the fault propagation path.

In step 76, the failure mode is classified using the result obtained in step 75 and further using the estimated state value at that time and the state values of all vectors. In step 77, the failure candidates are comprehensively determined and ranked using the frequency of the number of failure output terminals, the number of failure propagation paths, etc., and the failure candidate locations, state values, and failure modes that further satisfy the failure state, and weighted as the failure candidate accuracy I do.

In particular, the failure modes are classified based on the estimated failure propagation paths shown in steps 75 to 77, and the frequency and state value of each failure candidate are checked to determine the total likelihood of the failure. Is weighted.

FIGS. 26 and 27 are flow charts showing details of the present embodiment. FIG. 26 shows the entirety of the path estimation and the weighting of the failure estimation part for each combinational circuit, and FIG. 27 shows the path weighting method estimated for each combinational circuit.

The sequential circuit fault diagnosis method of the present invention uses all expected value information for all vectors of all flip-flops in an LSI prepared in advance, pass / fail output information of an actual tester, and connection information of all circuits. In addition, the fault estimation is performed by dividing the circuit into flip-flops and combinational circuits, and a fault propagation estimation path and a fault candidate point are obtained for each combinational circuit. The fault candidates are weighted based on the number of output terminals or the number of fault propagation paths existing between the output terminals and the fault propagation terminals.

First, in the procedure of the combinational circuit extraction 40, LS
Trace the input from the fault output of I or the output terminal that is already estimated to be faulty, and trace the output from the obtained input terminal or flip-flop output terminal to the output terminal or flip-flop input terminal. obtain. Further, a circuit is extracted in the input direction from the obtained output terminal or flip-flop input terminal to obtain a combinational circuit.

The obtained state estimation of the combinational circuit includes combinational circuit input terminal state estimations 41 to 43, and conditions to be obtained are different from each other. Combinational circuit input terminal state estimation 41
FIG. 28A shows an input terminal estimation procedure when it is assumed that no fault exists in the combinational circuit as shown in FIG. 28A and a fault signal is propagated from the preceding combinational circuit.

Combinational circuit input terminal state estimation 42, 43
Is based on the assumption that a fault exists in the combinational circuit.
Furthermore, the number of failures is divided into two. One is a combination circuit input terminal state estimation 42, which is a procedure for obtaining an estimated failure location and a combined circuit input state estimated value when it is assumed that one failure exists in the combination circuit as shown in FIG. is there. The other is a combinational circuit input terminal state estimation 43, which finds an estimated failure location and a combined circuit input terminal state estimation value when it is assumed that two failures exist in the combinational circuit as shown in FIG. Procedure.

In each procedure of the combinational circuit input terminal state estimation 41 to 43, a special example is a case where an estimated value of the combinational circuit input terminal state is not obtained and only one or two failure estimation parts are obtained. . In this case, there is no input state that satisfies the combination circuit output terminal state, indicating that no fault has propagated from the preceding combination circuit.

Simplification is performed using the input estimation vector obtained by the combinational circuit input terminal estimation 41, simplification of the Boolean algebra is performed in the procedure of the route selection 44 (45), and only the terminals having the same state value are collected. Reduce the number. As a simplification, an input state estimation value that satisfies the output of the combinational circuit is first examined to determine whether the state value of the terminal is 0 or 1, that is, a terminal and a state value that are not directly involved in a failure at the output of the combinational circuit. Are extracted, and terminals having the same state value are collected regardless of whether the terminal is in a failure state or not.

That is, terminals having different values are extracted, and if it is one terminal, it is 0/1, and if it is two terminals, it is 00/01/10 /
In the case where all four state values of 11 are included, replacing the terminal state values with "X" is a simplification. Further, as shown in FIG. 29, the route selection procedure also reduces the number of estimated vectors by grouping vectors having the same fault propagation terminal and state value of the obtained combined circuit estimation input vectors. After estimating the state of the combinational circuit input section in step 72, the procedure returns to the combinational circuit extraction 71 because it is necessary to go back to the preceding combinational circuit.

The new estimated fault candidate points obtained by the combinational circuit input terminal state estimations 42 and 43 are as follows.
In the judgment procedure 46, it is judged whether or not the total number of the fault candidate points obtained under the same condition and the newly obtained fault candidate points is 2 or less. It is deleted in the procedure of the fault candidate point and propagation path deletion 47. If the number is 2 or less, a new registration is performed according to the procedure of new extraction failure candidate point registration 48. Combinational circuit input estimation vector simplification, route selection 49
In (50), the number of estimated vectors obtained is reduced in the same manner as in step 44 (45).

In step 51, it is determined whether or not the fault candidate point has been extracted for all the vectors, and whether or not the estimated fault propagation input terminal from the preceding stage can be obtained any more as described in the special example above. It is determined whether or not to complete the estimation. If the estimation has not been completed,
Extract the fault propagation terminal from the combinational circuit input estimation vector,
Further, the procedure returns to the procedure of the combinational circuit extraction 40 and goes back to the preceding combinational circuit. When all the estimations have been completed in step 51 and the failure candidate points have been completely extracted, the occurrence frequency of each candidate point is obtained by a failure candidate point frequency calculation 52, and a location match weighting is performed by a weight 63.

Further, state value determination & simulation 64
To estimate the failure mode and improve its accuracy, and create a failure estimation list. In this operation, simulation is performed to ensure that the fault location matches its state value with the normal terminal output, and by examining the state value of each fault estimation candidate location, 0/1 degeneration, open, bridge fault, etc. Can be determined.

FIG. 27 shows a detailed procedure of the above procedures 52 and 63. Path estimation for each combinational circuit in step 78 is performed in steps
51, a route reconfiguration procedure up to the output terminal 7
In step 9, the connection relation between the estimated propagation paths for each combinational circuit is searched for as shown in FIG. Is reconfigured as a route.

Further, in the procedure of the frequency calculation 80 for the number of failure output terminals for each candidate point, the frequency is obtained as shown by the circled numbers in FIG. The fault output frequency in FIG. 30 means that the fault propagation estimated path reconstructed in the step 79 is traced in order from the fault output terminal, and the path is subdivided for each branch and connection of the path. Whether the operation is started from the failure output terminal is sequentially stored.

In the figure, the circled numbers indicate the starting failure output terminal numbers, and indicate that the failure output frequency numbers are 2 and 3, respectively. In other words, the failure output terminal cannot be satisfied by the candidate on the path indicated by, and the failure candidate on the failure propagation path indicated by satisfies the failure output to all of them at the same time. This operation makes it possible to associate each fault output terminal with a fault propagation path and a fault estimated location at each time point.

In the procedure 81, the procedure 80 is repeated for all the vectors, the frequency of the fault output terminal of the fault candidate location for each fault propagation path is obtained from all the fault outputs, and the result of storing the fault output terminal in the procedure 80 which is started is stored. The frequency of the number of fault output terminals of the entire vector is obtained based on In step 82, it is determined whether or not the failure output terminal frequencies for all the vectors have been obtained. In step 83, rearrangement is performed in descending order of the frequency of the number of failure output terminals obtained in step 81, and weighting of failure candidate portions is performed.
In other words, it is highly probable that a failure candidate location that satisfies more actual failure outputs among all failure candidates is true, and this operation selectively extracts the failure location.

As still another embodiment, there is a weighting method shown in FIG. The procedure shown in FIG.
Is another method for realizing the procedures 52 and 63 of FIG. Step 78
26 means the steps 40 to 51 in FIG. 26. In the path reconfiguration procedure 79 up to the output terminal, the newly extracted fault candidate point in step 48 and the path obtained by the fault propagation path registration are used. As shown in FIG. 30, the connection relation between the estimated propagation paths for each combinational circuit is searched and reconfigured as a path to a fault output terminal.

