[0001] This relates to integrated circuit capacitor measurement and testing.

BACKGROUND

[0002] The controlled and successful manufacture of integrated circuits requires evaluations, including measurement, testing, reliability and predictability of various parameters and behavior in the manufactured devices. One particular parameter example is capacitance, including the capacitance of structures that are intended to be capacitors in the circuit function itself. Variations in capacitance may be affected or caused by manufacturing variations, temperature dependence, voltage dependence, device structure and other manufacturing parameters and operating conditions, both in a given structure and over a population of manufactured integrated circuits, including variations in capacitance among capacitors within a given integrated circuit.

[0003] Capacitance variations and capacitor mismatch have been addressed in the design of modern analog-to-digital converters. Examples of calibration and correction techniques are described in U.S. Patent Nos. 7,136,006, 6,891,486 and 8,686,744, all three of which are commonly assigned herewith, and in Tan et al, "Error Correction Techniques for High-Performance Differential A/D Converters", J. Solid-State Circ, Vol. 25, No. 6 (IEEE, 1990), pp. 1318 - 27, all of which are incorporated herein by reference.

[0004] For purposes of calibration, trimming and process control, measurement of capacitor behavior is useful in manufactured devices, such as in wafer form along with functional and parametric electrical test. For such purposes, and for additional considerations such as circuit longevity, viability and operational limit determination, stress testing of circuit elements also may be useful.

[0005] FIG. 1 illustrates a conventional circuit for measuring mismatch between capacitors Ci and C ₂ , such as by evaluating the capacitance of one (or each) capacitor relative to the other. Capacitors Ci and C ₂ are connected in series between terminals Vi and V ₂ . In practice, capacitor C ₂ may be a "reference" capacitor, against which the capacitance of capacitor Ci is to be measured. A node V _INT between capacitors Ci and C ₂ is connected to the gate of a p-channel metal oxide semiconductor (MOS) transistor 14, the drain of which is at ground and the source of which is connected through a current source 12 to a bias voltage VDD- In this example, the body of transistor 14 is connected to its source.

[0006] In operation, current source 12 is biased to produce a constant current I _ls and bias voltage VDD is sufficiently positive (relative to the ground voltage at the drain of transistor 14) to place transistor 14 in saturation. Transistor 14 operates as a "source follower" under those conditions, because transistor 14 is in saturation, and the constant source-drain current Ii forces the transistor gate-to-source voltage VGS to be constant. Accordingly, ideally output voltage VOUT (or VOUT( as designated over time) at the source of transistor 14 follows changes in the voltage at its gate, which

[0007] To perform measurement of the relative capacitances of capacitors Ci and C ₂ , the voltage at node V ₂ is held constant (e.g., at ground), and the voltage at node Vi is ramped over time by increasing linearly from a starting voltage (e.g., ground) to a higher voltage. The voltage at intermediate node VINT will respond to the ramped voltage Vi by also ramping, but at a flatter slope according to the voltage divider of capacitors Ci and C ₂ , as shown in:

V _INT (t) = Vl(t) ^■ (^-) Equation 1

Accordingly, Equation 1 defines the slope the expected increasing voltage at node VINT as ( ^l ).

Moreover, and also ideally, the slope of the output voltage VOUT from the source follower of transistor 14 increases with this same slope as the ramping voltage VINT( , SO the expected slope for the rise of VOUT is as shown in:

S = ^Cl Equation 2

Ci +C ₂

[0008] As a result of the preceding, in response to the ramped voltage at node V _ls the voltage VOUT( may be measured and its slope determined, from which the capacitances of capacitors Ci and C ₂ can be determined according to:

C ₂ (l-S) .

— = Equation 3

c s ⁿ

[0009] From Equation 3, if nominally the capacitances of capacitors C _\ and C ₂ are equal, then ideally the ratio of Equation 3 will equal one. Or, if the nominal capacitances are accurate, then the ideal ratio thereof should be confirmed by Equation 3 and by evaluating the slope of VOUT( - However, in practice, the behavior of the source follower circuit of FIG. 1 is not ideal, especially in modern sub-micron transistors. In the circuit of FIG. 1 , the drain-to-source voltage of transistor 14 changes as the voltage at node V _INT (and VOU _T at the transistor source) increases. This modulation of the drain-to-source voltage causes some of the changes in the gate voltage to be consumed in charging or discharging parasitic junction capacitances in the device. Furthermore, because of the mechanism of drain-induced barrier lowering, transistor threshold voltages modulate in response to changes in drain-to-body node voltage. These effects cause the slope of output voltage VOU _T ( to not solely reflect the relative capacitances of capacitors Ci and C ₂ , but the ratio also will reflect capacitive effects and also variations in the threshold voltage of transistor 14 over the duration of the measurement. The resulting output voltage VOU _T ( will thus include non-linearities, which can be substantial. The resulting inaccuracy in capacitance measurement is incompatible with capacitors such as those intended for certain precision circuits.

