Signal transmission and reception method for array antenna based on tgv glass substrate

Abstract

This invention discloses a signal transceiver method for an array antenna based on a TGV glass substrate, specifically relating to the field of signal transceiver technology. The method involves rapidly coarsely scanning the entire sample to identify potentially defective areas, concentrating high-resolution, narrow-beamlength accumulated resources on these areas, while rapidly sweeping low-risk background areas with a wide beam. Based on the substrate temperature and bandwidth occupancy status, the method predicts temperature rise in real time, actively shutting down redundant array elements in the background area to release thermal margin and processing bandwidth. During load reduction, interpolation and inter-frame fusion are used to maintain image continuity. This allows for an overall detection speed that meets production line cycle time requirements without sacrificing the resolution of the fine inspection area, while simultaneously avoiding performance degradation of the transceiver link caused by overheating of the TGV substrate.

Description

Technical Field

[0001] This invention relates to the field of signal transceiver technology, and more specifically, to a method for signal transceiver of an array antenna based on a TGV glass substrate. Background Technology

[0002] TGV glass substrate array antennas are used for high-frequency imaging inspection of semiconductor packaging defects due to their low loss and high flatness. To identify micron-sized bubbles in MCU films or submicron cracks on the chip surface, millimeter-wave / Asia-Pacific Hertz broadband signals and increased antenna aperture are required to improve lateral resolution. However, this reduces the field of view for a single frame scan, necessitating point-by-point scanning and long signal accumulation to synthesize a complete image. At production lines with a production cycle of several wafers per second, the large number of spatial sampling points and accumulation time required for high resolution result in an imaging frame rate far lower than the required inspection speed.

[0003] Attempts to shorten integration time by increasing transmission power can lead to localized overheating due to the extremely low thermal conductivity of the TGV glass substrate, causing performance drift in the transceiver chip and even damaging the device under test. Constrained by receiver thermal noise and transmission safety power, improved spatial resolution inevitably sacrifices detection speed. Existing transceiver methods struggle to meet high-speed online inspection requirements without reducing defect sensitivity. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an array antenna signal transmission and reception method based on a TGV glass substrate to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The method for transmitting and receiving signals using an array antenna based on a TGV glass substrate involves transmitting a broadband signal and acquiring multi-channel echoes via a TGV array antenna, while simultaneously obtaining the substrate temperature and processing bandwidth status. The echoes are then rapidly imaged with low spatial resolution to obtain a global reflectivity map of the sample, which is compared with a benchmark to determine the areas to be inspected in detail. The beam scheduling priority is then calculated based on the degree of defect suspicion in each area. A time-division scanning control sequence is generated based on the priority and the current temperature status; During the execution of the sequence, the temperature rise is predicted based on the fitting relationship between the transmission duty cycle, bandwidth usage and temperature. When the predicted temperature reaches the critical value, the load reduction ratio of each region is obtained, and the corresponding transmit and receive channels are dynamically shut down and the bandwidth is released according to the ratio. Interpolation and inter-frame fusion are performed on the sparse echoes in the background area during the load reduction period. After the temperature drops to a safe threshold, full-function acquisition is restored, and a seamless detection image that maintains the target resolution for the fine inspection area is output.

[0006] In a preferred embodiment, obtaining the global reflectance map of the sample specifically includes: during the transmission of a broadband signal and the acquisition of the echo, reading the temperature sensor reading integrated on the back of the glass substrate as the substrate temperature, and reading the proportion of the processing bandwidth already occupied in the baseband processing unit as the processing bandwidth status. For the echo of each receiving channel, only a range-to-Fourier transform is performed to obtain a one-dimensional range image. On a two-dimensional grid parallel to the sample plane, incoherent projection superposition is performed using the pre-stored radiation patterns and range attenuation compensation coefficients of each array element to obtain the global reflectivity map.

[0007] In a preferred embodiment, the step of calculating beam scheduling priority based on the defect suspicion degree of each region specifically includes: obtaining a difference map by pixel-by-pixel difference between the global reflectivity map and the defect-free reflectivity reference template, and marking connected regions whose difference amplitude exceeds the defect threshold as regions to be inspected in detail. For each region to be inspected, the normalized ratio of the difference amplitude, the ratio of the region area to the minimum defect area, and the mean cosine of the observation angle of the region center to each array element are extracted and weighted to synthesize a defect suspicion score. All regions to be inspected are sorted according to the score, and a dynamic threshold is calculated in combination with the current substrate temperature and the proportion of the occupied processing bandwidth. Regions with scores not lower than the dynamic threshold are included in the high priority, and the rest are low priority, forming a beam scheduling priority.

[0008] In a preferred embodiment, the dynamic threshold is based on a base threshold when the substrate temperature is not higher than a safe temperature reference value and the bandwidth occupancy ratio is not higher than a reference ratio.

[0009] In a preferred embodiment, generating the time-division scanning control sequence specifically includes: for high-priority regions, calculating the phase shift value and amplitude weighting coefficient according to their center coordinates to form a focused narrow beam, and performing multiple coherent accumulations with the maximum available bandwidth and long frequency sweep period; For low-priority regions, wide beams and reduced array element subsets are matched according to the bounding box size to perform a single frequency sweep with reduced sweep bandwidth and shortened sweep period. The scanning tasks for each region are arranged into time-division sequences according to priority and spatial proximity. The cumulative temperature rise during the execution of the sequence is estimated by combining the thermal resistance model. When the estimated temperature rise exceeds the limit, the tasks in low-priority regions are reduced first. If the temperature still exceeds the limit, the number of accumulation times in high-priority regions is reduced until the temperature rise is below the safe limit temperature throughout the entire process.

[0010] In a preferred embodiment, the number of accumulations in the high-priority region is dynamically determined based on the region defect suspicion score and the current substrate temperature. The maximum number of accumulations is multiplied by the ratio of the region score to the maximum score in all high-priority regions, and then multiplied by a temperature-related backoff factor and rounded down.

