Chip surface flatness detection method and system

Abstract

The invention relates to the technical field of semiconductor testing, and discloses a chip surface flatness detection method and system. According to the method, an air pressure difference sensing method is taken as a core, a multi-distance stepped approximation dynamic calibration process is designed, a fitting method with physical constraint is adopted, an air gap backpressure response model which strictly follows a physical rule and is high in extrapolation is constructed, the accuracy and the stability of a mapping relation between an air pressure signal and microscopic distance change are ensured, and the accuracy and the reliability of the air gap backpressure response model are improved. A theoretical basis is laid for high-precision triggering, a double-threshold parallel triggering mechanism is designed, a hard real-time control bus is combined, microsecond-level synchronous response from air pressure signal sudden change to platform motion locking is achieved, an analog signal sensed by air pressure is reliably converted into a high-precision height value, and the high-precision height value is obtained. The problem of triggering position drifting caused by response delay or signal jitter in a traditional method is solved, and therefore the precision and efficiency of chip surface flatness detection can be remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of semiconductor testing technology, and in particular to a method and system for detecting chip surface flatness. Background Technology

[0002] In the field of semiconductor packaging and advanced manufacturing, chip packaging warpage or surface micromorphology is a key indicator affecting its subsequent stacking, bonding and long-term reliability. High-precision chip surface flatness detection is crucial for improving packaging yield and product performance.

[0003] Currently common non-contact flatness measurement technologies, such as optical interferometry, are highly accurate, but the equipment is expensive, the measurement speed is slow, and they are sensitive to environmental vibrations, making it difficult to integrate them into online production lines for rapid inspection. Capacitive sensing methods are not suitable for insulating materials, and image recognition methods have extremely high requirements for image quality. Therefore, these methods face multiple constraints in terms of cost, efficiency, and applicability in industrial online high-speed and highly adaptable inspection scenarios. Summary of the Invention

[0004] The main objective of this invention is to provide a method and system for detecting chip surface flatness, aiming to solve the technical problems in the prior art.

[0005] This invention proposes a method for detecting chip surface flatness, applied to chip flatness detection equipment, comprising: Acquire the dynamic error calibration dataset and probe attribute parameters of the flatness detection device, and acquire the chip material parameters, packaging material parameters and preset sampling point coordinate dataset of the chip; The safety error envelope value is obtained based on the dynamic error calibration dataset, and risk-weighted compensation is performed on the safety error envelope value based on the chip material parameters and the probe attribute parameters to obtain the safety descent height value. Multi-distance back pressure characteristic calibration was performed on the stepped standard gauge block to obtain the original synchronous calibration data stream; An air gap back pressure response model is constructed based on the original synchronous calibration data stream, and the air pressure trigger threshold and rate of change threshold are obtained according to the air gap back pressure response model. An ordered sampling path is obtained based on the chip material parameters, packaging material parameters, and the preset sampling point coordinate dataset, wherein the ordered sampling path includes multiple measured sampling points; Based on the safe descent altitude and the ordered sampling path, differential air pressure measurement is performed on the measured sampling points to obtain real-time feedback air pressure signals. Based on the real-time feedback air pressure signal, air pressure trigger threshold and the rate of change threshold, a critical trigger command is obtained to control the carrier platform to stop moving, and the absolute height value of the carrier platform at the time of triggering the critical trigger command is obtained by the grating ruler. The surface profile height deviation set of the chip is obtained based on the absolute height values ​​corresponding to multiple measured sampling points.

[0006] Preferably, the step of obtaining the safety error envelope value based on the dynamic error calibration dataset, and performing risk-weighted compensation on the safety error envelope value based on the chip material parameters and the probe attribute parameters to obtain the safety descent height value includes: The positioning deviation dataset of the bearing platform, the stagnation impact data sequence of the air pressure sensing probe, and the triaxial acceleration time series data are obtained based on the dynamic error calibration dataset. The triaxial acceleration time series data are then processed by time-domain quadratic integration and vector synthesis to obtain the vibration displacement amplitude. The stagnation impact amplitude is obtained from the positioning deviation dataset, and statistical analysis is performed on the positioning deviation dataset to obtain the inherent positioning deviation and random positioning error. The safety error envelope value is obtained by using the classification peak superposition method based on the inherent positioning deviation, random positioning error, stationary impact amplitude, and vibration displacement amplitude. The maximum thickness of the chip is obtained based on the chip material parameters, and the quasi-static collision risk index is obtained based on the chip material parameters and the probe attribute parameters. Obtain the safety factor mapping rule table, and obtain the comprehensive safety margin based on the quasi-static collision risk index, the safety factor mapping rule table, and the safety error envelope value; Obtain the probe chuck reference height difference, and obtain the safe descent height value based on the probe chuck reference height difference, the maximum chip thickness value, and the comprehensive safety margin.

[0007] Preferably, the step of constructing an air gap back pressure response model based on the original synchronous calibration data stream, and obtaining the air pressure trigger threshold and the rate of change threshold according to the air gap back pressure response model, includes: Extract the test feedback air pressure signal sequence and test platform height sequence corresponding to each step plane from the original synchronous calibration data stream; The test feedback air pressure signal sequence is smoothed to obtain a smoothed test air pressure signal sequence, and the test signal change rate sequence is obtained based on the smoothed test air pressure signal sequence. Obtain the positive inflection threshold, and identify abrupt changes in the rate of change of the test signal based on the test signal change rate sequence and the positive inflection threshold to obtain the characteristic inflection moment; Extract the platform height reading corresponding to the characteristic turning point from the test platform height sequence to obtain the trigger platform height reading, and obtain the steady-state air pressure characteristic value based on the characteristic turning point and the smooth air pressure signal sequence; Data reconstruction is performed on multiple height readings of the triggering platform and the corresponding steady-state air pressure characteristic values ​​to obtain a set of air gap back pressure calibration data points. The air gap back pressure calibration data point set is then fitted using a nonlinear least squares optimization algorithm to obtain an air gap back pressure response model. The sensitivity function is obtained based on the air gap back pressure response model, and an optimization algorithm is used to efficiently optimize the measurement point based on the sensitivity function to obtain the critical sensing air gap. The critical sensing air gap is substituted into the air gap back pressure response model for calculation to obtain the air pressure trigger threshold, and the rate of change threshold is obtained according to the air gap back pressure response model and the critical sensing air gap.

[0008] Preferably, the step of obtaining an ordered sampling path based on the chip material parameters, packaging material parameters, and the preset sampling point coordinate dataset, wherein the ordered sampling path includes multiple measured sampling points, includes: The Young's modulus ratio and theoretical curvature are obtained based on the chip material parameters and the packaging material parameters, and the chip aspect ratio is obtained based on the chip material parameters. The elongated shape determination threshold and the weak constraint determination threshold are obtained, and the dominant warping shape of the chip is determined based on the chip aspect ratio, Young's modulus ratio, elongated shape determination threshold and the weak constraint determination threshold to obtain the dominant warping mode. Obtain a sampling rule knowledge base, query the sampling rule knowledge base according to the dominant warping mode, retrieve and call the corresponding morphology sampling rule set, wherein the morphology sampling rule set includes key feature regions and their corresponding sampling priorities; Based on the key feature region, multiple preset sampling points in the preset sampling point coordinate dataset are screened to obtain multiple measured sampling points, and the multiple measured sampling points are sorted according to the sampling priority to obtain a preliminary sampling path sequence. The initial sampling path sequence is optimized by point order and smoothed by trajectory processing to obtain an ordered sampling path.