[0010] U.S. Patent 8,686,744, which is hereby incorporated herein by reference and is co-owned with this application, describes at least one example that also connects a node (existing between two series-connected capacitors) to a source follower transistor configuration, for purposes of testing for a mismatch in the capacitance value of the two capacitors. More specifically, a first ramping voltage source is applied across the two capacitors, while a second ramping voltage source (which increases at one half the rate of the first voltage source) is applied to the drain of the source follower. At the same time, the output of the source follower is monitored, which will provide a first slope proportional to a first of the two capacitors. The ratio of the first slope to the second slope may be evaluated to determine whether a match exists between the capacitance values of the two capacitors.

[0011] Capacitor reliability is an additional consideration in circuit design, use and establishing of operational specifications. Various models and testing have been used in view of these considerations, where certain such models are typically based on dielectric breakdown. Testing also is sometimes attempted, but accurate measurement of small capacitance shift under electrical stress is difficult, and very limited data is available on how capacitors degrade over time.

SUMMARY

[0012] In described examples of a method of evaluating at least one parameter of a first capacitor, the method couples at least three capacitors in a capacitor network to a common node. First, the method applies a first voltage range to the capacitor network for causing a first voltage drop across the first capacitor, and it evaluates the at least one parameter in response to the first voltage range. Second, the method applies a second voltage range to the capacitor network for causing a second voltage drop across the first capacitor, the second voltage drop being greater than the first voltage drop, and it evaluates the at least one parameter in response to the second voltage range.

BRIEF DESCRIPTION OF THE DRAWINGS

[0013] FIG. 1 illustrates a conventional circuit for measuring mismatch between capacitors Ci and C ₂ .

[0014] FIG. 2 is a block diagram of a capacitor evaluation configuration.

[0015] FIG. 3 is a flowchart of an example method of operation of the evaluation configuration of FIG. 2.

[0016] FIG. 4 illustrates an example embodiment of the evaluation configuration of FIG. 2.

[0017] FIG. 5a illustrates the example embodiment of FIG. 4 where capacitors Ci and C ₂ are selected, while capacitor C ₃ is allowed to float.

[0018] FIG. 5b illustrates the example embodiment of FIG. 5a, but where the capacitor input voltages are reversed for terminals Vi and V ₂ .

[0019] FIG. 5c illustrates the example embodiment of FIG. 4 where capacitors Ci and C ₃ are selected, while capacitor C ₂ is allowed to float.

[0020] FIG. 5d illustrates the example embodiment of FIG. 4 where capacitor C ₂ is to be stress tested.

[0021] FIG. 5e illustrates the example embodiment of FIG. 4 where the post-stressed capacitor C ₂ is re-evaluated at nominal levels.

[0022] FIG. 6 illustrates a first process from FIG. 3 in more detail.

[0023] FIG. 7 is a plot of example signals from the configuration of FIG. 5a.

[0024] FIG. 8 illustrates a second process from FIG. 3 in more detail.

[0025] FIG. 9 illustrates a third process from FIG. 3 in more detail.

[0026] FIG. 10 is a plot of example signals from the configuration of FIG. 5d.

[0027] FIG. 11 illustrates the delta of the FIG. 10 slopes of V _0UTI and V ₀ U _T2 .

[0028] FIGS. 12 and 13 are plots of various voltage signals for what might be expected from resistive changes in a stressed capacitor.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

[0029] FIG. 2 is a block diagram of a capacitor evaluation configuration 20, which includes a capacitive network 22 and buffer 24 formed as part of an integrated circuit 25, where integrated circuit 25 may represent any of various different devices and may include configuration 20 for purposes of design, testing and specification confirmation. Integrated circuit 25 is further connected in various ways to automated test equipment 26, as may be achieved during circuit design or later at verification or operational testing. In an example embodiment, automated test equipment 26 may be embodied by internal circuitry within an integrated circuit (e.g., apart or including integrated circuit 25), or laboratory bench equipment similarly may interface with integrated circuit 25. Also, in an example embodiment, capacitive network 22 includes at least three capacitors C _ls C ₂ and C ₃ , each with a same nominal capacitance (although differing values could be used, given the teachings with respect to ratios and other aspects of this document). Each of these capacitors has a first terminal coupled to a mutual intermediate node VINT, and node VINT is also connected as an input to buffer 24. Buffer 24 provides an output terminal VOUT, which is coupled to and monitored by automated test equipment 26. Automated test equipment 26 also has respective connections to the second terminal of each of capacitors C _{l s} C ₂ and C ₃ , and for ease of reference each such second terminal is referred to herein as an input by receiving a respective voltage V _ls V ₂ and V ₃ from automated test equipment 26. For example, capacitor C has one terminal connected to VINT and an opposing terminal Vi receiving that voltage from automated test equipment 26. Similarly, capacitor C ₂ has an input terminal/voltage V ₂ , and capacitor C ₃ has an input terminal/voltage V ₃ . Also, terminals V _ls V ₂ , V ₃ and VOUT may be realized by test pads for coupling to equipment 26.

