A semiconductor chip automated test optimization system

Abstract

The application provides a kind of semiconductor chip automation test optimization system, belong to the technical field of measurement electric variable. Including acquisition module, for establishing the time sequence of electric variable arranged in time sequence, drift analysis module, for calculating short-term fluctuation amplitude value, long-term drift slope and high-frequency noise amplitude based on the time sequence of electric variable arranged in time sequence, contact discrimination module, for constructing contact impedance coefficient based on short-term fluctuation amplitude value, long-term drift slope and high-frequency noise amplitude in the jth sliding window, and evaluation and optimization, data marking module, for marking the electric variable measurement data corresponding to the test channel of unstable contact evaluation as low confidence. The system can test the i test channel of temperature sensor chip, more accurate, and calculate short-term fluctuation amplitude value, long-term drift slope and high-frequency noise amplitude at the same time, multi-source data participation, improve test accuracy.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement technology, and more specifically, to an automated testing optimization system for semiconductor chips. Background Technology

[0002] With the widespread application of semiconductor chips in the industrial field, the reliability and accuracy testing of temperature sensor chips has become particularly critical. During the chip electrical performance testing process, it is usually necessary to collect and analyze the electrical variable signals of multiple test channels to evaluate the operating characteristics of the device.

[0003] Current testing systems acquire instantaneous fluctuation signals from the test channels of temperature sensor chips, preprocess the signals, and then process the acquired signal data as a whole for anomaly detection. However, current testing systems only analyze the overall test data of temperature sensor chips and rely on single signal characteristics, such as instantaneous fluctuations, to determine contact stability. They cannot analyze each test channel independently, nor can they simultaneously capture the combined effects of short-term fluctuations, long-term drift, and high-frequency noise. This results in the system being unable to detect minor and complex contact anomalies in a timely manner, reducing test accuracy.

[0004] Therefore, we propose an optimized automated testing system for semiconductor chips to solve this problem. Summary of the Invention

[0005] To overcome the above deficiencies, the present invention provides an optimized system for automated testing of semiconductor chips that overcomes or at least partially solves the above technical problems.

[0006] This invention is implemented as follows:

[0007] This invention provides an automated testing optimization system for semiconductor chips, comprising a data acquisition module, a drift analysis module, a contact discrimination module, and a data marking module;

[0008] The acquisition module is used to acquire contact impedance data and raw electrical variable measurement data of the i-th test channel of the temperature sensor chip at multiple times through a high-precision analog-to-digital converter chip, and to establish a time series of electrical variables arranged in chronological order.

[0009] The drift analysis module is used to calculate short-time fluctuation amplitude values ​​based on time series of electrical variables arranged in chronological order. Long-term drift slope and the high-frequency noise amplitude within j sliding windows ;

[0010] The contact discrimination module is used to determine the contact based on short-term fluctuation amplitude values. Long-term drift slope and the high-frequency noise amplitude within the j-th sliding window Constructing contact resistance coefficient And conduct evaluation and optimization;

[0011] The data labeling module is used to label electrical variable measurement data corresponding to test channels that are found to have unstable contact as having low confidence.

[0012] In a preferred embodiment, the acquisition module includes a region division unit, an acquisition unit, a data sorting unit, and a signal preprocessing unit;

[0013] The region division unit is used to divide the temperature sensor chip test channel into several channels, and label them as the first test channel, the second test channel, ... the i-th test channel, respectively.

[0014] The acquisition unit is used to periodically sample the i-th test channel of the temperature sensor chip multiple times using a high-precision analog-to-digital converter chip with a resolution of 18 bits or more and a sampling rate of 1 kHz or more, and to acquire contact resistance data and original electrical variable measurement data at different time points.

[0015] The data sorting unit is used to sort the contact impedance data and the original measurement data of electrical variables collected at different time points according to the sampling time timestamp, forming a time-continuous electrical variable time series arranged in chronological order.

[0016] The signal preprocessing unit is used to filter and denoise the time series of electrical variables arranged in chronological order.

[0017] In a preferred embodiment, the drift analysis module includes a calculation unit and an evaluation unit;

[0018] The calculation unit is used to calculate the short-time fluctuation amplitude value based on the processed time series of electrical variables arranged in chronological order. Long-term drift slope and the high-frequency noise amplitude within the j-th sliding window ;

[0019] Short-term fluctuation range The specific method of obtaining it is as follows:

[0020] The processed contact impedance time series is divided into multiple short-time sliding windows of length N, where the data points within the time period corresponding to the j-th sliding window are... for;

[0021] ;

[0022] Calculate the short-term fluctuation amplitude value based on the data points within the time period corresponding to the j-th sliding window. .

[0023] In a preferred embodiment, the evaluation unit includes a short-term fluctuation amplitude evaluation subunit. ;

[0024] The short-term fluctuation amplitude assessment subunit is used to preset the short-term fluctuation amplitude threshold. And set the short-term fluctuation threshold With short-term fluctuation amplitude value Compare the data and generate short-term fluctuation amplitude assessment instructions, including:

[0025] when > When the signal fluctuation is abnormal within the j-th sliding window, a short-term fluctuation amplitude strategy is generated, including extending the sampling time of the i-th test channel by 20%-32% to improve signal stability and lowering the digital filter cutoff frequency of the acquired signal by 10%-25% to enhance anti-interference capability.

[0026] when ≤ When the signal fluctuation is normal within the j-th sliding window, the sampling time and filtering parameters for the i-th test channel remain unchanged, and the fluctuation trend of subsequent sliding windows is continuously monitored.

