A method for evaluating bonding quality of semiconductor device bonding wires

Abstract

The present invention relates to the technical field of measuring electrical variables, specifically a method for assessing the bonding quality of semiconductor device bonding wires, comprising the following steps: applying a pulse current to the semiconductor device bonding wire using a resistance pulse measuring device, while recording the instantaneous voltage response of the bonding point to generate voltage response data. The present invention achieves in-depth detection and analysis of the dynamic characteristics of the bonding wire resistance by applying a pulse current and recording the instantaneous voltage response of the bonding point, thereby capturing tiny resistance fluctuations at the bonding connection; based on the dynamic resistance data, further frequency scanning measurement is performed to obtain the detailed variation pattern of resistance at different frequencies, intuitively presenting the resistance stability performance of the bonding wire; subsequently, with the help of precise positioning of the abnormal capacitance area around the bonding point, combined with real-time tracking of the current and voltage response characteristics under dynamic load conditions, the reliability performance of the bonding connection under complex environmental conditions is systematically revealed.

Description

Technical Field

[0001] The present invention relates to the technical field of measuring electrical variables, and in particular to a method for evaluating the bonding quality of a semiconductor device bonding wire. Background Art

[0002] Bond quality assessment of semiconductor device bonding wires refers to a method for evaluating the connection quality of bonding wires in semiconductor devices. Its main purpose is to determine the reliability of the bonding connection by detecting and analyzing the electrical characteristics of the bonding wires, and to promptly detect possible connection defects or failure risks.

[0003] Existing technologies primarily focus on static measurement and simple analysis of electrical variables such as current, voltage, resistance, and capacitance. They fail to delve into the in-depth detection of transient electrical characteristics under dynamic conditions, ignore the changes in the device's electrical response during actual operation under dynamic load conditions, and lack correlation analysis between frequency characteristics and abnormal capacitance areas. This limitation makes it difficult for existing technologies to identify subtle damage within bonding points and potential quality risks, making it difficult to effectively detect and warn of early failure risks or microscopic defects in a timely manner, increasing potential safety hazards and failure risks for semiconductor devices. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a bonding quality assessment method for semiconductor device bonding wires.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: a method for evaluating the bonding quality of a semiconductor device bonding wire, comprising the following steps:

[0006] Applying a pulse current to a semiconductor device bonding wire using a resistance pulse measurement device while simultaneously recording an instantaneous voltage response of the bonding point to generate voltage response data; analyzing a dynamic change in resistance of the bonding point under the action of the pulse current based on the voltage response data to obtain a resistance dynamic characteristic analysis result;

[0007] Based on the resistance dynamic characteristic analysis result, frequency scanning is performed on the curve of the resistance dynamic characteristic analysis result, and the resistance value at different frequencies is measured to generate frequency response data; by evaluating the frequency response data, the resistance stability of the bonding wire is calculated to obtain a resistance stability evaluation result;

[0008] Measure the capacitance around the bonding point, record the capacitance value, identify areas with abnormal capacitance compared to normal bonding, and obtain abnormal capacitance analysis results;

[0009] Based on the abnormal capacitance analysis results, a dynamic load electrical test is applied to record the voltage and current responses under varying load conditions to generate load response data; by analyzing the load response data, the bonding quality is evaluated to obtain a comprehensive bonding quality assessment result.

[0010] Preferably, the step of acquiring the voltage response data is:

[0011] Apply pulse current to the bonding wire of the semiconductor device, record the instantaneous voltage response of the bonding point, obtain the voltage value under each pulse current, and organize the voltage response data to form a voltage response data set;

[0012] Based on the voltage response data set, performing time series analysis on the voltage signal, extracting the maximum voltage value and the minimum voltage value in each pulse cycle, and generating voltage response characteristic data;

[0013] Based on the voltage response characteristic data, the frequency characteristics of the voltage fluctuation are analyzed, the changing trend of the voltage response at different frequencies is identified, and the voltage response data is generated.

[0014] Preferably, the steps of obtaining the resistance dynamic characteristic analysis result are:

[0015] Based on the voltage response data, extract the voltage change sequence under each pulse current, perform time window division, extract the maximum value, minimum value and change amplitude of the voltage in each pulse period, and obtain a processed voltage change data set;

[0016] The dynamic change value of the resistance is calculated based on the processed voltage change data set. The calculation formula is:

[0017] ;

[0018] in, is the dynamic change value of the resistance, is the instantaneous voltage change, is the time interval, is the maximum voltage value within the cycle, is the minimum voltage value within the cycle, is the magnitude of the applied pulse current, is the duration of current action;

[0019] Based on the dynamic change value of the resistance, the change trend of the resistance over time is analyzed, the periodic fluctuation characteristics of the resistance are identified, and the resistance dynamic characteristic analysis result is obtained.

[0020] Preferably, the steps of acquiring the frequency response data are:

[0021] Based on the resistance dynamic characteristic analysis results, multiple frequency intervals are set for scanning to obtain the resistance value corresponding to each frequency point, thereby obtaining frequency scanning result data;

[0022] According to the frequency scanning result data, the variation trend of the resistance at different frequencies is analyzed, and the frequency response data is obtained by comparing the resistance values at different frequencies.

[0023] Preferably, the steps of obtaining the resistance stability evaluation result are:

[0024] Based on the frequency response data, extract the resistance value at each frequency point, analyze the fluctuation of the resistance at each frequency point through Fourier transform, calculate the standard deviation and fluctuation amplitude of the resistance at each frequency, and obtain resistance change rate data;

[0025] According to the resistance change rate and frequency response data, the stability of the resistor is calculated using the following formula:

[0026] ;

[0027] in, is the resistance stability evaluation result, is the resistance value at the qth frequency point, is the average resistance at all frequency points, is the voltage change at the qth frequency point, is the magnitude of the applied pulse current, is the total number of frequency response data, is the change amplitude of the resistance value at the qth frequency point, and are the maximum and minimum resistance values, is the duration of current action;

[0028] Based on the resistance stability evaluation results, the resistance variation trend at different frequencies is analyzed, the frequency range with poor resistance stability is identified, and the resistance stability analysis results are generated.

[0029] Preferably, the steps for obtaining the abnormal capacitance analysis result are:

[0030] Measure the capacitance around the bonding point, record the capacitance value of each measurement point, and record the measurement time point to obtain a capacitance value record list and a time record list;

[0031] Based on the capacitance value record list and the time record list, the change rate of the capacitance value at each point relative to the previous measurement point is calculated using the following formula:

[0032] ;

[0033] in, is the rate of change, For the The capacitance value of the point, is the capacitance value of the previous point, is the standard deviation of the capacitance values of adjacent points, is the time interval, It is a frequency calculation based on time difference, and the calculation method is ;

[0034] Based on the change rate, an area where the change rate exceeds a set threshold is identified, and the area exceeding the set threshold is marked as abnormal, thereby obtaining an abnormal capacitance analysis result.

[0035] Preferably, the steps of acquiring the load response data are:

[0036] Based on the abnormal capacitance analysis results, a load change instruction is executed on the target area, the set resistance value is switched in sequence, and the instantaneous voltage value and current value under each load state are recorded respectively to generate a voltage response record and a current response record;

[0037] Based on the voltage response record and the current response record, the voltage values and the current values collected under each load state are structured and arranged, and time linear alignment is performed according to the load change sequence to generate load response data.

