Semiconductor test alignment method based on developing calibration and semiconductor test system

Abstract

The application discloses a semiconductor test alignment method based on developing calibration and a semiconductor test system. The method comprises the following steps: controlling a plurality of detection probes to press down on calibration medium covering corresponding test acupoints to form physical needle tracks, and obtaining actual alignment coordinates of each detection probe in a global coordinate system through visual recognition; in a test stage, a mechanical hand is controlled to grab a semiconductor chip to be tested, a slip deviation of the semiconductor chip relative to the mechanical hand is obtained through image recognition, and a design position deviation of a test pad relative to a geometric center of the semiconductor chip is obtained; for any target test acupoint, target motion coordinates of the mechanical hand are calculated based on the slip deviation, the design position deviation and the actual alignment coordinates of the corresponding detection probe, and the semiconductor chip is placed. The above method does not need to be equipped with an independent micro-displacement mechanism for each detection probe, and eliminates the problems of space interference, high cost and maintenance difficulty caused by multi-station dense installation of mechanical mechanisms.

Description

Technical Field

[0001] This invention relates to the field of semiconductor testing technology, and in particular to a semiconductor testing alignment method and semiconductor testing system based on development calibration. Background Technology

[0002] In the semiconductor packaging and testing process, the testing equipment needs to precisely place the semiconductor chip under test in the designated position within the test socket and control the probe holder to press down, ensuring that the tiny detection probes on the probe holder make precise contact with the chip's pins (pads) or test points for electrical performance testing. The alignment accuracy between the actual dwell point of the detection probe and the chip's test point directly determines the reliability of the test results. Any misalignment or poor contact will lead to misjudgment of yield or even scratching the chip.

[0003] During the assembly and operation of actual production equipment, due to factors such as machining tolerances and assembly errors, the actual physical landing point of the test probe often deviates from the center coordinates of the test base material. Traditional alignment correction methods typically employ a strategy of dynamically adjusting the test probe. That is, after detecting a positional deviation, a micro-motion module (such as a precision motor, miniature slide rail, micrometer adjusting screw, etc.) is equipped on the pin card assembly to move the pin card or test probe mechanically, attempting to align the test probe with the test point of the chip.

[0004] However, with the continuous advancement of semiconductor manufacturing processes, the aforementioned traditional technology relying on mechanical adjustment pins has revealed many shortcomings in practical applications: First, to improve mass production efficiency, modern semiconductor testing equipment generally adopts a multi-station (e.g., 8 or more slots) simultaneous testing mode. To achieve independent high-precision alignment of each station, each test probe must be equipped with an independent micro-mechanical adjustment mechanism. Densely installing multiple complex motion mechanisms in the extremely limited internal space of the equipment is not only extremely expensive, but also makes it easy for spatial interference and physical collisions to occur between components, resulting in enormous difficulties in equipment assembly and maintenance.

[0005] Secondly, wherever mechanical transmission exists, inherent physical tolerances such as backlash and friction are unavoidable. Currently, semiconductor testing often requires micrometer-level precision (e.g., within 20 micrometers or even 10 micrometers). Under these extreme requirements, mechanical fine-tuning mechanisms often face the dilemma of mismatch between commanded and actual displacements. Repeated mechanical fine-tuning easily leads to tolerance accumulation, resulting in increasingly larger errors with each adjustment, making it difficult to reach the ideal precision limit. Summary of the Invention

[0006] The purpose of this invention is to provide a semiconductor testing alignment method and semiconductor testing system based on development calibration that can achieve micron-level automatic and precise alignment in a multi-station environment without relying on a mechanical adjustment mechanism at the end of the detection probe, with a simpler structure.

[0007] To achieve the above objectives, the present invention provides a semiconductor testing alignment method based on development calibration, comprising: A testing machine is provided, wherein the testing machine has a plurality of test points for receiving semiconductor chips under test; Multiple detection probes are controlled to press down onto the calibration medium covering the corresponding acupoints to form multiple physical needle marks. Acquire stitch images of each of the physical stitches, and process the stitch images based on visual recognition to obtain the actual alignment coordinates of each detection probe in the global coordinate system; During the testing phase, the robotic arm is controlled to grasp the semiconductor chip to be tested, and the slippage deviation of the semiconductor chip relative to the robotic arm is obtained through image recognition. Obtain the design position deviation of the test pads on the semiconductor chip relative to the geometric center of the semiconductor chip; For any target acupoint, the target motion coordinates of the robotic arm are calculated based on the slip deviation, the design position deviation, and the actual alignment coordinates of the corresponding detection probe. The robotic arm is then controlled to move to the target motion coordinates to place the semiconductor chip.

