Shmoo test method and device and storage medium

Abstract

The invention belongs to the field of integrated circuit testing, and discloses a Shmoo testing method and device and a storage medium. The method comprises the steps of determining to-be-optimized Shmoo test parameters of an integrated circuit and defining a search range; an optimization objective function is established, an initial test parameter set is generated, and initial test or simulation is executed to obtain initial fitness data; executing an iterative loop of the grey wolf optimization algorithm, calculating a fitness value of a current test parameter set in each iteration, identifying an optimal test parameter, a suboptimal test parameter and a third optimal test parameter, and updating an algorithm control parameter; updating other test parameters according to the position information of the three test parameters, and carrying out boundary processing on the updated test parameters to keep the test parameters in a predefined search range; and when the termination condition is met, outputting the optimal test parameter as an optimal Shmoo test parameter combination, thereby determining the performance boundary of the integrated circuit or generating a Shmoo graph, and realizing more comprehensive and accurate Shmoo test.

Description

Technical Field

[0001] This invention relates to the field of integrated circuit testing, and in particular to a Shmoo testing method, apparatus, and storage medium. Background Technology

[0002] Shmoo testing is an important method for integrated circuit reliability verification and parameter optimization. The term "Shmoo diagram" originated in the early 1970s, named for its resemblance to a pumpkin-shaped cartoon character called Shmoo. Shmoo testing systematically adjusts test conditions such as voltage, frequency, and temperature in a multi-dimensional parameter space to obtain the chip's performance under different operating conditions, generating a graphical representation of the pass / fail operation regions for intuitive display, thereby determining the optimal operating window and extreme operating conditions. This testing method can characterize the fundamental electrical characteristics of transistors within an integrated circuit within their performance envelope, helping to determine robustness and stability during manufacturing, enabling production with minimal yield loss.

[0003] Traditional Shmoo testing employs a grid traversal method, which tests point by point within a predefined parameter range at fixed steps. While this method can fully cover the parameter space, it suffers from excessively long testing times and high costs. With the increasing complexity of integrated circuits and the growing dimensionality of test parameters, the parameter space grows exponentially, making full-scale traversal testing insufficient for practical engineering needs.

[0004] To improve testing efficiency, existing technologies attempt to introduce various optimization algorithms to reduce the number of test points. Some methods employ random sampling strategies, selecting test points through probabilistic methods; however, these methods are prone to missing critical regions, leading to unreliable test results. Other methods use traditional heuristic algorithms such as genetic algorithms or particle swarm optimization for parameter optimization. However, these algorithms generally suffer from slow convergence speed and a tendency to get trapped in local optima when dealing with high-dimensional parameter spaces. Especially in complex multi-peak Shmoo testing scenarios, it is difficult to balance the algorithm's global search capability and local optimization capability, often failing to find the truly optimal parameter combination within a limited number of tests.

[0005] Secondly, during the algorithm execution process, the existing methods mostly adopt linear decreasing or fixed step size control parameter update methods. This simple control strategy cannot dynamically adjust the exploration intensity according to the search state. It is easy to converge too early in the early stage of the search, and when it is close to the optimal test parameters, it may exhibit oscillation and non-convergence.

[0006] A more prominent problem is that existing optimization algorithms lack effective local optimum escape mechanisms and adaptive position update strategies. In actual Shmoo tests, the parameter space often contains multiple local optimum regions. When the algorithm converges to a local optimum, the lack of a dynamic perturbation mechanism to allow it to escape the local optimum and re-enter the global search ultimately prevents the test results from finding the truly optimal parameter combination. Summary of the Invention

[0007] The purpose of this invention is to provide a Shmoo testing method, apparatus, and storage medium to solve the problems of long testing time, high testing cost, multi-parameter interdependence, insufficient convergence speed, and insufficient testing accuracy in existing Shmoo testing methods.

[0008] To address the aforementioned technical problems, in a first aspect, the present invention provides a Shmoo testing method, comprising: Determine the Shmoo test parameters to be optimized for the integrated circuit, and define the search range for each Shmoo test parameter to be optimized; Establish an optimization objective function, generate an initial test parameter set, and perform preliminary tests or simulations to obtain initial fitness data; The iterative loop of the Grey Wolf optimization algorithm is executed. In each iteration, the fitness value of the current test parameter set is calculated, the optimal test parameters, the second-best test parameters, and the third-best test parameters are identified, and the algorithm control parameters are updated. Update other test parameters based on the position information of the optimal test parameter, the second-best test parameter, and the third-best test parameter; Boundary processing is performed on the updated test parameters to keep all test parameters within the predefined search range; When the termination condition is met, the optimal test parameters are output as the optimal Shmoo test parameter combination, and the performance boundary of the integrated circuit is determined or a Shmoo diagram is generated based on the optimal test parameters.

