Method of design for testability, test sequence generation method and semiconductor integrated circuit

Abstract

Flip-flops (FFs) to replace with scan FFs are selected for an integrated circuit designed at the gate level in order that the integrated circuit has an n-fold line-up structure. All FFs in an integrated circuit are temporarily selected as FFs to replace with scan FFs Each FF to replace with a scan FF is temporarily selected as a FF to replace with a non-scan flip-flop, and the structure of the integrated circuit is checked if it has an n-folded line-up structure and if so, then the FF is selected as a FF to replace with a non-scan flip-flop. For an integrated circuit designed at the gate level, flip-flops to replace with scan flip-flops are selected in order that the integrated circuit has an n-fold line-up structure, without recognizing load/hold FFs as self-loop structure FF. Thereafter, FFs to replace with scan FFs are selected in such a way as to facilitate testing on load/hold FFs. The present invention guarantees high fault efficiency in identifying FFs to replace with scan FFs and achieves a higher compaction rate than conventional technology.

Description

BACKGROUND OF THE INVENTION

The present invention relates to design for testability and test sequence generation for integrated circuits (LSIs).

Scan design is a conventionally used design as a typical design for testability technology. In a scan design method, flip-flops (FFs) in a logic-designed integrated circuit are replaced by scan FFs and hence can be controlled (scan-in) and observed (scan-out) directly from outside and the problem of sequential circuits is simplified into that of combinational circuits, for test sequence generation facilitation. One such technology is described in Digital Systems Testing and Testable Design, Chapter 9, Design For Testability, published in 1990 by Computer Science Press.

Scan design may be classified into two types, namely a full scan design method and a partial scan design method. In the former design method, all FFs in a circuit are replaced by scan FFs. On the other hand, in the latter design method, only some FFs in a circuit are replaced by scan FFs. A method of identifying (selecting) scan FFs in the partial scan design is fully discussed in a paper, entitled An Exact Algorithm for Selecting Partial Scan Flip-Flops, DAC (Design Automation Conference), pp.81-86, 1994 as well as in its references.

Additionally, as to test sequence generation for sequential circuits, test sequence compaction is fully described in a paper entitled Dynamic Test Compaction for Synchronous Sequential Circuits using Static Compaction Technique, FTCS (Fault Tolerant Computing Symposium), pp. 53-61, 1996 as well as in its references.

The above-noted prior art techniques however have their respective problems. A conventional partial scan design method produces the problem that in identification of FFs to replace with scan FFs it is not always possible to guarantee sufficiently high fault efficiency, i.e., 95% or more. Additionally a conventional test sequence compaction method for sequential circuits also produces the problem that it is poor in compaction rate.

SUMMARY OF THE INVENTION

An object of the invention is to provide an improved method of design for testability to guarantee a high fault efficiency in identifying FFs to replace with scan FFs. Another object of the invention is to provide an improved method of test sequence generation able to achieve a higher compaction rate in comparison with the prior art techniques.

This invention provides a method of design for testability for modifying design of an integrated circuit designed at gate level, in order to attain testability after manufacture, comprising the step of selecting flip-flops to replace with scan flip-flops among flip-flops in said integrated circuit in order that said integrated circuit has an n-fold line-up structure.

It is preferred that the method of design for testability comprises:

a full scan step of temporarily selecting about all flip-flops in said integrated circuit as flip-flops to replace with scan flip-flops; and

a non-scan flip-flop selection step of examining, for each of the flip-flops temporarily selected as flip-flops to replace with scan flip-flops by said full scan step, that is, temporary scan flip-flop, whether said integrated circuit has an n-fold line-up structure or not in assuming each said temporary scan flip-flop as flip-flops to replace with non-scan flip-flops, and temporarily selecting, when said integrated circuit has an n-fold line-up structure by said assumption, each said temporary scan flip-flop as a flip-flop to replace with a non-scan flip-flop,

wherein the flip-flops temporarily selected as flip-flops to replace with scan flip-flops by said full scan step and said non-scan flip-flop selection step are finally selected as flip-flops to replace with scan flip-flops.

This invention provides another method of design for testability for modifying design of an integrated circuit designed at gate level in order to attain testability after manufacture, comprising:

a first step of selecting, with recognizing load/hold flip-flop as not having a self-loop structure, flip-flops to replace with scan flip-flops among flip-flops in said integrated circuit in order that said integrated circuit has an n-fold line-up structure; and

a second step of selecting flip-flops to replace with scan flip-flops so as to attain testability on load/hold flip-flops, in said integrated circuit with the flip-flops to replace with scan flip-flops selected at said first step.

It is preferred that in the method of design for testability said second step includes:

performing timeframe expansion, based on the state justification of load/hold flip-flops, about said integrated circuit with the flip-flops to replace with scan flip-flops selected at said first step, and selecting flip-flops to replace with scan flip-flops from said timeframe expansion.

This invention provides a method of test sequence generation for an integrated circuit, comprising:

a buffer length set step of setting a buffer length for a buffer for storing a test sequence; and

a test sequence compaction step of generating a test sequence for said integrated circuit, in performing sequentially compaction storage of test sequences for respective faults in buffers having said buffer length set in said buffer length set step.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flowchart showing procedures in a method of design for testability according to a first embodiment of the present invention.

FIG. 2 is an illustration of an integrated circuit (LSI) which becomes an object of the method of design for testability of the first embodiment.

FIGS. 3 ( a )-( e ) show timeframe expansions when FF r 1 of the integrated circuit of FIG. 2 is selected as a FF to replace with a non-scan FF.

FIGS. 4 ( a )-( d ) show timeframe expansions when FF r 2 of the integrated circuit of FIG. 2 is selected as a FF to replace with a non-scan FF.

FIG. 5 shows a timeframe expansion when FF r 4 of the integrated circuit of FIG. 2 is selected as a FF to replace with a non-scan FF.

FIG. 6 is an illustration of a method of design for testability of the first embodiment applied to the integrated circuit of FIG. 2 .

FIGS. 7 ( a ) and 7 ( b ) show respectively L/H FFs.

FIG. 8 is a flowchart showing procedures in a method of design for testability according to a second embodiment of the present invention.

FIG. 9 is a flowchart showing the details of step SB 4 of FIG. 8 .

FIG. 10 is an illustration of a sequential circuit which becomes an object of the method of design for testability of the second embodiment.

FIG. 11 is a diagram obtained by determining FFs to replace with scan FFs in order that the sequential circuit of FIG. 10 has a 2-fold line-up structure.

FIGS. 12 ( a )-( c ) illustrate step SB 4 on a primary output PO 1 , with respect to the circuit of FIG. 11 .

FIGS. 13 ( a )-( c ) illustrate step SB 4 on a primary output PO 2 , with respect to the circuit of FIG. 11 .

FIGS. 14 ( a ) and ( b ) illustrate step SB 4 on a pseudo-primary output PPO 5 , with respect to the circuit of FIG. 11 .

FIGS. 15 ( a ) and ( b ) illustrate step SB 4 on a pseudo-primary output PPO 3 , with respect to the circuit of FIG. 11 .

FIGS. 16 ( a ) and ( b ) illustrate step SB 4 on a pseudo-primary output PPO 4 , with respect to the circuit of FIG. 11 .

FIGS. 17 ( a )-( c ) illustrate step SB 4 on a pseudo-primary output PPO 7 , with respect to the circuit of FIG. 11 .

FIG. 18 is a flowchart showing procedures in a method of design for testability according to a third embodiment of the present invention.

FIGS. 19 ( a ) and ( b ) are illustrations of sequential circuits which become an object of the method of design for testability of the third embodiment.

FIG. 20 is a flowchart showing procedures in a method of design for testability according to a fourth embodiment of the present invention.

FIG. 21 is a flowchart showing the details of step SD 4 of FIG. 20 .

FIGS. 22 ( a ) and ( b ) are illustrations of sequential circuits which become an object of the method of design for testability of the fourth embodiment.

FIGS. 23 ( a ) and ( b ) show FFs having a self-loop structure and is a diagram showing a method of how expanded pure L/H FFs are identified.

FIGS. 24 ( a ) and ( b ) show FFs having a self-loop structure and is a diagram showing a method of how expanded pure L/H FFs are identified.

FIG. 25 , comprised of FIGS. 25 ( a ) and ( b ), shows timeframe expansions of the circuit of FIG. 23 , wherein FIG. 25 ( a ) is an illustration of a normal timeframe expansion while FIG. 25 ( b ) is an illustration of a timeframe expansion making utilization of a hold signal line.

FIG. 26 is an illustration of a sequential circuit which becomes an object of the method of design for testability of the fourth embodiment.

FIGS. 27 ( a )-( c ) are diagrams each giving an indication of whether each FF 1 and FF 2 is an expanded pure L/H FF with reference to the circuit of FIG. 26 .

FIGS. 28 ( a )-( d ) are diagrams each giving an indication of whether each FF 3 and FF 4 is an expanded pure L/H FF with reference to the circuit of FIG. 26 .

FIGS. 29 ( a ) and ( b ) are diagrams showing results obtained by performing one-timeframe expansion on the primary output PO 1 , with respect to the circuit of FIG. 26 .

FIGS. 30 ( a ) and ( b ) are diagrams showing results obtained by performing, based on the results of FIG. 29 , a next one-timeframe expansion.

FIG. 31 is a diagram showing a result obtained by performing, based on the results of FIG. 30 , another next one-timeframe expansion.

FIGS. 32 ( a ) and ( b ) graphically show FF relations prepared as to the circuit of FIG. 2 .

FIG. 33 shows reachable data prepared based on the FF of FIG. 32 ( b ).

FIGS. 34 ( a )-( d ) show timeframe expansions when FF r 1 is selected as a FF to replace with a non-scan FF, with reference to the FF of FIG. 32 ( b ).

FIGS. 35 ( a ) and ( b ) show timeframe expansions when FF r 2 is selected as a FF to replace with a non-scan FF, with reference to the FF of FIG. 32 ( b ).

FIGS. 36 shows a timeframe expansion when FF r 4 selected as a FF to replace with a non-scan FF, with reference to the FF of FIG. 32 ( b ).

FIG. 37 is an illustration of a logic circuit which becomes an object of a method of design for testability as a seventh embodiment of the present invention.

FIG. 38 is an illustration of a result of the method of design for testability of the seventh embodiment when applied to the circuit of FIG. 37 .

FIG. 39 is a flowchart showing procedures in a method of design for testability according to an eighth embodiment of the present invention.

FIG. 40 is a FF relation graph representing a circuit which becomes an object of the method of design for testability of the eighth embodiment.

FIG. 41 ( a ) is a diagram obtained by determining FFs to replace with scan FFs in such a way as to implement a 1-fold line-up structure with reference to the FF of FIG. 40 and FIG. 41 ( b ) shows a timeframe expansion for the result of FIG. 41 ( a ).

