Method and apparatus of testing and analyzing CMOS integrated circuit

[0002] 1. Field of the Invention

[0003] The present invention relates to a method and apparatus of testing and analyzing a complementary metal oxide semiconductor (CMOS) integrated circuit.

[0004] 2. Description of the Related Art

[0005] Japanese Unexamined Patent Publication (Kokai) No. 8-271584, Japanese Unexamined Patent Publication (Kokai) No. 9-211088, U.S. Pat. No. 5,392,293, U.S. Pat. No. 5,519,333, and U.S. Pat. No. 5,889,408 disclose testing (I_DDQtesting) using a drain-to-drain quiescent power supply current (I_DDQ) of a CMOS integrated circuit.

[0006] In I_DDQtesting, measures the quiescent power supply current of a CMOS integrated circuit, is measured, and a good or defective of the CMOS integrated circuit is determined by using the measured current. Note that a CMOS integrated circuit under test is also called a device under test (DUT).

[0007] A quiescent power supply current I_DDQincludes a leakage current (intrinsic leakage current) flowing even in a good device and a defect current caused by a defect. That is, a quiescent power supply current I_DDQ, (hereinafter also called a quiescent power supply current I_Q) can be expressed by the summation of the intrinsic quiescent power supply current I_Fand defect current I_Dand can be shown by the following formula (1):

[0008] I_Q=I_F+I_D  (1)

[0009] The intrinsic leakage current I_Fcan be expressed by the summation of the leakage current (FET leakage current) I_Tgenerated due to the structure of a metal oxide semiconductor field effect transistor (MOSFET) and the leakage current (circuit leakage current) caused due to a circuit operation.

[0010] The circuit leakage current is generated by an analog circuit, pull-up current, bus collision, etc. Generation of the circuit leakage current is however avoided when I_DDQtesting for measuring the quiescent power supply current I_Qand therefore it can be disregarded. Therefore, the following formula (2) can represent the defect current I_D:

[0011] I_D=I_Q−I_T  (2)

[0012] In the I_DDQtesting, the FET leakage current I_Tof the CMOS integrated circuit under test is unknown and therefore estimated by some method or another. The good or defect of the CMOS integrated circuit is judged by using the measured quiescent power supply current I_Q.

[0013] The main defect current I_Din the quiescent power supply current I_Qis current generated by internal shorts among gates, sources, drains, and well of the FET and a bridging between interconnection patterns. The value of the defect current I_Ddepends on the power supply voltage and equivalent resistance value.

[0014] FIG. 1 is a graph of the relative frequency distribution of the maximum I_DDQ(the maximum value of the quiescent power supply current I_Q) in two types of integrated circuits as an example. The maximum I_DDQof an abscissa is broken down into 0 to 100 μA, 100 to 200 μA, . . . , 700 to 800 μA, and 800 μA or more.

[0015] The effective gate length Leff of the type A and B CMOS integrated circuits is 0.5 μm. A maximum I_DDQof more than 30 μA is shown. The distribution shown in FIG. 1 substantially corresponds to the distribution of the defect current I_D.

[0016] Further, A. E. Gattiker and W. Maly, “Toward Understanding ‘Iddq-Only’ Fails”, in Int. Test Conf., pp.174-183, IEEE, 1998, states that there is an indirectly generated defect current, in addition to the defect current generated directly by the bridging. This indirect defect current is generated because when a potential of a signal line falls between the power-supply voltage and the ground potential for some reason or another, the p type MOSFET and n type MOSFET driven by this signal line become on in state at the same time, and a penetration current flows.

[0017] It is necessary to apply a different voltage to the two ends of a bridge forming a defect in order to cause generation of the defect current I_D.

[0018] Since it is impossible to set the potential required for all presumed failures all at once, a test signal of a test pattern prepared by an automatic test pattern generator (ATPG) or a test signal of a test pattern for a function test is applied and the quiescent power supply current I_Qmeasured when a terminal becomes the necessary potential. Note that a measurement point for measuring the quiescent power supply current I_Qis called as a strobe (strobe point).

[0019] The strobe point (defect detection strobe point) which can detect a defect differs according to the cause of the defect.

[0020] Since bridging between the power supply line for supplying a power supply voltage V_Dand a ground line has no relation with the potential of the signal line, the defect is detected at all strobe points and the defect current becomes the same.

[0021] Note that there are few defect detection strobe points which can detect bridging of interconnections between cells. Further, the number of defect detection strobe points which can detect a short of the MOSFET in the cell is a levels between the two.

[0022] According to A. Keshavarzi, K. Roy, and C. F. Hawkins, “Intrinsic Leakage in Low Power Deep Submicron CMOS ICs”, in Int. Test Conf., pp. 146-155, IEEE, 1997; A. Ferre and J. Figueras, “On Estimating Bounds of the Quiescent Power Supply Current for I_DDQTesting”, in VLSI Test Sym., pp. 106-111, IEEE, 1996; and P. C. Maxwell and J. R. Rearick, “Estimation of Defect-Free IDDQ in Submicron Circuits Using Switch Level Simulation”, in Int. Test Conf., pp. 882-889, IEEE, 1998, the leakage current (EFT leakage current) I_Tof a defect-free normal MOSFET can be classified based on the ON/OFF state of the MOSFET, the current path, etc.

[0023] FIG. 2 is a graph showing a classification of the leakage current of a MOSFET.

[0024] FIG. 2 shows a feeding power route, causes of occurrence, conduction conditions, and an approximation formula of the leakage current with respect to cases L₁, to L₃.

[0025] In the case L₁, the feeding power route is between a well and base (substrate), the cause of occurrence is an inverse bias of a pn junction, and the conduction condition is an ON state.

[0026] In the case L₂, the feeding power route is between a drain and well, the cause of occurrence is the inverse bias of the pn junction, and the conduction condition is an OFF state.

[0027] In the case L₃, the feeding route is between the drain and source, the cause of occurrence is a weak inverse, and the conduction condition is an OFF state.

[0028] The leakage current of the case L₁is always generated if the well and base are inverse biases and is constant regardless of the strobe (strobe point).

[0029] The leakage current of the case L₂is generated in a case where the MOSFET is in an OFF state and the drain thereof is connected to the power supply line and/or the ground line.

[0030] The leakage currents of the case L₃is generated in a case where the MOSFET is in an off state and the drain and the source thereof are connected to the power supply line and/or the ground line.

[0031] If the leakage currents of the cases Li (i =1 to 3) in each MOSFET in a CMOS integrated circuit are averaged for each case, the leakage currents L_N1and L_N1in per n type MOSFET and p type MOSFET are found.

[0032] The FET leakage current IT can be expressed by the following formula (3) using the numbers N_N1and N_P1of the n type MOSFETs and the p type MOSFETs in which the leakage current of the case L₁has been generated: $\begin{array}{cc}{I}_T=\sum _{F=1}^3\left(N_{\mathrm{Ni}}\times I_{\mathrm{Mi}}+N_{\mathrm{Pi}}\times I_{\mathrm{Pi}}\right)& \left(3\right)\end{array}$

[0033] FIG. 3 is a chart illustrating the distribution of the maximum I_DDQin an CMOS integrated circuit and shows the frequency distribution for every 0.1 μA. The effective gate length Leff of the CMOS integrated circuit is 0.5 μm.

[0034] In FIG. 3, a CMOS integrated circuit having a maximum I_DDQof less than 1 μA is judged as good and a CMOS integrated circuit having a maximum I_DDQof 1 μA or more is judged as defective.

[0035] Note that the broken line shows a curve of an average value of 0.2 μA and a standard deviation σ of 0.14 μA. In a good CMOS integrated circuit having a maximum I_DDQof below 1 μA, the distribution of the FET leakage current is the normal distribution or similar to the normal distribution.

[0036] Note that if 1(≈0.2÷6×0.14) μA of exactly 6σ from the average value is made the maximum FET leakage current assuming a variation in current between lots, a quiescent power supply current I_Qabove that is considered to be due to a defect. That is, 1 μA is the threshold value of judgement of good or defective.

[0037] When the threshold value of judgement of good or defective is set to 1 μA in this way, a defect current of up to a maximum of 0.9 (=1 to 0.1) μA in a CMOS integrated circuit having an FET leakage current of 0.1 μA in the chart of FIG. 3 is missed.

[0038] Note that in the chart of FIG. 3, defective CMOS integrated circuits having a maximum I_DDQof 1 to 4 μA are substantially uniformly distributed. It is assumed that they are distributed at a range below 1 μA as well. However, the number of defective CMOS integrated circuits of less than 1 μA is small, so the failure to catch such defective CMOS integrated circuits in the testing can be disregarded.

[0039] FIG. 4 is a graph of the distribution of the FET leakage current in two good CMOS integrated circuits. Note that in a good CMOS integrated circuit, the defect currents I_Dis considered of a degree which can be disregarded and so the quiescent power supply current I_DDQof a good CMOS integrated circuit is considered the FET leakage current. The effective gate length Leff of the two CMOS integrated circuits LSI-1 and LSI-2 is 0.25 μm.

[0040] The abscissa of FIG. 4 shows the consecutive numbering of the strobe points (strobe number), while the ordinate shows the magnitude of the FET leakage current at each strobe point.

[0041] The FET leakage current of the CMOS integrated circuit LSI-1 changes up to ±10 μA about approximately 73 μA, while the FET leakage current of the CMOS integrated circuit LSI-2 changes up to ±5 μA about approximately 29 μA.

[0042] FIG. 5 is a chart of the relative frequency distribution of an FET leakage current at all 800 strobe points in the CMOS integrated circuit LSI-1 and LSI-2 of FIG. 4. The abscissa is graduated into units of 10 μA.

[0043] In the chart of FIG. 5, two different variations are seen.

[0044] The first variation is the variation among strobe problems in the same CMOS integrated circuit as seen from 25 to 35 μA.

[0045] The second variation is the variation between CMOS integrated circuits such as the center of frequency distribution being around 73 μA in the CMOS integrated circuit LSI-1 and around 29 μA in the CMOS integrated circuit LSI-2.

