Fault simulation method and fault simulator for semiconductor integrated circuit

[0002] The invention relates to a fault simulation method and a fault simulator which prepare a fault list occurring in a semiconductor integrated circuit which can be detected with a test pattern sequence.

[0003] According to a conventional practice, a fault simulation of a semiconductor integrated circuit comprises assuming a fault within the integrated circuit, calculating an output value from an output terminal in response to a given test pattern by a logic simulation, determining a fault in which a change occurs as compared with an output value of a fault-free arrangement, assembling such results into a table of correspondence between assumed faults and input/output logic values, which is commonly referred to as a fault dictionary, thus preparing a fault list which can be detected with individual test patterns. When testing a semiconductor integrated circuit, each test pattern is applied to the input of the semiconductor integrated circuit, and the resulting output value and the input are used in making a reference to the fault dictionary in order to estimate whether or not there is a fault in the integrated circuit or/and where in the integrated circuit the fault is located.

[0004] In order to accommodate for a fault which does not give rise to a wrong logic such as a short-circuit fault or a current leak fault, there is proposed a fault simulation method which combines the logic simulation with IDDQ (quiescent power supply current) testing technique. According to this method, a logic signal value on a signal line within the integrated circuit which would occur in response to a test pattern is calculated by the logic simulation. By assuming faults within the integrated circuit to enumerate a variety of faults which satisfy requirements for the occurrence of an abnormality in IDDQ, a list of detectable faults is prepared. When a test pattern is input to the integrated circuit, each signal line in the integrated circuit normally assumes either “0” or “1” logic value. Accordingly, if a short-circuit fault occurs between a signal line having a logic value of “0” and a signal line having a logic value of “1” in the integrated circuit in response to the application of the test patter, an IDDQ abnormality occurs in the integrated circuit. In this manner, for any combination of a signal line having a logic value of “0” and a signal line having a logic value of “1” in the integrated circuit, a short-circuit fault which occurs between the both kinds of signal lines can be detected by the IDDQ testing technique in response to the input test pattern. When these short-circuit faults are enumerated, a fault list which can be detected with the IDDQ testing technique can be prepared.

[0005] However, the fault simulation method which employs the logic simulation can only deal with a fault model in which a signal line is fixed to a given state (either “0” or “1”, or a single stuck-at fault, namely stuck-at 0 or stuck-at 1). This prevents a sensitive simulation of a multiple stuck-at fault in which a plurality of signal lines are fixed to either “0” or “1”,a delay fault, a short-circuit fault between signal lines or the like. Accordingly, this fault simulation method cannot prepare a list of detectable faults for these faults.

[0006] Another difficulty of the fault simulation method which combines the logic simulation with the IDDQ testing technique is the fact that it is incapable of preparing a list of detectable faults for those faults related to a transient phenomenon of a semiconductor integrated circuit such as a delay fault or an open fault which influences upon a delay time or an abnormality in a local or global process parameter (such as sheet resistance, an oxide film thickness of the like) for example, inasmuch as the IDDQ testing technique is directed to determining a power supply current in the stable condition of the semiconductor integrated circuit or is principally directed to short-circuit faults in the circuit.

[0007] Accordingly, there is a need for a fault simulation method capable of preparing a fault list which are detectable with a test pattern sequence for faults including a delay fault, an open fault or a parametric abnormality fault in an integrated circuit.

[0008] It is an object of the invention to provide a fault simulation method and a fault simulator capable of preparing a fault list which are detectable with test pattern sequence for delay faults, open faults and parametric abnormality faults in a semiconductor integrated circuit, by using the IDDT (transient power supply current) testing technique which affords a high level of observability and capable of testing transient phenomena in the circuit in combination with a transition simulation.

[0009] According to the present invention, there is provided a method of preparing a fault list which are detectable with input test pattern sequence, comprising a step of deriving a test pattern sequence formed by two or more test patterns and which is to be applied to a semiconductor integrated circuit under test, a step of performing a transition simulation of an operation of the semiconductor integrated circuit under test when each test pattern in the derived test pattern sequence is applied thereto to calculate a train of transition signal values occurring on signal lines within the semiconductor integrated circuit under test, and a step of preparing a fault list which can be detected by the transient power supply current testing which uses the test pattern sequence, by utilizing the train of transition signal values on signal lines which are calculated by the transition simulation.

[0010] With this method, it is possible to prepare a fault list which are detectable by the transient power supply current testing which uses given test patterns, for those faults which have been difficult to detect in the prior art, namely, delay faults and open faults which give rise to delay faults, thus allowing a substantial improvement in the efficiency of testing against delay faults and open faults.

[0011] In one form of the step of preparing the fault list, the fault list is prepared in unit of a delay fault of a logic gate.

[0012] In another form of the step of preparing the fault list, the fault list is prepared in unit of an open fault in a signal line.

[0013] In a further form of the step of preparing the fault list, the fault list is prepared in unit of a path delay fault on a signal transmission path.

[0014] The present invention also provides a fault simulator which prepares a fault list which are detectable with an input test pattern sequence, comprising a test pattern sequence selector for deriving a test pattern sequence formed by two or more test patterns and which is to be applied to a semiconductor integrated circuit under test, a transition simulator which receives the derived test pattern sequence and performs a transition simulation of the operation of the semiconductor integrated circuit under test when each test pattern of the sequence is input thereto to calculate a train of transition signal values occurring on signal lines within the semiconductor integrated circuit under test, and a fault list preparing unit for preparing a fault list which are detectable by the transient power supply current testing which uses the test pattern sequence, by utilizing the train of transition signal values on signal lines which are calculated by the transition simulator.

[0015] The fault simulator allows a fault list which are detectable by the transient power supply current testing which uses certain test pattern to be prepared, for a delay fault or an open fault which leads to a delay fault, either of which has been difficult in the prior art to detect, thus permitting the testing efficiency for the delay faults and the open faults to be significantly improved.

[0016] In the description to follow, the principle of the present invention will be described in terms of a CMOS integrated circuit which is a most common semiconductor integrated circuit.

[0017] Transient current of CMOS logic gate

[0018] FIG. 1 shows a transient response of a CMOS inverter shown in FIGS. 1cand d. The transient response has been determined with a circuit simulator. FIG. 1ashows a response of an output voltage V_OUTwith respect to an input voltage V_INin a transient condition, while FIG. 1bshows a response of a current I_DDwhich flows from a power supply into the CMOS inverter. The current I_DDis referred to as a transient current. When the input IN of the inverter transitions from “1” to “0” (see FIG. 1c), n-MOS and P-MOS are only momentarily turned on as long as the input voltage is higher than the threshold voltage of n-MOS and lower than the threshold voltage of P-MOS, whereby a short-circuit current Is flows from a power supply terminal T_VDto a ground GND. An output OUT from an output signal line of the inverter then transitions from “0” to “1”, and accordingly, at the same time as the short-circuit current I_S, a current I_Cwhich is associated with a capacitance charging of a parasitic capacitance C_loadconnected to the output signal line of the inverter flows from the power supply terminal T_VDto the parasitic capacitance C_load. Accordingly, when a falling transition (denoted by suffix “f”) occurs on the input IN of the inverter, the transient current I_Gfwhich flows into the inverter is given by the sum of the short-circuit current I_Sfand the capacitance charging current I_C.

[0019] I_Gf=I_Sf+I_C  (1)

[0020] On the other hand, when the input IN transitions from “0” to “1” or when the output transitions from “1” to “0” (denoted by suffix “r”) (see FIG. 1d), a capacitance discharge occurs from the parasitic capacitance C_loadconnected to the output signal line, thus producing a capacitance discharge current I_D, but the current I_Grwhich flows from the power supply terminal T_VDto the inverter comprises only the short-circuit current I_Sr. Accordingly, a peak of the current T_VDis slightly smaller than the peak of the transient current I_Gfwhich occurs during the falling transition as shown in FIG. 1b.

[0021] I_Gr=I_Sr  (2)

[0022] The CMOS inverter has a transfer characteristic which is such that a triangular pulse current I_Sis produced in response to a change in the input voltage V_IN, as shown in FIG. 2a. Accordingly, when a rising transition occurs on the input of the CMOS inverter, the short-circuit current waveform I_Srwhich flow through the CMOS inverter can be approximated by a triangular pulse, as indicated by “I_S” in FIG. 2b, assuming a ramp transition for the input voltage V_IN. The short-circuit current waveform I_Srwith respect to the rising transition of the input signal as indicated in FIG. 2bcan be given by the following approximations: $\begin{array}{cc}{I}_{\mathrm{Sr}}=\{\begin{array}{cc}0,& t\le \frac{V_{\mathrm{THN}}}{V_{\mathrm{DD}}}{t}_r\\ \frac{V_{\mathrm{DD}}·I_{S\max }}{\left(V_{\mathrm{SP}}-V_{\mathrm{THN}}\right)·t_r}t-\frac{V_{\mathrm{THN}}·I_{S\max }}{\left(V_{\mathrm{SP}}-V_{\mathrm{THN}}\right)},& \frac{V_{\mathrm{THN}}}{V_{\mathrm{DD}}}{t}_r<t\le \frac{V_{\mathrm{SP}}}{V_{\mathrm{DD}}}{t}_r\\ \frac{V_{\mathrm{DD}}·I_{S\max }}{\left(V_{\mathrm{SP}}-V_{\mathrm{DD}}+V_{\mathrm{THP}}\right)·t_r}t-\frac{\left(V_{\mathrm{DD}}-V_{\mathrm{THP}}\right)·I_{S\max }}{\left(V_{\mathrm{SP}}-V_{\mathrm{DD}}+V_{\mathrm{THP}}\right)},& \frac{V_{\mathrm{SP}}}{V_{\mathrm{DD}}}{t}_r<t\le \frac{V_{\mathrm{DD}}-V_{\mathrm{THP}}}{V_{\mathrm{DD}}}{t}_r\\ 0,& t\ge \frac{V_{\mathrm{DD}}-V_{\mathrm{THP}}}{V_{\mathrm{DD}}}{t}_r\end{array}& \left(3\right)\end{array}$

[0023] where I_Smaxrepresents a maximum value of the transient current (short-circuit current) which flows into the CMOS inverter, V_DDrepresents a power supply voltage, V_THNa threshold voltage of n-MOS transistor, V_THPa threshold voltage of a P-MOS transistor and t_ra time interval for the rising transition of the input signal. It is to be noted that V_THPis indicated in its absolute magnitude. To simplify the approximations, it is assumed that the transition of the input voltage V_INbegins at time 0, and the input voltage reaches to V_DDat time t_rwhich represents the completion of the transition.