FIG. 42 ( a ) is a diagram obtained by determining FFs to replace with scan FFs in such a way as to implement a 2-fold line-up structure with reference to the FF of FIG. 40 and FIG. 42 ( b ) shows a timeframe expansion for the result of FIG. 42 ( a ).

FIG. 43 ( a ) is a diagram obtained by determining FFs to replace with scan FFs in such a way as to implement a 3-fold line-up structure with reference to the FF of FIG. 40 and FIG. 43 ( b ) shows a timeframe expansion for the result of FIG. 43 ( a ).

FIG. 44 is a diagram schematically showing an integrated circuit composed of three identical blocks A.

FIGS. 45 ( a ) and ( b ) are flowcharts respectively showing procedures in a method of design for testability according to a ninth embodiment of the invention.

FIG. 46 shows a result obtained by performing step SF 1 of FIG. 45 on a certain circuit block.

FIG. 47 shows a result obtained by performing step SF 1 of FIG. 45 on a certain circuit block.

FIG. 48 is a flowchart showing procedures in a method of test sequence generation according to a tenth embodiment of the invention.

FIG. 49 is a flowchart showing the details of step SG 11 of FIG. 48 .

FIGS. 50 ( a ) and ( b ) are diagrams useful in understanding the method of test sequence generation of the tenth embodiment.

FIGS. 51 ( a ) and ( b ) are diagrams useful in understanding the method of test sequence generation of the tenth embodiment.

FIGS. 52 ( a ) and ( b ) are diagrams useful in understanding the method of test sequence generation of the tenth embodiment.

FIG. 53 is a diagram useful in understanding the method of test sequence generation of the tenth embodiment.

FIG. 54 is a flowchart showing procedures in a method of test sequence generation according to an eleventh embodiment of the invention.

FIG. 55 is a flowchart showing the details of step SH 13 of FIG. 54 .

FIGS. 56 ( a ) and ( b ) are diagrams useful in understanding the method of test sequence generation of the eleventh embodiment.

FIGS. 57 ( a ) and ( b ) are diagrams useful in understanding the method of test sequence generation of the eleventh embodiment.

FIGS. 58 ( a ) and ( b ) are diagrams useful in understanding the method of test sequence generation of the eleventh embodiment.

FIGS. 59 ( a ) and ( b ) are diagrams useful in understanding the method of test sequence generation of the eleventh embodiment.

DETAILED DESCRIPTION OF THE INVENTION

Embodiment 1

FIG. 1 is a flowchart showing procedures in a method of design for testability according to a first embodiment of the present invention. In FIG. 1 , in step SA 1 as a full scan step, all the flip-flops (FFs) in a target integrated circuit are temporarily selected as FFs to replace with scan FFs. Note that such a FF temporarily selected as FF to replace with scan FF is hereinafter called temporary scan FF. At steps SA 2 to SA 8 , for each temporary scan FF, it is examined whether the target integrated circuit has an n-fold line-up structure or not in assuming each the temporary scan FFs is selected as a FF to replace with a non-scan FF (step SA 7 ), and if such examination shows that the target integrated circuit has an n-fold line-up structure, the temporary scan FF in question is temporarily selected as a FF to replace with a non-scan FF. At step SA 9 , temporary scan FFs, i.e., FFs temporarily selected as FFs to replace with scan FFs by steps SA 1 to SA 8 are finally selected as flip-flops to replace with scan FF.

An n-fold line-up structure means a circuit structure in which all paths between a flip-flop and an output have n or less sequential depths. A circuit having the n-fold line-up structure circuit has a characteristic that at most n identical FFs exist in timeframe expansion with respect to a primary output. For instance, a 1-fold line-up structure (when n is equal to 1) has a characteristic that the number of identical FFs is one in timeframe expansion with respect to a primary output. In accordance with the present embodiment, the selecting of FFs to replace with scan FFs is made in order that a target integrated circuit has an n-fold line-up structure, wherein an input and an output of a FF to replace with a scan FF are assumed as a pseudo-primary output and as a pseudo-primary input, respectively.

SA 2 is the step of recognizing a flip-flop having a self-loop structure. When a flip-flop selected at step SA 4 is recognized as a FF having a self-loop structure at step SA 2 , the procedure returns to step SA 3 (step SA 5 ). In other words, step SA 7 is omitted for the FF having a self-loop structure, leaving the FF remained as temporary scan FF.

The method of design for testability of this embodiment will be described with reference to FIGS. 2-6 . Here, n is 1 in the n-fold line-up structure, and in step SA 7 it is examined whether a target integrated circuit has a 1-fold line-up structure.

FIG. 2 represents a target integrated circuit (a sequential circuit) for the method of design for testability of this embodiment. A-E are combinational circuits. r 1 -r 4 are flip-flops (FFs). PI 1 and PI 2 are primary inputs. PO 1 and PO 2 are primary outputs. Interconnections among elements are represented by arrows. All of the clocks of FF r 1 -FF r 4 are coupled to the same clock input, the description of which is omitted in FIG. 2 .

With respect to a temporary scan FF, its data input and output are assumed as a pseudo-primary output and as a pseudo-primary input, respectively. PPOi and PPIi represent, respectively, a pseudo-primary output and a pseudo-primary input, the character i being indicative of a corresponding FF. For example, when FF r 1 is a temporary scan FF, its data input and output are assumed as a pseudo-primary output PPO 1 and as a pseudo-primary input PPI 1 , respectively.

At step SA 1 , all flip-flops in FIG. 2 , i.e., FF r 1 -r 4 , are selected as temporary scan FFs. At step SA 2 , the circuit structure of FIG. 2 is analyzed and the result shows that FF r 3 is the only FF that has a self-loop structure.

Hereinafter, steps SA 4 -SA 8 are performed on each of FF r 1 -r 4 .

At step SA 4 , FF r 1 is selected. Since FF r 1 does not have a self-loop structure, at step SA 6 FF r 1 is temporarily selected as a FF to replace with a non-scan FF, and at step SA 7 the circuit is examined whether to have a 1-fold line-up structure.

FIG. 3 shows a timeframe expansion when FF r 1 is temporarily selected as a FF to replace with a non-scan FF. At step SA 7 c, timeframe expansion is performed with respect to the primary output PO 1 selected at step SA 7 b (FIG. 3 ( a )). Since there exists no FF in the expanded timeframe, the procedure returns to step SA 7 a from step SA 7 d. Then the primary output PO 2 is selected and timeframe expansion is performed (FIG. 3 ( b )). Since there exists no FF in the expanded timeframe, the procedure again returns to step SA 7 a. Then the pseudo-primary output PPO 2 (i.e., the data input of FF r 2 as a temporary scan FF) is selected, and timeframe expansion is performed (FIG. 3 ( c )). Since there exists only one FF r 1 in the expanded timeframe, the procedure again returns to step SA 7 a. Then the primary output PPO 3 (i.e., the data input of FF r 3 as a temporary scan FF) is selected and timeframe expansion is performed (FIG. 3 ( d )). Since there exists only one FF r 1 in the expanded timeframe, the procedure again returns to step SA 7 a. Then the pseudo-primary output PPO 4 (i.e., the data input of FF r 4 as a temporary scan FF) is selected, and timeframe expansion is performed (FIG. 3 ( e )). Since there exists no FF in the expanded timeframe, the procedure again returns to step SA 7 a. Timeframe expansion processing is now completed with respect to all the primary outputs and all the pseudo-primary outputs. The procedure therefore returns to step SA 3 . At step SA 4 , a next FF is selected.

At step SA 4 , FF r 2 is selected. Since FF r 2 does not have a self-loop structure, at step SA 6 FF r 2 is temporarily selected as a FF to replace with a non-scan FF. Step SA 7 examines whether the circuit has a 1-fold line-up structure.

FIG. 4 shows a timeframe expansion when FF r 2 is temporarily selected as a FF to replace with a non-scan FF. At step SA 7 c, timeframe expansion is performed with respect to the primary output PO 1 selected at step SA 7 b (FIG. 4 ( a )). Since there exists only one FF r 1 and one FF r 2 in the expanded timeframe, the procedure returns to step SA 7 a from step SA 7 d. Then, the primary output PO 2 is selected, and timeframe expansion is performed (FIG. 4 ( b )). Since there exists no FF in the expanded timeframe, the procedure again returns to step SA 7 a. Then, the pseudo-primary output PPO 3 is selected, and timeframe expansion is performed (FIG. 4 ( c )). Since there exists only one FF r 1 in the expanded timeframe, the procedure again returns to step SA 7 a. Then, the pseudo-primary output PPO 4 is selected, and timeframe expansion is performed (FIG. 4 ( d )). Since there exist only one FF r 1 and only one FF r 2 in the expanded timeframe, the procedure again returns to step SA 7 a. Timeframe expansion processing is now completed with respect to all the primary outputs and all the pseudo-primary outputs. The procedure therefore returns to step SA 3 . At step SA 4 , a next FF is selected.

At step SA 4 , FF r 3 is selected. However, since FF r 3 was already recognized as a FF having a self-loop structure, the procedure returns to step SA 3 from step SA 5 . At step SA 4 , a next FF is selected.

At step SA 4 , FF r 4 is selected. FF r 4 does not have a self-loop structure. Accordingly, at step SA 6 FF r 4 is temporarily selected as a FF to replace with a non-scan FF. Step SA 7 examines whether the circuit has a 1-fold line-up structure.

FIG. 5 shows a timeframe expansion when FF r 4 is temporarily selected as a FF to replace with a non-scan FF. At step SA 7 c, timeframe expansion is performed with respect to the output terminal PO selected at step SA 7 b. There exist two FFs r 1 and two FFs r 2 in the expanded timeframe, which means that the circuit does not have a 1-fold line-up structure. The procedure moves to step SA 8 from step SA 7 d, and FF r 4 is selected as a temporary scan FF.

Steps SA 4 -SA 8 has been performed for all the FF r 1 -r 4 . The temporary scan FFs, i.e., FF r 3 and FF r 4 , are finally selected as flip-flops to replace with FFs at step SA 9 , and the procedure is finished.

FIG. 6 shows a result of the method of design for testability of the present embodiment for the integrated circuit of FIG. 2 . In FIG. 6 , FF r 3 and r 4 which are finally selected as flip-flops to replace with, are really replaced with scan FFs. SI represents a scan-in primary input. SO represents a scan-out primary output. SE represents a scan enable primary input. A scan path is constructed such that it extends from the scan-in primary input SI to the scan-out primary output SO by way of a scan input DT and a scan output Q of the scan FF r 4 and a scan input DT and a scan output Q of the scan FF r 3 .

In accordance with the present method of design for testability, FFs to replace with scan FFs are determined in order that a sequential circuit has an n-fold line-up structure, hence, high fault efficiency can be guaranteed at scan FF identification.