[0030] FIG. 3 is a flowchart of an example method 10 of operation of evaluation configuration 20 of FIG. 2. The method 10 permits automated test equipment 26 to control the electrical configuration of capacitors in capacitive network 22, to selectively bias them in alternative ways, and to test the resultant voltage and effects of the configurations and biases. More particularly, such parameters are evaluated at node VINT, which is isolated by buffer 24 to reduce effects of the testing equipment during such evaluations, so the parameters are also comparably testable via terminal VOUT, while reducing the effects of the testing on the voltage that appears at node VINT- [0031] According to method 10, a first process 200 measures the nominal capacitance (relative) of each of capacitors C _ls C ₂ and C ₃ or certain types of tunnel-FETs. As discussed below in connection with FIG. 6, in one example, process 200 is performed by pairing two of the three capacitors in capacitive network 22 at a time and measuring the capacitance of each capacitor in such a pair under a nominal voltage, and then repeating until all capacitors are at least one time measured. Next, a second process 300 applies a stress voltage, in excess of the nominal voltage, to at least one capacitor in capacitive network 22. For example, the nominal voltage of process 200 may be a part/device/circuit specified value for which a certain percentage of yield is expected (e.g., an operational specification), while the stress voltage of process 300 is some greater percentage (e.g., one hundred percent) more than the nominal voltage. Accordingly, the process 300 stress voltage is intended to "stress" the limits of design and operation of the stressed capacitor. The method 10 concludes with a third process 400, where observations and analyses occur with respect to the voltage increase across the one stressed capacitor from process 300, such as an additional nominal measurement of its capacitance. Also, other parameters associated with the stressed capacitor from process 300 may be evaluated. In any event, with the repeated nominal evaluation of process 400, an example embodiment evaluates and observes any change in parameter behavior of the stressed capacitor as compared to the comparable parameter from process 200. Any such change may evidence or suggest optional additional testing, to determine any failure of the stressed capacitor. For example, the capacitance of the stressed capacitor may increase or decrease. As another example, the stressed capacitor may develop a decreased resistance through its dielectric, such as manifested by leakage across the capacitor after it has been stressed.

[0032] FIG. 4 illustrates the evaluation configuration 20 of FIG. 2, with additional schematic details for buffer 24. In the example embodiment of FIG. 4, buffer 24 includes devices and connections from the above-incorporated U.S. Patent 8,686,744. In this example, the devices and connections include a p-channel MOS transistor 28, having its gate connected to intermediate node V _{I T} - The drain of transistor 28 is connected to a reference voltage terminal V _R , and the source of transistor 28 is connected to terminal VOU _T and connected through a current source 30 to a power supply voltage V _DD - The body node of transistor 28 is connected to its source node, or alternatively to a substrate connection if desired. Current source 30 is a conventional current source device, such as a MOS transistor biased by a reference voltage to conduct a substantially constant current; a current mirror or other circuit for providing a substantially constant current also may be used. While not shown in the above-discussed FIG. 2, FIG. 4 also provides a connection of automated test equipment 26 to buffer 24 (e.g., to reference terminal V _R ) to facilitate the buffering effect.

[0033] Transistor 28 alternatively may be realized as an n-channel MOS transistor, in which case output terminal VOU _T would be connected to the source of that n-channel device.

[0034] In operation of buffer 24, transistor 28 is biased into its saturation region and operates as a source follower device by virtue of current source 30 providing a substantially constant source-drain current Ii. This constant source-drain current Ii causes transistor 28 to have a constant gate-to- source voltage. Changes in the voltage at the gate of transistor 28 (presented at intermediate node VI T) are thus reflected directly at output terminal VOUT- The precision with which the voltage at output terminal VOUT tracks the voltage at intermediate node VINT is better than the approach of FIG. 1 and depends on the operation of current source 30 to provide constant current Accordingly, for purposes of this description, the "substantially constant" current l to be provided by current source 30 refers to a current that is sufficiently constant to meet the desired precision of the capacitance testing and evaluation described herein.