[0009] Preferably, the step of performing differential pressure measurement on the measured sampling points based on the safe descent altitude and the ordered sampling path to obtain a real-time feedback pressure signal includes: The chip is transferred to the chuck by a robotic arm, and the chuck's vacuum adsorption function is activated to fix the chip in place. Extract the horizontal coordinates of the current measured sampling point from the ordered sampling path, and control the carrying platform to move horizontally according to the horizontal coordinates through the system control center, so that the current measured sampling point moves directly below the barometric pressure sensor. The system control center controls the opening of the first solenoid valve and controls the air pressure sensing probe to descend vertically according to the safe descent height value. The system control center controls the opening of the second solenoid valve and controls the carrier moving platform to drive the chip upward to perform a vertical uniform linear approximation motion. High-frequency signal acquisition is performed synchronously to obtain the original gas pressure acquisition signal sequence. The original gas pressure acquisition signal sequence is then filtered and scaled to obtain the real-time feedback gas pressure signal.

[0010] Preferably, the step of obtaining a critical trigger command based on the real-time feedback air pressure signal, the air pressure trigger threshold, and the rate of change threshold to control the carrier platform to stop moving, and obtaining the absolute height value of the carrier platform at the moment of triggering the critical trigger command through a grating ruler, includes: The real-time feedback air pressure signal is filtered in real time to obtain a smooth signal stream, and a first-order difference operation is performed on the smooth signal stream to obtain a real-time signal change rate sequence. The smooth signal stream includes multiple smooth air pressure sample values, and the real-time signal change rate sequence includes multiple instantaneous signal change rates. Parallel triggering condition monitoring is performed based on the smoothed signal stream, real-time signal change rate sequence, smoothed air pressure sample value, and instantaneous signal change rate. The parallel triggering conditions include a level triggering condition represented by the smoothed air pressure sample value not being less than the air pressure triggering threshold, and an edge triggering condition represented by the instantaneous signal change rate not being less than the change rate threshold. If, during the parallel trigger condition monitoring process, either the level trigger condition or the edge trigger condition is met for the first time, it is determined that a valid trigger event has occurred, and a critical trigger instruction is generated. The critical trigger command is sent to the motion controller in real time to trigger the motion controller to immediately generate an emergency stop command and lock the position of the carrying platform. The absolute height value of the carrying platform at the moment the critical trigger command is triggered is obtained by the grating ruler.

[0011] This application also provides a chip surface flatness detection system, including multiple modules, which are used to implement the steps of the chip surface flatness detection method described above.

[0012] Preferably, the module includes multiple units, which are used to implement the steps of the chip surface flatness detection method described above.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described chip surface flatness detection method.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described chip surface flatness detection method.

[0015] The beneficial effects of this invention are as follows: Firstly, this invention uses the differential pressure sensing method as its core. Through system dynamic error calibration and material risk analysis, it innovatively proposes a risk-weighted safety height calculation model. This allows the safety distance of the pressure probe to adaptively adjust when it approaches the chip, maximizing measurement efficiency while eliminating collision risks and overcoming the inherent contradiction between safety and efficiency in traditional methods. Secondly, by designing a multi-distance stepped approximation dynamic calibration process and employing a fitting method with physical constraints, a gap back pressure response model that strictly follows physical laws and has strong extrapolation properties is constructed. This ensures the accuracy and stability of the mapping relationship between the pressure signal and microscopic distance changes, laying a theoretical foundation for high-precision triggering. Thirdly, by designing a dual-threshold parallel triggering mechanism and combining it with a hard real-time control bus, a microsecond-level synchronous response from sudden changes in the pressure signal to platform motion locking is achieved. This reliably converts the analog signal of pressure sensing into a high-precision height value, solving the trigger position drift problem caused by response delay or signal jitter in traditional methods. This significantly improves the accuracy and efficiency of chip surface flatness detection. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a method flow according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the system structure according to an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application.

[0019] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] like Figure 1 As shown, this application provides a chip surface flatness detection method, applied to a chip flatness detection device, comprising: S1. Obtain the dynamic error calibration dataset and probe attribute parameters of the flatness detection device, and obtain the chip material parameters, packaging material parameters and preset sampling point coordinate dataset of the chip. S2. Obtain the safety error envelope value based on the dynamic error calibration dataset, and perform risk-weighted compensation on the safety error envelope value based on the chip material parameters and the probe attribute parameters to obtain the safety descent height value; S3. Perform multi-distance back pressure characteristic calibration on the stepped standard gauge block to obtain the original synchronous calibration data stream; S4. Construct an air gap back pressure response model based on the original synchronous calibration data stream, and obtain the air pressure trigger threshold and change rate threshold according to the air gap back pressure response model; S5. Obtain an ordered sampling path based on the chip material parameters, packaging material parameters and the preset sampling point coordinate dataset, wherein the ordered sampling path includes multiple measured sampling points; S6. Based on the safe descent height value and the ordered sampling path, perform air pressure differential sensing measurement on the measured sampling point to obtain real-time feedback air pressure signal; S7. Based on the real-time feedback air pressure signal, air pressure trigger threshold and the rate of change threshold, a critical trigger command is obtained to control the carrier platform to stop moving, and the absolute height value of the carrier platform at the moment of triggering the critical trigger command is obtained by the grating ruler. S8. Obtain the surface contour height deviation set of the chip based on the absolute height values ​​corresponding to multiple measured sampling points.