The algorithm of the present embodiment can be applied easily to methods of design for testability for an integrated circuit designed at register transfer level (RTL). At first, full scan step is performed in which all registers in an RTL-designed integrated circuit are temporarily selected as registers to replace with scan registers. Then, non-scan register selection step is performed in which, for each temporary scan register selected by the full scan step, it is examined whether the RTL-designed integrated circuit has an n-fold line-up structure or not in assuming each the temporary scan register is selected as a register to replace with a non-scan register, and if such examination shows that the RTL-designed integrated circuits has an n-fold line-up structure, each the temporary scan registers in question is temporarily selected as a register to replace with a non-scan register. And, the registers temporarily selected as registers to replace with scan registers by the full scan step and the non-scan register selecting step are finally selected as registers to replace with scan registers.

Embodiment 2

FIG. 7 shows a load/hold FF (L/H FF). As shown in FIG. 7 ( a ), the L/H FF switches between load mode and hold mode by a load/hold signal applied at its load/hold selection input LH. In the load mode, data is loaded at the D input, while in the hold mode the currently-held data continues to be held. Use of L/H FFs reduces clock signal lines and power consumption. As shown in FIG. 7 ( b ), the L/H FF may be described by a combination of a selector and a DFF. The selector selects, using the load/hold signal as a selection signal, a primary input or an output of the DFF and provides whichever is selected. The output of the selector serves as an input of the DFF. In other words, the L/H FF has a self-loop structure.

In accordance with a method of design for testability of the present embodiment, on condition that an L/H FF having a self-loop structure shown in FIG. 7 is recognized as a FF not having a self-loop structure, partial scan design is performed based on the state justification of load/hold flip-flops.

FIG. 8 is a flowchart showing procedures in the method of design for testability according to the present embodiment. In FIG. 8 , SB 1 is the step of selecting FFs to replace with scan FFs in order that a target integrated circuit has an n-fold line-up structure. This step SB 1 is similar to the method of design for testability of the first embodiment, but it differs from the first embodiment method as follows. That is, in step SB 1 , selection of FFs to replace with scan FFs is made, with recognizing L/H FF as flip-flop not having a self-loop structure.

At steps SB 2 to SB 4 , partial scan design is performed based on the L/H FF state justification. At step SB 3 , an output is selected among untreated primary outputs and pseudo-primary outputs. At step SB 4 , timeframe expansion, based on the L/H FF state justification, is performed with respect to the output selected at step SB 3 , to select FFs to replace with scan FFs. When steps SB 3 and SB 4 are performed on all of the primary outputs and the pseudo-primary outputs, the present design for testability is completed (step SB 2 ).

FIG. 9 is a flowchart showing the details of step SB 4 in the present method of design for testability shown in FIG. 8 . At this step SB 4 , the determining of FFs to replace with scan FFs is made in order that test sequences may become generative in a predetermined number of timeframes. In FIG. 9 , SB 4 a is the step of setting 1 as an initial value for a timeframe number i and setting the upper limit of the timeframe number i as a designation timeframe number t. SB 4 b is the step of determining whether processing is finished for timeframes equal in number to the designation timeframe number t. If finished, then the procedure moves to step SB 4 m, while if not, the procedure moves to SB 4 c. SB 4 c is the step of searching for the current timeframe i either from the primary output or pseudo-primary output or from the data input of a FF existing in the previous timeframe. SB 4 d is the step of examining whether there exists a FF in the timeframe i searched at step SB 4 c. If a FF exists in the timeframe i, the procedure advances to step SB 4 e, while if not, the procedure is finished. SB 4 e is the step of counting the number of FFs in the current timeframe i. SB 4 f is the step of examining whether the timeframe number i is different from the designation timeframe number t and there exists an L/H FF in the current timeframe i. If the numbers i and t disagree with each other and there exists an L/H FF in the current timeframe i, the procedure moves to step SB 4 g. If not, the procedure returns to step SB 4 b by step SB 41 . SB 4 g is the step of performing state justification on L/H FFs found to exist in the current timeframe i. SB 4 h is the step of examining whether L/H FF state justification at step SB 4 g succeeds. If every L/H FF state justification is successful, the procedure moves to step SB 4 i, while if not, the procedure moves to step SB 4 j. SB 4 i is the step of determining a search direction in a subsequent timeframe for a successfully state justified L/H FF. SB 4 j is the step of determining all the L/H FFs in the current timeframe i as FFs to replace with scan L/H. SB 4 k is the step of subtracting the number of L/H FFs determined as FFs to replace with scan L/H FFs at step SB 4 j from the number of FFs included in the current timeframe i. SB 4 m is the step of making a decision that all FFs, which exist in a timeframe with the minimum number of FFs counted at steps SB 4 d and SB 4 k, are selected as FFs to replace with scan FFs.

Now, a method of design for testability of the present embodiment shown in FIGS. 8 and 9 will be described, referring to FIGS. 10-17 .

FIG. 10 shows a sequential circuit as an object of the present method of design for testability. A-I are combinational circuits. r 1 -r 7 are FFs. PI 1 -PI 4 are primary inputs. PO 1 -PO 2 are primary outputs. r 3 , r 4 , r 6 , and r 7 are L/H FFs. Each arrow represents interconnection among these elements. Clocks of FF r 1 -FF r 7 are all coupled to the same clock input, the description of which is omitted in FIG. 10 .

For the circuit of FIG. 10 , step SB 1 determines FFs to replace with scan FFs in order that the circuit has an n-fold line-up structure. FIG. 11 shows a result obtained by determining FFs to replace with scan FFs in order that the circuit of FIG. 10 has a 2-fold line-up structure. FIG. 11 shows that only FF r 5 is selected as a FF to replace with a scan FF. Note that at step SB 1 , L/H FFs r 3 , r 4 , r 6 , and r 7 are not selected as FFs to replace with scan FFS, since they are recognized as not having a self-loop structure.

At steps SB 2 -SB 4 , based on the state justification of L/H FFs, partial scan design is performed for a circuit shown in FIG. 11 . At step SB 4 , according to the FIG. 9 flowchart, scannization based on the L/H FF state justification is performed, with respect to primary outputs or pseudo-primary outputs selected at step SB 3 . At step SB 4 , an L/H FF is regarded as a circuit composed of a selector and a DFF shown in FIG. 7 ( b ). In this case, load mode is a situation that the S input of the selector is “0”, while hold mode is a situation that the selector S input is “1”. Here, at step SB 4 a, the designation timeframe number t is set at 3.

At step SB 3 , the primary output PO 1 is selected, and step SB 4 is executed.

FIG. 12 shows the processing at step SB 4 for the primary output PO 1 . As shown in FIG. 12 ( a ), the search from the primary output PO 1 is performed, which results in getting to L/H FFS r 4 , r 6 (step SB 4 c ). The number of FFs in this timeframe 1 is two (step SB 4 e ). Since the timeframe number i is 1 and therefore is different from the designation timeframe number t (=3) and there exist L/H FF r 4 and L/H FF r 6 in the current timeframe (step SB 4 f ), L/H FFs r 4 , r 6 are state justified (step SB 4 g ). Suppose here that “0” state justification at D inputs of FF r 4 , r 6 is carried out with success and L/H FF r 4 and L/H FF r 6 are determined to be in hold mode and in load mode, respectively. As search directions in the next timeframe, the input side of hold mode is set for L/H FF r 4 and the input side of load mode is set for L/H FF r 6 (step SB 4 i ).

As shown in FIG. 12 ( b ), the hold mode input side of L/H FF r 4 and the load mode input side of L/H FF r 6 are searched, which results in getting to FF r 1 -FF r 4 (step SB 4 c ). The number of FFs in this timeframe 2 is four (step SB 4 e ). Since the timeframe number i is 2 and therefore is different from the designation timeframe number t (=3) and there exist L/H FF r 3 and L/H FF r 4 in the current timeframe (step SB 4 f ), L/H FF r 3 and L/H FF r 4 are state justified (step SB 4 g ). Suppose here that “0” state justification at D inputs of FF r 3 , r 4 is unsuccessful. In such a case, these L/H FF r 3 and L/H FF r 4 are selected as FFs to replace with scan FFs (step SB 4 j ), and the number of FFs present in the current timeframe 2 is reduced from 4 down to 2.

Next, as shown in FIG. 12 ( c ), the input side of each FF r 1 and FF r 2 is searched, which results in getting to no FFs but to the primary inputs PI 2 and PI 3 . The processing at step SB 4 for the primary output PO 1 is therefore finished (step SB 4 d ). As a result, it is determined that L/H FF r 3 and L/H FF r 4 are FFs to replace with scan FFs.

Next, the primary output PO 2 is selected at step SB 3 , and the processing at step SB 4 is carried out.

FIG. 13 shows the processing at step SB 4 for the primary output PO 2 . As shown in FIG. 13 ( a ), the search from the primary output PO 2 is performed, which results in getting to L/H FF r 7 (step SB 4 c ). The number of FFs in this timeframe 1 is one (step SB 4 e ). Since the timeframe number i is 1 and therefore is different from the designation timeframe number t (=3) and there exists L/H FF r 7 in the current timeframe (step SB 4 f ), L/H FF r 7 is state justified (step SB 4 g ). Suppose here that logical “0” state justification is carried out with success and L/H FF r 7 is determined to be in load mode. As a search direction in the next timeframe, the input side of load mode is set for L/H FF r 7 (step SB 4 i ).

As shown in FIG. 13 ( b ), the load mode input side of L/H FF r 7 is searched, which results in getting to FF r 6 (step SB 4 c ). The number of FFs in this timeframe 2 is one (step SB 4 e ). Since the timeframe number i is 2 and therefore is different from the designation timeframe number t and there exists L/H FF r 6 in the current timeframe (step SB 4 f ), L/H FF r 6 is state justified (step SB 4 g ). Suppose here that logical “0” state justification is carried out with success and L/H FF r 6 is determined to be in load mode. As a search direction in the next timeframe, the input side of load mode is set for L/H FF r 6 (step SB 4 i ).

Next, as shown in FIG. 13 ( c ), the input side of L/H FF r 6 in load mode is searched, which results in getting to FF r 2 (step SB 4 c ). The number of FFs in this timeframe 3 is one (step SB 4 e ). Since the timeframe number i is 3 and therefore agrees with the designation timeframe number t (step SB 4 f ), timeframe expansion is here finished (step SB 4 b ), and FFs which exist in a timeframe with the minimum number of FFs are determined as FFs to replace with scan FFs (step SB 4 m ). FIG. 13 shows that each timeframe has one FF, in other words each timeframe has the same number of FFs. Accordingly, FF r 7 in the timeframe 1 is determined as a FF to replace with a scan FF.

Next, at step SB 3 , the pseudo-primary output PPO 5 (i.e., the data input of FF r 5 selected as a FF to replace with a scan FF at step SB 1 ) is selected, and the step SB 4 is executed.

