Probe card pressure self-adaptive control pressure sensor integration method and system

Abstract

The present application relates to the technical field of semiconductor testing, and particularly relates to a probe card pressure self-adaptive control pressure sensor integration method and system, a pressure sensing unit array is arranged at the probe card to collect capacitance signals, and spatial distribution and time sequence changes are jointly analyzed to obtain pressure field state information, based on the information, an abnormal area and a transmission path are identified to obtain a pressure abnormal source positioning result, according to the positioning result, adjustment requirements are decomposed into global adjustment instructions and partition adjustment instructions, priorities are allocated to obtain a hierarchical regulation and control instruction set, a mapping knowledge base is established by applying different pressure modes for multiple times and is loaded to a regulation and control strategy matching unit, after receiving the positioning result, the unit queries the knowledge base to issue instructions for execution, self-adaptive and accurate control of the probe card pressure is realized, and the pressure uniformity of the contact interface and the test reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of semiconductor testing technology, and in particular to a method and system for integrating pressure sensors with adaptive pressure control of probe cards. Background Technology

[0002] In the field of semiconductor wafer testing, probe cards, as core interface components, directly affect test yield and equipment lifespan due to the pressure uniformity of their contact interfaces. To ensure reliable electrical contact between the probes and chip pads, the conventional approach is to use a mechanical leveling mechanism combined with overall clamping force control, and to perform static calibration of the probe card's flatness using a multi-axis actuator. During testing, some solutions place a small number of discrete piezoelectric sensors on the edges or critical locations of the probe card to monitor the overall contact pressure level. The output signals of these sensors are typically used only to determine whether a preset coarse threshold has been exceeded, triggering overall lifting or tilt correction. In addition, technicians preset a set of fixed pressure adjustment parameters based on experience or offline simulation data, which are reused when testing different batches of chips.

[0003] One of the core flaws of this conventional approach is the lack of fine-grained perception of the pressure field. The extremely low density of discrete sensor placement only reflects pressure values ​​at local measurement points, failing to construct a continuous pressure distribution map across the contact interface. Therefore, it is difficult to accurately identify minute anomalies where pressure exceeds uniformity requirements, let alone trace the transmission paths of these anomalies in the horizontal and vertical directions. When local pressure shifts are caused by the combined deformation of multi-layered structures, the information obtained from a limited number of measurement points is insufficient to distinguish whether the root cause of the pressure anomaly is overall tilt or a difference in local probe height.

[0004] Another drawback is the singularity and lag of pressure control strategies. Most existing technologies rely on overall adjustment commands, correcting flatness by uniformly adjusting the displacement of actuators on each axis, but cannot implement tiered responses for abnormal areas at different spatial scales. For example, a localized peak pressure zone at the center of a probe card and the overall tilt at its edges should require different compensation actions, but traditional methods can only perform a general leveling of the entire area, resulting in the averaging of local problems and even introducing new uneven areas. Furthermore, the determination of adjustment parameters often relies on manual experience or fixed value tables, lacking an adaptive learning mechanism from historical pressure anomaly characteristics to optimized control commands. This makes it impossible to quickly identify and specifically address recurring anomaly patterns, limiting both control efficiency and accuracy. Summary of the Invention

[0005] This invention provides a pressure sensor integration method and system for probe card pressure adaptive control, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a pressure sensor integration method for probe card pressure adaptive control, comprising:

[0007] A pressure sensing unit array is set at the probe card, and the capacitance signal output by the pressure sensing unit array is collected. By jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal, the pressure field state information is obtained.

[0008] Based on the pressure field state information, abnormal areas in the probe card contact interface where the pressure exceeds the uniformity threshold and the transmission paths of the abnormal areas in the vertical and horizontal directions are identified to obtain the pressure anomaly source location result.

[0009] Based on the spatial scale and transmission path of the abnormal area in the pressure anomaly source location result, the pressure adjustment requirement is decomposed into a global adjustment command for the overall tilt attitude of the probe card and a partitioned adjustment command for the local abnormal area, and time priority is assigned to each to obtain a hierarchical control command set.

[0010] Apply different pressure modes multiple times and collect the corresponding pressure field state information and pressure anomaly source location results to establish a mapping knowledge base between pressure anomaly characteristics and the optimal hierarchical control instruction set, and load the mapping knowledge base into the control strategy matching unit;

[0011] After receiving the pressure anomaly source location result, the control strategy matching unit queries the mapping knowledge base to obtain the corresponding hierarchical control instruction set, sends it to the multi-axis pressure actuator, and executes it.

[0012] The capacitance signal output from the pressure sensing unit array is acquired, and the pressure field state information is obtained by jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal, including:

[0013] The pressure sensing unit array is spatially divided into multiple sub-regions and local statistical analysis is performed on the capacitance signal to obtain the average pressure amplitude of each sub-region. The difference between the average pressure amplitudes of adjacent sub-regions is calculated to construct a pressure spatial gradient field. The pressure concentration feature locations in the pressure spatial gradient field where the gradient amplitude exceeds a preset gradient threshold are extracted.

[0014] The capacitance signal is sampled in time series, the rate of change of capacitance signal of each sensing unit in a continuous time window is calculated, and the sensing unit whose rate of change of capacitance signal exceeds a preset rate of change threshold is identified as the pressure dynamic fluctuation characteristic position and the spectral component of the corresponding rate of change of capacitance signal is recorded.

[0015] By combining the spectral components corresponding to the spatially overlapping regions of the pressure concentration feature locations and the pressure dynamic fluctuation feature locations, the pressure evolution rate and oscillation period of the overlapping regions are analyzed.

[0016] The overlapping region, the pressure evolution rate, the oscillation period, the gradient direction information of the pressure spatial gradient field, and the frequency characteristic information of the spectral components are combined to form a comprehensive feature vector characterizing the spatial non-uniformity and temporal evolution trend of the pressure field. The pressure field state information is generated based on the comprehensive feature vector.

[0017] Based on the pressure field state information, abnormal regions in the probe card contact interface where the pressure exceeds the uniformity threshold are identified, along with the transmission paths of these abnormal regions in the vertical and horizontal directions. The pressure anomaly source localization results include:

[0018] The Moran index of the pressure amplitude between each sensing unit in the pressure sensing unit array at the probe card contact interface and the surrounding neighboring sensing units is used to identify the position of the sensing unit where the sign of the Moran index changes as the pressure spatial aggregation mode transition boundary. The average difference of the pressure amplitude of the sensing units on both sides of the pressure spatial aggregation mode transition boundary is calculated as the uniformity threshold.

[0019] When the deviation between the pressure amplitude of each sensing unit within the pressure concentration feature location and the average pressure amplitude exceeds the uniformity threshold, the pressure concentration feature location is confirmed as an abnormal area.

[0020] Based on the gradient direction information of the pressure spatial gradient field, a horizontal transmission path is constructed by tracing the sensor unit sequence with progressively decreasing pressure amplitude outward along the gradient direction of the abnormal region; based on the pressure evolution rate and the oscillation period, the frequency bands of the spectral components corresponding to the abnormal region are divided and the energy ratio of each frequency band is calculated; based on the energy ratio, the attenuation characteristics of pressure transmission to the inner layer in the vertical layered structure of the identification probe card are constructed to form a vertical transmission path.

[0021] The spatial location information of the abnormal area, the coordinate information of each sensing unit sequence in the horizontal transmission path, and the energy proportion information of each frequency band in the vertical transmission path are combined to generate the pressure anomaly source location result.

[0022] Based on the pressure evolution rate and the oscillation period, the spectral components corresponding to the abnormal region are divided into frequency bands and the energy ratio of each frequency band is calculated. Based on the energy ratio, the attenuation characteristics of pressure transmission to the inner layer in the vertical layered structure of the identification probe card are used to construct the vertical transmission path, including:

[0023] The characteristic frequency of pressure amplitude change with time in the abnormal region is calculated based on the pressure evolution rate. The periodic frequency component of pressure oscillation in the abnormal region is calculated based on the oscillation period. The characteristic frequency and the periodic frequency component are used as the frequency band division boundary. The spectral component corresponding to the abnormal region is divided into multiple frequency bands. The sum of the squares of the amplitudes of each frequency component in each frequency band is calculated as the frequency band energy. The ratio of the frequency band energy to the total energy is calculated as the energy proportion of each frequency band.

[0024] A mapping relationship between the energy proportion of each layer and frequency band in the vertical layered structure of the probe card is established. The energy proportion of each frequency band is mapped to the pressure transmission characteristics of the probe card contact interface layer, intermediate support layer, and substrate layer, respectively. Based on the order of the energy proportion of each frequency band from high to low, the dominant path of pressure transmission in the vertical layered structure from the probe card contact interface layer to the intermediate support layer and then to the substrate layer is determined. The energy proportion of each frequency band in the dominant path is extracted to construct the attenuation sequence of pressure transmission layer by layer in the vertical direction as the transmission path in the vertical direction.