[0024] Similarly, the short-circuit current waveform I_Sfof the CMOS inverter in response to the falling transition of the input signal can be given by the following approximations: $\begin{array}{cc}{I}_{\mathrm{Sf}}=\{\begin{array}{cc}0,& t\le \frac{V_{\mathrm{THP}}}{V_{\mathrm{DD}}}{t}_f\\ \frac{V_{\mathrm{DD}}·I_{S\max }}{\left(V_{\mathrm{DD}}-V_{\mathrm{THP}}-V_{\mathrm{SP}}\right)·t_f}t-\frac{V_{\mathrm{THP}}·I_{S\max }}{\left(V_{\mathrm{DD}}-V_{\mathrm{THP}}-V_{\mathrm{SP}}\right)},& \frac{V_{\mathrm{THP}}}{V_{\mathrm{DD}}}{t}_f<t\le \frac{V_{\mathrm{DD}}-V_{\mathrm{SP}}}{V_{\mathrm{DD}}}{t}_f\\ \frac{V_{\mathrm{DD}}·I_{S\max }}{\left(V_{\mathrm{THN}}-V_{\mathrm{SP}}\right)·t_f}t-\frac{\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}\right)·I_{S\max }}{\left(V_{\mathrm{THN}}-V_{\mathrm{SP}}\right)},& \frac{V_{\mathrm{DD}}-V_{\mathrm{SP}}}{V_{\mathrm{DD}}}{t}_f<t\le \frac{V_{\mathrm{DD}}-V_{\mathrm{THN}}}{V_{\mathrm{DD}}}{t}_f\\ 0,& t\ge \frac{V_{\mathrm{DD}}-V_{\mathrm{THN}}}{V_{\mathrm{DD}}}{t}_f\end{array}& \left(4\right)\end{array}$

[0025] where t_frepresents a time interval for the falling transition of the input signal. For the sake of convenience, times for the start of the rising transition of the power supply current, the maximum value I_Smaxand the end of the falling transition are indicated in parentheses in FIG. 2b.

[0026] Denoting a voltage change on the output signal line by v_out(t), the charging current I_Cto the parasitic capacitance C_loadconnected to the output signal line of the CMOS inverter is given by the following equation: $\begin{array}{cc}{I}_C=C_{\mathrm{load}}\frac{d{v}_{\mathrm{out}}\left(t\right)}{dt}& \left(5\right)\end{array}$

[0027] Such equations can be similarly determined for logic gates other than the inverter.

[0028] Assuming that the transient current I_Gwhich flows into a logic gate is substantially short-circuit current, it can be approximated by a triangular pulse as indicated by I_sin FIG. 2b. In actuality, the transient current waveform I_Gof the CMOS inverter represents a triangular pulse as shown in FIG. 1b. Accordingly, the transient current I_Gof the logic gate increases monotonously until the maximum value I_Smaxis reached whereupon it decreases monotonously, as shown in FIG. 2b. The transient current I_Greaches its maximum value I_Smaxwhen the input voltage V_INreaches the switching voltage V_SPfor the logic gate. Specifically, as shown in FIG. 2b, the time when the current I_Greaches its peak coincides with the time of input transition of the logic gate. Because the logic gate has a delay time, the time for the output transition of the logic gate will be slightly lagging with respect to the time for the input transition. Thus, the time when the current I_Greaches its peak slightly precedes the time for the output transition of the logic gate. In this instance, the falling edge of the transient current waveform I_Gmay be considered as coincident with the time for the output transition. In addition, the transient current waveform I_Gof the logic gate has a pulse width which is proportional to the transition time of the input voltage (for example the time interval t_rfor the rising transition).

[0029] In the above description, it has been assumed that the transient current I_Gwhich flows into the logic gate is substantially a short-circuit current I_S. However, with a miniaturization of a CMOS manufacturing process, the wiring delay will be more dominant over the gate delay. This means that assuming a constant transition time interval for the input voltage, the proportion of the charging current I_Cto the output signal line will be greater than the proportion of the short-circuit current I_Sin the transient current I_Gwhich flows into the CMOS logic gate. Accordingly, the time when the transient current waveform of the logic gate reaches its peak depends on the ratio of the charging current I_Cto the short-circuit current I_S. When I_Cis less than I_S, the peak of the transient current waveform I_Gcoincides with the peak of I_S. Because the peak of I_Scoincides with the transition time interval of the input voltage, it follows that the peak of I_Cprecedes the time for the output transition of the logic gate. Conversely, if I_Cis greater than Is, the peak of the transient current waveform coincides with the peak of I_C. Because the charging current I_Crelates to the voltage transition on the output signal line, the peak of I_Gis substantially coincident with the time for the output transition of the logic gate.

[0030] It will be seen from the equations (1), (2), (3) and (4) that the pulse width of the transient current waveform I_Gof the logic gate is proportional to the transition time of the input voltage (for example, the rising transition time t_r). Here it is to be noted that because a change in the transition time of the output voltage is sufficiently small relative to a change in a transition time of the input voltage, I_Cterm appearing in the equation (1) has been neglected.

[0031] Integral of transient power supply current

[0032] Using the equations (3) and (4), time integrals Q_Srand Q_Sfof the short-circuit currents I_Srand I_Sfare given as follows: $\begin{array}{cc}{Q}_{\mathrm{Sr}}={\int }_{-\infty }^{\infty }{I}_{\mathrm{Sr}}dt=\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}{t}_r& \left(6\right)\\ Q_{\mathrm{Sf}}={\int }_{-\infty }^{\infty }{I}_{\mathrm{Sf}}dt=\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}{t}_f& \left(7\right)\end{array}$

[0033] Accordingly, the integral Q_Sof the short-circuit current through the logic gate as it is switched is given as follows: $\begin{array}{cc}{Q}_S\equiv {\int }_{-\infty }^{\infty }{I}_{S}dt=\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}{t}_T\propto t_T& \left(8\right)\end{array}$

[0034] where t_Trepresents a transition time for the input signal. Thus, the integral Q_Sof the short-circuit current I_S(I_Sror I_Sf) which flows into the logic gate is proportional to the time t_Tfor the input transition of the logic gate. It will also be seen that Q_Sdoes not depend on whether the transition of the input signal is rising or falling.

[0035] From the equation (5), it is seen that the integral Q_Cof the charging current I_Cto an output load capacitance C_loadof a CMOS inverter is given by the equation indicated below. $\begin{array}{cc}\begin{array}{c}{Q}_C={\int }_{-\infty }^{\infty }{I}_{C}dt={\int }_{-\infty }^{\infty }{C}_{\mathrm{load}}\frac{d{v}_{\mathrm{out}}\left(t\right)}{dt}dt\\ ={C_{\mathrm{load}}\left[v_{\mathrm{out}}\left(t\right)\right]}_{-\infty }^{\infty }=C_{\mathrm{load}}\left(V_{\mathrm{DD}}-0\right)=C_{\mathrm{load}}{V}_{\mathrm{DD}}\end{array}& \left(9\right)\end{array}$

[0036] and it is seen that it does not depend on the time t_Tfor the input transition of the CMOS inverter.

[0037] Consequently, the integrals Q_Gfand Q_Grof the transient currents I_Gfand I_Grpassing through the logic gate are determined from the equations (1), (2), (8) and (9) as follows: $\begin{array}{cc}\begin{array}{c}{Q}_{\mathrm{Gf}}={\int }_{-\infty }^{\infty }\left(I_{\mathrm{Sf}}+I_C\right)dt=\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}{t}_T+\\ C_{\mathrm{load}}{V}_{\mathrm{DD}}\propto t_T\end{array}& \left(10\right)\\ Q_{G_r}={\int }_{-\infty }^{\infty }{I}_{\mathrm{Sr}}dt=\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}{t}_T\propto t_T& \left(11\right)\end{array}$

[0038] Thus, an integral of the transient current through the logic gate is proportional to the time for the input transition of the logic gate.

[0039] Transient power supply current I_DDIin CMOS integrated circuit

[0040] A power supply current in a CMOS integrated circuit represents a power supply current which flows into the CMOS integrated circuit, and is represented by a sum of currents which flow through individual logic gates within the integrated circuit.

[0041] A CMOS integrated circuit shown in FIG. 3acomprises a combination of four (G₁, G₂, G₃, G₄) of the inverter shown in FIG. 1cand which are connected in tandem. Transient currents I_G1, I_G2, I_G3, I_G4which flow through the inverters G₁, G₂, G₃, G₄are normally fed from a single power supply terminal T_VD. Accordingly, the transient power supply current response of the integrated circuit is represented by a sum of transient currents flowing through the individual logic gates as shown in FIG. 3cand as given below. $\begin{array}{cc}{I}_{\mathrm{DDT}}=\sum _{n=1}^N{I}_{\mathrm{Gn}}& \left(12\right)\end{array}$

[0042] where N represents the number of logic gates which are switched by input test pattern sequence. In the example shown in FIG. 3a, N=4. FIG. 3bshows a relationship between an input voltage and an output voltage of each logic gate.

[0043] Accordingly, when there is a fault in the integrated circuit which causes a change in the transient current of the logic gate and when there is a test pattern sequence which is capable of activating the fault, the fault can be detected by an observation of I_DDTof the integrated circuit.

[0044] An integrated value Q_DDTof the transient power supply current I_DDTcan be represented as a sum of integrated values Q_Gn(1≦n≦N) of currents which pass through individual logic gates. $\begin{array}{cc}\begin{array}{c}{Q}_{\mathrm{DDT}}={\int }_{-\infty }^{\infty }{I}_{\mathrm{DDT}}dt={\int }_{-\infty }^{\infty }\left(\sum _{n=1}^N{I}_{\mathrm{Gn}}\right)dt\\ =\sum _{n=1}^N{\int }_{-\infty }^{\infty }{I}_{\mathrm{Gn}}dt=\sum _{n=1}^N{Q}_{\mathrm{Gn}}\end{array}& \left(13\right)\end{array}$

[0045] In the example of FIG. 3a, the integrated value Q_DDTof the transient power supply current I_DDTis represented as a sum of integrated values (Q_G1, Q_G2, Q_G3, Q_G4) of the currents (I_G1, I_G2, I_G3, I_G4) passing through the individual inverters. Accordingly, if there is a fault within the integrated circuit which causes a transient current through the logic gate or its time integral to change and if there is a test pattern which is capable of activating the fault, the fault can be detected from an observation of the integrated value Q_DDTof the transient power supply current I_DDTof the integrated circuit.

[0046] Detection of gate delay fault

[0047] A gate delay fault is a fault which causes a signal propagation delay from an input to an output of a logic gate to increase. A fault which causes a rising transition of an output from the logic gate to retard is referred to as a slow-to-rise fault while a fault which causes a falling transition to retard is referred to as a slow-to-fall fault. When a gate delay fault occurs through a logic gate disposed within a CMOS integrated circuit, the output transition time of the logic gate is retarded. Accordingly, a logic gate which receives at its input an output from the faulty logic gate will have a retarded timing for the peak of the transient current. Accordingly, if a test pattern sequence could activate a faulty logic gate and switch another logic gate which is driven by the faulty gate, the gate delay fault can be detected from an observation of a lag in the current peak of I_DDTof the CMOS integrated circuit.