FIG. 14 shows the processing at step SB 4 for the pseudo-primary output PPO 5 . As shown in FIG. 14 ( a ), search from the pseudo-primary output PPO 5 is performed, which results in getting to FF r 2 (step SB 4 c ). The number of FFs in this timeframe 1 is one (step SB 4 e ). Since this timeframe has no L/H FFs (step SB 4 f ), the next timeframe is expanded. As shown in FIG. 14 ( b ), since search from the input side of FF r 2 got to no FFs but to the primary input PI 3 , the processing at step SB 4 for the pseudo-primary output PPO 5 is finished (step SB 4 d ).

Next, at step SB 3 , the pseudo-primary output PPO 3 (i.e., the data input of a DFF forming L/H FF r 3 selected as a FF to replace with a scan FF at step SB 4 for the primary output PO 1 ) is selected, and the processing at step SB 4 is executed.

FIG. 15 shows the processing at step SB 4 for the pseudo-primary output PPO 3 . As shown in FIG. 15 ( a ), search from the pseudo-primary output PPO 3 , which results in getting to FF r 1 and FF r 2 (step SB 4 c ). The number of FFs in this timeframe 1 is two (step SB 4 e ). There exist no L/H FFs in the timeframe 1 (step SB 4 f ) and the next timeframe is therefore searched. As shown in FIG. 15 ( b ), since search from the input side of each FF r 1 , r 2 got to no FFs but to the primary inputs PI 2 and PI 3 , the processing at step SB 4 for the pseudo-primary output PPO 3 is finished (step SB 4 d ).

Next, at step SB 3 , the pseudo-primary output PPO 4 (i.e., the data input of a DFF forming L/H FF r 4 selected as a FF to replace with a scan FF at step SB 4 for the primary output PO 1 ) is selected, and the processing at step SB 4 is executed.

FIG. 16 shows the processing at step SB 4 for the pseudo-primary output PPO 4 . As shown in FIG. 16 ( a ), search from the pseudo-primary output PPO 4 is performed, which results in getting to FF r 1 and FF r 2 (step SB 4 c ). The number of FFs in this timeframe 1 is two (step SB 4 e ). There exist no L/H FFs in the timeframe 1 (step SB 4 f ) and the next timeframe is therefore searched. As shown in FIG. 16 ( b ), since the previous searching of the input side of each FF r 1 and FF r 2 got to no FFs but to the primary inputs PI 2 and PI 3 , the processing at step SB 4 with respect to the pseudo-primary output PPO 4 is finished (step SB 4 d ).

Next, at step SB 3 , the pseudo-primary output PPO 7 (i.e., the data input of a DFF forming L/H FF r 7 selected as a FF to replace with a scan FF at step SB 4 for the primary output PO 2 ) is selected, and the processing at step SB 4 is executed.

FIG. 17 shows the processing at step SB 4 for the pseudo-primary output PPO 7 . As shown in FIG. 17 ( a ), search from the pseudo-primary output PPO 7 is performed, which results in getting to FF r 6 (step SB 4 c ). The number of FFs in this timeframe 1 is one (step SB 4 e ). Since the timeframe number i is 1 and therefore is different from the designation timeframe number t and there exists L/H FF r 6 in the current timeframe (step SB 4 f ), L/H FF r 6 is state justified (step SB 4 g ). Suppose here that logical “0” state justification is carried out with success and it is determined that L/H FF r 6 is set to load mode. As a search direction in the next timeframe, the input side of load mode is set for L/H FF r 6 .

As shown in FIG. 17 ( b ), the load mode input side of L/H FF r 6 is searched to get to FF r 1 , r 2 (step SB 4 c ). The number of FFs in this timeframe 2 is two (step SB 4 e ). Since there exist no L/H FFs in the timeframe 2 , a search for the next timeframe is made.

As shown in FIG. 17 ( c ), search from the inside of each FF r 1 , r 2 got to no FFs but to the primary inputs PI 2 and PI 3 . The processing at step SB 4 for the pseudo-primary output PPO 7 is therefore finished.

Since the processing at step SB 4 is finished for all the primary outputs and all the pseudo-primary outputs (step SB 2 ), the present design for testability is completed. As a result, FF r 5 is selected as a FF to replace with scan FF at step SB 1 and L/H FF r 3 , FF r 4 , and FF r 7 are selected as FFs to replace with scan FFs at step SB 4 .

As described above, the present embodiment includes selecting FFs to replace with scan FFs in order that a integrated circuit has an n-fold line-up structure with recognizing L/H FF as FF not having a self-loop structure, and selecting, based on the L/H FF state justification, FFs to replace with scan FFS in order that test sequences may become generative in a predetermined number of timeframes. Such arrangement guarantees high fault efficiency for a sequential circuit including many L/H FFs in selecting FFs to replace with scan FFs at relatively low scannization rate.

Here, the purpose of assigning “0” at a plurality of L/H FFs in state justification is to determine whether each L/H FF is state justified in a load mode or in a hold mode. Thus, the other logic value, instead of “0”, may be assigned in state justification.

Embodiment 3

FIG. 18 is a flowchart showing procedures in a method of design for testability according to a third embodiment of the present invention. As shown in FIG. 18 , SC 1 is the step of selecting FFs to replace with scan FFs in order that a target integrated circuit has an n-fold line-up structure, and is similar to the step SB 1 of the second embodiment. In other words, with recognizing L/H FF as FF not having a self-loop structure, FFs to replace with scan FFs are selected such that a target circuit has an n-fold line-up structure. L/H FFs, not selected as FFs to replace with scan FFs at step SC 1 , are called here “pure L/H FFs”.

SC 2 is the step of determining whether each step SC 3 and SC 4 is-performed for all pure L/H FFs. When steps SC 3 and SC 4 are finished for all the pure L/H FFs, processing is finished. SC 3 is the step of selecting one of untreated pure L/H FFs. SC 4 is the step of selecting, as a FF to replace with a scan FF, a FF reachable from the load/hold selection input of the pure L/H FF selected at step SC 3 by passing merely through combinational circuits.

The method of design for testability of this embodiment shown in FIG. 18 will be described with reference to FIG. 19 .

FIG. 19 is a diagram showing a sequential circuit as an object of the present method of design for testability. FF 1 -FF 8 are L/H FFs. FF 9 -FF 15 are normal FFs (non-L/H FFS). A-D are combinational circuits. PI 1 -PI 8 are primary inputs. PO is a primary output. The combinational circuit C is a 1-hot decoder of which only one output provides a different value from the other outputs.

First, at step SC 1 , FFs to replace with scan FFs are selected in order that the circuit of FIG. 19 has an n-fold line-up structure. Here, n is equal to 1, that is, FFs to replace with scan FFs are selected such that the circuit of FIG. 19 has a 1-fold line-up structure circuit. At this time, L/H FFs 1 - 8 are recognized as not having a self-loop structure, which results in selecting only FF 15 as a FF to replace with a scan FF.

Next, at steps SC 2 -SC 4 , FF 9 -FF 12 are selected as FFs to replace with scan FFs because these FF 9 -FF 12 are reachable from the load/hold selection inputs of L/H FF 1 -FF 8 by passing merely through combinational circuits.

As a result, arbitrary setting of load/hold signals is possible through scan paths, for L/H FF 1 -FF 8 . Therefore, most of faults can be detected with load mode, which facilitates test sequence generation. Additionally, in some circuit structures, it is possible to reduce the scan rate less than selecting all L/H FFs as FFs to replace with scan FFs by regarding them as having self-loop structure FFs.

Embodiment 4

A fourth embodiment of the invention is the method of obtaining a load mode ratio and examining the validity of the method of design for testability of the third embodiment by the load mode ratio. The load mode ratio expresses the ratio of L/H FFs which have been state justified with load mode to all the L/H FFs.

FIG. 20 is a flowchart showing procedures in a method of design for testability according to the fourth embodiment. SD 1 is the step of selecting FFs to replace with scan FFs using the method of design for testability of the third embodiment. SD 2 is the step of setting a load mode ratio threshold P and initializing values M and L used for obtaining load mode ratio. At steps SD 3 and SD 4 , timeframe expansion is performed based on the state justification of L/H FFs, with respect to each primary output and each pseudo-primary output and the values M and L are computed, wherein the value M is the total number of L/H FFs in the expanded timeframe and the value L is the total number of L/H FFs that are state justified with load mode. Step SD 5 is the step of determining whether the load mode rate (i.e., L/M) is greater than the threshold P. If L/M is greater than P, the design for testability of step SD 1 is determined to be valid, and the result thereof is used, while if not, the design for testability of step SD 1 is determined to be invalid, and at step SD 6 another design for testability, e.g., according to the second embodiment based on the L/H FF state justification is performed.

FIG. 21 is a flowchart showing the details of step SD 4 of the method of design for testability shown in FIG. 20 . SD 4 a is the step of performing a one-timeframe expansion, with respect to the primary output selected at step SD 3 or the input of a FF of the previous timeframe. For L/H FF, when it is state justified with load mode, expansion is performed with respect to the load mode input and select input of it, while state justified with hold mode, expansion is performed with respect to the hold mode input and select input of it. If step SD 4 b shows that there exist FFs in the current timeframe, then the procedure moves to step SD 4 c, while if not, step SD 4 is finished.

At step SD 4 c the number of L/H FFs in the current timeframe is set as m, then the procedure returns to SD 4 a when m is equal to zero (step SD 4 d ). At step SD 4 e, all the L/H FFs in the current timeframe are state justified. When such state justification is successful, the number of L/H FFs state justified with load mode, is set as 1 at step SD 4 g, while when unsuccessful, 0 is set as 1 at step SD 4 h. At step SD 4 i, m is added to M and 1 is added to L.

Load mode rate computation is now described based on the circuits shown in FIGS. 19 and 22 . Note that the load mode rate threshold P is 80%.

For the circuit of FIG. 19 , step SD 1 selects FF 9 -FF 12 and FF 15 as FFs to replace with scan FFs, similarly to the third embodiment. All the L/H FFs that appear in the timeframe with respect to the primary output PO, i.e., FF 1 -FF 8 , can be state justified with load mode because FF 9 -FF 12 are FFs to replace with scan FFs. Furthermore, all the L/H FFs that appear in the timeframe with respect to the pseudo-primary output PPO 15 (the input of FF 15 selected as a FF to replace with a scan FF), i.e., FF 1 -FF 8 can be state justified with load mode because FF 9 -FF 12 are FFs to replace with scan FFs. Accordingly, the load mode ratio in this case is 100% from L/M=16/16. In other words, since the load mode ratio is greater than the threshold P, the result of the step SD 1 is adopted.

In FIG. 22 , FF 1 -FF 4 are L/H FFs. FF 5 and FF 6 are normal FFs (non-L/H FFs). A-D are combinational circuits. PI 1 -PI 6 are primary inputs. PO is a primary output. Note that the combinational circuit C represents a 1-hot decoder.