[0011] In a preferred embodiment, the step of comparing with a benchmark to determine the region to be inspected specifically includes: sorting the difference magnitudes of all pixels in the difference map from largest to smallest, and taking the value corresponding to the 95th percentile as a candidate threshold; if the candidate threshold is lower than a preset absolute lower limit, then the absolute lower limit is used as the final defect threshold; if the candidate threshold is higher than the absolute lower limit, then the candidate threshold is used as the final defect threshold.

[0012] In a preferred embodiment, obtaining the regional load reduction ratio specifically includes: traversing all low-priority regions, evaluating whether the composite coverage formed by the main lobe and side lobes of the remaining active array element beams in the region after shutting down its current wide-beam array element subset meets the preset coverage integrity lower limit, and marking it as a region that can be deloaded. The number of shut-off array elements and the total amount of processing bandwidth that can be released in all deloadable regions are counted. A uniform deload ratio is calculated. Array elements and corresponding scanning time slots in the deloadable regions are shut down in sequence according to the deload ratio, and the processing bandwidth is recovered.

[0013] In a preferred embodiment, the interpolation and inter-frame fusion of the sparse echoes in the background region during the load reduction period specifically includes: for the sampling positions corresponding to the turned-off array elements, performing inverse distance weighted interpolation using the echo amplitudes of the active array elements in the region with the inverse cube of the spatial distance as the weight; The last full-element acquisition frame before load reduction is used as the historical reference. The interpolation result is compared with the corresponding position of the historical reference frame. If the difference does not exceed the tolerance threshold based on the fluctuation statistics of the historical frame, the historical value is used as the substitute. The interpolated background region data is stitched together with the high-resolution data of the high-priority region according to spatial coordinates to generate a complete detection image.

[0014] In a preferred embodiment, the step of restoring full-function acquisition after the temperature drops back to the safety threshold specifically includes: when the substrate temperature is less than or equal to the safety critical temperature minus the preset hysteresis, and the bandwidth occupancy ratio is lower than the reference ratio, restoring the array element bias and signal link channel by channel in the reverse order of shutdown, and re-inserting the scan time slot.

[0015] The technical effects and advantages of this invention are as follows: 1. This invention solves the problem that in online defect detection of semiconductor packaging using TGV glass substrate array antennas, it is impossible to simultaneously achieve high spatial resolution imaging and high production line throughput per second due to spatial bandwidth accumulation and the extremely low thermal conductivity of glass. By rapidly coarsely scanning the entire sample to identify suspected defect areas, high-resolution narrow beamlength accumulation resources are concentrated on these areas, while low-risk background areas are quickly swept over with a wide beam. The temperature rise is predicted in real time based on the substrate temperature and bandwidth occupancy status, and redundant array elements in the background area are actively shut down to release thermal margin and processing bandwidth. During the load reduction period, interpolation and inter-frame fusion are used to maintain image continuity. Thus, without sacrificing the resolution of the fine inspection area, the overall detection speed is increased to meet the production line throughput requirements, while avoiding the degradation of transceiver link performance caused by overheating of the TGV substrate.

[0016] 2. This invention combines rapid coarse scanning of the entire sample with fine scanning of suspicious areas in a hierarchical scheduling manner. At the source of the detection process, it compresses the spatial range required for high-resolution imaging, so that the limited scanning time and thermal budget are concentrated on the local areas where defects are suspected. Thus, without sacrificing defect sensitivity, the overall detection frame rate is increased to a level that matches the production line cycle time.

[0017] 3. This invention introduces substrate temperature and bandwidth occupancy status as constraint variables for beam scheduling in real time. Driven by temperature rise prediction that is jointly fitted by transmit duty cycle and processing bandwidth, it utilizes multi-beam redundant field of view in low-priority areas to achieve precise load reduction in different regions. This transforms the inherent physical disadvantage of low thermal conductivity of TGV glass substrates into actively manageable system parameters. The load reduction operation only temporarily shuts down background area array elements and corresponding processing bandwidth that contribute little to defect judgment. Thus, while maintaining uninterrupted full-function acquisition of narrow beam length accumulation in high-priority fine inspection areas, it actively curbs substrate temperature rise and avoids missed or misjudged defects due to amplitude and phase drift caused by overheating in the transceiver link.

[0018] 4. During the load reduction period, the present invention performs spatial interpolation and inter-frame constraint correction on the sparse echoes in the background area, and restores the full array acquisition by using a gradual weight fusion method after the temperature drops to a safe threshold. This ensures that the detection images from before, during and after the load reduction to the recovery are uninterrupted in time and have no splicing traces in space. The fine inspection area maintains the target resolution throughout the process, and the background area will not introduce false defects or miss real changes due to the load reduction. Attached Figure Description

[0019] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0020] 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.

[0021] Example: Figure 1 The present invention provides a signal transmission and reception method for an array antenna based on a TGV glass substrate, comprising the following steps: Broadband signals are transmitted and multi-channel echoes are acquired through a TGV array antenna. The substrate temperature and processing bandwidth status are obtained simultaneously. The echoes are rapidly imaged with low spatial resolution to obtain a global reflectivity map of the sample. The map is then compared with a benchmark to determine the areas to be inspected in detail. The beam scheduling priority is calculated based on the degree of defect suspicion in each area. Based on the priority and the current temperature state, a time-division scanning control sequence is generated. For high-priority regions, a narrow beam, long frequency sweep and multiple accumulations are used to obtain high resolution, while for low-priority regions, a wide beam and short frequency sweep are used to quickly sweep over them. During the execution of the sequence, the temperature rise is predicted according to the fitting relationship between the transmit duty cycle, bandwidth occupancy and temperature. When the predicted temperature reaches the critical value, the feasibility of suspending some array elements or reducing the repetition rate is evaluated by using the multi-beam redundant field of view of the low priority area, the load reduction ratio of each area is obtained, and the corresponding transmit and receive channels are dynamically shut down and the bandwidth is released according to the ratio. Interpolation and inter-frame fusion are performed on the sparse echoes in the background area during the load reduction period. After the temperature drops to a safe threshold, full-function acquisition is restored, and a seamless detection image that maintains the target resolution for the fine inspection area is output.

