Electrical Circuit Test Apparatus and Method

This invention pertains to the testing of an electrical circuit and particularly to an apparatus and test method adapted to detect faults, particularly stuck-at faults at the input pins of combinational circuits by a simple method based on verification of Walsh coefficients.

Fault detection in electrical circuits is an important factor in quality control of the manufacturing processes by which circuits of ever decreasing size and increasing complexity are manufactured in ever increasing quantities. For a general description of fault testing, including one particular approach thereto, reference may be made, for example, to U.S. Pat. No. 4,074,851--Eichelberger et al. This patent is of further interest, with respect to the present invention, for its disclosure of means for conversion from sequential circuit to combinational circuit logic to facilitate testing. The present invention is directed to apparatus and a method for testing a combinational circuit, but would find further application in the testing of a sequential circuit including such logic conversion means. U.S. Pat. No. 4,039,813--Kregness is also of interest for its disclosure of an apparatus and method for diagnosing a digital circuit wherein a test step counter is utilized in conjunction with a register, memory, and error comparator, which translates detected errors into a displayable code for maintenance isolation.

Of more specific interest with respect to the present invention is certain literature references dealing with specific input-output analyses as fault detection techniques. In this regard, reference may be made to Bennetts and Hurst, RADEMACHER-WALSH SPECTRAL TECHNIQUES, A NEW TOOL FOR DIGITAL NETWORK FAULT DIAGNOSIS?, Digest of Papers, Seventh Annual International Conference on Fault Tolerant Computing, IEEE Catalog No. 77CH1223-7C (1977).

Also of interest are Savir, SYNDROME-TESTABLE DESIGN OF COMBINATIONAL CIRCUITS, Digest of Papers, The Ninth Annual International Symposium on Fault Tolerant Computing, IEEE Publication No. 79CH1396-1C, Tzidon et al, A PRACTICAL APPROACH TO FAULT DETECTION IN COMBINATIONAL NETWORKS, IEEE Transactions on Computers, Volume C-27, No. 10, October, 1978, pages 968-971.

The test approach as set forth in these papers is limited as to the kinds of faults that can be detected and is therefore more restrictive in application than the present invention.

Accordingly, notwithstanding these prior art teachings, there remains a need for simplified test methods and apparatus capable of quickly detecting faults, particularly stuck-at pin faults in an electrical circuit, and particularly a need for such a test method and apparatus which is sufficiently simple to facilitate incorporation of the test circuit in the circuit to be tested.

These objects are met by an apparatus and method for testing a combinational circuit having a plurality (n) of input pins by generating a plurality of electrical input signal patterns, the patterns consisting of the set of all binary numbers that can be formed from n digits. Each of these signal patterns are addressed sequentially to the input leads of the circuit to be tested and also to a parity signal generating means which emits a parity signal, either 0 or 1, indicative of whether the number of zeros in each input signal pattern is odd or even. This output of the circuit under test and the parity signal are then addressed to a reversible counter means, set at 0 to begin with, which then adds or substracts a unit in response to whether the circuit output and parity signal are both the same or different. If the parity signal is one for an even number, then the reversible counter in which the output signal and parity signal are compared adds one each time the output signal and parity signal are the same and subtracts one each time they are different. Due to the nature of Walsh coefficients describing a combinational function, in a circuit with stuck-at input faults, the final result in the reversible counter will be zero, and if not fault, it will be some other number peculiar to the function on hand.

Some combinational circuits will inherently produce a 0 when pin fault-free, but such a circuit may conveniently be modified to produce a no fault result other than zero.

If the parity signal is zero, in response to an input signal pattern having an even number of zeros, then the reversible counter subtracts one each time the parity signal and circuit output signal are the same and adds one when they are different. The same result, zero for a faulty circuit, except for certain specific circuits requiring adaptation, will be seen.

In the preferred form of the present invention, the input signal pattern generation means also includes a timing means to signal the reversible counter at the beginning of the test sequence and to zero it, and also to signal it at the conclusion of the test sequence, so that a final comparison to indicate fault or no fault, can be made at that time. In the preferred form of the present invention, the test apparatus may be incorporated as part of the circuit to be tested, in which case it is permanently attached by appropriate gating and typically mounted on the same circuit mounting means. This is particularly desirable in complex, miniaturized circuits manufactured in high quantity.

