Method of Characterizing Critical Timing Paths and Analyzing Timing Related Failure Modes in Very Large Scale Integrated Circuits

This invention relates generally to test systems for integrated cicuits, and, more particularly, to a laser testing method for characterizing critical timing paths and analyzing timing related failure modes in high clock rate very large scale integrated microcircuits.

One of the most important requirements in the design and production of complex integrated microcircuits is to understand and control critical timing paths in the circuit. Timing analysis for portions of a microcircuit may be performed with circuit simulators, however, the complexity and shear size of a very large scale integrated microcircuit tends to place limitations on the use of such simulators. For example, such simulators may not include mechanisms for modeling "macro" effect on circuit performance such as substrate current distribution, power supply bussing, temperature patterns and patterns of gate delays which depend on starting materials or processing. Furthmore, there is a need for experimental measurement techniques for verifying such simulations done to identify critical internal timing paths in finished microcircuits.

In addition, a need exists in the area of acceptance testing of finished products. When performing electrical testing on large scale integrated circuits (microcircuits) or very large scale integrated circuits (microcircuits) the majority of the internal timing paths are not characterized with respect to internal timing conditions. Functional testing has done little more than validate the device truth table and perhaps test for certain data pattern sensitivities. While this has been sufficient for small scale technologies it does not provide an adequate performance screen for more complex devices. Even further, there exists a need for techniques useful for analyzing failed parts which exhibit frequency sensitivity failure modes. These may be much more difficult to analyze than catastrophic failures because the failure site may not have any distinguishing visual characteristics. Consequently, internal node electrical testing techniques are essential in locating the failure site. Non-destructive, automated techniques are therefore clearly needed in this area.

A basic problem in satisfying all of these needs is the difficulty of making reliable electrical measurements internally in a operating circuit. Low capacitance mechanical die probing is widely used but it is difficult to do quickly and non-destructively at many internal circuit nodes and would be difficult to automate.

The present invention overcomes the problems encountered in the past and as set forth in detail hereinabove by providing an integrated microcircuit testing technique based upon the injection of laser photocurrent at the drain junction of a transistor which controls the logic state of the internal node of interest in the integrated microcircuit.

The method of the present invention utilizes a source of electromagnetic radiation, preferably in the form of a helium-neon laser, which is focused onto the drain junction of a transistor of the integrated microcircuit under test. Photogenerated carriers in the silicon making up the transistor are collected by the illuminated drain junction depending upon the logic state of the transistor. This effectively loads the node and perturbs logic level transistion times during testing.

More specifically, the localized injection of electromagnetic radiation producing a photocurrent at the drain junction of the transistor at specific times during the testing procedure increases the logic level transistion times associated with that particular node. This causes an increase in the minimum operating power supply and/or a decrease in the maximum operating frequency at which the microcircuit will properly function. Consideration of these parameters and the level of photocurrent provide a measurement related to the worst case timing margin which occurs during the functional test of the integrated microcircuit. Such measurements may be repeated on different transistors and on different microcircuits to gain knowledge about the effectiveness of design, fabrication process drifts, sensitivity to failure mechanisms and sensitivity to marginal operating conditions related to the environment such as temperature or electrical noise.

The utilization of a laser beam within the present invention allows for a non-destructive, non-contacting means for perturbing the operation of the internal circuitry in a microcircuit. The laser beam accomplishes this end in a controlled manner so as to obtain a measurement related to the timing margin associated with an internal node that may be deep within the microcircuit. Furthermore, as a result of the unique system of testing provided by the present invention, information can be obtained about the microcircuit without a detailed knowledge of the circuit itself. Consequently, a user of the testing system of the present invention who needs information with respect to a microcircuit, but who did not design this particular microcircuit and may not understand its detailed operation, may still effectively use the present testing system.

Stated more succinctly, the present invention solves the problem of characterizing critical timing paths in complex devices for which detail design data is not available in a procedure that is non-contacting and non-destructive. Since the timing margins measured are related mathematically to critical physical parameters of the transistors tested, the invention provides, for the first time, a non-contacting means for identifying internal transistors with abnormal physical parameters. In addition, the system of the present invention measures true device behavior and is not limited by imperfect models that are used in timing analysis by computer simulation.

It is therefore an object of this invention to provide a method for analyzing very large scale integrated microcircuits.

It is another object of this invention to provide a laser testing method for very large scale integrated microcircuits which is completely non-destructive and non-contacting.

It is a further object of this invention to provide a laser testing method for microcircuits capable of obtaining a measurement related to the timing margin associated with an internal node that may be deep within the microcircuit.

It is still another object of this invention to provide a laser testing method for microcircuits which is capable of characterizing critical timing paths in complex microcircuits.

It is an even further object of this invention to provide a laser testing method for very large scale integrated microcircuits which utilizes conventional, currently available components during the testing procedure.