For the circuit of FIG. 22 , step SD 1 selects FF 5 and FF 6 , which are reachable from the load/hold selection inputs of FF 1 -FF 4 which are L/H FFs by passing merely through combinational circuits as FFs to replace with scan FFs. Timeframe expansion is performed with respect to the primary output PO, and state justification is performed for L/H FFs in the expanded timeframe, i.e., FF 1 -FF 4 . Since the combinational circuit C is a 1-hot decoder, for example, FF 1 is state justified with load mode while the other FFs (FF 2 -FF 4 ) are state justified with hold mode. In this case, the load mode rate is 25% from L/M=1/4. In other words, since the load mode rate is less than the threshold P, the result of step SD 1 is not adopted, Instead, another design for testability, for example, a design for testability based on the L/H FF state justification of the second embodiment, is performed at step SD 6 .

Embodiment 5

In accordance with a fifth embodiment of the invention, a pre-step is performed on a target circuit before the method of design for testability of the second or third embodiment. Specifically, a pre-step is performed to recognize a FF having a self-loop structure other than L/H FF and capable of being state justified without assignment of values to its own output, as an extended pure L/H FF. Then, the method of design for testability of the second or third embodiment is performed in assuming such an extended pure L/H FF as L/H FF. As a result of such arrangement, it becomes possible to reduce the scan rate for the circuit including great numbers of FFs having a self-loop structure other than the L/H FFs.

A method of recognizing extended pure L/H FF will be described.

FIGS. 23 and 24 show FFs having a self-loop structure, and are used for explaining how extended pure L/H FFs are recognized.

In FIG. 23 ( a ), the setting of “0” at an input not forming a loop, of AND gate 1 b is sufficient to justify logical value “0” at D input of FF 1 a. Also, referring to FIG. 23 ( b ), the setting of “1” at inputs not forming a loop, of AND gate 1 b and OR gate 1 c is sufficient to justify logical value “1” at D input of FF 1 a. In other words, FF 1 a can be state justified, with its Q output not assigned a value. In such a case, FF 1 a is recognized as an extended pure L/H FF.

In FIG. 24 ( a ), the setting of “0” at an input not forming a loop, of AND gate 2 b is sufficient to justify logical value “0” at D input of FF 2 a. While referring to FIG. 24 ( b ), to justify logical value “1” at D input of FF 2 a, not only an input not forming a loop of AND gate 2 b but also an output of NOT gate 2 c are required to be set at “1”, hence, it becomes necessary to assign “0” to Q output of FF 2 a. In such a case, FF 2 a is not recognized as an extended pure L/H FF.

A hold signal line is defined for FF recognized as an extended pure L/H FF. After logical value “0” state justification, the first signal line at which the logical value is not set in search from D input via loop is defined as a 0 hold signal line. Also, after logical value “1” state justification, the first signal line at which the logical value is not ste in search from D input via loop is defined as a 1 hold signal line. Thereafter, the 0 hold signal line or the 1 hold signal line, whichever is farther away from D input, is defined as a hold signal line for the extended pure L/H FF. For instance, in the case of FIG. 23 , the 0 hold signal line is a signal line 1 d connecting output of OR gate 1 c and input of AND gate 1 b and the 1 hold signal line is a signal line 1 e connecting Q output of FF 1 a and input of OR gate 1 c. The 0 hold signal line 1 d or the 1 hold signal line 1 e, whichever is farther away from D input, i.e., the 1 hold signal line 1 e, is defined as a hold signal line for the extended pure L/H FF 1 a.

The object of defining a hold signal line is to have timeframe expansion result in a finite number of timeframes. FIG. 25 shows a timeframe expansion of the circuit shown in FIG. 23 . Whereas FIG. 25 ( a ) shows a normal timeframe expansion, FIG. 25 ( b ) shows a timeframe expansion using a hold signal line. In the normal timeframes expansion shown in FIG. 25 ( a ), timeframes are infinitely expanded for the circuit having a loop. Accordingly, in order to achieve a finite number of timeframes, FFs having self-loop structures are required to be selected as FF to replace with scan FF. On the other hand, in the timeframe expansion using a hold signal line shown in FIG. 25 ( b ), when an extended pure L/H FF is state justified with load mode, such a FF per se disappeared in a subsequent timeframe by performing no hold signal line expansion. This achieves a definite number of timeframes, which makes it possible to select FFs to replace with scan FFs.

FIG. 26 is a diagram showing a sequential circuit which is an object of the present method of design for testability. FF 1 -FF 4 are FFs having self-loop structures. FF 5 -FF 11 are FFs not having self-loop structures. PI 1 -PI 3 are primary inputs. PO is a primary output. A and B are combinational circuits.

In the first place, for the FIG. 26 sequential circuit, the pre-step is performed to determine whether each FF 1 -FF 4 having a self-loop structure is an extended pure L/H FF or not.

As shown in FIG. 27 ( a ), FF 1 is not recognized as an extended pure L/H FF, because “0” must be assigned to the Q input in “0” state justification at the D input. Also, as shown in FIGS. 27 ( b ) and ( c ), FF 2 is recognized as an extended pure L/H FF, because any value need not be assigned to the Q output, in not only “0” state justification but also “1” state justification at the D input. A 0 hold signal line 3 a that is farther away from the D input, is defined as a hold signal line for FF 2 .

As shown in FIGS. 28 ( a ) and ( b ), FF 3 is recognized as an extended pure L/H FF, because any value need not be assigned to the Q output, in not only “0” state justification but also “1” state justification at the D input. A 0 hold signal line 4 a that is farther away from the D input, is defined as a hold signal line for FF 3 . As shown in FIGS. 28 ( c ) and ( d ), FF 4 is recognized as an extended pure L/H FF, because any value need not be assigned to the Q output, in not only “0” state justification but also “1” state justification at the D input. A 0 hold signal line 5 a that is farther away from the D input, is defined as a hold signal line for FF 4 .

Next, the method of design for testability of the second embodiment shown in FIG. 8 is performed for the circuit of FIG. 26 . Here, at step SB 1 is performed in assuming FF 2 -FF 4 recognized as extended pure L/H FFs in the pre-step as FFs not having self-loop structures.

First of all, at step SB 1 , FFs to replace with scan FFs are selected in order that the circuit of FIG. 26 has a 1-fold line-up structure. Here, FF 1 having a self-loop structure and not recognized as an extended pure L/H FF in the pre-step is selected as a FF to replace with a scan FF.

Second, step SB 4 is performed with respect to the primary output PO. Here, designation timeframe number t is three.

FIG. 29 shows a one-timeframe expansion (time T) with respect to the primary output PO. As shown in FIG. 29 ( a ), the number of FFs in this timeframe is three since FF 2 -FF 4 exist. As shown in FIG. 29 ( b ), the “0” state justifications is performed at the D inputs of FF 2 -FF 4 recognized as extended pure L/H FFs. It is judged that FF 2 and FF 3 were state justified with load mode because it is unnecessary to assign any value to their own outputs. And, it is judged that FF 4 was state justified with hold mode because it is necessary to assign “0” to its own output.

FIG. 30 shows a next one-timeframe expansion (time T−1) with respect to the primary output PO. As shown in FIG. 30 ( a ), the number of FFs in this timeframe is six since FF 4 and FF 7 -FF 11 exist. As shown in FIG. 30 ( b ), the “0” state justification is performed at the D input of FF 4 recognized as an extended pure L/H FF. It is judged that FF 4 was state justified with load mode because it is unnecessary to assign any value to its own Q output.

FIG. 31 shows a further next one-timeframe expansion (time T−2) with respect to the primary output PO. As shown in FIG. 31 , the number of FFs in this timeframe is two since FF 10 and FF 11 having no self-loop structures exist. Three timeframes are expanded with respect to the primary output PO, and FF 10 and FF 11 existing in the timeframe of time T−2 with the minimum number of FFs are selected as FFs to replace with scan FFs.

Conventionally, all self-loop structure FFs are selected as FFs to replace with scan FFs for guaranteeing high fault efficiency. On the other hand, in accordance with the present embodiment, the method of design for testability of the second embodiment is performed, in assuming a FF having a self-loop structure but recognized as an extended pure L/H FF, as an L/H FF. This reduces the scan rate, maintaining high fault efficiency for circuits including many FFs having self-loop structures other than L/H FF. Method of design for testability for circuits in which extended pure L/H FFs are recognized is not limited to that of the second embodiment. For example, the method of design for testability of the third embodiment may be used instead of that of the second embodiment.

Embodiment 6

In a sixth embodiment of the invention, a FF relation graph is prepared as a pre-step of the method of design for testability of the first embodiment. The FF relation graph shows the interconnection relation, reachable by passing merely through combinational circuits, among FFs, primary inputs, and primary outputs in a target integrated circuit. Timeframe expansion is performed to select FFs to replace with scan FFs, based on the FF relation graph. This reduces the amount of computation for timeframe expansion very much.

Also, as a pre-step of the method of design for testability of the first embodiment, a reachable primary output and a reachable pseudo-primary output for a FF are found. And a reachable FFs that belong to paths from the FF to the reachable primary output or the reachable pseudo-primary output are found. Here, data about reachable primary outputs, reachable pseudo-primary outputs and reachable FFs for respective FFs are called as “reachable data”. In examining whether or not the integrated circuit has an n-fold line-up structure in assuming a temporary scan FF as a FF to replace with a non-scan FF, timeframe expansion is performed, by referring to the reachable data, only with respect to the reachable primary output and the reachable pseudo-primary output for the temporary scan FF and a data input of temporary scan FF among reachable flip-flops for the temporary scan FF. This reduces the amount of computation for timeframe expansion very much.

FIGS. 32 ( a ) and ( b ) are FF relation graphs prepared for sequential circuit of FIG. 2 . r 1 -r 4 are FFs. PI 1 and PI 2 are primary inputs. PO 1 and PO 2 are primary outputs. The FF relation graph of FIG. 32 ( a ) represents the sequential circuit of FIG. 2 as is. Having a self-loop structure, FF r 3 is selected as a FF to replace with a scan FF. Then, as shown in FIG. 32 ( b ), FF r 3 may be deleted from the FF relation graph, and a pseudo-primary output PPO 3 replaces the input of FF r 3 , and a pseudo-primary input PPI 3 replaces the output of FF r 3 .

FIG. 33 shows reachable data prepared based on the FF relation graph of FIG. 32 ( b ). The reachable data of FIG. 33 shows that for FF r 1 , PO 1 is a reachable primary output, PPO 3 a reachable pseudo-primary output, and FF r 2 , r 4 reachable FFs.

Based on the FF relation graph of FIG. 32 ( b ) and the reachable data of FIG. 33 , the method of design for testability of the first embodiment shown in FIG. 1 is performed. Here, step SA 7 examines whether the circuit has a 1-fold line-up structure.

At step SA 1 , all the FFs in the FF relation graph of FIG. 32 ( b ), i.e., FF r 1 , r 2 , and r 4 , are temporarily selected as FFs to replace with scan FFs. FFs ri selected as FFs to replace with scan FFs are assumed as a pseudo-primary output PPOi and a pseudo-primary input PPIi.