[0048] Detection of open fault

[0049] Next an open fault which causes a delay fault will be defined. An open fault refers to an unintended electrical discontinuity, specifically indicating that a signal line is divided into two or more distinct signal lines. An open fault includes a breaking as may be caused by a poor contact due to an absence of metal or the presence of an oxide film, a breaking of a metal wiring due to poor patterning or etching, and a breaking in a diffusion layer or a polysilicon due to a mask failure. An open fault can be categorized into two types, including an open fault in which a wiring L is divided in two spaced portions L1 and L2, as shown in FIG. 4aand where an input V_INto one end of the wiring L cannot appear at an output V_OUTat the other end of the wiring L to cause a “logic fault” as shown in FIG. 4b, and another open fault in which the divided wiling portions L1 and L2 are spaced very closely as shown in FIG. 4cand where an input signal V_INapplied to one end of the wiring L flows as a tunnel current and a transition signal appears at the other end of the wiring L with a delay to cause a “delay fault” as shown in FIG. 4d. An open fault which causes a logic fault involves a large scale of disconnection, so that there is no current flow if a voltage is applied across the signal lines (divided wiring portions L1 and L2) located on the opposite sides of the fault, and accordingly, there takes place no charging or discharge of the parasitic capacitance C_loadin response to a signal transition, thus causing a logic fault in which the logic is fixed to a given value. By contrast in an open fault which causes a delay fault, a very weak current flows when a voltage is applied across the signal lines (L2 and L2) located on the opposite sides of the fault, but the magnitude of the current is less than a normal current flow, retarding the charging or discharge of the parasitic capacitance C_loadin response to the signal transition, thus increasing a delay time of the circuit. An open fault which causes a delay fault includes a resistive open fault in which a resistance between signal lines L1 and L2 becomes larger than a normal value as a result of a poor contact or the resistance of the signal line L becomes larger than a normal value as a result of a failure occurring in the signal line L, and a minuscule open fault (<100 nm) in which a very weak leak current flows across two broken signal line portions L1 and L2 by the tunnel effect. A tunnel current through a minuscule open fault is described, for example, in C. L. Henderson, J. M. Soden and C. F. Hawkins, “The Behavior and Testing Implications of CMOS IC Logic Gate Open Circuits” Proceedings of IEEE International Test Conference, pp.302-310, 1991. An open fault which causes a delay fault is dealt with herein and will be simply referred to as an open fault.

[0050] With the open fault, there is a small current flow through the fault, and thus it can be modeled by resistive element R_openhaving an increased resistance. FIG. 5ashows an example of a CMOS integrated circuit having an open fault The CMOS integrated circuit comprises a pair of inverters G₁, G₂, and an open fault exists in a signal line M which connects between the inverters G₁and G₂. The location of disconnection can be regarded as being equivalent to a connection through the resistor R_open. When the inverter G₁switches to produce a signal transition on the signal line M, the presence of the open fault causes a time interval required for an input parasitic capacitor C_inof the inverter G₂to be charged or discharged increases (in proportion to the product of the resistance of the open fault and the parasitic capacitance), and accordingly, a signal transition on a signal line M′ which is located on the inverter G₂side of the equivalent resistive element R_openis retarded as shown in FIG. 5crelative to an output from the inverter G₁(FIG. 5b). Accordingly, as shown in FIG. 6, the transient current waveform (shown in dotted lines) of the inverter G₂having the open fault on its input side has a greater width of the current pulse and a greater time integral of the transient current as compared with the transient current waveform (shown in solid line) of the normal inverter. Accordingly, by observing I_DDTor Q_DDTof the CMOS integrated circuit the presence of an open fault in the input stage of the inverter G₂can be detected. An open fault which exists on the output stage of the inverter G₁is equivalent to an open fault present on the input stage of the inverter G₂, and thus can be similarly detected. For an open fault on any signal line in an integrated circuit generally, if an open fault can be activated by a test pattern sequence to cause a switching of a logic gate which is driven by the activated open fault, such open fault can be detected from an observation of I_DDTor Q_DDTof the integrated circuit.

[0051] Path delay fault

[0052] A route within an integrated circuit through which a signal is propagated is referred to as a path. When a delay time from a start point of the path (input signal line) to an end point (output signal line), which is referred to herein as path delay time, becomes greater (or less) than a given value, the integrated circuit exhibits an abnormal behavior, which is referred to as a path delay fault. A path delay fault will now be defined.

[0053] Consider that for a CMOS logic circuit that a path P={g₀, g₁. . . , g_m} is activated using a test pattern sequence T=<v₁,v₂> meaning that it contains two test patterns v₁and v₂and that a voltage signal v₂follows a voltage signal v₁. g₀represents an input signal line of the path P while g₁,g₂, . . . , g_mrepresent output signal lines of logic gates G₁, G₂, . . . , G_mon the path P. At the same time, g₀,g₁, . . . ,g_m−1represent input signal lines to the logic gates G₁,G₂, . . . , G_mon the path P. Denoting the time for a signal transition on each signal line g₀,g₁. . . , g_m(the time when the voltage signal passes V_DD/2) by τ₀,τ₁, . . . τ_m, the gate delay time t_gdt(1≦i≦m) of the respective logic gates G₁,G₂, . . . , G_mon the path P is given as follows:

[0054] t_gdt=τ₁−τ_1−l  (14)

[0055] Accordingly, the path delay time t_pdof the path P is given as a sum of the gate delay times t_gdi, as follows: $\begin{array}{cc}{t}_{\mathrm{pd}}=\sum _{i=1}^m{t}_{\mathrm{gdi}}={\tau }_m-{\tau }_0& \left(15\right)\end{array}$

[0056] However, the actual gate delay time t_gdtvaries under the influence of a fault as follows:

[0057] t_gdt=t_gdt,typδ₁, 1≦i≦m   (16)

[0058] where t_gdt,typrepresents a typical value of the gate delay time of the logic gate G_i, and δ_irepresents a variation component in the gate delay time. For example, a breaking fault causes the gate delay time of only a faulty logic gate to increase without increasing the delay time of other logic gates. A parametric fault causes the delay time of every logic gate to increase. With the variation in the gate delay time, the path delay time t_pdsimilarly varies as indicated by an equation given below. $\begin{array}{cc}{t}_{\mathrm{pd}}=t_{\mathrm{pd},\mathrm{typ}}+\Delta =\sum _{i=1}^m\left(t_{\mathrm{gdi},\mathrm{typ}}+{\delta }_i\right)& \left(17\right)\end{array}$

[0059] where t_pd,typrepresents a typical value of a path delay time for the path P and Δ represents a variation component in the path delay time.

[0060] FIG. 7 is a schematic illustration of the fundamental principle of a delay fault testing technique. In order for a semiconductor integrated circuit (DUT) under test shown in FIG. 7ato operate normally, a signal transition which occurs in an input latch must pass through a path P in the semiconductor integrated circuit under test to be transmitted to an output latch in a given time interval. Accordingly, it will be seen from the relationship between an input V_INand an output V_OUTand their relationship with respect to a system clock as shown in FIG. 7b, that the path delay time t_pdof the path P must satisfy the following requirement:

[0061] t_pd+T_SU<T_CLK−T_SKW  (18)

[0062] where T_SUrepresents a set-up time for the signal, T_CLK.the period of the system clock and T_SKWa clock skew of the system clock. The clock skew T_SKWis a jitter of the system clock and represents an amount by which the edge of the system clock changes in the positive or negative direction. The equation (18) can be modified as follows:

[0063] t_pd<T_CLK−T_SKW−T_SU≡T′  (19)

[0064] Thus, the path delay time t_pdof the path P must be less than a time interval T′ which results when margins such as the set-at time T_SUand the clock skew T_SKWare subtracted from the clock period T_CLK. If t_pdis greater than T′, the signal transmission along the path P cannot catch up with the system clock, and the circuit cannot operate properly. This condition is defined as a delay fault. In other words, the path P is defined as having a delay fault when t_pdis greater than a given time interval T′. T′ represents an upper limit of a permissible delay time.

[0065] Detection of path delay (utilizing a pulse width of transient power supply current)

[0066] Since the peak or the falling edge of the transient current waveform of the logic gate corresponds to the time for the output transition of the logic gate, it follows that the last peak (or the last falling edge) of the transient power supply current waveform of the CMOS integrated circuit coincides with the time of the output transition of the logic gate which is the last to be switched in the CMOS integrated circuit Accordingly, by detecting the last peak (or the last falling edge) of the transient power supply current waveform of the CMOS integrated circuit, and comparing the time of detection against the time for the input transition, a path delay time for the integrated circuit can be determined. The time for the last falling edge of the transient power supply current can be determined as a maximum value of the time when the transient power supply current assumes a given current value, for example, from the time for the input transition on the path of the integrated circuit. This current value represents a value of the power supply current when the output voltage from the last logic gate on the path under test reaches a value equal to one-half the supply voltage V_DD, and can be determined by the circuit simulation of the circuit under test or from statistical data obtained with an actual device.

[0067] A path delay fault on the path under test can be detected by comparing the determined path delay time against a given time (such as a period T_CLKof a system clock, for example).

[0068] A technique for detecting a delay fault utilizing the pulse width of a transient power supply current mentioned above will now be described. According to this technique, the pulse width of the power supply current waveform of a circuit under test is measured and compared against a given time interval. The fundamental principle of the technique is illustrated in FIG. 8.

[0069] Assume that in a CMOS logic circuit, a plurality of paths P₁, P₂, . . . , P_nare activated using test pattern sequence T=<v₁, v₂> including two test patterns v₁, v₂. Denoting the time when a j-th logic gate, as counted from the input of a path P_i, is switched by τ_ij, it will be noted that the number of logic gates depends on each of the paths P₁,P₂, . . . , P_n, and the time τ_maxfor the output transition of a logic gate G_finalwhich is the last to be switched among the paths P₁,P₂, . . . , P_nis given by the following equation: $\begin{array}{cc}{\tau }_{\max }=\underset{i,j}{\max }\left\{{\tau }_{\mathrm{ij}}\right\},1\le i\le n,1\le j& \left(20\right)\end{array}$

[0070] Thus, a maximum value of path delay time t_pd.maxamong the paths P₁,P₂, . . . , P_nis determined as a time interval between τ_maxand time τ₀for the input transition, as indicated below.

[0071] t_pd.max=τ_max−τ₀  (21)

[0072] On the other hand, the pulse width t_PWof the transient power supply current waveform of the CMOS logic circuit is defined as a time interval between the time τ₀for the signal transition of the circuit input and the τ_IDDof the last peak (falling edge) of the transient power supply current waveform. Thus

[0073] t_PW≡τ_IDD−τ₀  (22)

[0074] As mentioned previously, the time τ_IDDfor the last peak of the transient power supply current waveform either coincides with the time τ_maxfor the output transition of the logic gate G_finalwhich is the last to be switched or precedes τ_max. Thus, the pulse width t_PWof the transient power supply current waveform corresponds to the delay time t_pd.maxof the path P which is activated by a test pattern T.

