US 5,369,604 AGrant
Test Plan Generation for Analog Integrated Circuits
Issue Date:1994-11-29
•9 Claims
•2 Drawing Sheets
Abstract
A method of generating a test plan for a circuit designed with blocks of analog, digital, or mixed signal components. Each block is treated as a separate functional unit, with a test having block inputs that are set to predetermined values. A matrix of circuit equations is set up to determine what circuit inputs will result in these block inputs. The required number of equations is obtained by identifying any circuit inputs that need to be set heuristically.
Metadata
Assignee
- Texas Instruments Incorporated
Inventors
- Naiknaware Ravindranath
- G. N. Nandakumar
- Kasa S. Rao
Application Information
Application Number:US 0139033
Filing Date:1993-02-05
Priority Date:1993-02-05
Art Unit:234
Classifications
IPC:
G05B 1902
Field of Search:
364324578;580;481;48573.1;158 R
Patent Drawings (2 sheets)
Description
TECHNICAL FIELD OF THE INVENTION
This invention relates to testing of integrated circuits, and more particularly to a method for automatically generating a test plan for an integrated circuit having analog and mixed signal components.
BACKGROUND OF THE INVENTION
The number of components that can be integrated onto a semiconductor chip has been increasing rapidly. At the same time, the number of different types of integrable components has also increased. Today's integrated circuits (IC's) may have analog, digital, or mixed signal components or some combination of these types.
Testing analog and mixed signal IC's poses problems different from those encountered in testing digital integrated circuits. Analog and mixed signal IC's deal with analog signals, and the transistors as well as other components may or may not operate in a linear manner.
Analog circuit testing is generally "functional" in that it attempts to associate defects in the circuit with failure of the circuit to perform its intended function. One approach to functional testing is to conceptualize the circuit as a set of functional modules. This reduces the complexity of the test operation as compared to testing each primitive component of the circuit.
When the circuit is not integrated, block functional testing is not especially difficult. Each block is tested separately by applying input signal values and measuring outputs.
For IC's, functional testing is easily implemented in "end to end" testing of the entire circuit, because the circuit inputs and outputs are available for applying input signals and measuring outputs. However, for block level testing, where a block has one or more inputs or outputs that are not the same as the circuit inputs or outputs, test values cannot be physically applied to the block inputs and block outputs cannot be physically measured.
One solution to block functional testing is to manually propagate block inputs and outputs to circuit inputs and outputs. In the past, this has been done manually, but for complex circuit that computational burden is large. Automated techniques for backward and forward propagation have also been described. U.S. patent Ser. No. 970,973, now abandoned entitled "A Method for Generating Analog Test Plans", assigned to Texas Instruments, Incorporated, describes a method for testing analog integrated circuits, using forward and backward path propagation. Test values to be applied to block inputs and expected values at block outputs are attempted to be propagated to the circuit inputs and outputs.
Existing test plan generation methods fail to provide tests under some circumstances. For example, they do not provide tests that require current setting, ac analysis, or transient analysis. They do not provide tests for circuits with feedback. A need exists for a method for generating all types of tests for all types of analog circuits.
SUMMARY OF THE INVENTION
A first aspect of the invention is a method of using circuit simulation techniques to generate a test plan for a circuit comprised of blocks of analog, digital, or mixed signal components, or some combination of these components. The method treats each block separately, and performs an iterative process of attempting tests, methods of tests, and heuristic solutions. Accordingly, a first step is selecting a block of the circuit as the current block under test. Then, a first test to be performed on said block is selected. The test designates certain inputs that are to be set with predetermined test values and certain outputs that are to be measured and compared to expected values. The block's inputs are set by assigning the test values to them. To reduce the number of unknown variables, so that a solvable matrix can be formed, it is determined if any circuit inputs need to be identified and set heuristically. If so, these inputs are identified by using heuristic rules, and set to some value. A matrix of circuit equations is formed, which can be solved for circuit input values that will result in the test values at the block inputs. Now, the matrix can be solved to determine values to be applied to circuit inputs during testing. The matrix solution also includes expected circuit outputs. The test plan lists these circuit input values, and the circuit output(s) to be measured.
The method of the invention provides a set of circuit input values that are "consistent with" block inputs, in the sense that these circuit inputs will result in desired block inputs. Likewise, at least one circuit output is provided that is consistent with an expected block output. These consistent values to be applied and measured at the circuit pins may be different from the values specified for block pins by the test model.