[0009] Furthermore, the initial test parameter set is generated using a chaotic tent mapping method. A chaotic sequence is generated by iterating through the tent mapping function, and the chaotic sequence is mapped to the search range of each Shmoo test parameter to generate the initial test parameter set.

[0010] Furthermore, the process of performing preliminary tests or simulations to obtain initial fitness data includes: The parameters in the initial test parameter set are sent to an automated test device to perform a Shmoo test; or a pre-established integrated circuit behavior model is used to simulate and predict the parameters in the initial test parameter set. Record the test results corresponding to each set of test parameters. The test results include pass or fail status, performance indicators and bit error rate. Calculate the corresponding initial fitness value based on the test results of each set of test parameters.

[0011] Furthermore, calculating the fitness value of the current test parameter set in each iteration includes: The fitness value of each set of test parameters is calculated based on the test results of each set of test parameters; The current set of test parameters is sorted according to the fitness value of each set of test parameters. The test parameter with the highest fitness value is identified as the optimal test parameter, the test parameter with the second highest fitness value is identified as the suboptimal test parameter, and the test parameter with the third highest fitness value is identified as the third optimal test parameter.

[0012] Furthermore, the update algorithm control parameters employ a nonlinear convergence factor, which is a nonlinear function based on a Gaussian distribution curve. This allows the algorithm control parameters to maintain a large value in the early stages of the search in the Shmoo test parameter space for global exploration, and then gradually decrease as they approach the optimal test parameters to accelerate local convergence.

[0013] Furthermore, updating other test parameters based on the position information of the optimal test parameter, the second-best test parameter, and the third-best test parameter includes: Using the position update mathematical model of the gray wolf optimization algorithm, three candidate position vectors are calculated for the current test parameter to move towards the optimal test parameter, the second-best test parameter, and the third-best test parameter, respectively. The three candidate position vectors are weighted and summed to obtain the final position update vector; Update the position of the current test parameter according to the position update vector.

[0014] Furthermore, the method also includes an adaptive neighborhood mutation mechanism: Monitor the convergence state of the test parameter set and calculate the rate of change of the fitness value of the optimal test parameter within a continuous preset number of algebras; When the rate of change of the fitness value is less than a preset threshold, it is determined that the test parameter set may be trapped in a local optimum. A random perturbation based on the current position is applied to some test parameters. The amplitude of the random perturbation is adaptively adjusted with the number of iterations. The perturbation amplitude is larger in the early stage of the iteration and gradually decreases in the later stage of the iteration.

[0015] Furthermore, the boundary processing includes: Check if the updated test parameters are within the defined search range; For test parameter values ​​that are outside the search range, boundary correction is performed. A boundary truncation strategy is used to limit the parameter values ​​outside the range to the boundary values ​​of the search range, or a bounce strategy is used to remap them into the effective range.

[0016] Furthermore, the termination condition includes at least one of the following conditions: The number of iterations has reached the preset maximum number of iterations; The fitness value of the optimal test parameter changes less than a preset threshold within a consecutive preset number of algebras; The fitness value of the optimal test parameters reaches or exceeds the preset optimization target threshold.

[0017] Secondly, the present invention provides a Shmoo testing apparatus, comprising: The initialization configuration module is used to define the Shmoo test parameters to be optimized, the search range, and the optimization objective function, and uses the chaotic tent mapping method to generate the initial test parameter set and obtain the initial fitness data. The optimization execution module is used to execute the iterative loop of the Grey Wolf optimization algorithm. It includes a fitness calculation submodule, a parameter identification submodule, a control parameter update submodule, a position update submodule, a boundary processing submodule, and a convergence judgment submodule. The control parameter update submodule adopts a nonlinear convergence factor, and the position update submodule includes a dynamic proportional weighting unit and an adaptive neighborhood mutation unit. The test interface module connects to automated test equipment or simulation systems via a data bus and is used to send test parameters and receive test results. The output module is used to output the optimal combination of test parameters when the termination condition is met, and to generate a Shmoo plot and analysis report.