[0027] In a preferred embodiment, the long-term drift slope The specific method of obtaining it is as follows:

[0028] Based on the j-th sliding window, the window contains a set of sampling points;

[0029] , This represents the total number of sampling points;

[0030] In the formula, Represented as the Unix timestamp of the Pth sampling point. This represents the contact impedance value at the Pth sampling point;

[0031] Calculate the mean of the contact impedance Unix timestamps within the j-th sliding window. ;

[0032] ;

[0033] Calculate the mean value of the contact impedance within the j-th sliding window. ;

[0034] ;

[0035] Then, the contact impedance time series data are fitted using the least squares method, and the fitting function is:

[0036] ;

[0037] In the formula, The intercept is... This represents the long-term drift slope;

[0038] Next, the long-term drift slope of the fitted line is calculated according to the least squares criterion. ;

[0039] .

[0040] In a preferred embodiment, the evaluation unit further includes a long-term drift slope evaluation subunit;

[0041] The long-term drift slope evaluation subunit is used to preset the long-term drift slope threshold. and the long-term drift slope threshold With long-term drift slope Compare these parameters to generate long-term drift slope assessment instructions, including:

[0042] when > When the i-th test channel exhibits a long-term drift anomaly, a long-term drift slope strategy is generated, including increasing the pressure of the i-th test channel by 9%-23% to enhance contact stability, reducing the drift amplitude, increasing the contact impedance sampling frequency by 28%-37% to capture impedance changes more promptly and respond quickly to abnormal drift, and extending the sampling time of the i-th test channel by 28%-44% to improve data stability and acquisition accuracy.

[0043] when ≤ When the i-th test channel is tested normally, the current sampling frequency is maintained, and the drift slope and contact impedance are continuously monitored.

[0044] In a preferred embodiment, the high-frequency noise amplitude within the j-th sliding window The specific method of obtaining it is as follows:

[0045] Bandpass filtering is applied to the processed electrical variable time series arranged in chronological order to extract high-frequency components. Commonly used bandpass filter parameters are set according to experimental or system requirements, and generally a frequency band higher than the short-time fluctuation frequency range is selected.

[0046] Based on the data points within the time period corresponding to the j-th sliding window for;

[0047] ;

[0048] For the k-th data point in the j-th sliding window, calculate the peak-to-peak value to obtain the high-frequency noise amplitude within the j-th sliding window. .

[0049] In a preferred embodiment, the evaluation unit further includes a high-frequency noise evaluation subunit;

[0050] The high-frequency noise evaluation subunit is used to preset the high-frequency noise threshold. and set the high-frequency noise threshold With high frequency noise amplitude Compare and generate high-frequency noise assessment instructions, including:

[0051] when > When the signal in the j-th sliding window of the i-th test channel is abnormal, a high-frequency noise generation strategy is adopted, including lowering the cutoff frequency of the digital low-pass filter by 15%-23% to enhance the suppression of high-frequency noise, extending the sampling time of the i-th test channel by 27%-33% to improve the effective signal-to-noise ratio, and using multiple repeated sampling and taking the middle finger instead of single sampling to reduce the impact of random interference.

[0052] when ≤ When the signal in the j-th sliding window of the i-th test channel is normal, the current sampling time and the current digital filtering parameters are maintained to avoid excessive blue-green waves causing loss of signal details.

[0053] In a preferred embodiment, the contact discrimination module includes an association unit and an analysis unit;

[0054] The associated unit is used to convert the short-term fluctuation amplitude value Long-term drift slope and high frequency noise amplitude The contact resistance coefficient, after being dimensionless, is calculated using the following formula. ;

[0055] The analysis unit is used to preset the contact impedance threshold. and the contact impedance threshold With contact resistance coefficient Compare and generate contact resistance assessment instructions, including:

[0056] when > When the contact state of the i-th test channel is abnormal, a contact impedance strategy is generated, including extending the sampling time of the i-th test channel by 34%-56% to improve sampling stability, increasing the length of the j-th sliding window by 55%-67%, halving the step size to improve trend estimation accuracy, and lowering the digital filter cutoff frequency by 19%-37% to improve noise suppression.

[0057] when ≤ When the i-th test channel is in normal contact status, the current sampling frequency and filter plate parameters are maintained, and normal data acquisition and analysis continue to ensure continuous and stable data.

[0058] In a preferred embodiment, the data tagging module includes tagging units;

[0059] The marking unit is used for contact impedance evaluation instructions, including:

[0060] when < ≤ *1.5 indicates an abnormal contact in the i-th test channel, and the corresponding electrical variable measurement data is marked as low confidence -1;

[0061] when *1.5< ≤ When *2 is used, it indicates that the i-th test channel has an abnormal contact, and the corresponding electrical variable measurement data is marked as low confidence -2.

[0062] then > When *2, it indicates that the i-th test channel has an abnormal contact, and the corresponding electrical variable measurement data is marked as low confidence -3.

[0063] The present invention provides an optimized automated testing system for semiconductor chips, the advantages of which include:

[0064] By acquiring contact impedance and electrical variable data of the i-th test channel of the temperature sensor chip using a high-precision analog-to-digital converter chip, and establishing a time-continuous contact impedance data sequence and electrical variable time sequence, it helps to accurately capture the electrical characteristics of the chip at different time points in the i-th test channel, improving test accuracy. Furthermore, the drift analysis module can simultaneously calculate the short-term fluctuation amplitude, long-term drift slope, and high-frequency noise amplitude, thereby performing a comprehensive dynamic analysis of the contact impedance signal, quickly identifying signal drift and noise characteristics. Moreover, based on the short-term fluctuation, long-term drift, and high-frequency noise, a contact impedance coefficient is constructed, and the contact discrimination module can also comprehensively evaluate the contact state of the i-th test channel, ensuring that subsequent analysis and processing rely only on reliable data, thus improving the overall reliability of the system. Attached Figure Description

[0065] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0066] Figure 1 This is a system block diagram of an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Example 1, referring to Figure 1 The present invention provides a technical solution: an automated testing optimization system for semiconductor chips, comprising a data acquisition module, a drift analysis module, a contact discrimination module, and a data marking module;

[0069] The acquisition module is used to acquire contact impedance data and raw electrical variable measurement data of the i-th test channel of the temperature sensor chip at multiple times through a high-precision analog-to-digital converter chip, and to establish a time series of electrical variables arranged in chronological order.