[0022] As described in steps S1-S8 above, the chip flatness detection device of this invention consists of an integrated hardware system. The various parts of this system work together to achieve precise measurement. Specifically, it includes: a system control center, a pressure sensing probe, a signal control module, a support platform and its control module, and a grating ruler. The system control center, as the core processing and instruction unit, is responsible for executing all logical steps of the method of this invention and sending control commands to each module. It receives data from the signal control module and makes judgments, while simultaneously precisely controlling the platform movement through the support platform control module. The pressure sensing probe, as the core sensing element, includes a lifting air source input port, a pressure output port, and a feedback pressure input port. The lifting air source input port receives external air, the pressure output port outputs a constant sensing pressure to the tiny air gap formed between the probe end face and the chip surface to establish a measurement field, and the feedback pressure input port receives air from the aforementioned measurement air gap. The returned back pressure signal changes with the air gap. The signal control module, acting as the hub for air pressure monitoring and conversion, functions to: control the air pressure solenoid valve connected to the air pressure sensing probe to precisely open, close, or regulate airflow; receive the air pressure signal from the probe's feedback air pressure input port and convert it into a high-resolution real-time feedback air pressure signal for acquisition and analysis by the system control center; the carrier platform is equipped with a vacuum adsorption chuck for fixing the chip, and is driven by a high-precision electric motor driven by a driver, allowing for three-dimensional movement on a linear slide rail; the carrier platform control module receives the target three-dimensional coordinate position command from the system control center and drives the carrier platform to move precisely to the designated position, with its movement ultimately fed back by a high-precision grating ruler; the grating ruler is a high-precision position sensor rigidly connected to the carrier platform, used to measure and feedback the absolute position (height) of the carrier platform in real time, and is a key component for obtaining the absolute height value at the trigger moment; This invention establishes a multi-dimensional quantitative perception of flatness detection by acquiring the dynamic error calibration dataset and probe attribute parameters of the flatness detection device, as well as the chip material parameters, packaging material parameters, and preset sampling point coordinate dataset of the chip under test. This invention obtains the safety error envelope value based on the dynamic error calibration dataset and performs risk-weighted compensation on the safety error envelope value based on the chip material parameters and probe attribute parameters to obtain the safety descent height value. This achieves adaptive risk adjustment of the safety margin, which not only ensures the absolute safety of the chip when approaching the probe and guarantees that the probe is always safely positioned above the highest point on the chip surface, but also avoids the waste of measurement stroke caused by setting the safety descent height value too conservatively. This invention fixes a stepped standard gauge block with a known height difference to a support platform. For each stepped plane, after controlling the probe to be positioned at a safe height, a uniform approach operation is performed on the platform under a constant pressure field, while simultaneously acquiring the feedback air pressure signal sequence and the corresponding platform height sequence. This obtains the original synchronous calibration data stream corresponding to each stepped plane. This process simulates the physical process of the probe gradually approaching the chip surface in real measurement. The acquired data directly reflects the complete response characteristics of the system under dynamic approximation, rather than static multi-point calibration, thus laying the foundation for the subsequent establishment of a high-fidelity dynamic response model. Based on the original synchronous calibration data stream, this invention intelligently identifies the characteristic transition moment when the air pressure signal enters the nonlinear rapid increase region from the linear slowly varying region, extracts the trigger platform height reading and steady-state air pressure characteristic value corresponding to each stepped plane, and thus constructs the air pressure characteristic value. The air gap-back pressure calibration data set is used, and a constrained nonlinear least squares fitting is employed to obtain an air gap back pressure response model that strictly follows physical laws. Further analysis of the model optimizes and determines the critical sensing air gap and its corresponding air pressure trigger threshold while maximizing the overall quality factor that ensures high sensitivity and high signal-to-noise ratio. The rate of change threshold is also derived, enabling trigger judgment to possess both high sensitivity and strong anti-interference capabilities, thus improving measurement reliability from the source. This invention obtains an ordered sampling path based on chip material parameters, packaging material parameters, and a preset sampling point coordinate dataset, breaking away from the traditional blind scanning mode of uniform sampling. It achieves predictive intelligent inspection based on physical mechanisms. By predicting the most likely deformation characteristics of the chip in advance and specifically optimizing the distribution and access order of sampling points, it can more accurately capture key morphologies on the chip surface, significantly improving detection efficiency. This invention performs differential pressure measurement on each sampling point on the chip surface based on a pre-calculated safe descent height and an optimized, ordered sampling path. This process is not a simple one-time descent, but follows a strict, layered, progressive control logic: First, the support platform carries the chip horizontally to directly below the target point; then, the pressure sensing probe descends vertically to the safe descent height, ensuring a safe physical gap between the probe tip and the highest point of the chip surface; next, a stable axisymmetric constant pressure field is established between the probe and the chip surface; finally, the control platform drives the chip to approach the probe at a constant low speed in a straight line. In this process, the air pressure in the probe's back pressure chamber changes with minute variations in the air gap. Real-time feedback air pressure signals are acquired through high-frequency acquisition, thus encoding the microscopic spatial displacement changes into a continuously measurable analog signal. This operation stably maps microscopic spatial distance changes (nanometer to micrometer scale) into high-resolution time-series electrical signal changes. The system then performs real-time filtering on the feedback air pressure signal to obtain a smooth signal stream and calculates its first-order difference to obtain a real-time signal change rate sequence. This sequence is monitored in parallel using a dual-threshold parallel triggering mechanism with "OR" logic. When either condition is met for the first time, the system immediately determines it as a valid trigger event and generates a critical trigger command. This command is then transmitted in real-time... Industrial buses (such as EtherCAT) send data to the motion controller with the highest priority, triggering it to execute an "emergency stop" and lock the motor position within a control cycle (typically <1ms). Almost simultaneously with locking, the system reads the absolute position reading of a high-precision grating ruler rigidly connected to the platform. This reading represents the absolute height of the platform at the trigger moment. Based on the absolute height values ​​corresponding to all measured sampling points, the arithmetic mean is calculated to obtain the reference surface height value. Then, for each measured sampling point, the difference between its corresponding absolute height value and the reference surface height value is calculated to obtain the contour height deviation value. A positive value indicates a convexity, and a negative value indicates a concaveness. After traversing all measured sampling points and performing this calculation, the output is a set consisting of the contour height deviation values ​​of all measured sampling points, which is the surface contour height deviation set. Through the surface contour height deviation set, the overall topographic undulation characteristics of the chip surface can be intuitively and quantitatively evaluated, local high and low points can be located, and it can be determined whether its flatness meets the process specification requirements. This invention utilizes the core principle of constant air gap during triggering to transform the complex absolute surface height measurement into a relative comparison of high-precision platform coordinates. Through simple reference plane subtraction and constant conversion, the core data characterizing surface flatness can be extracted efficiently and directly, which is beneficial to improving the accuracy and efficiency of chip surface flatness detection.

[0023] In one embodiment, step S2, which involves obtaining a safety error envelope value based on the dynamic error calibration dataset and performing risk-weighted compensation on the safety error envelope value based on the chip material parameters and the probe attribute parameters to obtain a safety descent height value, includes: S21. Based on the dynamic error calibration dataset, obtain the positioning deviation dataset of the bearing platform, the stagnation impact data sequence of the air pressure sensing probe, and the triaxial acceleration time series data, and perform time-domain quadratic integration and vector synthesis processing on the triaxial acceleration time series data to obtain the vibration displacement amplitude. S22. Obtain the stagnation impact amplitude based on the positioning deviation dataset, and perform statistical analysis on the positioning deviation dataset to obtain the inherent positioning deviation and random positioning error; S23. The safety error envelope value is obtained by using the classification peak superposition method based on the inherent positioning deviation, random positioning error, stagnant impact amplitude and vibration displacement amplitude; S24. Obtain the maximum thickness value of the chip based on the chip material parameters, and obtain the quasi-static collision risk index based on the chip material parameters and the probe attribute parameters; S25. Obtain the safety factor mapping rule table, and obtain the comprehensive safety margin based on the quasi-static collision risk index, the safety factor mapping rule table, and the safety error envelope value; S26. Obtain the probe chuck reference height difference, and obtain the safe descent height value based on the probe chuck reference height difference, the maximum chip thickness value, and the comprehensive safety margin.