[0035] FIGS. 5a through 5e illustrate the application of various voltages by automated test equipment 26 to capacitive network 22 of FIG. 4, although the dotted rectangle and designation of 22 are removed in these latter Figures to simplify the illustration and discussion. Moreover, FIGS. 6, 8 and 9 further elaborate on processes 100, 200 and 300, respectively, of FIG. 3, and correspond to various illustrations of FIGS. 5a through 5f.

[0036] FIG. 6 illustrates process 200 from FIG. 3 in more detail, and starts with a step 202 in which two of the three capacitors in capacitive network 22 are selected by equipment 26. This capacitor selection may be achieved by establishing the connection of FIG. 5a, where the selected capacitors in that example are Ci and C ₂ . Accordingly, in this example, capacitor C ₃ is allowed to float (or, alternatively, its terminal V ₃ could be connected to the reference voltage, VR).

[0037] Next in FIG. 6, step 204 measures the capacitance of the two capacitors selected in step 202, either in absolute or relative manner, using equipment 26 and at a nominal voltage range. An example of this step is shown in FIG. 5 a, and the operation associated with FIG. 5 a is similar to the two-capacitor configuration and operation in U.S. Patent 8,686,744. More particularly, terminal V ₂ is biased to a reference voltage such as ground (shown in FIG. 5 a as 0 volts), and power supply voltage VDD is applied to current source 30. Meanwhile, measurement of the relative capacitances of capacitors Ci and C ₂ is performed by automated test equipment 26 ramping the voltage applied to bias terminal Vi at a selected time rate of change, beginning from a low voltage (such as that applied to terminal V ₂ ) and increasing to a nominal voltage (shown in FIG. 5a as 6 volts). Simultaneously, and over the same period of the ramped voltage at terminal V _ls automated test equipment 26 applies a ramping voltage to bias reference terminal VR, also beginning from a low voltage such as ground. The time rate of change of the voltage applied to terminal VR relative to that of the voltage applied to terminal Vi is selected to maintain a substantially constant drain-to- source voltage drop across transistor 28, preferably by ramping the voltage at terminal VR to equal the expected voltage increase at node V _I N _T , shown in step 204 as E {V _I N _T }. Assuming that capacitors Ci and C ₂ are expected to have an equal nominal capacitance, then they should equally divide the ramping 0 to 6 volts being applied to terminal Vi, thereby causing the voltage at node V _INT (i.e., across capacitor C ₂ ) to ramp from 0 to 3 volts; thus, the ramping voltage at terminal V _R (as applied by equipment 26) is also preferably applied to ramp from 0 to 3 volts, at the same time that the voltage applied to V ramps from 0 to 6 volts (i.e., at the same time the voltage divided to V _INT is expected, based on the equal capacitance of Ci and C ₂ , to ramp from 0 to 3 volts).

[0038] Looking at the preceding operation in additional detail, the nominal capacitor voltage divider ratio Cp (presented by capacitors Ci and C ₂ ) is determined (preferably a priori), such as from the circuit design or as based on actual measurements of dielectric properties and feature sizes for the lot or wafer of the instance of evaluation configuration 20. The voltage divider ratio Cp determines the rate the voltage at intermediate node V _INT will ramp relative to the ramping of the applied voltage at terminal Vi. Particularly, and according to circuit analysis, the voltage at intermediate node and C ₂ can be derived as shown in:

where V^t) is the time-dependent (i.e., ramping) voltage at terminal V _LS V _INT ( is the time-dependent voltage at intermediate node V _INT , and C and C ₂ are the nominal capacitances values of the respective capacitors. Accordingly, the capacitor voltage divider ratio Cp can be readily derived from the nominal capacitances Ci and C ₂ , as shown in:

Equation 5

From Equation 5, when (as in the example embodiment) the capacitors and C ₂ (or C ₃ , discussed below) have the same capacitance, then the voltage divider ratio Cp is readily calculable per Equation 5, as shown in:

Equation 5.1

Thus, Equation 5.1 confirms that, for equal value capacitors, the voltage divider ratio is anticipated to be one half. Accordingly, for a ramping voltage range applied across the series connection of those capacitors, the voltage divided to the intermediate node (V _INT ) between them should be one half of the range.

[0039] Further, the reference voltage applied to terminal V _R (by equipment 26) is selected to match, in value and time, the voltage expected to occur at V _INT - After the time rate of change (i.e., slope) of the voltage to be applied to terminal Vi is selected or otherwise identified, the time rate of change (i.e., slope) of the voltage to be applied to terminal V _R is then determined as the product of voltage divider ratio Cp and the slope of the voltage ramp at terminal Vi.