[0025] Based on the spatial scale and transmission path of the abnormal region in the pressure anomaly source localization results, the pressure adjustment requirement is decomposed into a global adjustment command for the overall tilt attitude of the probe card and a regional adjustment command for the local abnormal region. These are then assigned time priorities, resulting in a hierarchical control command set including:

[0026] Calculate the spatial offset vector between the geometric center of the abnormal region and the geometric center of the probe card contact interface, and perform principal component decomposition to extract the dominant offset direction;

[0027] When the spatial scale of the abnormal region exceeds a preset area ratio threshold of the probe card contact interface area and the angle between the dominant offset direction and the edge of the probe card contact interface is less than a preset angle threshold, a global adjustment command for the overall tilt attitude of the probe card is generated; otherwise, the boundary of the influence range of the abnormal region is determined based on the horizontal transmission path, and the attenuation coefficient of the pressure transmission depth to the internal layered structure of the probe card is calculated based on the energy ratio information of each frequency band in the vertical transmission path, and a partition adjustment command for the local abnormal region is generated.

[0028] Set timing constraints for the execution of the partition adjustment instruction after the global adjustment instruction has been executed;

[0029] The global adjustment instruction, the partition adjustment instruction, and the timing constraint are combined to generate the hierarchical control instruction set.

[0030] By applying different pressure modes multiple times and collecting the corresponding pressure field state information and pressure anomaly source location results, a mapping knowledge base is established between pressure anomaly characteristics and the optimal hierarchical control instruction set, including:

[0031] By applying uniform pressure mode, unilateral tilt pressure mode, and local concentrated pressure mode through a multi-axis pressure actuator, the pressure field state information corresponding to each pressure mode is collected. The spatial location information of the abnormal area corresponding to the pressure field state information is used to construct a spatial feature vector, the coordinate information of each sensing unit sequence in the horizontal transmission path is used to construct a transmission path feature vector, and the energy ratio information of each frequency band in the vertical transmission path is used to construct a spectral feature vector, thereby obtaining the pressure anomaly characteristics.

[0032] The execution parameters of the hierarchical control instruction set are adjusted according to the pressure anomaly characteristics corresponding to each pressure mode and executed repeatedly. The standard deviation of pressure amplitude in the pressure field state information after execution is calculated, and the hierarchical control instruction set with the smallest standard deviation of pressure amplitude is marked as the optimal hierarchical control instruction set for the corresponding pressure anomaly characteristics.

[0033] The variance contribution of the spatial feature vector, the transmission path feature vector, and the spectral feature vector is calculated. Based on the variance contribution, each feature vector is weighted and fused to generate a pressure anomaly feature index key. The corresponding optimal hierarchical control instruction set is used as the mapping value of the pressure anomaly feature index key to establish a mapping entry. Multiple mapping entries are organized into an index structure based on the pressure anomaly feature index key to form the mapping knowledge base.

[0034] A second aspect of the present invention provides a pressure sensor integrated system for probe card pressure adaptive control, comprising:

[0035] The pressure sensing unit is used to set up a pressure sensing unit array at the probe card, collect the capacitance signal output by the pressure sensing unit array, and obtain pressure field state information by jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal.

[0036] An anomaly localization unit is used to identify, based on the pressure field state information, an anomaly region in the probe card contact interface where the pressure exceeds the uniformity threshold and the transmission path of the anomaly region in the vertical and horizontal directions, thereby obtaining the pressure anomaly source localization result.

[0037] The hierarchical control unit is used to decompose the pressure adjustment requirement into a global adjustment command for the overall tilt attitude of the probe card and a partitioned adjustment command for the local abnormal area based on the spatial scale and transmission path of the abnormal area in the pressure anomaly source location result, and to assign time priority to each command to obtain a hierarchical control command set.

[0038] The mapping library unit is used to apply different pressure modes multiple times and collect the corresponding pressure field state information and pressure anomaly source location results, establish a mapping knowledge base between pressure anomaly characteristics and the optimal hierarchical control instruction set, and load the mapping knowledge base into the control strategy matching unit.

[0039] The control matching unit is used to query the mapping knowledge base to obtain the corresponding hierarchical control instruction set after the control strategy matching unit receives the pressure anomaly source location result, and then send it to the multi-axis pressure actuator for execution.

[0040] A third aspect of the present invention provides an electronic device, comprising:

[0041] processor;

[0042] Memory used to store processor-executable instructions;

[0043] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0044] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0045] The beneficial effects of this application are as follows:

[0046] The pressure sensing unit array performs joint analysis of the spatial distribution and temporal changes of the capacitance signal, enabling high-resolution reconstruction of the pressure field state at the probe card contact interface. This significantly improves the accuracy and real-time performance of pressure detection, avoiding the blind spots of traditional single-point detection. The transmission paths of abnormal regions in the vertical and horizontal directions are precisely located, allowing the identification of pressure anomaly sources to shift from empirical judgment to quantitative analysis, providing clear targets for subsequent adjustments.

[0047] The decomposition of global and regional adjustment commands, combined with timing priority allocation, enables coordinated correction of tilt attitude and local areas, avoiding excessive intervention of normal areas by global adjustments. The hierarchical control command set enables the actuator to respond quickly to pressure deviations of different scales, keeping the uniformity of the contact interface within a small fluctuation range, significantly reducing probe card wear and chip test failures caused by uneven pressure.

[0048] The mapping knowledge base is built through training under multiple stress patterns, directly linking abnormal stress characteristics with optimal control strategies. The control strategy matching unit queries the knowledge base in real time and issues instructions, eliminating the need for iterative trial-and-error processes and enabling predictive and memory-based adaptive stress adjustment. This method reduces adjustment time by several times, while improving the long-term stability and test yield of the probe card and reducing maintenance costs. Attached Figure Description

[0049] Figure 1 A schematic diagram of the pressure sensor integration method for probe card pressure adaptive control;

[0050] Figure 2 This is a schematic diagram of the pressure field state information extraction process. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0053] Figure 1 This is a schematic flowchart of the pressure sensor integration method for probe card pressure adaptive control according to an embodiment of the present invention.

[0054] The pressure sensor integration method for probe card pressure adaptive control includes:

[0055] A pressure sensing unit array is set at the probe card, and the capacitance signal output by the pressure sensing unit array is collected. By jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal, the pressure field state information is obtained.

[0056] Based on the pressure field state information, abnormal areas in the probe card contact interface where the pressure exceeds the uniformity threshold and the transmission paths of the abnormal areas in the vertical and horizontal directions are identified to obtain the pressure anomaly source location result.

[0057] Based on the spatial scale and transmission path of the abnormal area in the pressure anomaly source location result, the pressure adjustment requirement is decomposed into a global adjustment command for the overall tilt attitude of the probe card and a partitioned adjustment command for the local abnormal area, and time priority is assigned to each to obtain a hierarchical control command set.

[0058] Apply different pressure modes multiple times and collect the corresponding pressure field state information and pressure anomaly source location results to establish a mapping knowledge base between pressure anomaly characteristics and the optimal hierarchical control instruction set, and load the mapping knowledge base into the control strategy matching unit;

[0059] After receiving the pressure anomaly source location result, the control strategy matching unit queries the mapping knowledge base to obtain the corresponding hierarchical control instruction set, sends it to the multi-axis pressure actuator, and executes it.

[0060] In one optional implementation, the capacitance signal output by the pressure sensing unit array is acquired, and the pressure field state information is obtained by jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal, including:

[0061] The pressure sensing unit array is spatially divided into multiple sub-regions and local statistical analysis is performed on the capacitance signal to obtain the average pressure amplitude of each sub-region. The difference between the average pressure amplitudes of adjacent sub-regions is calculated to construct a pressure spatial gradient field. The pressure concentration feature locations in the pressure spatial gradient field where the gradient amplitude exceeds a preset gradient threshold are extracted.

[0062] The capacitance signal is sampled in time series, the rate of change of capacitance signal of each sensing unit in a continuous time window is calculated, and the sensing unit whose rate of change of capacitance signal exceeds a preset rate of change threshold is identified as the pressure dynamic fluctuation characteristic position and the spectral component of the corresponding rate of change of capacitance signal is recorded.

[0063] By combining the spectral components corresponding to the spatially overlapping regions of the pressure concentration feature locations and the pressure dynamic fluctuation feature locations, the pressure evolution rate and oscillation period of the overlapping regions are analyzed.

[0064] The overlapping region, the pressure evolution rate, the oscillation period, the gradient direction information of the pressure spatial gradient field, and the frequency characteristic information of the spectral components are combined to form a comprehensive feature vector characterizing the spatial non-uniformity and temporal evolution trend of the pressure field. The pressure field state information is generated based on the comprehensive feature vector.

[0065] like Figure 2 As shown, the method includes:

[0066] A pressure sensing unit array is arranged in the contact area between the probe card and the wafer chip. The array uses a flexible substrate to support the capacitive pressure sensing units. The spacing between the sensing units is determined according to the probe distribution density, typically ranging from 0.3 mm to 1 mm. Each pressure sensing unit consists of two layers of metal electrodes and an intermediate dielectric layer. When pressure is applied, the thickness of the dielectric layer decreases, resulting in an increase in capacitance. The change in capacitance is linearly related to the applied pressure. The sensing unit array is connected to a signal conditioning module via signal lines led out from the flexible circuit board. The signal conditioning module converts the weak capacitance change signal into a voltage signal and amplifies and filters it.