[0024] As described in steps S21-S26 above, this invention acquires a vertical dynamic error calibration dataset of a flatness detection device. This vertical dynamic error calibration dataset refers to a set of heterogeneous, multidimensional data obtained through a specially designed dynamic calibration process. It is used to quantify and characterize the various dynamic error characteristics generated by the bearing platform and the barometric pressure sensor probe in the vertical direction under simulated working conditions. This includes the bearing platform's positioning deviation dataset, as well as the barometric pressure sensor probe's stagnation impact data sequence and triaxial acceleration time series data. Specifically, the positioning deviation dataset refers to the data set generated by repeatedly controlling the bearing platform to descend to the same target height under load, recording the height deviation between the actual stopping position and the target position each time. The stagnation impact data sequence refers to the data set generated by controlling the barometric pressure sensor probe to descend to the same target height under high-frequency displacement sensor monitoring. A single descent and lock-up operation is performed, recording a set of complete waveform data of the probe end face displacement changing over time. The maximum positive overshoot exceeding the final stable position is extracted from the stagnation impact data sequence as the stagnation impact amplitude. Simultaneously, the time elapsed from the start of motion to the displacement stabilizing within ±1% of the final value is recorded as the stagnation process duration. The triaxial acceleration time series data refers to the acceleration values ​​in three mutually perpendicular directions (usually defined as X, Y, and Z axes) measured and recorded by an accelerometer mounted on the probe base during at least one complete working cycle under simulated system operation conditions, i.e., when the air valves open and close according to the working cycle and the platform simulates start and stop. These values ​​are arranged in strict chronological order to form a data set. By acquiring the above data, the structural vibration information induced by the system's own operation can be quantified. Traditional safety descent height design typically uses static accuracy parameters from equipment specifications. This method fails to consider the dynamic coupling effects generated during actual system operation, potentially leading to inappropriate safety boundary settings: overly conservative settings affect measurement efficiency, while overly conservative settings fail to adequately cover inherent system errors and disturbances, posing a risk of probe-chip surface collision. Next, the triaxial acceleration time-series data undergoes time-domain quadratic integration and vector synthesis: first, each axis's data is integrated twice to obtain the displacement time-series; then, the vector magnitude of the triaxial displacement is calculated to obtain equivalent vibration displacement time-series data. The maximum absolute value of this equivalent vibration displacement time-series data is extracted as the vibration displacement amplitude, representing the maximum equivalent vertical displacement that vibration may cause. Through mathematical statistical analysis of the positioning deviation dataset, its mean and standard deviation are calculated. The absolute value of the mean is used as the inherent positioning deviation, and the standard deviation as the random positioning error. Through… "Calculate the safety error envelope value, where, Indicates the safety error envelope value. Indicates inherent positioning deviation. This indicates random positioning error. Indicates the magnitude of the stalled impact. It represents the vibration displacement amplitude. By performing a classification peak superposition operation, the systematic deviation and the statistical random boundary are arithmetically superimposed and then superimposed with two independent instantaneous physical peaks, so as to obtain a single quantitative index that can comprehensively and conservatively cover the total displacement deviation caused by all key dynamic error sources under the theoretically most unfavorable combination. This invention obtains the Vickers hardness, yield strength, Young's modulus, and yield strength of the chip material through chip material parameters and probe property parameters, respectively, and then uses " "Calculate the quasi-static collision risk index, where, This indicates the quasi-static collision risk index. Indicates the Vickers hardness of the chip material. This indicates the Young's modulus of the probe material. Indicates the yield strength of the chip material. Indicates the yield strength of the probe material. This represents the first weighting coefficient. This represents the second weighting coefficient, which is obtained through finite element simulation and collision experiments. Because high-hardness chip materials (such as silicon) are prone to cracking or breakage during collisions, and low-modulus probes may experience increased local stress due to deformation, therefore… "This item is used to assess the risk of brittle damage. The higher this ratio, the higher the quasi-static collision risk index. If the yield strength of the chip material is much lower than that of the probe, the collision may cause irreversible indentation or material transfer on the chip surface. Therefore," This step assesses the risks of plastic deformation and adhesion. Ultimately, based on the calculated quasi-static collision risk index, it determines the discrete risk level corresponding to the current probe-chip material pair and obtains a safety factor mapping rule table. This table is a deterministic correspondence table that maps three discrete risk levels ("low," "medium," and "high") to a set of preset, numerically increasing anti-collision risk coefficients. The anti-collision risk coefficient is obtained by querying the safety factor mapping rule table based on the determined discrete risk level. This coefficient is used to proportionally amplify the safety error envelope value S, calculated based on the system's dynamic error, according to the risk level of the current material pair. The calculation of the safety error envelope value and the anti-collision risk coefficient... The product yields the comprehensive safety margin. This comprehensive safety margin is obtained by adding a safety buffer based on the severity of collision consequences to the pure error envelope, thus theoretically constructing a more intelligent final safety boundary that can simultaneously resist inherent dynamic errors of the system and prevent high-consequence collision risks caused by differences in material properties. The maximum chip thickness is obtained based on the chip material parameters; this value is the upper limit allowed by the product standard. The probe chuck reference height difference is then obtained. This reference height difference refers to the absolute distance difference in the vertical direction between the center point of the measuring end face of the barometric pressure sensor probe at its defined "mechanical origin" and the upper surface of the chuck on the support platform (when no object is placed). This is achieved through… "Calculate the safe descent altitude, where, Indicates the safe descent altitude. Indicates the reference height difference of the probe chuck. This indicates the maximum thickness of the chip. The safe descent height value, representing the overall safety margin, refers to the absolutely safe axial displacement of the air pressure sensing probe as it descends from its mechanical origin towards the chip. The safe descent height value determined through this process ensures that the probe can be safely positioned above the highest point of the surface when any compliant chip is measured, thus providing fundamental safety assurance for subsequent precision measurements.

[0025] In one embodiment, step S4, which involves constructing an air gap back pressure response model based on the original synchronous calibration data stream and obtaining the air pressure trigger threshold and the rate of change threshold according to the air gap back pressure response model, includes: S41. Extract the test feedback air pressure signal sequence and test platform height sequence corresponding to each step plane from the original synchronous calibration data stream; S42. The test feedback air pressure signal sequence is smoothed to obtain a test smoothed air pressure signal sequence, and the test signal change rate sequence is obtained based on the test smoothed air pressure signal sequence. S43. Obtain the positive turning point determination threshold, and identify the abrupt change in the rate of change of the test signal based on the test signal change rate sequence and the positive turning point determination threshold to obtain the characteristic turning point moment; S44. Extract the platform height reading corresponding to the characteristic turning point from the test platform height sequence to obtain the trigger platform height reading, and obtain the steady-state air pressure characteristic value according to the characteristic turning point and the smooth air pressure signal sequence. S45. Data reconstruction is performed on multiple height readings of the triggering platform and the corresponding steady-state air pressure characteristic values ​​to obtain a set of air gap back pressure calibration data points. The air gap back pressure calibration data point set is then fitted using a nonlinear least squares optimization algorithm to obtain an air gap back pressure response model. S46. Obtain the sensitivity function based on the air gap back pressure response model, and use an optimization algorithm to efficiently optimize the measurement point based on the sensitivity function to obtain the critical sensing air gap; S47. Substitute the critical sensing air gap into the air gap back pressure response model for calculation to obtain the air pressure trigger threshold, and obtain the rate of change threshold according to the air gap back pressure response model and the critical sensing air gap.