[0070] The drift analysis module is used to calculate short-time fluctuation amplitude values ​​based on time series of electrical variables arranged in chronological order.

[0071] Long-term drift slope and the high-frequency noise amplitude within j sliding windows ;

[0072] The contact discrimination module is used to determine the contact based on short-term fluctuation amplitude values. Long-term drift slope and the high-frequency noise amplitude within the j-th sliding window Constructing contact resistance coefficient And conduct evaluation and optimization;

[0073] The data labeling module is used to label electrical variable measurement data corresponding to test channels that are found to have unstable contact as having low confidence.

[0074] In this embodiment, the contact impedance data and electrical variable data of the temperature sensor are collected by a high-precision analog-to-digital converter chip, and a time series of electrical variables with continuous time is established. This helps to accurately capture the electrical characteristics of the chip at different time points and improve the test accuracy. The drift analysis module can simultaneously calculate the short-term fluctuation amplitude value, long-term drift slope and high-frequency noise amplitude, thereby performing a comprehensive dynamic analysis of the contact impedance signal and quickly identifying signal drift and noise characteristics.

[0075] Based on short-term fluctuations, long-term drift, and high-frequency noise, the contact impedance coefficient is constructed. The contact discrimination module can comprehensively evaluate the contact status of the test channel and generate optimization strategies to achieve automatic adjustment and improvement of the contact status. The data marking module can mark the electrical variable measurement data of the test channel with unstable contact with low confidence, ensuring that subsequent analysis and processing rely only on reliable data, thereby improving the overall reliability of the system and the credibility of the test results.

[0076] Example 2 is an explanation of Example 1; please refer to it. Figure 1 Specifically, the acquisition module includes a region division unit, an acquisition unit, a data sorting unit, and a signal preprocessing unit;

[0077] The region division unit is used to divide the temperature sensor chip test channel into several channels, and label them as the first test channel, the second test channel, ... the i-th test channel, respectively.

[0078] The acquisition unit is used to periodically sample the i-th test channel of the temperature sensor chip multiple times using a high-precision analog-to-digital converter chip with a resolution of 18 bits or more and a sampling rate of 1 kHz or more, and to acquire contact resistance data and original electrical variable measurement data at different time points.

[0079] The data sorting unit is used to sort the contact impedance data and the original measurement data of electrical variables collected at different time points according to the sampling time timestamp, forming a time-continuous electrical variable time series arranged in chronological order.

[0080] The signal preprocessing unit is used to filter the time series of electrical variables arranged in chronological order, and to remove high-frequency noise and transient interference using digital filtering algorithms, specifically including: using digital filtering algorithms, such as median filters;

[0081] The aim is to effectively suppress pulse-type transient interference and avoid the distortion problem of the mean method;

[0082] First, set the window length, such as 3, 5, or 7. For each sampling point, combine it with the two points before and after it to form a window. If there are less than two sampling points at the start and end of the sequence, use boundary value copying to fill in the missing data.

[0083] Then, the data in the window is sorted from smallest to largest, and the median value is taken as the filtering result for that point.

[0084] In this embodiment, the temperature sensor chip test channels are precisely divided by a region division unit, and a high-precision analog-to-digital converter chip with a resolution of ≥18 bits and a sampling rate of ≥1kHz is used to achieve high-precision, periodic multiple sampling of each test channel. This ensures that the collected contact impedance and electrical variable raw measurement data have high resolution and high fidelity. The data sorting unit sorts the data collected at different time points according to the timestamp of the sampling time, forming a time-continuous electrical variable time series arranged in chronological order. This ensures the temporal integrity of the data and provides a reliable data foundation for subsequent drift analysis, fluctuation calculation and trend prediction.

[0085] The acquisition module ensures the reliability and consistency of the output signal through precise segmentation, stable sampling, time series processing, and high-quality preprocessing. This provides high-quality input data for the drift analysis module, contact discrimination module, and data marking module, fundamentally improving the performance and robustness of the entire test optimization system.

[0086] Example 3 is an explanation of Example 1; please refer to the provided text. Figure 1 Specifically, the drift analysis module includes a calculation unit and an evaluation unit;

[0087] The calculation unit is used to calculate the short-time fluctuation amplitude value based on the processed time series of electrical variables arranged in chronological order. Long-term drift slope and the high-frequency noise amplitude within j sliding windows value;

[0088] Short-term fluctuation range The specific method of obtaining it is as follows:

[0089] The processed contact impedance time series is divided into multiple short-time sliding windows of length N, where the data points within the time period corresponding to the j-th sliding window are... for;

[0090] ;

[0091] Calculate the short-term fluctuation amplitude value based on the data points within the time period corresponding to the j-th sliding window. ;

[0092] ;

[0093] In the formula, It represents the maximum value of the j-th sliding window data point within the sampling time. It represents the minimum value of the j-th sliding window data point within the sampling time;

[0094] For example, suppose the sampled value within the j-th sliding window is ;

[0095] The maximum value here is 1.3, and the minimum value is 0.9;

[0096] The short-term fluctuation range is 1.3 - 0.9 = 0.4.