[0067] During data acquisition, a multiplexer sequentially selects each sensing unit, and a capacitance-to-digital converter (CDC) chip converts the capacitance signal into a digital value at a sampling frequency ranging from 100Hz to 1000Hz. The sampling frequency is determined based on the dynamic response time of the probe card's contact process and must satisfy the Nyquist sampling theorem to capture the transient characteristics of pressure changes. The acquired capacitance digital signal is calibrated and converted into a pressure amplitude. The calibration process is achieved by measuring the capacitance output under a known pressure load and establishing a fitting curve.

[0068] The pressure sensing unit array is spatially divided into multiple sub-regions. The specific division method can be determined based on the geometry of the probe card and the distribution density of the sensing units. For example, when the sensing unit array is arranged in a 16×16 matrix, it can be divided into 16 sub-regions of 4×4, with each sub-region containing 4×4 sensing units. After division, local statistical analysis is performed on the capacitance signals output by all sensing units in each sub-region. The statistical analysis includes calculating the arithmetic mean of all capacitance signals in the sub-region, which is used as the average pressure amplitude of that sub-region. In this way, the originally discrete sensing unit signals can be transformed into a pressure amplitude distribution that is representative of the region.

[0069] After obtaining the average pressure amplitude of each sub-region, the difference in average pressure amplitude between adjacent sub-regions is calculated. Adjacency relationships include horizontal, vertical, and diagonal adjacency. For two horizontally adjacent sub-regions, the difference in their average pressure amplitude reflects the horizontal pressure change trend; for two vertically adjacent sub-regions, the difference reflects the vertical pressure change trend. The differences in average pressure amplitude between all adjacent sub-regions are organized according to their spatial location to construct a pressure spatial gradient field. Each location in this gradient field corresponds to a gradient vector, the direction of which points in the direction of increasing pressure, and the magnitude of the gradient vector represents the degree of pressure change.

[0070] In the pressure gradient field, locations where the gradient amplitude exceeds a preset gradient threshold are extracted as pressure concentration feature locations. The preset gradient threshold can be determined according to the probe card design specifications and testing requirements, and is usually set to 1.5 to 2 times the normal pressure gradient amplitude. Pressure concentration feature locations often correspond to areas of uneven pressure distribution on the probe card contact interface, caused by local deformation of the probe card, inconsistent probe height, or unevenness of the tested chip surface. Recording the spatial coordinates of these pressure concentration feature locations, as well as the corresponding gradient amplitude and gradient direction, provides a spatial basis for subsequent pressure anomaly source localization.

[0071] Based on spatial analysis, the capacitance signal is sampled over time. The sampling frequency is determined according to the dynamic characteristics of the probe card contact process, typically set between 100Hz and 1000Hz. For each sensing unit, multiple capacitance signal samples are acquired within consecutive time windows. The length of the time window can be set from 50 milliseconds to 200 milliseconds to ensure that the dynamic process of pressure change can be captured. Within each time window, the rate of change of the capacitance signal is calculated. The rate of change of the capacitance signal is defined as the ratio of the increment of the capacitance signal within the time window to the length of the time window; this parameter reflects the speed at which the pressure changes over time.

[0072] Sensing units whose capacitance signal change rate exceeds a preset threshold are identified as pressure dynamic fluctuation characteristic locations. The preset threshold can be determined based on the stability requirements of the probe card contact process, typically set to 2 to 3 times the normal change rate. Pressure dynamic fluctuation characteristic locations often correspond to transient pressure fluctuations occurring during probe card contact, caused by mechanical vibration, friction between the probe and the pad, or microscopic slippage at the contact interface. The spatial coordinates of these pressure dynamic fluctuation characteristic locations and the corresponding capacitance signal change rate values ​​are recorded.

[0073] Spectral analysis was performed on the capacitance signal corresponding to the characteristic locations of dynamic pressure fluctuations. A Fast Fourier Transform (FFT) was used to convert the time-domain capacitance signal into frequency-domain spectral components. These spectral components contain amplitude and phase information for different frequency components. By analyzing these spectral components, the main frequency characteristics of the pressure fluctuations can be identified. For example, low-frequency components typically correspond to slow pressure changes across the probe card as a whole, mid-frequency components correspond to periodic oscillations in localized areas, and high-frequency components correspond to microscopic friction or impacts at the contact interface. The top three frequency components with the largest amplitudes and their corresponding amplitudes in the spectral components corresponding to each characteristic location of dynamic pressure fluctuations were recorded.

[0074] This study combines the spatial overlap of pressure concentration and dynamic pressure fluctuation characteristics. Overlapping regions are spatial locations that simultaneously satisfy both pressure concentration and dynamic pressure fluctuation characteristics. These regions are often the most critical anomaly areas at the probe card contact interface because they exhibit not only uneven pressure distribution but also dynamic pressure fluctuations. For each overlapping region, its corresponding spectral components are extracted to analyze the frequency characteristics of pressure fluctuations in that region.

[0075] The pressure evolution rate and oscillation period of the overlapping region were analyzed. The pressure evolution rate is defined as the average rate of change of pressure amplitude over time within the overlapping region, which can be obtained by weighted averaging of the capacitance signal change rates of all sensing units within the region. The oscillation period is defined as the main period of pressure fluctuations within the overlapping region, which can be obtained by taking the reciprocal of the frequency component with the largest amplitude in the spectral components. The pressure evolution rate reflects the speed of pressure change in the region, and the oscillation period reflects the time scale of pressure fluctuations in the region. These two parameters are of great significance for judging the severity of pressure anomalies and formulating corresponding control strategies.

[0076] A comprehensive feature vector is constructed by combining the spatial coordinates of the overlapping region, the pressure evolution rate, the oscillation period, the gradient direction information of the pressure spatial gradient field, and the frequency characteristics of the spectral components. This comprehensive feature vector is a multi-dimensional vector, where each dimension corresponds to a feature parameter. For example, the comprehensive feature vector can be represented as a vector containing multiple elements, including the x and y coordinates of the center of the overlapping region, the pressure evolution rate, the oscillation period, the angle of the gradient direction, and the frequency and amplitude of the dominant frequency components. This comprehensive feature vector fully describes the spatial non-uniformity and temporal evolution trend of the pressure field, providing a complete information foundation for subsequent pressure anomaly source localization and control strategy formulation.

[0077] Pressure field state information is generated based on comprehensive feature vectors. This information provides a comprehensive description of the pressure distribution at the probe card's contact interface, including the number and distribution of pressure concentration features, the number and distribution of dynamic pressure fluctuation features, the number and characteristic parameters of overlapping regions, the overall distribution pattern of the pressure spatial gradient field, and the frequency characteristics of pressure fluctuations. The pressure field state information is stored in a structured data format, facilitating processing by subsequent pressure anomaly source localization algorithms and control strategy matching algorithms. This joint analysis method of spatial distribution features and temporal variation features allows for the extraction of key information reflecting the essential characteristics of the pressure field from complex capacitance signals, providing accurate and reliable state perception capabilities for probe card pressure adaptive control.

[0078] In practical applications, the spatial resolution and temporal sampling frequency of the pressure sensing unit array need to be optimized based on the specific specifications and testing requirements of the probe card. Higher spatial resolution captures richer details of the pressure distribution, but also increases the computational load for data processing. Higher temporal sampling frequency captures more refined dynamic changes in pressure, but also increases the burden of data storage and transmission. In actual engineering implementation, a trade-off must be struck between sensing accuracy and computational efficiency, selecting appropriate spatial resolution and temporal sampling frequency to ensure timely and accurate acquisition of pressure field state information, providing effective support for the probe card's adaptive pressure control.

[0079] In one optional implementation, based on the pressure field state information, abnormal regions in the probe card contact interface where the pressure exceeds the uniformity threshold and the transmission paths of the abnormal regions in the vertical and horizontal directions are identified to obtain the pressure anomaly source localization result, including:

[0080] The Moran index of the pressure amplitude between each sensing unit in the pressure sensing unit array at the probe card contact interface and the surrounding neighboring sensing units is used to identify the position of the sensing unit where the sign of the Moran index changes as the pressure spatial aggregation mode transition boundary. The average difference of the pressure amplitude of the sensing units on both sides of the pressure spatial aggregation mode transition boundary is calculated as the uniformity threshold.

[0081] When the deviation between the pressure amplitude of each sensing unit within the pressure concentration feature location and the average pressure amplitude exceeds the uniformity threshold, the pressure concentration feature location is confirmed as an abnormal area.

[0082] Based on the gradient direction information of the pressure spatial gradient field, a horizontal transmission path is constructed by tracing the sequence of sensing units with progressively decreasing pressure amplitude outward along the gradient direction of the abnormal region.