[0038] Preferably, the steps for obtaining the comprehensive evaluation result of the bonding quality are:

[0039] Filter complete cycle sample segments from the load response data, extract the peak voltage, peak current, baseline voltage and baseline current in each segment, and simultaneously record the response lag time after the load switching moment and the time point when the voltage extreme value appears in the cycle to obtain a peak value and timing parameter group;

[0040] Based on the peak value and timing parameter set, the bond integrity index is calculated using the following formula:

[0041] ;

[0042] in, is the bond integrity index, represents the peak voltage in a single cycle load response, represents the peak current in a single cycle load response, represents the single cycle baseline voltage, represents the single cycle baseline current, Indicates the lag time from load switching to the start of voltage response. Indicates the time point when the voltage peak occurs within the cycle;

[0043] Based on the bond integrity index, the bond integrity indices of multiple test cycles are called to construct a comparison matrix, and horizontal interval analysis and vertical trend profile analysis are performed to determine whether there are abnormal sections in all test cycles, thereby forming a comprehensive bond quality assessment result.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are:

[0045] The present invention achieves in-depth detection and analysis of the dynamic characteristics of the resistance of the bonding wire by applying a pulse current and recording the instantaneous voltage response of the bonding point, thereby capturing the tiny resistance fluctuations at the bonding connection; based on the dynamic resistance data, frequency scanning measurement is further implemented to obtain the detailed variation pattern of the resistance at different frequencies, and intuitively present the resistance stability performance of the bonding wire; then, with the help of precise positioning of the abnormal capacitance area around the bonding point, combined with the real-time tracking of the current and voltage response characteristics under dynamic load conditions, the reliability performance of the bonding connection under complex environmental conditions is systematically revealed, the detection capability of bonding defects and hidden dangers is improved, the accuracy and stability of the bonding quality assessment are enhanced, the probability of potential failures is reduced, and the operating reliability and overall performance stability of the semiconductor device are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0048] See also Figure 1 The present invention provides a technical solution, a method for evaluating the bonding quality of a semiconductor device bonding wire, comprising the following steps:

[0049] A resistance pulse measurement device is used to apply a pulse current to the bonding wire of a semiconductor device, while simultaneously recording the instantaneous voltage response of the bonding point to generate voltage response data. Based on the voltage response data, the dynamic change in resistance of the bonding point under the action of the pulse current is analyzed to obtain the resistance dynamic characteristic analysis results.

[0050] Based on the resistance dynamic characteristics analysis results, the curve of the resistance dynamic characteristics analysis results is frequency scanned to measure the resistance values at different frequencies and generate frequency response data; by evaluating the frequency response data, the resistance stability of the bonding wire is calculated to obtain the resistance stability evaluation result;

[0051] Measure the capacitance around the bonding point, record the capacitance value, identify areas with abnormal capacitance compared to normal bonding, and obtain abnormal capacitance analysis results;

[0052] Based on the abnormal capacitance analysis results, dynamic load electrical testing is applied to record the voltage and current responses under varying load conditions to generate load response data. By analyzing the load response data, the bonding quality is evaluated to obtain a comprehensive bonding quality assessment result.

[0053] The steps for obtaining voltage response data are:

[0054] Apply pulse current to the bonding wire of the semiconductor device, record the instantaneous voltage response of the bonding point, obtain the voltage value under each pulse current, and organize the voltage response data to form a voltage response data set;

[0055] Based on the voltage response data set, the voltage signal is analyzed in time series, the maximum voltage value and the minimum voltage value in each pulse cycle are extracted, and the voltage response characteristic data is generated;

[0056] Based on the voltage response characteristic data, the frequency characteristics of voltage fluctuations are analyzed, the changing trend of voltage response at different frequencies is identified, and voltage response data is generated.

[0057] Specifically, based on the steps of applying pulse current to the bonding wire of the semiconductor device, the pulse current range is first set in the test bench to 0.5A to 2A. The range is calculated based on the empirical data of small-batch tests conducted in advance. The situation where the current below 0.5A cannot reflect the complete conduction characteristics and the current above 2A is likely to cause excessive heating of the device is selected as the basis for judgment, and the pulse duration is fixed between 2ms and 10ms. The specific value is measured through testing. A pulse width exceeding 10ms will cause a local temperature rise of more than 30°C, and when it is less than 2ms, the voltage sampling is unstable. According to the above settings, the pulse current output is executed and the instantaneous voltage response of the bonding wire is sampled. The sampling frequency can be set between 1000 and 5000 times per second. 2000 times is used as an example value here and obtained by the control experiment. When the sampling frequency is lower than 1000 times, missed sampling may occur, while when it is higher than 5000 times, data redundancy may occur, resulting in excessive computing resources being occupied in subsequent analysis. A series of data entries are generated by recording the real-time voltage values corresponding to the bond points in each pulse cycle. If the monitored voltage value exceeds 10V, it is marked as an abnormal record. The 10V upper limit is based on the safe data range obtained from the device withstand voltage test (usually it can still maintain normal status at 12V but with a 2V margin reserved). In this process, each voltage value is compared with the valid range of 0V to 10V. The mean of the voltage peak distribution is selected from the ten groups of test results accumulated in the laboratory and verified with the safety range of historical records. Finally, all the measured voltage values are aggregated and sorted in sequence by pulse cycle to form a voltage response data set.

[0058] Based on the voltage response data set, in the process of performing time series analysis on the voltage signal and extracting the maximum and minimum voltage values within each pulse cycle, it is necessary to first determine the division method of a single pulse cycle. The pulse start time is used as the reference point and the current stops as a complete cycle. The cycle length is consistent with the aforementioned pulse duration. Each pulse cycle interval is located by comparing the position of the acquisition timestamp in each voltage response record, and the voltage extreme value is queried in a sequential scanning manner within each interval. Records with a voltage greater than 8V are marked. The threshold of 8V is determined by comparing the 10V safety range collected above and referring to the spike shape during actual measurement. It is estimated to distinguish the difference between transient spikes and sustained increases. If a record higher than 8V appears, it will be determined whether it is greater than 8V for more than two consecutive times. If it is greater than 8V twice in a row, it will be marked as a special extreme value segment. Then, after the entire cycle scan is completed, the maximum voltage value, minimum voltage value and related positioning information will be sorted into a characteristic list. At the same time, these extreme value information will be correlated and compared with the voltage response data set obtained in the previous article, and combined with the timestamps of the start and end points of each cycle for re-verification. If missed sampling or sudden change values occur, the equipment status during the sampling process must be re-evaluated under the premise of confirming the accuracy of the sampling instrument, and finally the voltage response characteristic data will be generated.

[0059] Based on the voltage response characteristic data, in the process of analyzing the frequency characteristics of voltage fluctuations and identifying the voltage response change trend at different frequencies, the complete voltage change sequence is first obtained in the time domain according to the pulse period length confirmed in the previous article, and then the discrete Fourier transform method is used to observe the amplitude distribution of each frequency component. Specifically, the frequency scanning range can be set from 50Hz to 10kHz. This range is obtained by comparing reference materials. If it is lower than 50Hz, it belongs to the extremely low frequency band and is easily confused with the environmental power frequency interference. After exceeding 10kHz, the signal amplitude attenuates significantly. Therefore, the corresponding voltage amplitude is calculated at each discrete frequency point and the amplitude increase or decrease with frequency is obtained. Change list, if it is found that the amplitude at a certain frequency point exceeds 1.2 times the known mean, it will be marked separately. The 1.2 times ratio is a standard determined after multiple measurements of relative fluctuation levels in historical experiments, and a floating step size between 0.1 and 0.3 is introduced to track other frequency points near this frequency point to verify its true amplitude distribution. At the same time, a scanning scheme with a frequency increment of 1 Hz is established to refine the location of the suspicious frequency band. Finally, these analysis records are combined and the amplitude information in each frequency interval is compared with the timing characteristics within the pulse period to obtain a result table containing the correspondence between amplitude and frequency band, and generate voltage response data.