[0008] Preferably, the method for obtaining the actual alignment coordinates includes: The robotic arm is controlled to adsorb and transport the calibration medium to a position above the first image acquisition device for image acquisition, thereby obtaining an image of the first medium. The first medium image is subjected to image recognition processing to obtain the first coordinates of the geometric center of the calibration medium in this state; The robotic arm is controlled to place the calibration medium at the preset center coordinates of the corresponding test acupoint, and the detection probe is controlled to press down to form the physical needle mark; The robotic arm is controlled to return to the preset center coordinates, re-adsorb the calibration medium, and transport it above the first image acquisition device for image acquisition, so as to obtain a second medium image. The second medium image is subjected to image recognition processing to obtain the second coordinates of the geometric center of the calibration medium in this state; Calculate the amount of physical and mechanical slippage generated by the calibration medium during the pressing of the detection probe based on the first coordinate and the second coordinate. When calculating the actual alignment coordinates based on the stitch image, the physical mechanical slippage is deducted.

[0009] Preferably, the method for obtaining the actual alignment coordinates further includes: After obtaining the second coordinates, the robot arm is controlled to transport and place the calibration medium on a calibration platform, and the second image acquisition device located above the calibration platform is used to acquire an image of the calibration medium to obtain the stitch image. The needle mark image is processed by image recognition to obtain the third coordinate of the geometric center of the calibration medium in this state, and the needle mark coordinate of the physical needle mark; The actual alignment coordinates are calculated based on the first coordinate, the second coordinate, the third coordinate, the needle mark coordinate, and the preset center coordinate.

[0010] Preferably, the formula for calculating the actual alignment coordinates (Xtlp, Ytlp) is as follows: Xtlp = Xt1 + (Xlp1 - Xmu1) * Kup - (Xmd2 - Xmd1) * Kdown Ytlp = Yt1 + (Ylp1 - Ymu1) * Kup - (Ymd2 - Ymd1) * Kdown Wherein, (Xt1, Yt1) are the preset center coordinates, (Xmu1, Ymu1) are the third coordinates, (Xlp1, Ylp1) are the needle mark coordinates, (Xmd1, Ymd1) are the first coordinates, (Xmd2, Ymd2) are the second coordinates, Kup is the pixel resolution coefficient of the second image acquisition device, and Kdown is the pixel resolution coefficient of the first image acquisition device.

[0011] Preferably, during the testing phase, while obtaining the sliding deviation of the semiconductor chip relative to the robotic arm through image recognition, the angular deviation of the semiconductor chip relative to the robotic arm is also obtained. Before controlling the robotic arm to move to the target motion coordinate, the method further includes: controlling the robotic arm to perform rotational correction based on the angle deviation.

[0012] Preferably, the calibration medium comprises a thin film medium that covers the test acupoint and is capable of producing identifiable physical or chemical changes under pressure.

[0013] Preferably, any one of pressure-sensitive developing paper, developing colorant coated on a physical surface, toner coating, or a dark metal plate that has undergone hard anodizing.

[0014] Preferably, the plurality of test acupoints on the test machine are arranged in an array; The actual alignment coordinates of the detection probes are calculated and stored for each of the multiple test acupoints to establish a coordinate compensation data table for all the test acupoints.

[0015] The present invention also provides a semiconductor testing system, comprising: The testing machine has multiple test points for receiving semiconductor chips under test; A probe group located above the testing machine, the probe group comprising multiple detection probes respectively corresponding to each of the testing acupoints; A robotic arm is used to grasp and move calibration media and semiconductor chips to be tested; A controller, connected to the probe group and the robotic arm, performs semiconductor chip loading control based on the semiconductor testing alignment method described above.