[0018] Thirdly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the Shmoo testing method described in the first aspect of this application. Compared with the prior art, the present invention has at least the following beneficial effects: This invention transforms the multidimensional parameter optimization problem of Shmoo testing into a problem solved using the Grey Wolf Optimization Algorithm. By executing the iterative loop of the Grey Wolf Optimization Algorithm, the fitness value is calculated and the optimal, suboptimal, and tertiary optimal test parameters are identified in each iteration. The location information of these three levels of high-quality test parameters is used to guide the update of other test parameters. This allows the algorithm to effectively capture and optimize the complex interdependencies and linkage effects between multiple parameters, avoiding suboptimal test parameters or misjudgments caused by neglecting parameter linkage in traditional methods. Furthermore, boundary handling ensures that all test parameters remain within the defined search range, and flexible termination conditions achieve a good balance between convergence speed and test parameter accuracy. This enables faster and more accurate identification of the pass / fail boundaries and optimal operating points of integrated circuits, thereby outputting the optimal combination of test parameters and generating a test spectrum, achieving a more comprehensive and accurate Shmoo test. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the Shmoo testing method in one embodiment of the present invention. Detailed Implementation

[0020] Based on the teachings of this specification, those skilled in the art can form new technical solutions by combining different implementation methods without creating technical contradictions. Such variations should be considered to fall within the protection scope of this patent.

[0021] The following is a more detailed description of a Shmoo testing method, apparatus, and storage medium according to the present invention, with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. It should be understood that those skilled in the art can modify the invention described herein while still achieving its advantageous effects. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the invention.

[0022] The invention is described more specifically by way of example in the following paragraphs with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0023] Example 1 like Figure 1 As shown in the figure, this invention proposes a Shmoo testing method. This method introduces the Grey Wolf Optimizer (GWO) algorithm into the Shmoo testing of integrated circuits. It replaces the traditional exhaustive grid search with an intelligent search strategy, reducing the number of test points and time required for Shmoo testing in a high-dimensional parameter space. The Grey Wolf Optimizer (GWO) is a swarm intelligence optimization algorithm that simulates the social hierarchy and hunting behavior of grey wolves. It guides the entire wolf pack in a collaborative search through three levels of leader wolves: α, β, and δ. It has advantages such as simple structure, few parameters, fast convergence speed, and strong global search capability.

[0024] In this embodiment, the Shmoo testing method includes: Step S1: Determine the Shmoo test parameters to be optimized for the integrated circuit, and define the search range for each Shmoo test parameter to be optimized.

[0025] In Shmoo testing, it's necessary to calculate the operating state of an integrated circuit under multiple parameter combinations. Common test parameters include, but are not limited to, power supply voltage, clock frequency, temperature, input signal swing, and bias current. In this step, based on the characteristics of the integrated circuit under test and the testing requirements, the test parameters that need optimization are determined. For example, for a digital integrated circuit, it might be necessary to optimize the power supply voltage and clock frequency; however, for an analog integrated circuit, it might be necessary to optimize multiple parameters such as power supply voltage, temperature, and bias current.

[0026] For each Shmoo test parameter to be optimized, a search range needs to be defined, i.e., the upper and lower limits of the parameter. The determination of the search range is based on the integrated circuit design specifications, process tolerances, and the requirements of the actual application scenario.

[0027] Step S2: Establish the optimization objective function, generate the initial test parameter set, and perform preliminary tests or simulations to obtain initial fitness data.

[0028] In this embodiment, optimizing the objective function is key to transforming the Shmoo test problem into an optimization problem. Different forms of objective functions can be established depending on the test objective. Fitness data is typically associated with the optimization objective function, reflecting the quality of the test parameter combination; a higher fitness value indicates that the parameter combination is closer to the optimal test parameters.

[0029] In this embodiment, the design of the optimization objective function depends on the specific testing requirements. For example, if the testing objective is to find the highest operating frequency of an integrated circuit chip at the lowest voltage, then the optimization objective function (also known as the fitness function) Fitness can be defined as: Fitness = ω1 × f clk +ω2×1 / V dd , where f clk V is the frequency at which the current test point passes. dd ω1 and ω2 are the power supply voltage and weighting coefficients, respectively.

[0030] If the test result is unsuccessful, the fitness value needs to be penalized by setting it to a minimum value (e.g., 0). By maximizing the fitness function, the Grey Wolf optimization algorithm can guide the test parameter configuration to converge towards the feasible region of low voltage and high frequency, and gradually identify the optimal operating point that can reduce power consumption while ensuring high performance during the iterative optimization process.

[0031] Step S3: Execute the iterative loop of the Grey Wolf optimization algorithm. In each iteration, calculate the fitness value of the current test parameter set, identify the optimal test parameters, the second-best test parameters, and the third-best test parameters, and update the algorithm control parameters.