[0060] The steps to obtain the resistance dynamic characteristics analysis results are:

[0061] Based on the voltage response data, the voltage change sequence under each pulse current is extracted, time window division is performed, and the maximum, minimum and change amplitude of the voltage in each pulse cycle are extracted to obtain a processed voltage change data set;

[0062] According to the processed voltage change data set, the dynamic change value of the resistance is calculated using the following formula:

[0063] ;

[0064] in, is the dynamic change value of resistance, is the instantaneous voltage change, is the time interval, is the maximum voltage value within the cycle, is the minimum voltage value within the cycle, is the magnitude of the applied pulse current, is the duration of current action;

[0065] Based on the dynamic change value of resistance, the change trend of resistance over time is analyzed, the periodic fluctuation characteristics of resistance are identified, and the dynamic characteristic analysis results of resistance are obtained.

[0066] Specifically, based on the voltage response data obtained earlier, the pulse signal records are selected and referred to their corresponding timestamps to extract and disassemble each pulse current action segment one by one. First, the start time is calibrated at the beginning of the pulse and the end time is calibrated at the end of the pulse. During this period, the voltage value is inspected in steps of 0.5ms. For example, the voltage is judged whether it is in a safe range by comparing it with the interval of 0V to 10V. If a certain sampled instantaneous voltage value exceeds 10V, it is marked as abnormal. The 10V standard comes from a large number of previous measurements and comparisons combined with the device withstand voltage limit setting. When it is found that three consecutive samples have exceeded 10V, it will be marked as abnormal. Manual verification is performed to check whether there is signal drift in the recording link. The confirmed effective voltage value is compared with the pulse start time to the end time to form a time window. All voltage values in the time window are further compared point by point, and the maximum value, minimum value and the difference between the maximum value and the minimum value are extracted as the variation amplitude. After inspection, the corresponding maximum value, minimum value and variation amplitude in each pulse cycle are sorted out separately. If it is found that a pulse cycle is missed, the corresponding section is supplemented by comparing the timestamp again. The extraction results of all pulse cycles are summarized one by one and the pulse number is used as the index to finally obtain the processed voltage change data set.

[0067] formula: The benefit of the formula is that it introduces the instantaneous voltage change and duration of current action The voltage extreme value difference and time interval are integrated into the numerator and denominator for comprehensive calculation. This can take into account both instantaneous and periodic characteristics when evaluating the resistance state, achieving a three-dimensional characterization of the dynamic change of resistance, thereby obtaining an accurate assessment of the bonding wire state in a wider range of pulse current environments.

[0068] The parameter acquisition steps are as follows: when performing a timed scan of the pulse cycle, the evenly divided interval between the start and end times of each pulse is used as the basic interval, the number of valid voltage samples in each interval is counted, and the single-segment sampling duration is obtained by dividing the pulse duration by the number of samples. Then, the same operation is performed on all pulse cycles and their time intervals are averaged to obtain the duration. For example, if about 30 valid sampling points are collected for each of the 15 pulse cycles in a test, the single segment time is , if the duration of a pulse cycle If the time of a single segment is 10ms, then it is about 0.333ms. After multiple calculations, the average value of 0.350ms is taken as ;

[0069] and The parameter acquisition step is to directly reference the extreme value information in each pulse cycle from the processed voltage change data set obtained above, and use the difference judgment method to mark the most prominent peak in the cycle as , the lowest valley is marked as In actual measurement, through inspection, it can be known that the maximum voltage value is distributed between 5V and 9V. In this example, the maximum value of a certain cycle is 7.2V and the minimum value is 0.8V. Comparison and observation are made in adjacent cycles. Once an abnormal record of the maximum value exceeding 9V or the minimum value being lower than 0V occurs for two consecutive cycles, the overall distribution of the previous sampling is used to check again. In this example, V. V;

[0070] The parameter acquisition step is to measure the current value of each pulse in real time when the pulse current is output, and select the current mean or median value of multiple pulses to define If the single test is within the pre-set range of 0.5A to 1.5A, 1.0A is selected as the pulse current intensity for this stage based on the previous hardware capability assessment. In this example, A;

[0071] The parameter acquisition step is to set the duration of the pulse generator directly from the hardware, which usually varies from 1ms to 10ms. In this example, ms is the duration of continuous pulse action;

[0072] Calculation process:

[0073] The first step is to bring in the known values V. ms, V. V. A. ms;

[0074] The second step is to calculate the numerator first. ,in Need to be converted into seconds for calculation, here s, that is V·s, plus V, the numerator is approximately ;

[0075] Step 3: Calculate the denominator ,in A·s, V·s, the sum of the denominators is ;

[0076] Step 4: Divide the numerator and denominator to get In this scenario, if the unit is ohm, the result can be regarded as 2245.69Ω. However, due to the dimensional combination of parameters such as voltage, current, and time in their specific expression, it is necessary to re-verify it in combination with the equivalent conversion method defined earlier and indicate it in the analysis report;

[0077] The results show that under the sample parameters, the dynamic change in resistance within the pulse cycle is 2245.69Ω. If this value is higher than 2000Ω, it indicates that the device has relatively high fluctuations or unstable internal contact during this cycle. If it is lower than 500Ω, it indicates that the contact state is relatively stable. The above value range is refined from more than 50 measurements and can be used as a basis for subsequent judgment and screening.

[0078] Based on the dynamic change value of resistance, the corresponding DR value is collected after each pulse cycle and recorded in a set of sequences. A step-by-step time series comparison is performed on these sequences. First, the cycles are numbered from 1 to N on the time axis and arranged in sequence. The DR value corresponding to each number is compared with the reference range of 0Ω to 5000Ω. If the DR values of two consecutive numbers exceed 3000Ω, the hardware operation record is checked and the pulse current intensity and duration mentioned above are compared to confirm whether there is overcurrent or abnormal ambient temperature. If the DR value is between 200Ω and 1500Ω, it is temporarily identified as a normal section and repeated detection is repeated for five cycles for tracking. If the DR value is between 15 The range between 100Ω and 3000Ω is set as the key focus segment and the corresponding pulse current size is compared again with the actual current record collected. The DR value corresponding to each cycle and the actual running time of the detected pulse are summarized together to form a fluctuation trend sequence, and the fluctuation trend sequence is superimposed in the time domain to confirm whether there is a stage-by-stage high phenomenon. For example, from the 10th to the 15th cycle, it may be continuously between 2500Ω and 2800Ω. At this time, it is necessary to trace the DR value distribution in the previous corresponding cycle. If continuous growth occurs in this segment, it is marked as an abnormal segment and the resistance contact quality is manually checked. Finally, after comparing the complete time series, the dynamic characteristic analysis results of the resistance are obtained.