Further, in the procedure of the frequency calculation 84 of the number of fault propagation paths for each candidate point, the frequency is obtained as shown by the circled numbers in FIG. The fault output frequency in FIG. 32 means that the fault propagation estimated route reconstructed in the step 79 is traced in order from the fault output terminal, and the route is subdivided for each branch and connection of the route. Whether the operation is started from the failure output terminal is sequentially stored. The circled numbers in the figure are the starting fault output terminal numbers, and the numbers indicate the fault output frequency numbers 2 and 3, respectively. At this time, if a branch occurs when tracing from the output terminal side, add a '
It shows the necessary conditions for establishing the propagation path on the output side.

That is, FIG. 32 shows a case where the fault propagation is realized only when the fault is propagated simultaneously with “′”. Similarly, when the paths are coupled when tracing from the output terminal side In the case of the same failure output terminal, the state of 'is added together, so that if the failure output terminal is the same, the state of' is also stored. ´ ”)
Since "" + "" = "from the path connection of"
+ "=, Which is the result. In the case of FIG. 32, since the fault propagation path as shown in FIG. Satisfying this failure output,
The path marked with cannot be a failure candidate.

According to this operation, it is possible to relate in detail the fault propagation path and the necessary condition of the fault output for each fault estimated location at each time point, and it is possible to narrow down the fault location with high accuracy.

In the procedure 85, the procedure 84 is repeated for all the vectors, the number of fault propagation paths at the fault candidate locations for each fault propagation path is obtained from all the fault outputs, and the output terminal of the procedure 84 is started from which output terminal and passed through which path. Based on the stored result, the frequency of the number of fault propagation paths of the entire vector is obtained. Step 86
Then, it is determined whether the frequency of the number of fault propagation paths in all the vectors has been obtained. In step 87, rearrangement is performed in descending order of the frequency of the number of fault propagation paths obtained in step 85, and weighting is performed on the fault candidate portions. By this operation, a candidate fault location satisfying the fault output is selectively extracted.

As still another embodiment of the procedure for calculating the frequency of the number of fault propagation paths for each candidate point, there is also a method shown in FIG. Route estimation procedure 7 for each combinational circuit in the method shown in FIG.
8 means steps 40 to 49 in FIG.
Then, the fault propagation path is reconfigured from the output terminal side to the input terminal side, and further, the path branching and connection points are recognized and subdivided.
FIG. 34 is a block diagram showing the procedure of FIG. 33. In FIG. 34, AE are subdivided routes.

In FIG. 34, there are two connection points of the route, that is, the junction of D and E, and the junction of A and C, and shows the correspondence between the failure mode and the failure candidate with respect to the inter-path condition at each connection point. FIG.
The meaning of the table in FIG. 4 will be described using (AorC) and (DorE) as examples. In the fault estimation of this method, as shown in FIG. 26, the combinational circuit goes back sequentially from the fault output to the input direction. I do. The satisfaction condition calculation procedure 88 for each fault propagation path is a procedure for finding a condition that satisfies the fault propagation on the subsequent path, and the satisfaction condition DorE,
AorC is a condition for finding the condition DorE that satisfies the fault propagation, and a condition AorC that satisfies the fault propagation E.

That is, the conditions for establishing the fault propagation at each branch point of the fault propagation path are shown in parentheses, and the conditions are also separated by ",". Therefore, (A
The meaning of (AorC) of (orC) and (DorE) indicates that when a fault propagates to the path E, the condition for satisfying the fault of the path E is that the fault propagates to the path A or the path C. If a fault exists on the route E, there is no need for a fault on the route A or the route C. As the failure candidates satisfying this condition, A, B, C, D, E, A & B, A & C, A & D, A
& E, B & C, B & D, B & E, C & D, C & E, D &
E is given.

Similarly, in step 88, the inter-path condition (Aor
C) and (D & E) are obtained as follows. , D & E = B & E, C & D = C & B =
B & D, and then = (A & B) or (C & B) = (A & B) orB = B.

Therefore, the route B, the route A & the route B, the route A & the route D, the route B & the route C, the route B & the route D, the route B & the route E, the route C & the route D, the route D & the route E are obtained from the above equations.
Are failure candidates and are 1 in the table of FIG. On the other hand, 0 in the table indicates that it is not established as a failure candidate.

In the failure candidate classification procedure 89 according to the failure mode, the obtained failure candidates are classified into a single failure and a double failure. In the example of FIG. 34, the case where 1 is displayed in the route A to the route E is a single fault candidate, and the route A & route B to the route D & route E are double fault candidates. (Aor
Looking at (C) and (DorE), no matter where a fault exists in the paths A to E, the fault propagation is satisfied.
The routes A to E are classified as single failure candidates in this procedure. Further, all of the route A & route B to route D & route E are 1, which indicates that a double failure may occur in any combination of the route A to route E.

In step 90, it is determined whether or not the frequency calculation has been completed for all the candidate points of all the vectors. If the frequency calculation has not been completed, the process returns to step 79, and the process returns to another combination circuit that has not performed the frequency calculation. The same processing as described above is performed. If it has been completed, the procedure proceeds to step 91. In step 91, the failure modes in all the vectors are calculated based on the failure candidates for each failure mode obtained in step 89. In step 89, the faults are classified into single faults and double faults. However, since single faults include single stuck-at faults and open faults, single stuck-at faults based on fault candidate points from all vectors and their state values are used. Classify faults and open faults.

In the classification method, the determination is made based on whether or not the fault state values and the normal state values of all the vectors always become a constant state value. If the state value always becomes constant, it is regarded as a 0/1 stuck-at fault, and if the state value takes both values of 0/1 depending on the time, it is regarded as an open failure. A double fault is classified as a bridge fault. Based on the failure candidate points for each failure mode obtained in step 91, the route frequency for each failure candidate point in all vectors is calculated in step 92, and weighting is performed in descending order of frequency.

FIG. 35 shows still another embodiment. FIG.
33 are the same as steps 78 to 90 in FIG. 33. In step 79, the fault propagation path is reconfigured from the output terminal side to the input terminal side. I do. Procedure 88 is a procedure for obtaining a condition that satisfies the fault propagation on the path at the branch point and the connection point.

In the fault candidate classification procedure 89 based on the number of fault locations, the obtained fault candidates are classified into single faults and double faults. In step 90, it is determined whether or not the frequency calculation for all the candidate points of all the vectors has been completed. If it has not been completed, the process returns to the step 79, and the above is performed for another combinational circuit that has not performed the frequency calculation. The same processing is performed. If it has been completed, the procedure proceeds to step 93.

In step 93, the frequency for each fault candidate is calculated using the fault candidate classification for each number of fault locations of all the vectors obtained in step 89, and rearrangement is performed in descending order of frequency. Further, in step 94, the failure mode is checked for all vectors, and the frequency of each failure candidate is calculated for each failure mode.
The operations of the procedures 93 and 94 are procedures for obtaining the failure mode more precisely. In the case where a single fault is obtained in the classification for each number of faults obtained in step 89, not only an open fault but also a double fault may be included, so it is necessary to further classify the fault. Perform this operation.

As described in the procedure 93, even if the fault is a single fault, the possibility of an open fault is high unless the fault estimated state value and the normal estimated state value are always constant. However,
As shown in FIG. 36, the state where the estimated state value is not constant throughout all the vectors may be a bridge fault between candidate fault points obtained as a single fault equivalent fault location. × in FIG. 36
The portions marked with "" indicate a failure location that is presumed to have the same state at the same time.

Therefore, as a means for extracting a bridge fault from a single fault location, a fault location whose state value is not constant is extracted through all vectors, and a frequency calculation is performed for each fault location. The estimated state values are compared with each other to extract a fault location in the same state at the same time. Then, any two locations having a frequency equal to or higher than a certain threshold are selected from the extracted failure locations, and the frequency in the case where a failure exists at the same time is obtained. By this operation, it is possible to extract and select a bridge fault that satisfies the output fault state more frequently from those assumed to be open faults.