[0046] FIG. 6 is a chart illustrating the relative frequency distribution of FET leakage current in a case where the average value of the quiescent power supply current in a CMOS integrated circuit (average I_DDQ) is considered the FET leakage current. The chart of FIG. 6 illustrates the distribution of 32 good CMOS integrated circuits randomly sampled from two lots. The current is widely distributed from 0 to 350 μA.

[0047] The above judgement of good or defective using as the threshold the estimated maximum value of the FET leakage current is liable to miss catching many defects.

[0048] As one example, if 85μA is made the threshold value in the chart of FIG. 5, the occurrence of a defect current of 60 (=85 to 25) μA in a CMOS integrated circuit LSI-1 having an FET leakage current of 25 μA cannot be detected.

[0049] In this way, in the above I_DDQtesting measuring the quiescent power supply current of a CMOS integrated circuit and judging that circuit to be defective when the measured value is above a threshold, as described in T. W. Williams, R. H. Dennard, and R. Kapur, “Iddq Test: Sensitivity Analysis of Scalings”, in Int. Test Conf., pp. 786-792. IEEE, 1996, sometimes accurate judgement is difficult.

[0050] For example, this is because, when the interconnection pattern of the CMOS integrated circuit is fine, the leakage current of a MOSFET (FET leakage current) increases exponentially along with the fineness. Specifically, this is because the FET leakage current in the case L₃increases exponentially due to the shortening of the effective gate length Leff and the fall in the threshold voltage V_th.

[0051] Therefore, various methods for reducing FET leakage current during I_DDQtesting have been proposed.

[0052] One method of reducing the FET leakage current in I_DDQtesting is to lower the FET leakage current when strobing so as to reduce the threshold. The low power supply voltage method, low temperature measurement method, and a well bias method are known for this.

[0053] The low power supply voltage method uses the fact that the leakage current falls by the approximation formulas of cases L₁to L₃shown in FIG. 2 when lowering the power supply voltage V_D. However, since the power supply voltage can only be reduced to an extent where the circuit does not malfunction, the rate of reduction of the FET leakage current is low.

[0054] Further, several milliseconds (ms) are necessary to change the power supply voltage V_Dbefore and after strobing. The test cost is increased along with this increase of testing time.

[0055] Moreover, according to A. E. Gattiker and W. Maly, “Toward Understanding ‘Iddq-Only’ Fails”, in Int. Test Conf., pp. 174-183, IEEE, 1998, there are cases where the fault penetration current disappears when lowering the power supply voltage V_Dand this may lead to defects being missed in the testing.

[0056] This method uses the fact that the FET leakage current falls by the approximation formula of the case L₃shown in FIG. 2 when lowering the operating temperature.

[0057] The lower limit temperature is determined by the guaranteed reliability and the cost of the test system for maintaining the low temperature and test system. The limit in civilian use low temperature maintaining systems is about 0οC, so the rate of reduction of the FET leakage current is low. Further, the expense of the system and its running costs pile up and increase costs.

[0058] The well bias method is described in a paper of A. Keshavarzi, K. Roy, and C. F. Hawkins, “Intrinsic Leakage in Low Power Deep Submicron CMOS ICs”, in Int. Test Conf., pp. 146-155, IEEE, 1997.

[0059] The leakage current falls by the approximation formulas of the cases L₁to L₃shown in FIG. 2 when the threshold voltage V_this raised. Further, the threshold voltage rises when a bias voltage is applied between the source and well.

[0060] In this method, since interconnections are added to apply the bias voltage, the chip of the CMOS integrated circuit on the chip is increased and the cost rises.

[0061] Further, the rate of reduction of the leakage current is very susceptible to variations in the effective gate length Leff, so the FET leakage current fluctuates along with increasing miniaturization.

[0062] On the other hand, methods have been proposed for judging good or defective or detecting defect current in a state with a large FET leakage current. For example, there are the signature analysis method, delta method, and upper and lower limit method.

[0063] However, in these methods, it suffers from that detection of defect currents less than the variation of the FET leakage current among strobes is impossible, defects where the same defect current is generated at all the strobe points miss being caught, and estimation of the defect current is difficult.

[0064] The signature analysis method is disclosed in a paper of A. E. Gattiker and W. Mary, “Current Signatures”, in VLSI Test Sym., pp. 112-117. IEEE, 1996.

[0065] In this method, since the measured quiescent power supply currents are rearranged in order of magnitude and good or defective is judged based on existence of any sharp steps, it is not too well suited for mass production.

[0066] The delta method is described in a paper of C. Thibeault, “On the Comparison of ΔI_DDQand I_DDQTesting”, in VLSI Test Sym., pp. 143-150, IEEE, 1999.

[0067] In this method, defective CMOS integrated circuits are separated by the amount of change of the quiescent power supply current, so a large amount of statistical processing is necessary. Therefore, this method is not too well suited to a mass production process.

[0068] The upper and lower limit method is disclosed in U.S. Pat. No. 5,914,615, Jun. 1999, B. Chess, “Method of Improving the Good or Defective and Efficiency of Iddq Testing”.

[0069] In this method, the upper and lower limit are set according to the average value of the measured quiescent power supply currents and are used to detect defective CMOS integrated circuits.

[0070] However, since variations among CMOS integrated circuits and variations among strobes are not differentiated, error sometimes occurs in the upper limit value and lower limit value calculated from the average value of the quiescent power supply currents.

[0071] A first object of the present invention is to provide a method and apparatus of testing a CMOS integrated circuit capable of judging the good or defective of a CMOS integrated circuit having a large quiescent power supply current and large variations.

[0072] A second object of the present invention is to provide a method and apparatus of analyzing a CMOS integrated circuit capable of detecting a defect current of a CMOS integrated circuit having a large quiescent power supply current and large variations.

[0073] According to a first aspect of the present invention, there is provided a method of testing a CMOS integrated circuit comprising the steps of: applying a test signal to a CMOS integrated circuit under test and measuring a quiescent power supply current at part or all of the strobe points among a predetermined plurality of strobe points; and judging the good or defective of the CMOS integrated circuit under test based on average value ratios of quiescent power supply currents of the plurality of strobe points calculated in advance for a good CMOS integrated circuit, the measured values at the part or all of the strobe points, and the average value of the measured values.

[0074] Preferably the step of judging the good or defective based on the average value ratios, the measured values, and the average value comprises the steps of calculating defect current estimated value P_Dj: I_Qj−R_j×Iq corresponding to the strobe points based on the average value ratios R_jof the quiescent power supply currents at the plurality of strobe points, the measured values I_Qjat the part or all of the strobe points, and the average value Iq; and judging the good or defective of the CMOS integrated circuit under test based on the calculated defect current estimates P_Djand allowable error E of the measured values of the quiescent power supply current.

[0075] Alternatively, the step of judging good or defective based on the defect current estimated value P_Djand the allowable error E judges the CMOS integrated circuit under test as a defect when an absolute value of a calculated defect current estimate P_DJis larger than an absolute value of the allowable error E.

[0076] Alternatively, the step of judging good or defective based on the defect current estimates P_Djand the allowable error E comprises the steps of calculating a variation rate P_STQ=Iq×(R_b-R_S)/(I_Qb/- I_Qs) when each of a plurality of absolute values of the plurality of defect current estimated value P_DJcorresponding to all of the strobe points is not more than the absolute value of the allowable error E and of judging the CMOS integrated circuit under test as a defect when the variation rate P_STQis greater than 1. Here, R_bis the biggest average value ratio among the average value ratios R_j, I_Qbis the measured value of the strobe point corresponding to the biggest average value ratio R_b, R_sis the smallest average value ratio among the average value ratios R_j, and I_Qsis the measured value of the strobe point corresponding to the smallest average value ratio R_s.

[0077] In the latter case, the step of judging good or defective based on the defect current estimated value P_Djand the allowable error E may further comprise a step of judging the CMOS integrated circuit under test as good when the variation rate P_STQis less than 1.

[0078] Preferably, the allowable error E of the measured quiescent power supply current is whichever of a product (E_R×Iq) of a maximum error rate E_Ramong error rates E_Rjof average value ratios R_jat the plurality of strobe points and the average value Iq and of a maximum measurement error E_Mhas a larger absolute value.

[0079] Preferably, the average value is defined as an arithmetic average value of the measured values at part or all of the strobe points.

[0080] Alternatively, the average value is defined as ratio (I_Q1/R₁) between a measured quiescent power supply current I_Q1at a predetermined strobe point of the part or all of the strobe points and the average value ratio R₁of the quiescent power supply current at the predetermined strobe point.

[0081] In this case, preferably the step of judging good or defective based on the average value ratios, the measured values, and the average value comprises the steps of calculating defect current predictions P_Q1j=I_Qj−R_j×(I_Q1/R₁) corresponding to the strobe points based on the average value ratios R_jof the quiescent power supply currents at the plurality of strobe points, the measured values I_Qjat the part or all of the strobe points, and the ratio(I_Q1/R₁) and judging good or defective based on the calculated defect current predictions P_Q1jand the allowable error E of the measured values of the quiescent power supply current.

[0082] More preferably, the step of judging good or defective based on the defect current predictions P_Q1jand the allowable error E comprises a step of judging the CMOS integrated circuit under test as a defect when an absolute value of a calculated defect current prediction P_Q1jis larger than an absolute value of the allowable error E.

[0083] In the above case, preferably, the allowable error E of the measured quiescent power supply current is whichever of a product (E_R×I_Q1/R₁) of a maximum error ratio E_Ramong a plurality of error ratios E_Rjof the average value ratios R_jof the quiescent power supply current at the plurality of strobe points and the ratio (I_Q1/R₁) and of the maximum measurement error E_Mhas the larger absolute value.