For a better understanding of the present invention, together with other and further objects thereof, reference is made to the following description, taken in conjunction with the accompanying drawings and its scope will be pointed out in the appended claims.

FIG. 1 is a schematic illustration of the laser test apparatus utilized with the method of the present invention capable of analyzing very large scale integrated microcircuits;

FIG. 2 is an electrical circuit diagram representative of a timing control unit which can be utilized with the laser test apparatus of the method of the present invention;

FIG. 3 is an electrical circuit diagram of a bit stream buffer unit which can be utilized with the laser test apparatus of the method of the present invention;

FIG. 4 is an electrical circuit diagram of a photocurrent sampling unit which can be utilized with the laser test apparatus of the method of the present invention; and

FIG. 5 is a circuit diagram of a simplified circuit utilized for analyzing the effect of injected photocurrent on high to low propagation delay within the method of the present invention.

In order to more clearly understand the method of the present invention reference is first made to the laser testing apparatus 10 utilized with the present method. More particularly, laser testing apparatus 10 incroporates therein a source of electromagnetic radiation preferably in the form of a helium-neon laser 12. Laser 12 preferably operates at 0.5 milliwatts and at a wavelength of 6328 angstroms. However, it should be realized that the above-mentioned parameters of power and wavelength for laser 12, although desirable, may vary within the confines of the present invention.

Laser 12 is optically aligned or positioned within apparatus 10 so as to direct its electromagnetic beam of radiation 14 through a modulator 16. An example of such a modulator would be a Crystal Technology 3080 acousto-optic modulator having a 1380B modulator driver or, more simply stated, an acousto-optic Bragg cell. The output 18 from modultor 16 is directed (deflected) by any conventional reflecting means in the form of mirror 20 to a set of X-Y scanning mirrors 22. More particularly, scanning mirrors 22 can be in the form of a General Scanning CCX-650T optical galvanometer scanner drive amplifierand a General Scanning G-100PBT galvenometer mirror positioning device.

Optically aligned with the output 24 of scanning mirrors 22 is a metallurgical microscope 26, preferably in the form of a Leitz Orthoplan microscope. Micro 26 is capable of focusing the beam 28 output therefrom to an extremely fine spot (about 2 microns in diameter) on the surface of the microcircuit 30 under test.

As illustrated in FIG. 1 of the drawings the microcircuit 30 is plugged into a socket forming part of two directional (X-Y) stepper stages 32. The X-Y stepper stages 32 can be in the form of Aerotech ATS302MNSW stepper stages each having an optical encoder. The socket forming part of X-Y stepper stages 32 is wired to a conventional integrated circuit tester 34. The integrated circuit tester 34 may be, for example, in the form of a Tektronix S-3260 automatic IC tester having a 1803 tester station. Circuit tester 34 is capable of running functional tests on microcircuit 30 using various power supply voltages and operating frequencies.

Also, electrically interconnected with the various components mentioned above is an electronic control circuit 36 made up of a mirror control unit 38, a timing control unit 40, a bit stream buffer unit 42, a photocurrent sampling unit 44, and a stage control unit 46; and a microcomputer 48, all of which will be described in greater detail hereinbelow. It is the purpose of the electronic control circuit 36 to convert the microcircuit supply current to a voltage, to sample this voltage in synchronism with the beam of electromagnetic radiation 28 from laser 12, to subtract, threshold detect and to convert these measurements to logic levels, to store sequences of thousands of these logic level measurement, to gate the laser 12 on the off, to interface these units to microcomputer 48 and to also interface microcomputer 48 to the integrated circuit tester 34.

More specifically, mirror control unit 38 is interconnected to scanning mirrors 22 so as to control the movement thereof while the stage control unit 46 is interconnected to the X-Y stages 32 for its control. The remaining three control units, in the form of timing control unit 40, bit stream buffer unit 42 and photocurrent sampling unit 44 are illustrated schematically in FIGS. 2, 3 and 4 of the drawings, respectively.

These particular units 40, 42 and 44 may be varied within the scope of this inventio, however, typical units are illustrated in FIGS. 2-4 in order to provide an example of operative control units capable of use within the method of the present invention.

Within timing control unit 40 shown in FIG. 2 of the drawings all resistors are 10,000 ohm variable resistors while all capacitors are 0.01 microfarads, with integrated circuits 74123 being conventional dual retriggerable resettable multivibrators and integrated circuits 74121 being conventional multivibrators. Bit stream buffer unit 42 illustrated in FIG. 3 of the drawings incorporates therein integrated circuits AM2833 and 4013 which are in the form of conventional 1024 bit static shift registers and D type flip flops, respectively. In photocurrent sampling unit 44 of FIG. 4 of the drawings the following conventional integrated circuits are used: OP-27 (operational amplifier), SMP 10 (sample and hold amplifier) and CMP-05 (comparator).