FIG. 34 shows a timeframe expansion in selecting FF r 1 as a FF to replace with a non-scan FF. The reachable data of FIG. 33 shows that PO 1 is the only reachable primary output and PPO 3 is the only reachable pseudo-primary output for FF r 1 . As shown in FIGS. 34 ( a ), ( b ), timeframe expansion is performed with respect to PO 1 and PPO 3 , respectively. Additionally, the reachable data of FIG. 33 shows that FF r 2 , r 4 are reachable FFs for FF r 1 , and FF r 1 , r 2 are temporary scan FFs. As shown in FIG. 34 ( c ), ( d ), timeframe expansion is performed with respect to the data inputs of FF r 2 , r 4 , i.e., the pseudo-primary outputs PPO 2 and PPO 4 , respectively. No timeframe expansion is performed with respect to primary outputs or pseudo-primary outputs other than PO 1 , PPO 2 , PPO 3 , and PPO 4 . Two or more identical FFs does not appear in each expanded timeframe, which means that this circuit has a 1-fold line-up structure. Accordingly, FF r 1 is temporarily selected as a FF to replace with a non-scan FF.

FIG. 35 shows a timeframe expansion in selecting FF r 2 as a FF to replace with a no-scan FF. The reachable reachable data of FIG. 33 shows that PO 1 is the only reachable primary output for FF r 2 . As shown in FIG. 35 ( a ), timeframe expansion is performed with respect to PO 1 . Additionally, the reachable data of FIG. 33 shows that FF r 4 is a reachable FF for FF r 2 , and FF r 4 is temporary scan FF. As shown in FIG. 35 ( b ), timeframe expansion is performed with respect to the data input of FF r 4 , i.e., the pseudo-primary output PPO 4 . No timeframe expansion is performed with respect to primary outputs or pseudo-primary outputs other than PO 1 and PPO 4 . Two or more identical FFs does not appear in each expanded timeframe, which means that this circuit has a 1-fold line-up structure. Accordingly, a temporary determination is made that FF r 2 is a FF to replace with a non-scan FF.

FIG. 36 shows a timeframe expansion in selecting FF r 4 as a FF to replace with a non-scan FF. The reachable data of FIG. 33 shows that PO 1 is the only reachable primary output for FF r 4 . Accordingly, timeframe expansion is performed with respect to PO 1 . Two FF r 1 s and two FF r 2 s appear in the expanded timeframe, which means that the circuit does not have a 1-fold line-up structure. Accordingly, FF r 4 is selected as a temporary scan FF.

As a result of the above-described processing, FF r 3 , r 4 are selected as temporary scan FFs. These FFS r 3 , r 4 are finally selected as flip-flop to replace with scan flip flops.

In accordance with the present embodiment, timeframe expansion is performed for a FF relation graph composed of three gates, instead of a sequential circuit of nine gates. Additionally, the use of the reachable data reduces the number of timeframe expansion from ten down to seven. This reduces the amount of computation for timeframe expansion very much.

Also, for the other designs for testability, the use of the FF relation graph reduces the amount of computation very much.

Embodiment 7

A seventh embodiment of the invention is directed to a method of design for testability for a logic circuit that is an integrated circuit designed at gate level. This method is of selecting FFs to replace with scan FFs in order that in assuming data inputs and data outputs of FFs to replace with scan FFs respectively as pseudo-primary outputs and as pseudo-primary inputs, the numbers of gates may be less than or equal to n in respective paths from a primary input or pseudo-primary input to a primary output or pseudo-primary output, wherein n is zero or a natural number.

The method of design for testability of this embodiment will be described with reference to FIG. 37 . FIG. 37 shows a target logic circuit for the method of design for testability of this embodiment. FF 1 -FF 7 are flip-flops. PI 1 and PI 2 are primary inputs. PO is a primary output. Here, for the circuit of FIG. 37 , FFs to replace with scan FFs are selected in order that the number of gates may be less than or equal to six in respective paths from the primary input or pseudo-primary input to the primary output or pseudo-primary output.

The circuit of FIG. 37 includes two loops, namely a loop 1 of FF 1 →FF 5 →FF 6 →FF 1 wherein the number of gates is ten and a loop 2 of FF 2 →FF 5 →FF 2 wherein the number of gates is two. Here, to break the loop 1 with more gates than the loop 2 , FFs to replace with scan FFs are selected.

Suppose that FF 1 is selected as a FF to replace with a scan FF among the FFs forming the loop 1 . The maximum number of gates (MNG) are: ten in a path from FF 1 to FF 1 , infinite in a path from FF 1 to PO, four in a path from PI 1 to FF 1 , eight in a path from PI 2 to FF 1 , infinite in a path from PI 1 to PO, and infinite in a path from PI 2 to PO. Likewise, suppose that FF 5 is selected as a FF to replace with a scan FF. The MNGs are: ten in a path from FF 5 to FF 5 , eight in a path from PI 1 to FF 5 , four in a path from FF 5 to PO, two in a path from PI 2 to FF 5 , and nine in a path from PI 2 to PO. Next, suppose that FF 6 is selected as a FF to replace with a scan FF. The MNGs are: ten in a path from FF 6 to FF 6 ; ten in a path from PI 1 to FF 6 , two in a path from FF 6 to PO, infinite in a path from PI 2 to FF 6 , infinite in a path from PI 1 to PO, and infinite in a path from PI 2 to PO.

Selecting FF 1 or FF 6 as a FF to replace with a scan FF results in leaving a path whose MNG is infinite, while selecting FF 5 as a FF to replace with a scan FF results in leaving no paths whose MNGs are infinite. Accordingly, FF 5 is first selected as a FF to replace with a scan FF.

When FF 5 is a scan FF, the circuit of FIG. 37 has four paths whose MNGs exceed six, that is, a path of FF 5 →FF 6 →FF 1 →FF 5 (MNG=10), a path of PI 1 →FF 1 →FF 5 (MNG=8), a path of PI 2 →FF 3 →FF 7 →PO (MNG=9), and a path of PI 2 →FF 2 →FF 7 →PO (MNG=7). Here, in selecting each of FFs in these four paths as a FF to replace with a scan FF, the MNGs of these paths are found,

Selecting FF 6 as a FF to replace with a scan FF provides four paths whose MNGs exceed six: a path of FF 6 →FF 1 →FF 5 (MNG=8), a path of PI 1 →FF 1 →FF 5 (MNG=8), a path of PI 2 →FF 3 →FF 7 →PO (MNG=9), and a path of PI 2 →FF 2 →FF 7 →PO (MNG=7). Selecting FF 1 as a FF to replace with a scan FF provides two paths whose MNGs exceed six: a path of PI 2 →FF 3 →FF 7 →PO (MNG=9) and a path of PI 2 →FF 2 →FF 7 →PO (MNG=7). Selecting FF 3 as a FF to replace with a scan FF provides three paths whose MNGs exceed six: a path of FF 5 →FF 6 →FF 1 →FF 5 (MNG=10), a path of PI 1 →FF 1 →FF 5 (MNG=8), and a path of PI 2 →FF 2 →FF 7 →PO (MNG=7). Selecting FF 2 as a FF to replace with a scan FF provides three paths whose MNGs exceed six: a path of FF 5 →FF 6 →FF 1 →FF 5 (MNG=10), a path of PI 1 →FF 1 →FF 5 (MNG=8), and a path of PI 2 →FF 3 →FF 7 →PO (MNG=9). Selecting FF 7 as a FF to replace with a scan FF provides three paths whose MNGs exceed six: a path of FF 5 →FF 6 →FF 1 →FF 5 (MNG=10), a path of PI 1 →FF 1 →FF 5 (MNG=8), and a path of PI 2 →FF 3 →FF 7 →PO (MNG=9).

Selecting FF 1 as a FF to replace with scan FF achieves the minimum number of paths whose MNGs exceed six. Accordingly, FF 1 is finally selected as FF to replace with scan FF.

Additionally, selecting FF 2 or FF 3 as a temporary scan FF results in leaving no paths whose MNGs exceed of six. Accordingly, for example, FF 2 is finally selected as FF to replace with scan FF.

FIG. 38 shows the result of the method of design for testability of this embodiment to the circuit of FIG. 37 . As shown in FIG. 38 , FF 1 , FF 2 , and FF 5 are selected as FFs to replace with scan FFs. As a result, the circuit has a no-closed circuit structure and the numbers of gates are less than or equal to six in respective paths from the primary input or pseudo-primary input to the primary output or pseudo-primary output. This facilitates test input generation in the circuit after insert as scan path.

Embodiment 8

FIG. 39 is a flowchart showing procedures in a method of design for testability according to an eighth embodiment of the present invention. The present method of design for testability is of selecting FFs to replace with scan FFs in order that a target circuit has a n-fold line-up structure, and of determining the value of n in order to reduce the test sequence generation time and achieve a sufficiently high fault efficiency while keeping the scan rate low.

Generally, for the circuit of an n-fold line-up structure, the scan rate increases, while the test sequence generation time gets shorter and the fault efficiency gets higher as the value of n decreases. On the other hand, as the value of n increases, the scan rate decreases, while the test sequence generation time gets longer the fault efficiency gets higher. Accordingly, it is required to determine an optimum value of n by balancing the scan rate with the test sequence generation time and the fault efficiency.

The results of experiments by the inventor of the present invention show that, when the value of n increases, the test sequence generation time tends to abruptly increase from a specific value. In addition, such a specific value varies by the circuit structures. For this reason, it is difficult to select an optimum value for n.

Further examination by the inventor shows as follows. When timeframe expansion is performed for a circuit after design for testability with respect to each primary output and each pseudo-primary output in assuming data inputs and data outputs of scan FFs, respectively, as pseudo-primary outputs and as pseudo-primary inputs, there is a high correlation between the number of FFs or gates in the expanded timeframe and the test sequence generation time and the fault efficiency. Here, the number of FFs in the expanded timeframe is called as the number of test sequence generation influence FFs, and the number of gates in the expanded timeframe is called as the number of test sequence generation influence gates. In other words, experiment by the inventor confirms that when the test sequence generation time abruptly increases, the number of test sequence generation influence FFs or the number of test sequence generation influence gates likewise abruptly increases.

The invention of the present embodiment makes use of the number of test sequence generation influence FFs or the number of test sequence generation influence gates, as an index for determining the value of n.

The present method of design for testability of FIG. 39 will be described by reference to FIGS. 40-43 .

FIG. 40 is a FF relation graph representing a target circuit of the present method of design for testability. A-L are FFs. PI 1 -PI 4 are primary inputs. PO is a primary output.

At step SE 1 , 1.6 is set as a threshold P. At step SE 2 , n is initialized to 1. At step SE 3 , FFs to replace with scan FFs are selected in order that the circuit of FIG. 40 has a 1-fold line-up structure.