[0040] The effects of the above aspects are further discussed in connection with the example of FIG. 5a and the plot in FIG. 7, where FIG. 7 shows time across its horizontal axis and voltage across its vertical axis. As described above, the voltage divider ratio C _P corresponds to the fraction of the voltage at terminal Vi that appears at intermediate node V _INT - Accordingly, if the slope over time of the voltage at terminal Vi is S, then the slope of the voltage at intermediate terminal V _INT will correspond to the product of the voltage divider ratio Cp and the slope S. And, in the example embodiment, the example C _P =½ (see Equation 5.1). Thus, the voltage expected at V _INT will be ½(V!), and (as mentioned above) this voltage is therefore applied to terminal V _R . Therefore, the voltage at reference terminal V _R is ramped by automated test equipment 26 at the same rate of change as expected of the voltage at terminal V _INT - Accordingly, in the example of FIG. 5 a, the capacitor input voltage applied by equipment 26 to terminal Vi ramps over a period of time from 0 to 6 volts, and the voltage applied by equipment 26 to reference terminal V _R ramps from 0 to 3 volts over that same period. Both of these signals are also shown in the plot of FIG. 7.

[0041] As described above, transistor 28 operates as a source follower, given that a constant source-drain current is being supplied by current source 30. With transistor 28 in its saturation region, which occurs upon application of a sufficiently high supply voltage V _DD , the gate-to-source voltage of transistor 28 will remain constant. As the voltage at intermediate node V _INT increases over time as the voltage at terminal Vi is ramped, the output voltage at terminal VOU _T increases over time. And because the voltage at terminal V _R at the drain of transistor 28 is ramping at the same rate as the expected voltage at intermediate node V _INT , the drain-to-source voltage of transistor 28 will remain constant. Thus, the voltages at terminals Vi and V _R are simultaneously ramped by equipment 26, during which time the voltage at terminal VOU _T is measured. Thus, ramping of the applied voltage at terminal V _R simultaneously with the voltage at terminal Vi (V _R at the reduced slope relative to \ corresponding to voltage divider ratio C _P ) causes the drain-to-source voltage of transistor 28 to remain substantially constant, even as the source voltage (at terminal VOU _T ) rises with the rising voltage at intermediate node V _INT - By maintaining both the drain-to-source voltage and the gate-to-source voltage constant, the parasitic capacitances presented by transistor 28 to intermediate node V _INT remain constant over the applied voltage range, thereby accomplishing the intended isolating or buffering effect of buffer 24. Accordingly, the resulting source voltage at terminal VOUT, as shown in the plot of FIG. 7, is thus not non-linearly affected by the charging and discharging of these device parasitic capacitance. Also, shifting of the threshold voltage of transistor 28 due to drain effects is greatly reduced. Consequently, the following measurement of capacitive match (or mismatch) is more precise as compared to other techniques.

[0042] The slope of time-dependent voltage VOUT( ^m response to the ramped voltage at terminal Vi is used by equipment 26 to determine the relative capacitances of Ci and C ₂ as shown in:

(jrj = Equation 6

In operation, automated test equipment 26 (or other alternative circuitry or methodology) determines the slope S of the measured output voltage VOUT( - From that slope S, equipment 26 solves for the relative ratio of capacitances C ₂ /Ci. To the extent that this ratio deviates from that expected based on the nominal capacitances Ci and C ₂ , such deviation will correspond to the capacitive mismatch between capacitors Ci and C ₂ . For example, in this embodiment where C ₁ =C ₂ , the expected slope would be is readily calculable per Equation 6, as shown in:

(¾) = ! = ^ ; (1 - 5) = 5; 5 = i Equation 6.1

From Equation 6.1 , when C ₁ =C ₂ , the expected slope S of the output voltage VOUT( would be ½, and a capacitive mismatch is detected between capacitors Ci and C ₂ (i.e., with respect to capacitance, Ci≠C ₂ ) to the extent that the measured slope differs from ½.

[0043] FIG. 5b illustrates an additional or alternative nominal measurement, where the capacitor input voltages are reversed for terminals Vi and V ₂ . Accordingly, equipment 26 applies 0 volts to terminal Vi and the ramping voltage of 0 to 6 volts to terminal V ₂ . The aspects described above will again apply to FIG. 5b, with VINT here representing (relative to ground) the voltage across C _\ . If the configuration of FIG. 5b is conducted in addition to that of FIG. 5a, then the final value at VOUT should be the same for both configurations, thereby confirming that C ₁ =C ₂ . If the values of VOUT differ for each, then a mismatch exists between those capacitance values, as again may be determined from the slope of VOUT for either configuration.