A technical advantage of the invention is that it automatically generates tests for linear and non-linear analog sub-circuits within an integrated circuit. It can be used with various types of functional tests, including DC, AC, and transient response analysis. On the average, the generation of the test plan takes less time than previous automated methods. This is at least in part due to the use of functional block models as a basis for test plan generation, rather than transistor level models. It generates tests that could not be generated with previous methods, such as those that have current values to be set at block inputs.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 illustrates the basic steps of generating a test plan in accordance with the invention.
FIG. 2 illustrates the details of the test generation step of FIG. 1.
FIG. 3 is an example of an analog circuit to be tested using the method of the invention.
FIG. 4 is a graph of the paths and nodes of the circuit of FIG. 3.
COPYRIGHT
A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by any one of the patent disclosure, as it appears in the Patent and Trademark Office patent files or records, but otherwise reserves all copyright rights whatsoever.
DETAILED DESCRIPTION OF THE INVENTION
I. Test Plan Generation
FIG. 1 illustrates the basic steps of the process of generating a test program in accordance with the invention. Each of the steps of FIG. 1 will be explained in general terms in this Part I. In Part II, the same steps are described in terms of a simple example circuit.
As is evident from the following description, this method may be implemented with a test generator program for a computer, which accesses certain pre-defined data files and libraries and generates a test plan. Many of the circuit descriptions, functional models, and test models used as input for the test plan generator can be derived from existing libraries in the form of computer databases. Various parsing routines can be used to adapt data in these libraries to the syntax of the test plan generator.
In general, the method applies to an analog or partly analog modular circuit, in which sub-circuits are defined in terms of functional blocks. For implementation of the computer-based method, a circuit description, a set of functional models for its blocks, and a set of test models for these blocks are input to a test plan generator. Block by block, the functional and test models are used to determine the relationship between known inputs and expected outputs for each block. For each block, the test plan generator determines what values should be applied to the circuit inputs to produce those block inputs, as well as the value at the circuit output that will result from the block output. In other words, the circuit inputs will result in the values listed in the test model at the block inputs. In this manner, the test plan generator produces a set of consistent circuit values to be applied and measured at the circuit level when the block is under test. The calculation of the circuit values is done by means of circuit simulation, using a matrix of circuit equations. The result is a test plan, in which known circuit inputs can be applied to an IC, and the resulting outputs compared to expected results for the circuit.
In accordance with the preceding overview, it is assumed that the circuit to be tested has a modular design, in which each module can be described with a high level functional model. These-modules are referred to herein as "blocks". The functional models can be derived or are available from electronic design software packages.
As indicated in step 11, the IC under test is expressed in terms of a schematic description, interpretable by computer. Known methods of generating such a description may be used, such as schematic capture software, circuit synthesis software, or hardware description languages.
The circuit is expressed in terms of "blocks". One block is the "circuit block" which represents the entire circuit. Other blocks represent sub-circuits. Each sub-circuit block identifies its own block inputs and block outputs, so that circuit simulation can be performed with test voltages and currents set at block inputs. As explained below, the circuit simulation is in the form of a matrix of equations representing the circuit and the unknown currents and voltages in the circuit. The matrix is solved for these unknown values.
For purposes of this description, block input and output nodes, as well as those of the circuit, are referred to as "pins". It should be understood, however, that although all pins of the circuit are physically accessible, those of the blocks may be either internal or the same as those of the circuit.
In step 12, each block of the circuit is described with a functional model of its operation, which describes the behavior of the block across its inputs and outputs. The functional models are independent of the block under test, in the sense that two blocks in the circuit that are internally the same, i.e., have the same type of components, will have the same functional model with different values to be associated with the model's parameters. For example, two resistors would have the same functional model, but can have different resistance values. Or, as another example, all operational amplifiers (op amps) in a circuit might have similar input versus output characteristics, but can have different CMRR, open loop gain, etc., parameters.
The functional models may be created using techniques known in the art of electronic circuit analysis. They will generally be a set of mathematical parameters and equations describing the characteristics of the block. For example, the Shockly equation might be used to describe the operation of a diode. For use with a computer as in the invention, known modeling languages can be used. Examples of commercially available models are those provided by the SPICE circuit simulation program or written in the MAST analog behavioral description language used for the SABER simulator.