[0084] In the above case, alternatively, the step of judging good or defective based on the average value ratios, the measured values, and the average value comprises the steps of calculating an upper limit threshold I_QU=(1+f×σ_R)× I_Q1/R₁based on the average value ratio R₁of the quiescent power supply current at the predetermined strobe point, the ratio (I_Q1/R₁), and a constant f of 4 to 7; and judging the CMOS integrated circuit under test as a defect when a measured quiescent power supply current is larger than the upper limit threshold I_QU, where σ_Ris the square root of the average value of (R_j−1)²at the plurality of strobe points.

[0085] Here, preferably the step of judging good or defective based on the average value ratios, the measured values, and the average value further comprises the steps of calculating a lower limit threshold I_QL=(1−f×σ_R)×I_Q1/R₁; judging the CMOS integrated circuit under test as a defect when a measured value is below the lower limit threshold I_QL; and judging the CMOS integrated circuit under test as good when each of the measured values at the part or all of the strobe points between the lower limit threshold I_QLand the upper limit threshold I_QU.

[0086] Here, alternatively, the value of the constant f is 5 to 6.

[0087] In the above case, alternatively, the step of judging good or defective based on the average value ratios, the measured values, and the average value comprises the steps of calculating a lower limit threshold I_QL=(1−f×σ_R)×I_Q1/R₁based on the average value ratio R₁of the quiescent power supply current at the predetermined strobe point, the ratio (I_Q1/R₁), and a constant f of 4 to 7 and judging the CMOS integrated circuit under test as a defect when the measured quiescent power supply current is smaller than the lower limit threshold I_QL, where σ_Ris the square root of the average value of (R_j−1)²at the plurality of strobe points. Here, preferably, the value of the constant f is 5 to 6.

[0088] In the above case, alternatively, the predetermined strobe point comprises a first strobe point of the part or all of the strobe points.

[0089] According to the present invention, there is also provided an apparatus for testing a CMOS integrated circuit comprising: a means for applying a test signal to a CMOS integrated circuit under test and measuring a quiescent power supply current at part or all of the strobe points among a predetermined plurality of strobe points; and a means for judging the good or defective of the CMOS integrated circuit under test based on average value ratios of quiescent power supply currents of said plurality of strobe points calculated in advance for a good CMOS integrated circuit, the measured values at the part or all of the strobe points, and the average value of the measured values.

[0090] According to a second aspect of the present invention, there is provided a method of analyzing a CMOS integrated circuit, comprising the steps of applying a test signal to a CMOS integrated circuit under test and measuring a quiescent power supply current at a predetermined plurality of strobe points; calculating defect current estimated values P_Dj=I_Qj−R_j×Iq corresponding to the strobe points based on average value ratios R_jof quiescent power supply currents at the plurality of strobe points calculated in advance for a good CMOS integrated circuit, measured values I_Qjat the plurality of strobe points, and an average value Iq of the measured values; calculating a summation P_DSUMof negative value defect current estimated value P_DJamong the calculated defect current estimated values P_DJ; calculating a summation R_SUMof the average value ratios R_jfor the strobe points corresponding to the negative value defect current estimated value P_Djamong the plurality of strobe points; and calculating an absolute value |P_DSUM/P_SUM| of a ratio of the summation P_DSUMof the defect current estimated value P_Djand the summation R_SUMof the average value ratios R_j.

[0091] Preferably, the method further comprises the steps of calculating first current values I_Dj=P_Dj+Rj+R_j×|P_DSUM/R_SUM| for strobe points corresponding to a positive value defect current estimated value P_Djamong the plurality of strobe points and detecting first current values which are larger than an absolute value of an allowable error E of a measured value of the quiescent power supply current among the calculated first current values I_Djas defect current values.

[0092] Alternatively, the method further comprises the steps of calculating a second current value I_D=(1−1/P_STQ)×Iq based on a variation rate P_STQ=Iq×(R_b−R_s)/(I_Qb−I_Qs) when each of the calculated first current values I_Djis not more than the absolute value of the allowable error E and detecting that each of the measured values at the plurality of strobe points includes a defect current value indicated by the second current value I_Dwhen the second current value I_Dis larger than the absolute value of the allowable error E, where R_bis the biggest average value ratio among the average value ratios R_j, I_Qbis the measured value at the strobe point corresponding to the biggest average value ratio R_b, R_sis the smallest average value ratio among the average value ratios R_j, and I_Qsis the measured value at the strobe point corresponding to the smallest average value ratio R_s.

[0093] Here, preferably, the method further comprises the step of detecting that each of the measured values at the plurality of strobe points does not include a defect current value when the second current value I_Dis not more than the absolute value of the allowable error E.

[0094] Preferably the allowable error E of the measured quiescent power supply current is whichever of a product (E_R×Iq) of a maximum error rate E_Ramong error rates E_RJof average value ratios R_jof the quiescent power supply current at the plurality of strobe points and the average value Iq and of a maximum measurement error E_Mhas a larger absolute value.

[0095] Preferably, the average value is defined as an arithmetic average value of the measured values at the plurality of strobe points.

[0096] Preferably, the average value ratios of the quiescent power supply currents at the plurality of strobe points are ratios between the plurality of measured values obtained by applying the test signal to a good CMOS integrated circuit and measuring the quiescent power supply current at a plurality of strobe points and the average value of the plurality of measured values which are then in turn averaged for a plurality of good CMOS integrated circuits or for more than a plurality of good CMOS integrated circuits.

[0097] According to the present invention, there is provided an apparatus for analyzing a CMOS integrated circuit, comprising: a means for applying a test signal to a CMOS integrated circuit under test and measuring a quiescent power supply current at a predetermined plurality of strobe points; a means for calculating defect current estimated values P_Dj=I_Qj−R_j×Iq corresponding to the strobe points based on average value ratios R_jof quiescent power supply currents at the plurality of strobe points calculated in advance for a good CMOS integrated circuit, measured values I_Qjat the plurality of strobe points, and an average value Iq of the measured values; a means for calculating a summation P_DSUMof negative value defect current estimated value P_Djamong the calculated defect current estimated values P_Dj; a means for calculating a summation R_SUMof the average value ratios R_jfor the strobe points corresponding to the negative value defect current estimated values P_Djamong the plurality of strobe points; and a means for calculating an absolute value |P_DSUM/P_SUMof a ratio of the summation P_DSUMof the defect current estimated value P_Djand the summation R_SUMof the average value ratios R_j.

[0098] In both the method and apparatus of testing and the method and apparatus of analyzing a CMOS integrated circuit according to the present invention, preferably, the test signal has a plurality of test patterns and the method further comprises a step of switching test patterns between adjacent strobe points of the plurality of strobe points.

[0099] That is, in the method and apparatus of testing a CMOS integrated circuit according to the present invention, since use is made of average value ratios of quiescent power supply currents at a plurality of strobe points calculated in advance for a good CMOS integrated circuit and an average value of measured values of the quiescent power supply currents, it is possible to determine the distribution of quiescent power supply currents in a good CMOS integrated circuit with reference to the average value of the quiescent power supply currents and therefore learn the ranges which measured values can take when the CMOS integrated circuit under test is a good one from the average value ratios of the quiescent power supply currents and the average value of the measured values.

[0100] For example, the method and apparatus calculates defect current estimates P_Dj=I_Qj−R_j×Iq based on the average value ratios R_jof quiescent power supply currents at the plurality of strobe points, the measured values I_Qj, and the average value Iq of the measured values.

[0101] By multiplying the average value Iq of the measured values by an average value ratio R_j, it is possible to calculate the expected value of a quiescent power supply current I_DDQin a case where the CMOS integrated circuit under test is a good one. By subtracting this expected value (R_j×Iq) from a measured value I_Qj, it is possible to obtain a defect current estimate.

[0102] Note that a defect current estimate P_Djcan be predicted from a defect current I_Djat that strobe point, the summation of the defect currents at all strobe points, the total number m of the strobe points, and the average value ratio R_j, so a value not depending on the size of a measured value of the quiescent power supply current can be obtained.

[0103] Further, the method and apparatus for example calculates defect current predictions P_Q1j=I_Qj−R_j×I_Q1/R₁based on the average value ratios R_jof the quiescent power supply currents at a plurality of strobe points, the measured values I_Qj, and the ratio (I_Q1/R₁).

[0104] By multiplying the ratio (I_Q1/R₁) corresponding to the prediction of the average value of the measured values and an average value ratio R_j, it is possible to calculate an expected value of the quiescent power supply current I_DDQin a case where the CMOS integrated circuit under test is a good device. By subtracting this expected value (I_Q1×R_j/R₁) from a measured value I_Qj, it is possible to obtain a defect current prediction.

[0105] Note that the defect current prediction P_Q1jcan be projected from the defect currents I_Djand I_D1at strobe points and the average value ratios R_jand R₁, so a value not depending on the size of the measured value of the quiescent power supply current can therefore be obtained.

[0106] The method for example calculates an upper limit threshold I_QU=(1+f×σ_R)×I_Q1/R₁based on the average value ratio R₁of the quiescent power supply current at the predetermined strobe point, the ratio (I_Q1/R₁), and a constant f of 4 to 7.

[0107] By multiplying the ratio (I_Q1/R₁) corresponding to the prediction of average value of the measured values by (f×σ_R) relating to the standard deviation of the average value ratios R₁, it is possible to calculate the range of distribution of the quiescent power supply current I_DDQin the case where the CMOS integrated circuit under test is good. By adding this range (f×σ_R×I_Q1/R₁) and the ratio (I_Q1/R₁), it is possible to obtain the upper limit threshold.

[0108] σ_Rcan be calculated from an average value ratio R₁and the total number m of the strobe points, so a value not depending on the measured value of the quiescent power supply current can be obtained.

[0109] The method for example calculates the lower limit threshold I_QL=(1−f×σ_R)×I_Q1/R₁based on the average value ratio R₁of the quiescent power supply current at the predetermined strobe point, the ratio (I_Q1/R₁), and a constant f of 4 to 7.