Referring now specifically to the method of the present invention, it is pointed out that timing problems arise in a circuit (microcircuit) when a logic level transition at a gate output does not occur by the time it is sampled or stored by some other part of the circuit. A logic error occurs which propagates through the circuit. The difference between the actual time in which the gate output transition is complete and the maximum time allowed by the design and operating frequency is referred to as the timing "slack" or "margin. " For a particular gate it depends upon the gate's propagation delay and settling times of its inputs, which in turn depend on sums of other gate propagation delays. Each gate has a specific timing margin associated with each transition of its output during a full test of the circuit. The minimum of these margins is referred to as the cirucit's critical timing margin an its value determines the circuit's maximum operating frequency.

The maximum operating frequency of a circuit (microcircuit) is expressed as some complex function of equations for gate propagation delays and the boolean equations describing the logic function of the entire circuit. In the past to simulate and analyze the timing of all nodes in a complex device one would need a full knowledge of the circuit, perfect gate models and a large amount of simulation time. A good design analysis might also include simulations of "worst case" distributions of process related patterns of transistor parameters across the device where they are known to occur and have a significant effect on timing.

Within the method of the present invention, laser 12 is utilized to generate very localized quantities of free hole-electron pairs in silicon. These carriers normally diffuse away from the generation volume and simply recombine. However, if they encounter an electric field, specifically in the depletion region of a reversed bias junction, they may be separated, collected and forced onto an internal circuit node as additional junction leakage current. The effect of this photocurrent depends on the circuit, and often it affects transition times. Both rise and fall times can be increased or decreased in simple CMOS gates. In NMOS it is likely that fall times can only be decreased and that rise times can only be increased but this depends on the particular circuit design.

In order to develop a qualitative analysis of the effect of injected photocurrent (as with the laser beam in the method of the present invention) on the propagation delays of a CMOS inverter, a simple circuit model can be used which takes into account the primary factors which determine switching speeds. In so doing, it is possible to study the high to low transition and then consider the effect of photocurrent generated at the p-channel transistor drain to substrate junction. Similar arguments can be made for the low to high transition with photocurrent generated at the n-channel transistor drain.

When the output is switched from high to low, the n-channel transistor turns on and the p-channel transistor turns off. It can be assumed in such a simple model that this happens immediately and that afterwards the p-channel transistor acts as an open circuit and the n-channel transistor acts as a constant current source.

Reference is now made to FIG. 5 of the drawings which represents such a simplified circuit 50 for analyzing the effect of injected photocurrent on high to low propagation delay.

The output voltage decays from V.sub.dd as the on n-channel transistor discharges the node capacitance, here mainly the drain-substrate junction capacitances of the two transistors (C.sub.dp and C.sub.dn respectively). The linear portion of the output voltage is found to be ##EQU1## where C=C.sub.dp +C.sub.dn ; I.sub.n =saturatred channel current; I.sub.ph =photocurrent and t=time. If we define the propagation delay to be the amount of time required for the output to decay to one half of V.sub.dd, then the propagation delay T.sub.phl is ##EQU2## and substituting I.sub.n =K(V.sub.gs -V.sub.thn).sup.2 and V.sub.gs =V.sub.dd (true for a CMOS inverter), ##EQU3##

It is clear that when the magnitude of the photocurrent I.sub.ph approaches that of the channel current I.sub.n, the propagation delay can be increased greatly. This can be accomplished by increasing the injected photocurrent I.sub.ph or by decreasing the channel current by lowering the power supply voltage V.sub.dd closer to the threshold voltage V.sub.thn.

Based upon the above analysis the following detailed description of the method of the present invention can be more readily understood. It should be pointed out, however, that additional substantiation of the results produced by the method of the present invention can be found in a paper by the present inventors, Daniel J. Burns et al, "Reliability/Design Assessment by Internal-Node Timing-Margin Analysis Using Laser Photocurrent-Injection," IEEE 22nd Annual Proceedings Reliability Physics, April 1984, pp. 76-82, which is incorporated herein by reference.

Reference is once again made to FIG. 1 of the drawings for a fuller understanding of the particular method of the present invention. Initially, microcircuit 30 to be tested is plugged into the socket located on the X-Y stages 32 which are utilized to position a desired area of microcircuit 30 in the field of view of microscope 26. The X-Y scanning mirrors 22 under the control of mirror control unit 38 positions the electromagnetic beam of radiation 28 to a particular spot also referred to as the gate output (the drain or gate junction for MOS technology or the collector for bipolar technology, for example) of a transistor to be tested within microcircuit 30. This spot is focused to minimum diameter by microscope 26. Normally, the electromagnetic beam of radiation 28 is attenuated with a neutral density filter (not shown) during positioning in order to avoid exposure of microcircuit 30 to the strong beam of electromagnetic radiation 28.