In step 92, similarly to step 92 in FIG.
The frequency of the fault candidate points classified for each of the single 0/1 stuck-at fault, open fault, and bridge fault modes are rearranged, and the estimated fault location is weighted. At this time, regarding the bridging fault, both the estimated bridging fault location classified as a double fault in the procedure 89 and the estimated bridging fault location obtained in the procedure 94 are treated the same, and weighting is performed based on frequency.

FIG. 37 shows another embodiment. In the present embodiment, the failure estimation completion determination procedure 95 is added to the weighting sequence of FIG. 25. Since the failure estimation can be stopped when the weighting accuracy by frequency is obtained by this function, the calculation processing time can be reduced. Optimization can be achieved.

First, a failure output terminal is extracted from the failure output result of the tester, and a combination circuit is dynamically extracted in a combination circuit extraction procedure 71 using the terminal as a starting point. Step 7
2 is a procedure for obtaining an input terminal state from an output terminal state for each combinational circuit obtained in step 71, and step 73 registers an estimated fault propagation path inside the combinational circuit obtained when obtaining the input terminal state of the combinational circuit. I do. In order to go back through the circuits until the estimated state cannot be obtained in steps 71 to 73, it is controlled whether or not to continue the estimation for each combinational circuit in the circuit backward completion determination step 74. At this time, if another unestimated fault output terminal exists in the same test vector, the above procedure is repeated in the same manner.

When the fault estimation for each vector is completed in step 73, branching and connection recognition of the path are performed based on the estimated fault propagation path for each combinational circuit obtained in step 72, and the fault output terminal The fault propagation path up to is reconfigured.
By this operation, it is possible to classify a fault candidate point on the fault propagation path as a single fault point, a multiple fault point, or a fault propagation path instead of a fault candidate. In step 76, the failure mode is classified using the result obtained in step 75, and further using the estimated state value at that time and the state values of all vectors used in the estimation up to that time. In step 77, the failure candidates are comprehensively determined and ranked using the frequency of the number of failure output terminals, the number of failure propagation paths, etc., and the failure candidate locations, state values, and failure modes that further satisfy the failure state, and weighted as the failure candidate accuracy I do.

Further, in step 95, the frequency distribution of the number of fault output terminals and the number of fault propagation paths used as weights in step 77 is checked, and compared with a set threshold value. Then, the procedure returns to the procedure 71 using the new vector of the test result, and performs circuit traversal for each combinational circuit. (2) If the error is equal to or larger than the threshold value, it is determined that the fault candidate estimation accuracy is obtained, and the fault estimation is completed. A failure estimation termination determination by frequency determination is provided.

FIGS. 38 to 40 are obtained by inserting a failure estimation end determination procedure 95 into each of the above embodiments. Specifically, FIG. 38 shows the failure estimation termination determination procedure of step 95 in the frequency calculation procedure of FIG. 31, FIG. 39 shows the failure estimation termination determination procedure of procedure 95 in the weighting sequence using the inter-path condition of FIG. In FIG. 40, the failure estimation termination determination procedure of procedure 95 is added to the bridge failure detail determination sequence of FIG. 35, so that the calculation processing time can be shortened optimally. FIG. 41 is a block diagram of an apparatus having the function of FIG.

In this apparatus, first, a netlist to be subjected to fault estimation is input to the netlist management unit 206 and stored therein, and at the same time, the input / output terminals of the LSI and the flip-flops obtained by using the test vector for fault estimation are obtained. Input / output terminal for expected value, flip-flop expected value management unit 2
07 and stored. The estimated value / measured value management unit 209 records the output terminal state value of the test result.

The sequence control unit 205 includes
4, 210 to 213 are controlled, and particularly, queuing control between blocks and operation sequence control are performed. Each of the management units 206 to 209 performs information search, information registration, and deletion from each of the estimation operation function units, and performs unified management of data.

As described above, the net list management unit 206 stores and manages the net connection information of all the target LSI circuits and the connection information of the circuits divided from all the circuits into partial circuits. Then, connection information retrieval of elements and paths and registration / deletion of partial circuits are performed according to the request of each block.

Similarly, as described above, the input / output terminal and the flip-flop expected value management unit 207 also operate as a target LSI.
The expected values of the input / output terminals and internal flip-flops are stored and managed. This block searches and outputs the expected value in response to the expected value search request of the path and the terminal of the element from each block, and registers / deletes / corrects the state value and additionally registers the expected value of the internal element other than the flip-flop. / Delete function.

The failure candidate management unit 208 stores and manages the element names and path names of the candidate points obtained by the failure location estimation, their estimated state values, and their times. This block registers such information in response to a request for registration of a path or a terminal of an element from each block, and performs operations such as output of search results, data deletion, and data correction in response to a search / deletion / correction request. Have a function.

The failure candidate management unit 208 manages the failure candidates, while the estimated value management unit 209 manages the estimated time and its state value. Therefore, regarding the failure state, the failure candidate management unit and the estimated value management unit manage the information in cooperation with each other, and the estimated state management other than the failure state is performed by this block. This block manages the estimated time and status value results regardless of the failure / normal status, and based on the registered results for the time / status value registration / search / deletion / correction requests from each estimated block. Manage and output information.

Next, as for each of the failure estimation function blocks, the combinational circuit extraction unit 201 searches the net list management unit 206 for connection information while the failure candidate management unit 20
8 and a state value search by the estimated value management unit 209 to extract a combinational circuit considered to include the fault propagation path, and register the partial circuit in the netlist management unit.

The combinational circuit state estimating unit 202 obtains the estimated value and the input / output state value of the combinational circuit from the actually measured value managing unit 209 for the obtained combinational circuit, and The internal state is estimated while performing the circuit connection information search for the list management unit, and the state estimation is sequentially performed in the input direction to estimate the state value of the combination circuit input terminal. Then, the estimated value of the input state of the combinational circuit is registered in the estimated value / measured value management unit 209.

At the same time, the fault propagation path extracting section 203 retrieves circuit connection information from the netlist managing section based on the result obtained by the combinational circuit internal state estimating section and extracts a propagation path inside the combinational circuit. . At this time, the expected value of the combinational circuit input / output terminal is obtained from the input / output terminal / flip-flop expected value management unit 207, and the expected value of the internal circuit is obtained, thereby obtaining the estimated fault propagation path. Since this estimated fault propagation path can be a fault candidate, the fault candidate management unit 208 and the estimated time and estimated state value are registered in the estimated value and measured value management unit 209.

The path selector 204 uses the plurality of combined circuit input state estimated values obtained by the combined circuit state estimator 202 to collect fault propagation terminals common to both terminals and state values, and selects a fault propagation path. The common terminals and state values compiled by the path selection unit 204 are registered as a common group in the estimated value / measured value management unit 209 regardless of whether the state value is faulty or normal.

The fault propagation path reconfiguring unit 210 searches the netlist managing unit 206 for circuit connection information based on the fault candidate points registered in the fault candidate managing unit 208 and the estimated state value for each combinational circuit. , Each fault candidate point and L
The fault propagation path to the fault output terminal of the SI is reconfigured, and the path between each fault candidate and the relationship with the fault output terminal are registered in the fault candidate management unit 208.

The estimated state value judging section 211 obtains the obtained fault candidate point, time, and estimated fault state value from the fault candidate managing section 208, and furthermore, an input / output terminal, a flip-flop expected value managing section 207, an estimated value, and an actual measurement value. The value management unit 209 searches the state values of the failure candidate points of all the vectors and determines whether the estimated state value is always constant.

The failure mode classifying section 212 determines a single 0/1 stuck-at fault based on the state judgment obtained by the estimated state judging section.
Classify open faults and bridge faults. At this time, the fault propagation path reconfiguring section 210 searches for a path between each fault candidate registered in the fault management section 208 and a relation with a fault output terminal, and finds an open fault and a bridge fault between two points on the fault propagation path. It is determined whether there is a possibility, and detailed classification is performed.
The classified failure mode for each failure candidate point is registered in the failure candidate management unit 208 together with the state value searched again.