[0044] Returning to FIG. 7, and having extensively detailed its step 204 in connection with the illustrations of FIGS. 5a and 5b, a next step 206 may cause additional iterations to be made, preferably for all different pairings of potential capacitors in capacitive network 22. For example, with FIGS. 5a and 5b having paired capacitors Ci and C ₂ , step 206 will determine that not all capacitors in the network (e.g., having three capacitors) have been paired, in which case method 200 repeats steps 202 and 204, with respect to another pair of capacitors. FIG. 5c illustrates the pairing by equipment 26 of capacitors Ci and C ₃ , with capacitor C ₂ floating. Again, one capacitor (e.g., Ci) receives a fixed terminal voltage, while during a period of time the other in the pair (e.g., C ₃ ) receives a ramped voltage, while during the same period reference VR is ramped at E {VINT} . Moreover, equipment 26 monitors VOUT( during this time and determines its slope S, from which a determination is made whether that slope represents the expected relative ratio of capacitances C ₃ /Ci. To the extent (if any) that this ratio deviates from that expected based on the nominal capacitances Ci and C ₃ , such deviation will correspond to the capacitive mismatch between capacitors Ci and C ₃ . The above steps also may be repeated with respect to capacitors C ₂ and C ₃ .

[0045] Completing FIG. 6, after evaluating each pair of capacitors in network 22 as described above, step 206 determines that no other capacitors require such evaluation, and process 200 completes. As shown in FIG. 2, after process 200 completes, a next process 300 is undertaken.

[0046] FIG. 8 illustrates additional steps in connection with the capacitor stress process 300 of FIG. 2, as further discussed in connection with FIG. 5d. In step 302, equipment 26 configures capacitors in network 22, so that a capacitor to be stress tested has a lesser amount of capacitance than the combined (or equivalent) capacitance of other capacitors that will be included in the stress testing. For example, in FIG. 5d, assume capacitor C ₂ is to be stress tested; accordingly, step 302 configures two other capacitors in network 22, which in this example includes only two other capacitors Ci and C ₃ , so that the capacitance of these two configured capacitors collectively exceeds the to-be stressed capacitor C ₂ . In an example embodiment, this relative capacitance is achieved by connecting the configured capacitors Ci and C ₃ in parallel, which is electrically achieved by connecting a same potential to their respective terminals Vi and V ₃ . Thus, in FIG. 5d, a common voltage of 0 volts is shown connected by equipment 26 to terminals Vi and V ₃ . Also, equipment 26 may include switching circuitry (not expressly shown) to connect terminals Vi and V ₃ directly to one another, while applying a common voltage to that connection, or each terminal may individually receive the mutual voltage level.

[0047] Next, in process 300, step 304 applies a ramped voltage level across capacitors, including the to-be stressed capacitor, where the voltage range is selected to nominally bias the stressed capacitor beyond its nominal value. In this example, capacitor C ₂ was nominally biased to 3 volts described above. Thus, in the example embodiment, and for purposes of stressing that capacitor, step 304 causes a bias greater than 3 volts. In one embodiment, the increased bias is achieved by the change in capacitance by configuring Ci and C ₃ as a parallel capacitance in series with C ₂ . Accordingly, C ₂ will necessarily drop additional voltage in this changed configuration. Moreover, the stressed voltage can be further increased by increasing the ramping voltage applied to V ₂ (e.g., greater than that used in step 204 of process 200). For example, in FIG. 5a, a nominal voltage range of 0 to 6 volts was applied to terminal Vi of capacitor C ₂ , resulting in a nominal bias across it of 3 volts. In contrast, in FIG. 5d an increased (i.e., stressing) voltage range of 0 to 9 volts is applied to terminal V ₂ of capacitor C ₂ , which will drop approximately 6 volts across capacitor C ₂ , to place it in a stressed condition.

[0048] Step 304, consistent in part with the preceding discussion of buffer 24, also applies a second ramped voltage to terminal V _R . However, step 304 differs from the earlier step 204 by taking into account that the configuration of FIG. 5e no longer has solely two equal capacitors in series. More specifically (e.g., referring to FIG. 5d), the resultant capacitive network is C ₂ connected in series with the parallel connection of Ci and C ₃ . Thus, a voltage divider is thereby created, and the voltage drop across a series capacitance in the divider is inversely proportional to the total capacitance value of the divider. As a result, the voltage ramp of node V _INT (i.e., the voltage across capacitor Ci and the parallel C ₃ ) is as shown in:

^VlNT ^ ⁼ ((c ₁₊ ⁽ S ₊ c ₂ ) ^Ec l ^{uation 7}

Moreover, because all three capacitance values are nominally equal (and were confirmed in process 200) in this example, the expected voltage V _INT will rise to 3.0 volts, after the voltage to terminal V ₂ reaches its maximum of 9.0 volts, as shown in:

^V = (τ^¾^) [9 - 0] = (9) = 3 Equation 7.1

With the expectation that V _INT will ramp from 0 to 3 volts during the same period that V ₂ ramps from 0 to 9 volts, equipment 26 likewise ramps the reference voltage V _R from 0 to 3 volts (i.e., to E {V _INT }). If the earlier steps determine a mismatch in any of capacitors C _ls C ₂ and C ₃ , then adjustments can be made, given the teachings herein, to adjust the ramping of V _R accordingly.