Typically, the functional models for the blocks of a particular circuit will be already stored in a library of functional models. Thus, in practice, step 12 will be implemented by accessing this library and parsing the data to form a functional model data structure for each block.
In step 13, a test model for each block of the circuit is created. A block's test model gives information about the tests that may be performed on that block. As is the case for functional models, any two or more identical blocks will have the same test model. These test models assume that the block is a stand-alone circuit, with all inputs and outputs accessible for application of test voltages and currents.
Depending on the type of block, its test model might have more than one test. Furthermore, each test might have more than one method. Within each method, certain block parameters are "set values" and other parameters are "measured values". The set values are set by assigning them specific voltage or current values. The measured values represent values that would be measured while the block is operating. The measured values are compared to an expected output value, or used to calculate some other value that is compared to an expected output value. If comparison is favorable, the block can be assumed to be operating correctly. The test type may be dc, ac, or transient. For ac tests, the test method may also specify frequency values.
As with the functional models, the test models are parsed or otherwise processed to create a test model data structure for each block.
Step 14 is creating a set of rules for heuristically determining whether any circuit inputs are to be identified and set during the generation of the test for each block in the circuit. If there are such circuit inputs, they are identified and set. This step ensures that sufficient equations are available for solving a circuit matrix. Details of this step are discussed below in connection with FIG. 2.
In step 15, the test plan generator performs an iterative, block-by-block, test generation process. determines how to relate the tests for each block to the actual input and output pins of the circuit. Paths from the non-heuristically identified circuit inputs to block inputs that are to be set with test values are identified. Also, a path from the block output that is to be measured for a test output is identified. If there are mixed signal blocks in the paths, appropriate pins are identified and set. Details of this step, in which each block has different variables for simulating the circuit with a different matrix, are explained below in connection with FIG. 2.
In step 16, the output of the test plan generator is a set of tests for each block in the circuit that has analog or mixed signals. Each test lists the current, voltage and frequency values to be set, as well as values to be measured, at circuit pins. This test plan can be converted to a test program specific to available test equipment.
Referring now to FIG. 2, the process of generating the test plan is iterative, such that each block of the circuit under test is separately analyzed. For each block, different tests, as well as different methods within each test, may be attempted.
The process of FIG. 2 may be implemented with a computer program to be run on any computer processing device. It is assumed that the computer program is ready with the data structures discussed above in connection with FIG. 1, in particular, a circuit description, a set of functional models, and a set of test models.
In step 21, a first block of the circuit is selected from the circuit description for analysis. Because the method of this invention is designed for analog testing, blocks that have completely digital inputs and outputs are not considered. However, mixed signal blocks can be tested insofar as digital inputs and outputs will be assigned high or low voltage values.
In step 22a, a first test for that block is selected. In step 22b, a method from that test is selected.
In step 23, the test model is used to provide a set of test input values for the block under test. These values are assigned to appropriate block input nodes.
The block inputs to which the test values are assigned may or may not be directly accessible by means of the inputs and outputs of the IC. In other words, if the block is "internal" one or more of its block inputs or outputs will not be the same as the circuit input or output pins. Often, the number of circuit inputs and outputs exceeds the number of block inputs and outputs. For a given set of block inputs, consistent values at circuit inputs can be calculated by means of the matrix calculation of steps 27a -27c. However, because there are usually more circuit pins than block pins, some pins must be set before the matrix can be solved. A feature of the invention is that these pins are heuristically identified to provide the best circuit simulation.
Also, as part of step 23, circuit power pins are identified and set. Values for these pins can be provided by the circuit database.
In step 24, circuit paths are "sensitized" for simulation purposes. For this, each measured value from the test model is assigned to an appropriate output node and a path to a circuit output selected. A list of all circuit outputs connected to the block output is generated, i.e., an "output pin list".
A feature of the invention is that paths containing mixed signal blocks may also be sensitized. It is necessary to find any mixed signal blocks in the path to a circuit pin and to set certain pins of the mixed signal block so that a circuit simulation can be performed. More specifically, to select a path, step 24 operates on each pin of the block under test. If the pin is a circuit input, no further action is taken. If the pin is a circuit output, it is included in the output pin list.