[0075] t_PW=τ_IDD−τ₀≦τ_max−τ₀=t_pd.max  (23)

[0076] If t_PWis greater than the upper limit T′ of the permissible delay time, it follows that

[0077] T′<t_PW≦t_{pd max}  (24)

[0078] Thus, the signal transmission cannot catch up with the system clock on the path having the greatest delay time t_pd.max. Thus, there exists a delay fault in the circuit. In this manner, t_PWgreater than T′ indicates the presence of a delay fault on some one of the activated paths, while t_PWless than T′ indicates the absence of a delay fault on any activated path.

[0079] t_PW≦T′ no delay fault

[0080] t_PW>T′ delay fault present   (25)

[0081] In this manner, a delay fault in the circuit can be tested by comparing the pulse width t_PWof the transient power supply current waveform against a given time T′.

[0082] Delay fault detecting technique (utilizing the instantaneous value of the transient power supply current)

[0083] Because the transient power supply current of a logic gate decreases monotonously after its peak as shown in FIG. 1b, it will be seen that the power supply current of the CMOS integrated circuit shown in FIG. 3cdecreases monotonously subsequent to the time for the output transition of a logic gate which is the last to be switched in the integrated circuit. In other words, in a fault-free CMOS integrated circuit, denoting the time for the output transition of the last logic gate to be switched by τ_maxand the instantaneous value of the transient power supply current at time τ_maxby I′, the transient power supply current of the CMOS integrated circuit cannot become greater than I′ subsequent to τ_max.

[0084] Using this principle, a path delay fault in a circuit under test can be detected by measuring the instantaneous value of a transient power supply current of a CMOS integrated circuit at a given time. The current value I′ which is used as a criterion for the judgment of a fault detection is a value of the power supply current at the time when an output from the last logic gate on a path under test assumes one-half the supply voltage, and can be determined by a circuit simulation of the circuit under test or from statistical data using an actual device.

[0085] A technique for detecting a delay fault utilizing the instantaneous value of the transient power supply current will be described. According to this technique, the instantaneous value of the transient power supply current of the circuit under test at a given time is measured, and is compared against the value of the transient power supply current of a golden circuit which is free from a delay fault. The fundamental principle of the technique is illustrated in FIG. 9.

[0086] Assume that in a CMOS logic circuit, a plurality of paths P₁,P₂, . . . ,P_nare activated by test pattern sequence T=<v₁,v₂>. Denoting the time when a j-th logic gate as counted from the input of a path P_iby τ_ij, the time τ_maxfor the output transition of a logic gate G_finalwhich is the last to be switched among the paths P₁,P₂, . . . ,P_nis given by the following equation: $\begin{array}{cc}{\tau }_{\max }=\underset{i,j}{\max }\left\{{\tau }_{\mathrm{ij}}\right\},1\le i\le n,1\le j& \left(26\right)\end{array}$

[0087] Accordingly, a maximum value t_pd.maxof the path delay time among the paths P₁,P₂, . . . ,P_ncan be determined as a time interval between τ_maxand a time τ₀for the input transition, as indicated below.

[0088] t_{pd max}=τ_max−τ₀  (27)

[0089] As mentioned previously, because the time for the output transition of a logic gate coincides with the time for the peak or the falling edge of the transient power supply current of the logic gate, it follows that τ_maxcorresponds to a time τ_IDDfor the last peak or the falling edge of the transient power supply current waveform I_DDTof the circuit. The power supply current I_Gof the logic gate can be approximated by a triangular wave and G_finalrepresents the last gate to be switched, and hence, there is no logic circuit which has a peak of the power supply current subsequent to τ_max. Accordingly, a power supply current waveform function i_DDT(t) is a monotonously decreasing function at time t which is t≧τ_max. Thus, denoting the time function for the power supply current waveform by i_DDT(t) and the instantaneous value of the power supply current at a time τ_maxby I′, the latter is given as follows:

[0090] I′≡i_DDT(τ_max)   (28)

[0091] It follows that at time t which is t≧τ_max,

[0092] i_DDT(t)≦i_DDT(τ_max)=I′,t≧τ_max  (29)

[0093] In order for the circuit to operate properly, t_pd.maxmust be less than the upper limit T′ of the delay time (which is equal to T_CLK−T_SKEW=T_SUas indicated in the equation (19). Thus

[0094] t_pd.max=τ_max−τ₀<T′  (30)

[0095] Accordingly, in the absence of a fault in the circuit, it follows from the equation (29) that at time t which is t=T′+τ₀>τ_max

[0096] i_DDT(T′+τ₀)≦I′  (31)

[0097] If the instantaneous value of I_DDTat T′+τ₀is greater than I′ or

[0098] i_DDT(T′+τ₀)>I′=i_DDT(τ_max)   (32)

[0099] it follows from the equation (29) that because T′+τ₀cannot be greater than τ_max

[0100] τ_max>T′+τ₀  (33)

[0101] t_pd.max=τ_max−τ₀>T′  (34)

[0102] This means that on a path having the greatest delay time t_pd.max, the signal transmission cannot catch up with the system clock. Thus the presence of a delay fault in the circuit is indicated. In is manner, the fact that a value of the transient power supply current I_DDT(T′+τ₀) at time T′+τ₀is greater than I′ indicates the presence of a delay fault in one of the activated paths. Conversely, the fact that i_DDT(T′+τ₀) is less than I′ indicates the absence of a delay fault on any activated path.

[0103] i_DDT(T′+τ₀)≦I′ no delay fault

[0104] i_DDT(T′+τ₀)>I′ delay fault present   (35)

[0105] As discussed above, a delay fault in the circuit can be detected by comparing the instantaneous value of I_DDTat a given time against I_DDTlevel of a fault-free circuit.

[0106] Detection of path delay fault (utilizing a time integral of transient power supply current

[0107] In addition, by using the integrated value Q_DDTof the transient power supply current I_DDT, a path delay fault which is caused by an open fault can be detected. A technique for detecting a path delay fault by utilizing an integrated value of the transient power supply current will now be described. According to this technique, an integrated value of a transient power supply current through a circuit under test is measured, and is compared against a given value to evaluate a path delay fault.

[0108] The integrated value Q_DDTof the transient power supply current I_DDTis represented as a sum of integrated values Q_Gn(1≦n≦N) of currents passing through individual logic gates. $\begin{array}{cc}\begin{array}{c}{Q}_{\mathrm{DDT}}={\int }_{-\infty }^{\infty }{I}_{\mathrm{DDT}}dt={\int }_{-\infty }^{\infty }\left(\sum _{n=1}^N{I}_{\mathrm{Gn}}\right)dt\\ =\sum _{n=1}^N{\int }_{-\infty }^{\infty }{I}_{\mathrm{Gn}}dt=\sum _{n=1}^N{Q}_{\mathrm{Gn}}\end{array}& \left(13\right)\end{array}$

[0109] Since the integral Q_Gn(1≦n≦N) of the current passing through each logic gate is proportional to the time t_Tn(1≦n≦N) for the input transition of each logic gate, as indicated by the equation (10) or (11), Q_DDTis given by a linear polynomial of t_Tn(1≦n≦N). In the example shown in FIG. 3a, Q_DDTis given by a linear polynomial 36, indicated below of the times (t_T1,t_T2,t_T3,t_T4) for the input transitions of the respective inverters (G₁,G₂,G₃,G₄). $\begin{array}{cc}{Q}_{\mathrm{DDT}}=\sum _{n=1}^N{Q}_{\mathrm{Gn}}=\sum _{n=1}^N{Q}_{\mathrm{Sn}}+\sum _{n=1}^N{Q}_{\mathrm{Cn}}=\sum _{n=1}^N{a}_n{t}_{\mathrm{Tn}}+b& \left(36\right)\end{array}$

[0110] In the equation (36), an represents a factor of proportionality between the integral Q_Snof the short-circuit current through the logic gate G_nand the time t_Tnfor the input transition of the logic gate G_n, and b a constant term which is represented by a sum of charging currents Q_Cnwhich flow into the respective logic gates.

[0111] An open fault can be modeled by a large resistance R_opensince a very weak current flows through the fault. FIG. 10ashows an example of CMOS inverter having an open fault at its input. When a signal transition as shown in FIG. 10boccurs on an input signal line A, a signal transition occurring on an signal line A′ which is located subsequent to the location of the fault is retarded due to the open fault as indicated in FIG. 10c. At this time, denoting the resistance of the open fault by R_openand the parasitic capacitance at the input of the inverter by C_in, the time interval t_Tfor the signal transition on the signal line A′ is given as indicated below

[0112] t_T≈t_T,typ+2.2R_openC_in  (37)

[0113] where t_T.typrepresents a typical value of the transition time for the input signal when there is no fault. The transition time t_Tis determined here as a time interval for the voltage to rise from 0.1 V_DDto 0.9 V_DD(or to fall from 0.9 V_DDto 0.1 V_DD) 2.2 R_openC_inrepresents a time interval in order for a voltage across C_inrises from 0.1 V_DDto 0.9 V_DD, and is determined from log_e(0.9 V_DD/0.1 V_DD)×R_openC_m. Thus, an increment in the transition time for the input signal of the inverter is proportional to the resistance R_openof the open fault. Accordingly, when there is an open fault on the input of a k-th inverter on the path under test the integral Q_DDTof the power supply current of the CMOS integrated circuit is determined from the equations (36) and (37) as indicated by an equation (38). Thus, Q_DDTlinearly changes in accordance with the resistance R_openof the open fault and the increment is proportional to the resistance R_openof the open fault. $\begin{array}{cc}\begin{array}{c}{Q}_{\mathrm{DDT}}=\sum _{n=1}^N{a}_n{t}_{\mathrm{Tn}}+b=\left(\sum _{n=1}^N{a}_n{t}_{\mathrm{Tn},\mathrm{typ}}+b\right)+2.2a_k{C}_{\mathrm{in}}{R}_{\mathrm{open}}\\ =Q_{\mathrm{DDT},\mathrm{typ}}+2.2a_k{C}_{\mathrm{in}}{R}_{\mathrm{open}}\propto R_{\mathrm{open}}\end{array}& \left(38\right)\end{array}$

[0114] where Q_{DDT typ}represents a typical value of the integral of the power supply current when there is no fault. 2.2a_kC_inR_openappearing as a second term on the right hand side of the equation (38) is an addition which is based on the input open fault of the k-th inverter. The equation (38) shows a coincidence with a result of simulation performed on a change in Q_DDTwith respect to R_openas shown in FIG. 11. FIG. 11 is a plot of a change in Q_DDTwith respect to the resistance R_openof the open fault when there is an open fault in the input signal line IN2 of the inverter G₂in the circuit shown in FIG. 3.