[0079] The steps to obtain frequency response data are:

[0080] Based on the analysis results of the dynamic characteristics of the resistor, multiple frequency intervals are set for scanning to obtain the resistance value corresponding to each frequency point and obtain the frequency scanning result data;

[0081] Based on the frequency scanning result data, the changing trend of resistance at different frequencies is analyzed, and the frequency response data is obtained by comparing the resistance values at different frequencies.

[0082] Specifically, based on the results of the dynamic characteristics analysis of the resistor, multiple frequency intervals are set for scanning to obtain the resistance value corresponding to each frequency point. First, it is necessary to clarify the starting and ending ends of the frequency under experimental conditions. For example, the starting frequency is set at 50Hz, the ending frequency is set at 10kHz, and the frequency increments are 100Hz between 50Hz and 10kHz. These values are obtained after summarizing and analyzing the bandwidth distribution of the pulse signal collected previously and the possible resonance points of the device in the low-frequency to high-frequency range. If there is not enough data or hardware accumulated for the performance of the device in the ultra-high frequency band in the early stage, If the scanning capability of more than 10kHz is not available, a preliminary exploration can be conducted at a larger interval of 0.1kHz or 1kHz, and then the scanning range can be narrowed down according to the concentration of the monitored resistance fluctuation distribution. During the process, the observed resistance value needs to be compared with the pre-established safety interval at each scanning frequency point. For example, if the safety interval is set to 0Ω to 2kΩ and the rated range of the device itself is between 0Ω and 3kΩ, the records in the data that exceed 2kΩ but are less than 3kΩ can be defined as moderate anomalies and manually proofread. At the same time, the records that exceed 3kΩ can be defined as major anomalies and listed separately. The boundary value between moderate anomaly and major anomaly is set based on the data of more than 30 repeated experiments and the statistical results of device overcurrent failures. During the scanning process, the current and voltage amplitude at each frequency point must be recorded to assist in verifying whether the high resistance value comes from excessive current or unexpected vibration effects. If it is found that the resistance of multiple consecutive frequency points exceeds 2kΩ, cross-check the resistance dynamic characteristic analysis results obtained previously to confirm whether there is a deviation caused by the pulse duration or high ambient temperature. If a significant increase in resistance is observed at a certain frequency point, a more detailed frequency step is performed again with that point as the center. The scan is then performed, with the step size reduced from 100Hz to 10Hz and point-by-point comparisons performed above and below 50Hz. During this period, the resistance values at adjacent frequencies are differentially calculated and summarized in a temporary record sequence. It is then necessary to check whether periods of continuously increasing differences correspond to fluctuations in the same batch of devices or in the same time period due to internal ambient temperature fluctuations. If significant increases are still shown after repeated checks, an independent abnormality flag is established and the relevant current and voltage data are reviewed simultaneously. Finally, all scanned resistance data are arranged in ascending frequency order and recorded in a frequency-resistance comparison table to obtain the frequency scan result data.

[0083] According to the frequency scanning result data, the changing trend of resistance at different frequencies is analyzed. By comparing the resistance values at different frequencies, it is necessary to first find the resistance increase or decrease between adjacent frequency points from the frequency-resistance comparison table obtained earlier and summarize them. Whenever the resistance difference between two consecutive frequency points exceeds 500Ω, they should be marked separately. The 500Ω threshold is established with reference to multiple batches of experimental data before. If more than three consecutive increases or decreases cross this threshold, the spectrum distribution is locally amplified to further confirm whether the device has abnormal fluctuations in the corresponding frequency band. Then all frequency points above 2000Ω should be separated and matched with the previously recorded resistance dynamic characteristics analysis results to see whether they show concentrated growth in a certain range from low frequency to high frequency. In order to distinguish background noise or equipment errors, The measured voltage is synchronously compared with the actual current reference value. For example, when testing pulse current in the range of 0.5A to 1.5A, if the measured value deviates significantly from a certain level above 1.0A, it is necessary to recheck the timing of the pulse generation and the power output condition to avoid misjudging the fluctuation caused by unstable power supply as unstable resistance. If the data is confirmed to be normal, the frequency bands in the high resistance range are sorted and a digital record of the increase and decrease trends is attached. For some small fluctuations between 1kHz and 2kHz, the ambient temperature, humidity or other external factors can be combined in the next analysis to verify whether there is any interaction. After completing these statistics, the comparison results of the frequency band and the resistance value are integrated and summarized into a frequency comparison table based on the highest, lowest and average resistance values corresponding to each scanning frequency point to obtain the frequency response data.

[0084] The steps to obtain the resistance stability evaluation results are:

[0085] Based on the frequency response data, the resistance value at each frequency point is extracted, and the fluctuation of the resistance at each frequency point is analyzed by Fourier transform. The standard deviation and fluctuation amplitude of the resistance at each frequency are calculated to obtain the resistance change rate data;

[0086] According to the resistance change rate and frequency response data, the stability of the resistor is calculated using the following formula:

[0087] ;

[0088] in, is the resistance stability evaluation result, is the resistance value at the qth frequency point, is the average resistance at all frequency points, is the voltage change at the qth frequency point, is the magnitude of the applied pulse current, is the total number of frequency response data, is the change amplitude of the resistance value at the qth frequency point, and are the maximum and minimum resistance values, is the duration of current action;

[0089] Based on the resistance stability evaluation results, the resistance variation trend at different frequencies is analyzed, the frequency range with poor resistance stability is identified, and the resistance stability analysis results are generated.

[0090] Specifically, based on the frequency response data obtained previously, first select the resistance value of each frequency point in the complete measurement cycle and list it, then pair the sampling time label and instantaneous voltage record with the same number for the resistance value corresponding to each frequency point, and determine the time domain fluctuation of the resistance value at the frequency by traversing point by point. Then, use Fourier transform to map the time domain fluctuation to the frequency domain form. In the process, the time series is divided according to the number of sampling points and the corresponding frequency components are matched to ensure that the frequency resolution is in a reasonable range of 0.1Hz to 1Hz. The setting of this range refers to the frequency step strategy obtained previously. If no obvious amplitude anomaly can be located in a smaller frequency band, the step is selected to be 1Hz. Otherwise, it can be reduced to 0.1Hz for finer-grained analysis. At the same time, the resistance value corresponding to the frequency band is compared point by point with the valid range established previously. For example, it is judged whether it exceeds the standard between 0Ω and 3000Ω, and all Records exceeding 3000Ω are marked as outliers. This method confirms the abnormal distribution of resistance values at each frequency point. Amplitude calculations are then used to calculate the standard deviation and fluctuation range of the resistance values. The standard deviation can be statistically determined within a given batch. The resistance measurements at each frequency at different observation times are compared against their means, and the sum of the squared deviations from the mean is calculated. This sum is then divided by the number of measurements and the square root is taken. If the standard deviation at a frequency point is greater than 100Ω, it is given special attention in subsequent analysis. The 100Ω threshold is set based on the summary statistics of multiple batches of measurement records. In this statistics, the standard deviations of adjacent frequency bands are compared, and it is found that a value below 100Ω indicates relative stability, while a value above 100Ω indicates more severe fluctuations. Similarly, the calculation of the fluctuation range also requires the introduction of the difference between the maximum and minimum values for further processing. These results are summed up to form the resistance change rate data.

[0091] formula: The benefit of the formula is that it can achieve a more comprehensive quantitative evaluation of the resistance stability by comprehensively using the average resistance value and the voltage change at each frequency point, combined with multiple parameters such as current size and resistance change amplitude, especially the use of logarithmic function. To characterize the impact of extreme value ratio on system stability, this ratio factor is combined with other linear terms, thus taking into account both the overall distribution and individual extreme fluctuations in the evaluation process.