[0083] Based on the pressure evolution rate and the oscillation period, the spectral components corresponding to the abnormal region are divided into frequency bands and the energy ratio of each frequency band is calculated. Based on the energy ratio, the attenuation characteristics of pressure transmission to the inner layer in the vertical layered structure of the identification probe card are used to construct the vertical transmission path.

[0084] The spatial location information of the abnormal area, the coordinate information of each sensing unit sequence in the horizontal transmission path, and the energy proportion information of each frequency band in the vertical transmission path are combined to generate the pressure anomaly source location result.

[0085] For any sensing unit on the probe card's contact interface, its neighborhood is determined. This neighborhood can be a four-neighborhood, an eight-neighborhood, or a larger spatial window, the specific choice depending on the density of the sensing unit array and the size of the probe card. In a typical application scenario, if the spacing between the sensing unit arrays is 0.5 mm, an eight-neighborhood can be chosen as the calculation range, encompassing the eight adjacent sensing units surrounding the target sensing unit. For each sensing unit, its pressure amplitude is calculated as the product of the pressure amplitudes of its neighboring sensing units, and these products are then summed with weights. The weighting coefficients are determined based on the spatial distance; sensing units closer to each other have a higher weight. The weighted sum is divided by the variance of the pressure amplitudes of all sensing units to obtain the Moran's index for that sensing unit.

[0086] The Moran's index has a clear physical meaning. A positive value indicates a positive correlation between the pressure amplitude of the sensing unit and its neighboring units, meaning the pressure amplitude in the surrounding area is similar, forming a pressure aggregation pattern. A negative value indicates a negative correlation, meaning the pressure amplitude of the sensing unit differs significantly from the surrounding area, forming a pressure dispersion pattern. By traversing the entire array of sensing units and calculating the Moran's index for each unit, a spatial distribution pattern can be observed. In areas with relatively uniform pressure distribution, the Moran's index remains a stable positive value or close to zero. However, at locations where the pressure distribution changes abruptly, the Moran's index changes from positive to negative, or vice versa. These sign changes mark the boundaries of transitions in spatial pressure aggregation patterns.

[0087] After identifying the pressure spatial aggregation pattern transition boundary, it is necessary to further quantify the degree of pressure difference on both sides of the boundary to determine the uniformity threshold. Along the transition boundary, the sensing units on one side of the boundary are divided into one group, and the sensing units on the other side are divided into another group. The average pressure amplitude of the sensing units in both groups is calculated, and the difference between the two average values ​​is the average pressure amplitude difference at that boundary location. Statistics are performed on all identified transition boundaries on the probe card contact interface, and the median or weighted average of the average pressure amplitude differences at all boundary locations is taken as the global uniformity threshold. This threshold reflects the critical degree of difference in pressure distribution on the probe card contact interface as it transitions from a uniform to a non-uniform state.

[0088] After determining the uniformity threshold, anomaly detection is performed on these locations by combining the pressure concentration feature locations extracted from the pressure field state information. For each pressure concentration feature location, the average pressure amplitude of all sensing units within that location is calculated as the representative pressure level of the region. Simultaneously, the deviation between the pressure amplitude of each sensing unit within that region and the average pressure amplitude of the region is calculated. If the deviation of the pressure amplitude of a certain sensing unit from the regional average exceeds the uniformity threshold, then the location of that sensing unit is considered to have a pressure anomaly. Furthermore, the proportion of sensing units exceeding the uniformity threshold within the pressure concentration feature location is statistically analyzed to determine the proportion of the total number of sensing units in that region. When this proportion exceeds a preset detection ratio, such as 30%, the pressure concentration feature location is confirmed as an abnormal region. This detection method considers both the pressure deviation of individual sensing units and the distribution density of abnormal sensing units within the region, avoiding misjudgments caused by noise from individual sensing units.

[0089] After identifying the anomalous region, it is necessary to trace the horizontal transmission path of pressure. The pressure spatial gradient field provides information on the direction of pressure amplitude change in space. For each sensing unit within the anomalous region, adjacent sensing units are searched outwards according to their gradient direction. If the pressure amplitude of an adjacent sensing unit is lower than that of the current sensing unit, the adjacent sensing unit is added to the transmission path sequence. The tracing continues outwards along the gradient direction until a sensing unit whose pressure amplitude no longer decreases is encountered, or the boundary of the sensing unit array is reached. Through this layer-by-layer tracing method, multiple sensing unit sequences radiating outwards from the anomalous region can be constructed, each sequence representing a horizontal transmission path. These paths reflect the spatial distribution characteristics of pressure diffusion from the anomalous region to the surrounding area.

[0090] Based on the horizontal pressure transmission path, it is also necessary to analyze the vertical pressure transmission characteristics. Probe cards typically consist of a multi-layered structure, including a probe layer, an elastic support layer, and a printed circuit board layer. As pressure is transmitted from the contact interface to the inner layers, it attenuates due to differences in the elastic modulus, thickness, and interlayer contact characteristics of the materials in each layer. To identify this vertical transmission characteristic, the pressure evolution rate and oscillation period information from the pressure field state information are utilized. The pressure evolution rate reflects how quickly the pressure changes over time, while the oscillation period reflects the periodicity of pressure fluctuations. Spectral analysis is performed on the pressure time-series signal corresponding to the abnormal region, dividing the spectral components into multiple frequency bands according to frequency range. The low-frequency band corresponds to slow pressure changes and is usually related to the overall mechanical response of the probe card. The mid-frequency band corresponds to periodic pressure fluctuations and is related to the resonance characteristics between the internal layers of the probe card. The high-frequency band corresponds to rapid pressure changes and is usually related to local micro-motions at the contact interface.

[0091] The energy percentage of each frequency band is calculated, which is the sum of the energy of all frequency components within that band divided by the total spectral energy. The energy percentage reflects the contribution of different frequency bands to the pressure signal. A higher energy percentage in the low-frequency band indicates that pressure is mainly transmitted to the inner layer in a slowly varying manner, suggesting that the probe card's inner layer has good pressure buffering capabilities. A higher energy percentage in the high-frequency band indicates that pressure is transmitted to the inner layer in a rapidly fluctuating manner, suggesting that the probe card's inner layer is more sensitive to pressure and has weaker attenuation characteristics. By analyzing the energy percentage of each frequency band, the pressure transmission path in the probe card's vertical layered structure can be deduced. For example, if the low-frequency energy percentage gradually increases from the contact interface towards the inner layer, while the high-frequency energy percentage gradually decreases, it indicates that during pressure transmission to the inner layer, high-frequency components are absorbed layer by layer, while low-frequency components are retained, forming a significant frequency-selective attenuation characteristic. This attenuation characteristic constitutes the vertical transmission path.

[0092] After constructing the horizontal and vertical transmission paths, the spatial location information of the anomaly region, the coordinate information of each sensor unit sequence in the horizontal transmission path, and the energy proportion information of each frequency band in the vertical transmission path are combined to generate the pressure anomaly source localization result. The spatial location information of the anomaly region includes the center coordinates, boundary range, and number of sensor units within the region. The coordinate information of the horizontal transmission path is represented in the form of sensor unit sequences, with each sequence recording the coordinates of sensor units extending outward from the anomaly region and the corresponding pressure amplitude. The energy proportion information of the vertical transmission path is represented in the form of frequency band energy distribution, recording the frequency range of each band and the corresponding energy proportion value. This information together constitutes a complete description of the pressure anomaly source localization result, providing a precise basis for subsequent pressure adjustments.

[0093] In practical applications, the results of pressure anomaly source localization can be organized using data structures. For example, a nested dictionary structure can be employed: the first layer records the identifier and spatial location information of the anomaly region; the second layer records the sequence set of horizontal transmission paths; and the third layer records the frequency band energy distribution of the vertical transmission paths. This structured representation facilitates subsequent data retrieval and the generation of control commands. Simultaneously, the pressure anomaly source localization results can be visualized, marking the anomaly region on a planar map of the probe card contact interface, using different colored arrows to represent the direction and intensity of the horizontal transmission paths, and using bar charts to represent the frequency band energy distribution of the vertical transmission paths. This visualization helps operators intuitively understand the spatial distribution characteristics and transmission mechanisms of pressure anomalies, providing a reference for manual intervention.

[0094] In one optional implementation, based on the pressure evolution rate and the oscillation period, the spectral components corresponding to the abnormal region are divided into frequency bands and the energy proportion of each frequency band is calculated. Based on the energy proportion, the attenuation characteristics of pressure transmission to the inner layer in the vertical layered structure of the identification probe card are used to construct a vertical transmission path, including:

[0095] The characteristic frequency of pressure amplitude change with time in the abnormal region is calculated based on the pressure evolution rate. The periodic frequency component of pressure oscillation in the abnormal region is calculated based on the oscillation period. The characteristic frequency and the periodic frequency component are used as the frequency band division boundary. The spectral component corresponding to the abnormal region is divided into multiple frequency bands. The sum of the squares of the amplitudes of each frequency component in each frequency band is calculated as the frequency band energy. The ratio of the frequency band energy to the total energy is calculated as the energy proportion of each frequency band.