[0097] The following are the short-term fluctuation amplitude values. The sample data example table is shown below;

[0098] Sliding window number Data points within the time period corresponding to the j-th sliding window (Ω) The maximum value of the j-th sliding window data point within the sampling time. (Ω) Minimum value of the j-th sliding window data point within the sampling time (Ω) Short-term fluctuation range (Ω) 1 0.95,1.00,1.05,0.98,1.02 1.05 0.95 0.10 2 1.10,1.12,1.15,1.08,1.09 1.15 1.08 0.07 3 1.20,1.25,1.18,1.22,1.21 1.25 1.18 0.07 4 1.30,1.28,1.33,1.31,1.29 1.33 1.28 0.05 5 1.35,1.38,1.37,1.36,1.34 1.38 1.34 0.04

[0099] In this embodiment, by dividing the processed contact impedance time series into multiple short-time sliding windows of length N, and calculating the difference between the maximum and minimum values ​​of the data points within the j-th window, the fluctuation amplitude of the test channel within a short time range can be quantified intuitively and quickly, thereby effectively reflecting the instantaneous instability of the contact impedance. The calculation of the short-time fluctuation amplitude does not rely on complex frequency domain transformation, but directly utilizes the extreme value difference of the time domain data, which can keenly capture local abrupt signals caused by poor contact, noise impact, etc., and improve the sensitivity of detecting small and rapid fluctuations.

[0100] The short-term fluctuation amplitude, along with the long-term drift slope and high-frequency noise amplitude, constitutes an important input parameter for the contact impedance coefficient. Accurate acquisition of this parameter lays a solid foundation for subsequent comprehensive evaluation and optimization strategy generation.

[0101] Example 4 is an explanation of Example 1; please refer to the provided text. Figure 1 Specifically, the evaluation unit includes a short-term fluctuation amplitude evaluation subunit. ;

[0102] The short-term fluctuation amplitude assessment subunit is used to preset the short-term fluctuation amplitude threshold. ,

[0103] By collecting signals under multiple similar operating conditions, the distribution range of short-term fluctuation amplitude values ​​is recorded. The upper limit of normal fluctuation is extracted from the distribution range, and combined with historical data for reference, the threshold of short-term fluctuation amplitude is determined. The value is 0.06;

[0104] and short-term fluctuation threshold With short-term fluctuation amplitude value Compare the data and generate short-term fluctuation amplitude assessment instructions, including:

[0105] when > When the signal fluctuation is abnormal within the j-th sliding window, a short-term fluctuation amplitude strategy is generated, including extending the sampling time of the i-th test channel by 20%-32% to improve signal stability and lowering the digital filter cutoff frequency of the acquired signal by 10%-25% to enhance anti-interference capability.

[0106] when ≤ When the signal fluctuation is normal within the j-th sliding window, the sampling time and filtering parameters for the i-th test channel remain unchanged, and the fluctuation trend of subsequent sliding windows is continuously monitored.

[0107] Based on the sliding window number, continue to construct short-term fluctuation amplitude thresholds. With short-term fluctuation amplitude value A sample table of comparison data is shown below;

[0108] Sliding window number Short-term fluctuation range (Ω) Short-term fluctuation threshold Evaluation results 1 0.10 0.06 An anomaly occurred; a short-term volatility strategy was executed. 2 0.07 0.06 An anomaly occurred; a short-term volatility strategy was executed. 3 0.07 0.06 An anomaly occurred; a short-term volatility strategy was executed. 4 0.05 0.06 normal 5 0.04 0.06 normal

[0109] In this embodiment, by setting a short-term fluctuation amplitude threshold and comparing it with the actual calculated short-term fluctuation amplitude value in real time, it is possible to determine whether the signal fluctuation is abnormal at the window level, thereby achieving rapid identification of short-term instability of contact impedance. Combined with the mechanism of dynamically adjusting sampling time and digital filtering parameters, the system can adapt to changes in hardware contact state and maintain high measurement stability and data reliability in the test environment.

[0110] Example 5 is an explanation of Example 1; please refer to it. Figure 1 Specifically, the long-term drift slope The specific method of obtaining it is as follows:

[0111] Based on the j-th sliding window, the window contains a set of sampling points;

[0112] , This represents the total number of sampling points;

[0113] In the formula, Represented as the Unix timestamp of the Pth sampling point. This represents the contact impedance value at the Pth sampling point;

[0114] Calculate the mean of the contact impedance Unix timestamps within the j-th sliding window. ;

[0115] ;

[0116] Calculate the mean value of the contact impedance within the j-th sliding window. ;

[0117] ;

[0118] Then, the contact impedance time series data are fitted using the least squares method, and the fitting function is:

[0119] ;

[0120] In the formula, The intercept is... This represents the long-term drift slope;

[0121] Next, the long-term drift slope of the fitted line is calculated according to the least squares criterion. ;

[0122] .

[0123] It should be noted that the long-term drift slope It's not the slope of a simple straight line drawn from two points, but rather the slope of a least-squares fitted line. Based on the least-squares criterion, by minimizing the sum of the squared distances between all sampling points and the fitted line, the long-term drift slope can be derived. Without first writing out the fitting function, it's impossible to clearly define which line the slope refers to. The data may be curved or noisy, so only by clearly defining the slope of the fitted line can the calculation formula have a basis.