If the block pin is not a circuit pin, all blocks connected to it are identified. Then, for each of these blocks, it is determined whether any pin of that block is a "control" type pin, and if so, the block is identified as a mixed signal block. If this pin is an output pin, any one input pin and this output pin will be activated. If this pin is an input pin, any one output pin and this input pin are sensitized. If any sensitized pin is a circuit pin, step 24 stops here. If no pin of the block is a control type pin, the block is not a mixed signal block, and if its connected pin is a circuit output, that output is included in the output pin list.
For each input pin of the block under test, the above routine is performed once for each block output. For each output pin of the block under test, the function is called once for each block input. Appendix A is an example of an implementation of step 24.
As a result of step 24, all circuit outputs connected to the block under test through purely analog blocks are to be measured. In this case, the output pin list will contain all circuit outputs. For any mixed signal block in the circuit, only a particular path through that block is activated and is of importance for measuring an output value.
Step 25 applies if there are mixed signal blocks in any path to or from the block under test to the circuit inputs or outputs, as determined in step 24. If so, high or low values are set on digital pins, such that the path will be active during simulation. Appendix B describes a sensitivity test that may be used as part of step 25.
Step 26 is heuristically determining which, if any, circuit pins are to be set prior to the matrix solution. To determine if there are such circuit pins, the number of block inputs that will have set values, m, is subtracted from the number of circuit inputs, n. The difference is the number of circuit pins that are to be heuristically set. The number of set values, m, will be the total of block inputs set during step 24, whether by the test model or during power pin setting.
Once it is determined that one or more circuit pins are to be heuristically set, various approaches may be used for identifying them. One approach is a "distance" approach that involves finding distances between each circuit pin and the block under test. The farthest away circuit pin is to be set. The following are exemplary rules for these distance heuristics:
1. The distance across any two pins of a block not under test is equal to the number of pins of that block.
2. The distance across any two pins of the block under test is 1.
3. The primary inputs of the block under test are excluded.
4. The inputs of the blocks following the block under test are preferred to be excluded.
A more detailed discussed of an algorithm for these heuristics is set out in Part III below.
The heuristically identified circuit pin(s) are then assigned an input value. Where, as is the case in this description, KCL equations are used for the matrix, this value will represent an input voltage. The value to which a heuristically determined input is set can be any value within the range of the input pin. This range may be specified, such as in the test block model or elsewhere. For example, if the range is -4.5 volts to 4.5 volts, the voltage may be set anywhere within that range, including 0.
As a result of step 26, a number of heuristic circuit pins have set values. The remaining circuit pins are those that will receive values set by a matrix calculation. This permits formation of a sufficient number of circuit equations to feasibly be solved, i.e., the number of unknowns does not exceed the number of equations.
Step 26 may be used to heuristically identify and set either circuit inputs or outputs. For example, if an output is farthest from the block it may be assigned a value for purposes of matrix solution. However, for purposes of test plan generation, if this is to be a measured output, it will be designated as a value to be measured rather than set.
In step 27a, a matrix of equations for solving the circuit is formed. In general, the matrix resembles a conventional circuit simulation matrix. However, the variables for which it is solved are different from those of conventional simulation. Some known variables are set heuristically and others are represented in the matrix by block inputs. By solving the matrix, consistent circuit input values that will result in the desired test input values at the block under test are calculated.
For forming the matrix, KCL equations are extracted from a circuit graph of paths and nodes. These equations are stored in the form of a matrix. Each row of the matrix corresponds to a node and each column corresponds to a path, as described by the circuit graph. Matrix entries can be -1, 0, or 1. If the path is not connected to the node, the entry is 0. If current on the path is entering the node, the entry is 1. If current on the path is leaving the node, the entry is -1. A KCL equation is available for each node that is not a circuit input or ground pin. Additional equations are supplied by the functional models for the blocks.
Mathematically speaking, a system of n equations, having n unknowns, is formed. As described in step 26, when there is a fewer number of available equations than circuit unknowns, some unknowns are set heuristically so that the number of unknowns is the same as the number of available equations. Thus, the number of equations in the matrix is the sum of the number of KCL equations plus a number of equations derived from blocks of the circuit.
In step 27b, the matrix is attempted to be solved. The techniques used to solve the matrix include traditional numerical methods. Appendix C describes a solution routine in detail. In general, a column matrix, G(x(k)), is set up, and its derivative square matrix, J, is calculated. During each iteration of a Newton-Raphson procedure, the G matrix and J matrix are updated by solving the functional model of the block under test.