[0115] A gate delay time t_gdof a logic gate is proportional to the time t_Tfor the transition of an input signal as indicated by an equation (39) below. ( see equations 4.52 and 4.53, “Principles of CMOS VLSI Design-A Systems Perspective”, Second Edition, Addison-Weely Publishing Company, issued 1999, pp. 216-217) $\begin{array}{cc}{t}_{\mathrm{gd}}=t_{\mathrm{gd},\mathrm{step}}+\frac{1}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)t_T& \left(39\right)\end{array}$

[0116] where t_gd.steprepresents a delay time of a fault free inverter with respect to a step input at transition time 0, and V_THa threshold voltage for PMOS or n-MOS, it being understood that V_TH=V_THNfor a rising transition of an input, and V_TH=V_THPfor a falling transition of an input. Accordingly, since a gate delay time t_gdof a logic gate having an open fault which can be modeled by a resistance R_openon an input signal line is given by the equation (37), it can be determined by substitution of the equation (37) into the equation (39), as indicated below. $\begin{array}{cc}\begin{array}{c}{t}_{\mathrm{gd}}=t_{\mathrm{gd},\mathrm{step}}+\frac{t_T}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)\\ =t_{\mathrm{gd},\mathrm{step}}+\frac{t_{T,\mathrm{typ}}+2.2R_{\mathrm{open}}{C}_{\mathrm{in}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)\\ =t_{\mathrm{gd},\mathrm{step}}+\frac{t_{T,\mathrm{typ}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)+\frac{2.2C_{\mathrm{in}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)R_{\mathrm{open}}\\ =t_{\mathrm{gd},\mathrm{typ}}+\frac{2.2C_{\mathrm{in}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)R_{\mathrm{open}}\propto R_{\mathrm{open}}\end{array}& \left(40\right)\end{array}$

[0117] where t_gd.typis a typical value of the gate delay time of a fault-free logic gate. Specifically, the gate delay time t_gdof a logic gate having an open fault varies depending on the resistance R_openof the fault, and the increment δof the gate delay time is proportional to the resistance R_openof the fault. Consequently, when there is a braking fault on the input of any logic gate on the path under test, the path delay time t_pdof the path under test is also proportional to R_open. This is indicated by an equation (41) given below which is obtained by the substitution of the equation (40) into the equation (17). $\begin{array}{cc}\begin{array}{c}{t}_{\mathrm{pd}}=\sum _{i=1}^m{t}_{\mathrm{gdi}}\\ =\sum _{i=1}^m{t}_{\mathrm{gdi},\mathrm{typ}}+\frac{2.2C_{\mathrm{ink}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)R_{\mathrm{open}}\\ =t_{\mathrm{pd},\mathrm{typ}}+\frac{2.2C_{\mathrm{ink}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)R_{\mathrm{open}}\propto R_{\mathrm{open}}\end{array}& \left(41\right)\end{array}$

[0118] This agrees with the result of simulation conducted on a change of t_pdwith respect to R_openas indicated in FIG. 12, which is a plot of a change of t_pdwith respect to the resistance R_openof an open fault when there is an open fault in the input signal line IN2 of an inverter G₂in the circuit shown in FIG. 3a.

[0119] When an open fault exist on the input of a logic gate G_kon the path P, the integral Q_Skof the short-circuit current of the gate G_kis determined from the equations (8) and (37) as given below $\begin{array}{c}{Q}_{\mathrm{Sk}}=\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}{t}_{\mathrm{Tk}}\\ =\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}\left(t_{\mathrm{Tk},\mathrm{typ}}+2.2R_{\mathrm{open}}{C}_{\mathrm{ink}}\right)\\ =\frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}·t_{\mathrm{Tk},\mathrm{typ}}+\\ \frac{I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{2V_{\mathrm{DD}}}·2.2R_{\mathrm{open}}{C}_{\mathrm{ink}}\\ =Q_{\mathrm{Sk},\mathrm{typ}}+\frac{2.2I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)C_{\mathrm{ink}}}{2V_{\mathrm{DD}}}·R_{\mathrm{open}}\end{array}$

[0120] Accordingly, the integral Q_DDTof the transient power supply current of the integrated circuit is determined according to the equation (36) as given below. $\begin{array}{cc}\begin{array}{c}{Q}_{\mathrm{DDT}}=\sum _{n=1}^N{Q}_{\mathrm{Gn}}\\ =\sum _{n=1}^N{Q}_{\mathrm{Sn}}+\sum _{n=1}^N{Q}_{\mathrm{Cn}}\\ =\sum _{n=k}^N{Q}_{\mathrm{Sn},\mathrm{typ}}+Q_{\mathrm{Sk},\mathrm{typ}}+\\ \frac{2.2I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)C_{\mathrm{ink}}}{2V_{\mathrm{DD}}}·R_{\mathrm{open}}+\sum _{n=k}^N{Q}_{\mathrm{Cn}}\\ =\sum _{n=1}^N{Q}_{\mathrm{Sn},\mathrm{typ}}+\sum _{n=1}^N{Q}_{\mathrm{Cn}}+\\ \frac{2.2I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)C_{\mathrm{ink}}}{2V_{\mathrm{DD}}}·R_{\mathrm{open}}\\ =Q_{\mathrm{DDT},\mathrm{typ}}+\frac{2.2I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)C_{\mathrm{ink}}}{2V_{\mathrm{DD}}}·R_{\mathrm{open}}\end{array}& \left(42\right)\end{array}$

[0121] Thus, the integral Q_DDTof the transient power supply current of the integrated circuit is also proportional to the resistance R_openof the open fault.

[0122] Thus, it will be seen from the equations (41) and (42) that the delay time t_pdof the path P having an open fault changes linearly with respect to the integral Q_DDTof the transient power supply current of the CMOS integrated circuit, and this agrees with the result of a simulation conducted on the change of t_pdwith respect to Q_DDTas indicated in FIG. 13, which is a plot of a change t_pdwith respect to the integral Q_DDTof the transient power supply current when an open fault is present on the input signal line IN2 of an inverter G₂in the circuit shown in FIG. 3a.

[0123] The substitution of R_openwhich is determined from the equation (42) into the equation (41) yields an equation (43). $\begin{array}{cc}\begin{array}{c}{t}_{\mathrm{pd}}=t_{\mathrm{pd},\mathrm{typ}}+\frac{2.2C_{\mathrm{ink}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)R_{\mathrm{open}}\\ t_{\mathrm{pd},\mathrm{typ}}+\frac{2.2C_{\mathrm{ink}}}{6}\left(1-2\frac{V_{\mathrm{TH}}}{V_{\mathrm{DD}}}\right)·\\ \frac{\left(Q_{\mathrm{DDT}}-Q_{\mathrm{DDT},\mathrm{typ}}\right)·2V_{\mathrm{DD}}}{2.2I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)C_{\mathrm{ink}}}\\ =t_{\mathrm{pd},\mathrm{typ}}+\frac{V_{\mathrm{DD}}-2V_{\mathrm{TH}}}{3I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}(Q_{\mathrm{DDT}}-\\ Q_{\mathrm{DDT},\mathrm{typ}})\end{array}& \left(43\right)\end{array}$

[0124] Denoting the integral of the transient power supply current when the path delay time t_pdis equal to the upper limit T′ of the permissible delay time by Q_max, an equation (44), given below, is obtained by solving the equation (43) for Q_maxby putting t_pd=T′ and Q_DDT=Q_max. $\begin{array}{cc}{Q}_{\max }=Q_{\mathrm{DDT},\mathrm{typ}}+\frac{3I_{S\max }\left(V_{\mathrm{DD}}-V_{\mathrm{THN}}-V_{\mathrm{THP}}\right)}{V_{\mathrm{DD}}-2V_{\mathrm{TH}}}\left(T^{\prime }-t_{\mathrm{pd},\mathrm{typ}}\right)& \left(44\right)\end{array}$

[0125] The Q_maxrepresents the upper limit of the integral Q_DDTof the transient power supply current of the CMOS integrated circuit which is free from a delay fault. Thus it can be determined that there is no path delay fault in the CMOS integrated circuit when Q_DDTis less than Q_maxwhile there exists a path delay fault due to a break fault in the CMOS integrated circuit when Q_DDTis greater than Q_max.

[0126] Q_DDT≦Q_maxno delay fault

[0127] Q_DDT>Q_maxdelay fault present   (45)

[0128] As discussed above, by comparing the integral Q_DDTof the transient power supply current against the given value Q_max, a delay fault in the circuit can be tested The given value Q_maxcan be determined by a circuit simulation or from the equation (45) using statistical data.

[0129] Features of a delay fault test by way of the transient power supply current

[0130] The transient power supply current is a transient current which flows through a power supply pin of an integrated circuit, and a higher observability is assured for this transient current than for a voltage signal. Accordingly, a delay fault testing method which employs the transient power supply current is capable of assuring a high observability of a delay fault than with a delay fault testing technique which employs a voltage signal. By way of example, a delay fault testing technique which employs a voltage signal can not detect a delay fault unless the voltage signal is transmitted to an output signal line of an integrated circuit. By contrast with a delay fault testing technique which employs a transient power supply current signal, a transient power supply current signal having a pulse width which corresponds to a delay time on a path on which the voltage signal is transmitted is observable if the voltage signal is not transmitted to an output signal line of the integrated circuit, and thus is capable of detecting a delay fault. Concomitantly, there is no need to transmit a voltage signal to an output signal line of an integrated circuit with the delay fault testing technique which employs the transient power supply current signal and accordingly, limitations imposed upon generating test patterns are reduced with this technique in comparison to the generation of test patterns for a conventional delay fault testing technique which employs a voltage signal where it is necessary to transmit the voltage signal to the output signal line of the integrated circuit. This allows the generation of test patterns to be facilitated. In extreme cases, if test pattern sequence are selected at random, the delay fault testing technique which employs the transient power supply current signal is capable of detecting a delay fault on a path which is activated by selected test pattern sequence.

[0131] Fault simulation method (for gate delay fault)

[0132] The fault simulation method according to the present invention which prepares a fault list in unit of a gate delay fault will now be described. FIG. 14 shows an example of a MOS integrated circuit under test. The integrated circuit under test includes four input terminals x₁, x₂, x₃, x₄, two output terminals z₁, z₂, five internal signal nodes n₁, n₂, n₃, n₄, n₅, five logic gates G₁, G₂, G₃, G₄, G₅and output buffers G₆, G₇. Input terminals x₁and x₂are connected to two inputs of NAND gate G₁, the output of which is connected through the signal node n₁, to the input of the inverter gate G₂and to one input of OR gate G₃. The output from the inverter gate G₂is connected through the node n₂to one input of AND gate G₄, and the output from the gate G₃is connected through the node n₃to the other input of AND gate G₄and to one input of AND gate G₅. The input terminal x₃is connected to the other input of OR gate G₃, and the input terminal x₄is connected to the other input of AND gate G₅. Outputs from AND gates G₄and G₅are connected through the nodes n₄and n₅, respectively, and further through the output buffers G₆, G₇to the output terminals z₁and z₂, respectively. The power supply terminals of the gates G₁to G₇are connected to a common power supply, while this is not shown.