[0016] Compared to existing technologies, the semiconductor testing alignment method provided by the above-mentioned technical solution no longer dynamically moves the detection probes. Instead, it visually calibrates the physical landing points of each detection probe into actual alignment coordinates in a global coordinate system using the physical stitches of the calibration medium. This transforms the probe position deviations caused by manufacturing and assembly into known static benchmarks. During the testing phase, by controlling the placement of the semiconductor chip by the robotic arm, and comprehensively compensating for the random slip deviation of the semiconductor chip and the design position deviation of the test pads, the chip test pads can be precisely aligned with the actual landing points of the probes. Therefore, there is no need to equip each detection probe with an independent micro-motion mechanism, fundamentally eliminating the spatial interference, high cost, and maintenance problems caused by densely installed mechanical mechanisms at multiple workstations. It also avoids backlash, friction, and tolerance accumulation in mechanical transmission, achieving micron-level high-precision and high-stability alignment, while improving the efficiency and yield of simultaneous multi-station testing. Attached Figure Description

[0017] Figure 1 This is a flowchart of the semiconductor testing alignment method in an embodiment of the present invention.

[0018] Figure 2 This is a flowchart illustrating the actual alignment coordinate method in an embodiment of the invention. Detailed Implementation

[0019] To illustrate the technical content, structural features, objectives, and effects of the present invention in detail, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0020] This embodiment provides a semiconductor testing alignment method based on development calibration, which is mainly applied in the semiconductor packaging and testing process. It accurately places the semiconductor chip under test into the test slot of the testing machine, so that the detection probe on the testing machine can make precise contact with the test pads on the semiconductor chip, thereby completing the electrical performance test.

[0021] First, it should be noted that the testing machine has multiple test slots for receiving semiconductor chips under test. These test slots support the semiconductor chip, and each test slot has a detection probe positioned above it. During testing, the detection probe presses down to contact the test pads of the semiconductor chip. The number of test slots can be determined according to the scale of the testing station, for example, 8, 16, or more, to accommodate the mass production needs of multi-station simultaneous testing.

[0022] Based on this, such as Figure 1 The alignment method in this embodiment includes the following steps: S1: Controls the movement of multiple detection probes, which press down onto the calibration medium covering the corresponding test acupoints to form multiple physical needle marks.

[0023] During this calibration phase, a calibration medium is applied to each acupoint. Under the pressure of the probe, this medium leaves identifiable physical needle marks at the contact points. Due to individual differences in the actual landing points of each probe, the positions of the physical needle marks left on the calibration medium also vary. These physical needle marks are the physical mapping of the actual landing points of the probes.

[0024] S2: Obtain the stitch images of each of the physical needle marks, and process the stitch images based on visual recognition to obtain the actual alignment coordinates of each of the detection probes in the global coordinate system.

[0025] A calibration medium with physical needle marks is placed within the field of view of an image acquisition device to acquire needle mark images. Image recognition algorithms (such as spot localization algorithms based on grayscale threshold segmentation, edge detection, or template matching) are used to identify the pixel positions of each physical needle mark in the image. Then, combined with the pixel resolution coefficient of the image acquisition device and the placement reference of the calibration medium in the test acupoint, the pixel positions are converted into coordinates in the global coordinate system, thus obtaining the actual alignment coordinates of each detection probe.

[0026] A global coordinate system refers to a coordinate system established with a fixed reference point on the testing machine as the origin, which can uniformly describe the movement of the robotic arm and the positions of each testing acupoint. Through this step, the landing point of the detection probe, which was originally unknown due to machining and assembly errors, is transformed into a set of known coordinate data that can be used for subsequent calculations.

[0027] S3: During the testing phase, the robotic arm is controlled to grasp the semiconductor chip to be tested, and the slip deviation of the semiconductor chip relative to the robotic arm is obtained through image recognition.

[0028] When a robotic arm picks up semiconductor chips from a tray, due to the limitations of repeatability and positioning accuracy inherent in the picking action itself, as well as slight differences in the initial placement of the chips in the tray, the geometric center of the picked chip is often not perfectly aligned with the robotic arm's picking reference point (e.g., the center of the suction nozzle). Instead, there is a random positional offset, known as slip deviation. By using an image recognition device to photograph and identify the semiconductor chip in its picked state, the actual position of the chip relative to the robotic arm can be measured, and the slip deviation can be calculated. This slip deviation is a random quantity that may vary with each pick and must be measured before each placement.