[0022] In this embodiment of the invention, a broadband signal is transmitted and multi-channel echoes are acquired via a TGV array antenna, while simultaneously obtaining the substrate temperature and processing bandwidth status. The echoes are then subjected to low spatial resolution rapid imaging to obtain a global reflectivity map of the sample, which is compared with a benchmark to determine the areas to be inspected in detail. Furthermore, beam scheduling priority is calculated based on the degree of defect suspicion in each area, specifically including: Through the transmission channel of the TGV array antenna, a linear frequency modulated continuous wave is radiated outward with a pulse sequence whose duty cycle is configurable. The duty cycle of the pulse sequence is set by the baseband processing unit according to a preset low-resolution fast imaging mode, so that the average transmission power is lower than that of the full-function detection mode. Each receiving unit synchronously acquires the scattered echoes from the sample surface, and the echoes are the raw array element domain data without beamforming; During data acquisition, readings from a temperature sensor integrated on the back of the glass substrate, located directly below the power amplifier chip and baseband processing chip, are read in parallel to obtain the substrate temperature. Simultaneously, it reads the currently occupied processing bandwidth percentage from the bandwidth monitoring register of the baseband processing unit. As a processing bandwidth status, where The processing bandwidth already used. Total processing bandwidth; A range-to-fast Fourier transform is performed only once on the time-domain scattered echo of each receiving channel to obtain a one-dimensional range image. Each range gate in the one-dimensional range image corresponds to a range cell along the radar line of sight. The radiation pattern and range attenuation compensation curve of each array element obtained by pre-measurement or simulation are stored in the memory. For each grid point in a preset two-dimensional grid parallel to the plane of the sample under test, the following incoherent projection superposition is performed: Based on the spatial geometric relationship between the grid point and the nth receiving array element, the slant range from the grid point to the array element is calculated. The amplitude value corresponding to the slant range is extracted from the one-dimensional range image of the nth channel and multiplied by the radiation gain and range attenuation compensation coefficient of the array element in the direction of the grid point to obtain the contribution value of the nth array element to the grid point. The contribution values ​​of all array elements to the grid point are accumulated and used as the pixel value of the global reflectivity map at the grid point. The global reflectivity map is obtained by traversing all grid points. During the projection superposition process, phase weighting and beamforming operations between array elements are not performed to exchange spatial resolution for computational speed. It should be noted that the radiation gain and distance attenuation compensation coefficients are not fixed values, but are determined point by point based on the pre-calibrated array antenna pattern and measurement geometry.

[0023] The radiation gain refers to the normalized pattern amplitude value of the nth receiving element in the direction of the two-dimensional grid point. This value is obtained in advance by: measuring the near-field or far-field pattern of the TGV array antenna in a microwave anechoic chamber, recording the complex gain of each element at different elevation and azimuth angles, and storing the normalized amplitude portion in a two-dimensional lookup table; for cases where measurement conditions are not available, the active element pattern of each element can be extracted using full-wave electromagnetic simulation as a substitute. During incoherent projection superposition, the corresponding elevation and azimuth angles are calculated based on the geometric relationship between the grid point and the element, and the radiation gain value in that direction is obtained by looking up the table or interpolation.

[0024] The distance attenuation compensation coefficient is used to offset the amplitude difference introduced by the two-way propagation loss of electromagnetic waves from the array element to the sample surface and back to the array element with different distances. Under free space conditions, the two-way attenuation is inversely proportional to the fourth power of the distance, so the compensation coefficient is taken as the fourth power of the slant distance from the grid point to the array element.

[0025] The global reflectance map is compared with a defect-free reflectance reference template obtained and stored by performing the same rapid imaging process on known defect-free samples of the same type using the same array. Point-by-point difference calculation is performed at each pixel position of the two-dimensional grid to obtain a difference map. For connected pixel regions in the difference map whose difference amplitude exceeds a preset defect threshold, the regions are marked according to their center coordinates, bounding box size and number of pixels contained therein. The marking results are determined as regions to be inspected in detail. In one embodiment of this application, the defect threshold is set in an adaptive manner: the difference amplitude of all pixels in the difference map is sorted from largest to smallest, and the value corresponding to the 95th percentile is taken as a candidate threshold; if the candidate threshold is lower than the system's preset absolute lower limit Tmin, then Tmin is used as the final defect threshold; if the candidate threshold is higher than Tmin, then the candidate threshold is used as the final defect threshold, thereby achieving a balance between background noise fluctuations and the detection rate of minor defects. For each marked region k to be inspected, the following three sets of features are extracted to calculate its defect suspicion score: The first feature is the normalized ratio of the mean difference magnitude of all pixels in the difference map within that region to the reflectance difference corresponding to the maximum resolvable defect set in the current detection task. The reflectance difference corresponding to the largest resolvable defect is determined by simulation or measurement of the reflectance coefficient between the intact area and the completely debonded or material-abrupt area on the sample surface; the second feature is the area of ​​this region. The minimum defect area required to be detected by the current inspection task The ratio; the third feature is the average of the cosine values ​​of the observation angle of the nth element in the TGV array relative to the center position of the region, where the observation angle is the angle between the line connecting the element and the center of the region and the normal of the element. The defect suspicion score of the area k to be inspected is weighted and synthesized according to the following formula. : In the formula, This is the normalized ratio of the average reflectance difference in the k-th region to be examined. The area of ​​the region, This represents the minimum defect area required to be detected in the current inspection task. Let cosine be the angle of observation of the nth array element to the center of the region. The total number of array elements. These are weighting coefficients determined through offline training based on historical defect statistics of the production line, with default values ​​of 0.5, 0.3, and 0.2 respectively; In one embodiment disclosed in this application, taking the detection of air bubbles in a certain MCU film as an example: a region A to be inspected in detail... ,area Minimum defect area Three times that of the array, its location directly below the array makes the mean cosine of the observation angle of each array element close to 0.95; another area to be examined in detail, B. ,area Also Three times that of the array, but its proximity to the array edge resulted in a mean cosine of the observed angle of only 0.65; using , , Calculation, region of Region B Region A received a higher score due to its more favorable observation location and will be given priority in high-resolution scanning resources in subsequent beam scheduling.