[0110] By multiplying the ratio (I_Q1/R₁) corresponding to the prediction of the average value of the measured values by (f×σ_R) relating to the standard deviation of the average value ratios R₁, it is possible to calculate the range of distribution of the quiescent power supply current I_DDQthe case where the CMOS integrated circuit under test is good. By subtracting this range (f×σ_R×I_Q1/R₁) from the ratio (I_Q1/R₁), it is possible to obtain the lower limit threshold.

[0111] σ_Rcan be calculated from the average value ratio R₁and the total number m of the strobe points, so a value not depending on the measured value of the quiescent power supply current can be obtained.

[0112] The method of analyzing a CMOS integrated circuit according to the present invention calculates defect current estimates P_Dj=I_Qj−R_j×Iq based on the average value ratios R_jof quiescent power supply currents at a plurality of strobe points, measured values I_Qj, and an average value Iq.

[0113] By multiplying the average value Iq of the measured values by an average value ratio R_j, it is possible to calculate the expected value of a quiescent power supply current I_DDQin the case where the CMOS integrated circuit under test is good. By subtracting this expected value (R_j×Iq) from a measured value I_Qj, it is possible to obtain a defect current estimate.

[0114] By calculating the absolute value |R_DSUM/R_SUM| of a ratio between the summation P_DSUMof the negative value defect current estimated value P_Djand the summation R_SUMof the average value ratios R_Jfor the strobe points corresponding to the negative value defect current estimates P_DJ, it is possible to obtain a variation ΔIq showing the difference between the average value of the measured values and the average value of FET leakage currents.

[0115] As explained above, according to the methods of testing and analyzing a CMOS integrated circuit of the present invention, it is possible to detect and analyze defect currents even with a CMOS integrated circuit with a large quiescent power supply current and large variations.

[0116] These and other objects and features of the present invention will become clearer from the following description of the preferred embodiments given with reference to the accompanying drawings, in which:

[0117] FIG. 1 is a graph illustrating the relative frequency distribution of the maximum I_DDQin two kinds CMOS integrated circuit;

[0118] FIG. 2 is a graph showing classifications of leakage currents of a MOSFET;

[0119] FIG. 3 is a graph illustrating the distribution of the maximum I_DDQin a CMOS integrated circuit;

[0120] FIG. 4 is a graph of the distribution of leakage currents in two good CMOS integrated circuit;

[0121] FIG. 5 is a graph of the relative frequency distribution of the FET leakage currents at all strobe points in the CMOS integrated circuit of FIG. 4;

[0122] FIG. 6 is a graph illustrating the relative frequency distribution of the FET leakage current with respect to 32 good CMOS integrated circuits;

[0123] FIG. 7 is a graph of another example of distribution of an estimated error ratio of a reference average value ratio;

[0124] FIG. 8 is a graph of another example of distribution of an estimated error ratio of a reference average value ratio;

[0125] FIG. 9 is a schematic block diagram of the configuration of a test device for the method of testing a CMOS integrated circuit according to the present invention;

[0126] FIGS. 10A to 10F are schematic flow chart for Illustrating the test method which the test device in FIG. 9 performs in relation to the average value method;

[0127] FIGS. 11A and 11B are schematic flow chart for illustrating the test method which the test device in FIG. 9 performs in relation to the one point method; and

[0128] FIG. 12 is a schematic flow chart for Illustrating the test method which the test device in FIG. 9 performs in relation to the upper and lower limit method of the one point method.

[0129] Below, preferred embodiments will be described with reference to the accompanying drawings.

[0130] First, an average value ratio R_jand the calculation of the same will be described.

[0131] The FET leakage current I_Tj(Z =1 to m) at each of m number of strobes can be expressed by the following formula (4) from the above formula (3): $\begin{array}{cc}{I}_{\mathrm{Tj}}=N_{\mathrm{N3j}}\times I_{\mathrm{N3}}+N_{\mathrm{P3j}}\times I_{\mathrm{P3}}& \text{(5-1)}\\ I_t=\frac{1}{m}\sum _{j=1}^m{I}_{\mathrm{Tj}}=\frac{1}{m}\sum _{j=1}^m\left(N_{\mathrm{N3j}}\times I_{\mathrm{N3}}+N_{\mathrm{P3j}}\times I_{\mathrm{P3}}\right)=n_{\mathrm{N3}}\times I_{\mathrm{N3}}+n_{\mathrm{P3}}\times I_{\mathrm{P3}}& \text{(5-2)}\\ n_{\mathrm{N3}}=\frac{1}{m}\sum _{j=1}^m{N}_{\mathrm{N3j}}& \text{(5-3)}\\ n_{\mathrm{P3}}=\frac{1}{m}\sum _{j=1}^m{N}_{\mathrm{P3j}}& \text{(5-4)}\end{array}$

[0132] Here, N_Nijand N_Pijare made N_Niand N_Pi(i=1 to 3) at the strobe number Z.

[0133] Note that N_Niand N_piare the numbers of the n type MOSFETs and p type MOSFETs where leakage currents of the case L_i(i=1 to 3) occur as mentioned above.

[0134] The FET leakage current I_Tjcan be expressed by the following formula (5-1) since the case L₃becomes predominant along with miniaturization. The average value It of the FET leakage currents I_Tjcan be expressed by the following formula (5-2): $\begin{array}{cc}{I}_{\mathrm{Tj}}=\sum _{i=1}^3\left(N_{\mathrm{Nij}}\times I_{\mathrm{Ni}}+N_{\mathrm{Pij}}\times I_{\mathrm{Pi}}\right)& \left(4\right)\end{array}$

[0135] Here, n_N3is the average of the number N_N3jat all strobes (average number) as shown in the above formula (5-3). Further, n_P3is the average of the number N_P3jat all strobes (average number) as shown in the above formula (5-4).

[0136] The average value ratio R_jis defined by the following formula (6).

[0137] R_j=I_Tj/It  (6)

[0138] When formula (5-2) is entered into formula (6), the following formula (7) is obtained: $\begin{array}{cc}{R}_j=\frac{N_{\mathrm{N3j}}\times I_{\mathrm{N3}}+N_{\mathrm{P3j}}\times I_{\mathrm{P3}}}{n_{\mathrm{N3}}\times I_{\mathrm{P3}}+n_{\mathrm{P3}}\times I_{\mathrm{P3}}}=\frac{N_{\mathrm{N3j}}}{n_{\mathrm{N3}}+n_{\mathrm{P3}}\times \frac{I_{\mathrm{P3}}}{I_{\mathrm{N3}}}}+\frac{N_{\mathrm{P3j}}}{n_{\mathrm{N3}}\times \frac{I_{\mathrm{N3}}}{I_{\mathrm{P3}}}+n_{\mathrm{P3}}}& \left(7\right)\end{array}$

[0139] The numbers N_N3jand N_P3jare unambiguously determined by the full terminal voltage of the MOSFETs in the CMOS integrated circuit. Manufacturing variations are irrelevant. That is, if the layout of the circuitry of the CMOS integrated circuit and test pattern are the same, these numbers become the same in all CMOS integrated circuits.

[0140] In this case, the average numbers n_N3and n_P3become the same in all CMOS integrated circuits.

[0141] Further, the average value ratios R_jbecome the same in all CMOS integrated circuits regardless of the size of the FET leakage currents so long as the ratio of the FET leakage currents I_N3and I_P3is constant.

[0142] From the above formula (6), the FET leakage current I_Tjcan be expressed by the following formula (8):

[0143] Further, the dispersion (σ_T)²of the FET leakage currents with respect to all strobes is expressed by the following formula (9-1), while the standard deviation σ_Tis expressed by the following formula (9-2): $\begin{array}{cc}{I}_{\mathrm{Tj}}=R_j\times \mathrm{It}& \left(8\right)\\ {\left({\sigma }_T\right)}^2=\frac{1}{m}\sum _{j=1}^m{\left(I_{\mathrm{Tj}}-I_t\right)}^2& \text{(9-1)}\\ {\sigma }_T=\sqrt{\frac{1}{m}\sum _{j=1}^m{\left(I_{\mathrm{Tj}}-I_t\right)}^2}& \text{(9-2)}\end{array}$

[0144] When entering formula (8) into the above formulas (9-1) and (9-2), the following formulas (10-1), (10-2), and (10-3) are obtained: $\begin{array}{cc}{\left({\sigma }_T\right)}^2=\frac{1}{m}\sum _{j=1}^m{\left(R_j\times I_t-I_t\right)}^2=\frac{{\left(I_t\right)}^2}{m}\sum _{j=1}^m{\left(R_j-1\right)}^2& \text{(10-1)}\\ {\sigma }_T=I_t\sqrt{\frac{1}{m}\sum _{j=1}^m{\left(R_j-1\right)}^2}=I_t\times {\sigma }_R& \text{(10-2)}\\ {\sigma }_T=\sqrt{\frac{1}{m}\sum _{j=1}^m{\left(R_j-1\right)}^2}& \text{(10-3)}\end{array}$

[0145] The standard deviation σ_Tis directly proportional to the average value It of the FET leakage currents. The proportional coefficient σ_Rdoes not depend on the size of the FET leakage currents I_Tj.

[0146] The reference average value ratio (standard average value ratio) R_j^REFis obtained from circuit simulation or FET leakage current simulation or obtained from measured values etc. of the quiescent power supply currents in a plurality of good CMOS integrated circuits. When the average value It can be estimated from measurement data, the FET leakage currents I_Tjcan be calculated from the above formula (8).

[0147] The error rate (estimated error ratio) E_Rjof the reference average value ratio R_j^REFis defined by the following formula (11):

[0148] E_Rj=(R_j^REF×It−I_Tj)/I_Tj  (11)

[0149] FIGS. 7 and 8 are charts illustrating the distributions of the estimate error rates of reference average value ratios. The average value ratios R_j( Z=1 to m) of 32 CMOS integrated circuits of FIG. 6 were calculated for each strobe point, a reference average value ratio R_j^REFwas calculated by averaging the 32, and the estimated error rate E_Rjof the reference average value ratio R_j^REFwas plotted by a white circle. Each plot in the figures corresponds to one strobe point. The abscissas show the FET leakage currents (μA), while the ordinates show the estimated error rates (%).