[0092] The steps to obtain the parameters are as follows: In the resistance sequence obtained by the frequency scan above, number them in order from small to large frequency. , will The resistance corresponding to the frequency point is marked as , which is derived from actual measurements and obtained through the Fourier transform mentioned above, each All of the steps must be followed to remove invalid sampling points. If the resistance measurement value corresponding to the 10th frequency point in a certain test is 1200Ω, it can be recorded as Ω.

[0093] The steps to obtain the parameters are: In the same test batch, the cumulative sum is divided by the total quantity , the obtained value is the average resistance , during the acquisition process, it is necessary to ensure that all Have a consistent measurement basis and exclude extreme outliers, for example, by measuring 50 frequency points in a sample After summing up, the cumulative result is 58000Ω, then Ω.

[0094] The parameter acquisition steps are: at each frequency point, the instantaneous voltage change is recorded and sampled multiple times, and then the voltage differences after these samples are weighted or averaged to form the final , it is necessary to count the voltage change information collected at each frequency point in the early stage, for example, At a frequency point, if the actual measured voltage fluctuates within the range of 1.0V to 1.5V, a final voltage change can be obtained by performing differential integration on the continuous waveforms in this interval. If the final result is 0.40V, it is recorded as V.

[0095] The steps for obtaining the parameters are as follows: obtain the current size through the real-time detection record of the pulse current obtained previously, measure each pulse stage separately, and then take a stable segment or average value from all the pulse data to define For example, if multiple sets of measurements are performed within the range of 0.5A to 2A allowed by the device, and the final value of the pulse current applied this time is 1.2A, then in the subsequent calculation formula, A is brought in.

[0096] The steps to obtain the parameters are: the total number of frequency response data, that is, the number of frequency points actually scanned within the frequency range. For example, in the range of 50Hz to 10kHz, if it is set to scan once every 200Hz and 50 valid frequency measurements are actually completed after verification, then .

[0097] The steps to obtain the parameters are: Frequency points are measured to investigate the change of resistance value in different measurement periods, and the difference between the maximum value and the minimum value is recorded as For example, if the maximum resistance value is 1250Ω and the minimum resistance value is 1200Ω at the same frequency point within 6 seconds, then Ω.

[0098] and The steps to obtain the parameters are as follows: in the entire frequency scanning range, traverse the resistance measurement values corresponding to all frequency points and find the maximum and minimum values, and mark the maximum value as The smallest one is marked as For example, the resistance values at 50 frequency points range from 1000Ω to 1600Ω. If the maximum value is 1600Ω, then Ω, if the minimum value is 950Ω then Ω.

[0099] The parameters are obtained by the following steps: the duration of the pulse current is determined by the device before it is applied, and the precise duration is obtained by the timer and recorded in the measurement description. For example, if the pulse duration is 5ms in this batch of tests, then ms.

[0100] Calculation process:

[0101] The first step is to substitute the above parameter values into the formula, for example, 、 Ω、 Ω、 V. A. Ω、 Ω、 Ω、 ms;

[0102] The second step is to calculate the sub-items in the formula: Right now , here Ω, multiply by 0.40 to get 16.0 (V·Ω), and then divide by 1.2A to get approximately 13.33 (V·Ω / A);

[0103] Step 3: Calculate Right now , first The natural logarithm is approximately equal to 0.525, multiplied by 50 to get 26.25, and then divided by 0.005 (i.e. 5ms is converted to 0.005s) to equal 5250;

[0104] Step 4: For all From 1 to Perform similar operations on the values and then find the average, add up the results and divide by And add the last item to get , if only one value is shown in the example If we add it together with the other 49 similar results and divide it by 50, we can get the overall value;

[0105] The results show that larger values mean lower resistance stability. If the result exceeds 2000, it means that there are significant fluctuations within the current frequency scanning range. If it is less than 500, it can be considered that the performance of this batch of resistors is relatively stable. In different test environments or different device types, the threshold value also needs to be adjusted and recorded with reference to historical measurement data.

[0106] Based on the resistance stability evaluation results, first sort out the stability values corresponding to each frequency point and summarize them in sections according to the frequency span of 50Hz to 10kHz. For example, 50Hz to 500Hz is divided into low frequency band, 500Hz to 2kHz is divided into medium frequency band, and 2kHz to 10kHz is divided into high frequency band. The average and peak values of the stability values are counted within each band, and then compared with the predetermined stability threshold range. For example, if the value is greater than 2500 as an unstable range, below 2500 and above 800 as a suspicious range, and below 800 as a relatively stable range, the frequency range with a value exceeding 2500 is marked as an unstable point and reversely queried through the inspection record to confirm the previous resistance fluctuation rate and instantaneous voltage. Check whether there are obvious anomalies in the voltage difference, and count the number of frequency bands in the suspicious range to see whether they appear in clusters in the low-frequency band or the high-frequency band. If multiple adjacent frequency points are found to appear in the unstable range or the suspicious range, perform manual review and refer to the actual amplitude of the applied pulse current. If similar results appear in multiple acquisitions between 0.5A and 2A, it can be said that this batch of devices has significant resistance changes in this frequency range. List these frequency points in the unstable or suspicious range and their corresponding fluctuation information, and then combine them with other experimental conditions such as ambient temperature and electrode material batch number for re-checking. Finally, summarize the frequency ranges with poor resistance stability and identify specific ranges that need attention to generate resistance stability analysis results.

[0107] The steps to obtain abnormal capacitance analysis results are as follows:

[0108] Measure the capacitance around the bonding point, record the capacitance value of each measurement point, and record the measurement time point to obtain a capacitance value record list and a time record list;

[0109] Based on the capacitance value record list and time record list, calculate the rate of change of the capacitance value at each point relative to the previous measurement point. The calculation formula is:

[0110] ;

[0111] in, is the rate of change, For the The capacitance value of the point, is the capacitance value of the previous point, is the standard deviation of the capacitance values of adjacent points, is the time interval, It is a frequency calculation based on time difference, and the calculation method is ;

[0112] Based on the rate of change, the area where the rate of change exceeds a set threshold is identified, and the area exceeding the set threshold is marked as abnormal, thereby obtaining an abnormal capacitance analysis result.

[0113] Specifically, to measure the capacitance around the bonding point, it is first necessary to select several discrete measurement points around the pad based on the previously defined measurement object and spatial position. Each measurement point needs to be marked with a reference to determine its relative position. A capacitance meter is used to collect capacitance data for each point, and the range is pre-set according to the effective range of 0.01pF to 100pF. This range is obtained by statistically analyzing the capacitance measurement data of thirty batches of similar processes in the early stage. The minimum value is approximately between 0.02pF and 0.03pF, and the maximum value is distributed in the range of tens of pF. The range is covered to 100pF and a safety margin is reserved. The measurement frequency of each measurement point is then set to once every two seconds and the corresponding timestamp is recorded. When performing continuous measurements, if the capacitance of a certain point exceeds 80pF, it is viewed separately as a high-value mark. 80pF comes from the significantly high range found in multiple experiments. If the capacitance is lower than 0.02pF, further inspection is performed on the connection. The system checks whether the wiring or probe contact is stable. By inspecting each measurement record, the capacitance value is compared with the valid range of 0pF to 100pF one by one, and all qualified measurement values are automatically summarized in a capacitance record table. Each capacitance value is marked with a second-level timestamp and a ternary index structure with the measurement point number is established. In the event of a sudden increase or decrease in capacitance value, the calibration status of the measuring instrument is checked to determine whether the instrument has drifted. If the instrument is normal, it is confirmed that the capacitance of the measurement point has indeed changed significantly in the current period. When dealing with large-scale fluctuations, each measurement result is compared with the average of the previous ten times. If the difference is greater than 5pF, it is stored in a single column. 5pF is the cutoff value estimated after observing more than 70 fluctuation curves and obtained through multiple comparisons. After multiple cycles of measurement, a continuous capacitance value sequence and the corresponding measurement time point sequence are finally obtained for all measurement points. They are arranged in the order of measurement, resulting in a capacitance value record list and a time record list.