[0124] Based on the sliding window numbering, continue to construct the long-term drift slope. The sample data example table is shown below;

[0125] Sliding window number Total number of sampling points Unix timestamps (s) of the sampling points Contact resistance value (Ω) Mean of contact impedance Unix timestamps within the j-th sliding window (s) The average contact impedance value within the j-th sliding window (Ω) intercept (Ω) Long-term drift slope (Ω / s) 1 5 10、20、30、40、50 1.00、1.02、1.03、1.05、1.06 30 1.03 0.97 0.0021 2 5 60、70、80、90、100 1.07、1.08、1.10、1.12、1.13 80 1.10 1.05 0.0020 3 5 110、120、130、140、150 1.14、1.15、1.17、1.18、1.20 130 1.17 1.12 0.0021 4 5 160、170、180、190、200 1.21、1.22、1.24、1.25、1.27 180 1.24 1.19 0.0020 5 5 210、220、230、240、250 1.28、1.30、1.31、1.33、1.35 230 1.31 1.26 0.0020

[0126] In this embodiment, by calculating the mean of the contact impedance data within a sliding window and fitting it using the least squares method, the long-term drift slope can be accurately extracted, intuitively reflecting the overall trend of contact impedance change over time. This avoids accidental misjudgments caused by relying on single-point measurements. By averaging the timestamp and impedance values ​​and performing linear fitting, short-term random fluctuations and transient interference signals can be effectively smoothed out, making the calculated slope more reflective of the true changes in the contact state.

[0127] Example 6 is an explanation of Example 1; please refer to the provided text. Figure 1 Specifically, the evaluation unit also includes a long-term drift slope evaluation subunit;

[0128] The long-term drift slope evaluation subunit is used to preset the long-term drift slope threshold. ,

[0129] By collecting long-term stable data under normal equipment operation and interference-free conditions, signals are collected over a relatively long period, such as several hours to several days. Then, linear fitting is performed on the data of the j-th sliding window to obtain a slope sequence. The mean and standard deviation are then calculated based on the slope sequence. Data exceeding the mean and standard deviation are analyzed to determine if abnormal fluctuations exist, thereby deriving a suitable long-term drift slope threshold. And for the long-term drift slope threshold The value is 0.0020;

[0130] and the long-term drift slope threshold With long-term drift slope Compare these parameters to generate long-term drift slope assessment instructions, including:

[0131] when > When the i-th test channel exhibits a long-term drift anomaly, a long-term drift slope strategy is generated, including increasing the pressure of the i-th test channel by 9%-23% to enhance contact stability, reducing the drift amplitude, increasing the contact impedance sampling frequency by 28%-37% to capture impedance changes more promptly and respond quickly to abnormal drift, and extending the sampling time of the i-th test channel by 28%-44% to improve data stability and acquisition accuracy.

[0132] when ≤ When the i-th test channel is tested normally, the current sampling frequency is maintained, and the drift slope and contact impedance are continuously monitored.

[0133] Based on the sliding window number, we continue to construct the long-term drift slope threshold. With long-term drift slope A sample table of comparison data is shown below;

[0134] Sliding window number Long-term drift slope (Ω / s) Long-term drift slope threshold Evaluation results 1 0.0021 0.0020 Anomaly, executing a long-term drift slope strategy. 2 0.0020 0.0020 normal 3 0.0021 0.0020 Anomaly, executing a long-term drift slope strategy. 4 0.0020 0.0020 normal 5 0.0020 0.0020 normal

[0135] In this embodiment, by comparing the long-term drift slope with a preset threshold, the contact degradation trend of the i-th test channel can be identified in advance before the impedance value exceeds the standard, thus achieving early warning and reducing the risk of test failure. When the drift slope exceeds the threshold, the system will automatically execute a combination of strategies such as increasing pressure, increasing sampling frequency, and extending sampling time, which can simultaneously improve the drift problem from three dimensions: contact stability, data acquisition speed, and acquisition accuracy, thereby improving test reliability.

[0136] Example 7 is an explanation of Example 1; please refer to it. Figure 1 Specifically, the high-frequency noise amplitude within the j-th sliding window The specific method of obtaining it is as follows:

[0137] Bandpass filtering is applied to the processed electrical variable time series arranged in chronological order to extract high-frequency components. Commonly used bandpass filter parameters are set according to experimental or system requirements, and generally a frequency band higher than the short-time fluctuation frequency range is selected.

[0138] Based on the data points within the time period corresponding to the j-th sliding window for;

[0139] ;

[0140] For the k-th data point in the j-th sliding window, calculate the peak-to-peak value to obtain the high-frequency noise amplitude within the j-th sliding window. ;

[0141] ;

[0142] In the formula, k takes values ​​of 1, 2, 3, ..., T, where T represents the total number of sampling points;

[0143] , representing the maximum signal value at the k-th data point in the j-th sliding window;

[0144] , which represents the minimum signal value at the k-th data point in the j-th sliding window.

[0145] The following are the sliding window numbers; continue constructing the high-frequency noise amplitude within the j-th sliding window. The sample data example table is shown below;

[0146] Sliding window number Data points within the time period corresponding to the j-th sliding window (Ω) High-frequency components The maximum signal value of the k-th data point in the j-th sliding window. The minimum signal value of the k-th data point in the j-th sliding window. High-frequency noise amplitude within the j-th sliding window 1 0.02、0.01、0.03、0.015、0.018 0.03 0.01 0.02 2 0.025、0.028、0.03、0.022、0.027 0.03 0.022 0.008 3 0.03、0.035、0.028、0.032、0.031 0.035 0.028 0.007 4 0.04、0.038、0.043、0.041、0.039 0.043 0.038 0.005 5 0.045、0.048、0.047、0.046、0.044 0.048 0.044 0.004

[0147] In this embodiment, by applying bandpass filtering to the time series of electrical variables arranged in chronological order, high-frequency components above the short-term fluctuation frequency range can be effectively separated, avoiding interference from low-frequency drift or DC offset on noise analysis and improving the accuracy of high-frequency noise amplitude calculation. By adopting a sliding window method (the j-th sliding window), the changes in high-frequency noise can be captured in segments in the time domain, which can reflect the overall trend and detect instantaneous interference peaks, thereby improving the timeliness and local sensitivity of noise monitoring.