[0096] A mapping relationship between the energy proportion of each layer and frequency band in the vertical layered structure of the probe card is established. The energy proportion of each frequency band is mapped to the pressure transmission characteristics of the probe card contact interface layer, intermediate support layer, and substrate layer, respectively. Based on the order of the energy proportion of each frequency band from high to low, the dominant path of pressure transmission in the vertical layered structure from the probe card contact interface layer to the intermediate support layer and then to the substrate layer is determined. The energy proportion of each frequency band in the dominant path is extracted to construct the attenuation sequence of pressure transmission layer by layer in the vertical direction as the transmission path in the vertical direction.

[0097] After obtaining the pressure evolution rate and oscillation period of the anomalous region, further analysis of the pressure signal distribution characteristics in the frequency domain is needed to reveal the pressure transmission mechanism within the vertical layered structure of the probe card. The pressure evolution rate reflects the rate of change of pressure amplitude over time within the anomalous region, and can be obtained by performing a first-order difference on the pressure time series and calculating the root mean square of the difference. Specifically, for a specific location within the anomalous region at consecutive time points... Collected pressure value sequence Calculate the pressure change between adjacent time points. ,in Values ​​range from 1 to Then calculate the pressure evolution rate. The higher the rate value, the more drastic the pressure change, and the higher the proportion of high-frequency components in the corresponding spectrum.

[0098] The oscillation period describes the time interval of periodic fluctuations in the pressure signal. By performing autocorrelation analysis on the pressure time series, the time delay at which the first significant peak of the autocorrelation function appears is identified; this delay is the oscillation period. The reciprocal of the oscillation period corresponds to the oscillation frequency. This frequency reflects the dominant frequency of the periodic component in the pressure signal. During probe card testing, oscillation phenomena typically originate from the elastic deformation recovery at the probe-wafer contact interface, mechanical resonance of the probe card's internal structure, or coupling with external vibration interference.

[0099] Based on the pressure evolution rate and oscillation period, the key boundaries for frequency band division can be determined. The characteristic frequencies corresponding to the pressure evolution rate are then... As the first dividing boundary, in This represents the standard deviation of pressure values ​​within the abnormal region; this frequency marks the boundary between rapidly changing and slowly changing pressure components. The oscillation frequency... As the second dividing boundary, this frequency marks the center frequency of the periodic oscillation component. Based on these two boundaries, the spectrum is divided into low-frequency, mid-frequency, and high-frequency bands. The low-frequency band ranges from 0 to... It mainly contains a trend component of slowly changing pressure; the mid-frequency range is arrive It includes components that exhibit transitional pressure changes; the high-frequency range is... The above includes rapid pressure oscillations and noise components.

[0100] A Fast Fourier Transform (FFT) is performed on the pressure time series of the anomalous region to obtain its frequency domain representation. Within each frequency band, the amplitudes of all frequency components are extracted, and the sum of the squares of the amplitudes is calculated as the energy of that frequency band. Assume that the low-frequency band contains... Each frequency component has an amplitude of [number] respectively. Low-frequency energy Similarly, calculate the mid-frequency energy. With high-frequency energy Total energy The energy percentages of each frequency band are as follows: , , The energy percentage reflects the degree to which different frequency components contribute to the pressure signal, and also indirectly reflects the distribution characteristics of pressure in different structural layers.

[0101] The vertical layered structure of a probe card typically comprises three main layers: the probe card contact interface layer, the intermediate support layer, and the substrate layer. The probe card contact interface layer, located on the outermost side, directly contacts the wafer. This layer experiences the most dramatic pressure changes and has the fastest response time, corresponding to the high-frequency components of the spectrum. The intermediate support layer, situated between the contact interface layer and the substrate layer, acts as a buffer and transfer layer. Its response to pressure exhibits some hysteresis and attenuation, corresponding to the mid-frequency components of the spectrum. The substrate layer, located on the innermost side, provides overall rigid support. Pressure changes in this layer are the slowest and have the smallest amplitude, corresponding to the low-frequency components of the spectrum.

[0102] Establish a mapping relationship between the energy proportion of each frequency band and the vertical layered structure. When the energy proportion of the high-frequency band... At its maximum, this indicates that the pressure is mainly concentrated at the probe card contact interface layer, and has not yet been transmitted to the inner layers or the transmission is weak. (The text then abruptly shifts to a discussion of mid-frequency energy percentage.) At its maximum, this indicates that pressure has been transferred from the contact interface layer to the intermediate support layer, resulting in a significant dynamic response at that layer. When the low-frequency energy proportion... At its maximum, this indicates that the pressure has been deeply transmitted to the substrate layer, and the entire probe card structure is in a relatively stable stress state. By comparing the energy proportions of the three frequency bands, the vertical transmission depth of the pressure and the dominant layer can be determined.

[0103] Based on the ranking of energy proportions in each frequency band from high to low, the dominant pressure transmission path is determined. If the ranking result is... The dominant path is from the probe card contact interface layer to the intermediate support layer, but it has not yet fully reached the substrate layer. At this point, the pressure anomaly is mainly caused by localized unevenness at the contact interface. If the sorting result is... If the dominant path is that the pressure has already formed a dominant response in the intermediate support layer, it is caused by insufficient stiffness or uneven distribution of the support structure. If the sorting result is... If the pressure has been transmitted to the substrate layer, it is caused by a problem with the flatness of the substrate or a deviation in the overall mounting posture.

[0104] The energy percentage of each frequency band in the dominant path is extracted to construct an attenuation sequence of pressure transmission layer by layer in the vertical direction. This attenuation sequence is arranged according to the physical order of pressure transmission, i.e., from the probe card contact interface layer to the intermediate support layer and then to the substrate layer. Let the energy percentage of the high-frequency band corresponding to the probe card contact interface layer be... The intermediate support layer corresponds to the mid-frequency energy ratio. The energy percentage of the substrate layer in the low-frequency band The decay sequence is then expressed as The numerical trend of this sequence reflects the energy attenuation law during the vertical transmission of pressure. If... Much larger and This indicates that pressure is largely absorbed or reflected at the contact interface layer, resulting in low efficiency in its transmission to the inner layers. If the three values ​​are close, it indicates that pressure is transmitted relatively evenly between the layers, and the structure has weak damping characteristics.

[0105] The attenuation sequence can be further quantized into attenuation coefficients. The attenuation coefficient is defined from the probe card contact interface layer to the intermediate support layer. Attenuation coefficient from the intermediate support layer to the substrate layer A smaller attenuation coefficient indicates greater energy loss during pressure transmission between layers. Analyzing the attenuation coefficient can identify weak points in the probe card structure. For example, if... Significantly smaller than This indicates that the intermediate support layer has a strong barrier effect against pressure, and the material or structural design of the support layer needs to be optimized.

[0106] In practical applications, for the same anomalous region, the frequency band division and energy percentage calculation can be repeated at different times to observe the temporal evolution characteristics of the attenuation sequence. If the attenuation sequence changes significantly within a short period, it indicates that the pressure transmission path is unstable, suggesting contact loosening or structural fatigue. If the attenuation sequence remains stable, it indicates that the pressure transmission path is fixed, and targeted regulation can be implemented for that path.

[0107] The vertical transmission path constructed using the above method not only reveals the pressure transmission mechanism within the probe card but also provides a basis for subsequent layered control. Different control strategies can be designed for different transmission path characteristics. For example, when the pressure is mainly concentrated in the contact interface layer, the contact pressure distribution of the probe can be adjusted first; when the pressure has been transmitted to the substrate layer, the overall mounting orientation of the probe card needs to be adjusted.

[0108] In one optional implementation, based on the spatial scale and transmission path of the abnormal region in the pressure anomaly source localization result, the pressure adjustment requirement is decomposed into a global adjustment command for the overall tilt attitude of the probe card and a partitioned adjustment command for the local abnormal region, and time priority is assigned to each, resulting in a hierarchical control command set including:

[0109] Calculate the spatial offset vector between the geometric center of the abnormal region and the geometric center of the probe card contact interface, and perform principal component decomposition to extract the dominant offset direction;

[0110] When the spatial scale of the abnormal region exceeds a preset area ratio threshold of the probe card contact interface area and the angle between the dominant offset direction and the edge of the probe card contact interface is less than a preset angle threshold, a global adjustment command for the overall tilt attitude of the probe card is generated; otherwise, the boundary of the influence range of the abnormal region is determined based on the horizontal transmission path, and the attenuation coefficient of the pressure transmission depth to the internal layered structure of the probe card is calculated based on the energy ratio information of each frequency band in the vertical transmission path, and a partition adjustment command for the local abnormal region is generated.

[0111] Set timing constraints for the execution of the partition adjustment instruction after the global adjustment instruction has been executed;

[0112] The global adjustment instruction, the partition adjustment instruction, and the timing constraint are combined to generate the hierarchical control instruction set.