The frequency weighting section 213 is provided for the failure management section 208.
State value, failure mode classification,
Search the route between each fault candidate and the relationship with the fault output terminal,
Based on the frequency of the number of fault output terminals or the number of estimated fault propagation paths, comprehensively determine the likelihood of estimation of each fault candidate point,
Ranking is performed for each failure mode.

FIG. 42 is a block diagram of another apparatus, and the same parts as those in FIG. 41 are denoted by the same reference numerals. Basically, the function is the same as that of the example of FIG. 41, but the netlist management unit for storing the netlist in the previous embodiment, the input / output terminal, and the input / output terminal for storing the expected value of the flip-flop in advance A flip-flop expected value management unit, an estimated value for recording the output terminal state value of the test result, an actually measured value management unit, and a failure management unit for further storing a failure candidate of the estimation result as a server on the network. Have.

Each of the management servers 310 to 313 has a management server 201 to 201.
The information search, information registration, and deletion requests from the clients of the estimation operation function units 20, 210 to 213 are received via the network, the processing is performed based on the requests, and a processing completion notification is sent to each requesting client. Returns and performs unified management of data.

The sequence control unit 205 includes
4, and performs operation control of each estimation operation function of 210 to 213, and particularly controls queuing between blocks and operation sequence control. Combinational circuit extractor 2, which is each state estimator
01, combinational circuit state estimator 202, fault propagation path extractor 203, path selector 204, fault propagation path reconstructor 2
10, estimated state value determination unit 211, failure mode classification unit 21
2. The frequency weighting unit 213 is equivalent to each of the state estimation function units described above.

[0154]

According to the present invention, a combinational circuit is extracted from the output side of an LSI, and a fault propagation value at an input boundary of the combinational circuit extracted for all failures is sequentially estimated and traced back. Therefore, compared to the case where a fault is assumed for all the signal lines of all circuits and a fault simulation is performed using all the vectors, the estimation operation of only the necessary circuits is sufficient, and there is an effect that the operation amount can be greatly reduced. .

Further, since the fault propagation path estimation in each combinational circuit is limited to the extracted combinational circuit, the circuit scale is greatly reduced as compared with the simulation of the entire circuit, and the amount of calculation can be greatly reduced. Since the fault propagation path extraction in the combinational circuit can be performed only by comparing the failure simulation result of the combinational circuit input part for failure estimation with the simulation result in a normal case, the amount of calculation can be reduced and the reconvergence in the combinational circuit can be reduced. The fault propagation path can be extracted for the circuit.

In the propagation path estimation in the combinational circuit according to the present method, estimation is performed by assuming that there is a fault inside the circuit and that there is one or two faults. .

In addition, the fault propagation value at the input boundary of the combinational circuit is estimated for all the fail vectors, and the failure propagation estimation path is further narrowed down to the preceding combinational circuit. In the case of multiple faults, it is possible to assume both independent fault propagation and propagation that influence each other, and it is unlikely that an estimation error occurs. Regarding timing faults, a critical path is extracted from the information obtained as a state fault as a result, and whether or not a timing fault occurs is estimated.

Furthermore, by simplifying and estimating the input state of each of the combinational circuits and by selecting the fault propagation path, the number of paths going back and the combinational circuits to be extracted can be greatly reduced, and the amount of calculation can be reduced.

In addition, it is possible to perform estimation with an optimum sequence even for a device incorporating a scan path or a device incorporating a ROM and a RAM according to the present method, so that the time required for failure can be greatly reduced. effective.

Further, a combinational circuit which is dynamically estimated to include a fault propagation path is sequentially extracted from an actual fault output terminal, a state is estimated, and a path is selected to estimate an LSI fault location. The time required for the estimation can be greatly reduced as compared with the method using other failure dictionaries. It is fully expected that this will become more significant as the scale of LSIs increases.

The reason is that, in the case of the conventional method using the fault dictionary, a fault dictionary is created by making a fault simulation for all the expected nodes and performing a fault simulation.
When the SI becomes large in scale, the operation time is greatly increased. On the other hand, in the present method, since the estimation operation is performed on the combinational circuit including the path estimated to have propagated the fault, the operation target of the simulation is the SI operation. It is considered that the circuit is not a whole but a partial combinational circuit, and the operation time does not increase significantly as in the failure dictionary method even when the scale of the LSI is increased.

Further, since it depends on the circuit configuration and the logic depth from the output terminal of the fault location, a fault at a location with a relatively shallow logic depth can greatly reduce the estimation time. First
Although the effect of (1) described that the calculation time was shortened by the estimation sequence, further reduction can be achieved. In addition, if it is desired to increase the fault accuracy, the number of test vectors used for the estimation can be increased and the trade-off between the calculation time and the estimation accuracy can be controlled.

The reason is that, in the case of a fault dictionary, the location of a fault cannot be narrowed down unless a relationship between all fault assumption points and its outputs is prepared as a search dictionary. However, in this method, a circuit is dynamically extracted from a fault output terminal. Since it goes back in the input direction, if the location of the fault location is shallow from the output terminal, the location of the fault location is quick, so that the fault location can be narrowed down in a relatively short time regardless of the scale of the LSI.

Furthermore, if it is desired to increase the fault accuracy, it is possible to increase the number of test vectors used for the estimation, so that a trade-off between the calculation time and the estimation accuracy can be controlled.

The reason is that failure estimation is performed by sequentially reading test results, and frequency numbers are ranked and weighted using only the obtained failure estimation results.
This is because, if it is desired to improve the estimation accuracy, it is possible to easily read the test result until the required level is reached and stop the estimation when the result is satisfied.

Further, unlike the fault dictionary method, faults are assumed up to double faults, which are bridge faults, so that a single stuck-at fault, open fault, and bridge fault can be estimated with relatively high accuracy.

The reason is that an estimated fault propagation path is determined assuming a double fault, and a double fault is assumed on the obtained path, so that faults can be assumed within a narrower range and accuracy is improved. It is. In addition, since the estimated fault propagation path obtained for each combinational circuit is reconstructed and subdivided and associated from the fault candidate point to the fault output terminal, the weight of the node near the output terminal can be prevented from becoming high, and each fault mode can be avoided. This is because accuracy is improved. On the other hand, in the case of a fault dictionary, if the fault simulation is extended to a double fault, it is necessary to make a fault assumption with a combination of numbers close to the square of the number of nets, and the number of combinations increases explosively. Is not realistic in terms of processing time.

[Brief description of the drawings]

FIG. 1 is a flowchart showing a first embodiment of the present invention.

2A and 2B are diagrams in which a fault propagation path inside a combinational circuit is estimated, wherein FIG. 2A shows a case where no fault exists inside the combinational circuit, FIG. 2B shows a case where a fault exists inside the combinational circuit, and FIG. () Is a diagram when two failures exist inside the combinational circuit.

FIG. 3 is a modification of the flowchart of FIG. 1;

FIG. 4 is a diagram showing a simplified operation at the time of estimating a combination circuit input terminal state.

FIG. 5 is a flowchart showing a fault propagation path output sequence (comparison for each node state estimation) when it is assumed that there is no fault location in the combinational circuit.

FIG. 6 is a flowchart showing a fault propagation path output sequence (comparison after estimating the state of the input terminal of the combinational circuit) when it is assumed that there is no failure point in the combinational circuit.

FIG. 7 is a flowchart showing a fault propagation path output sequence (comparison for each node state estimation) when it is assumed that a fault location exists in the combinational circuit.

FIG. 8 is a flowchart showing a fault propagation path output sequence (comparison with a simulation result after estimating the state of the input terminals of the combinational circuit) when it is assumed that a failure point exists in the combinational circuit.

FIGS. 9A and 9B are diagrams showing extraction of a fault propagation path in a combinational circuit, and FIG. 9A is a diagram for finding a fault propagation path when a fault location is assumed to exist in the combinational circuit (comparison for each node state estimation); FIG. 6B is a diagram for obtaining a fault propagation path when a fault location is assumed to exist in the combinational circuit (comparison with a simulation result after estimating the input terminal state of the combinational circuit).