[0133] Here it is assumed that the gates G₁, G₂, G₃, G₄, G₅, G₆and G₇have gate delays (propagation delays) equal to 1, 1, 3, 2, 1, 1 and 1, respectively. An example of results of a fault simulation which is performed upon the MOS integrated circuit under test is illustrated in FIG. 15. In FIG. 15, the first column, as counted from the left, represents an identifier of a test pattern sequence; the second column shows input signals applied to the input terminals x₁, x₂, x₃, x₄of the CMOS integrated circuit under test; the third column shows a train of transition signal values which appear on the internal signal nodes n,, n₂, n₃, n₄, n, of the CMOS integrated circuit under test when each test pattern sequence is applied to the input terminals x₁, x₂, x₃, x₄; and the fourth column shows a train of transition signal values which occur on the output terminals z₁, z₂of the CMOS integrated circuit under test which result from the transition signal values on these nodes. Signals “L”, “H”, “R”, and “F” entered in the second, the third and the fourth column in FIG. 15 represent a normally low level signal <“0”, “0”> (it being understood that a first element within represents an initial signal value and a second element represents a final signal value), a normally high level signal <“1”,“1”>, a signal which rises from a low level to a high level <“0”, “1”> and a signal which falls from a high level to a low level <“1”, “0”>, respectively.

[0134] Thus each test pattern sequence comprises two test patterns. For example, a test pattern sequence T1=“LLR” signifies that x₁, x₂, x₃=<“000”, “001”>. Numerals entered in parentheses below corresponding signal values represent a transition time of each signal as referenced to the transition time of the signal at the input terminal which is chosen to be 0. For example, R(3) indicates the occurrence of a rising transition signal (“R”) at time 3.

[0135] The fifth column in FIG. 15 shows a set of delay faults which can be detected during the transient power supply current testing or a fault list when each test pattern sequence is applied to the integrated circuit under test. When a logic gate has a gate delay fault, the output transition time of the logic gate is retarded. Concomitantly, a logic gate which is connected at its input to the output of this faulty logic gate has the timing of a peak of its transient current retarded, whereby the transient power supply current of the integrated circuit under test exhibits an abnormality. In this manner, from an observation to see whether or not an abnormality occurs in the transient power supply current when a test pattern sequence is applied, it is possible to determine whether or not a logic gate which undergoes a switching operation (either a rising or falling transition) in response to the input test pattern sequence is faulty as a logic gate which is connected at its input to the output signal line from the first mentioned logic gate undergoes a switching operation.

[0136] By way of example, when a test pattern sequence T2 shown in FIG. 15 is applied to the CMOS integrated circuit under test shown in FIG. 14, there occurs a transition signal F(1) at an output node n₁, of a logic gate G₁in the CMOS integrated circuit under test, as shown in FIG. 16. Similarly, transition signals R(2), and F(4) occur at output nodes n₂and n₃of logic gates G₂and G₃, respectively. As a result of the transition signal R(2) occurring at the node n₂and the transition signal F(4) occurring at the node n₃, there occur at the output node n₄of the logic gate G₄a rising transition R at time 4 and a failing transition F at time 6. Thus, the transition signals R(4)-F(6) occur at the node n₄. Similarly, the transition signals R(5)-F(7) occur at the output terminal z₁of the output buffer G₆. Accordingly, when there exists a slow-to-fall fault in either one of the logic gates G₁, G₃and G₄or when there is a slow-to-rise fault in the logic gate G₂or G₄, an abnormality in the transient power supply current is observed during the transient power supply current testing which uses the test pattern sequence T2. In other words, the transient power supply current testing which uses the test pattern sequence T2 allows a slow-to-fall fault in either one of the logic gates G₁, G₃and G₄and a slow-to-rise fault in either logic gate G₂or G₄to be detected. Thus, a fault list which is determined by the test pattern sequence T2 can be determined as {G₁F, G₃F, G₄F, G₂R, G₄R} from the described I_DDTfault simulation, as indicated in the row of test pattern sequence T2 in FIG. 15. It will be understood that “F” and “R” in the fault list represent a slow-to-fall fault and a slow-to-rise fault, respectively.

[0137] It will be understood from the foregoing that a fault list, in unit of a gate delay fault which are detectable with a test pattern sequence can be prepared.

[0138] I_DDTfault simulation method (for open faults)

[0139] I_DDTfault simulation method which prepares a fault list in unit of a open fault will now be described. FIG. 17 shows CMOS integrated circuit under test. This circuit is the same as shown in FIG. 14, but signal lines connecting between input terminals x₁, x₂, x₃, x₄and logic gates as well as signal lines connecting between logic gates are designated by signal lines m₁, m₂, . . . , m₁₃. It is understood that the signal lines include input and output signal lines as well, and a branch signal line is treated as a distinct signal line. Output signal lines m₁₂and m₁₃are connected to output buffers G₆, G₇, respectively.

[0140] An example of results of I_DDTfault simulation which is performed upon the CMOS integrated circuit under test is illustrated in FIG. 18. In FIG. 18, the first column, as counted from the left, represents an identifier for a test pattern sequence; the second column show input signals applied to the input terminals x₁, x₂, x₃, x₄of the CMOS integrated circuit under test; the third column show signals occurring on signal lines m₁, m₂, . . . , m₁₃of the CMOS integrated circuit under test when each test pattern sequence is applied thereto; and the fourth column show signals appearing on the output terminals z₁, z₂of the CMOS integrated circuit under test when each test pattern sequence is applied thereto. Signals “L”, “H”, “R” and “F” remain the same as described above in connection with FIG. 15. For example, a test pattern sequence T=“RLLL” signifies that x₁,x₂,x₃,x₄=<“0000”, “1000”>. The fifth column in FIG. 18 show a set of signal lines having open faults which are detectable by the transient power supply current testing when each test pattern sequence is applied to the CMOS integrated circuit under test, or a fault list. When a signal line in the integrated circuit has an open fault, the switching operation of a logic gate which is connected at its input to the faulty signal fine will be retarded, accompanying a change in the transient power supply current waveform of the logic circuit and thus exhibiting an abnormality in the transient power supply current of the integrated circuit under test. Accordingly, from an observation to see whether or not an abnormality occurs in the transient power supply current when a test pattern sequence is applied, it is possible to determine whether or not a signal line which undergoes a switching operation in response to the input test pattern sequence is faulty when a logic gate which is connected at its input to the faulty signal fine undergoes a switching operation.

[0141] By way of example, when a test pattern sequence T2 shown in FIG. 18 is applied to a CMOS integrated circuit under test shown in FIG. 17, a transition signal F(0) occurs on a signal lines m₁in the CMOS integrated circuit under test, and transition signals F(1), F(1) and F(1) occur on signal lines m₅, m₆and m₇, respectively, as shown in FIG. 19. Similarly, transition signals R(2), F(4), F(4), F(4), and R(4)-F(6) occur on signal lines m₈, m₉, m₁₀,m₁₁and m₁₂, respectively, and transition signals R(5)-F(7) occur on the output terminal z₁. A transition signal F(4) occurs on the signal line m₁₁, but there occurs no transition in the output of the logic gate G₅which is connected at its input to the signal line m₁₁. Accordingly, whenever there is an open fault in either one of the signal lines m₁, m₅, m₆, m₇, m₈, m₉, m₁₀and m₁₂, there occurs a transition in the output of a logic gate which is connected at its input to such signal line. Accordingly, an abnormality in the transient power supply current is observed during the transient power supply current testing which uses the test pattern sequence T2 if an open fault exists on one of these signal lines. In other words, the transient power supply current testing which uses the test pattern sequence T2 allows the presence of an open fault on the signal lines m₁, m₅, m₆, m₇, m₈, m₉m₁₀and m₁₂to be detected. Thus, a fault list which is obtained with the test pattern sequence T2 is determined by the described I_DDTfault simulation to be {m₁, m₅, m₆, m₇, m₈, m₉, m₁₀, m₁₂},as indicated in the row of the test pattern sequence T2 in FIG. 18. The list of open faults is indicated in terms of signal lines on which an open fault is likely to occur.

[0142] It will be understood from the foregoing that a fault list, in unit of an open fault, which are detectable with a test pattern sequence can be prepared. I_DDTfault simulation method mentioned above is not limited to the detection of open faults on signal lines which connect between logic gates, but is equally applicable to open faults on signal lines which are located within logic gates, by assuming the presence of a fault on a signal line within a logic gate.

[0143] I_DDTfault simulation method (for path delay faults)

[0144] The fault simulation method according to the present invention which prepares a fault list in unit of a path delay fault will now be described. An example of results of a fault simulation which is performed upon the CMOS integrated circuit under test shown in FIG. 14 is illustrated in FIG. 20. In FIG. 20, the first, the second, the third and the fourth column have the same meaning as in FIG. 15. The fifth column show a set of path delay faults (a fault list) which are detectable by the transient power supply current testing when each test pattern sequence is applied to the CMOS integrated circuit under test. When the integrated circuit under test has a path delay fault, an output transition time from the faulty path is retarded. Concomitantly, a change in the output transition time of every or any one of logic gates located on the faulty path causes the timing of a peak in the transient current of a next stage logic gate which receives the output from such logic gate at its input to be retarded, whereby the transient power supply current of the integrated circuit under test exhibits an abnormality. Accordingly, by applying a test pattern sequence to the integrated circuit under test and observing whether or not an abnormality occurs in the transient power supply current of the integrated circuit under test, it is possible to determine whether or not a path delay fault has occurred on any signal propagation path on which every logic gate on this path undergoes a switching operation (either rising or falling transition) in response to the input test pattern sequence.

[0145] By way of example, when the test pattern sequence T1 is applied to the CMOS integrated circuit under test as shown in FIG. 14, there occurs a transition signal F(1) at the output node n₁of the logic gate G₁disposed within the CMOS integrated circuit under test, as shown in FIG. 21a. Similarly, signals R(2), R(4) and R(5) occur at the outputs of the logic gates G₂, G₄and the output buffer G₆, respectively. In this manner, because every logic gate located on a signal propagation path {x₁, n₁, n₂, n₄, z₁} is subject to a switching operation, it follows that whenever there is a path delay fault on the signal propagation path {x₁, n₁, n₂, n₄, z₁}, an abnormality in the transient power supply current is observed during the transient power supply current testing which uses the test pattern sequence T1. In other words, the transient power supply current testing which uses the test pattern sequence T1 allows a path delay fault or faults on the signal propagation path {x₁, n₁, n₂, n₄, z₁} to be detected. In this manner, a detected fault list which can be obtained with the test pattern sequence T1 can be determined by the described fault simulation to be as {{x₁, n₁, n₂, n₄, z₁}}.