[0150] In FIG. 7, the FET leakage currents are distributed from about 290 to 360 μA. In FIG. 8, the FET leakage currents are distributed from about 58 to 80 μA. Despite the fact that the centers of the distribution differ, the estimated error rate E_Rjis kept to within about ±6% and is constant or is almost constant without depending on the average value It.

[0151] Note that in the chart of FIG. 6, about 70 μA was the peak and the estimated error was ±4.2 (=70×6%) μA

[0152] This estimated error is within 6 μA, that is, the maximum measurement error E_Mof an ammeter for measuring the quiescent power supply current I_Q.

[0153] When the estimated error rate E_Rjis within ±E_Ror when the maximum error rate comprised of the estimated error rate E_Rjwith the largest absolute value is designated as E_R, (R_j^REF−|E_R|)×It≦( R_j^REF+|E_R|)×It stands.

[0154] Note that, separate from the estimated error, the larger of the absolute value of the measurement error E_Mdecided by the range of the ammeter of the measurement unit for measuring the quiescent power supply current I_DDQand the absolute value of the estimated error (E_R×It) is used as the error of the criteria of judgement, that is, the allowable error E.

[0155] The method of estimating the average value It of the FET leakage currents I_Tjfrom the quiescent power supply current I_Qwhich may contain a defect current is shown next.

[0156] When the above formula (1) is applied to the strobe point of the strobe number Z (strobe point Z), the measured values I_Qjcorresponding to the strobe point Z is shown by the following formula (12-1), the average value Iq of the measured values I_Qjat all strobe points is shown by the following formula (12-2), and the ΔIq shown by the following formula (12-3) can be obtained. $\begin{array}{cc}{I}_{\mathrm{Qj}}=I_{\mathrm{Tj}}+I_{\mathrm{Dj}}& \text{(12-1)}\\ I_q=\frac{1}{m}\sum _{j=1}^m{I}_{\mathrm{Qj}}=\frac{1}{m}\sum _{j=1}^m(I_{\mathrm{Tj}}+I_{\mathrm{Dj}}=\frac{1}{m}\sum _{j=1}^m{I}_{\mathrm{Tj}}+\frac{1}{m}\sum _{k=1}^d{I}_{\mathrm{Dk}}=I_t+\Delta I_q& \text{(12-2)}\\ \Delta I_q=\frac{1}{m}\sum _{k=1}^d{I}_{\mathrm{Dk}}& \text{(12-3)}\end{array}$

[0157] Here, I_Dk(k=1 to d) is the defeat current I_Dat a succession of d number of strobe points among the m number of the strobe points.

[0158] Further, ΔIq is the defect current averaged for the number of points m. As shown in the above formula (12-2), it may be considered the variation superposed on the average value It of the FET leakage currents at the average value Iq.

[0159] Moreover, the average value It of the FET leakage currents can be expressed by the following formula (13).

[0160] It=Iq−ΔIq  (13)

[0161] When entering the above formula (13) in formula (8), I_TjR_j×(Iq−ΔIq) can be obtained. When applying this formula to the above formula (2), the following formula (14) can be obtained:

[0162] I_Dj=I_Qj−R_j×(Iq−ΔIq)

[0163] =I_Qj−R_j×Iq+R_j×ΔIq  (14)

[0164] Here, the defect current estimated value P_Djis defined as P_Dj=I_Qj−R_j×Iq.

[0165] This defect current estimated value P_Djcan be calculated during the testing of the CMOS integrated circuit from an average value ratio R_jcalculated in advance and the measured quiescent power supply current I_Q(measured value I_Qj).

[0166] From the above formula (14), the following formula (15) can be obtained regarding the defect current estimate P_DJ: $\begin{array}{cc}{P}_{\mathrm{Dj}}=I_{\mathrm{Dj}}-R_j\left(\frac{1}{m}\sum _{k=1}^d{I}_{\mathrm{Dk}}\right)& \left(15\right)\end{array}$

[0167] The right side of the above formula (15) depends on only the defect current at the strobe point, the constant R_j, the summation of the defect currents, and the total number m of the strobe points and does not depend on the measured value I_Qjof the quiescent power supply current and the average value Iq.

[0168] The defect current estimated value P_Djbecomes smaller from the defect current I_Djby exactly the second term when there is a defect current I_Djat the strobe point Z.

[0169] On the other hand, when there is no defect current at the strobe point Z, the defect current estimate becomes a negative value of the amount of the second term of the right side and shows that a defect current is flowing at other than the strobe point Z.

[0170] In this way, when the defect current estimated value P_Djis above E, it means that there is a defect current at the strobe point Z, while when the defect current estimated value P_DJis below −E, it means that there is a defect current at other than the strobe point Z.

[0171] Here, the value of the allowable error E used for comparison is made for example the value max(|E_R×Iq|, |E_H|) of the larger of the absolute value of E_R×Iq and the absolute value of the maximum measurement error E_M.

[0172] The condition for judging the CMOS integrated circuit as good may also be made that −E≦P_Dj≦E stands for all strobe points.

[0173] However, when there is a defect current of the same size at all strobe points, the following formula (16) can be obtained as a special case from the above formula (15). Here, I_Dkis I_Dand d is m.

[0174] P_Dj=(1−R_j)×I_D  (16)

[0175] Therefore, the defect current estimate P_Djbecomes the maximum at the strobe point s where the average value ratio R_jbecomes the smallest value R_s. Further, the minimum value I_DMINof the defect currents I_Dcan be expressed by the following formula (17):

[0176] I_DMIN=E/(1−R_s)  (17)

[0177] When the defect rate unusually rises in the testing of CMOS integrated circuits or when controlling good or defective, detection of the defect current value is often needed.

[0178] In the above formula (15), when finding the negative value defect current estimates P_DJ, that is, the sums P_DSUMof the defect current estimates P_Djat strobe points h (=1 to q) where there is no defect current, the following formulas (18-1), (18-2), and (18-3) can be obtained: $\begin{array}{cc}{P}_{\mathrm{DSUM}}=\Delta I\sum _{h=1}^g\mathrm{Rh}& \text{(18-1)}\\ \Delta I_q=\frac{-P_{\mathrm{DSUM}}}{R_{\mathrm{SUM}}}& \text{(18-2)}\\ R_{\mathrm{SUM}}=\sum _{h=1}^g\mathrm{Rh}& \text{(18-3)}\end{array}$

[0179] The value of the above formula (18-3) is the summation R_SUMof the average value ratios R_hat the strobe points h (=1 to q) and can be calculated, therefore it is also possible to calculate ΔIq.

[0180] Thus, the following formula (19) can be obtained at the strobe points k (=1 to d) where the defect current estimated value P_Djare above 0 by using the above formula (15), and the defect currents I_Dkcan be calculated.

[0181] I_Dk=P_Dk+R_k×ΔIq  (19)

[0182] Here, the value of the quiescent power supply current I_Qat the strobe point b where the average value ratio R_jbecomes the biggest value (biggest average value ratio) R_bamong all strobe points is designated as I_Qb, while the value of the quiescent power supply current I_Qat the strobe point where the average value ratio R_jbecomes the smallest value (smallest average value ratio) R_sis designated as I_Qs.

[0183] Further, the quiescent power supply current I_Qbis assumed to have the FET leakage current I_Tband the defect current I_Db, and the quiescent power supply current I_Qsis assumed to have the FET leakage current I_Tsand defect current I_Ds.

[0184] The amplitude ratio R_SQof the amplitude difference of the quiescent power supply currents I_Qband I_Qsat the strobe points b and s and the average value Iq is expressed by the following formula (20):

[0185] R_SQ=(I_Qb−I_Qs)/Iq

[0186] ={(I_Tb+IDb) (I_Ts+I_Ds)}/Iq  (20)

[0187] The amplitude ratio R_SQwhen the same defect current I_Dexists at all the strobe points is shown by the following formula (21): $\begin{array}{cc}{R}_{\mathrm{SQ}}=\left\{\left(I_{\mathrm{Tb}}+I_D\right)-\left(I_{\mathrm{Ts}}+T_D\right)\right\}/\mathrm{Iq}=\left(I_{\mathrm{Tb}}-I_{\mathrm{Ts}}\right)/\left(\mathrm{It}+I_D\right)=\left(R_b-R_s\right)\times \mathrm{It}/\left(\mathrm{It}+I_D\right)& \left(21\right)\end{array}$

[0188] If R_STis defined as being equal to R_b-R_sand the amplitude ratio variation rate P_STQis defined as R_ST/R_SQ, this amplitude ratio variation rate P_STQcan be calculated during testing and the following formula (22) can be obtained from the above formula (21):

[0189] P_STQ=1+I_D/It  (22)

[0190] From the above formula (22), an amplitude ratio variation rate P_STQof above 1 can be used as the condition to show the existence of a defect current I_D, and a CMOS integrated circuit can be judged as a defect based on this.

[0191] On the other hand, an amplitude ratio variation rate P_STQof less than or equal to 1 can be used as the condition to show the lack of a defect current I_D, and a CMOS integrated circuit can be judged as good based on this.

[0192] By entering It=Iq-I_Dinto the above formula (22), the following formula (23) can be obtained and the defect current I_Dcan be found.

[0193] I_D=(1—1/P_STQ)×Iq  (23)

[0194] In the average value method, the measured values of the quiescent power supply currents are averaged until obtaining the necessary judgment error, so m number of strobe points are necessary.

[0195] However, although the accuracy is reduced, the average value It of the FET leakage currents I_Tjcan also be estimated from one measured value of the quiescent power supply current I_Q.

[0196] First, a ratio R_Qjof a quiescent power supply current I_Qjand an average value ratio R_jis set as R_Qj=I_Qj/R_j.