FIG. 10 is a flowchart showing a second embodiment of the present invention.

FIG. 11 is a diagram showing a route selection method for estimating the input state of a combinational circuit, in which all vectors are combined for each fail terminal.

FIG. 12 is a diagram showing a case where all vectors are similarly grouped for each fail terminal and the order of appearance of the fail is selected.

FIG. 13 is a diagram showing a case in which all vectors are similarly grouped for each fail terminal, and the vectors having failed are selected in descending order.

FIG. 14 is a diagram showing a case in which all vectors are combined for each fail terminal and a plurality of terminals are selected in descending order of the number of failed vectors.

FIG. 15 is a table showing the relationship between the estimated number of fault locations in the combination circuit and the failure type in the estimation result for each combination circuit.

FIG. 16 is a table showing a relationship between a failure mode, a failure location, and a state.

FIG. 17 is a diagram for extracting candidate fault points by superposition.

FIG. 18 is a diagram illustrating calculation of a failure candidate point appearance frequency.

FIG. 19 is an explanatory diagram of a case where a latch is included in an estimated failure candidate location.

FIG. 20 is a flowchart for determining a timing failure.

FIG. 21 is an explanatory diagram of a case where a scan FF is included in a circuit.

FIG. 22 is a diagram showing a backward direction when a memory cell is reached.

FIG. 23 is a flowchart of memory cell retrace.

FIG. 24 is a modification of the flowchart of FIG. 10;

FIG. 25 is a schematic flowchart of weighting according to another embodiment of the present invention;

FIG. 26 is a detailed flowchart of a fault location estimating method according to the embodiment of the present invention.

FIG. 27 is a detailed flowchart of the weighting sequence of the frequency of the number of failure output terminals in FIG. 25;

28A shows a state of fault propagation when there is no fault in the combinational circuit, FIG. 28B shows a state of fault propagation when there is a single fault in the combinational circuit, and FIG. It is a figure which shows each failure propagation situation when there is a bridge failure.

FIG. 29 is a diagram illustrating route selection.

FIG. 30 is an explanatory diagram for weighting the frequency of failure output terminals.

FIG. 31 is a detailed flowchart of a weighting sequence for the frequency of the number of fault propagation paths.

FIG. 32 is an explanatory diagram of weighting the frequency of the number of fault propagation paths.

FIG. 33 is a detailed flowchart of a weighting sequence using an inter-path condition.

FIG. 34 is an explanatory diagram of weighting using an inter-path condition.

FIG. 35 is a detailed flowchart of a detailed judgment weighting sequence of a bridging fault.

FIG. 36 is an explanatory diagram of a failure candidate of a detailed determination weighting of a bridge failure.

FIG. 37 is a weighting flowchart for failure estimation completion determination by frequency determination.

FIG. 38 is a detailed flowchart of a weighting sequence based on frequency determination.

FIG. 39 is a detailed flowchart of a weighting sequence using an inter-path condition based on frequency determination.

FIG. 40 is a detailed flowchart of a bridge failure detailed determination weighting sequence based on frequency determination.

FIG. 41 is a functional block diagram showing an embodiment of a failure point estimation device using the present weighting method.

FIG. 42 is a functional block diagram showing a client / server type fault location estimating apparatus embodiment using the weighting method.

[Explanation of symbols]

REFERENCE SIGNS LIST 100 combinational circuit 101 input latch 102 output latch 201 combinational circuit extraction unit 202 combinational circuit state estimation unit 203 fault propagation path extraction unit 204 path selection unit 205 sequence control unit 206 netlist management unit 207 input / output terminal, flip-flop expected value management unit 208 Fault candidate management unit 209 Estimated value, measured value management unit 210 Fault propagation path reconstruction unit 211 Estimated state value judgment unit 212 Failure mode classification unit 213 Frequency weighting unit 310 Netlist management server 311 Input / output terminal, expected flip-flop value management Server 312 Failure candidate management server 313 Estimated value, measured value management server

Claims (29)