[0049] Continuing with the example of FIG. 5d, if the biasing of terminals V ₂ and V _R cause node V _INT to ramp from 0 to 3 volts, then the remainder of the voltage is across the divider. Accordingly, the voltage across the stressed capacitor C ₂ will ramp to a total of 6 volts. Thus, whereas capacitor C ₂ dropped 3.0 volts during nominal measurement process 200, stress process 300 (and step 304 thereof) doubles that drop voltage to 6.0 volts, thereby stressing the capacitor to twice its nominal bias. While the example demonstrates a stress of a 100% increase in bias, other values (greater than the nominal bias value) could be implemented.

[0050] FIG. 9 illustrates additional steps in connection with the observation and analyses of process 400 of FIG. 2, as further discussed in connection with the configuration of FIG. 5e and the plots of FIGS. 10 through 13.

[0051] In step 402, equipment 26 monitors Vour(t) as the ramped capacitor input voltage is applied (e.g., to terminal V ₂ ) and the ramped reference voltage is applied to terminal VR. Generally, and without any catastrophic failure of the post-stressed capacitor (e.g., C ₂ ), the signals will take the form shown in FIG. 10, which (like FIG. 7) shows time across its horizontal axis and voltage across its vertical axis. Therefore, FIG. 10 confirms various aspects of the discussion above. For example, during a same time period, V ₂ ramps from 0 to 9 volts, while both VR and VINT ramp from 0 to approximately 3 volts. Moreover, because both the gate-to-source and source-to-drain potentials are held relatively constant due to the ramp at VR tracking the ramp at VINT, OUT has approximately the same slope as VINT (and VR), but is higher by approximately 0.5 volts due to the threshold voltage of transistor 28. In all events, any significant departure (either in slope or amplitude) from the expected VOUT of FIG. 10 could indicate a catastrophic failure of the stressed capacitor.

[0052] In step 404, equipment 26 repeats the nominal measure discussed earlier in connection with step 204 of FIG. 6, but the step 404 is taken with respect to the now post-stressed capacitor, which is the capacitor that was stressed by process 300 (and step 304 thereof). For example, refer to the earlier discussion and FIG. 5a. However, in an example embodiment, the nominal measure is taken twice, applying the nominal capacitor input voltage (e.g., 0 to 6 volts) in a first instance to the stressed capacitor (to C ₂ , through terminal V ₂ , as shown in FIG. 5a) to produce a corresponding output indicated herein as VOUTI, and in a second instance by applying it to the non-stressed capacitor (e.g., to C _ls through terminal Y _\ ) to produce a corresponding output indicated herein as ou _T2 - Both VOUTI for the resultant voltage across the non-stressed capacitor and Voun for the resultant voltage across the post-stressed capacitor are shown by example in the plot of FIG. 10.

[0053] In step 406, equipment 26 compares the slope of VOUTI and VOUTC, as may be appreciated again by the example plot in FIG. 10. Therefore, if the post-stressed capacitor has had a change in its operational parameters due to the stress process 300, its slope should differ from that measured in the earlier nominal process 200. Moreover, assuming that the post-stressed and non-stressed capacitors originally had the same capacitance before the stress process 300, then the slopes of VOU _TI and VOU _T2 should be substantially the same if no change occurred in the post-stressed capacitor. However, in the plot of FIG. 10, the slopes of VOU _TI and VOU _T2 diverge, particularly as the input voltage (V ₂ or Vi) increases. This difference, or delta, is evaluated by equipment 26, as shown by the example plot in FIG. 1 1. A greater delta indicates a greater change in the post-stressed capacitor, as resulting from the stress process 300.

[0054] In step 408, equipment 26 determines whether the post-stressed capacitor parameters evaluated in any of steps 402 through 406 fall within expected ranges. Such ranges may be established, given various considerations including the parameter(s) tested, design specification, process variation, capacitor sizing and type. For example, equipment 26 may compare the delta of FIG. 1 1 to some threshold, so a delta below the threshold indicates no or little device failure, while a delta above the threshold indicates a failure, such as an increase or decrease in capacitance value as a result of the stress, or the formation of leakage resistance in the capacitor. Other parameters likewise may be considered. If all parameters are within expectation, method 400 may complete.