[0026] All areas to be inspected will be scored according to the aforementioned defect suspicion level. Sort by highest to lowest; based on the currently acquired substrate temperature. and the proportion of processing bandwidth already used Calculate the dynamic threshold using the following formula. : In the formula, The substrate temperature should not exceed the reference temperature. And the bandwidth usage ratio is not higher than the reference ratio. The basic threshold value at that time, This is a safe temperature reference value determined based on the thermal expansion coefficient of the TGV glass substrate and the inflection point of antenna phase consistency degradation. This refers to the maximum acceptable bandwidth occupancy ratio for the baseband processing unit to maintain data flow without frame loss in real-time acquisition mode. and The temperature increment coefficient and bandwidth increment coefficient are obtained through joint calibration of system thermal simulation and production line cycle time. All those that meet the requirements... The areas to be inspected are marked as high priority and added to the beam scheduling priority queue, while the remaining areas are marked as low priority, thus generating the beam scheduling priority. Under this mechanism, when the substrate temperature... No more than And bandwidth usage ratio No more than Dynamic threshold equal to the basic threshold The queue includes all defects with a suspicion score of no less than [a certain value]. The region; when either the substrate temperature or the bandwidth occupancy ratio exceeds its reference value, the dynamic threshold increases linearly with the magnitude of the exceedance, and the queue only retains regions whose defect suspicion score matches the current thermal load and bandwidth pressure; the beam scheduling priority is directly used to guide the decision-making of allocating narrow beam length accumulation or wide beam fast sweep mode to different regions in the subsequent time-division scanning control sequence.

[0027] In this embodiment of the invention, a time-division scanning control sequence is generated based on the priority and the current temperature state. For high-priority regions, a narrow beam, long frequency sweep, and multiple accumulations are used to obtain high resolution; for low-priority regions, a wide beam, short frequency sweep, and rapid sweep are used. Specifically, this includes: Read all regions to be inspected that have been marked as high priority and low priority from the beam scheduling priority queue, and obtain the current substrate temperature. and the proportion of processing bandwidth already used For each high-priority region, based on its central spatial coordinates and the relative positions of each element of the TGV array, the phase shift values ​​of each transmitting element required for the array's transmitted beam to form in-phase superposition at the central point are calculated according to the free space propagation model. At the same time, the amplitude weighting coefficients of each element are calculated according to the Taylor weighted distribution to suppress sidelobes. The resulting focused narrow beam has a half-power beamwidth that does not exceed the lateral resolution angle corresponding to the minimum defect size required by the current detection task. The lateral resolution angle is determined by the ratio of the minimum defect size to the vertical distance from the array to the sample surface. In one embodiment disclosed in this application, the process of calculating the phase shift value based on the free space propagation model is as follows: a three-dimensional rectangular coordinate system is established with the center of the TGV array plane as the origin, and the spatial coordinates of the nth array element are obtained. and the target coordinates of the center of the high-priority region. ,in The vertical working distance from the array plane to the sample surface is obtained through laser rangefinding calibration during system assembly; the slant distance from this array element to the target center is calculated. Select the slant distance from the center element of the array or a preset reference element to the target center. As a phase reference, the required phase shift value for the nth transmitting element for: In the formula, This represents the free-space wavelength corresponding to the center frequency of the transmitted signal. Taking a scenario with an operating frequency of 60 GHz, a wavelength of λ = 5 mm, a 4×4 uniform array, an element spacing of 3 mm, a center at the origin, and a target at coordinates (10 mm, 8 mm, 20 mm) as an example, the phase shift value for each channel can be obtained by substituting each element into the above formula. This calculation is completed in the baseband processing unit through table lookup or real-time computation, and the result is directly written to the phase shifter register to achieve in-phase focusing of the transmitted beam at the target center.

[0028] In one embodiment of this application, the calculation of amplitude weighting coefficients for each array element based on a Taylor weighted distribution to suppress sidelobes specifically employs a -30dB Taylor pattern synthesis design. For an N-element transmitting antenna linear array arranged at equal intervals along the surface of a TGV glass substrate, the element spacing is taken as the free space half-wavelength corresponding to the operating center frequency. First, the near-zero broadening factor σ of the Taylor distribution is calculated based on the selected -30dB sidelobe level. Then, a continuous Taylor aperture distribution is generated from the normalized position of each element in the array and σ. Finally, the current excitation amplitude of each element is sampled at the element position to form a set of normalized amplitude weighting coefficients that decrease symmetrically from the center of the array to both sides. Taking the case of an operating frequency of 77GHz and the number of array elements N=16 as an example, the calculated centrally symmetrical normalized amplitude weighting coefficients are 1.00, 0.98, 0.94, 0.87, 0.78, 0.68, 0.56, and 0.45, respectively, with the coefficient of the outermost element being approximately 0.32. The weighting coefficients are directly fed into the control terminal of the variable gain amplifier of each transmission channel during the transmission beam scanning. This ensures that the formed focused narrow beam meets the requirement of the minimum defect lateral resolution angle corresponding to the main lobe width, while stabilizing the peak sidelobe level below -30dB, thereby suppressing the interference of sidelobe region reflected echoes on the imaging of the current fine inspection area.