[0032] The Gray Wolf Optimization Algorithm optimizes the search by simulating the social hierarchy and hunting behavior of gray wolves. Within a wolf pack, wolves are divided into four social tiers: α wolves (leaders), β wolves (supporters), δ wolves (subordinates), and ω wolves (the remaining members). In the optimization algorithm, the test parameter combination with the highest fitness value corresponds to the α wolf, the second highest fitness value to the β wolf, the third highest fitness value to the δ wolf, and the remaining test parameter combinations to the ω wolf. These three test parameter combinations represent the optimal search area found in the current iteration and will guide other test parameter combinations towards these areas.

[0033] Step S4: Update other test parameters based on the position information of the optimal test parameter, the second-best test parameter, and the third-best test parameter.

[0034] In the gray wolf optimization algorithm, ω wolf (and the remaining test parameter combinations) is updated based on the positions of α wolf, β wolf, and δ wolf. This process simulates the behavior of lower-ranking wolves following higher-ranking wolves during a hunt. This update method ensures that the movement of each test parameter is not only influenced by a single high-quality parameter, but also comprehensively considers the guidance of three different levels of high-quality parameters. This avoids premature convergence caused by over-reliance on a single parameter, and improves the algorithm's robustness and global search capability.

[0035] Step S5: Perform boundary processing on the updated test parameters to keep all test parameters within the predefined search range.

[0036] During parameter updates, due to randomness and the magnitude of shifts, some test parameters may exceed the predefined search range. Parameter values ​​outside this range are invalid in practical applications and may correspond to unsafe operating conditions or configurations exceeding hardware capabilities. Therefore, boundary handling is necessary to correct out-of-range parameter values ​​back to the valid range.

[0037] Step S6: When the termination condition is met, output the optimal test parameters as the optimal Shmoo test parameter combination, and determine the performance boundary of the integrated circuit or generate a Shmoo diagram based on the optimal test parameters.

[0038] Termination conditions are used to determine whether the optimization process should end. A reasonable termination condition setting can achieve a balance between optimization effectiveness and computational cost. When the termination condition is met, a local refinement scan centered on the optimal test parameters is performed to determine the boundary between success and failure, or to generate a Shmoo graph.

[0039] Furthermore, in step S2 of this embodiment, the initial test parameter set is generated using the chaotic tent mapping method. A chaotic sequence is generated by iterating the tent mapping function, and the chaotic sequence is mapped to the search range of each Shmoo test parameter to generate the initial test parameter set.

[0040] Chaotic tent maps are a simple yet effective chaotic system characterized by ergodicity, randomness, and sensitivity to initial conditions. By iterating through the tent map function, well-ergodic chaotic sequences can be generated. Compared to random sequences, chaotic sequences exhibit better uniformity and ergodicity, providing a more comprehensive coverage of the parameter space. The initial test parameter set generated in this way is more evenly distributed and comprehensively covered in the parameter space, providing a better starting point for subsequent optimization. Compared to traditional random initialization, chaotic initialization significantly improves the quality of the initial population, reduces the number of iterations required for optimization, and enhances the algorithm's convergence speed and global search capability.

[0041] In step S2 of this embodiment, performing preliminary testing or simulation to obtain initial fitness data includes: The parameters in the initial test parameter set are sent to automated testing equipment to perform Shmoo testing. Alternatively, a pre-established integrated circuit behavior model can be used to simulate and predict the parameters in the initial test parameter set. The former can obtain realistic and accurate test data, but the testing time and cost are relatively high; the latter is fast and low-cost, but its accuracy depends on the accuracy of the model. In practical applications, rapid optimization can be performed through simulation first, followed by actual testing and verification at key parameter points.

[0042] Record the test results corresponding to each set of test parameters. The test results include pass or fail status, performance indicators and bit error rate.

[0043] Calculate the corresponding initial fitness value based on the test results of each set of test parameters.

[0044] In this way, the Shmoo test problem is transformed into a numerical optimization problem that can be handled by the Grey Wolf optimization algorithm.

[0045] In step S3, calculating the fitness value of the current test parameter set in each iteration includes: The fitness value of each set of test parameters is calculated based on the test results, similar to step S2. Since the test parameters are continuously updated during iteration, it is necessary to test or simulate the newly generated test parameters to obtain fitness values. To reduce testing costs, an incremental testing strategy can be adopted, testing only the changed test parameters, or interpolation, machine learning, and other methods can be used to predict the fitness values ​​of untested parameter points.