[0026] As described in steps S41-S47 above, this invention extracts the test feedback air pressure signal sequence and test platform height sequence corresponding to the step plane from the original synchronous calibration data stream, and performs moving average or low-pass filtering on the test feedback air pressure signal sequence to suppress high-frequency noise, obtaining a smoothed test air pressure signal sequence. The first-order difference of the smoothed test air pressure signal sequence is calculated to obtain the test signal change rate sequence, thus directly reflecting the sensitivity of air pressure change with the distance between the probe and the step surface. A positive turning point judgment threshold is set based on the system noise level: in a stable state where the calibration system is stationary and there is no approximation action, background feedback air pressure signals are collected for a period of time, and the standard deviation of its signal change rate sequence is calculated. Three to five times this standard deviation is set as the positive turning point judgment threshold to ensure effective differentiation between noise fluctuations and true signal turning points. During the approximation process, when the value of the test signal change rate sequence first jumps from a state below the positive turning point judgment threshold and continues to exceed the positive turning point judgment threshold in multiple consecutive sampling points thereafter... When the time is right, this moment is determined as the characteristic transition moment when the air pressure signal enters the "nonlinear rapid increase region" from the "linear gradual change region". The height of the test platform corresponding to this characteristic transition moment is recorded as the trigger platform height reading. At the same time, starting from this characteristic transition moment, a test smooth air pressure signal sequence of a preset duration is extracted, such as 100 milliseconds. The arithmetic mean of this data sequence is calculated, and this average value is recorded as the steady-state air pressure characteristic value corresponding to the step plane. This steady-state air pressure characteristic value eliminates the fluctuation of the trigger transient and represents the stable response of the system at this fixed distance. The trigger platform height reading and the corresponding steady-state air pressure characteristic value are used as calibration data pairs corresponding to each step plane. This invention obtains the probe end face height and the chuck reference plane offset. The probe end face height refers to the vertical height value of the measuring end face of the air pressure sensing probe under the unified absolute mechanical reference system of the equipment (such as the base reference plane). The chuck reference plane offset refers to the height difference between the upper surface of the chuck and the absolute mechanical reference plane of the equipment. "Calculate the actual physical distance, where, Indicates the first The actual physical distance corresponding to each step plane Indicates the height of the probe end face. Indicates the first The platform height readings corresponding to each step plane Indicates the offset of the chuck reference plane. Indicates the first The vertical height value corresponding to each step plane is used to convert the platform coordinate system data in the calibration data pair into the physical distance data required for modeling, and to construct the air gap back pressure calibration data point set based on the actual physical distance and its corresponding steady-state air pressure characteristic value. Based on the physical mechanism of nonlinear saturation response of air pressure under narrow air gaps due to airflow limitation, an exponential decay function was selected as the candidate basic model. After selecting the model expression, a constrained nonlinear least squares optimization algorithm was used for fitting. This algorithm aims to minimize the sum of squared residuals between the model's predicted and measured values. Simultaneously, during the iterative solution process, the constraint condition that the first derivative is always negative is enforced. Through this constrained fitting strategy, inviolable physical prior knowledge is embedded into the data-driven process, thereby ensuring that the final model is accurate in any extrapolation or interpolation. During the measurement, no non-physical responses are generated, such as the anomalous interval where the air gap decreases but the air pressure drops instead, which greatly enhances the extrapolation reliability and generalization ability of the model. After the optimization algorithm converges, a set of optimal model parameters is obtained. Substituting these parameters into the selected function form yields the air gap back pressure response model. For the sake of rigor, residual analysis is further used to check the randomness of the fitting residuals to determine systematic bias, thereby outputting a continuous mathematical model that accurately matches the experimental data and strictly follows the physical laws. This model constitutes the theoretical basis and conversion benchmark for the entire air pressure difference sensing system to achieve quantitative measurement. By taking the first derivative of the air gap back pressure response model, the sensitivity function is obtained. This sensitivity function is a mathematical function that takes the distance between the probe and the surface as the independent variable and the absolute value of the first derivative of the model at that distance as its function value. The sensitivity function quantitatively describes the drastic response of the system pressure output signal to distance changes throughout the entire measurement range, i.e., the distribution of sensitivity as a function of distance. Simultaneously, it is essential to ensure that the signal at the selected operating point is sufficiently strong to distinguish it from the system's background noise, but not so strong as to approach the sensor's measurement upper limit, lest the space for variation be lost. Therefore, regions with excessively weak or strong signals need to be avoided. Within the effective distance domain of the air gap back pressure response model, each possible distance point is comprehensively evaluated using the following method: obtaining the sensitivity corresponding to the distance value through the sensitivity function, and simultaneously obtaining the model-predicted air pressure value and system noise level at that distance point. The system noise level is estimated by taking the average value and calculating the difference between the model-predicted air pressure value and the system noise level estimate to obtain the signal-to-noise margin. The product of the sensitivity at that point and the signal-to-noise margin is calculated to obtain the overall quality factor at that point, which reflects both the sensitivity of the point to distance changes and the discriminability of the signal itself. Subsequently, a one-dimensional optimization algorithm (such as the golden section search) is used to find a specific distance value within the domain that maximizes this quality factor function. This distance value is determined as the critical sensing air gap. The critical sensing air gap is substituted into the air gap back pressure response model for calculation, and the obtained air pressure value is the air pressure trigger threshold. The first derivative analysis of the air gap back pressure response model is performed near the critical sensing air gap. The product of the absolute value of the derivative and the preset safety factor is taken as the rate of change threshold. The rate of change threshold is proportional to the intrinsic sensitivity of the system at the optimal operating point, which can effectively capture the real rapid changes in the signal and suppress small fluctuations caused by noise.

[0027] In one embodiment, step S5, which involves obtaining an ordered sampling path based on the chip material parameters, packaging material parameters, and the preset sampling point coordinate dataset, wherein the ordered sampling path includes multiple measured sampling points, includes: S51. Obtain the Young's modulus ratio and theoretical curvature based on the chip material parameters and the packaging material parameters, and obtain the chip aspect ratio based on the chip material parameters; S52. Obtain the elongated shape determination threshold and the weak constraint determination threshold, and determine the dominant warping shape of the chip based on the chip aspect ratio, Young's modulus ratio, elongated shape determination threshold and the weak constraint determination threshold to obtain the dominant warping mode. S53. Obtain the sampling rule knowledge base, query the sampling rule knowledge base according to the dominant warping mode, retrieve and call the corresponding morphology sampling rule set, wherein the morphology sampling rule set includes key feature regions and their corresponding sampling priorities; S54. Based on the key feature region, multiple preset sampling points in the preset sampling point coordinate dataset are screened to obtain multiple measured sampling points, and the multiple measured sampling points are sorted according to the sampling priority to obtain a preliminary sampling path sequence. S55. Perform point order optimization and trajectory smoothing on the preliminary sampling path sequence to obtain an ordered sampling path.