FIG. 41 ( a ) shows a result of selecting FFs to replace with scan FFs in order that the circuit of FIG. 40 has a 1-fold line-up structure. As shown in FIG. 41 ( a ), FF C, E, F, and L with hatched are selected as FFs to replace with scan FFs. At step SE 4 , as shown in FIG. 41 ( b ), timeframe expansion is performed in assuming the data inputs of FF C, E, F and L as pseudo-primary outputs PPOC, PPOE, PPOF and PPOL, respectively, and the data outputs thereof as pseudo-primary inputs PPIC, PPIE, PPIF and PPIL, respectively, with respect to each PO, PPOC, PPOE, PPOF, PPOL to find the number of test sequence generation influence FFs, AIF( 1 ). Here, AIF( 1 ) is 12. Since the sequential depth of the circuit=2, the procedure advances to step SE 7 . The value of n becomes 2 by addition of 1 thereto.

The procedure returns to step SE 3 . Since n=2, FFs to replace with scan FFs are selected in order that the circuit of FIG. 40 , has a 2-fold line-up structure. FIG. 42 ( a ) shows a result of selecting FFs to replace with scan FFs in order that the circuit of FIG. 40 has a 2-fold line-up structure. As shown in FIG. 42 ( a ), FF C, F, and L are selected as FFs to replace with scan FFs. At step SE 4 , as shown in FIG. 42 ( b ), timeframe expansion is performed, in assuming the data inputs of FF C, F and L as PPOC, PPOF and PPOL, respectively, and the data outputs thereof as PPIC, PPIF and PPIL, respectively, with respect to each PO, PPOC, PPOF, PPOL to find the number of test sequence generation influence FFs, AIF( 2 ). Here, AIF( 2 ) is 15. Since AIF( 2 )/AIF( 1 )=15/12=1.2<P=1.6 at step SE 5 , the procedure advances to step SE 6 . Since the sequential depth is 3, the procedure moves to step SE 7 . The value of n becomes 3 by addition of 1 thereto.

The procedure returns to step SE 3 . Since n=3, FFs to replace with scan FFs are selected in order that the circuit, of FIG. 40 has a 3-fold line-up structure. FIG. 43 ( a ) shows a result of selecting FFs to replace with scan FFs in order that the circuit of FIG. 40 has a 3-fold line-up structure. As shown in FIG. 43 ( a ), FF C and L are selected as FFs to replace with scan FFs. At step SE 4 , as shown in FIG. 43 ( b ), timeframe expansion is performed in assuming the data inputs of FF C and L as PPOC and PPOL, respectively, and the data outputs thereof as PPIC and PPIL, respectively, with respect to each PO, PPOC and PPOL to find the number of test sequence generation influence FFs, AIF( 3 ). Here AIF( 3 ) is 24. Since AIF( 3 )/AIF( 1 )=24/12=2.0>P=1.6 at step SE 5 , the procedure advances to step SE 8 . The value of n becomes 2 by subtraction of 1 therefrom, and the procedure is finished. In other words, the 2-fold line-up structure is selected as optimal.

Embodiment 9

FIG. 44 is a model diagram showing an integrated circuit of three identical blocks A. In design for testability for an integrated circuit composed of a plurality of blocks, selecting FFs to replace with scan FFs to attain testability in each block will not always lead to attain testability for the integrated circuit. For instance, when the blocks have a loop structure as shown in FIG. 44 , each block having n-fold line-up structure will not always lead to allow the integrated circuit to have n-fold line-up structure.

Then, in the ninth embodiment of the present invention, before selecting FFs to replace with scan FFs in each block, a FF reachable from block output by passing merely through combinational circuits is selected as FF to replace with scan FF in each block. In the present embodiment, a FF reachable from a block output by passing merely through combinational circuits is called output part FF.

FIG. 45 is a flowchart showing procedures in a method of design for testability of the present embodiment. As shown in FIG. 45 , step SF 1 is a pre-step of selecting the output part FF as FF to replace with scan FF in each block. Thereafter, scan FF selection for the n-fold line-up structure as described in the first embodiment is performed in each block (step SF 2 ), as shown in FIG. 45 ( a ), or scan FF selection based on the state justification of L/H FFs as described in the second embodiment is performed in each block (step SF 3 ), as shown in FIG. 45 ( b ).

FIG. 46 shows an example result of step SF 1 on a specific circuit block. A-I are combinational circuits. r 1 -r 7 are FFs. BPI 1 -BPI 4 are block inputs. BPO 1 and BPO 2 are block outputs. FFs reachable from the block output BPO 1 by passing merely through combinational circuits are FF r 4 and r 6 . And, FFs reachable from the block output BPO 2 by passing merely through combinational circuits are FF r 5 and r 7 . Accordingly, FF r 4 , r 5 , r 6 , and r 7 are recognized as output part FFs, as shown in FIG. 46 and these FFs are then selected as FFs to replace with scan FFs. Thereafter, step SF 2 is performed, as in the first embodiment.

FIG. 47 shows another example result of step SF 1 on a specific circuit block. A-I are combinational circuits. r 1 , r 2 and r 5 are FFs. r 3 , r 4 , r 6 , and r 7 are L/H FFs. BPI 1 -BPI 4 are block inputs. BPO 1 and BPO 2 are block outputs. FFs reachable from the block output BPO 1 by passing merely through combinational circuits are FF r 4 and r 6 . And, FFs reachable from the block output BPO 2 by passing merely through combinational circuits are FF r 5 and r 7 . Accordingly, FF r 4 , r 5 , r 6 , and r 7 are recognized as output part FFs, as shown in FIG. 47 and these FFs are then selected as FFs to replace with scan FFs. Thereafter, step SF 3 is performed, as in the second embodiment.

In accordance with the present embodiment, output part FF that is reachable from the block output by passing merely through combinational circuits is first selected as FF to replace with scan FF, in each block and then FFs to replace with scan FFs are selected in each block by methods of design for testability of the first or second embodiment. This guarantees high fault efficiency at selecting FFs to replace with scan FFs. And, test input pattern generation complexity can be easier from circuit scale level down to block scale level.

Also, instead of output part FF, input part FF that is reachable from the block input by passing merely through combinational circuits may be selected as FFs to replace with scan FFs. However, since in general the number of input part FFs is greater than the number of output part FFs, it is preferred that output part FFs are selected as FFs to replace with scan FFs.

Embodiment 10

FIG. 48 is a flowchart showing procedures in a method of test sequence generation according to a tenth embodiment of the present invention. In accordance with the present method of test sequence generation, test sequence is generated for a given circuit while sequentially compaction storing test sequences generated for respective faults in buffers.

In FIG. 48 , SG 1 is the step of examining the given circuit for a loop. In the presence of a loop, the procedure moves to step SG 2 , while in the absence of a loop, the procedure moves to step SG 3 . SG 2 is the step of setting a predetermined value as buffer length, providing a single buffer having such a buffer length, and initializing all data thereof to “don't care”. SG 3 is a step of setting a value obtained by addition of 1 to the sequential depth of the circuit as buffer length, providing a single buffer having such a buffer length, and initializing all data thereof to “don't care”.

After setting of the buffer lengths at steps SG 2 or SG 3 , the maximum number of buffers is designated at step SG 4 .

SG 5 is the step of determining whether test sequence has been generated (ATPG) for every fault. If YES, the procedure advances to step SG 6 , while if NO, the procedure advances to step SG 7 . At step SG 6 , a fault simulation is performed using all the test sequences stored in a group of buffers.

SG 7 is the step of selecting a single fault as a target fault for which no test sequences have been generated. SG 8 is the step of generating a test sequence for the target fault selected at step SG 7 .

SG 9 is the step of determining whether the test sequence generated at step SG 8 has a greater length than buffer length set at step SG 2 or SG 3 . If YES, the procedure moves to step SG 10 , while if NO, the procedure moves to step SG 11 . Note that in the absence of a loop in the circuit the test sequence length does not exceed the buffer length, which is set to the sequential depth plus 1 at step SG 3 . At step SG 10 , a fault simulation is performed using the generated test sequence. At step SG 11 , the generated test sequence is compaction stored in the buffer. The details of step SG 11 are described later.

SG 12 is the step of determining whether the number of buffers is greater than the maximum number of buffers designated at step SG 4 at compaction storage at step SG 11 . If YES, the procedure moves to step SG 13 , while if NO, the procedure returns to step SG 5 . At step SG 13 , a fault simulation is preformed using a test sequence held in a buffer with the minimum number of “don't care”s and this buffer is deleted.

FIG. 49 is a flowchart showing the details of step SG 11 of FIG. 48 . SG 11 a is the step of determining whether compaction storage of the test sequence generated at step SG 8 has been tried to every buffer. If compaction storage has been tried to every buffer, the procedure advances to step SG 11 , while if not, the procedure moves to step SG 11 b.

SG 11 b is the step of selecting one buffer to which compaction storage has not been tried. SG 11 c is the step of determining whether compaction storage has been tried on every start position, for the buffer selected at step SG 11 b. If YES, the procedure returns to step SG 11 a, while if not, the procedure advances to step SG 11 d. SG 11 d is the step of selecting a start position to which compaction storage is tried. It is possible to select (BL−TPL+1) start positions per buffer, where BL is the buffer length and TPL the test sequence length.

SG 11 e is the step of determining whether compaction storage in the buffer selected at step SG 11 b is successful from the start position selected at step SG 11 d. Step SG 11 e determines whether the test sequence can be compaction stored based on the compaction rule, where the test sequence's “0” is compactible with “0” or “don't care” but incompactible with “1” in-buffer, and the test sequence's “1” is compactible with “1” or “don't care” but incompactible with logical “0” in-buffer.

When step SG 11 e determines that the test sequence can be compaction stored, the test sequence is compaction stored in the buffer based on the foregoing compaction rules at step SG 11 f.

When the test sequence cannot be compaction stored in any buffer, at step SG 11 g a new buffer in which to store the test sequence is additionally provided. SG 11 h is the step of sorting the buffers in order of the number of “don't care”s.

The method of test sequence generation of FIG. 48 and FIG. 49 will be described by reference to FIGS. 50-53 . Here, suppose that a target circuit has no loops and has a sequential depth of 3, and has four faults, i.e., FAULT-a, FAULT-b, FAULT-c, and FAULT-d, as undetected faults. In FIGS. 50-53 , X 1 -X 3 are primary inputs. A and B are buffers. In buffers, 0 represents logic value “0”. 1 represents logic value “1”. X represents “don't care”.

Since the target circuit has no loop, at step SG 3 , a value of 3 obtained by addition of 1 to the sequential depth is set as buffer length. As shown in FIG. 50 ( a ), buffer A whose buffer length is 3 is generated and all data of buffer A are initialized to “don't care”. At step SG 4 , a value of 2 is designated as the maximum number of buffers.