Other defects, including stuck-at lead fault defects, may be detected by similar analysis of other Walsh coefficients of the circuit. For that purpose, in the apparatus and method of the present invention, there may be incorporated a means for selecting less than all leads for parity determination and a second counter means for comparing the output signal, in response to the selected input signal patterns, to a separate parity signal generating means, generating a parity signal in response to each of the signal patterns, the result being compared to determine whether the output and the parity signal are the same or different. In this case, however, a faulty circuit may produce a number other than zero, depending on the circuit, and each circuit must be separately analyzed to determine the correct result for the subset. Still further, the same input signal generation means and parity signal generation means may be used in a separate test sequence, or plurality of sequences, to generate parity over less than all binary leads. The same reversible counter being used in each case. The nature of the fault detection technique utilizing less than all binary numbers of the input lead number is such that the final content of the reversible counter will be a predetermined number when fault-free and the criterion acceptance is whether or not that number has been accumulated.

For a better understanding of the present invention, reference may be made to the detailed description thereof which follows, taken in conjunction with the accompanying in the figures and the appended claims, wherein:

FIG. 1 is a block diagram of the preferred embodiment of the present invention;

FIG. 2 is a logic diagram of the selective circuit utilized to select a subset of the input driving signals;

FIG. 3 is also a logic diagram of the modified circuit which is used when the unit under test inherently produces a zero when the input pins are fault free.

This invention is described in detail with reference to the accompanying figures which are diagramatic circuit illustrations of the test apparatus and method of the present invention.

In the FIG. 1 it is shown that an input signal generation means (ISGM), which is also called herein a driving counter, is connected for addressing input signals, of specific patterns, to a unit under test (UUT). In the diagramatic circuit illustration, the unit under test 10 includes three pins or input positions a, b and c. Input signal leads a, b and c are also connected to a parity signal generation means (PSGM) 14 which determines whether the number of zeros in a particular simultaneously generated pattern of signals in input leads a, b and c, is odd or even, in which case the parity signal generation means 14 produces an output signal, either zero or one, depending on the convention used, to a reversible counter means (RC) 16. The reversible counter 16 is also adapted to receive the output signal from the unit under test 10 and in response thereto to produce an output, indicative of whether the signal from the unit under test 10 is the same as or different than the parity signal produced by the parity signal generation means 14.

In the circuit as illustrated in FIG. 1 and in the preferred form of the present invention, the input signal generation means also called driving counter, 12 is also adapted to produce test initiation and test completion signals to the reversible counter 16, in order to zero the reversible counter 16 at the beginning of the test and to hold the reversible counter at the final resultant number, after incrementing and decrementing in the course of the test sequence. The final output is zero if there is any stuck-at fault on any input lead or any combination thereof. Additionally, other faults may manifest themselves as a zero output. Still further a fault-free circuit will produce a particular number and still other faults will manifest themselves as deviations to a number other than zero and such faults may be separately detected in the same test sequence.

By way of example and as further explanation of the underlying theory upon which the present invention is based, following is a detailed description of a particular theoretical circuit under test.

An n-input combinational circuit under test is driven by an n-bit counter, of conventional design, that

a. generates all 2.sup.n distinct patterns, each exactly once;

b. generates a parity signal p with value 0(1) if the present pattern of counter bits or a selected portion has an odd (even) number of zeros;

c. Signals the beginning and end of the pattern generation process.

The response z of the unit under test is scored by a reversible counter that, when in the run mode, operates in accordance with the following table:

For each new number generated by the ISGM, also called the driving counter, one such arithmetic operation is performed by the reversible counter (RC). Prior to the start of the test, the driving counter causes the RC to be preset to zero and after the last number has been generated by the driving counter the contents of the RC are examined by the output or comparison circuit. In the case where p is derived from all of the bits in the driving counter 12, if the RC 16 has contents of zero, i.e., all bits of the binary output are zero, the circuit under test 10 is declared to be faulty; otherwise the circuit is declared acceptable. This represents the "C.sub.ALL test".

FIG. 2 illustrates a further extension of such test apparatus and method, where p may be derived from a proper subset of the bits control 0, 1, 2 and 3 in the driving counter, and the final contents of the RC are compared with a stored number N. If the contents of the RC differ from N, the unit under test is declared faulty. Thus, the presence of a binary 1 at control signal 3 allows the input signal 3 to pass through to the exclusive OR gate 28.

Similarly, the other signals 0, 1, and 2 will be allowed through when control signals 0, 1, and 2 are at a binary 1 level.