(57) [Claims] 1. An LSI which uses all expected value information for all test vectors of all flip-flops in a previously prepared LSI, pass / fail output information of an actual tester, and connection information of all circuits. A fault location estimation method for a sequential circuit in which the circuit is divided into the flip-flop and the combinational circuit to estimate a failure, wherein a failure propagation estimation value at an input boundary of the combinational circuit is determined for each fail vector. The combinational circuit is extracted in the input direction from the actual fail output pin or the flip-flop input line which is presumed to have failed to the input pin or the flip-flop output of the LSI, and further from the input of the extracted combinational circuit to the output direction. The circuit is extracted up to the output pin or flip-flop input of the LSI, and the output pin or the Using the procedure for extracting the combinational circuit from the input terminal of the flip-flop in the input direction and the estimated failure output at the output unit of the combinational circuit and the state value of the estimated normal output, (1) a failure in the combinational circuit When it is assumed that there is no fault, (2) When it is assumed that a fault exists in the combinational circuit, the estimation is performed. When it is estimated that no fault exists in the combinational circuit, the signal propagates from the preceding stage to the input terminal of the combinational circuit. The state value of the signal is estimated, and when it is estimated that a fault exists inside the combinational circuit, the signal is propagated from the preceding stage to the estimated fault location in the combinational circuit and the extracted combinational circuit input terminal in the same state. If it can be estimated that the fault has occurred, the procedure for estimating the state value of the combinational circuit input terminal signal is also used. A procedure for obtaining, a procedure for simplifying the Boolean algebra using all of the plurality of estimated input vectors at the obtained combination circuit input terminal, and a fault propagation by comparing the simplified estimated input vector with an expected value. A procedure for classifying each terminal into status values of terminals, normal terminals, and terminals that are not involved in output state determination, and obtaining a state at that time for each terminal, and when a fault propagation estimation terminal is present in the obtained estimation result, A procedure for re-executing from the first combinational circuit extraction procedure, and a procedure for determining the coincidence of all the estimated fault locations in the combinational circuit obtained by the estimation for each combinational circuit and extracting the same location satisfying all the states. A fault location estimation method for a sequential circuit, wherein a final failure estimation list is created. 2. A method according to claim 1, wherein in place of the procedure for obtaining the estimated fault location and the estimated input vector of the combination circuit input terminal, it is assumed that no fault exists in the combination circuit.
The same procedure as in claim 1 is performed using the state values of the estimated fault output and the estimated normal output at the output section of the combinational circuit, and when it is assumed that a fault exists in the combinational circuit, the combinational circuit output section The procedure for estimating the state of the input terminal of the combinational circuit using only the state value of the estimated failure terminal in step (a), and performing the logic simulation for each estimation vector using each obtained estimation vector A procedure for obtaining a state value for each estimating vector, performing a logic simulation by giving an expected value to the input terminal of the combinational circuit, and comparing the obtained node state value for each vector with a fault for each estimated vector. A procedure for extracting a propagation path, a path in which a fault has propagated to the obtained fault propagation path other than the estimated failure output terminal of the combinational circuit output unit used first, and the path and the estimated failure output If the terminal intersects with the terminal, all fault propagation paths from the intersection to the combinational circuit input terminal are extracted and deleted from the obtained fault propagation path to determine the fault location and the estimated input vector. 2. A method according to claim 1, further comprising the step of obtaining an estimated input vector of the sequential circuit, if any. 3. A procedure for determining the estimated input vector, wherein (1), by following the operation of estimating the combinational circuit inside the failure point, similar combinations within the circuit to obtain the combination circuit output state (2) A procedure for summing the number of fault locations already obtained by the fault estimation operation and the number newly obtained by fault location estimation in the combinational circuit; and a stuck-at fault based on the sum of the estimated fault locations obtained by the summation. Open faults, short-circuit faults, timing faults, etc. are classified by assumed fault type, and by comparing with the estimated number of fault locations specified for each fault type, a new estimated fault location exceeding the estimated fault number specified And simultaneously deleting the input estimation vector of the combinational circuit input terminal obtained in the same state. Fault location estimation method of the sequential circuit according to claim 1, characterized in that so as to reduce the number of circuit input estimation vector. 4. The method according to claim 1, wherein said simplified input vector is an expected value.
Fault propagation terminal, normal terminal, output state determined by comparison with
Are classified into each state value of the terminal that is not related to
Instead of the procedure for determining the state for each terminal, a fault propagation input terminal is extracted by comparing the estimated value of each combination vector with the expected value of the combination circuit input terminal using all of the plurality of estimated input vectors at the combination circuit input terminal. A step of classifying all the vectors of the estimated combinational circuit input terminals for each of the extracted fault propagation terminals, and examining and classifying all the estimation vectors of the combinational circuit input terminals classified for each of the failure propagation terminals. The states of the terminals that correspond to all the estimated vectors obtained are as follows: (1) the state that is in agreement with the expected value and its state value; and (2) the state that is different from the expected value and its state. If there is a vector whose state matches the expected value and a vector that does not match the expected value in the classified estimated vectors, (3) the fault may have propagated Expresses sex and its state Reduce the estimated number of vectors of the combinational circuit input terminal by expressing the combinational circuit input terminal as a fault propagation terminal, a normal terminal,
2. A method according to claim 1, further comprising the step of classifying the terminal into a terminal not involved in determining the output state. 5. The method according to claim 1, wherein said simplified input vector is an expected value.
Fault propagation terminal, normal terminal, output state determined by comparison with
Are classified into each state value of the terminal that is not related to
In the procedure for determining the state for each terminal, the procedure for extracting the fault propagation input terminal by comparing the expected value of the combinational circuit input terminal for each estimation vector using all of the plurality of estimated input vectors at the combinational circuit input terminal later, using all the estimated input vector obtained in claim 1, wherein the fault propagation and counting procedure for counting a frequency of occurrence of a failure for each input terminal, only base Kuta order that terminal from more frequency of occurrence pin is a failure Is extracted, and the state of the terminal that matches from all of the extracted estimation vectors is represented by (1) a state that is in agreement with the expected value and its state value, and (2) different from the expected value. If there is a vector that represents a state and its state value, and each of the classified estimating vectors does not match a vector whose state matches the expected value, (3) the fault propagates Possible The expressed and represented in the state value, reducing the estimated number vector combination circuit input terminal, fault propagation terminal and a normal terminal of the combination circuit input terminal,
A classification procedure for classifying the input state into terminals that are not involved in the output state determination; and a procedure for repeating the counting procedure and the classification procedure until all the estimated input vectors are covered. 5. The method according to claim 4, wherein the number of circuit input estimation vectors is reduced. Wherein said combining circuit input terminal fault propagation terminals, normally the terminal, in the procedure for classifying the terminal that is not involved in the output state determination, the failure for each of the estimated input vector failure propagation input terminals obtained with all after step for counting the occurrence frequency, extracts the fault propagation terminal based on certain predetermined threshold value from the obtained frequency, the extracted plurality of fault propagation estimation terminal is estimated to be all faults an extraction procedure for extracting the Kuta, the state of the pins that match over from the extracted base Kuta, (1) the expected value and - represents a match with that state and its state value is different from the (2) the expected value (3) If there is a vector whose state matches the expected value and a vector that does not match the expected value in the classified estimated vectors, Propagating Possibility represents and reduce the estimated number vector combination circuit input terminal expressed in that state values, fault propagation terminal combination circuit input terminal, a normal terminal,
A classification procedure to classify the terminal that is not involved in the output state determination, and a procedure is repeated until the extracting order and the classification procedure to cover all the estimated input vector, combined obtained for each input terminal state estimation for each combination circuit 6. The method according to claim 5, wherein the number of circuit input estimation vectors is reduced. Wherein said combining circuit input terminal fault propagation terminal and normal terminal, in the procedure for classifying the terminal that is not involved in the output state determination, the use of all estimated input vector, resulting failure propagation input terminals every failure After the procedure of counting the appearance frequency, the fault propagation terminals are extracted in descending order of the obtained appearance frequency using a certain number of terminals as a threshold, and it is estimated that all of the extracted fault propagation estimation terminals are faulty. An extraction procedure for extracting the extracted vector, and a state of the terminal corresponding to the extracted vector, (1) representing a state corresponding to the expected value and its state value, and (2) different from the expected value. (3) If there is a vector whose state matches the expected value and a vector that does not match the expected value in the classified estimated vector, Biography And represents a possibility has been and reducing the estimated base number Kuta combination circuit input terminal represented by its state value, the combinational circuit failure propagation pin an input terminal, a normal terminal,
A classification procedure to classify the terminal that is not involved in the output state determination, and a procedure is repeated until the extraction procedure and the classification procedure to cover all the estimated input vector, combined obtained for each input terminal state estimation for each combination circuit 6. The method according to claim 5, wherein the number of estimated circuit input vectors is reduced. 8. A fault propagation terminal is extracted based on a certain threshold value based on the occurrence frequency according to claim 6, and a vector in which all of the extracted fault propagation estimation terminals are estimated to be faulty is extracted. a first extraction procedure to the certain fixed number of terminals to extract the fault propagation terminal of the frequency of occurrence in decreasing order as a threshold, base Kuta plurality of failure propagation estimation terminal that is the extracted was estimated to be all faults Second extraction hand to extract