[0113] To calculate the geometric center of the abnormal region, the coordinates of all pressure sensing units marked as abnormal in the pressure sensing unit array are weighted and averaged, with the weight being the pressure deviation value of each sensing unit. This yields the coordinates of the geometric center of the abnormal region. Simultaneously calculate the geometric center coordinates of the probe card contact interface. This coordinate is typically the center point of the rectangular area of ​​the contact interface. The spatial offset vector between them is... The magnitude of the vector It reflects the degree of deviation of the abnormal region from the center of the contact interface, while the vector direction indicates the spatial distribution trend of the abnormal region.

[0114] Principal component decomposition is performed on the spatial offset vector to extract the dominant offset direction. A coordinate matrix is ​​constructed from the coordinates of all abnormal sensing units, and covariance analysis is performed on this matrix to calculate the eigenvalues ​​and eigenvectors. The eigenvector corresponding to the largest eigenvalue is the dominant offset direction. This direction reflects the main spatial extension direction of the anomaly region. The angles between the dominant offset direction and each edge of the probe card contact interface are calculated, and the minimum angle value is taken. As a basis for judgment.

[0115] Determine the spatial scale of the abnormal region and count the total number of sensor units marked as abnormal. Calculate its proportion of the total number of sensing units within the probe card contact interface. proportion This ratio reflects the spatial coverage of the abnormal area. A preset area ratio threshold is set. The value typically ranges from 0.3 to 0.5, meaning that when the abnormal area covers more than 30% to 50% of the contact interface, the abnormality is considered to have a global characteristic. A preset angle threshold is also set. The value is usually between 15 and 25 degrees, and is used to determine whether abnormal areas are distributed along the edge of the contact interface.

[0116] When satisfied and Under these two conditions, the pressure anomaly is determined to be primarily caused by the overall tilt attitude of the probe card. A global adjustment command is then generated to adjust the overall attitude of the probe card. This global adjustment command contains three key parameters: tilt adjustment axis, tilt adjustment angle, and adjustment execution rate. The tilt adjustment axis is determined based on the dominant offset direction. Once determined, select an axis perpendicular to this direction as the rotation axis. The tilt adjustment angle is based on the magnitude of the spatial offset vector. Based on the probe card thickness estimation, the angle adjustment needs to compensate for the height difference of the contact surface caused by tilting. The adjustment execution rate should be set to a slower rate to avoid impact loads caused by rapid adjustments.

[0117] When the above conditions are not met, the pressure anomaly is determined to be a localized anomaly, requiring the generation of a zone adjustment command. The boundary of the anomaly's influence area is determined based on the horizontal transmission path. The horizontal transmission path describes how the pressure anomaly spreads within the contact interface plane. Starting from the geometric center of the anomaly area, it expands outward layer by layer along the horizontal direction. When the pressure deviation value decays to within the uniformity threshold, this location is the boundary of the influence area. By connecting the boundary points in each direction, a closed contour of the anomaly's influence area is formed. All sensing units within this contour constitute the target area requiring zone adjustment.

[0118] The attenuation characteristics of pressure transmission to the internal layered structure of the probe card are calculated based on the vertical transmission path. The energy proportion information of each frequency band in the vertical transmission path was obtained during the pressure anomaly source localization phase. Low-frequency energy proportion... This reflects the intensity of pressure anomaly transmission in deep structures, and the proportion of high-frequency energy. This reflects the degree of pressure anomaly concentration at the surface. A pressure transmission depth attenuation coefficient is defined. This coefficient describes the rate attenuation of pressure anomalies with increasing depth. There is a correlation between the attenuation coefficient and the energy percentage: when... A larger value indicates that the pressure anomaly can penetrate deep into the internal layered structure of the probe card, resulting in a higher attenuation coefficient. Smaller; when A higher value indicates that the pressure anomaly is mainly concentrated on the surface of the contact interface, and the attenuation coefficient is [not specified]. The value is relatively large. By establishing a mapping relationship between energy percentage and attenuation coefficient, it is possible to determine the relationship based on... , , Numerical calculations yielded .

[0119] According to the attenuation coefficient Based on the influence range of the abnormal area, a zonal adjustment command is generated. This command includes three key parameters: adjustment area coordinates, adjustment pressure amplitude, and adjustment depth of action. The adjustment area coordinates are determined by the boundary of the influence range, mapping the closed contour to the coordinate system of the execution unit of the multi-axis pressure actuator. The adjustment pressure amplitude is determined based on the average pressure deviation value within the abnormal area; the pressure value to be compensated is equal to the difference between the target pressure value and the current average pressure value. The adjustment depth of action is determined based on the attenuation coefficient. Confirmed, when When the probe is small, an adjustment force needs to be applied to the internal support structure of the probe card, and the adjustment depth is relatively large; when When the force is large, only the surface of the contact interface needs to be adjusted, and the depth of adjustment is small.

[0120] After generating the global adjustment command and the zonal adjustment command, it is necessary to set the timing constraints between them. Since the overall tilt attitude of the probe card affects the local pressure distribution, the global adjustment command must be executed first to eliminate the overall attitude deviation, and then the zonal adjustment command must be executed to finely adjust the local abnormal areas. The timing constraint rule is set as follows: the execution trigger condition for the zonal adjustment command is that the global adjustment command is completed and the pressure field state information is updated. In specific implementation, a completion flag is set in the global adjustment command. When the multi-axis pressure actuator completes the global adjustment action and returns a completion signal, the pressure sensing unit array is triggered to perform a new round of acquisition to obtain the adjusted pressure field state information. The pressure field state information before and after adjustment is compared to verify the global adjustment effect. If local abnormal areas still exist after the global adjustment, the execution process of the zonal adjustment command is initiated.

[0121] Global adjustment commands, zone adjustment commands, and timing constraint rules are structured to form a hierarchical control command set. This command set is stored using a hierarchical data structure. The first layer is the global adjustment command layer, containing fields such as tilt adjustment axis, tilt adjustment angle, adjustment execution rate, and completion flag. The second layer is the zone adjustment command layer, containing fields such as adjustment area coordinate array, adjustment pressure amplitude array, adjustment depth array, and execution trigger conditions. The third layer is the timing constraint layer, defining the dependencies and execution order between commands at each layer. The hierarchical control command set is encapsulated in a standardized format, facilitating parsing by the control strategy matching unit and execution by multi-axis pressure actuators.

[0122] In practical applications, the generation process of the hierarchical control instruction set needs to consider the mechanical structural characteristics of the probe card and the operational capabilities of the multi-axis pressure actuator. For large-size probe cards, the angular resolution of the overall tilt adjustment is limited by the minimum step angle of the actuator, requiring the calculated theoretical adjustment angle to be quantized into executable discrete angle values. For high-density sensor unit arrays, the partitioned adjustment instructions involve a large number of local adjustment regions, requiring the merging of adjacent small-scale abnormal regions to reduce the number of instructions and improve execution efficiency. Simultaneously, the waiting time in the timing constraints needs to be set according to the viscoelastic properties of the probe card material to ensure that the pressure field reaches a steady state after global adjustment before partitioned adjustment, avoiding misjudgments caused by transitional state interference.

[0123] In one optional implementation, applying different pressure modes multiple times and collecting corresponding pressure field state information and pressure anomaly source location results, and establishing a mapping knowledge base between pressure anomaly characteristics and the optimal hierarchical control instruction set includes:

[0124] By applying uniform pressure mode, unilateral tilt pressure mode, and local concentrated pressure mode through a multi-axis pressure actuator, the pressure field state information corresponding to each pressure mode is collected. The spatial location information of the abnormal area corresponding to the pressure field state information is used to construct a spatial feature vector, the coordinate information of each sensing unit sequence in the horizontal transmission path is used to construct a transmission path feature vector, and the energy ratio information of each frequency band in the vertical transmission path is used to construct a spectral feature vector, thereby obtaining the pressure anomaly characteristics.

[0125] The execution parameters of the hierarchical control instruction set are adjusted according to the pressure anomaly characteristics corresponding to each pressure mode and executed repeatedly. The standard deviation of pressure amplitude in the pressure field state information after execution is calculated, and the hierarchical control instruction set with the smallest standard deviation of pressure amplitude is marked as the optimal hierarchical control instruction set for the corresponding pressure anomaly characteristics.

[0126] The variance contribution of the spatial feature vector, the transmission path feature vector, and the spectral feature vector is calculated. Based on the variance contribution, each feature vector is weighted and fused to generate a pressure anomaly feature index key. The corresponding optimal hierarchical control instruction set is used as the mapping value of the pressure anomaly feature index key to establish a mapping entry. Multiple mapping entries are organized into an index structure based on the pressure anomaly feature index key to form the mapping knowledge base.