[0029] S4: Obtain the design position deviation of the test pads on the semiconductor chip relative to the geometric center of the semiconductor chip.

[0030] The test pads on a semiconductor chip, which are used to contact the test probes, are positioned on the semiconductor chip layout by design. They are usually not located at the geometric center of the semiconductor chip, but rather have a fixed offset relative to the geometric center. This offset is the design position deviation.

[0031] Unlike slip deviation, design position deviation is a constant known quantity for the same type of semiconductor chip. It can be directly read from the design data or measured and stored in advance, without the need to be repeatedly identified each time it is placed.

[0032] S5: For any target acupoint, based on the slip deviation, the design position deviation, and the actual alignment coordinates of the corresponding detection probe, calculate the target motion coordinates of the robotic arm, and control the robotic arm to move to the target motion coordinates to place the semiconductor chip.

[0033] For a target test point, the actual alignment coordinates of its corresponding detection probe have been obtained in step S2 above. These coordinates are the target point that the semiconductor chip test pad ultimately needs to be aligned with. To ensure that the test pad lands on this target point, the robotic arm needs to compensate for two types of data simultaneously during placement: one is the sliding deviation of the semiconductor chip relative to the robotic arm, and the other is the design position deviation of the test pad relative to the geometric center of the semiconductor chip. By superimposing these two deviations on the actual alignment coordinates, the target motion coordinates that the robotic arm should reach can be calculated. After the robotic arm moves to the target motion coordinates and places the semiconductor chip, the test pad of the semiconductor chip will accurately land at the actual landing point of the detection probe.

[0034] Let the actual alignment coordinates of the probe corresponding to a certain acupoint be (Xtlp, Ytlp), which is the target point that the test pad needs to reach; let the slip deviation be (Δxs, Δys), representing the offset of the geometric center of the semiconductor chip relative to the robot's reference point; let the design position deviation be (Δxd, Δyd), representing the offset of the test pad relative to the geometric center of the semiconductor chip. When the robot's reference point is located at coordinates (X, Y), the actual landing point of the test pad is (X+Δxs+Δxd, Y+Δys+Δyd). By setting the actual landing point of the test pad equal to the target point, we can inversely solve for the target motion coordinates that the robot's reference point should reach: X_target = Xtlp Δxs Δxd, Y_target = Ytlp Δys Δyd. The robotic arm moves and places itself according to this coordinate, thus completing a precise alignment.

[0035] Those skilled in the art should understand that the compensation order and superposition form of the aforementioned slip deviation and design position deviation can be equivalently adjusted according to the definition of the actual coordinate system, as long as the final landing point of the test pad is consistent with the actual alignment coordinate of the detection probe, and is not limited to any specific positive or negative sign expression. Furthermore, the robotic arm can be a vacuum suction nozzle pick-and-place mechanism or a micro-gripper grasping mechanism, as long as it can grasp, move, and place the semiconductor chip; the image recognition can be achieved using an industrial camera combined with a vision algorithm, or it can be achieved using other sensing methods capable of acquiring object position information.

[0036] Because this embodiment calibrates the actual landing point of the detection probe to the known coordinates in the global coordinate system in one go, and completes the alignment only by controlling the placement coordinates of the robot during the testing phase, there is no need to equip each test acupoint with an independent micro-motion displacement mechanism to move the detection probe. This avoids the problems of spatial interference, high cost and maintenance difficulties caused by densely installing multiple sets of mechanical motion mechanisms in a limited equipment space in the scenario of simultaneous testing at multiple workstations. At the same time, since the alignment compensation is completed entirely at the coordinate calculation level, it no longer relies on mechanical fine-tuning transmission, thus eliminating the inherent backlash, friction and tolerance accumulation caused by multiple fine-tuning of mechanical transmission, so that the alignment accuracy can stably reach the micron level.

[0037] On the other hand, considering that the calibration medium may move slightly during the pressing of the detection probe, thus introducing errors, a mechanism for measuring and subtracting the physical and mechanical slippage of the calibration medium is introduced when obtaining the actual alignment coordinates. For example... Figure 2 The method for obtaining the actual alignment coordinates specifically includes the following steps: S21: Control the robotic arm to adsorb and transport the calibration medium to the top of the first image acquisition device for image acquisition, so as to obtain the first medium image.