If the matrix is not able to be solved, step 26 is repeated with different heuristic rules or different heuristically derived pins or with different values on the same heuristically derived pins. Step 27a is repeated to form a new matrix, and step 27b is repeated to attempt to solve it. If the program is unable to create any test for the block, it outputs a message to the user so that design engineers can redesign the circuit to make it testable.
Step 28 is performed when step 27b results in a solution to the matrix. The solution of the matrix provides circuit input values that are consistent with the block input values provided by the block test model. It also provides at least one circuit output value consistent with the expected block output value provided by the test model. In other words, because the block voltages, circuit voltages, and all currents in the circuit block, form a part of the circuit matrix, all these values become determined. The test method, with the measured circuit output(s) and with the circuit inputs heuristically set in step 26 or calculated in step 27b, is added to the test plan.
In step 29a, it is determined whether another test method remains to be tried for the block. If so, steps 23-28 can be repeated for that method. Step 29a is optional, in the sense that if one test method is successful, it may be expedient to immediately go to step 29b to determine if there are additional tests for the block, or to step 29c to determine if there are additional blocks to be tested.
The above-described process is especially useful for analog and mixed signal integrated circuits. If the circuit has digital blocks, they can be masked during test generation. Also, the method could be used for functional testing of digital blocks, where instead of logic decisions, high and low voltages are set and measured.
II. Example Circuit for Test Plan Generation
FIG. 3 is a schematic diagram of a simple circuit, for which a test plan may be generated in accordance with the invention. It is referred to herein as a "cascaded op amp" circuit, comprised of two operational amplifiers with feedback and input resistors.
In accordance with step 11 of FIG. 1, the circuit is described in terms of a circuit block, CASCADE, and two internal blocks, RESISTOR and OP AMP. The internal blocks correspond to functional models that are available for op amps and resistors. Using these blocks, the circuit description could be:
FIG. 4 is an equivalent directional graph of the circuit of FIG. 3, which better illustrates its paths and nodes. The circuit has 6 paths and 8 nodes. Each node is labeled in FIG. 4 with both a variable name and a numerical identifier.
In accordance with step 12 of FIG. 1, each different type of block is associated with a functional model. Thus, the resistor blocks will have the same functional model, as will the op amp blocks.
This example makes use of models available from the commercially available SABER-MAST set of electrical models. For the op amp blocks, where the inputs are at nodes A and B and the output at node C, the model is expressed as:
In accordance with step 13 of FIG. 1, test models for each type of block are derived. For the op amp blocks, there are two tests, one for output voltage, Voo, and one for gain:
At this point, the inputs for the test plan generator, as illustrated in FIG. 1, are complete. Referring now to FIG. 2, each block of the cascaded op amp circuit is separately processed to obtain a test plan for that block.
The following description assumes that the block currently being processed is the RES1 block of FIG. 3. Thus, in accordance with step 21 of FIG. 2, the selected block is RES1.
In step 22a, a first test for the resistor is selected for consideration. This is the R test of the resistor test model described above. It has only one method, and step is selecting that method.
In step 23, the values to the block nodes are set in accordance with the test model. Thus, IN1, which corresponds to node A of RES1 is set to 2 volts. IN.sub.-- MINUS, which corresponds to node B of RES1, is set to 0 volts.
In step 24, a path is traced from the block output to a circuit output. This identifies OUT2 as a node at which a circuit measurement will be made during testing. Also, GND is identified as a power pin and set to 0.
Step 25 is forming the matrix for the circuit. There are four nodes other than circuit input nodes. Referring to FIG. 4, these nodes are OUT2, IN.sub.-- MINUS, OUT, and IN.sub.-- MINUS2. The KCL equations for these nodes, where imn represents the current, i, between nodes m and n, are:
The KCL matrix for these equations is set up such that each row represents a KCL at a different node. Each column represents a path. Thus, where node 4 (GND) is set at 0 volts:
An additional 8 equations are available to represent circuit blocks. Referring to the functional models set out above, each resistor contributes one equation, and each op amp contributes two. Thus:
Thus, the total number of available equations is 12. However, the number of unknown variables is 13: i05, i56, i64, i67, i73, i34, v1, v2, v6, v7, v3, i1 (op amp 1), i1 (op amp 2).