[0146] In another example, when a test patterns sequence T2 is applied to the CMOS integrated circuit under test, a transition signal F(1) occurs at the output node n₁, of the logic gate G₁within the CMOS integrated circuit under test, and similarly, signals R(2), F(4), R(4)-F(6), R(5)-F(7) occur at the outputs of the logic gates G₂, G₃, G₄and the output buffer G₆, as shown in FIG. 21b. Since every logic gate located on signal propagation paths {x₁, n₁, n₂, n₄, z₁} and {x₁, n₁, n₃, n₄, z₁} is subject to a switching operation, if there is a path delay fault on either signal propagation path {x₁, n₁, n₂, n₄, z₁} or {x₁, n₁, n₃, n₄, z₁}, an abnormality in the transient power supply current is observed during the transient power supply current testing which uses the test pattern sequence T2. If a path delay time on the signal propagation path {x₁, n₁, n₂, n₄, z₁} increases, the rising transition R(5) at the output signal line z₁will be retarded, causing a likelihood that a signal transition may not occur at z₁. This prevents the path delay time on the signal propagation path {x₁, n₁, n₂, n₄, z₁} from being determined by the transient power supply current testing, resulting in a failure to detect a path delay fault on the signal propagation path {x₁, n₁, n₂, n₄, z₁}. Thus, a path delay fault or faults which can be detected by the transient power supply current testing which uses the test pattern sequence T2 are only those path delay faults which occur on the signal propagation path {x₁, n₁, n₃, n₄, z₁}. Consequently, a detected fault list which can be determined with the test pattern sequence T2 is determined from the described fault simulation to be {{x₁, n₁, n₃, n₄, z₁}}.

[0147] Generally, when there occur a plurality of transition signals on a signal line in response to a test pattern sequence, a path delay fault which is detectable by the transient power supply current testing is only a path delay fault on a signal transmission path which gives rise to a transition signal on the signal line, that is the same as a last transition signal on the signal line. For example, it may be assumed that transition signals R(a)-F(b)R(c)-F(d)-R(e) occur on a signal line in response to a test pattern sequence. Here it is assumed that transition times a, b, c, d and e are related such that a<b<c<d<e. In this instance, a signal transmission path which can be tested by the transient power supply current testing is limited to only a path which produce transition signals R(a), R(c) and R(e).

[0148] It will be understood from the foregoing that a fault list in unit of a path delay fault which are detectable with a test pattern sequence can be prepared. A signal transmission path having a path delay fault which is registered with a fault list is not limited to those paths which extend from an input terminal to an output terminal of a circuit under test, but may also include a signal transmission path such as {x₁, n₁} or {x₁, n₁, m₃} in the semiconductor integrated circuit shown in FIG. 14 which does not reach the output terminal.

[0149] In addition, the fault simulation method and the fault simulator according to the present invention is not limited to in its application to a CMOS integrated circuit, but is equally applicable to other types of semiconductor integrated circuits.

[0150] FIG. 1ais a plot indicating an input voltage V_INand an output voltage V_OUTin a typical CMOS inverter against time, FIG. 1bgraphically shows an example of the transient response of power supply current I_DD, FIG. 1cshows the CMOS inverter circuit and also showing a plot of the power supply current which flows upon the rising transition of the output and FIG. 1dshows the CMOS inverter circuit together with a plot of the power supply current which flows upon the falling transition of the output;

[0151] FIGS. 2aand 2bgraphically show a typical example of the transient response of CMOS logic gate, FIG. 2ashowing an input voltage V_IN, an output voltage V_OUTand the transfer characteristic of the supply current I_S, FIG. 2billustrating an approximated waveform of the transient current;

[0152] FIG. 3ais a circuit diagram of an exemplary CMOS integrated circuit, FIG. 3bis a plot graphically illustrating an input voltage and an output voltage of the integrated circuit against time, and FIG. 3cis a plot of the transient power supply current response I_DDTagainst time;

[0153] FIG. 4ais a schematic illustration of a perfect open fault, FIG. 4bis an exemplary illustration of an input/output with respect to the location of a fault, FIG. 4cillustrates a model of a delay open fault, and FIG. 4dillustrates an example of input/output with respect to the fault location;

[0154] FIG. 5ais a circuit diagram of an exemplary MOS integrated circuit having an open fault, and FIGS. 5band 5cgraphically show examples of output waveforms when there is no open fault and when there is an open fault;

[0155] FIG. 6 graphically shows an exemplary transient power supply current response of CMOS integrated circuit having an open fault;

[0156] FIG. 7ais a schematic illustration of the fundamental principle of the path delay fault testing method, FIG. 7bgraphically shows input/output voltage, and FIG. 7cgraphically shows a corresponding system clock;

[0157] FIG. 8 illustrates the principle of the transient power supply current testing technique which utilizes the pulse width of the transient power supply current, FIG. 8aindicating an input/output voltage plotted against time and FIG. 8bindicating a transient power supply current plotted against time;

[0158] FIG. 9 illustrates the principle of the transient power supply current testing technique which utilizes the instantaneous value of the transient power supply current, FIG. 9aindicating an input/output voltage plotted against time and FIG. 9bindicating a transient power supply current plotted against time;

[0159] FIG. 10ais a circuit diagram indicating a model for a very small open fault which is present on an input signal line of a CMOS inverter, FIG. 10bgraphically shows an exemplary output waveform when there is no fault, and FIG. 10cgraphically shows an exemplary output waveform when there is a fault;

[0160] FIG. 11 graphically shows a change in the integral of the transient power supply current of CMOS integrated circuit plotted against the resistance of a very small open fault present on a path under test of CMOS integrated circuit;

[0161] FIG. 12 graphically shows a change in the path delay time of a path under test plotted against the resistance of a very small open fault present on a path under test of CMOS integrated circuit;

[0162] FIG. 13 graphically shows the linearity of the relationship between the path delay time of the path under test and the integral of the transient power supply current of CMOS integrated circuit when the presence of a very small open fault is assumed on the path under test of CMOS integrated circuit;

[0163] FIG. 14 is a circuit diagram of an exemplary CMOS integrated circuit under test which is used to illustrate the fault simulation method according to the present invention;

[0164] FIG. 15 is a chart showing an example of results of the fault simulation performed upon the circuit shown in FIG. 14 when considering gate delay faults;

[0165] FIG. 16 is a circuit diagram, illustrating results of the fault simulation which is performed upon the circuit shown in FIG. 14 when a test pattern sequence is applied thereto;

[0166] FIG. 17 is a circuit diagram of an exemplary CMOS integrated circuit under test when open faults are considered;

[0167] FIG. 18 is a chart showing an example of results of the fault simulation according to the present invention which is applied to open faults;

[0168] FIG. 19 is a circuit diagram illustrating results of transition simulation applied to the circuit shown in FIG. 17 when the test pattern sequence T2 is applied thereto;

[0169] FIG. 20 is a chart showing an example of results of the fault simulation according to the present invention when path delay faults are considered;

[0170] FIGS. 21aand 21bare circuit diagrams illustrating results of the transition simulation applied to the circuit shown in FIG. 14, showing signals appearing at various paths when a test pattern sequence T1 and T2 are applied thereto;

[0171] FIG. 22 is an illustration of the functional arrangement of the fault simulator according to the present invention;

[0172] FIG. 23 is a flow chart according to an embodiment of the fault simulation method according to the present invention;

[0173] FIG. 24 is a flow chart of a specific procedure of the fault list preparing step shown in FIG. 23;

[0174] FIG. 25 is a flow chart showing an exemplary procedure when the fault list preparing step shown in FIG. 23 is applied for preparing a fault list in unit of a gate delay fault;

[0175] FIG. 26 is a flow chart of an exemplary procedure when the fault list preparing step shown in FIG. 23 is applied for preparing a fault list in unit of an open fault; and

[0176] FIG. 27 is a flow chart of an exemplary procedure when the fault list preparing step shown in FIG. 23 is applied for preparing a fault list in unit of a path delay fault.

[0177] Several embodiments of the present invention will now be described.

[0178] FIG. 22 shows an exemplary functional arrangement of a fault simulator according to an embodiment of the present invention. The simulator 100 comprises a test pattern sequence selector 101 which selects a test pattern sequence, formed by two or more test patterns and which is used in preparing a fault list, from a list of test pattern sequence 104; a transition simulator 102 which performs a simulation of transitions occurring in an integral circuit under test when the selected test pattern sequence from the selector 101 is input thereto to calculate a train of transition signal values on signal lines within the circuit, while discriminating between a rising transition and a falling transition, on the basis of circuit information of the integrated circuit under test which is available from a circuit information memory 105 and delay information of individual logic gates therein which is available from a delay information memory 106; a fault list preparing unit 103 which utilizes the train of transition signal values on various signal lines which are calculated by the transition simulator 102 to prepare a fault list which are detectable by the transient power supply current testing which uses the test pattern sequence for storage in a fault dictionary memory 107; and a controller 108 which operates the selector 101, the simulator 102 and the unit 103. Specifically, for a semiconductor integrated circuit under test having three input terminals, the test pattern sequence list 104 and the test pattern sequence selector 101 may comprise tree pseudo-random pattern sequence generators to provide random pattern sequence having mutually different initial values, but are synchronized with a common clock For each clock, each individual output from the three pseudo-random pattern sequence generators may be derived as a test pattern. Alternatively, a 3-bit counter may be operated with a clock, and a count represented by the three bits may be derived as test patterns. In addition, this may be implemented by a software. As a further alterative, required test patterns may be stored in a memory, and may be sequentially taken out. At any rate, the pseudo-random pattern sequence generator may have a number of shift stages or a sufficient number of test pattern sequence may be stored in the memory so that test patterns are available for providing sufficient data to enable a delay fault of any logic gate with a distinction between a rising and a falling transition, an open fault on any signal line and a delay fault on any path to be detected as a detectable fault. The test pattern sequence selector 101 is provided for supplying such a test pattern sequence to the transition simulator 102, the requirement upon the selector 101 being that it is capable of deriving a test pattern sequence from the test pattern sequence list 104 or generating it otherwise to feed the transition simulator 102.

[0179] When the transition simulator 102 is implemented in a software, the software sequentially performs calculations for each logic gate in the semiconductor integrated circuit under test using the circuit information from the memory 105 and the delay information from the memory 106 when each of the test pattern is applied to the semiconductor integrated circuit under test, thus determining transitions between logic values based on the results of these calculations, including a rising transition R, a falling transition F, no transition L or H together with a delay time from the application of the test pattern and storing these results for each signal line on which the transition signal value appears. This is repeated for each test pattern, thus easily obtaining a train of transition signal values occurring on various signal lines. A general purpose event driven logic simulator, such as VHDL System Simulator (VSS) from Synopsis, may be used for the transition simulator 102. Means 103 for preparing a fault list will be described later.

[0180] The operation of performing a fault simulation of the semiconductor integrated circuit using the fault simulator 100 according to the present invention will now be described. FIG. 23 shows an exemplary processing procedure for the fault simulation method of the present invention. Initially, at step 201, the fault simulator 100 prepares a list of test pattern sequence which are required for performing a fault simulation. Any test pattern is registered with the list. As mentioned previously, the list may not necessarily be prepared beforehand, but an arrangement may be made so that required test pattern sequence can be obtained from random pattern generators, for example.