[0029] Set the sweep bandwidth of the transmitted signal to the maximum available bandwidth supported by the system hardware. The frequency sweep period is set to , The value of ensures that the distance resolution meets the fine detection requirements under the maximum available bandwidth, and that the substrate temperature rise does not exceed the preset single-cycle temperature rise limit within a single frequency sweep cycle; at the receiving end, multiple frequency sweep echoes continuously acquired in this high-priority region are phase-aligned and superimposed in phase to complete coherent accumulation; wherein, the accumulation times allocated to each high-priority region are... Based on the degree of suspicion of defects in this area. With substrate temperature Determined dynamically according to the following formula: In the formula, The maximum number of accumulations allowed per single area is calculated by dividing the longest single-area dwell time tolerated by the production line cycle time by the single sweep frequency cycle. The result is obtained by rounding down. The maximum value of the defect suspicion score across all high-priority regions is used to normalize the score. The retreat factor is related to the substrate temperature. hour ,when hour ,in This is the temperature coefficient obtained by fitting the measured temperature rise rate of the glass substrate under different emission duty cycles. This refers to the safe temperature reference value; For each low-priority region, its bounding box size is matched with the beamwidth required for low-resolution coverage: using the diagonal length of the bounding box of the low-priority region as the coverage range, a lower limit of the beamwidth that meets the coverage range requirements is obtained, thereby determining the size of the array element subset participating in forming the wide beam; the array element subset is selected from the central region of the TGV array, and the number of array elements is minimized by constraining the lower limit of the beamwidth, so that the main lobe of the beam pattern covers the bounding box of the region, and array elements not participating in the subset are in standby or off state to save power; the sweep bandwidth of the wide beam is reduced to , The lower limit bandwidth value is taken to meet the requirements of low-resolution detection, and the frequency sweep period is shortened to... , according to The corresponding minimum sampling time is determined, and each low-priority region performs only a single frequency sweep without accumulation; After allocating the scanning mode parameters for all regions, the multiple accumulated scanning tasks of each high-priority region and the single frequency sweep tasks of each low-priority region are arranged into a time-division scanning control sequence, prioritizing high-priority regions and grouping regions of the same priority according to spatial proximity. The spatial proximity grouping refers to grouping regions of the same priority with adjacent spatial coordinates and beam pointing angle differences less than a preset angle threshold into the same scanning group, completing the scan within consecutive time slots to reduce the phase shifter re-establishment time caused by beam agility. Each time slot in the time-division scanning control sequence corresponds to a specific beam pointing, a set of array element phase shift values ​​and amplitude weighting coefficients, as well as corresponding frequency sweep bandwidth and frequency sweep period parameters. After the sequence is arranged, the cumulative temperature rise curve of the substrate during the entire sequence is estimated based on the transmission duration, transmission power and number of array elements activated for all time slots in the sequence, combined with the thermal resistance model of the TGV glass substrate. In one embodiment disclosed in this application, the method of predicting the cumulative temperature rise curve of the substrate during the entire sequence process using a thermal resistance model of the TGV glass substrate is constructed and executed in the following specific manner: The TGV glass substrate and its integrated main heat-generating components (including the power amplifier chip, baseband processing chip, and ferroelectric device bias power supply module) were discretized into a three-node lumped-parameter thermal resistance network in a thermal simulation environment. Node 1 is the junction layer of the power amplifier chip, node 2 is the junction layer of the baseband processing chip, and node 3 is the central region of the upper surface of the glass substrate. The thermal resistance parameters between each node were extracted by combining the transient thermal response curves of the measured glass substrate sample with finite element steady-state thermal simulation. Specifically, a thermal imager was used to record the heating process of the glass substrate under step power excitation at a rate of 30 frames per second, and the interfacial thermal resistance from node 1 to node 3 was inverted by fitting the time constant of the node temperature rise. Thermal resistance of node three to the environment The heat capacity of node two , , It is calculated by multiplying the material's specific heat capacity, volume, and density.

[0030] During operation, based on the current time-division scanning control sequence, the number of transmission channels activated, the transmission duty cycle, and the signal processing bandwidth for each time slot are converted into the heat generation power on node one and node two. and Input the system of ordinary differential equations for the thermal resistance network: , , ,in , , For each node relative to the ambient temperature The temperature rise value, This represents the interfacial thermal resistance between the baseband chip and the glass substrate. Let be the convective-radiative equivalent thermal resistance from the glass substrate to the air. The equations are solved using the fourth-order Runge-Kutta method with a step size of 0.1 seconds, yielding the following results. The discrete sequence, i.e. the predicted cumulative temperature rise curve of the substrate temperature over time.

[0031] If the predicted temperature at any point in the cumulative temperature rise curve exceeds the preset safety limit temperature... Then, perform task reduction: first, remove the lowest-scoring items from the sequence in the low-priority areas one by one; after each removal, re-estimate the temperature rise curve and compare it with the previous curve. Comparison; if the limit is still exceeded after removing all low-priority areas, then in order of the number of accumulated frequencies in high-priority areas from fewest to most, areas with more than 1 accumulated frequency will be reduced to single-frequency sweeps until the estimated temperature rise does not exceed the limit throughout the entire process. The final generated time-division scanning control sequence is directly used to drive the transmit waveform generator, phase shifter network and receive gating timing of the TGV array. High-priority areas are acquired using narrow beam, long frequency sweep and multiple accumulation methods to obtain high-resolution imaging data, while low-priority areas are quickly covered using wide beam and short frequency sweep. The imaging data and coverage data are stored in the corresponding storage partitions for subsequent processing.

[0032] In this embodiment of the invention, during the execution of the sequence, the temperature rise is predicted based on the fitting relationship between the transmit duty cycle, bandwidth occupancy, and temperature. When the predicted temperature reaches a critical value, the feasibility of suspending some array elements or reducing the repetition rate is assessed using the multi-beam redundant field of view in the low-priority region, obtaining the regional load reduction ratio, and dynamically shutting down the corresponding transmit and receive channels and releasing bandwidth according to the ratio. Specifically, this includes: While the time-division scanning control sequence is executed slot by slot, the number of transmit channels activated in the current slot and adjacent future slots, the pulse repetition frequency and pulse width of each transmit channel are read in real time from the timing controller of the baseband processing unit, and the transmit duty cycle of the current slot is calculated: The system reads the number of enabled transmit channels in the current time slot and the number of transmit channels to be enabled in the next adjacent time slot from the status register of the timing controller. The larger of these two values ​​is taken as the number of transmit channels enabled (Mactive) in the current evaluation period to cover the situation of sudden increase in transmit load during time slot switching. For each of the Mactive transmit channels, its pulse repetition frequency and pulse width are read. The reciprocal of the pulse repetition frequency is the pulse period of that channel. The pulse width is divided by the pulse period to obtain the transmit duty cycle of a single channel. The arithmetic mean of the transmit duty cycles of all Mactive channels is calculated. If the pulse repetition frequency and pulse width of each transmit channel in the array are uniformly configured, the transmit duty cycle of any channel is directly taken as the single channel duty cycle. The single channel duty cycle is multiplied by the ratio of Mactive to the total number of transmit channels Mtotal in the TGV array to obtain the transmit duty cycle of the current time slot.