During the testing procedure, laser 12 is controlled by timing control unit 40 and microcircuit 30 is functionally tested by the conventional IC tester 34. During this functional test, the logic level sequence of the specific transistor within the microcircuit under test is analyzed and stored by the electronic control circuit 36 and microcomputer 48. Stated more succinctly, the logic level sequence refers to the sequence of logic levels (HI or LOW) or electrical states (ON or OFF) which characterize the behavior of an individual internal transistor during the functional test of the overall microcircuit 30. For example, in the case of a microprocessor integrated microcircuit, such a functional test may consist of many instructions, each of which in turn consists of multiple clock periods. The transistor assumes different logic levels during various of these periods, according to the design of the microcircuit and the transistors' function in the circuit.

A functional test of typical microprocessor microcircuit 30 may consist of hundreds of instructions which together consist of perhaps 20,000 clock periods. During each of these individual clock periods, the transistor under test assumes one of two electrical states, ON or OFF. Determination is made of the state of the transistor under test during each clock period by making two measurements of device power supply current during each clock period; one with laser 12 off and one with laser 12 on. The difference between the two measurements indicates whether the electrical state of the particular transistor in microcircuit 30 is ON or OFF.

An OFF transistor will result in a large difference between the two measurements and an ON transistor will result in a very small difference. The concatenation of the result of this determination for all clock periods during the device test is referred to as the logic level sequence of the transistor under test.

The logic level sequence is measured in parts, each temporarily stored in a shift register and transferred to microcomputer 48 for storage and display. An example of a computer 48 and I/O boards which can be used with the present invention is the Intel MDS-220 microcomputer with National BLC-104 I/O board and National BLC-732 analog I/O board. An example of a typical software program which can be run with the above-mentioned microcomputer 48 is presented at the end of the specification and before the claims. When all the parts of the logic level sequence are measured, the complete logic level sequence is stored in microcomputer 48.

The timing margin measurement determined by the method of the present invention is then performed by functionally testing (with IC tester 34) microcircuit 30 a second time, using the same test vector set, but with the intensity of laser 12 gated by the previously measured and stored logic level sequence. Laser 12 is turned on only when the logic level sequence indicates that that the transistor under test will be OFF during the next clock period. When laser 12 is on and the transistor turns OFF, photogenerated carriers in the semiconductor are collected by the drain junction of the transistor under test as it becomes reverse biased and turns OFF. The effect of this collection of carriers is to slow down the logic level transition which is occurring on the circuit node because it opposes the action of other circuit elements in the device which are trying to switch the node voltage by supplying or sinking current to or from the node. The magnitude of the increase in switching time is related mathematically to transistor parameters (gain, threshold) and circuit parameters (node capacitance, current driving the node) and to the timing margin of the node, that is, how much time elapses between the time when switching is complete and the time when switching must be complete. If the transition time is increased to a point greater than that allowed by the design for proper operation, a logic level error will occur that will propagate through the device and cause the device to fail the functional test. It should also be noted that the intensity of laser 12 may be varied in order to establish different switching times.

Thus, this laser induced slowing of internal node switching time can be conventiently quantified because it can be measured externally as a change in certain overall device parameters such as minimum operating power supply voltage and maximum operating frequency. The higher the induced photocurrent, the more pronounced the effect. If the same level of photocurrent is used to test different transistors, a relative measure of the timing margins associated with those transistors may be obtained by determining the minimum power supply voltage or maximum operating frequency for which the integrated circuit operates correctly (no logic errors). Similarly, if the same level of power supply voltage is used to functionally test the integrated circuit with the beam applied to different transistors, a relative measure of the timing margins of those transistors may be obtained by determining the maximum level of photocurrent or maximum operating frequency for which the integrated circuit operates correctly (no logic errors). It is this relative ranking of timing margins associated with various transistors in microcircuit 30 which is of interest because it points to specific areas in the microcircuit 30 which should be redesigned to achieve higher performance (speed, temperature range, noise immunity).

It will be useful to compare patterns of timing margins among devices of the same and different circuit types and between different manufacturers in studying the "goodness" of designs. Another eventual goal to be achieved by use of the method of the present invention is to use this information to investigate the relationship between initial timing margin patterns (at fabrication) and long term reliability. Such an analysis could lead to an improved method of identifying and rejecting potentially unreliable devices (microcircuits).

Although this invention has been described with reference to a particular method, it will be understood that this invention is also capable of further and other variations within the spirit and scope of the appended claims. For example, it is possible within the scope of the present invention to illuminate by laser 12 the gate output which can encompass either a transistor, a gate or even possibly the entire integrated circuit. Furthermore, in some instances it may be desirable to keep laser 12 on continuously during the entire functional testing sequence, that is, not gated as determined by the transistor's logic state sequence. ##SPC1##