[0038] S22: Perform image recognition processing on the first medium image to obtain the first coordinates (Xmd1, Ymd1) of the geometric center of the calibration medium in this state.

[0039] The first image acquisition device here can be a camera set below the robot's handling path, used to capture the outline of the calibration medium when it is held by the robot, thereby determining the geometric center position of the calibration medium at this moment.

[0040] S23: Control the robotic arm to place the calibration medium at the preset center coordinates (Xt1, Yt1) of the corresponding test acupoint, and control the detection probe to press down to form the physical needle mark.

[0041] The preset center coordinates are the theoretical center position of the acupoint in the global coordinate system. The robotic arm places the calibration medium according to these coordinates, and then the probe is pressed down to leave a mark.

[0042] S24: Control the robotic arm to return to the preset center coordinates, re-adsorb the calibration medium, and transport it above the first image acquisition device to acquire an image of the second medium.

[0043] S25: Perform image recognition processing on the second medium image to obtain the second coordinates (Xmd2, Ymd2) of the geometric center of the calibration medium in this state.

[0044] When the probe is pressed down, it applies pressure to the calibration medium, which may cause the calibration medium to slide slightly relative to the test acupoint. Therefore, after pressing down, the calibration medium is re-absorbed and its geometric center in the field of view of the first image acquisition device is measured again. The change in the geometric center between the two measurements can reflect this sliding.

[0045] Based on the first coordinate and the second coordinate, the physical and mechanical slippage of the calibration medium during the pressing of the detection probe can be calculated. This slippage is (Xmd2) Xmd1, Ymd2 The positional change represented by Ymd1).

[0046] Then, when calculating the actual alignment coordinates based on the needle trace image, the physical mechanical slippage is deducted to eliminate the interference of the calibration medium's own movement on the actual landing point positioning of the detection probe.

[0047] Because the act of the probe leaving a mark may cause slight slippage in the calibration medium, without correction, the identified physical needle mark position will simultaneously include errors from both the actual landing point of the probe and the slippage of the calibration medium. This embodiment measures the geometric center of the calibration medium before and after pressing down, separating and subtracting the slippage of the calibration medium. This ensures that the final calculated alignment coordinates only reflect the actual landing point of the probe, improving calibration accuracy.

[0048] Furthermore, after obtaining the second coordinates, the stitch image is obtained and the actual alignment coordinates are fully calculated through the following steps.

[0049] S26: Control the robotic arm to transport and place the calibration medium on a calibration platform, and use a second image acquisition device located above the calibration platform to acquire an image of the calibration medium to obtain the stitch image.

[0050] The second image acquisition device can be a camera positioned above the calibration platform, with its field of view facing the calibration platform, for clearly reading the physical needle marks on the calibration medium.

[0051] S27: Perform image recognition processing on the stitch image to obtain the third coordinates (Xmu1, Ymu1) of the geometric center of the calibration medium in this state, and the coordinates (Xlp1, Ylp1) of the physical stitch.

[0052] In the same needle mark image, both the geometric center of the calibration medium and the center position of the physical needle mark are identified. Since they are on the same coordinate reference, it is convenient for subsequent relative calculations.

[0053] S28: Based on the first coordinate, the second coordinate, the third coordinate, the needle mark coordinate, and the preset center coordinate, the actual alignment coordinate is calculated.

[0054] Specifically, the formula for calculating the actual alignment coordinates (Xtlp, Ytlp) is as follows: Xtlp=Xt1+(Xlp1 Xmu1)*Kup (Xmd2 Xmd1)*Kdown Ytlp=Yt1+(Ylp1 Ymu1)*Kup (Ymd2 Ymd1)*Kdown Wherein, (Xt1, Yt1) are the preset center coordinates, (Xmu1, Ymu1) are the third coordinates, (Xlp1, Ylp1) are the needle mark coordinates, (Xmd1, Ymd1) are the first coordinates, (Xmd2, Ymd2) are the second coordinates, Kup is the pixel resolution coefficient of the second image acquisition device, and Kdown is the pixel resolution coefficient of the first image acquisition device.

[0055] The physical meaning of the above formula is as follows: (Xlp1) Xmu1) and (Ylp1) Ymu1) represents the pixel offset of the physical needle mark relative to the geometric center of the calibration medium. Multiplying it by the pixel resolution coefficient Kup of the second image acquisition device converts it into the actual distance of the detection probe landing point relative to the center of the calibration medium in the global coordinate system.