[0127] During the pressure mode application phase, three typical pressure modes are applied through a multi-axis pressure actuator. In the uniform pressure mode, each axial actuator outputs the same pressure setpoint, ensuring the probe card contact interface experiences a uniformly distributed pressure load under ideal conditions. The pressure field state information acquired in this mode can serve as a reference state. The unilateral tilt pressure mode adjusts the output pressure of the actuator on one side of the probe card, creating a linear pressure gradient distribution along a specific direction at the contact interface. This mode simulates tilted contact states caused by probe card mounting posture deviations or insufficient substrate flatness. The locally concentrated pressure mode applies a higher pressure load to a specific area of ​​the contact interface than the surrounding area, achieved by controlling some actuators to output larger pressure values. This mode simulates localized pressure concentration caused by probe wear, chip surface protrusions, or particulate contaminants. After each pressure mode is applied, the pressure sensing unit array acquires capacitance signals in real time and converts them into pressure field state information. This information includes the pressure amplitude at each sensing unit location, the spatial gradient of the pressure distribution, and the pressure evolution characteristics over time.

[0128] Based on the collected pressure field state information, an anomaly source localization algorithm is used to identify the anomaly region and its propagation path. The spatial location information of the anomaly region includes the geometric center coordinates, the boundary contour coordinates, and the offset distance and orientation angle of the anomaly region relative to the geometric center of the contact interface. This spatial location information is organized into a spatial feature vector. The dimension of this vector depends on the complexity of the anomalous region. For a single connected anomalous region, the spatial feature vector can be represented as a combination of parameters such as geometric center coordinates, equivalent radius, major axis direction angle, and shape irregularity. The horizontal transmission path reflects the diffusion characteristics of the pressure anomaly within the contact interface plane. The coordinate information of each sensing unit is extracted along the transmission path and arranged in order of expansion from the center of the anomaly source outwards to construct the transmission path feature vector. This vector contains not only the absolute coordinates of the sensing units, but also the spacing between adjacent sensing units, the curvature of the transmission path, and the relative positional relationship between the transmission path and the edge of the contact interface. The vertical transmission path reflects the longitudinal transmission characteristics of pressure anomalies in the multi-layer structure of the probe card. By performing spectral analysis on the pressure signals at different depths, the energy proportions of the low-frequency, mid-frequency, and high-frequency bands are obtained, and a spectral feature vector is constructed. The low-frequency energy proportion reflects the overall transmission efficiency of the pressure anomaly, the mid-frequency energy proportion reflects the damping characteristics of the intermediate support layer, and the high-frequency energy proportion is related to the local stiffness fluctuations of the contact interface. The spatial feature vector, transmission path feature vector, and spectral feature vector are combined to form a complete description of the pressure anomaly characteristics.

[0129] For each pressure mode and its corresponding pressure anomaly characteristics, an optimal hierarchical control instruction set needs to be determined through an iterative optimization process. The hierarchical control instruction set comprises two levels: global adjustment instructions and regional adjustment instructions. Global adjustment instructions control the overall tilt attitude of the probe card by adjusting the reference pressure offset of the multi-axis actuators. Regional adjustment instructions provide refined pressure compensation for local anomaly areas by adjusting the pressure increment of the actuators corresponding to the anomaly areas. Execution parameters include the pressure offset amplitude of the global adjustment instructions, the pressure increment amplitude of the regional adjustment instructions, the execution timing interval between the two levels of instructions, and the pressure adjustment rate limit. Initial execution parameters are estimated based on the amplitude and spatial range of the pressure anomaly characteristics, followed by an iterative optimization loop. In each iteration, the hierarchical control instruction set is issued to the multi-axis pressure actuators according to the current execution parameters and executed. After execution, pressure field state information is re-acquired, and the standard deviation of the pressure amplitude is calculated. As a quantitative indicator of pressure uniformity, the standard deviation of pressure amplitude is calculated by statistically analyzing the pressure measurements of all sensing units within the contact interface. A smaller value indicates a more uniform pressure distribution. Comparing the standard deviation of pressure amplitude under different execution parameters will... The hierarchical control instruction set corresponding to the execution parameter combination that reaches the minimum value is marked as the optimal hierarchical control instruction set for this pressure anomaly characteristic. The iterative optimization process can use optimization algorithms such as gradient descent or genetic algorithms to search for the global optimal solution in the execution parameter space. At the same time, a convergence criterion is set, and the iteration is terminated when the improvement of the standard deviation of the pressure amplitude in multiple consecutive iterations is less than a preset threshold.

[0130] To achieve efficient knowledge base querying and matching, pressure anomaly features need to be converted into structured index keys. The contribution of spatial feature vectors, transmission path feature vectors, and spectral feature vectors to the overall variance of pressure anomaly features is calculated. The variance contribution reflects the importance of each feature vector in distinguishing different pressure anomaly patterns. The calculation method involves performing principal component analysis on the feature vectors of all collected samples and extracting the proportion of the projected variance of each feature vector in the principal component space. Let the variance contribution of the spatial feature vector be... The variance contribution of the feature vector of the transmission path is The variance contribution of the spectral eigenvector is All three satisfy the normalization condition. The feature vectors are weighted and fused based on their variance contribution to generate a stress anomaly feature index key. The fusion process is achieved through vector concatenation and weighted normalization, ensuring that the index key retains the key information of each feature vector while highlighting feature dimensions with high variance contribution. The corresponding optimal hierarchical control instruction set is used as the mapping value for the index key of this pressure anomaly feature, establishing a mapping entry. The data structure of the mapping entry includes an index key field, a mapping value field, and an auxiliary information field. The auxiliary information field records metadata such as the pressure mode type, number of optimization iterations, and standard deviation of the final pressure amplitude corresponding to the mapping entry, facilitating subsequent knowledge base maintenance and updates.

[0131] Multiple mapping entries are organized into an index structure based on stress anomaly feature index keys, forming a mapping knowledge base. The index structure is implemented using data structures such as multidimensional index trees or hash tables, supporting fast query operations based on index keys. For continuous feature index keys, a... A spatial index is established using a tree structure. During a query, the Euclidean distance between the query index key and each index key in the knowledge base is calculated, returning the nearest mapping entries. For discrete feature index keys, a hash table structure is used to establish an exact matching index, and the corresponding mapping entry is directly located during a query using a hash function. To improve the generalization ability of the knowledge base, a similarity threshold mechanism is introduced into the index structure. When the distance between the query index key and the nearest neighbor index key in the knowledge base exceeds the similarity threshold, an interpolation or extrapolation strategy is triggered. A new hierarchical control instruction set is generated based on the mapping values ​​of multiple adjacent mapping entries, avoiding matching failures due to insufficient knowledge base coverage. After the mapping knowledge base is loaded into the control strategy matching unit, the control strategy matching unit receives the real-time pressure anomaly source location results, extracts the spatial location information, transmission path information, and spectrum information, generates a query index key according to the same feature extraction and weighted fusion process, executes the query operation in the mapping knowledge base, obtains the corresponding optimal hierarchical control instruction set, and sends it to the multi-axis pressure actuator for execution, realizing adaptive control of the pressure field.

[0132] A second aspect of the present invention provides a pressure sensor integrated system for probe card pressure adaptive control, comprising:

[0133] The pressure sensing unit is used to set up a pressure sensing unit array at the probe card, collect the capacitance signal output by the pressure sensing unit array, and obtain pressure field state information by jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal.

[0134] An anomaly localization unit is used to identify, based on the pressure field state information, an anomaly region in the probe card contact interface where the pressure exceeds the uniformity threshold and the transmission path of the anomaly region in the vertical and horizontal directions, thereby obtaining the pressure anomaly source localization result.

[0135] The hierarchical control unit is used to decompose the pressure adjustment requirement into a global adjustment command for the overall tilt attitude of the probe card and a partitioned adjustment command for the local abnormal area based on the spatial scale and transmission path of the abnormal area in the pressure anomaly source location result, and to assign time priority to each command to obtain a hierarchical control command set.

[0136] The mapping library unit is used to apply different pressure modes multiple times and collect the corresponding pressure field state information and pressure anomaly source location results, establish a mapping knowledge base between pressure anomaly characteristics and the optimal hierarchical control instruction set, and load the mapping knowledge base into the control strategy matching unit.

[0137] The control matching unit is used to query the mapping knowledge base to obtain the corresponding hierarchical control instruction set after the control strategy matching unit receives the pressure anomaly source location result, and then send it to the multi-axis pressure actuator for execution.

[0138] A third aspect of the present invention provides an electronic device, comprising:

[0139] processor;

[0140] Memory used to store processor-executable instructions;

[0141] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0142] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0143] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A pressure sensor integration method for probe card pressure adaptive control, characterized in that, include: A pressure sensing unit array is set at the probe card, and the capacitance signal output by the pressure sensing unit array is collected. By jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal, the pressure field state information is obtained. Based on the pressure field state information, abnormal areas in the probe card contact interface where the pressure exceeds the uniformity threshold and the transmission paths of the abnormal areas in the vertical and horizontal directions are identified to obtain the pressure anomaly source location result. Based on the spatial scale and transmission path of the abnormal area in the pressure anomaly source location result, the pressure adjustment requirement is decomposed into a global adjustment command for the overall tilt attitude of the probe card and a partitioned adjustment command for the local abnormal area, and time priority is assigned to each to obtain a hierarchical control command set. Apply different pressure modes multiple times and collect the corresponding pressure field state information and pressure anomaly source location results to establish a mapping knowledge base between pressure anomaly characteristics and the optimal hierarchical control instruction set, and load the mapping knowledge base into the control strategy matching unit; After receiving the pressure anomaly source location result, the control strategy matching unit queries the mapping knowledge base to obtain the corresponding hierarchical control instruction set, sends it to the multi-axis pressure actuator, and executes it.