[0046] The current set of test parameters is sorted according to their fitness values. The test parameter with the highest fitness value is identified as the optimal test parameter, the second highest fitness value as the suboptimal test parameter, and the third highest fitness value as the third optimal test parameter. The optimal test parameter corresponds to the α wolf, the suboptimal test parameter to the β wolf, and the third optimal test parameter to the δ wolf. The α wolf provides the primary search direction, while the β and δ wolves provide auxiliary search directions. Their synergistic effect allows the algorithm to accelerate convergence while maintaining global search capabilities. This three-tiered leadership structure is the core feature of the Gray Wolf Optimization Algorithm. Compared to algorithms guided by a single optimal test parameter, multi-tiered guidance better balances exploration and development, avoiding premature convergence to local optima.

[0047] In step S3 of this embodiment, the update algorithm control parameters adopt a nonlinear convergence factor. The nonlinear convergence factor adopts a nonlinear function based on the Gaussian distribution curve, so that the algorithm control parameters maintain a large value in the early stage of the search in the Shmoo test parameter space for global exploration, and gradually decrease as they approach the optimal test parameters to accelerate local convergence.

[0048] During the iteration process, the value of the nonlinear convergence factor decreases nonlinearly from 2 to 0. In the early stages of iteration (when the value of the nonlinear convergence factor is close to 2), the algorithm's control parameters maintain a relatively large value in the initial search of the Shmoo test parameter space, exhibiting strong global exploration capabilities and thus discovering multiple potential high-quality regions. In the later stages of iteration (when the value of the nonlinear convergence factor is close to 0), the convergence factor decreases rapidly according to Gaussian curve characteristics, causing the test parameters to quickly shift to a local exploration mode, accelerating local convergence and improving convergence speed and accuracy. By automatically adjusting the search strategy according to the optimization process, the nonlinear convergence factor avoids both the local optimum trap caused by premature convergence and the slow convergence problem caused by overexploration, significantly improving the overall performance of the algorithm.

[0049] In step S4, updating other test parameters based on the position information of the optimal test parameter, the second-best test parameter, and the third-best test parameter includes: Using the position update mathematical model of the Grey Wolf optimization algorithm, three candidate position vectors are calculated for the current test parameter to move towards the optimal test parameter, the second-best test parameter, and the third-best test parameter, respectively.

[0050] The behavior of a gray wolf encircling its prey during a hunt corresponds to the process of test parameters moving closer to optimal test parameters. This behavior is achieved by updating the position of each test parameter based on its distance to the currently known optimal test parameter (representing the best test plan found so far). The distance vector D between the test parameter and the target optimal test parameter is calculated as follows: D=|C×X p (t)-X(t)|, where X(t) represents the position vector of the current test parameter at the t-th iteration, X p (t) represents the position vector of the target high-quality test parameters.

[0051] The position update equation for the test parameters is: X(t+1)=X p (t)-A×D.

[0052] Where A = 2a × r1 - a, C = 2 × r2; a is the convergence factor parameter. In this embodiment, the convergence factor parameter a adopts a non-linear decreasing strategy. Its value decreases non-linearly from 2 to 0 according to a non-linear function based on a Gaussian distribution curve as the number of iterations increases (while in traditional algorithms it is usually linearly decreasing). Using non-linear decreasing enables the Shmoo testing method to maintain a high search capability for a longer period in the early stage, and then decreases rapidly in the later stage to enter local convergence. A is a coefficient vector that controls the step size and direction of the current test parameter configuration moving towards the target high-quality test parameter configuration. Its value range decreases as the convergence factor a decreases. C is a weight coefficient vector that provides random weights for the influence of the target high-quality test parameters. Its value range is between [0, 2]. X(t + 1) is the new position of the test parameter configuration in the next iteration (t + 1), representing the updated test parameter combination. r1 and r2 are random vectors uniformly distributed between [0,1], with a dimension equal to the number of test parameters. They generate new random values ​​independently each time a position is updated, introducing randomness into the algorithm and enhancing the diversity of the test parameter population.

[0053] Since the hunting process is primarily guided by alpha, beta, and delta wolves, they are considered to have the best knowledge of the prey's potential location. In the optimization of test parameters, these three wolves correspond to the optimal, second-best, and third-best test parameters in the current population, respectively. All other ω wolves (with ordinary test parameters) update their positions based on the combined effect of these three optimal test parameters.

[0054] In the Shmoo testing method, the position update mathematical model is used with the Grey Wolf optimization algorithm to calculate the distances between the current test parameters and the optimal, second-best, and third-best test parameters, respectively. The calculation formula is as follows: D1 = |C1×X1-X|; D2 = |C2×X2-X|; D3 = |C3×X3-X|.