[0033] The system reads the currently occupied processing bandwidth ratio from the bandwidth monitoring register; reads the current substrate temperature from the temperature sensor; and inputs the transmit duty cycle, processing bandwidth ratio, and substrate temperature into a pre-fitted temperature rise prediction model obtained from system measured data. This model is a lookup table and linear interpolation model established by measuring the temperature rise curves of the TGV glass substrate during continuous operation under different transmit duty cycles and bandwidth occupancy combinations, and then fitting the data to obtain the correspondence between the transmit duty cycle, bandwidth occupancy ratio, and substrate temperature rise rate. Starting with the current substrate temperature, the model iteratively accumulates the temperature rise across time slots according to the estimated transmit duty cycle for each time slot, outputting the predicted substrate temperature after a specified time period. When the predicted substrate temperature exceeds a preset safety critical temperature... When this occurs, a thermal performance warning is triggered and a load reduction assessment process is initiated. The safety critical temperature is determined by combining the upper limit of the junction temperature safety of the power amplifier chip on the TGV glass substrate and the maximum temperature gradient allowed for the phase consistency of the antenna array, taking the lower of the two. The load reduction assessment process first traverses all low-priority regions in the beam scheduling priority queue, and performs a redundancy field-of-view analysis for each low-priority region: based on the spatial bounding box of the low-priority region and the positions and beam configurations of the currently active array elements in the TGV array, using the pre-stored radiation pattern data of each array element, it calculates the composite radiation power distribution formed by the main lobe and side lobes of the high-priority region beam currently being executed by the remaining active array elements in the low-priority region after all the array element subsets participating in wide beam coverage in the region are turned off; if the value of the composite radiation power distribution at any pixel point in the region is not lower than the minimum detectable power sensitivity corresponding to the pixel position, then the pixel is determined to be covered; the proportion of the number of covered pixels to the total number of pixels in the region is counted, and if this proportion is not lower than the preset coverage integrity lower limit, then the coverage integrity assessment process is considered complete. If the condition is met, the low-priority area is determined to have redundant coverage, marked as a deloadable area, and the currently used subset of array elements and the occupied scan time slots are recorded; if the condition is not met, the original array element configuration and time slots of the low-priority area are retained, and no deload processing is performed. After completing the redundant field-of-view analysis of all low-priority regions, the total number of shut-off array elements and the total amount of releaseable processing bandwidth corresponding to all deloadable regions are counted. The total amount of releaseable processing bandwidth is the sum of the beamforming and filtering bandwidth occupied by the deloadable regions in the baseband processing unit. The uniform deload ratio for each deloadable region is calculated using the following formula. : In the formula, This is a load reduction ratio adjustment coefficient calibrated based on the relationship between the heat capacity of the glass substrate and the emission power and the rate of temperature rise. Total processing bandwidth To prevent small constants with a denominator of zero, This is the preset safety critical temperature; According to the aforementioned unified load reduction ratio For each deloadable area, dynamic shutdown is performed: In the subset of shut-off array elements corresponding to that area, array elements are selected sequentially in ascending order of their coverage contribution to the area. The number of elements is rounded down to the nearest r. On the transmit channel side, the power amplifier bias is cut off via a gating signal. On the receive channel side, the connection between the low-noise amplifier output and the baseband processing link is disconnected via a switch, putting the array element into a sleep state. Simultaneously, the scanning time slots originally occupied by that area are removed from the sequence, and the released time slots are used by high-priority areas to extend their accumulation time or complete scanning earlier. The baseband processing unit synchronously recovers the processing bandwidth occupied by beamforming and filtering operations corresponding to the shut-off array elements, ensuring that the proportion of the occupied processing bandwidth is reduced. The corresponding decrease; immediately after the shutdown operation, the proportion of pixels covered by the remaining array elements in that area is recalculated, and if it is lower than... Then, a small number of array elements are replenished from the end of the shutdown sequence until the condition is met. until; The dynamic shutdown operation continuously monitors the substrate temperature and the proportion of processing bandwidth already used during execution. And U fell below the reference ratio At that time, the bias and signal link of the array elements are restored channel by channel in the reverse order of shutdown, and the corresponding scan time slots are re-inserted into the sequence so that all deloadable areas are restored to normal acquisition mode; the preset hysteresis The thermal time constant of the glass substrate is set to avoid frequent switching near the critical temperature.

[0034] In this embodiment of the invention, sparse echoes in the background region during the load reduction period are interpolated and fused with inter-frame data. After the temperature drops to a safe threshold, full-function acquisition is resumed, and a seamless detection image that maintains the target resolution for the fine detection area is output. Specifically, this includes: During the load reduction operation, the receiving channels of the switched-off array elements in the load reduction area are in a dormant state, and only the remaining active array elements in the area provide sparse spatially sampled echo data; the baseband processing unit continues to perform range-to-fast Fourier transform for each active array element in the area to obtain a one-dimensional range profile of each active array element in a single frequency sweep, and the amplitude value of the one-dimensional range profile constitutes the sparse sampling point set of the area in the current frequency sweep frame; For the original sampling positions corresponding to the shut-off array elements within the deloadable region, interpolation reconstruction and inter-frame constraints are performed from both spatial and temporal dimensions.