[0028] As described in steps S51-S55 above, this invention obtains the chip planar dimensions, chip substrate Young's modulus, and chip thermal expansion coefficient based on chip material parameters, and obtains the encapsulation curing temperature, encapsulation material Young's modulus, encapsulation material thermal expansion coefficient, and encapsulation thickness based on encapsulation material parameters. Simultaneously, a preset curing environment reference temperature is obtained, and the difference between the encapsulation curing temperature and the curing environment reference temperature is calculated to obtain the encapsulation curing temperature difference. For a simple dual-material beam model, through " "Calculate the thermal mismatch strain, where, Indicates thermal mismatch strain. Indicates the coefficient of thermal expansion of the chip. This represents the coefficient of thermal expansion of the packaging material. This indicates the temperature difference during encapsulation and curing; based on the chip's planar dimensions, the length, width, and thickness are obtained, and the ratio of the package thickness to the chip thickness is calculated to obtain the thickness ratio; the ratio of the Young's modulus of the package material to the Young's modulus of the chip substrate is calculated to obtain the Young's modulus ratio, which is then obtained through " "Calculate the theoretical curvature, where, Indicates the theoretical curvature. Indicates thermal mismatch strain. Indicates the thickness ratio. Indicates Young's modulus ratio. The chip thickness is indicated by the theoretical curvature. When the theoretical curvature is greater than "0", it indicates that the chip is bent towards the package (usually concave); when the theoretical curvature is less than "0", it indicates that the chip is bent away from the package (usually convex). The chip aspect ratio is calculated based on the chip material parameters, and the thresholds for determining elongated shape and weak constraints are obtained. If both the chip aspect ratio and the Young's modulus ratio are satisfied, the dominant warping mode is directly determined to be corner warping. For chips that do not meet the corner warping condition, if the theoretical curvature is not equal to "0", the dominant warping mode is determined to be global bending. If the theoretical curvature is equal to "0", it can be determined that there is no obvious warping, but this situation rarely occurs when there is a temperature difference and thermal mismatch. The sampling rule knowledge base is acquired. This knowledge base is a pre-built and stored dataset that associates specific warp pattern identifiers (such as "global curvature" or "corner upturn") with a set of optimal spatial sampling rules tailored to that pattern. Based on the dominant warp pattern, the uniquely corresponding shape sampling rule set is retrieved and invoked by querying the sampling rule knowledge base. The shape sampling rule set is a structured set of instructions used to guide the measurement system to perform efficient spatial sampling on specific shape surfaces. It includes sampling key feature regions and their corresponding sampling priorities. The process iterates through each coordinate point in the pre-set sampling point coordinate dataset, calculating its relationship with each key feature region. The system determines the spatial relationship of the key feature regions. If a coordinate point is located within the geometric boundary of any key feature region, or its spatial distance from that region is less than a preset tolerance threshold, then the point is identified as a measured sampling point and retained. Otherwise, the point is marked as redundant and removed. After filtering, the system sorts all measured sampling points according to the sampling priority corresponding to each key feature region: all points belonging to higher priority regions are arranged at the beginning of the path sequence. For multiple points within the same priority region, they are locally sorted according to their original numbering order or spatial proximity principle (such as based on their X or Y coordinates). After this spatial matching and filtering... The selection and priority sorting process yields a preliminary sampling path sequence. Optimization algorithms, such as simulated annealing, genetic algorithms, or the Lin-Kernighan heuristic, are used to search the possible point sequence arrangement space. This involves repeatedly swapping, inserting, or reversing segments in the point sequence, and calculating in real-time the estimated total travel distance required for the platform to sequentially access all points under each new arrangement, as well as the estimated total time to satisfy acceleration and deceleration constraints, in order to find the point sequence arrangement that minimizes the total time. For the optimized point sequence, spline interpolation or Bézier curve fitting is used to generate smooth transition curves between adjacent points, replacing direct linear interpolation and thus avoiding errors at points. When a sudden change in velocity direction occurs, to ensure smooth movement and reduce vibration, an ordered sampling path is obtained after the above point sequence optimization and trajectory smoothing calculation. This invention predicts the most likely warping mode by analyzing the chip's material and process parameters, and dynamically generates an optimized measurement route that can accurately capture its main morphological features with the fewest sampling points and the shortest movement path, thereby significantly improving detection efficiency. It also improves defect detection rate and measurement reliability by increasing the sampling density in high-risk areas, which is conducive to realizing the leap from uniform blind scanning to predictive intelligent inspection, so as to solve the problem that the traditional uniform grid sampling method is inefficient and may miss key deformation features.

[0029] In one embodiment, step S6, which involves performing differential pressure measurement on the measured sampling points based on the safe descent altitude and the ordered sampling path to obtain a real-time feedback pressure signal, includes: S61. The chip is transferred to the chuck by the robotic arm, and the vacuum adsorption function of the chuck is activated to fix the chip. S62. Extract the horizontal coordinates of the current measured sampling point from the ordered sampling path, and control the carrying platform to move horizontally according to the horizontal coordinates through the system control center, so that the current measured sampling point moves directly below the barometric pressure sensor. S63. The first solenoid valve is opened by controlling the system control center, and the air pressure sensing probe is controlled to descend vertically according to the safe descent height value. S64. The second solenoid valve is opened by controlling the system control center, and the carrier moving platform is controlled to drive the chip to perform a vertical uniform linear approximation motion upward. S65. Simultaneously perform high-frequency signal acquisition to obtain the original gas pressure acquisition signal sequence, and filter and scale the original gas pressure acquisition signal sequence to obtain the real-time feedback gas pressure signal.

[0030] As described in steps S61-S65 above, the present invention uses a robotic arm to transfer the chip to the top of the support platform and activates the vacuum adsorption function of the chuck on the support platform to fix the chip; then, the horizontal coordinates of the current measured sampling point are obtained through an ordered sampling path; then, horizontal and vertical safety positioning is performed: the first solenoid valve is opened by the system control center, and layered motion control is performed using the horizontal coordinates of the current measured sampling point and the safe descent height as the moving target; the support platform is first horizontally aligned and then vertically descended by the system control center, so that the current measured sampling point is moved directly below the barometric pressure sensing probe; subsequently, the second solenoid valve is opened by the system control center, thereby positioning the probe end face and... A safe initial distance and a stable axisymmetric measurement pressure field are established between the chip surfaces. Then, at a preset constant speed, the carrier platform drives the chip to perform uniform linear approximation motion through precise uniform motion control. During the continuous ascent of the chip by the carrier platform, high-frequency signal acquisition is performed synchronously through the signal control module to obtain the original gas pressure acquisition signal sequence. Through digital signal processing, such as filtering and scaling conversion, a filtered and calibrated real-time feedback gas pressure signal is obtained. This process of the present invention forcibly stipulates a strict action sequence of "first safe positioning, then establishing a constant pressure field, and then uniform approximation and synchronous acquisition", thereby creatively mapping changes in microscopic spatial distance into high-resolution time-series electrical signals.