[0055] Where D1, D2, and D3 are the distance vectors between the current test parameter and the optimal, suboptimal, and third-optimal test parameters (i.e., the distances between ω wolf and α wolf, β wolf, and δ wolf); vectors X1, X2, and X3 represent the position vectors of the optimal, suboptimal, and third-optimal test parameters, respectively; C1, C2, and C3 are independently generated weight coefficient vectors; X1 is the position vector of the optimal test parameter (i.e., the position of α wolf); X2 is the position vector of the suboptimal test parameter (i.e., the position of β wolf); X3 is the position vector of the third-optimal test parameter (i.e., the position of δ wolf); and X is the position vector of the current test parameter (i.e., a certain ω wolf position).

[0056] Guided by the optimal test parameters, the suboptimal test parameters, and the third optimal test parameters, three candidate position vectors are calculated for the current test parameters to move towards the optimal, suboptimal, and third optimal test parameters, respectively. The calculation formula is as follows: X a =X1-A1×D1; X b =X2-A2×D2; X c =X3-A3×D3.

[0057] Where A1, A2, and A3 are coefficient vectors; X a X b X c These represent three candidate position vectors that represent the movement of the current test parameter relative to the optimal test parameter, the second-best test parameter, and the third-best test parameter, respectively.

[0058] This update method ensures that the movement of each test parameter is not only influenced by a single test parameter, but also comprehensively considers the guidance of three different levels of test parameters. This avoids premature convergence caused by over-reliance on a single test parameter, improving the robustness and global search capability of the Shmoo testing method. Furthermore, the introduction of random vectors A and C achieves a balance between deterministic and random search in the Shmoo testing method, effectively utilizing known test parameter information while exploring unknown regions of the parameter space.

[0059] The three candidate position vectors are weighted and summed to obtain the final position update vector X. d .

[0060] Update vector X according to the position update vector X d Update the position of the current test parameters to complete one iteration of position update.

[0061] As a preferred embodiment, the position update vector X is calculated. d The average weighting strategy is adopted, i.e.: X d =(X a +X b +X c ) / 3.

[0062] It should be noted that in other embodiments, the vector X can also be used. a X b and X c (i.e., α wolf, β wolf, and δ wolf) The current fitness values ​​are dynamically assigned weight coefficients ω1, ω2, and ω3 to calculate the position update vector X. d That is: X d =ω1X a +ω2X b +ω3X c In this way, test parameters with higher fitness values ​​play a greater role in position updates, thereby making fuller use of the information from high-quality test parameters to guide algorithm convergence (i.e., further strengthening the guiding role of the better wolf pack).

[0063] Furthermore, in step S4 of this embodiment, the method further includes an adaptive neighborhood mutation mechanism: Monitor the convergence state of the test parameter set and calculate the rate of change of the fitness value of the optimal test parameter within a continuous preset number of algebras.

[0064] When the rate of change of the fitness value is less than a preset threshold, it is determined that the test parameter set may be trapped in a local optimum.

[0065] A random perturbation based on the current position is applied to some test parameters. The amplitude of the random perturbation is adaptively adjusted with the number of iterations. The perturbation amplitude is larger in the early stage of the iteration and gradually decreases in the later stage of the iteration.

[0066] The adaptive neighborhood mutation mechanism significantly enhances the algorithm's ability to escape local optima by monitoring the convergence state and applying adaptively adjusted random perturbations at appropriate times. When the algorithm stagnates, the mutation mechanism introduces new search directions, reactivating population diversity and enabling the algorithm to continue exploring other potential high-quality regions, increasing the probability of finding the globally optimal test parameters. Simultaneously, the adaptive adjustment of the perturbation amplitude ensures that appropriate perturbation strengths are used at different optimization stages, avoiding excessive destruction of existing information in the early stages and preventing excessive perturbation in the later stages that leads to convergence failure.

[0067] In this embodiment, the boundary processing includes: Check if the updated test parameters are within the defined search range.

[0068] For test parameter values ​​that are outside the search range, boundary correction is performed. A boundary truncation strategy is used to limit the parameter values ​​outside the range to the boundary values ​​of the search range, or a bounce strategy is used to remap them into the effective range.

[0069] In practical applications, appropriate boundary handling strategies can be selected based on the characteristics of the problem. For problems where optimal test parameters may exist near the boundary, a truncation strategy is more suitable; for problems that require maintaining population diversity, a bounce strategy is more appropriate. A hybrid strategy can also be used, such as truncation for some parameters and bounce for others.