2. The method according to claim 1, characterized in that, The capacitance signal output from the pressure sensing unit array is acquired, and the pressure field state information is obtained by jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal, including: The pressure sensing unit array is spatially divided into multiple sub-regions and local statistical analysis is performed on the capacitance signal to obtain the average pressure amplitude of each sub-region. The difference between the average pressure amplitudes of adjacent sub-regions is calculated to construct a pressure spatial gradient field. The pressure concentration feature locations in the pressure spatial gradient field where the gradient amplitude exceeds a preset gradient threshold are extracted. The capacitance signal is sampled in time series, the rate of change of capacitance signal of each sensing unit in a continuous time window is calculated, and the sensing unit whose rate of change of capacitance signal exceeds a preset rate of change threshold is identified as the pressure dynamic fluctuation characteristic position and the spectral component of the corresponding rate of change of capacitance signal is recorded. By combining the spectral components corresponding to the spatially overlapping regions of the pressure concentration feature locations and the pressure dynamic fluctuation feature locations, the pressure evolution rate and oscillation period of the overlapping regions are analyzed. The overlapping region, the pressure evolution rate, the oscillation period, the gradient direction information of the pressure spatial gradient field, and the frequency characteristic information of the spectral components are combined to form a comprehensive feature vector characterizing the spatial non-uniformity and temporal evolution trend of the pressure field. The pressure field state information is generated based on the comprehensive feature vector.

3. The method according to claim 2, characterized in that, Based on the pressure field state information, abnormal regions in the probe card contact interface where the pressure exceeds the uniformity threshold are identified, along with the transmission paths of these abnormal regions in the vertical and horizontal directions. The pressure anomaly source localization results include: The Moran index of the pressure amplitude between each sensing unit in the pressure sensing unit array at the probe card contact interface and the surrounding neighboring sensing units is used to identify the position of the sensing unit where the sign of the Moran index changes as the pressure spatial aggregation mode transition boundary. The average difference of the pressure amplitude of the sensing units on both sides of the pressure spatial aggregation mode transition boundary is calculated as the uniformity threshold. When the deviation between the pressure amplitude of each sensing unit within the pressure concentration feature location and the average pressure amplitude exceeds the uniformity threshold, the pressure concentration feature location is confirmed as an abnormal area. Based on the gradient direction information of the pressure spatial gradient field, a horizontal transmission path is constructed by tracing the sensor unit sequence with progressively decreasing pressure amplitude outward along the gradient direction of the abnormal region; based on the pressure evolution rate and the oscillation period, the frequency bands of the spectral components corresponding to the abnormal region are divided and the energy ratio of each frequency band is calculated; based on the energy ratio, the attenuation characteristics of pressure transmission to the inner layer in the vertical layered structure of the identification probe card are constructed to form a vertical transmission path. The spatial location information of the abnormal area, the coordinate information of each sensing unit sequence in the horizontal transmission path, and the energy proportion information of each frequency band in the vertical transmission path are combined to generate the pressure anomaly source location result.

4. The method according to claim 3, characterized in that, Based on the pressure evolution rate and the oscillation period, the spectral components corresponding to the abnormal region are divided into frequency bands and the energy ratio of each frequency band is calculated. Based on the energy ratio, the attenuation characteristics of pressure transmission to the inner layer in the vertical layered structure of the identification probe card are used to construct the vertical transmission path, including: The characteristic frequency of pressure amplitude change with time in the abnormal region is calculated based on the pressure evolution rate. The periodic frequency component of pressure oscillation in the abnormal region is calculated based on the oscillation period. The characteristic frequency and the periodic frequency component are used as the frequency band division boundary. The spectral component corresponding to the abnormal region is divided into multiple frequency bands. The sum of the squares of the amplitudes of each frequency component in each frequency band is calculated as the frequency band energy. The ratio of the frequency band energy to the total energy is calculated as the energy proportion of each frequency band. A mapping relationship between the energy proportion of each layer and frequency band in the vertical layered structure of the probe card is established. The energy proportion of each frequency band is mapped to the pressure transmission characteristics of the probe card contact interface layer, intermediate support layer, and substrate layer, respectively. Based on the order of the energy proportion of each frequency band from high to low, the dominant path of pressure transmission in the vertical layered structure from the probe card contact interface layer to the intermediate support layer and then to the substrate layer is determined. The energy proportion of each frequency band in the dominant path is extracted to construct the attenuation sequence of pressure transmission layer by layer in the vertical direction as the transmission path in the vertical direction.

5. The method according to claim 1, characterized in that, Based on the spatial scale and transmission path of the abnormal region in the pressure anomaly source localization results, the pressure adjustment requirement is decomposed into a global adjustment command for the overall tilt attitude of the probe card and a regional adjustment command for the local abnormal region. These are then assigned time priorities, resulting in a hierarchical control command set including: Calculate the spatial offset vector between the geometric center of the abnormal region and the geometric center of the probe card contact interface, and perform principal component decomposition to extract the dominant offset direction; When the spatial scale of the abnormal region exceeds a preset area ratio threshold of the probe card contact interface area and the angle between the dominant offset direction and the edge of the probe card contact interface is less than a preset angle threshold, a global adjustment command for the overall tilt attitude of the probe card is generated; otherwise, the boundary of the influence range of the abnormal region is determined based on the horizontal transmission path, and the attenuation coefficient of the pressure transmission depth to the internal layered structure of the probe card is calculated based on the energy ratio information of each frequency band in the vertical transmission path, and a partition adjustment command for the local abnormal region is generated. Set timing constraints for the execution of the partition adjustment instruction after the global adjustment instruction has been executed; The global adjustment instruction, the partition adjustment instruction, and the timing constraint are combined to generate the hierarchical control instruction set.

6. The method according to claim 1, characterized in that, By applying different pressure modes multiple times and collecting the corresponding pressure field state information and pressure anomaly source location results, a mapping knowledge base is established between pressure anomaly characteristics and the optimal hierarchical control instruction set, including: By applying uniform pressure mode, unilateral tilt pressure mode, and local concentrated pressure mode through a multi-axis pressure actuator, the pressure field state information corresponding to each pressure mode is collected. The spatial location information of the abnormal area corresponding to the pressure field state information is used to construct a spatial feature vector, the coordinate information of each sensing unit sequence in the horizontal transmission path is used to construct a transmission path feature vector, and the energy ratio information of each frequency band in the vertical transmission path is used to construct a spectral feature vector, thereby obtaining the pressure anomaly characteristics. The execution parameters of the hierarchical control instruction set are adjusted according to the pressure anomaly characteristics corresponding to each pressure mode and executed repeatedly. The standard deviation of pressure amplitude in the pressure field state information after execution is calculated, and the hierarchical control instruction set with the smallest standard deviation of pressure amplitude is marked as the optimal hierarchical control instruction set for the corresponding pressure anomaly characteristics. The variance contribution of the spatial feature vector, the transmission path feature vector, and the spectral feature vector is calculated. Based on the variance contribution, each feature vector is weighted and fused to generate a pressure anomaly feature index key. The corresponding optimal hierarchical control instruction set is used as the mapping value of the pressure anomaly feature index key to establish a mapping entry. Multiple mapping entries are organized into an index structure based on the pressure anomaly feature index key to form the mapping knowledge base.

7. A pressure sensor integrated system for probe card pressure adaptive control, used to implement the method as described in any one of claims 1-6, characterized in that, include: The pressure sensing unit is used to set up a pressure sensing unit array at the probe card, collect the capacitance signal output by the pressure sensing unit array, and obtain pressure field state information by jointly analyzing the spatial distribution characteristics and temporal variation characteristics of the capacitance signal. An anomaly localization unit is used to identify, based on the pressure field state information, an anomaly region in the probe card contact interface where the pressure exceeds the uniformity threshold and the transmission path of the anomaly region in the vertical and horizontal directions, thereby obtaining the pressure anomaly source localization result. The hierarchical control unit is used to decompose the pressure adjustment requirement into a global adjustment command for the overall tilt attitude of the probe card and a partitioned adjustment command for the local abnormal area based on the spatial scale and transmission path of the abnormal area in the pressure anomaly source location result, and to assign time priority to each command to obtain a hierarchical control command set. The mapping library unit is used to apply different pressure modes multiple times and collect the corresponding pressure field state information and pressure anomaly source location results, establish a mapping knowledge base between pressure anomaly characteristics and the optimal hierarchical control instruction set, and load the mapping knowledge base into the control strategy matching unit. The control matching unit is used to query the mapping knowledge base to obtain the corresponding hierarchical control instruction set after the control strategy matching unit receives the pressure anomaly source location result, and then send it to the multi-axis pressure actuator for execution.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images