The extraction of the estimated vector and the reduction of the estimated vector are repeated by the same procedure as the procedure of claim 6 for the terminal whose occurrence frequency for each fault propagation estimation terminal is equal to or more than a predetermined threshold value. If the frequency of occurrence of each failure propagation estimation terminal by extraction becomes smaller than the threshold value, the number of fixed terminal, wherein there is extracted the fault propagation terminal of the frequency of occurrence in decreasing order as a threshold, a plurality of which are the extracted 7. The method according to claim 6, further comprising the step of extracting and reducing vectors whose fault propagation estimation terminals are all assumed to be faulty, and repeating the process until all fault vectors are covered. Method. 9. The combination circuit input terminal obtained in claim 1, wherein the fault propagation terminal and the normal terminal are classified into the terminals not involved in determining the output state, and the state at that time is determined for each terminal. all of the plurality of estimated input vector after the procedure for simplification of Boolean algebra using in further claims
5, the counting procedure, the classification procedure, and the repeating procedure
Or the extraction procedure, the classification procedure and
The repeating procedure, or the extracting procedure of claim 7,
9. The method of claim 8, wherein the classifying step and the repeating step are performed.
The first extraction procedure, the second extraction procedure and the
9. The method according to claim 5 , further comprising the step of returning a fault. 10. Every time a combinational circuit is extracted in a procedure for finding a fault estimation location and an estimated input vector satisfying the output state of the combinational circuit, an expected value is used for an obtained combinational circuit input terminal, (1) In the combinational circuit, using the procedure for performing the logic simulation inside the circuit to obtain the expected value of each node in the combinational circuit and the state value of the estimated fault output and the estimated normal output at the output section of the combinational circuit, (2) When it is assumed that a fault exists in the combinational circuit, estimation is performed for each node sequentially from the output terminal to the input direction in each case, and the fault in the combinational circuit and the input terminal When estimating the state, every time the estimated value is obtained, the estimated value is checked against the expected value of each node in the combinational circuit. The fault propagation path from the output terminal to the input estimated state value obtained is extracted, and in the case of (2), the estimated fault estimated location obtained from the estimated fault output terminal and the estimated fault propagation path to the input estimated state value are extracted. Lottery
And a procedure for managing the estimated fault estimated location and the estimated fault propagation path obtained as a logical sum and the estimated fault estimated location and the estimated fault propagation path obtained as a logical product. After repeating the steps,
A step of classifying all the estimated fault locations in the combination circuit and all the estimated fault propagation paths obtained by the estimation for each combination circuit in the same time unit, and a step of classifying the obtained estimated fault location and the estimated fault propagation in the same time unit 2. A method according to claim 1, further comprising the step of: determining a locational match using a route and extracting the same location satisfying all states. Each combination circuit is extracted 11. A procedure for determining the estimated failure site and estimated input vector meet the output state of the combinational circuit, using the expected value for the obtained combination circuit input terminal, previously combined (1) In the combinational circuit, using the procedure for performing the logic simulation inside the circuit to obtain the expected value of each node in the combinational circuit and the state value of the estimated fault output and the estimated normal output at the output section of the combinational circuit, (2) When it is assumed that a fault exists in the combinational circuit, estimation is performed for each node sequentially from the output terminal to the input direction in each case, and the fault in the combinational circuit and the input terminal When estimating the state, a procedure of estimating the state by using a fixed value fixed to 0 and 1 and the state X which may be any of the states of 1 and 0 and simultaneously performing simplification, When performing matching with the expected value of each node in the obtained combinational circuit, the fixed values 0 and 1 are not treated as the matching judgment with the expected value or as the estimated fault propagation path.
Only in the case of (1), the estimated fault propagation path up to the input estimated state value obtained from the estimated fault output terminal is extracted, and in the case of (2), the fault propagation path is obtained from the estimated fault output terminal. Extraction procedure for extracting the estimated fault location and the estimated fault propagation path up to the input estimated state value, the fault estimated location obtained as a logical sum, the fault estimated location obtained as a logical product of the estimated fault propagation path, and the estimated fault propagation path And a procedure for calculating an expected value of each node in the combinational circuit.
The extraction procedure, the simplification procedure, and the management
After repeated the procedure, a procedure for classifying all combinations circuit estimating fault location and all fault propagation path obtained by the estimation of each said combination circuit in the same time unit, resulting in estimated for each of the combinational circuit 2. The same location extracting procedure according to claim 1, further comprising: determining a locational match using all the estimated fault locations in the combinational circuit and all the estimated fault propagation paths and satisfying all states. A method for estimating the failure location of a sequential circuit. 12. After the procedure for simplifying the Boolean algebra, a logic simulation of the combinational circuit is performed by using the obtained state value of the estimated fault location and the reduced input estimation vector. A procedure for obtaining the expected value of each node in the combinational circuit by performing a logic simulation using the normal state of the input terminal of the combinational circuit and assuming that no fault exists in the combinational circuit; An extraction procedure for extracting a node different from the expected value by comparing each node state based on the obtained estimated state with the obtained expected value and extracting an estimated fault propagation path. wherein, a management procedure for managing the estimated failure site and estimating fault propagation path is obtained as the estimated failure site and estimating fault propagation path and the logical product is obtained as a logical sum, each node-like Procedure for obtaining the said expected value of each node
, The extraction procedure, and the management procedure are repeatedly performed, and then all the estimated fault locations in the combination circuit and all the fault propagation paths obtained by the estimation for each combination circuit are classified into the same time unit. 2. The method according to claim 1, further comprising: comparing the estimated fault location classified at each same time with the estimated fault propagation path and extracting the same location satisfying all states. A method for estimating the failure location of a sequential circuit. 13. The method of claim 10 or claim 11, after the extraction procedure definitive to claim 12, all combinations circuit estimating fault location and all fault propagation path obtained by the estimation of each said combinational circuit The estimated fault location and the estimated fault propagation path obtained as a logical sum are classified as the same group, and the estimated fault location and the estimated fault propagation path obtained as a logical product are managed. Includes a procedure of obtaining the frequency of occurrence of the estimation result for each node of each circuit by superimposing the estimated fault location in the combinational circuit and all fault propagation paths, and weighting the fault estimated locations in descending order of frequency. The fault location estimation method for a sequential circuit according to any one of claims 10, 11, and 12. 14. A combination circuit for each combination circuit.
Estimated fault in the circuit and the estimated fault at the input terminal of the combinational circuit
After obtaining the propagation path, stuck-at faults for each combination circuit, open faults are classified as short-circuit failure or the like, the failure at each step from the obtained estimated fault location and combining circuit input estimate vector according to claim 10 or claim 11 A procedure for obtaining a propagation path, and repeating the estimation for each of the combinational circuits to obtain an estimated failure location in all the combinational circuits and all the failure propagation paths,
The estimated fault location and the estimated fault propagation path obtained as a logical sum are classified as the same group , and the estimated fault location and the estimated fault propagation path obtained as a logical product are managed. Estimating the occurrence frequency of the estimation result for each node of each circuit by superimposing the estimated fault location in the combinational circuit and all the fault propagation paths, and weighting the fault estimated locations in descending order of frequency. The method for estimating a fault location in a sequential circuit according to claim 10 or 11, wherein: 15. The combination circuit according to claim 14,
Is repeated for all the estimated fault locations in the combinational circuit.
After obtaining all the fault propagation paths , the fault is classified by the number of faults for each assumed fault such as stuck-at fault, open fault, short fault, timing fault, etc. Using the estimated fault location obtained and the obtained estimated fault propagation path, a procedure for obtaining the frequency of occurrence of the estimation result for each node, and a procedure for weighting the obtained frequency of occurrence and prioritizing the likelihood of failure And (1) a stuck-at fault if all are the same fault state values, and (2) a single fault location and each fault estimated state value If they are different, an open fault or a short fault with a node that is always in a normal state. (3) There are two estimated fault locations, and the state estimated value is normal and the other is faulty. Short-circuit failure if the state value is consistent, (4) if the estimated fault location comprises a flip-flop to flip-flop or equivalent fault location, if it has a case and a fault condition in which the estimated state value has a normal state Is Thailand