[0070] By handling boundary conditions, all test parameters remain within the predefined search range, representing a valid and feasible test configuration. This not only ensures the feasibility of the algorithm but also prevents test equipment or simulation systems from reporting errors due to parameters exceeding the range, thus improving the stability and reliability of the algorithm.

[0071] In this embodiment, the termination condition includes at least one of the following conditions: The number of iterations has reached the preset maximum number of iterations; If the fitness value of the optimal test parameter changes less than a preset threshold within a consecutive preset number of generations, it is used to detect whether the algorithm has converged. When the fitness value of the optimal test parameter hardly changes over several consecutive generations, it indicates that the algorithm has found a stable optimization result.

[0072] The fitness value of the optimal test parameters reaches or exceeds the preset optimization target threshold. When the theoretical optimal value is known or there is a clear performance target, the target threshold can be set, and the test can be terminated once it is reached.

[0073] At the end of each iteration, the three termination conditions mentioned above are checked sequentially. The optimization process terminates if any one of them is met, and the α wolf position of the current test parameters is taken as the optimal Shmoo test parameter combination. If the termination condition is not met, the process returns to step S3 and begins a new round of iteration. Through multiple iterations, the test parameter set continuously moves towards the optimal region, gradually approaching the true optimal test parameters. The entire optimization process is a dynamic search process, achieving efficient exploration of the complex parameter space through the cooperative evolution of the population.

[0074] This multi-condition termination mechanism takes into account computational cost control, convergence detection, and goal achievement, enabling the algorithm to terminate flexibly in different scenarios, ensuring optimization quality while avoiding overcomputation.

[0075] Based on the optimal parameter combination, a complete Shmoo plot can also be generated. By performing supplementary tests in the region near the optimal parameters, a precise contour of the failure boundary can be drawn. Traditional grid search methods require testing all grid points in the entire parameter space to draw a Shmoo plot, while the method of this invention first finds the key region through intelligent search, and then performs local dense testing in the key region, which greatly reduces the total number of test points.

[0076] Example 2 A Shmoo testing apparatus, operating using the Shmoo testing method described in Embodiment 1, includes: The initialization configuration module is used to define the Shmoo test parameters to be optimized, the search range, and the optimization objective function, and to generate the initial test parameter set and obtain the initial fitness data using the chaotic tent mapping method.

[0077] The optimization execution module is used to execute the iterative loop of the Grey Wolf optimization algorithm, including a fitness calculation submodule, a parameter identification submodule, a control parameter update submodule, a position update submodule, a boundary handling submodule, and a convergence judgment submodule. The control parameter update submodule uses a nonlinear convergence factor, and the position update submodule includes a dynamic proportional weighting unit and an adaptive neighborhood mutation unit.

[0078] The test interface module connects to automated test equipment or simulation systems via a data bus to send test parameters and receive test results.

[0079] The output module is used to output the optimal combination of test parameters when the termination condition is met, and to generate a Shmoo plot and analysis report.

[0080] Example 3 A computer-readable storage medium storing a computer program that, when executed by a processor, implements the Shmoo testing method as described in Embodiment 1.

[0081] In summary, this invention transforms the multidimensional parameter optimization problem of Shmoo testing into a problem solved using the Grey Wolf Optimization Algorithm. By executing the iterative loop of the Grey Wolf Optimization Algorithm, the fitness value is calculated and the optimal, suboptimal, and tertiary optimal test parameters are identified in each iteration. The location information of these three levels of high-quality test parameters is used to guide the update of other test parameters, enabling the algorithm to effectively capture and optimize the complex interdependencies and linkage effects among multiple parameters. This avoids suboptimal test parameters or misjudgments caused by neglecting parameter linkages in traditional methods. Furthermore, boundary handling ensures that all test parameters remain within the defined search range, and flexible termination conditions achieve a good balance between convergence speed and test parameter accuracy. This allows for faster and more accurate identification of the pass / fail boundaries and optimal operating points of integrated circuits, thereby outputting the optimal combination of test parameters and generating a test spectrum, achieving a more comprehensive and accurate Shmoo test.

[0082] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A Shmoo test method characterized by, The method comprises the following steps: determining Shmoo test parameters to be optimized of an integrated circuit, defining a search range of each Shmoo test parameter to be optimized; establishing an optimization objective function, generating an initial test parameter set, and performing preliminary testing or simulation to obtain initial fitness data; performing an iterative cycle of grey wolf optimization algorithm, calculating a fitness value of the current test parameter set in each iteration, identifying an optimal test parameter, a sub-optimal test parameter and a third optimal test parameter, and updating algorithm control parameters; updating other test parameters according to the position information of the optimal test parameter, the sub-optimal test parameter and the third optimal test parameter; performing boundary processing on the updated test parameters to keep all test parameters within the predefined search range; when the termination condition is met, outputting the optimal test parameter as the optimal Shmoo test parameter combination, and determining the performance boundary of the integrated circuit or generating a Shmoo chart based on the optimal test parameter.