[0031] In one embodiment, step S7, which involves obtaining a critical trigger command based on the real-time feedback air pressure signal, the air pressure trigger threshold, and the rate of change threshold to control the carrier platform to stop moving, and obtaining the absolute height value of the carrier platform at the moment of triggering the critical trigger command via a grating ruler, includes: S71. The real-time feedback air pressure signal is filtered in real time to obtain a smooth signal stream, and a first-order difference operation is performed on the smooth signal stream to obtain a real-time signal change rate sequence, wherein the smooth signal stream includes multiple smooth air pressure sampling values, and the real-time signal change rate sequence includes multiple instantaneous signal change rates. S72. Parallel triggering condition monitoring is performed based on the smoothed signal stream, the real-time signal change rate sequence, the smoothed air pressure sample value, and the instantaneous signal change rate. The parallel triggering conditions include the level triggering condition represented by the smoothed air pressure sample value not being less than the air pressure triggering threshold, and the edge triggering condition represented by the instantaneous signal change rate not being less than the change rate threshold. S73. If, during the parallel trigger condition monitoring process, either the level trigger condition or the edge trigger condition is met for the first time, it is determined that a valid trigger event has occurred, and a critical trigger instruction is generated. S74. The critical trigger command is sent to the motion controller in real time to trigger the motion controller to immediately generate an emergency stop command and lock the position of the carrying platform, and the absolute height value of the carrying platform at the moment the critical trigger command is triggered is obtained by the grating ruler.

[0032] As described in steps S71-S74 above, this invention establishes a real-time data buffer to continuously receive and temporarily store the data stream acquired by the high-speed ADC, ensuring no data loss and providing continuous input data for subsequent processing. Through data buffering and queue management, it outputs continuous real-time feedback pressure signals. A sliding window averaging algorithm or digital low-pass filtering is used to filter the real-time feedback pressure signals in real time, suppressing high-frequency noise and obtaining a smooth signal stream, thereby improving signal stability. Then, a first-order difference operation is performed on the smooth signal stream to obtain a real-time signal rate of change sequence, which directly characterizes... The system measures the instantaneous rate of change of air pressure. It loads the rate of change threshold and the air pressure trigger threshold into the real-time judgment logic unit and establishes two independent comparison conditions: a level trigger condition (the smoothed air pressure sample value is not less than the air pressure trigger threshold) and an edge trigger condition (the instantaneous signal change rate is not less than the rate of change threshold). The system monitors these two conditions in parallel in real time, using an "OR" logic, meaning that the instant either condition is first met is considered a valid trigger event. This design of the invention has high robustness: the level trigger condition ensures reliable triggering when the signal is rising smoothly; the edge trigger condition provides greater sensitivity. This invention captures the starting point of a sudden rise in signal intensity, counteracts signal delays or minor fluctuations, prevents false or missed triggers, and acquires a critical trigger command after a valid trigger event occurs. The critical trigger command is a digital pulse signal characterizing the occurrence of the trigger event. The critical trigger command is sent to the motion controller with the highest priority via a real-time industrial bus (such as EtherCAT). The motion controller responds to the command within one control cycle (typically less than 1 millisecond), immediately generating an "emergency stop" command, interrupting the servo motor's trajectory planning, and locking the motor position. At this moment, the upward movement of the platform is instantaneously stopped and the position is locked. While the motion controller latches the motor position, it simultaneously reads the absolute position of the grating ruler to obtain the absolute height value of the platform at the trigger moment. This invention employs a dual-condition judgment logic to design a complete, low-latency hard real-time processing chain from signal processing, condition judgment, real-time control to data capture, ensuring reliable capture of valid trigger events within microseconds and synchronous recording of the corresponding absolute spatial coordinates. This achieves accurate, fast, and reliable conversion from continuous analog signals to discrete "trigger-stop-record" events.

[0033] This application also provides a chip surface flatness detection system, including: The data acquisition module is used to acquire the dynamic error calibration dataset and probe attribute parameters of the flatness detection device, as well as the chip material parameters, packaging material parameters, and preset sampling point coordinate dataset of the chip. The safety planning module is used to obtain the safety error envelope value based on the dynamic error calibration dataset, and to perform risk weighting compensation on the safety error envelope value based on the chip material parameters and the probe attribute parameters to obtain the safety descent height value. The characteristic calibration module is used to perform multi-distance back pressure characteristic calibration on the stepped standard gauge block to obtain the original synchronous calibration data stream. The model building module is used to build an air gap back pressure response model based on the original synchronous calibration data stream, and to obtain the air pressure trigger threshold and the rate of change threshold according to the air gap back pressure response model. The path planning module is used to obtain an ordered sampling path based on the chip material parameters, packaging material parameters and the preset sampling point coordinate dataset, wherein the ordered sampling path includes multiple measured sampling points; The precision measurement module is used to perform differential pressure measurement on the measured sampling points based on the safe descent height and the ordered sampling path to obtain real-time feedback pressure signals. The trigger latch module is used to obtain a critical trigger command based on the real-time feedback air pressure signal, the air pressure trigger threshold and the rate of change threshold, so as to control the carrier platform to stop moving, and to obtain the absolute height value of the carrier platform at the time of triggering the critical trigger command through the grating ruler; The data analysis module is used to obtain the surface contour height deviation set of the chip based on the absolute height values ​​corresponding to multiple measured sampling points.

[0034] In one embodiment, the precision measurement module includes: The chip fixing unit is used to transfer the chip to the chuck by a robotic arm and activate the chuck's vacuum adsorption function to fix the chip. The horizontal alignment unit is used to extract the horizontal coordinates of the current measured sampling point from the ordered sampling path, and control the carrier platform to move horizontally according to the horizontal coordinates through the system control center, so that the current measured sampling point moves directly below the barometric pressure sensor probe. The probe positioning unit is used to control the opening of the first solenoid valve through the system control center, and to control the air pressure sensing probe to descend vertically according to the safe descent height value. The uniform speed approximation unit is used to control the opening of the second solenoid valve through the system control center, and to control the carrier moving platform to drive the chip to perform a vertical uniform speed linear approximation motion upward. The signal acquisition unit is used to synchronously acquire high-frequency signals to obtain the original gas pressure acquisition signal sequence, and to filter and scale the original gas pressure acquisition signal sequence to obtain the real-time feedback gas pressure signal.

[0035] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described chip surface flatness detection method.

[0036] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described chip surface flatness detection method.

[0037] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0038] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0039] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A chip surface flatness detection method applied to a chip flatness detection device, characterized in that, The method comprises the following steps: acquiring dynamic error calibration data set of flatness detection equipment, probe attribute parameters, and acquiring chip material parameters, packaging material parameters and preset sampling point coordinate data set of the chip; acquiring safety error envelope value according to the dynamic error calibration data set, and performing risk weighted compensation on the safety error envelope value based on the chip material parameters and the probe attribute parameters to obtain safety drop height value; performing multi-distance back pressure characteristic calibration on the stepped standard block to obtain original synchronous calibration data stream; constructing air gap back pressure response model based on the original synchronous calibration data stream, and acquiring air pressure trigger threshold and change rate threshold according to the air gap back pressure response model; acquiring ordered sampling path according to the chip material parameters, packaging material parameters and the preset sampling point coordinate data set, wherein the ordered sampling path comprises a plurality of measured sampling points; performing air pressure difference sensing on the measured sampling points according to the safety drop height value and the ordered sampling path to obtain real-time feedback air pressure signal; acquiring critical trigger instruction based on the real-time feedback air pressure signal, air pressure trigger threshold and the change rate threshold to control the bearing platform to stop moving, and acquiring absolute height value of the bearing platform at the time when the critical trigger instruction is triggered through the grating ruler; acquiring surface profile height deviation set of the chip according to the absolute height value corresponding to each of the plurality of measured sampling points.