[0148] The peak-to-peak value calculation method based on the maximum and minimum values ​​within the window can directly quantify the amplitude of high-frequency noise, avoiding complex processing such as spectrum analysis, improving real-time processing efficiency, and is suitable for online detection and rapid feedback scenarios. The center frequency and bandwidth of the bandpass filter can be flexibly set according to experimental or system requirements, and can adapt to the testing needs of different types of semiconductor chips, different packaging structures and different production lines, enhancing the adaptability and versatility of the system.

[0149] Example 8 is an explanation of Example 1; please refer to it. Figure 1 Specifically, the evaluation unit also includes a high-frequency noise evaluation subunit;

[0150] The high-frequency noise evaluation subunit is used to preset the high-frequency noise threshold. ,

[0151] By long-term acquisition of the signal from the i-th test channel under normal operating conditions, followed by preliminary filtering of the signal and extraction of high-frequency noise amplitude, the extracted data is analyzed. A preset threshold is then established and compared with and fine-tuned against historical thresholds to derive the high-frequency noise threshold. The value is 0.006;

[0152] And high frequency noise threshold With high frequency noise amplitude Compare and generate high-frequency noise assessment instructions, including:

[0153] when > When the signal in the j-th sliding window of the i-th test channel is abnormal, a high-frequency noise generation strategy is adopted, including lowering the cutoff frequency of the digital low-pass filter by 15%-23% to enhance the suppression of high-frequency noise, extending the sampling time of the i-th test channel by 27%-33% to improve the effective signal-to-noise ratio, and using multiple repeated sampling and taking the middle finger instead of single sampling to reduce the impact of random interference.

[0154] when ≤ When the signal in the j-th sliding window of the i-th test channel is normal, the current sampling time and the current digital filtering parameters are maintained to avoid excessive blue-green waves causing loss of signal details.

[0155] Based on the sliding window numbering, continue to construct the high-frequency noise threshold. With high frequency noise amplitude A sample table of comparison data is shown below;

[0156] Sliding window number High-frequency noise amplitude within the j-th sliding window High-frequency noise threshold Evaluation results 1 0.02 0.006 An anomaly occurred; a high-frequency noise strategy was executed. 2 0.008 0.006 An anomaly occurred; a high-frequency noise strategy was executed. 3 0.007 0.006 An anomaly occurred; a high-frequency noise strategy was executed. 4 0.005 0.006 normal 5 0.004 0.006 normal

[0157] In this embodiment, the high-frequency noise assessment subunit can dynamically adjust the sampling and filtering strategies according to the noise level during the test, achieving a balance between detection accuracy and processing speed. This adapts to the automated testing needs of semiconductor chips under different environments and operating conditions. The assessment results of the high-frequency noise amplitude can be used together with the short-term fluctuation amplitude and long-term drift slope to calculate the contact impedance coefficient, forming a multi-dimensional contact stability judgment mechanism. This allows for more accurate identification of contact anomalies and guidance for optimization strategies.

[0158] Example 9, this example is an explanation of Example 1, please refer to it. Figure 1 Specifically, the contact discrimination module includes an association unit and an analysis unit;

[0159] The associated unit is used to convert the short-term fluctuation amplitude value Long-term drift slope and high frequency noise amplitude The contact resistance coefficient, after being dimensionless, is calculated using the following formula. ;

[0160] ;

[0161] In the formula, , and Let be the weighting coefficient, satisfying , This is represented as a reference value for short-term fluctuation range. This is represented as a reference value for the long-term drift slope. This is represented as a reference value for the amplitude of high-frequency noise.

[0162] Preset 0.4 0.3 and It is 0.3. It is 0.1. It is 0.002. It is 0.02.

[0163] Analysis of the collected data revealed that short-term fluctuations directly reflect the significant impact of loosening of the contact surface, thus warranting a high weighting. The value is 0.4;

[0164] Similarly, in data analysis, the long-term drift slope affects long-term measurement accuracy, and the weighting is moderate, so it is important to... The value is 0.3;

[0165] Similarly, in data analysis, the amplitude of high-frequency noise affects the instantaneous accuracy and anti-interference ability of the signal, and the weighting is moderate, so it is important to... The value is 0.3;

[0166] Based on historical measurement data, the typical upper limit of short-term fluctuation amplitude under normal working conditions will be determined. The value is 0.1;

[0167] Based on historical measurement data, typical values ​​of long-term drift slope under normal conditions are statistically analyzed. The value is 0.002;

[0168] Based on historical measurement data, typical values ​​of high-frequency noise amplitude under normal conditions are statistically analyzed. The value is 0.02;

[0169] Based on the sliding window numbering, the contact impedance coefficient is further constructed. The sample data example table is shown below;

[0170] Sliding window number Short-term fluctuation range (Ω) Short-term fluctuation range reference value Long-term drift slope (Ω / s) Long-term drift slope reference value High-frequency noise amplitude within the j-th sliding window High-frequency noise amplitude reference value Contact resistance coefficient 1 0.10 0.1 0.0021 0.002 0.02 0.02 1.015 2 0.07 0.1 0.0020 0.002 0.008 0.02 0.7 3 0.07 0.1 0.0021 0.002 0.007 0.02 0.7 4 0.05 0.1 0.0020 0.002 0.005 0.02 0.575 5 0.04 0.1 0.0020 0.002 0.004 0.02 0.52

[0171] The analysis unit is used to preset the contact impedance threshold. ;