2. The Shmoo test method of claim 1, wherein, The initial test parameter set is generated by using a chaotic tent mapping method, generating a chaotic sequence by iterating a tent mapping function, and mapping the chaotic sequence to the search range of each Shmoo test parameter to generate the initial test parameter set; The initial fitness data is obtained by performing preliminary testing or simulation, which comprises the following steps: sending each group of parameters in the initial test parameter set to an automated test equipment to perform Shmoo testing, or simulating and predicting each group of parameters in the initial test parameter set by using a pre-established integrated circuit behavior model; recording the test results corresponding to each group of test parameters, which include pass or fail status, performance indicators and bit error rate; calculating the initial fitness value of each group of test parameters according to the test results of each group of test parameters.

3. The Shmoo test method of claim 1, wherein, The fitness value of the current test parameter set in each iteration is calculated, which comprises the following steps: calculating the fitness value of each group of test parameters according to the test results of each group of test parameters; sorting the current test parameter set according to the fitness value of each group of test parameters, identifying the test parameter with the highest fitness value as the optimal test parameter, the test parameter with the second highest fitness value as the sub-optimal test parameter, and the test parameter with the third highest fitness value as the third optimal test parameter.

4. The Shmoo test method of claim 1, wherein, The algorithm control parameters are updated by using a nonlinear convergence factor, which is a nonlinear function based on a Gaussian distribution change curve, so that the algorithm control parameters maintain a larger value at the initial stage of the search in the Shmoo test parameter space to perform global exploration, and gradually decrease when approaching the optimal test parameter to accelerate local convergence.

5. The Shmoo test method of claim 1, wherein, The other test parameters are updated according to the position information of the optimal test parameter, the sub-optimal test parameter and the third optimal test parameter, which comprises the following steps: calculating three candidate position vectors of the current test parameters moving to the optimal test parameter, the sub-optimal test parameter and the third optimal test parameter respectively by using the position update mathematical model of the grey wolf optimization algorithm; performing weighted summation on the three candidate position vectors to obtain a final position update vector; updating the position of the current test parameters according to the position update vector.

6. The Shmoo test method of claim 5, wherein, The method further comprises an adaptive neighborhood mutation mechanism. monitoring a convergence state of the test parameter set, calculating a rate of change of fitness value of the optimal test parameter in a preset number of continuous iterations; when the rate of change of fitness value is less than a preset threshold, determining that the test parameter set may fall into local optimum; applying a random disturbance based on the current position to part of the test parameters, the amplitude of the random disturbance being adaptively adjusted according to the iteration number, the disturbance amplitude being larger at the early stage of iteration and gradually decreasing at the later stage of iteration.

7. The Shmoo test method of claim 1, wherein, The boundary processing includes: checking whether the updated test parameters are within the defined search range; performing boundary correction on the test parameter values beyond the search range, limiting the parameter values beyond the range to the boundary values of the search range by using a boundary truncation strategy, or remapping to the effective range by using a rebound strategy.

8. The Shmoo test method of claim 1, wherein, The termination condition includes at least one of the following conditions: the number of iterations reaches a preset maximum number of iterations; the fitness value of the optimal test parameter in a preset number of continuous iterations changes by less than a preset threshold; the fitness value of the optimal test parameter reaches or exceeds a preset optimization target threshold.

9. A Shmoo testing apparatus operating with the Shmoo testing method according to any one of claims 1 to 8, characterized by It includes: an initialization configuration module for defining Shmoo test parameters to be optimized, search range and optimization target function, and generating an initial test parameter set by using a chaotic tent mapping method to obtain initial fitness data; an optimization execution module for executing an iteration loop of the grey wolf optimization algorithm, including a fitness calculation submodule, a parameter identification submodule, a control parameter update submodule, a position update submodule, a boundary processing submodule and a convergence judgment submodule, wherein the control parameter update submodule uses a nonlinear convergence factor, and the position update submodule includes a dynamic proportional weighting unit and an adaptive neighborhood mutation unit; a test interface module connected with an automated test equipment or a simulation system through a data bus, for sending test parameters and receiving test results; an output module for outputting the optimal test parameter combination when the termination condition is met, and generating a Shmoo chart and an analysis report.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to realize the Shmoo test method according to any one of claims 1 to 8.

Images