When p is determined by all of the bits in the driving counter, i.e., when a C.sub.ALL test is made, the present method of testing can be shown to detect all input stuck-at-constant-value faults (often called pin faults), in all possible combinations, singly or multiply. Moreover, since many of the possible faults in the interior circuit under test are equivalent to pin faults, this method will also detect these.

The C.sub.ALL test is illustrated by an example set forth in the following table, wherein the fault-free circuit is a "majority" (also called "2-out-of-338 or "voter") circuit, and the faulty unit has input c stuck-at 1.

The C.sub.ALL test can be used on any combinational circuit, except that a circuit which when fault free would end with the RC having a final value of zero must have its design somewhat modified. A particularly simple modification is shown in FIG. 3 and consists of the addition of a single new input q and an AND-circuit 42 such that it realizes the modified function F.sup.m, defined as

Here F is the n-variable function that the original circuit 40 was intended to realize and x.sub.1 * denotes either x.sub.i or x.sub.i. When q=0, F.sup.m =F, and this is the mode in which the circuit is used for its normal application. In testing, however, the terminal q is made to take on both the value 0 and the value 1. Thus the driving counter 12 has n+1 stages for testing an n-input modified logic block.

For internal stuck-at lead-fault detection, a scheme shown in the FIG. 2 is used and the detection of these faults can be made concurrently with the detection of pin faults. Typically, the same driving counter is used, a second reversible counter may be added, as the orginal one sequentially used, and its odd-even terminal is controlled by a second p signal that is derived from a proper subset of the bits in the driving counter.

In general, it can be seen that the apparatus and method of the present invention are relatively simple, test preparation requires negligible effort and for pin fault detection, one merely needs to determine in the parts design process if F, the object function, needs to be modified into F.sup.m, as specified above as shown in FIG. 3. The cost of the modification, if needed, is small and for purposes of defining the present invention, an "n input combinational circuit" is considered to be only an input combinational circuits in which C.sub.ALL is not zero when fault free, otherwise the circuit is modified to one where C.sub.ALL of F.sup.m is not zero and the n input pins thereafter consist of n+1, as compared to the unmodified circuit. Thus any combinational circuit may be tested directly or simply adapted for testing in accordance with the present invention.

Also, because of the simplicity of the apparatus and method of the present invention, the test components may easily be built into miniaturized circuits, and most particularly into integrated circuit chips at modest cost. Thus the technique offers a feasible means for built-in-test circuitry.

In the case of pin fault detection, the validity of the test procedure is shown to depend only on the function mechanized. In the case of internal leads, it also depends on the structure of the internal fan-out of the circuit. Accordingly, the method of the present invention is substantially independent of the detailed gate implementation and device technology used.

All networks with n or less input pins may be tested in accordance with the present invention with the same equipment having an n bit lay-out, and this one set of equipment may be time shared by a multiplicity of circuits located on a given chip or circuit board.

Where the present invention is used for a built-in test, it will be evident that the test components are permanently connected to (via appropriate gating) and typically mounted on a common circuit mounting means with the circuit or circuits to be tested.

It should be noted that the sequence in which the sequentially generated input signal patterns are generated is irrelevant, so that a variety of counter types may be used for generating this plurality of input signal patterns (i.e., polynominal binary, Gray code, etc.). Similarly, linear feedback shift registers with maximum cycle lengths may be used, providing provision is made to generate the all zero bit pattern. This may be advantageous in built-in test implementations where the LSSD (also called scan-in/scan-out) design approach is used, because the shift registers are already available.

The present method of testing is based on the verification of so-called Walsh coefficients. That is, the test determines whether or not the function realized by the network on hand has the same partial "Walsh spectrum" as the nominal function. A Walsh spectrum of a function of n variables consists of 2.sup.n coefficients that can be made complete, so that the set of Walsh coefficients can be made to uniquely represent the given function. Thus the Walsh spectrum is similar to the Fourier spectrum and both are examples of transformations. In the case of Walsh spectra, however, there is no physical interpretation similar to harmonics; the coefficients are merely numbers.

For purposes of comparison, reference may again be made to the Tzidon et al article referred to above. The approach there described differs from the test method and apparatus disclosed herein, in that it essentially tests a circuit output with a similar sequence of input test signals but without parity determination.

As will be apparent, the present invention is deemed to have significant utility in the rapid and efficient testing of an electrical circuit based on combinational logic, either in unmodified form or modified to a minor degree so as to adapt the circuit for test in accordance with the present invention. Moreover, this test method and apparatus is sufficiently simple to be adapted for integration as part of the circuit to be tested.