[0114] formula: The benefit of this formula is that it incorporates both the capacitance difference and the frequency element derived from the time interval into the evaluation process of the rate of change. Through the square root structure, the change in capacitance between consecutive measurement points and the time domain frequency component are quantified in parallel. This not only focuses on the adjacent differences in the capacitance values themselves, but also amplifies and measures the frequency factor generated over time, thereby taking into account both the amplitude mutation and the change rhythm when measuring capacitance anomalies.

[0115] The steps for obtaining the parameters are as follows: the capacitance value recorded at each measurement is read point by point and the ambient temperature is kept relatively constant during the same period. The availability and consistency of the data are ensured by repeated measurements. If the single-point measurement result is 15.5pF, it is recorded as pF.

[0116] The steps to obtain the parameters are: corresponding to the current point The capacitance value of the previous measurement can be directly found by the same sequence order as before. The capacitance value of each sample, for example, at the tenth measurement pF, then in the calculation hour pF, pF will be brought into the calculation.

[0117] The steps to obtain the parameters are: take the statistical standard deviation of the capacitance value difference between every two adjacent measurement points, in order to obtain , first pair all available measurement points, Form a difference list, and then use the standard deviation formula to process the difference list. For example, 29 adjacent differences are formed under 30 measurements, and these differences form a set ,in , the standard deviation can be calculated by reference , is the mean of the difference, and the final statistical result is pF.

[0118] The steps to obtain the parameters are: at each measurement point After collecting the capacitance value, record the time when it occurs and the same as the previous measurement point Subtract the time from the , according to the pre-set measurement frequency and measurement duration, if a measurement is taken every two seconds and the The first measurement occurs at 40 seconds, so The first measurement occurred at 38 seconds, s.

[0119] The steps to obtain the parameters are: The calculated frequency components are obtained by To convert, if s =0.5Hz, and the same method can be used for all measurement periods to obtain sequence.

[0120] Calculation process:

[0121] The first step is to select a measurement sequence, such as pF, pF, s and known pF, then pF;

[0122] The second step is to calculate , and then square it to get ;

[0123] Step 3, Hz, will , and square it to get ;

[0124] Step 4. , perform the same operation on other measurement points in the sequence and summarize all ;

[0125] The results show that It is approximately 0.604. When the value is greater than 1.0, it can be regarded as a drastic change in capacitance. If it is in the range of 0.2 to 1.0, it is regarded as a moderate fluctuation. Below 0.2 is a stable range. These judgment limits are refined by comparing dozens of similar measurement reports and determining the rationality of the values. Using different thresholds can also be applied to environments with higher or lower sensitivity to capacitance changes.

[0126] Based on the rate of change, first calculate the Perform continuous scanning and generate The sequence is arranged in increasing order of measurement, and then a pre-established threshold is selected to measure whether the rate of change is too large. For example, the threshold formed by 200 measurement points is When summarizing the series, set the threshold to 1.0 and all values above 1.0 will be The value is marked as obviously abnormal. The value of 1.0 is set by comparing the device test results for 20 times. When the value is greater than 1.0, the capacitance difference increases significantly and is often accompanied by multiple fluctuations at short time intervals. All records between 0.2 and 1.0 are considered as potential abnormal segments and require subsequent manual confirmation or extended monitoring. 0.2 is also based on the summary of multiple measurements in the stable phase. The measurement points or sections that meet all the abnormal conditions are collected to form an abnormal identification list. Finally, these abnormal measurement points or sections are summarized and processed to obtain the abnormal capacitance analysis results.

[0127] The steps to obtain load response data are:

[0128] Based on the abnormal capacitance analysis results, the load change command is executed on the target area, the set resistance value is switched in sequence, and the instantaneous voltage and current values under each load state are recorded respectively to generate voltage response records and current response records;

[0129] Based on the voltage response records and current response records, the voltage and current values collected under each load state are structured and organized, and time linear alignment is performed according to the load change sequence to generate load response data.

[0130] Specifically, to execute the load change instruction on the target area, it is first necessary to refer to the positions marked as abnormal or potentially abnormal in the abnormal capacitance analysis results obtained previously and organize the identification numbers corresponding to these positions, and then issue the load switching instruction to the target area in the order of the numbers on the experimental platform. The specific method is to pre-select several resistance values in the range of 0Ω to 1kΩ as candidate parameters, such as switching at several typical values such as 0Ω, 50Ω, 100Ω, 200Ω, 500Ω and 1kΩ, and confirm that the laboratory temperature is stable between 20℃ and 25℃ and the humidity is between 40% and 60% before switching. If it is detected that the temperature shows a trend of being higher than 25℃, environmental treatment is first performed to prevent thermal interference, and then the resistance value is switched step by step from small to large. After each switch is completed, the voltage and current are recorded instantaneously. The voltage monitoring range can be set between 0V and 24V and is consistent with the previous analysis of the rated voltage limit of the device. If the voltage is found to exceed 20V, it is marked as a high value and recorded with the historical high value. For comparison, the current monitoring range can be set between 0A and 5A. If the actual sampled value exceeds 3A, it is marked as high and repeated readings are performed for confirmation. All these monitoring boundaries are set based on safety limits or common ranges from multiple previous measurements. For example, in the statistical data of thirty tests, the range of 0V to 24V can cover more than 80% of the sampled values. The 3A threshold for high voltage is based on the rated continuous current tolerance of this type of device. The instantaneous voltage and current after each switching are scanned point by point and the timing difference between adjacent sampling moments is recorded to generate voltage and current response records. If a set of records contains a voltage difference greater than 5V between two consecutive samples with a sampling interval less than 10ms, it is marked as an abnormal jump. The 5V value is based on the laboratory's observation that most normal jumps remain around 3V during repeated testing, with a certain margin set. After all resistance value switching steps are completed, multiple sets of voltage and current response records corresponding to different load conditions are generated.