[0035] In the spatial dimension, for each grid point in the preset two-dimensional grid that falls within the region, the three closest active array elements are selected based on the spatial distance between the grid point and each active array element. Their amplitude values ​​in the current frequency sweep frame are weighted and averaged with the cube of the reciprocal of their respective distances as the weights to synthesize the reflectance estimate of the grid point, thereby obtaining the spatially interpolated regional reflectance map. In the time dimension, the complete frequency sweep frame obtained in the last full-element normal acquisition mode before load reduction in this region is extracted as the historical reference frame. The reflectance estimate of each grid point in the spatially interpolated regional reflectance map is subtracted from the reflectance value of the same grid point in the historical reference frame. If the difference exceeds the preset reflectance time-varying tolerance threshold, it is determined that the position has undergone a real change exceeding the noise fluctuation range during the load reduction period. In this case, the spatial interpolation result is adopted as the final reflectance value of the current frame. If the difference does not exceed the reflectance time-varying tolerance threshold, the reflectance value of the historical reference frame is used to replace the current interpolation result to suppress random jitter introduced by interpolation and maintain the image stability of the background region during the load reduction period.

[0036] The time-varying reflectance tolerance threshold is set based on three times the standard deviation of the pixel reflectance time fluctuation in the region under full acquisition mode for several consecutive frames before load reduction, so that only significant reflectance changes are retained as valid updates, and noise level fluctuations are replaced by historical values. After each frequency sweep frame is processed, the reflectivity data of the deloadable region after spatial interpolation and inter-frame constraint correction is stitched together with the high-resolution imaging data of the high-priority region within the same frame obtained by narrow beam length accumulation full-function acquisition according to their respective spatial coordinates on the preset two-dimensional grid to generate the complete detection image of the current frame. During stitching, for the same grid point, if the point falls on the overlapping boundary of the high-priority fine detection region and the deloadable background region, the measured pixel value of the high-priority region is used to cover the interpolated pixel value first, so as to ensure that the fine detection region is always presented with the actual high-resolution data acquired. During the load reduction operation, the above interpolation, correction, and splicing process is executed cyclically at intervals equal to the frame processing cycle of the baseband processing unit, while continuously monitoring the substrate temperature Tsub and the proportion of occupied processing bandwidth U; when Falling back to satisfaction And U fell below the reference ratio When the temperature has dropped to a safe threshold, the baseband processing unit issues a recovery command to restore the power amplifier bias and low noise amplifier signal links of the shut-down array elements in each channel in the reverse order of the shutdown sequence. The removed scan time slots are then reinserted into the time-division scanning control sequence, so that the deloadable area is restored to the normal acquisition mode of full array elements and full frequency sweep. Starting from the first full-element frequency sweep frame after the recovery command is executed, the deloadable area regenerates complete echo data acquired by all array elements. The reflectivity data of the recovery frame and the last frame during the deload period, which has been corrected by interpolation constraints, are then fused pixel by pixel with weighted fusion. The fusion weights change linearly according to the recovery frame number: the first frame after recovery uses a weight of 1 for the last frame during deload and a weight of 0 for the full acquisition frame; the weight ratio is adjusted to 0.5 to 0.5 for the second frame, 0 to 1 for the third frame, and from the fourth frame onwards, it is completely switched to full acquisition data. This eliminates the signal jump caused by the reopening of array elements and the establishment of links at the moment of recovery, and outputs a continuous detection image sequence without breaks, including full-function acquisition before deload, interpolation constraint maintenance during deload, and full-function acquisition after recovery. The fine detection area maintains the target resolution unchanged throughout the process using a narrow beam length accumulation mode.

[0037] This invention combines rapid coarse scanning of the entire sample with fine scanning of suspicious areas in a hierarchical scheduling manner. At the source of the detection process, it compresses the spatial range required for high-resolution imaging, allowing the limited scanning time and thermal budget to be concentrated on local areas where defects are truly suspected. This improves the overall detection frame rate to a level that matches the production line cycle time without sacrificing defect sensitivity.

[0038] This invention introduces substrate temperature and bandwidth occupancy status as constraint variables for beam scheduling in real time. Driven by temperature rise prediction that is jointly fitted by transmit duty cycle and processing bandwidth, it utilizes multi-beam redundant field of view in low-priority areas to achieve precise load reduction in different regions. This transforms the inherent physical disadvantage of low thermal conductivity of TGV glass substrates into actively manageable system parameters. The load reduction operation only temporarily shuts down background array elements and corresponding processing bandwidth that contribute little to defect judgment. Thus, while maintaining uninterrupted full-function acquisition of narrow beam length accumulation in high-priority fine inspection areas, it actively curbs substrate temperature rise and avoids missed or misjudged defects due to amplitude and phase drift caused by overheating in the transceiver link.

[0039] This invention ensures that the detection images from before, during and after the load reduction process are uninterrupted in time and without splicing traces in space. The fine inspection area maintains the target resolution throughout the process, and the background area will not introduce false defects or miss real changes due to the load reduction.

[0040] This invention solves the problem of the incompatibility between high spatial resolution imaging and high production line throughput per second in online defect detection of semiconductor packaging using TGV glass substrate array antennas due to spatial bandwidth accumulation and the extremely low thermal conductivity of glass. By rapidly coarsely scanning the entire sample to identify suspected defect areas, high-resolution narrow beamlength accumulation resources are concentrated on these areas, while low-risk background areas are quickly swept over with a wide beam. The temperature rise is predicted in real time based on the substrate temperature and bandwidth occupancy status, and redundant array elements in the background area are actively shut down to release thermal margin and processing bandwidth. During the load reduction period, interpolation and inter-frame fusion are used to maintain image continuity. Thus, without sacrificing the resolution of the fine inspection area, the overall detection speed is increased to meet the production line throughput requirements, while avoiding the degradation of transceiver link performance caused by overheating of the TGV substrate.

[0041] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0042] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0043] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for transmitting and receiving array antenna signals based on a TGV glass substrate, characterized in that: Broadband signals are transmitted and multi-channel echoes are acquired through a TGV array antenna. The substrate temperature and processing bandwidth status are obtained simultaneously. The echoes are rapidly imaged with low spatial resolution to obtain a global reflectivity map of the sample. The map is then compared with a benchmark to determine the areas to be inspected in detail. The beam scheduling priority is calculated based on the degree of defect suspicion in each area. A time-division scanning control sequence is generated based on the priority and the current temperature status; During the execution of the sequence, the temperature rise is predicted based on the fitting relationship between the transmission duty cycle, bandwidth usage and temperature. When the predicted temperature reaches the critical value, the load reduction ratio of each region is obtained, and the corresponding transmit and receive channels are dynamically shut down and the bandwidth is released according to the ratio. Interpolation and inter-frame fusion are performed on the sparse echoes in the background area during the load reduction period. After the temperature drops to a safe threshold, full-function acquisition is restored, and a seamless detection image that maintains the target resolution for the fine inspection area is output.