[0172] The contact impedance coefficient is calculated for each sliding window in historical data. Statistical analysis is performed on the contact impedance coefficient for each sliding window to obtain the mean, standard deviation, maximum value, and quantiles. Based on the overall fluctuation characteristics of the historical data, the value closest to the fluctuation point is selected. By appropriately increasing the safety margin, the contact impedance threshold is derived. The value is 0.6;

[0173] and contact impedance threshold The contact resistance coefficient is compared to generate a contact resistance assessment instruction, including:

[0174] when > When the contact state of the i-th test channel is abnormal, a contact impedance strategy is generated, including extending the sampling time of the i-th test channel by 34%-56% to improve sampling stability, increasing the length of the j-th sliding window by 55%-67%, halving the step size to improve trend estimation accuracy, and lowering the digital filter cutoff frequency by 19%-37% to improve noise suppression.

[0175] when ≤ When the i-th test channel is in normal contact status, the current sampling frequency and filter plate parameters are maintained, and normal data acquisition and analysis continue to ensure continuous and stable data.

[0176] Based on the sliding window number, the contact impedance threshold is further constructed. With contact resistance coefficient A sample table of comparison data is shown below;

[0177] Sliding window number Contact resistance coefficient Contact resistance threshold Evaluation results 1 1.015 0.6 Abnormal situation, execute contact impedance strategy 2 0.7 0.6 Abnormal situation, execute contact impedance strategy 3 0.7 0.6 Abnormal situation, execute contact impedance strategy 4 0.575 0.6 normal 5 0.52 0.6 normal

[0178] In this embodiment, short-term fluctuation amplitude, long-term drift slope, and high-frequency noise amplitude are correlated through an association unit, and the contact impedance coefficient is calculated after dimensionless processing. Compared with a single index judgment method, this can comprehensively reflect the contact stability of the test channel, reduce missed judgments and false judgments. By dynamically adjusting the sampling time, sliding window parameters, and filtering frequency, the system can balance acquisition accuracy and test speed while ensuring contact stability, adapting to test requirements under different working conditions. The contact discrimination module fuses the features of multiple sources for unified judgment and links with the automated optimization strategy, effectively improving the stability and data reliability of the entire semiconductor chip automated test system, providing a solid data foundation for subsequent analysis and quality assessment.

[0179] Example 10: This example is an explanation of Example 1. Please refer to the provided text. Figure 1 Specifically, the data tagging module includes tagging units;

[0180] The marking unit is used for contact impedance evaluation instructions, including:

[0181] when < ≤ *1.5 indicates an abnormal contact in the i-th test channel, and the corresponding electrical variable measurement data is marked as low confidence -1;

[0182] when *1.5< ≤ When *2 is used, it indicates that the i-th test channel has an abnormal contact, and the corresponding electrical variable measurement data is marked as low confidence -2.

[0183] then > When *2, it indicates that the i-th test channel has an abnormal contact, and the corresponding electrical variable measurement data is marked as low confidence -3.

[0184] In this embodiment, by comparing the contact impedance coefficient with multiples of a preset contact impedance threshold in a tiered manner and assigning them labels such as low confidence level -1, low confidence level -2, and low confidence level -3, a fine distinction of the degree of contact anomaly is achieved. This allows for the differentiation of minor, moderate, and severe contact anomalies, avoiding a "one-size-fits-all" judgment and improving the interpretability and traceability of the judgment results. By labeling different levels of low confidence data, the interference of abnormal data on overall trend judgment and model training can be effectively reduced, thereby improving the overall analysis accuracy of the system.

[0185] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value, it is acceptable.

[0186] The above formulas are all derived from software simulation using a large amount of data and are selected to be close to the actual values. The coefficients in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.

Claims

1. An automated testing optimization system for semiconductor chips, characterized in that, It includes a data acquisition module, a drift analysis module, a contact discrimination module, and a data tagging module; The acquisition module is used to acquire contact impedance data and raw electrical variable measurement data of the i-th test channel of the temperature sensor chip at multiple times through a high-precision analog-to-digital converter chip, and to establish a time series of electrical variables arranged in chronological order. The drift analysis module is used to calculate short-time fluctuation amplitude values ​​based on time series of electrical variables arranged in chronological order. Long-term drift slope and the high-frequency noise amplitude within the j-th sliding window ; The contact discrimination module is used to determine the contact based on short-term fluctuation amplitude values. Long-term drift slope and the high-frequency noise amplitude within the j-th sliding window Constructing contact resistance coefficient And conduct evaluation and optimization; The data labeling module is used to label electrical variable measurement data corresponding to test channels that are found to have unstable contact as having low confidence.

2. The semiconductor chip automated testing optimization system according to claim 1, characterized in that, The acquisition module includes a region division unit, an acquisition unit, a data sorting unit, and a signal preprocessing unit; The region division unit is used to divide the temperature sensor chip test channel into several channels, and label them as the first test channel, the second test channel, ... the i-th test channel, respectively. The acquisition unit is used to periodically sample the i-th test channel of the temperature sensor chip multiple times using a high-precision analog-to-digital converter chip with a resolution of 18 bits or more and a sampling rate of 1 kHz or more, and to acquire contact resistance data and original electrical variable measurement data at different time points. The data sorting unit is used to sort the contact impedance data and the original measurement data of electrical variables collected at different time points according to the sampling time timestamp, forming a time-continuous electrical variable time series arranged in chronological order. The signal preprocessing unit is used to filter and denoise the time series of electrical variables arranged in chronological order.