Step 26 begins with determining how many values are to be set heuristically. For this determination, the number of nodes with set values is subtracted from the number of circuit inputs. In this example, there are 4 circuit inputs and 3 set values. (As indicated above, the node GND is considered a circuit input and its value is set by default to 0). Thus, one value needs to be set heuristically.
Step 26 also involves identifying each node to be set. Using distance heuristics, the distances between the input to RES1 and other circuit inputs are calculated. Node 2 (IN3) has a distance of 9 units from RES1. Node 1 (IN2) has a distance of 3 units. Thus, node 2 is the farthest node from the block under test. It is set to an arbitrary value, such as 0. The remaining circuit inputs (IN2) is matched to a block inputs. Thus, the value to be solved for is IN2, which will provide an input signal to be applied during testing so that 0 volts will appear at IN.sub.-- MINUS. There are now 12 unknown variables and 12 matrix equations. The matrix may now be solved.
The result of steps 22-28 is a test plan for the block. Appendix D is a test plan generated for the example circuit. As indicated, for each test, the plan lists provides input values for the IN1, IN2, IN3, and GND. It also provides an expected value for the circuit output, OUT2. During testing, each test is performed and OUT2 measured. The measured value is then compared to the expected output value.
III. Distance Heuristics
One method for identifying circuit pins that should be set heuristically is to form a D matrix. This matrix has dimension x by x, where x is the number of nodes in the circuit.
The elements of the D matrix are filled with infinity values, i.e., very large numbers. Then, the D matrix is weighted such that
, where
d (i,j )=complexity of the block through which nodes i and j are connected
d(i,i)=0.
In other words, if the complexity of the block at (i,j) is less that infinity, that number is used instead of infinity.
Then, for each pair of pins of the block under test, zeros are entered for the corresponding element in D. As a result, all nodes around the block are "fused" together as a single block node. Then, the shortest path from this "fused" node to each node that is a circuit pin is calculated. An algorithm known as Dijkstra's algorithm may be used for this calculation. The result is a list of distances from the block to each circuit pin, from which the longest distance may be selected.
Other Embodiments
Although the invention has been described with reference to specific embodiments, this description is not meant to be construed in a limiting sense. Various modifications of the disclosed embodiments, as well as alternative embodiments, will be apparent to persons skilled in the art. It is, therefore, contemplated that the appended claims will cover all modifications that fall within the true scope of the invention. ##SPC1##
Claims
What is claimed is:
1. A method of using a computer to automatically generate a test plan for an integrated circuit comprised of blocks of analog or mixed signal components, wherein the number of unknown circuit voltages and currents is greater than the number of available circuit equations, comprising the steps of: loading a block functional description of the circuit into a computer; loading predefined block test models for each block of the integrated circuit into the computer; setting each functional blocks' inputs to values specified by the predefined block test model for them; automatically determining by the computer whether a sufficient number of block inputs are set by said setting step, such that the number of unknown circuit voltages and currents is the same as the number of available circuit equations; if said number of inputs is not sufficient, automatically identifying by the computer circuit pins to be set with a known voltage value; automatically setting by the computer each circuit pin identified during said identifying step with a voltage value; automatically forming a matrix of circuit equations by the computer whose unknown variables include voltages at those circuit pins not identified in said identifying step; automatically solving said matrix for said unknown variables by the computer; and listing values provided by said solving step as circuit input values to be applied during a test of said block.
2. The method of claim 1, wherein said identifying step is performed by locating a circuit input that is farthest from the block under test.
3. The method of claim 1, further comprising the step of identifying circuit outputs to be measured during a test of said block.
4. The method of claim 1, wherein said matrix includes unknown circuit outputs and further comprising the step of listing values provided by said solving step as circuit output values to be expected during a test of said block.
5. The method of claim 1, further comprising the step of setting values to power pins of said circuit before said solving step.
6. The method of claim 5, wherein said identifying step is performed if the number of inputs with values set in said setting steps is less than the number of circuit inputs.
7. The method of claim 1, further comprising the step of identifying mixed signal blocks in said circuit and of setting values of at least one input and at least one output of said mixed signal block.
8. The method of claim 1, further comprising the step of identifying and masking any digital blocks in said circuit.
9. The method of claim 1, wherein said step of forming a matrix is accomplished by generating KCL equations for said circuit and at least one functional equation for each block of said circuit.