[0197] When using the one point method at the first strobe point, that is, where Z is 1, the following formula (24) can be obtained. Note that the ratio R_Q1corresponds to the prediction of the average value Iq of measured values I_Qj. $\begin{array}{cc}{R}_{\mathrm{Q1}}=I_{\mathrm{Q1}}/R_{\mathrm{Q1}}=\left(I_{\mathrm{T1}}+I_{\mathrm{D1}}\right)/R_1=\left(I_{\mathrm{T1}}/R_1\right)+\left(I_{\mathrm{D1}}/R_1\right)& \left(24\right)\end{array}$

[0198] Here, since It is I_T1/R₁, the following formula (25) can be obtained from the above formula (24):

[0199] It=R_Q1−I_D1/R₁  (25)

[0200] Further, since I_Qjis I_Tj+I_Dj, the following formula (26) can be obtained from the above formula (25): $\begin{array}{cc}\begin{array}{c}{I}_{\mathrm{Dj}}=I_{\mathrm{Qj}}-I_{\mathrm{Tj}}\\ =I_{\mathrm{Qj}}-R_j\times \mathrm{It}\\ =I_{\mathrm{Qj}}-R_j\times \left(R_{\mathrm{Q1}}-I_{\mathrm{D1}}/R_1\right)\\ =I_{\mathrm{Qj}}-R_j\times R_{\mathrm{Q1}}+R_j\times \left(I_{\mathrm{D1}}/R_1\right)\end{array}& \left(26\right)\end{array}$

[0201] Here, when the defect current prediction P_Q1jis defined as I_Qj−R_j×R_Q1, the following formula (27) can be obtained from the above formula (26): $\begin{array}{cc}\begin{array}{c}{R}_{\mathrm{Q1j}}=I_{\mathrm{Qj}}-R_j\times R_{\mathrm{Q1}}\\ =I_{\mathrm{Dj}}-R_j\times \left(I_{\mathrm{D1}}/R_1\right)\end{array}& \left(27\right)\end{array}$

[0202] From formula (27),the defect current prediction P_Q1jbecomes a value which depends on only the defect current I_Djat a strobe point Z, the defect current I_D1at the first strobe point, the average value ratio R₁at the first strobe point, and an average value ratio R_jand does not depend on the measured values of the quiescent power supply currents or the average value Iq thereof.

[0203] Here, when the defect currents I_D1is 0, the condition for detection of a defect current at the strobe point becomes P_Q1j>E.

[0204] When the defect current I_D1is not 0, if P_Q1jis above E, there is a defect current I_Djat the strobe point Z.

[0205] When P_Q1jis below −E, there is a defect current I_Djat a strobe point other than the strobe point Z.

[0206] Note that the allowable error E used for comparison is made the value max(|E_R×R_Q1|,|E_H|) of the larger of the absolute value of E_R×R_Q1and the absolute value of the maximum measurement error E_M.

[0207] When the FET leakage currents are distributed by the normal distribution or similar to the normal distribution as shown in FIG. 5, each FET leakage current I_Tjis within about five or six times the standard deviation σ_Tfrom the average value. This may be utilized to judge the good or defective of the CMOS integrated circuit.

[0208] In the upper and lower limit method, an upper limit threshold I_QUand lower limit threshold I_QLare set as shown in the following formulas (28-1) and (28-2) by using a proportional coefficient σ_Rof the above formula (10-3): $\begin{array}{cc}\begin{array}{c}{I}_{\mathrm{QU}}=\left(1+6{\sigma }_R\right)\times R_{\mathrm{Q1}}\\ =\left(1+6{\sigma }_R\right)\times I_{\mathrm{Q1}}/R_1\end{array}& \text{(28-1)}\\ \begin{array}{c}{I}_{\mathrm{QL}}=\left(1-6{\sigma }_R\right)\times R_{\mathrm{Q1}}\\ =\left(1-6{\sigma }_R\right)\times I_{\mathrm{Q1}}/R_1\end{array}& \text{(28-2)}\end{array}$

[0209] In the upper and lower limit method, it is possible set the upper limit threshold I_Quand the lower limit threshold I_QLbased on the measured value at the first strobe point and therefore test at a high speed.

[0210] However, it is difficult to detect a defect current present in a CMOS integrated circuit having the lowest limit FET leakage current when the defect current is less than the range of the upper and lower limits (I_QU-I_QL) therefore this is effective when the range is small.

[0211] FIG. 9 is a schematic block diagram of the configuration of a testing device for the test method of CMOS integrated circuit according to the present invention.

[0212] This testing device 100 has a control unit 21, a signal drive unit 31, a measurement unit 32, a measured I_DDQtable memory 41, an average value ratio table memory 42, a good or defective judgement unit 43, an output-unit 44, and first to third memories 11 to 13.

[0213] The first memory 11 stores test pattern information, that is, information for showing a test pattern TP used for I_DDQtesting, and strobe information, that is, information showing strobe points being measured.

[0214] The second memory 12 stores the information of the average value ratios R_j(R_j^REF) found from the PET leakage current simulation before testing or the information of the average value ratios R_j(R_j^REF) found from measured values I_Qjobtained by measuring the quiescent power supply currents I_Qof a plurality of good CMOS integrated circuits.

[0215] The third memory 13 stores information for showing estimated error corresponding to the average value ratios R_j(R_j^REF) stored in the second memory or the estimated error rates E_Rjand information for showing the maximum 10 measurement error E_Mof the measurement unit 32.

[0216] The control unit 21 is a controller for controlling the entire testing device 100.

[0217] The control unit 21 is supplied with information showing the test pattern and information showing the strobe points from the first memory 11, information showing the average value ratios from the second memory 12, and information showing the estimated errors or the estimated error rates E_Rjand information showing the maximum measurement error E_Mfrom the third memory 13.

[0218] The control unit 21 supplies the information showing the test pattern and information showing the strobe points to the signal drive unit 31.

[0219] Further, the control unit 21 converts the information showing the average value ratios into an 25 average value ratio table and supplies it to the average value ratio table memory 42.

[0220] Moreover, the control unit 21 supplies the information showing the estimated errors or the estimated error rates E_Rjand information showing the maximum measurement error E_Mof the measurement unit 32 to the good or defective judgement unit 43.

[0221] The signal drive unit 31 applies a test signal of a predetermined test pattern to a CMOS integrated circuit 50 under test under the control of the control unit 21.

[0222] Further, the signal drive unit 31 supplies a timing signal showing the strobe points (strobe pulses) to the measurement unit 32 when a test signal is applied.

[0223] The measurement unit 32 measures the quiescent power supply currents I_Qof the CMOS integrated circuit 50 according to the strobe pulses, converts the measured values I_Qjinto the measured I_DDQtable, and supplies it to and stores it in the measured I_DDQtable memory 41.

[0224] The measured I_DDQtable memory 41 and the average value ratio memory 42 are constituted by memories readable at a high speed by designating the strobe number Z.

[0225] The control unit 21 activates the good or defective judgement unit 43 when finishing storing the measured values I_QJfor all the strobe points or the measured values I_Qjfor predetermined strobe points.

[0226] The good or defective judgement unit 43 judges the good or defective of the CMOS integrated circuit 50 based on the average value ratios R_j(R_j^REF) and the measured values I_Qjetc. and outputs a judgement signal showing the measurement results, judgement results, various calculation results, etc. to the output unit 44.

[0227] The output unit 44 has a display, printer, or other output device and displays on a display screen or prints on a print medium the measurement results, judgement results, various calculation results, etc. based on the judgement signal from the good or defective judgement unit 43.

[0228] The output unit 44 may also be constituted by a computer display and the control unit 21, the measured I_DDQtable memory 41, the average value ratio table memory 42, and the good or defective judgement unit 43 may be constituted by a computer or a microcomputer. The control unit 21 and the good or defective judgement unit 43 may also be provided as a single unit.

[0229] Next, the method of testing a CMOS integrated circuit which the test device 100 performs will be explained.

[0230] FIGS. 10A to 10C are schematic flow charts illustrating the test method which the test device 100 performs in relation to the average value method.

[0231] The test device 100 is made one for which the reference average value ratios R_j^REFand the error rates (estimated error rates) E_Rjhave been calculated in advance for a predetermined number of samples of CMOS integrated circuits and one for which the maximum measurement error E_Mis known in advance. Further, the reference average value ratios R_j^REFare used as the average value ratios R_jwhen testing.

[0232] First, at the start, the value of the strobe point variable is initially set to “1”, where after the routine proceeds to step S1.

[0233] At step S1, the signal drive unit 31 applies a test signal of a predetermined test pattern to the CMOS integrated circuit 50, and the measurement unit 32 measures the quiescent power supply currents at the strobe points Z to obtain the measured values I_Qj.

[0234] At step S2, the control unit 21 judges whether measurement of the quiescent power supply currents at all of the predetermined strobe points Z (=1 to m) has finished.

[0235] When the predetermined measurement has been finished, the measured values I_Qjare stored in the table memory 41 and supplied to the good or defective judgement unit 43, where after the routine proceeds to step S4.

[0236] When the predetermined measurement has not been finished, the routine proceeds to step S3, where the value of Z is incremented to increase it by exactly “1”, where after the routine returns to step S1 where the next measurement is performed.

[0237] At step S4, the good or defective judgement unit 43 calculates the average value Iq of the plurality of measured values I_Qj(Z=1 to m) obtained from the CMOS integrated circuit 50 at the strobe points.

[0238] The average value Iq is (I_Q1+I_Q2+. . .+I_Qm)/m. Note that, for example, m is made at least 2. The value of m may even be set to about 800 as an example.

[0239] It is also possible not to average all the measured values I_Qjbut rather to represent this by the average of the measured values of the first n number of strobe points. For example, n may be an integer larger than 2, n is less than m, and Iq may then be (I_Q1+I_Q2+. . . +I_Qn)/n.

[0240] At step S5, the good or defective judgement unit 43 calculates the corresponding defect current estimate P_Dj=I_Qj−R_j×Iq for each strobe point Z (Z =1 to m).

[0241] At step S6, the good or defective judgement unit 43 calculates the allowable error E. The allowable error E is made the larger of the absolute value of the estimated error rate E_R×Iq and the absolute value of the maximum measurement error E_M.