[0131] Based on the voltage response record and the current response record, it is necessary to first compare the sampling timestamps under each load state and splice them on the same timeline according to the order of load switching. To this end, the experiment start time of each record is read first and the execution time point of the load switching is determined. These time points are sorted in ascending order to form a load switching sequence table. For example, if six resistance values have been scheduled to be switched in ascending order between 0Ω and 1kΩ, it is necessary to mark that the 0Ω switch occurs at the 5th second, the 50Ω switch occurs at the 10th second, and so on. The sampling data with an interval of less than 50ms is regarded as the same load switching stage and the sampling data are concentrated in the same paragraph in the form of a queue. If the interval corresponding to a resistance switching is greater than 1s, the experimental records need to be compared again to check whether there is human intervention or instrument delay during this period. The interval between 50ms and 1s is set here mainly based on the circuit It is defined by the duration of response fluctuations that may occur when switching loads. For example, in more than 60 similar experiments, it was found that most switching actions would be completed within 100ms, and 1s was used as the limit check value. Subsequently, in each load switching interval, the collected voltage and current data were linearly aligned in the order of timestamps, and overlapping values that were read repeatedly at the same time were excluded, and the missing samples were marked. If the voltage or current in a certain instantaneous data differed by more than 10 times from the previous and next sampling points, the suspected impact point was marked and added to a separate sequence. The 10-fold difference comes from the statistical evaluation of the dynamic range of the device. Under experimental conditions where the voltage usually does not exceed 24V and the current does not exceed 5A, if an extreme jump occurs, it often means that the line or contact is abnormal. After all the processed data segments are pieced together one by one, a complete sequence can be formed on the time axis according to the load switching order to generate load response data.

[0132] The steps to obtain the comprehensive evaluation results of bonding quality are as follows:

[0133] Filter complete cycle sample segments from the load response data, extract the peak voltage, peak current, baseline voltage, and baseline current within each segment, and simultaneously record the response lag time after the load switching moment and the time point when the voltage extreme value appears in the cycle to obtain the peak value and timing parameter group;

[0134] Based on the peak and timing parameter set, the bond integrity index is calculated using the following formula:

[0135] ;

[0136] in, is the bond integrity index, represents the peak voltage in a single cycle load response, represents the peak current in a single cycle load response, represents the single cycle baseline voltage, represents the single cycle baseline current, Indicates the lag time from load switching to the start of voltage response. Indicates the time point when the voltage peak occurs within the cycle;

[0137] Based on the bond integrity index, the bond integrity index of multiple test cycles is called to construct a comparison matrix, and horizontal interval analysis and vertical trend profile analysis are performed to determine whether there are abnormal sections in all test cycles, forming a comprehensive assessment result of the bond quality.

[0138] Specifically, to filter complete cycle sample segments from the load response data, it is first necessary to compare the time axis of the previously recorded voltage and current sequences, and determine whether each cycle forms a complete closed loop between the start and end times. Then, each identified cycle is individually checked to see whether it contains the identification information of the load switching moment. If a certain cycle lacks data at the switching moment or the sampling is interrupted, the cycle is excluded and not included in the complete cycle category. Subsequently, all sampling points of voltage and current are traversed within each filtered cycle and compared one by one with the 0V to 24V range and the 0A to 5A range as the reference range. If the voltage or current values of some sampling points are found to be outside this range, they are marked as abnormal and the timing difference between the adjacent samples is checked. If the timing difference is less than 10ms and exceeds the pre-established threshold range three times in a row, the cycle is listed separately for manual confirmation. The source of this threshold is based on the statistics of multiple batch experiments, which show that the voltage and current of most normal cycles are concentrated within 24V and 5A and are large. The amplitude change usually occurs when the interval between two samples does not exceed 10ms. Therefore, this range is confirmed for rapid troubleshooting of abnormal jumps. After confirming that there is no abnormality in the cycle, the voltage is scanned point by point within the cycle and the peak point is recorded. That is, the time corresponding to the maximum voltage and the corresponding specific voltage value are found. If the peak occurs when the voltage is the same for multiple samples, the first time point at which the voltage is reached is selected to maintain uniqueness in subsequent processing. Next, the peak current of the current cycle is determined in the same way and the value and the time of occurrence are archived. Based on this pattern, the baseline is extracted for the more stable part of the voltage and current series. After comparing several adjacent samples, the voltage is selected as the baseline segment where the fluctuation does not exceed 0.2V within 1V and the duration is greater than 5ms. Similarly, the current is less than 0.1A and the segment where the stability does not exceed ±0.02A is the current baseline segment. These numerical thresholds are based on the average level extracted after multiple rounds of experiments and take into account a certain degree of redundancy. If the baseline voltage is successfully determined and baseline current Once the peak and baseline extraction is completed, the response lag time after the load switching moment is calculated. The duration from the switching command triggering to the first voltage jump is recorded as The sampling time when the voltage peak occurs in this cycle is defined as , continue to perform the same operation to traverse all complete cycles, and then summarize the peak value and timing parameter group.

[0139] formula: The benefit of the formula is that it incorporates the peak ratio of voltage to current, the relative relationship between baseline voltage and current, and the difference between lag time and peak occurrence time point into the measurement at the same time, thereby coupling the multi-dimensional factors of amplitude magnitude and response timing in the same expression, taking into account both the extreme values in amplitude and the delay characteristics in the time domain, and balancing different dimensions by using methods such as squaring, square root and piecewise multiplication, so that the formula can comprehensively measure instantaneous impacts and take into account slow changes in overall response when evaluating bonding quality.

[0140] The parameter acquisition steps are as follows: lock the highest voltage value in a single cycle through the peak voltage list obtained previously. For example, if the maximum voltage observed in a certain cycle is 12.8V, it is recorded as V.

[0141] The steps to obtain the parameters are as follows: corresponding to the peak current in a single cycle, find the maximum current value in the cycle in the current sequence previously screened. If multiple extreme values appear in parallel, select the one that appears earliest in time as the peak current. , in a set of tests, if the peak current in a certain cycle is measured to be 3.5A, then A.

[0142] The parameter acquisition steps are as follows: take the section with a longer duration in a single cycle and a voltage fluctuation less than the preset threshold value of 0.2V as the baseline section, record the stable voltage value in this section and take the average value to define it as For example, if the voltage fluctuates by 0.1V from 10ms to 15ms during the acquisition, then these sample values are accumulated and divided by the corresponding number of samples to obtain an average value of 1.05V, that is, V.

[0143] The steps to obtain the parameters are: Similarly, the current value in a single cycle remains stable and the change does not exceed 0.02A. The current sampling value of the segment is averaged and determined as For example, in a certain period of time, the current is stable at 0.05A with a range of 0.01A between 20ms and 30ms. The average value is about 0.06A, which can be expressed as A.

[0144] The steps for obtaining the parameters are as follows: the time from the issuance of the load switching operation instruction to the detection of the significant change in voltage is called the hysteresis time, which needs to be calculated by comparing the timestamp of the switching moment with the timestamp of the initial moment of the voltage jump. For example, if the switching command is issued at 5.000s and the voltage rises and falls significantly at 5.003s, then ms.

[0145] The parameters are obtained by: the time point when the voltage peak occurs in the cycle, and the difference between the cycle interval or the start time form For example, when the reference starting point of the cycle is 0ms, if the maximum voltage appears at 15ms, then ms.

[0146] Calculation process:

[0147] The first step is to bring in the measured data from a sample cycle: Assume V. A. V. A. ms, ms;

[0148] The third step is to calculate the numerator first. ,Right now ,in 、 , the first term in the numerator is , the second item , the sum of the numerators is approximately ;

[0149] Step 4: Calculate the denominator ,in ,then , so the denominator ;

[0150] Step 5. ;

[0151] The results show that when When the value is about 0.81, it means that the relationship between the peak voltage and peak current in this cycle and the comprehensive evaluation value brought by the response timing are at a medium level. The results are compared with each other and the high value interval is set as greater than 2.0 and the low value interval is set as less than 0.5. In subsequent screening or statistics, the periods exceeding 2.0 can be paid special attention, and the periods below 0.5 can be regarded as a relatively stable state for further observation.