3. The semiconductor chip automated testing optimization system according to claim 2, characterized in that, The drift analysis module includes a calculation unit and an evaluation unit; The calculation unit is used to calculate the short-time fluctuation amplitude value based on the processed time series of electrical variables arranged in chronological order. Long-term drift slope and the high-frequency noise amplitude within the j-th sliding window ; Short-term fluctuation range The specific method of obtaining it is as follows: The processed contact impedance time series is divided into multiple short-time sliding windows of length N, where the data points within the time period corresponding to the j-th sliding window are... for: ； Calculate the short-term fluctuation amplitude value based on the data points within the time period corresponding to the j-th sliding window. .

4. The semiconductor chip automated testing optimization system according to claim 3, characterized in that, The assessment unit includes a short-term fluctuation amplitude assessment subunit; The short-term fluctuation amplitude assessment subunit is used to preset the short-term fluctuation amplitude threshold. And set the short-term fluctuation threshold With short-term fluctuation amplitude value The comparison is performed to generate a short-term fluctuation amplitude assessment instruction, including: when > When the signal fluctuation is abnormal within the j-th sliding window, a short-term fluctuation amplitude strategy is generated, including extending the sampling time of the i-th test channel by 20%-32% and reducing the digital filter cutoff frequency of the acquired signal by 10%-25%. when ≤ When the signal fluctuation is normal within the j-th sliding window, the sampling time and filtering parameters for the i-th test channel remain unchanged, and the fluctuation trend of subsequent sliding windows is continuously monitored.

5. The semiconductor chip automated testing optimization system according to claim 4, characterized in that, The long-term drift slope The specific method of obtaining it is as follows: Based on the j-th sliding window, the window contains a set of sampling points; , This represents the total number of sampling points; In the formula, Represented as the Unix timestamp of the Pth sampling point. This represents the contact impedance value at the Pth sampling point; Calculate the mean of the contact impedance Unix timestamps within the j-th sliding window. ; ； Calculate the mean value of the contact impedance within the j-th sliding window. ; ； Then, the contact impedance time series data are fitted using the least squares method, and the fitting function is as follows: ； In the formula, The intercept is... This represents the long-term drift slope; Next, the long-term drift slope of the fitted line is calculated according to the least squares criterion. ; 。 6. The semiconductor chip automated testing optimization system according to claim 5, characterized in that, The evaluation unit also includes a long-term drift slope evaluation subunit; The long-term drift slope evaluation subunit is used to preset the long-term drift slope threshold. and the long-term drift slope threshold With long-term drift slope The comparison generates a long-term drift slope assessment instruction, including: when > When this occurs, it indicates that the i-th test channel has a long-term drift anomaly. A long-term drift slope strategy is generated, which includes increasing the pressure of the i-th test channel by 9%-23%, increasing the contact impedance sampling frequency by 28%-37%, and extending the sampling time of the i-th test channel by 28%-44%. when ≤ When the i-th test channel is tested normally, the current sampling frequency is maintained, and the drift slope and contact impedance are continuously monitored.

7. The semiconductor chip automated testing optimization system according to claim 6, characterized in that, The high-frequency noise amplitude within the j-th sliding window The specific method of obtaining it is as follows: Bandpass filtering is applied to the processed time series of electrical variables arranged in chronological order to extract high-frequency components. Commonly used bandpass filter parameters are set according to experimental or system requirements. Based on the data points within the time period corresponding to the j-th sliding window for: ； For the k-th data point in the j-th sliding window, calculate the peak-to-peak value to obtain the high-frequency noise amplitude within the j-th sliding window. .

8. The semiconductor chip automated testing optimization system according to claim 7, characterized in that, The evaluation unit also includes a high-frequency noise evaluation subunit; The high-frequency noise evaluation subunit is used to preset the high-frequency noise threshold. and set the high-frequency noise threshold With high frequency noise amplitude A comparison is performed to generate high-frequency noise evaluation instructions, including: when > When the signal in the j-th sliding window of the i-th test channel is abnormal, a high-frequency noise generation strategy is adopted, which includes lowering the cutoff frequency of the digital low-pass filter by 15%-23% and extending the sampling time of the i-th test channel by 27%-33%. when ≤ When the signal in the j-th sliding window of the i-th test channel is normal, the current sampling time and the current digital filtering parameters are maintained.

9. The semiconductor chip automated testing optimization system according to claim 8, characterized in that, The contact discrimination module includes an association unit and an analysis unit; The associated unit is used to convert the short-term fluctuation amplitude value Long-term drift slope and high frequency noise amplitude The contact resistance coefficient, after being dimensionless, is calculated using the following formula. ; The analysis unit is used to preset the contact impedance threshold. and the contact impedance threshold With contact resistance coefficient The comparison generates a contact resistance assessment instruction, including: when > When the contact state of the i-th test channel is abnormal, a contact impedance strategy is generated, which includes extending the sampling time of the i-th test channel by 34%-56%, increasing the length of the j-th sliding window by 55%-67%, and lowering the digital filter cutoff frequency by 19%-37%. when ≤ When the i-th test channel is in normal contact status, the current sampling frequency and filter plate parameters are maintained, and normal data acquisition and analysis continue to ensure continuous and stable data.

10. The semiconductor chip automated testing optimization system according to claim 9, characterized in that, The data tagging module includes a tagging unit; The marking unit, used for contact impedance evaluation instructions, includes: when < ≤ *1.5 indicates an abnormal contact in the i-th test channel, and the corresponding electrical variable measurement data is marked as low confidence -1; when *1.5< ≤ When *2 is used, it indicates that the i-th test channel has an abnormal contact, and the corresponding electrical variable measurement data is marked as low confidence -2. then > When *2, it indicates that the i-th test channel has an abnormal contact, and the corresponding electrical variable measurement data is marked as low confidence -3.

Images