[0242] At step S7, the good or defective judgement unit 43 sets the strobe point variable Z to “1”.

[0243] At step S8, the good or defective judgement unit 43 compares the defect current estimated value P_Djwith the allowable error E.

[0244] When P_Djis greater than E, the routine proceeds to step S9, where the good or defective judgement unit 43 detects that there is a defect current at the strobe point Z and judges that the CMOS integrated circuit 50 under test is a defect, the output unit 44 outputs the judgement results etc., and the test ends. Note that it is also possible to judge the existence of a defect current for all the strobe points.

[0245] When P_Djis less than E, the routine proceeds to step S10.

[0246] At step S10, the good or defective judgement unit 43 compares the defect current estimated value P_Djwith −E.

[0247] When P_Djis below −E, the routine proceeds to step 11, where the good or defective judgement unit 43 detects that there is a defect current at a strobe point other than the strobe point Z and judges that the CMOS integrated circuit 50 under test is a defect, the output unit 44 outputs the judgement results etc., and the test ends. Note that it is also possible to judge the existence of a defect current for all the strobe points.

[0248] When P_Djis not below −E, the routine proceeds to step S12.

[0249] At step 12, the good or defective judgement unit 43 judges whether or not the comparison of the defect current estimated value P_Djhas been finished for all the strobe points.

[0250] When the comparison is not finished, the routine proceeds to step 13, where the value of the variable Z is incremented to increase it by exactly “1”, then the test returns to step S8 where the next comparison is performed.

[0251] When the comparison is finished, the routine proceeds to step 14.

[0252] At step 14, the amplitude ratio variation rate P_STQis calculated. P_STQis Iq×(R_b−R_s)/(I_Qb−I_Qs).

[0253] Note that the average value ratio R_bis the biggest value of the average value ratio R_j. The measured value at the strobe point b at this time is I_Qb.

[0254] Moreover, the average value ratio R_sis the smallest value of the average value ratio R_j. The measured value at the strobe point s at this time is I_Qs.

[0255] At step 15, the good or defective judgement unit 43 judges whether or not the amplitude ratio variation rate P_STQis greater than 1.

[0256] When P_STQis above 1, the good or defective judgement unit 43 detects that there is a defeat current at all the strobe points Z (=1 to m) and judges that the CMOS integrated circuit 50 under test is a defect, the output unit 44 outputs the judgement results etc., and the test ends.

[0257] When P_STQis not greater than 1, the routine proceeds to step 17, where the good or defective judgement unit 43 judges that the CMOS integrated circuit 50 under test is good, the output unit 44 outputs the judgement results etc., and the test ends.

[0258] Next, the method of analyzing a CMOS integrated circuit will be explained.

[0259] FIGS. 10D to 10F are schematic flow charts illustrating the analyzing method which the test device 100 performs in relation to the average value method.

[0260] First, at the start, the value of the strobe point variable Z is initially set to “1”, where after the routine proceeds to step S21.

[0261] At step S21, the signal drive unit 31 applies a test signal of a predetermined test pattern to the CMOS integrated circuit 50, and the measurement unit 32 measures the quiescent power supply currents at the strobe points Z to obtain the measured values I_Qj.

[0262] At step S22, the control unit 21 judges whether measurement of the quiescent power supply currents at all of the predetermined strobe points Z (=1 to m) has finished.

[0263] When the predetermined measurement has been finished, the measured values I_Qjare stored in the table memory 41 and supplied to the good or defective judgement unit 43, where after the routine proceeds to step S24.

[0264] When the predetermined measurement has not been finished, the routine proceeds to step S23, where the value of Z is incremented to increase it by exactly “1”, where after the routine returns to step S21 where the next measurement is performed.

[0265] At step S24, the good or defective judgement unit 43 calculates the average value Iq of the plurality of measured values I_Qj(Z=1 to m) obtained from the CMOS integrated circuit 50 at the strobe points.

[0266] The average value Iq is (I_Q1+I_Q2+. . . +I_Qm)/m. Note that, for example, m is made at least 2. The value of m may even be set to about 800 as an example.

[0267] At step S25, the good or defective judgement unit 43 calculates the corresponding defect current estimated values P_Dj=I_Qj−R_j×Iq for each strobe point Z (Z =1 to m).

[0268] At step S26, the good or defective judgement unit 43 calculates the allowable error E. The allowable error E is made the larger of the absolute value of the estimated error rate E_R×Iq and the absolute value of the maximum measurement error E_M.

[0269] At step S27, the good or defective judgement unit 43 calculates the summation P_DSUMof the defeat current estimates P_Djand the summation R_SUMof the average value ratios R_j(R_j^REF) for the strobes h having defect current estimates P_Djof a negative value.

[0270] At step S28, the good or defective judgement unit 43 calculates the variation ΔIq of the average value Iq. ΔIq is −P_DSUM/R_SUM.

[0271] At step S29, first current values I_Dkare calculated for strobes k having defect current estimates P_Djlarger than the allowable error E. A first current value I_Dkis P_Dk+R_k×ΔIq.

[0272] At step S30, strobe points k having calculated first current values I_Dkabove the allowable error E are detected as including defect currents shown by the first current values I_Dkand the number of times of detection of the defect currents is counted.

[0273] At step 31, whether or not the number of times of detection of the defect currents is 0 is judged.

[0274] When the number of times of detection is not 0, the routine proceeds to step S36.

[0275] When the number of times of defection is 0, the routine proceeds to step S32 where a second current value I_Dis calculated.

[0276] The second current value I_Dis (1—1/R_STD)×Iq.

[0277] At step 33, whether the second current value I_Dis larger than the allowable error E is judged.

[0278] When the second current value I_Dis over E, the routine proceeds to step S34 where the fact that there is a defect current shown by the second current value I_Dat all the strobe points is detected.

[0279] When the second current value I_Dis not over E, the routine proceeds to step S35 where there is no defect current at any strobe point is detected.

[0280] At step S36, the output unit 44 displays or prints out the measurement results of the CMOS integrated circuit 50, or calculation results of the defect current value, etc.

[0281] FIGS. 11A and 11B are schematic flow charts illustrating the test method which the test device 100 performs in relation to the one point method.

[0282] The test device 100 is made one for which the reference average value ratios R_j^REFand the estimated error rates E_Rjhave been calculated in advance for a predetermined number of samples of CMOS integrated circuits and one for which the maximum measurement error E_Mis known in advance. Further, the reference average value ratios R_j^REFare used as the average value ratios R_jwhen testing.

[0283] First, at step S41, the signal drive unit 31 applies a test signal of a predetermined test pattern to the CMOS integrated circuit 50, and the measurement unit 32 measures the quiescent power supply current at the strobe point 1 to obtain the measured value I_Q1. This measured value I_Q1is stored in the table memory 41 and supplied to the good or defective judgement unit 43.

[0284] At step S42, the good or defective judgement unit 43 calculates a ratio R_Q1corresponding to the prediction of the average value Iq of the measured values I_Qj. The ratio R_Q1is I_Q1/R₁. Next, it sets the value of the variable Z to 2, that is, Z =2.

[0285] At step S43, the good or defective judgement unit 43 calculates the allowable error E. The allowable error E is the larger of the absolute value of the estimated error rate E_R×R_Q1and the absolute value of the maximum measurement error E_M.

[0286] At step S44, the signal drive unit 31 applies a test signal of a predetermined test pattern to the CMOS integrated circuit 50, and the measurement unit 32 measures the quiescent power supply current I_Qjat a strobe point Z(Z>1). The measured value I_Qjis stored in the table memory 41 and is supplied to the good or defective judgement unit 43.

[0287] At step S45, the good or defective judgement unit 43 calculates a defect current prediction P_Q1j. The defect current prediction P_Q1jis I_Qj−R_j×R_Q1.

[0288] At step S46, the good or defective judgement unit 43 compares the defect current prediction P_Q1jwith the allowable error E.

[0289] When P_Q1jis above E, the routine proceeds to step S47, where the good or defective judgement unit 43 detects the fact that there is a defect current at the strobe point Z and judges the CMOS integrated circuit 50 under test as a defect, the output unit 44 outputs the judgement results etc., and the test is ended. Note that it is also possible to judge the existence of a defect current for all the strobe points.

[0290] When P_Q1jis not above E, the routine proceeds to step S48.

[0291] At step S48, the good or defective judgement unit 43 compares the defect current prediction P_Q1jwith −E.

[0292] When P_Q1jis below −E, the routine proceeds to step S49, where the good or defective judgement unit 43 detects that there is a defect current at a strobe point other than the strobe point Z and judges the CMOS integrated circuit 50 as a defect, the output unit 44 outputs the judgement results etc., and the test ends. Note that it is also possible to judge the existence of a defect current for all the strobe points.

[0293] When P_Q1jis not below −E, the routine proceeds to step 50.

[0294] At step 50, the control unit 21 judges whether or not measurement of the quiescent power supply current has been finished for all the predetermined strobe points (Z =1 to m).

[0295] When the predetermined measurement is not finished, the routine proceeds to step S51, where the value of Z is incremented to increase it by exactly “1”, then the test returns to step S44 where the next measurement is performed.

[0296] When the predetermined measurement is finished, the routine proceeds to step S52, where the output unit 44 displays or prints out the judgement results of the good or defective of the CMOS integrated circuit, the measurement results, various calculation results, etc. It is also possible to have the good or defective judgement unit 43 judge the CMOS integrated circuit 50 under test as good and have the output unit 44 output that judgement result when for example −E≦P_Q1j≦E.

[0297] FIG. 12 is a schematic flow chart illustrating the test method which the test device 100 in relation to the upper and lower limit method in the one point method.

[0298] The test device 100 is made one for which the reference average value ratios R_j^REFhave been calculated in advance for a predetermined number of samples of CMOS integrated circuits and one for which the proportional coefficient σ_Rshown in the above formula (10-3) has been calculated in advance. Further, the reference average value ratios R_j^REFare used as the average value ratios R_jduring the test.