[0055] In step 408, if one or more parameters are outside of expectation (e.g., beyond a threshold), method 400 may continue to a step 410 for additional testing. In one example embodiment, an additional such test repeats the nominal measure for the post-stressed capacitor (e.g., with the terminal for the post-stressed capacitor (e.g., V ₂ ) set to 0 volts, and the terminal of the non stressed capacitor (e.g., Vi) ramping from 0 to 6 volts), to evaluate the voltage dropped across the post-stressed capacitor. However, step 410 further changes the ramping reference voltage V _R to a lesser value than used in step 404, where the reduction may be in an amount X, as shown in step 410. Thus, where step 404 applies (to terminal V _R ) a range starting from 0 volts up to E {V _INT }, step 410 applies V _R in a range starting from 0-X volts up to E {V _INT } -^. AS a result, the highest voltage applied to terminal V _R in the step 410 repeated nominal measure should be less (i.e., less) than the expected value that intermediate node V _INT will reach. For example, FIG. 5e illustrates this step, where X=l, so instead of equipment 26 ramping V _R from 0 to 3 volts as described above (in connection with FIG. 5e), it ramps terminal V _R from -1 to 2 volts. The basis for applying this lower ramping reference voltage V _R is further discussed below.

[0056] In step 412, equipment 26 determines whether VOU _T changes, or changes beyond some threshold, as between the step 404 nominal measure using V _R = E {V _INT } and the step 410 nominal measure using VR= E If such a change in VOUT does not occur, method 400 continues to step 414, which concludes that the change in VOUT slope of the post-stressed capacitor (detected in step 408), as compared to its value before stress, is primarily a change that has occurred in capacitance of the stressed capacitor. In contrast, if such a change in VOUT does occur, method 400 continues to step 416, which concludes that the change in VOUT slope of the post-stressed capacitor as compared to its value before stress, is primarily a change that has occurred in resistance of the stressed capacitor. The basis for the alternative determinations of steps 414 and 416 is further discussed below.

[0057] The conclusions in steps 414 and 416 are understood with reference to an additional discussion of FIG. 5e, given the values of Vi(t) and V _R in that Figure. Specifically, as Vi and V _R both ramp upward, if the post-stressed capacitor C ₂ is relatively unaffected by the stress process 300, then the expectation is: as Vi reaches its top 6 volts, 3 volts are dropped across C ₂ , so VINT will be at approximately 3 volts, and VOUT will therefore be at least 3 volts, if not up to one threshold voltage (e.g., in this example 0.5 volts) higher than the voltage dropped across C ₂ . However, if stress process 300 causes a sufficient resistive leak to be formed in capacitor C ₂ (e.g., tunnel resistance), then after Vi reaches its top 6 volts, the resistive leak through capacitor C ₂ provides a leakage path to ground, so VINT will drop from 3 volts, and VOUT will drop correspondingly. However, in this case, because transistor 28 is in saturation mode, then VOUT cannot fall below the top voltage of VR, namely 2 volts in this example. In any event, with the leakage resistance, the value of VOUT will (over time) drop toward the upper value of VR. Accordingly, a change in VOUT of this level, as detected by step 412 which compares VOUT of steps 402 and 410, causes method 400 to continue to step 416, concluding that the dominant change in the stressed capacitor is due to a resistance change. For example, FIGS. 12 and 13 illustrate a plot of the various voltage signals for what might be expected from resistive changes in a stressed capacitor, as indicated by the nominal test from step 404 of that circuit, where the time illustrated (or the duration of the imposed stress) is shorter in FIG. 12 than in FIG. 13. Therefore, in FIG. 12, VINT fails to rise to the expected level of 3.0 volts, which in FIG. 13 is further shown as a brief rise to approximately 2.5 volts followed by a drop-off to 2.0 volts. This failure of VINT to rise to 3.0 volts is further reflected and observable in the fact that VOUT fails to rise to the expected value of approximately 3.5 volts. Thus, the additional testing per step 410 and analyses of steps 412 and 416 may identify this drop in VINT, as evident from the low value of VQUT, as a post-stress resistive change in the capacitor. On the other hand, if step 412 determines the VOU _T of steps 402 and 410 are comparable to one another, such as within a threshold voltage of transistor 28 difference, then method 400 continues to step 414 to conclude that the dominant change in the stressed capacitor is due to a capacitance change.

[0058] Accordingly, various embodiments provide improvements to integrated circuit capacitor testing and measurement. Example embodiments may be applied to analyze various capacitor types, such as metal or junction or gate capacitors, but additional practical factors may be considered in determining the efficacy of example embodiment for certain capacitors, such as capacitors that are not well-matched or that have voltage variances (which could make the analysis either unrealistic or less useful). Example embodiments also may be used for comparing capacitors against each other.

[0059] Modifications are possible in the described embodiments, and other embodiments are possible, within the scope of the claims.