2. The signal transceiver method for an array antenna based on a TGV glass substrate according to claim 1, characterized in that: The process of obtaining the global reflectivity map of the sample specifically includes: during the transmission of a broadband signal and the acquisition of the echo, reading the temperature sensor reading integrated on the back of the glass substrate as the substrate temperature, and reading the proportion of the processing bandwidth already occupied in the baseband processing unit as the processing bandwidth status. For the echo of each receiving channel, only a range-to-Fourier transform is performed to obtain a one-dimensional range image. On a two-dimensional grid parallel to the sample plane, incoherent projection superposition is performed using the pre-stored radiation patterns and range attenuation compensation coefficients of each array element to obtain the global reflectivity map.

3. The signal transceiver method for an array antenna based on a TGV glass substrate according to claim 1, characterized in that: The step of calculating beam scheduling priority based on the defect suspicion level of each region specifically includes: taking the pixel-by-pixel difference between the global reflectivity map and the defect-free reflectivity reference template to obtain a difference map, and marking connected regions whose difference amplitude exceeds the defect threshold as regions to be inspected in detail; For each region to be inspected, the normalized ratio of the difference amplitude, the ratio of the region area to the minimum defect area, and the mean cosine of the observation angle of the region center to each array element are extracted and weighted to synthesize a defect suspicion score. All regions to be inspected are sorted according to the score, and a dynamic threshold is calculated in combination with the current substrate temperature and the proportion of the occupied processing bandwidth. Regions with scores not lower than the dynamic threshold are included in the high priority, and the rest are low priority, forming a beam scheduling priority.

4. The signal transceiver method for an array antenna based on a TGV glass substrate according to claim 3, characterized in that: The dynamic threshold is based on the base threshold when the substrate temperature is not higher than the safe temperature reference value and the bandwidth occupancy ratio is not higher than the reference ratio.

5. The method for transmitting and receiving array antenna signals based on a TGV glass substrate according to claim 1, characterized in that: The generation of the time-division scanning control sequence specifically includes: for high-priority regions, calculating the phase shift value and amplitude weighting coefficient according to their center coordinates to form a focused narrow beam, and performing multiple coherent accumulations with the maximum available bandwidth and long frequency sweep period; For low-priority regions, wide beams and reduced array element subsets are matched according to the bounding box size to perform a single frequency sweep with reduced sweep bandwidth and shortened sweep period. The scanning tasks for each region are arranged into time-division sequences according to priority and spatial proximity. The cumulative temperature rise during the execution of the sequence is estimated by combining the thermal resistance model. When the estimated temperature rise exceeds the limit, the tasks in low-priority regions are reduced first. If the temperature still exceeds the limit, the number of accumulation times in high-priority regions is reduced until the temperature rise is below the safe limit temperature throughout the entire process.

6. The signal transceiver method for an array antenna based on a TGV glass substrate according to claim 5, characterized in that: The number of accumulations in the high-priority region is dynamically determined based on the region defect suspicion score and the current substrate temperature. The maximum number of accumulations is multiplied by the ratio of the region score to the maximum score in all high-priority regions, and then multiplied by a temperature-related backoff factor and rounded down.

7. The method for transmitting and receiving array antenna signals based on a TGV glass substrate according to claim 3, characterized in that: The step of comparing with a benchmark to determine the area to be inspected specifically includes: sorting the difference magnitudes of all pixels in the difference map from largest to smallest, and taking the value corresponding to the 95th percentile as a candidate threshold; if the candidate threshold is lower than a preset absolute lower limit, then the absolute lower limit is used as the final defect threshold; if the candidate threshold is higher than the absolute lower limit, then the candidate threshold is used as the final defect threshold.

8. The signal transceiver method for an array antenna based on a TGV glass substrate according to claim 1, characterized in that: The process of obtaining the regional load reduction ratio specifically includes: traversing all low-priority regions, evaluating whether the composite coverage formed by the main lobe and side lobe of the remaining active array element beams in the region after shutting down its current wide-beam array element subset meets the preset coverage integrity lower limit, and marking it as a region that can be deloaded. The number of shut-off array elements and the total amount of processing bandwidth that can be released in all deloadable regions are counted. A uniform deload ratio is calculated. Array elements and corresponding scanning time slots in the deloadable regions are shut down in sequence according to the deload ratio, and the processing bandwidth is recovered.

9. The signal transceiver method for an array antenna based on a TGV glass substrate according to claim 1, characterized in that: The interpolation and inter-frame fusion of sparse echoes in the background region during the load reduction period specifically includes: for the sampling positions corresponding to the turned-off array elements, inverse distance weighted interpolation is performed using the echo amplitude of active array elements in the region with the inverse cube of the spatial distance as the weight. The last full-element acquisition frame before load reduction is used as the historical reference. The interpolation result is compared with the corresponding position of the historical reference frame. If the difference does not exceed the tolerance threshold based on the fluctuation statistics of the historical frame, the historical value is used as the substitute. The interpolated background region data is stitched together with the high-resolution data of the high-priority region according to spatial coordinates to generate a complete detection image.

10. The method for transmitting and receiving array antenna signals based on a TGV glass substrate according to claim 1, characterized in that: The process of restoring full-function acquisition after the temperature drops back to the safety threshold specifically includes: when the substrate temperature is less than or equal to the safety critical temperature minus the preset hysteresis, and the bandwidth occupancy ratio is lower than the reference ratio, the array element bias and signal link are restored channel by channel in the reverse order of shutdown, and the scan time slot is re-inserted.

Images