[0181] At step 202, the test pattern sequence selector 101 selects one of the test pattern sequence from the test pattern sequence list, for which it is desired to prepare a fault list Accordingly, one test pattern sequence is selected arbitrarily from the test pattern sequence list which is previously provided, or one test pattern sequence is obtained from the test pattern sequence generators. At step 203, the transition simulator 102 performs a transition simulation assuming that the test pattern sequence selected at step 102 has been input to the integrated circuit under test, using the circuit information of the semiconductor integrated circuit under test and delay information of logic gates, thus calculating a train of transition signal values which occur on various signal lines within the circuit. At step 204, the fault list preparing unit 103 prepares a fault list which are detectable with the selected test pattern sequence on the basis of the train of transition signal values occurring on signal lines within the circuit which are obtained at step 203.

[0182] Finally, at step 205, the controller 108 confirms to see whether or not there remains any test pattern sequence in the test pattern sequence list which has not yet been processed and also confirms whether or not there has been a sufficient number of test pattern sequence to enable the required detection of faults during a phase of deriving the test pattern sequence from the test pattern sequence list. If it is found that there is a test pattern sequence which is not yet processed or which has not been generated, the steps 202, 203, 204, and 205 are repeated. On the other hand, if there is no test pattern sequence which has not been processed or if a sufficient number of test pattern sequence have been generated, the operation is completed. The controller 108 exercises a control for sequentially operating the various units to perform the described steps in sequence. At step 204 where the fault list is prepared, the fault list may be prepared in unit of a gate delay fault, in unit of an open fault or in unit of a path delay fault. It is to be noted that the function of the fault simulator 100 can be served by using a computer which executes a program. Thus, the controller 108 may comprise CPU or a processor which executes a program stored in a memory, thus implementing the operation of the test pattern sequence selector 101, the transition simulator 102 and the fault list preparing unit 103 through the execution of the program. The operation of each of such units may be implemented by executing a subroutine, which may be partly assisted by a hardware.

[0183] FIG. 24 shows an exemplary processing procedure for preparing the fault list which takes place at step 204 in FIG. 23. Initially, at sep 301, one of possible faults in the semiconductor integrated circuit under test is selected for purpose of initialization. At next step 302, a confirmation is made to see whether or not the train of transition signal values which are calculated by the transition simulator 402 satisfy the requirement for detecting the fault according to the transient power supply current testing in the event a fault has occurred at the location which is set up as mentioned above. If the train of Position signal values satisfy the requirement for detecting the fault according to the transient power supply current testing, the selected location is registered with a fault list at step 303. However if it is found at step 302 that the train of transition signal values do not satisfy the requirement for detecting the fault according to the transient power supply current testing, the operation proceeds to step 304 where a confirmation is made to see whether or not there remains any other unprocessed location which is likely to cause a fault, and if there is an unprocessed location having a likelihood of fault, a next location which is likely to cause a fault is set up at step 305 before returning to step 302. In this manner, steps 302, 303, 304 and 305 are repeated until every location within the semiconductor integrated circuit under test which is likely to cause a fault is processed, and when there remains no unprocessed location, the operation is completed.

[0184] FIG. 25 shows an exemplary processing procedure for the preparation of the fault list which takes place at step 204 in FIG. 23 in case the fault list is prepared in unit of a gate delay fault. Initially, at step 401, one of gate delay faults (logic gate) which are likely to occur in a semiconductor integrated circuit under test is selected with a distinction of the direction in which the transition occurs for purpose of initialization. At step 402, a confirmation is made to see whether or not a logic gate having the gate delay fault (or a faulty logic gate) is subject to a switching operation in a proper manner including the direction of transition so as to conform to the gate delay fault which is set up by the initialization, on the basis of results of the transition simulation which is calculated by the transition simulator 102. If the faulty logic gate is properly subject to a switching operation, the operation proceeds to step 403, but if the faulty logic gate is not properly subject to a switching operation, the operation transfers to step 405.

[0185] At step 402, if the gate delay fault which is set up is a slow-to-rise fault, a confirmation is made to see whether or not the output from the faulty logic gate has a rising transition. Conversely, if the gate delay fault which is set up is a slow-to-fall fault, a confirmation is made to see whether or not the output from the faulty logic gate has a falling transition. At next step 403, a confirmation is made to see whether or not the output signal line of a logic gate (load logic gate) which input is connected to the output of the logic gate having the gate delay fault is subject to a switching operation, on the basis of results of the fault simulation which is calculated by the transition simulator 102. If the output signal line of the load logic gate is subject to a switching operation, the gate delay fault is registered with the fault list at step 404. On the other hand, if the output signal line of the load logic gate is not subject to a switching operation, the operation transfers to step 405. By way of example, if a gate delay fault G₁F is set up in the semiconductor integrated circuit under test shown in FIG. 14, a proper falling transition of the faulty logic gate G₁is confirmed by the test pattern sequence T1, T2 and T3 at step 402, and the fact that the output line of the load logic gate G₂is subject to a switching operation in response to the test pattern sequence T1, T2 and T3 is confirmed at step 403. Accordingly, G₁F is registered for each of the test pattern sequence T1, T2 and T3 or the test pattern sequence T1, T2 and T3 are registered with G₁F at step 404. Similarly, when G₁R is set up, a proper rising transition of the faulty logic gate G₁is confined by the test pattern sequence T4 and T5 and a switching operation of the output line of the load logic gate G₂is confirmed by the test pattern sequence T4 and T5. Accordingly, G₁R is registered with each of T4 and T5 or T4 and T5 are registered with G₁R

[0186] At step 405, a confirmation is made to see whether there remains any other unprocessed gate delay fault If there remains an unprocessed gate delay fault, next possible gate delay fault is set up at step 406, and the steps 402, 403, 404 and 405 are repeated. However, if there remains no unprocessed gate delay fault, the operation is completed.

[0187] FIG. 26 shows an exemplary processing procedure for preparing a fault list which takes place at step 204 in FIG. 23 when preparing a fault list in unit of an open fault Initially, at step 501, one of open faults which are likely to occur in a semiconductor integrated circuit under test is selected for purpose of initialization. At next step 502, a confirmation is made to see whether or not a signal line having the open fault which is set up (or a faulty signal line) is subject to a switching operation, on the basis of the results of the transition simulation which is calculated by the transition simulator 102. If the faulty signal line is subject to a switching operation, the operation proceeds to step 503, but if the faulty signal line is not subject to a switching operation, the operation transfers to step 505.

[0188] At step 503, a confirmation is made to see whether or not the output signal line of a logic gate which input is connected to the faulty signal line (or a load logic gate) is subject to a switching operation, on the basis of results of the transition simulation calculated by the transition simulator 102. If the output signal line of the load logic gate is subject to a switching operation, the open fault is registered with a fault list at step 504. However, if the output signal line of the load logic gate is not subject to a switching operation, the operation transfers to step 505. At step 505, a confirmation is made to see whether or not there remains any other unprocessed open fault If there remains an unprocessed open fault, a next possible open fault is set up at step 506, and then the steps 502, 503, 504 and 505 are repeated. If there remains no other unprocessed open fault, the operation is completed.

[0189] FIG. 27 shows an exemplary processing procedure for preparing a fault list which takes place at step 204 in FIG. 23 when preparing a fault list in unit of a path delay fault. Initially, at step 601, one of path delay faults which are likely to occur in a semiconductor integrated circuit under test is selected for purpose of initialization. At step 602, a confirmation is made to see whether or not every logic gate on a path having the selected path delay fault (or a faulty path) is subject to a switching operation, on the basis of the results of the transition simulation which is calculated by the transition simulator 102. If the every logic gate on the faulty path is subject to a switching operation, the operation proceeds to step 603, but if any one of the logic gates on the faulty path is not subject to a switching operation, the operation transfers to step 606.

[0190] At step 603, a confirmation is made to see whether or not there is a logic gate on the faulty path which is subject to a switching operation a plurality of times in response to a test pattern sequence which is selected at step 202 in FIG. 23, on the basis of results of the transition simulation which is calculated by the transition simulator 102. If there is a logic gate which is subject to a switching operation a plurality of times, the operation proceeds to step 604, but if there is no logic gate which is subject to a switching operation a plurality of times, the path delay fault is registered with a fault list at step 605.

[0191] At step 604, a confirmation is made to see whether or not an on-path input of every logic gate which is subject to a switching operation a plurality of times (or an input signal line of a logic gate on the faulty path) satisfies the requirement for detecting the fault according to the transient power supply current testing. If the on-path input satisfies the requirement for detecting the fault according to the transient power supply current testing, the path delay fault is registered with the fault list at step 605. However, if the on-path input does not satisfy the requirement for detecting the fault according to the transient power supply current testing, the operation transfers to step 606. Thus, what occurs at step 604 is to confirm whether or not the output transition signal corresponding to the on-path input to the logic gate is the same kind of transition signal as the last transition signal on the output signal line of the logic gate, the confirmation taking place for every logic gate which is subject to a switching operation a plurality of times. For example, the logic gate G₄shown in FIG. 21(b) is subject to a switching operation a plurality of times, and the output transition signal “F(6)” of the logic gate G₄which corresponds to the input transition signal “F(4)” of the on-path input n₃of the logic gate G₄coincides with the last transition signal “F(6)” of the logic gate G₄. Accordingly, the on-path input n₃satisfies the requirement for detecting a fault according to the transient power supply current testing. However, the last transition signal of the logic gate G₄which corresponds to the input transition signal “R(2)” on the on-path input n₂of the logic gate G₄is equal to “F(6)”, which is not of the same kind as “R(2)”.

[0192] Thus, the on-path input n₂does not satisfy the requirement for detecting a fault according to the transient power supply current testing.

[0193] At step 606, a confirmation is made to see whether or not there remains any other unprocessed path delay fault If there remains an unprocessed path delay fault, a next possible path delay fault is set up at step 607, and the steps 602, 603, 604, 605 and 606 are repeated. However, if there remains no other unprocessed path delay fault, the operation is completed.

[0194] The fault simulation method and the fault simulator according to the present invention are not limited in its application to a delay fault and an open fault, but may also be used for detecting other faults including a logic fault (stuck-at fault), a short-circuit fault and wrong parameters of MOS transistor, by suitably modifying the requirement for detecting the fault according to the transient power supply current testing and a fault model.

[0195] As discussed above, the fault simulation method and the fault simulator according to the present invention each allow a fault list to be prepared which can be detected by the transient power supply current testing with a given test pattern, even for a delay fault or an open fault which gives rise to a delay fault, either of which has been difficult in the prior art to detect, by using the transient power supply current testing technique which affords a high level of observability and which provide logic gate switching information. Accordingly, the efficiency of testing a delay fault or an open fault can be significantly improved.