[0152] Based on the bond integrity index, the results from multiple test cycles are collated. The values are generated in chronological order or in test batch order to generate a two-dimensional matrix. Each row can represent a cycle number and each column represents a test condition label. Put it into the corresponding position of the matrix, and then select several intervals in the horizontal dimension to determine whether there are too high or too low values. For example, use 0.5 and 2.0 as the dividing line to distinguish different segments. When most cycles in a segment If all values are higher than 2.0, they are marked as potential anomalies and compared with adjacent segments to check their evolution trend. In the vertical dimension, it can be traced whether each cycle appears continuously. The rise or fall phenomenon is checked and the peak distribution of voltage and current mentioned above is inspected. If a large range of rise or fall is confirmed, the corresponding segment is listed as the object of attention and the specific cycle number is indexed for verification. After completing all horizontal and vertical cross-analysis, the marked segment columns are summarized together and the final conclusion on whether there is an abnormality is output in the order of cycle time to form a comprehensive assessment result of the bonding quality.

[0153] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for evaluating the bonding quality of a semiconductor device bonding wire, characterized in that: The following steps are involved: Use a resistance pulse measurement device to apply a pulse current to the bonding wire of a semiconductor device and simultaneously record the instantaneous voltage response of the bonding point to generate voltage response data; Analyzing the dynamic change of resistance of the bonding point under the action of the pulse current according to the voltage response data to obtain a resistance dynamic characteristic analysis result; Based on the resistance dynamic characteristic analysis result, frequency scanning is performed on the curve of the resistance dynamic characteristic analysis result, and the resistance value at different frequencies is measured to generate frequency response data; by evaluating the frequency response data, the resistance stability of the bonding wire is calculated to obtain a resistance stability evaluation result; Measure the capacitance around the bonding point, record the capacitance value, identify areas with abnormal capacitance compared to normal bonding, and obtain abnormal capacitance analysis results; Based on the abnormal capacitance analysis results, a dynamic load electrical test is applied to record voltage and current responses under varying load conditions to generate load response data; and bonding quality is evaluated by analyzing the load response data to obtain a comprehensive bonding quality assessment result. The steps for obtaining the comprehensive evaluation results of bonding quality are as follows: Filter complete cycle sample segments from the load response data, extract the peak voltage, peak current, baseline voltage and baseline current in each segment, and simultaneously record the response lag time after the load switching moment and the time point when the voltage extreme value appears in the cycle to obtain a peak value and timing parameter group; Based on the peak value and timing parameter set, the bond integrity index is calculated using the following formula: ； in, is the bond integrity index, represents the peak voltage in a single cycle load response, represents the peak current in a single cycle load response, represents the single cycle baseline voltage, represents the single cycle baseline current, Indicates the lag time from load switching to the start of voltage response. Indicates the time point when the voltage peak occurs within the cycle; Based on the bond integrity index, the bond integrity indices of multiple test cycles are called to construct a comparison matrix, and horizontal interval analysis and vertical trend profile analysis are performed to determine whether there are abnormal sections in all test cycles, thereby forming a comprehensive bond quality assessment result.

2. The bonding quality assessment method of a semiconductor device bonding wire according to claim 1, wherein: The steps for acquiring the voltage response data are: Apply pulse current to the bonding wire of the semiconductor device, record the instantaneous voltage response of the bonding point, obtain the voltage value under each pulse current, and organize the voltage response data to form a voltage response data set; Based on the voltage response data set, performing time series analysis on the voltage signal, extracting the maximum voltage value and the minimum voltage value in each pulse cycle, and generating voltage response characteristic data; Based on the voltage response characteristic data, the frequency characteristics of the voltage fluctuation are analyzed, the changing trend of the voltage response at different frequencies is identified, and the voltage response data is generated.

3. The bonding quality assessment method of a semiconductor device bonding wire according to claim 1, wherein: The steps for obtaining the resistance dynamic characteristic analysis results are: Based on the voltage response data, extract the voltage change sequence under each pulse current, perform time window division, extract the maximum value, minimum value and change amplitude of the voltage in each pulse period, and obtain a processed voltage change data set; The dynamic change value of the resistance is calculated based on the processed voltage change data set. The calculation formula is: ； in, is the dynamic change value of the resistance, is the instantaneous voltage change, is the time interval, is the maximum voltage value within the cycle, is the minimum voltage value within the cycle, is the magnitude of the applied pulse current, is the duration of current action; Based on the dynamic change value of the resistance, the change trend of the resistance over time is analyzed, the periodic fluctuation characteristics of the resistance are identified, and the resistance dynamic characteristic analysis result is obtained.

4. The method for evaluating the bonding quality of a semiconductor device bonding wire according to claim 1, wherein: The steps for obtaining the frequency response data are: Based on the resistance dynamic characteristic analysis results, multiple frequency intervals are set for scanning to obtain the resistance value corresponding to each frequency point, thereby obtaining frequency scanning result data; According to the frequency scanning result data, the variation trend of the resistance at different frequencies is analyzed, and the frequency response data is obtained by comparing the resistance values at different frequencies.

5. The method for evaluating the bonding quality of a semiconductor device bonding wire according to claim 1, wherein: The steps for obtaining the resistance stability evaluation result are: Based on the frequency response data, extract the resistance value at each frequency point, analyze the fluctuation of the resistance at each frequency point through Fourier transform, calculate the standard deviation and fluctuation amplitude of the resistance at each frequency, and obtain resistance change rate data; According to the resistance change rate and frequency response data, the stability of the resistor is calculated using the following formula: ； in, is the resistance stability evaluation result, is the resistance value at the qth frequency point, is the average resistance at all frequency points, is the voltage change at the qth frequency point, is the magnitude of the applied pulse current, is the total number of frequency response data, is the change amplitude of the resistance value at the qth frequency point, and are the maximum and minimum resistance values, is the duration of current action; Based on the resistance stability evaluation results, the resistance variation trend at different frequencies is analyzed, the frequency range with poor resistance stability is identified, and the resistance stability analysis results are generated.

6. The method for evaluating the bonding quality of a semiconductor device bonding wire according to claim 1, wherein: The steps for obtaining the abnormal capacitance analysis result are as follows: Measure the capacitance around the bonding point, record the capacitance value of each measurement point, and record the measurement time point to obtain a capacitance value record list and a time record list; Based on the capacitance value record list and the time record list, the change rate of the capacitance value at each point relative to the previous measurement point is calculated using the following formula: ； in, is the rate of change, For the The capacitance value of the point, is the capacitance value of the previous point, is the standard deviation of the capacitance values of adjacent points, is the time interval, It is a frequency calculation based on time difference, and the calculation method is ; Based on the change rate, an area where the change rate exceeds a set threshold is identified, and the area exceeding the set threshold is marked as abnormal, thereby obtaining an abnormal capacitance analysis result.

7. The method for evaluating the bonding quality of a semiconductor device bonding wire according to claim 1, wherein: The steps for acquiring the load response data are: Based on the abnormal capacitance analysis results, a load change instruction is executed on the target area, the set resistance value is switched in sequence, and the instantaneous voltage value and current value under each load state are recorded respectively to generate a voltage response record and a current response record; Based on the voltage response record and the current response record, the voltage values and the current values collected under each load state are structured and arranged, and time linear alignment is performed according to the load change sequence to generate load response data.

Images