15. The method for estimating a fault location of a sequential circuit according to claim 14, further comprising a detailed classification procedure for performing a detailed classification of a mining fault . 16. A detailed classification procedure, in addition to said criteria timing failure (4), by obtaining the estimated fault propagation path from net connected to the flip-flop input with a condition that is estimated failure and 16. The fault location of a sequential circuit according to claim 15, comprising: a procedure for obtaining a critical path that causes a timing failure; and a procedure for determining whether the critical paths extracted as the timing failure are the same path. Estimation method. 17. The procedure for performing detailed classification from the fault estimated state value of each of the prioritized nodes according to claim 15, further comprising: A procedure for obtaining a critical path that becomes a timing failure by obtaining a fault propagation path; a logic simulation for the critical path extracted as the timing failure, and a procedure for calculating a propagation delay time of the critical path; Calculating the delay time in which the propagation delay time of the critical path has the swing of the device characteristic, and the obtained timing failure causes the propagation of all the critical paths longer than the obtained propagation delay time to fail. And determining whether or not the fault has occurred in the sequential circuit. Estimation method. 18. After all the steps of any one of claims 1 to 17, a fault simulation is performed on the obtained estimated fault location by using all test vectors given based on the estimated fault state. The method according to any one of claims 1 to 17, further comprising a step of determining whether all of the estimated failure locations and failure causes match. 19. The combination circuit extraction, after the steps of the combination circuit estimating, examining the expected value of the flip-flop clock input at that time, determines whether or not the output state of the Furippufurotsupu is updated at that time Based on the result of the update decision of the flip-flop, (1) If the time has been updated (test vector), use the estimated combinational circuit input terminal state value to further advance the combinational circuit (2) if not updated, find the last updated time before that time, keep the estimated result obtained up to the last updated time, and maintain the estimated result and AND state When the estimation result in the above is inconsistent with the estimation result, the procedure for deleting the estimation time after the estimation time (where the estimation time is the last updated time in the case of a flip-flop Fault location estimation method of the sequential circuit in accordance with claim 18, characterized in that it comprises a a means), a. 20. A procedure for previously searching a circuit connection information or a test vector to determine whether or not a scan path is included in all circuits, and determining whether or not a scan path itself has failed based on the scan path existence determination result. (1) a procedure for extracting a combinational circuit except for a clock line and a scan path line at the time of extracting a combinational circuit, tracking the path in the extracted combinational circuit, and estimating a fault; A procedure for extracting a combinational circuit including a line and a scan path line, tracking and also estimating a fault on the clock line and the scan path line, and (3) obtaining the estimation results of both the above (1) and (2). Performing unnecessary tracking of clock lines and scan path lines, including a step of comparing the scan path itself to determine whether or not there is a failure. 20. The method for estimating a fault location of a sequential circuit according to claim 1, wherein 21. A procedure for extracting a combinational circuit according to any one of claims 1 to 20, wherein R is a combinational circuit input terminal.
When the output terminals of the memory blocks such as the OM and the RAM are extracted, (1) When the output terminals of the extracted memory are the data lines, the combination circuit is formed so as to cover all the data lines of the same memory block. 21. A procedure for extracting by dividing into one or a plurality of pieces, and (2) a procedure for performing a normal combinational circuit extraction according to any one of claims 1 to 20 when the extracted output terminal of the memory is not a data line; A) a process of estimating a data line by estimating an input terminal state by a combinational circuit extracting a data line of a memory block in the step (b); and supplying the data to the memory at the same time from the obtained estimation result of the data multiple output line of the memory block. a procedure for determining whether the status value of the address lines is wrong that, from the determination result, correct (3) address lines If it can be estimated that,
A procedure for determining the bit failure in the address during writing, the procedure of estimating the (4) when it can be estimated that the address lines is incorrect, the supply address from the estimated results of the data line, (5) the (3 21. The method according to any one of claims 1 to 20, further comprising a procedure for performing both the procedure ( 3 ) and the procedure ( 4 ) if it is determined that both of the defects ( 4 ) and ( 4 ) can exist. Method for estimating the fault location of a sequential circuit. 22. A test result obtained by using all expected value information for all vectors of all flip-flops in an LSI prepared in advance, pass / fail output information of an actual tester, and connection information of all circuits. And dynamically extracting a combinational circuit sequentially from each fault output terminal of each vector, and estimating a combinational circuit input state value from the output state value or the estimated value of the extracted combinational circuit to thereby obtain each flip-flop in the LSI. The procedure for estimating the state value of the loop, extracting the terminal, element and its state value that are estimated to have propagated the fault from the difference from the expected value, and going back to the test vector from the fault propagation estimation terminal obtained in this way, A procedure in which the above operations are sequentially repeated, such as extracting a combinational circuit in the input direction, and a similar operation is performed on all test vectors, and A fault propagation estimating element and a procedure for obtaining a state value of the fault propagation estimating element, and a method for weighting the candidate, the circuit connection information being searched using the fault propagation estimating element and the state value in each of the combinational circuits. The procedure for reconstructing the connection relation as the fault propagation path to the output terminal and the state value of each estimated fault propagation point on the estimated fault propagation path are checked. (2) Select a node whose estimated state value changes with time as an estimated fault candidate point; (3) Select a node that requires two estimated fault locations at the same time as an estimated fault candidate point A procedure for counting the number of fault output terminals of the LSI reaching each selected estimated fault candidate point as a frequency, and faults in descending order of the counted frequency. A method of rearranging the order of appearance of the candidate points and weighting the candidate points. 23. Instead of the step of counting the number of fault output terminals of the LSI reaching each of the selected estimated fault candidate points as a frequency, each of the selected estimated fault candidate points and the LSI
23. The method according to claim 22, further comprising a step of counting the number of paths existing between the failure candidate terminal and the failure output terminal as a frequency, rearranging the appearance order of the failure candidate points in descending order of the counted frequency number, and performing weighting. Of weighting the estimated failure location. 24. After the procedure for retrieving circuit connection information and reconstructing a connection relation as a fault propagation path to an output terminal using the obtained fault propagation estimation elements and state values in each of the combinational circuits, (1) Expectation values at the same time are different and different at the same time, by examining the relation between the subdivided paths and examining whether or not to assume a short-circuit fault. Select as a short-circuit fault candidate between two points on the subdivision path; (2) For each fault candidate point on the path that branches into two paths and connects to one path, an arbitrary one on a different subdivision path (3) A candidate point on each of the estimated fault propagation paths that branches into three or more paths at the same time and is connected to one path is not selected as a short-circuit fault; Fault propagation established for each fault propagation path A procedure for largely classifying a failure mode for each of the estimated number of fault locations, and an estimated state value and a fault state of all vectors with respect to the assumed short-circuit fault location between two points. Determine estimates are always consistent, the procedure for recalculating the failure modes in detail, reorder the appearance rank fault candidate point in descending order of the number of failed output frequency of the said counting, and a procedure of weighting 23. The method according to claim 22, wherein the weighting of the estimated fault location is performed. 25. After all the procedures of claim 24 , the weighted frequency distribution is further checked and compared with a set threshold value. (1) If the frequency distribution is less than the threshold value, a new vector Back to the combinational circuit extraction procedure from the failure output terminal; (2) in the case of more than the threshold value, and terminates the estimated accuracy was obtained estimated failure; claim 2 having a failure estimating end determination procedure by frequency determination
4. A method for weighting a failure estimation portion described in 4 . 26. After the procedure of classifying the failure mode into large numbers for each of the estimated number of failure points, a procedure of calculating the frequency of the number of failure output terminals for each of the failure candidates in all the vectors; Inspection of the failure mode and recalculation of the frequency for the two points above to determine the bridge failure in detail, and rearrange the order of appearance of the failure candidate points in descending order of the counted number of failure output terminals 25. The method for weighting a failure estimation portion according to claim 24, further comprising: performing weighting. 27. After all the procedures of claim 26 , the weighted frequency distribution is further compared with a set threshold value which has been checked. (1) If the frequency distribution is less than the threshold value, a new vector Returning to the combinational circuit extraction procedure from the failure output terminal; (2) if the error is equal to or greater than the threshold value, it is determined that the estimation accuracy has been obtained; and the failure estimation is terminated;
26. A method of weighting a failure estimation point for each failure mode, comprising the method of weighting a failure estimation point described in 26 . 28. A net list of an LSI to be estimated is stored, and a net list divided into partial circuits can be registered together, a search and a deletion request are processed, and a unit of the net list is processed. A netlist management unit for managing, an input / output terminal / flip-flop expected value management unit for managing expected values of input / output terminals and flip-flops, and an estimated / actual value management unit for storing estimated values and measured values A failure candidate management unit that stores the failure candidate points, estimated failure state values, and failure mode and other estimation results obtained from each failure location estimation function; and dynamically extracts a combinational circuit from a failure output terminal or a failure estimation output terminal. A combinational circuit extractor, a combinational circuit state estimator for estimating a state value inside the combinational circuit, and an estimated fault propagation path from the state estimation result inside the combinational circuit. A fault propagation path extracting unit for extracting a path, a path selecting unit for selecting a fault propagation path using the obtained fault propagation terminal and the state value of the combinational circuit input terminal, and a reconfiguration of the obtained fault propagation path. The fault propagation path reconfiguration unit that performs the association from the fault location to the fault output terminal and checks the obtained estimated state of each fault candidate for all the vectors to determine whether the state value is constant or changed. An estimated state value determination unit for determining the number of failure points based on the result obtained in the state value determination, and a failure mode classification unit for performing failure mode classification based on the number of failure locations; and a failure output terminal number for each failure candidate point. Alternatively, a frequency weighting unit that counts the frequency of the number of fault propagation paths and weights each fault mode, and an overall estimation sheet such as control of the processing order of the fault estimation functions and queuing control between the fault estimation functions. Device for weighting the estimated failure site for each failure mode, characterized in that it comprises a sequence controller for controlling the Nsu. 29. A server comprising a netlist management unit, an input / output terminal, a flip-flop expected value management unit, an estimated / measured value management unit, and a failure candidate management unit as a server on a network. From the client of the propagation path extractor, fault propagation path extractor, path selector, fault propagation path reconstructor, estimated state value determiner, failure mode classifier, frequency weighter, and sequence controller that controls the entire estimated sequence The registration request and before
29. The apparatus according to claim 28, wherein the apparatus performs a process for a request for search, deletion, etc., and returns a process completion notification to the client.