Next, at step SG 8 , a test sequence is generated for FAULT-a selected at step SG 7 . At step SG 11 , the test sequence generated for FAULT-a is compaction stored in the buffer A. Such compaction storage is successful based on the foregoing compaction rules. Because 0 is the only possible start position since the buffer length is equal to the length of test sequence for FAULT-a and, in addition, all the data of the buffer A have been initialized to “don't care”. As a result, as shown in FIG. 50 ( b ), the test sequence for FAULT-a is compaction stored in the buffer A.

Subsequently, at step SG 8 , a test sequence is generated for FAULT-b selected at step SG 7 . The buffer length is equal to the length of the test sequence for FAULT-b and 0 is therefore the only possible start position. As shown in FIG. 51 ( a ), compaction storage based on the foregoing compaction rules is unsuccessful. A new buffer B in which to store the test sequence is additionally provided, as shown in FIG. 51 ( b ). The number of “don't care”s in the buffer A is four and the number of “don't care”s in the buffer B is five. Accordingly, the buffers A, B are sorted in the order of B, A.

Subsequently, at step SG 8 , a test sequence is generated for FAULT-c selected at step SG 7 . The test sequence generated for FAULT-c has a length of 2, and 0 and 1 are therefore the only possible start positions. As shown in FIG. 52 ( a ), based on the foregoing compaction rules, compaction storage of the test sequence is tried to both the buffers A and B. In the buffer B, compaction storage is unsuccessful, not only when the start position is 0 but also when the start position is 1. In the buffer A, compaction storage is unsuccessful when the start position is 0 but successful when the start position is 1. As a result, the test sequence generated for FAULT-c is compact stored in the buffer A with the start position at 1 (see FIG. 52 ( b )). The number of “don't care”s in the buffer A is one and the number of “don't care”s in the buffer B is five. Accordingly, the buffer B, A are sorted in the order of B, A.

Subsequently, at step SG 8 , a test sequence is generated for FAULT-d selected at step SG 7 . The length of the test sequence generated for FAULT-d is equal to the buffer length and 0 is therefore the only possible start position. Based on the foregoing compaction rules, compaction storage of the test sequence is tried to both the buffers A and B. Compaction storage is unsuccessful in both the buffers A and B. A new buffer C in which to store the test sequence is additionally provided. The number of “don't care”s in buffer A is one, and the number of “don't care”s in buffer B is five, and the number of “don't care”s in buffer C is three. Accordingly, the buffers A, B, C are sorted in the order of B, C, A.

Since the number of buffers exceeds 2 that is the maximum number of buffers (step SG 12 ), each “don't care” is randomly replaced with either “0” or “1” in a buffer with the minimum number of “don't care” s of the buffers A, B, and C, i.e., the buffer A, to execute a fault simulation. The buffer A is then deleted and the buffer C, in which the test sequence for FAULT-d is stored, serves now as a new buffer A (step SG 13 ).

Since test sequences are generated for all undetected faults (step SG 5 ), each “don't care” is randomly replaced with either “0” or “1” in the buffers A and B, to execute a fault simulation (step SG 6 ).

In accordance with the present embodiment, compaction is performed dynamically with buffers in test sequence generation. This achieves generation of shorter test sequences.

Embodiment 11

FIG. 54 is a flowchart showing procedures in a method of test sequence generation according to an eleventh embodiment of the present invention. In accordance with the present method of test sequence generation, test sequence is generated for a given circuit while sequentially compaction storing test sequences generated for respective faults in buffers. The present embodiment is different in algorithm for compaction storage of test sequences in buffer from the tenth embodiment.

In FIG. 54 , SH 1 is the step of examining the given circuit for a loop. In the presence of a loop, the procedure moves to step SH 2 , while in the absence of a loop, the procedure moves to step SH 3 . SH 2 is the step of setting a predetermined value L as buffer length, and providing N buffers having such a buffer length L, wherein the number N is a positive integer, and initializing all data thereof to “don't care”. SH 3 is the step of setting a value obtained by addition of 1 to the sequential depth of the circuit as buffer length, and providing N buffers having such a buffer length, and initializing all data thereof to “don't care”.

SH 4 is the step of determining whether the “don't care” ratio in the N buffers exceeds a predetermined upper limit P. If the “don't care” ratio exceeds the upper limit P, at step SH 5 , each “don't care” is replaced with either “0” or “1” in the N bummers to execute a fault simulation. At step SH 6 , all the data of the buffers are initialized to “don't care”. While, if the “don't care” ratio does not exceed the upper limit P, the procedure moves to step SH 7 .

SH 7 is the step of determining whether test sequences have been generated for all the faults. If YES, at step SH 8 , each “don't care” is replaced with either “0” or “1” to execute a fault simulation.

SH 9 is the step of selecting a single fault as a target fault for which no test sequences have been generated. SH 10 is the step of generating a test sequence for the target fault selected at step SH 9 .

SH 11 is the step of determining whether the test sequence generated at step SH 10 has a greater length than the buffer length set at step SH 2 or SH 3 . If YES, the procedure moves to step SH 12 , while if NO, the procedure moves to step SH 13 . Note that in the absence of a loop in the circuit the test sequence length does not exceed the buffer length, which is set to the sequential depth plus 1 at step SH 3 . At step SH 12 , a fault simulation is performed with the generated test sequence. In step SH 13 , the generated test sequence is compaction stored in the N buffers. The details of step SH 13 are described later. Step SH 14 determines whether such compaction storage is performed successfully. If YES, the procedure returns to step SH 4 , while if NO, a fault simulation is executed using this the test sequence at step SH 12 before returning to step SH 4 .

FIG. 55 is a flowchart showing the details of step SH 13 of FIG. 54 . In FIG. 55 , SH 13 a is the step of initializing a start position, i, to 0. SH 13 b -SH 13 e are the steps of determining whether the test sequence can be compaction stored in the N buffers, while incrementing the start position i. If YES, the test sequence is compaction stored in the N buffers, while if NO, processing is finished.

The method of test sequence generation of FIG. 54 and FIG. 55 will be described by reference to FIGS. 56-59 . Here, suppose that a target circuit has no loops and has a sequential depth of 2 . The number of buffers N is two. The upper limit P of the “don't care” ratio is 60%. The target circuit has four faults, i.e., FAULT-a, FAULT-b, FAULT-c, and FAULT-d, as undetected faults. In FIGS. 56-59 , X 1 -X 3 are primary inputs. In buffers, 0 represents logic value “0”. 1 represents logic value “1”. X represents “don't care”. Also, a set of N buffers is regarded as a single buffer in this embodiment.

Since the target circuit has no loops at SH 3 , a value of 3 obtained by addition of 1 to the sequential depth is set as buffer length. As shown in FIG. 56 ( a ), two buffers whose a buffer length is 3 are generated. These two buffers being regarded as a single buffer whose buffer length is 6, all data stored therein are initialized to “don't care”.

At step SH 4 , whether the “don't care” ratio exceeds the upper limit P is determined. Here, the “don't care” ratio is 0% and the procedure advances to step SH 7 .

Next, a test sequence is generated at step SH 10 for FAULT-a selected at step SH 9 . At step SH 13 , the test sequence thus generated is compaction stored in the buffer. As shown in FIG. 56 ( a ), such compaction storage is successful when the start position is 0. Therefore, the generated test sequence is compaction stored as shown in FIG. 56 ( b ). The “don't care” ratio is 5/18 which is smaller than the upper limit P (=60%).

Next, a test sequence is generated at step SH 10 for FAULT-b selected at step SH 9 . At step SH 13 , the test sequence thus generated is compaction stored in the buffer. As shown in FIG. 57 ( a ), such compaction storage is successful when the start position is 1. Therefore, the generated test sequence is compaction stored as shown in FIG. 57 ( b ). The “don't care” ratio is 8/18 which is smaller than the upper limit P (=60%).

Next, a test sequence is generated at step SH 10 for FAULT-c selected at step SH 9 . At step SH 13 , the test sequence thus generated is compaction stored in the buffer. As shown in FIG. 58 ( a ), such compaction storage is successful when the start position is 4. Therefore, the generated test sequence is compaction stored as shown in FIG. 58 ( b ). The “don't care” ratio is 12/18 which is greater than the upper limit P (=60%). A fault simulation is accordingly performed at step SH 5 , and all the buffer data are initialized to “don't care” at step SH 6 .

Subsequently to the above, a test sequence is generated at step SH 10 for FAULT-d selected at step SH 9 . At step SH 13 , the test sequence thus generated is compaction stored in the buffer. As shown in FIG. 59 ( a ), such compaction storage is successful when the start position is 0. Therefore, the generated test sequence is stored as shown in FIG. 59 ( b ). The “don't care” ratio is 6/18 which is smaller than the upper limit P (=60%).

In accordance with the present embodiment, it is determined whether further compaction storage can be performed successfully according, to “don't care” ratio in buffer. This achieves dynamic test sequence compaction with high efficiency. In addition, it becomes possible to perform compaction storage over a plurality of buffers by regarding a plurality of buffers as a single buffer, therefore compaction rates increases.

Claims

1. A method of test sequence generation for an integrated circuit, comprising:a buffer length set step of setting a buffer length for a buffer for storing a test sequence; and a test sequence compaction step of generating a test sequence for said integrated circuit, in performing sequentially compaction storage of test sequences for respective faults in buffers having said buffer length set in said buffer length set step.

2. The method of test sequence generation of claim 1, wherein said buffer length set step is the step of:examining said integrated circuit for a loop, and setting a predetermined value as the buffer length in the presence of a loop and setting a value, obtained by addition of 1 to the sequential depth of said integrated circuit, as the buffer length in the absence of a loop.

3. The method of test sequence generation of claim 1, wherein:said buffer length set step is the step of providing a single buffer having said buffer length, and said test sequence compaction step includes: a first step of generating a test sequence for a fault; a second step of performing compaction storage of the test sequence generated at said first step in said already provided buffer if such compaction storage is carried out successfully and, if not, providing a new buffer having said buffer length and storing said test sequence in said new buffer; and a third step of executing a fault simulation by using a test sequence stored in a buffer with the minimum number of “don't care's” when the number of buffers exceeds a given limit at said second step and deleting said buffer, wherein said first to third steps are iterated for respective faults.

4. The method of test sequence generation of claim 1, wherein:said buffer length set step is the step of providing a given number of buffers having said buffer length, and said test sequence compaction step includes: a first step of generating a test sequence for a fault; a second step of performing compaction storage of the test sequence generated in said first step in said given number of buffers if such compaction storage is carried out successfully and, if not, executing a fault simulation using said test sequence; and a third step of executing a fault simulation by using the test sequence stored in said given number of buffers when the “don't care” ratio in said given number of buffers exceeds a given limit at said second step and initializing all data in said given number of buffers to “don't care”, wherein said first to third steps are iterated for respective faults.

5. The method of test sequence generation according to claim 1, wherein said test sequence operates as input data to said integrated circuit.