[0299] First, at step S61, the signal drive unit 31 applies a test signal of a predetermined test pattern to the CMOS integrated circuit 50, and the measurement unit 32 measures the quiescent power supply current at the strobe point 1 to obtain the measured value I_Q1. This measured value I_Q1is stored in the table memory 41 and supplied to the good or defective judgement unit 43.

[0300] At step S62, the good or defective judgement unit 43 calculates a ratio R_Q1corresponding to the prediction of the average value Iq of the measured values I_Qj. The ratio R_Q1is I_Q1/R₁. Next, it sets the value of the variable Z to 2, that is, Z=2.

[0301] At step S63, the signal drive unit 31 applies a test signal of a predetermined test pattern to the CMOS integrated circuit 50, and the measurement unit 32 measures the quiescent power supply current I_Qjat a strobe point Z(Z>l) to obtain a measured value I_Qj. The measured value I_Qjis stored in the table memory 41 and is supplied to the good or defective judgement unit 43.

[0302] At step S64, the good or defective judgement unit 43 calculates an upper limit threshold I_QUand a lower limit threshold I_QL. The upper limit threshold I_QUis (1+6σ_R)×R_Q1and the lower limit threshold I_QLis (1−6σ_R)×R_Q1.

[0303] At step S65, the good or defective judgement unit 43 compares the measured value I_Qjwith the upper limit and lower limit thresholds I_QUand I_QL.

[0304] When I_Qj>I_QUor I_Qj<I_QLstands, the routine proceeds to step S66, where the good or defective judgement unit 43 detects that there is a defect current at the strobe point Z and judges the CMOS integrated circuit 50 under test as a defect, the output unit 44 outputs the judgement results etc., and the test ends. Note that it is also possible to judge the existence of a defect current for all the strobe points.

[0305] When I_Qj>I_QUand I_Qj<I_QLdo not stand, that is, when I_QL≦I_Qj≦I_QU, the routine proceed to step S67.

[0306] At step 67, the control unit 21 judges whether or not measurement of the quiescent power supply current has been finished for all the predetermined strobe points (Z =1 to m).

[0307] When the predetermined measurement is not finished, the test proceeds to step S68, where the value of the variable Z is incremented to increase it by exactly “1”, then the test returns to step S63 where the next measurement is performed.

[0308] When the predetermined measurement is finished, the routine proceeds to step S69, where the output unit 44 displays or prints out the judgement results of the good or defective of the CMOS integrated circuit, the measurement results, various calculation results, etc. It is also possible to have the good or defective judgement unit 43 judge the CMOS integrated circuit 50 under test as good and have the output unit 44 output that judgement result when for example I_QL≦I_Qj≦I_QU.

[0309] As explained above, in the method of testing a CMOS integrated circuit according to the present invention, there is no need to add any special element or structure to the CMOS integrated circuit 50 and therefore a rise in the manufacturing cost of the CMOS integrated circuit 50 due to testing can be suppressed.

[0310] Further, it is possible to use a general-purpose logic tester for the testing. In this regard as well, it is possible to suppress a rise in the manufacturing cost of the CMOS integrated circuit 50.

[0311] Moreover, it is possible to test in an actual operating mode or under conditions after shipment, to prevent occurrence of circuit failures due to testing, to prevent defects being missed, and to improve the quality (good or defective evaluation) and reliability of shipments of CMOS integrated circuits 50.

[0312] Furthermore, the good or defective and reliability of shipments of the integrated circuits can be improved by performing the method of testing according to the present invention on CMOS integrated circuit which would not be I_DDQtested otherwise in view of the cost of employing methods to reduce the leakage current.

[0313] Further, when lowering the good or defective judgement threshold of the quiescent power supply current I_Qto judge good or defective, the test method according to the present invention can be used to reduce the frequency by which good devices are judged as defects.

[0314] Moreover, by applying the test method according to the present invention to I_DDQtesting in which the effective gate length is less than about 0.25 μm, for example, about 0.25 to 0.1 μm or about 0.25 to 0.05 μm, it is possible to improve the good or defective of the CMOS integrated circuit.

[0315] Further, by applying the test method to a CMOS integrated circuit in which the quiescent power supply current I_Qis large due to a high temperature test and I_DDQtesting is not possible, for example, a CMOS integrated circuit having an effective gate length of 0.5 μm or more, it is possible to improve the good or defective.

[0316] Moreover, by using the test method along with other methods of reducing the FET leakage current, the good or defective can be even further improved.

[0317] Moreover, by applying the test method of the present invention to a standby test performing the I_DDQtesting on a single measurement point (standby current test) or a multiple point I_DDQtest performing the I_DDQtesting on several measurement points so as to perform the I_DDQtesting on even more points, the quality (good or defective) judgment can be even further improved.

[0318] Note that the present invention is not limited to the above embodiments and includes modifications within the scope of the claims.

[0319] That is, in the method of testing a CMOS integrated circuit according to the present invention, since use is made of average value ratios of quiescent power supply currents at a plurality of strobe points calculated in advance for a good CMOS integrated circuit and an average value of measured values of the quiescent power supply currents, it is possible to determine the distribution of quiescent power supply currents in a good CMOS integrated circuit with reference to the average value of the quiescent power supply currents and therefore learn the ranges which measured values can take when the CMOS integrated circuit under test is a good one from the average value ratios of the quiescent power supply currents and the average value of the measured values.

[0320] For example, the method calculates defect current estimates P_Dj=I_Qj−R_j×Iq based on the average value ratios R_jof quiescent power supply currents at the plurality of strobe points, the measured values I_Qj, and the average value Iq of the measured values.

[0321] By multiplying the average value Iq of the measured values by an average value ratio R_j, it is possible to calculate the expected value of a quiescent power supply current I_DDQin a case where the CMOS integrated circuit under test is a good one. By subtracting this expected value (R_j×Iq) from a measured value I_Qj, it is possible to obtain a defect current estimate.

[0322] Note that a defect current estimate P_Djcan be predicted from a defect current I_Djat that strobe point, the summation of the defect currents at all strobe points, the total number m of the strobe points, and the average value ratio R_j, so a value not depending on the size of a measured value of the quiescent power supply current can be obtained.

[0323] Further, the method for example calculates defect current predictions P_Q1j=I_Qj−R_j×I_Q1/R₁based on the average value ratios R_jof the quiescent power supply currents at a plurality of strobe points, the measured values I_Qj, and the ratio (I_Q1/R₁).

[0324] By multiplying the ratio (I_Q1/R₁) corresponding to the prediction of the average value of the measured values and an average value ratio R_j, it is possible to calculate an expected value of the quiescent power supply current I_DDQin a case where the CMOS integrated circuit under test is a good device. By subtracting this expected value (I_Q1×R_j/R₁) from a measured value I_Qj, it is possible to obtain a defect current prediction.

[0325] Note that, as explained above, the defect current prediction P_Q1jcan be projected from the defect currents I_Djand I_D1at strobe points and the average value ratios R_jand R₁, so a value not depending on the size of the measured value of the quiescent power supply current can therefore be obtained.

[0326] The method for example calculates an upper limit threshold I_QU=(1+f×σ_R)×I_Q1/R₁based on the average value ratio R₁of the quiescent power supply current at the predetermined strobe point, the ratio (I_Q1/R₁), and a constant f of 4 to 7.

[0327] By multiplying the ratio (I_Q1/R₁) corresponding to the prediction of average value of the measured values by (f×σ_R) relating to the standard deviation of the average value ratios R₁, it is possible to calculate the range of distribution of the quiescent power supply current I_DDQin the case where the CMOS integrated circuit under test is good. By adding this range (f×σ_R×I_Q1/R₁) and the ratio (I_Q1/R₁), it is possible to obtain the upper limit threshold.

[0328] σ_Rcan be calculated from an average value ratio R₁and the total number m of the strobe points, so a value not depending on the measured value of the quiescent power supply current can be obtained.

[0329] The method for example calculates the lower limit threshold I_QL=(1−f×σ_R)×I_Q1/R₁based on the average value ratio R₁of the quiescent power supply current at the predetermined strobe point, the ratio (I_Q1/R₁), and a constant f of 4 to 7.

[0330] By multiplying the ratio (I_Q1/R₁) corresponding to the prediction of the average value of the measured values by (f×σ_R) relating to the standard deviation of the average value ratios R₁, it is possible to calculate the range of distribution of the quiescent power supply current I_DDQthe case where the CMOS integrated circuit under test is good. By subtracting this range (f×σ_R×I_Q1/R₁) from the ratio (I_Q1/R₁), it is possible to obtain the lower limit threshold.

[0331] σ_Rcan be calculated from the average value ratio R₁and the total number m of the strobe points, so a value not depending on the measured value of the quiescent power supply current can be obtained.

[0332] The method of analyzing a CMOS integrated circuit according to the present invention calculates defect current estimates P_Dj=I_Qj−R_j×Iq based on the average value ratios R_jof quiescent power supply currents at a plurality of strobe points, measured values I_Qj, and an average value Iq.

[0333] By multiplying the average value Iq of the measured values by an average value ratio R_j, it is possible to calculate the expected value of a quiescent power supply current I_DDQin the case where the CMOS integrated circuit under test is good. By subtracting this expected value (R_j×Iq) from a measured value I_Qj, it is possible to obtain a defect current estimate.

[0334] By calculating the absolute value |R_DSUM/R_SUM| of a ratio between the summation P_DSUMof the negative value defect current estimated values P_Djand the summation R_SUMof the average value ratios R_Jfor the strobe points corresponding to the negative value defect current estimated values P_DJ, it is possible to obtain a variation ΔIq showing the difference between the average value of the measured values and the average value of FET leakage currents.

[0335] As explained above, according to the methods apparatus of testing and analyzing a CMOS integrated circuit of the present invention, it is possible to detect and analyze defect currents even with a CMOS integrated circuit with a large quiescent power supply current and large variations.