2. The method of claim 1, wherein the step of detecting the flatness of the wafer surface is performed by a flatness detector. The step of acquiring safety error envelope value according to the dynamic error calibration data set, and performing risk weighted compensation on the safety error envelope value based on the chip material parameters and the probe attribute parameters to obtain safety drop height value comprises: acquiring positioning deviation data set of the bearing platform, and stagnation impact data sequence and three-axis acceleration time series data of the air pressure sensing probe according to the dynamic error calibration data set, and performing time domain quadratic integration and vector synthesis processing on the three-axis acceleration time series data to obtain vibration displacement amplitude value; acquiring stagnation impact amplitude value according to the positioning deviation data set, and obtaining inherent positioning deviation and random positioning error by statistical analysis on the positioning deviation data set; acquiring safety error envelope value according to the inherent positioning deviation, random positioning error, stagnation impact amplitude value and the vibration displacement amplitude value by using classification peak superposition method; acquiring chip maximum thickness value according to the chip material parameters, and acquiring quasi-static collision risk index according to the chip material parameters and the probe attribute parameters; acquiring safety coefficient mapping rule table, and acquiring comprehensive safety margin according to the quasi-static collision risk index, the safety coefficient mapping rule table and the safety error envelope value; acquiring probe chuck reference height difference, and acquiring safety drop height value according to the probe chuck reference height difference, the chip maximum thickness value and the comprehensive safety margin.

3. The method of claim 1, wherein the step of detecting the flatness of the wafer surface is performed by a flatness detector. The step of constructing air gap back pressure response model based on the original synchronous calibration data stream, and acquiring air pressure trigger threshold and change rate threshold according to the air gap back pressure response model comprises: extracting test feedback air pressure signal sequence and test platform height sequence corresponding to each step plane from the original synchronous calibration data stream; Smooth the test feedback air pressure signal sequence to obtain a test smooth air pressure signal sequence, and acquire a test signal change rate sequence according to the test smooth air pressure signal sequence; Acquire a positive turning point determination threshold, and identify a test signal change rate mutation based on the test signal change rate sequence and the positive turning point determination threshold to obtain a feature turning point moment; Extract a platform height reading corresponding to the feature turning point moment from the test platform height sequence to obtain a trigger platform height reading, and acquire a steady-state air pressure feature value according to the feature turning point moment and the smooth air pressure signal sequence; Reconstruct data of a plurality of trigger platform height readings and the steady-state air pressure feature values corresponding thereto to obtain a gas gap back pressure calibration data point set, and fit the gas gap back pressure calibration data point set by using a nonlinear least squares optimization algorithm to obtain a gas gap back pressure response model; Acquire a sensitivity function according to the gas gap back pressure response model, and perform efficient measurement point optimization based on the sensitivity function by using an optimization algorithm to obtain a critical sensing gas gap; Substitute the critical sensing gas gap into the gas gap back pressure response model to obtain a gas pressure trigger threshold, and acquire a change rate threshold according to the gas gap back pressure response model and the critical sensing gas gap.

4. The method of claim 1, wherein the step of detecting the flatness of the wafer surface is performed by a flatness detector. The ordered sampling path is acquired according to the chip material parameters, the packaging material parameters, and the preset sampling point coordinate data set, wherein the ordered sampling path includes a plurality of measured sampling points, and the method comprises the following steps: The Young's modulus ratio and the theoretical curvature are acquired according to the chip material parameters and the packaging material parameters, and the chip length-width ratio is acquired according to the chip material parameters; The elongated appearance determination threshold and the weak constraint determination threshold are acquired, and the dominant warping appearance of the chip is determined based on the chip length-width ratio, the Young's modulus ratio, the elongated appearance determination threshold, and the weak constraint determination threshold to obtain a dominant warping mode; The sampling rule knowledge base is acquired, the sampling rule knowledge base is queried according to the dominant warping mode, and the corresponding appearance sampling rule set is searched and called, wherein the appearance sampling rule set includes a key feature region and a corresponding sampling priority thereof; The plurality of preset sampling points in the preset sampling point coordinate data set are screened based on the key feature region to obtain a plurality of measured sampling points, and the plurality of measured sampling points are sorted based on the sampling priority to obtain a preliminary sampling path sequence; The preliminary sampling path sequence is subjected to point sequence optimization and trajectory smoothing processing to obtain an ordered sampling path.

5. The method of claim 1, wherein the step of detecting the flatness of the wafer surface is performed by a flatness detector. The measured sampling points are subjected to air pressure difference sensing according to the safety falling height value and the ordered sampling path to obtain a real-time feedback air pressure signal, and the method comprises the following steps: The chip is delivered to the chuck by the mechanical arm, and the chuck vacuum adsorption function is turned on to fix the chip; The horizontal plane coordinates of the current measured sampling points are extracted from the ordered sampling path, and the horizontal plane coordinates are used to control the horizontal movement of the bearing platform by the system control center to move the current measured sampling points to the position directly below the air pressure sensing probe. The first electromagnetic valve is opened by the system control center, and the air pressure sensing probe is lowered in the vertical direction according to the safe falling height value; The second electromagnetic valve is opened by the system control center, and the chip is driven upward by the bearing moving platform to perform a vertical uniform-speed straight-line approaching motion. Synchronous high-frequency signal acquisition is performed to obtain an air pressure original acquisition signal sequence, and filtering and scaling conversion are performed on the air pressure original acquisition signal sequence to obtain a real-time feedback air pressure signal.

6. The method of claim 1, wherein the step of detecting the flatness of the wafer surface is performed by a flatness detector. The step of obtaining a critical trigger instruction based on the real-time feedback air pressure signal, an air pressure trigger threshold and a change rate threshold to control the bearing platform to stop moving, and obtaining an absolute height value of the bearing platform at the time when the critical trigger instruction is triggered by the grating ruler, comprises: Real-time filtering is performed on the real-time feedback air pressure signal to obtain a smooth signal stream, and first-order difference operation is performed on the smooth signal stream to obtain a real-time signal change rate sequence, wherein the smooth signal stream comprises a plurality of smooth air pressure sampling values, and the real-time signal change rate sequence comprises a plurality of instantaneous signal change rates. Parallel trigger condition monitoring is performed based on the smooth signal stream, the real-time signal change rate sequence, the smooth air pressure sampling value and the instantaneous signal change rate, wherein the parallel trigger condition comprises a level trigger condition that the smooth air pressure sampling value is not less than the air pressure trigger threshold, and an edge trigger condition that the instantaneous signal change rate is not less than the change rate threshold. If any one of the level trigger condition or the edge trigger condition is first satisfied in the parallel trigger condition monitoring process, it is determined that an effective trigger event occurs, and a critical trigger instruction is generated. The critical trigger instruction is sent to the motion controller in real time to trigger the motion controller to immediately generate an emergency stop command and lock the position of the bearing platform, and the absolute height value of the bearing platform at the time when the critical trigger instruction is triggered is obtained by the grating ruler.

7. A chip surface flatness detection system, characterized in that, The plurality of modules are used to implement the steps of the method of any one of claims 1 to 6.

8. The system of claim 7, wherein the system is configured to determine the flatness of the wafer surface by determining a difference between the first and second measurements. The plurality of units are used to implement the steps of the method of any one of claims 1 to 6. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 6.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

Images