[0056] (Xmd2 Xmd1) and (Ymd2) Ymd1) represents the pixel displacement of the calibration medium before and after pressing. Multiplying it by the pixel resolution coefficient Kdown of the first image acquisition device gives the physical and mechanical slippage of the calibration medium itself during the pressing process.

[0057] Adding the former to the preset center coordinates and subtracting the latter gives the actual landing point of the detection probe in the global coordinate system, which is also the actual alignment coordinate.

[0058] Let the preset center coordinates of a target acupoint be (Xt1, Yt1) = (100.000 mm, 100.000 mm). When the calibration medium is held, the first image acquisition device identifies the first coordinates (Xmd1, Ymd1) = (512, 480) (in pixels). After the detection probe is pressed down and leaves a mark, it is re-held and identified, resulting in the second coordinates (Xmd2, Ymd2) = (516, 483) (in pixels). The pixel resolution coefficient of the first image acquisition device is Kdown = 0.005 mm / pixel. When the calibration medium is moved to the calibration platform, the second image acquisition device identifies the third coordinates (Xmu1, Ymu1) = (640, 512), and the physical needle mark coordinates (Xlp1, Ylp1) = (648, 505) (both in pixels). The pixel resolution coefficient of the second image acquisition device is Kup = 0.004 mm / pixel. Substituting into the above formula: Xtlp = 100.000 + (648 640) × 0.004 (516 512) × 0.005 = 100.000 + 0.032 0.020 = 100.012 mm; Ytlp = 100.000 + (505 512) × 0.004 (483 480) × 0.005 = 100.000 0.028 0.015 = 99.957 mm.

[0059] Therefore, the actual alignment coordinates of the detection probe are (100.012 mm, 99.957 mm), meaning that the actual landing point of the detection probe is offset by +0.012 mm in the X direction and offset by +0.012 mm in the Y direction relative to the preset center coordinates. 0.043 mm. If the physical and mechanical slippage of the calibration medium is not deducted, errors of 0.020 mm and 0.015 mm will be introduced in the X and Y directions, respectively. Under the micron-level alignment requirements, this error is not negligible.

[0060] On the other hand, when a semiconductor chip is grasped by a robotic arm, in addition to the possible translational slippage of the geometric center relative to the robotic arm's reference point, the semiconductor chip itself may also rotate at a certain angle relative to the robotic arm's reference direction. If not corrected, even if the translational alignment is accurate, the orientation of the test pad array will still be inconsistent with the orientation of the detection probe array, resulting in some test pads and detection probes making misaligned contact.

[0061] Therefore, during the testing phase, while obtaining the sliding deviation of the semiconductor chip relative to the robotic arm through image recognition, the angular deviation of the semiconductor chip relative to the robotic arm is also obtained. This angular deviation can be obtained by identifying the edges, positioning marks, or test pad array arrangement direction in the semiconductor chip image and calculating the angle between them and the reference direction of the robotic arm.

[0062] Before controlling the robotic arm to move to the target motion coordinates, the method further includes: controlling the robotic arm to perform rotational correction based on the angle deviation, that is, driving the rotation axis of the robotic arm to return the semiconductor chip to the same orientation as the detection probe array, and then performing translational placement.

[0063] For example, if a +0.3° angular deviation is detected between the semiconductor chip and the robotic arm, the robotic arm will first rotate 0.3° in the opposite direction to straighten the semiconductor chip before placing it, and then complete the translational placement according to the target motion coordinates. In this way, the combination of translational alignment and angular alignment ensures that the test pad array is precisely matched with the detection probe array in both position and orientation.

[0064] On the other hand, the calibration medium includes a thin film covering the test acupoint, capable of producing identifiable physical or chemical changes under pressure. The core requirement is that when the detection probe is pressed down, the calibration medium leaves a physical needle mark at the contact point that contrasts sharply with the surrounding area and can be reliably identified by the image acquisition device, and the position of this needle mark accurately reflects the landing point of the detection probe. The physical change can be an indentation on the medium surface, plastic deformation, or the surface coating being broken to reveal the underlying color; the chemical change can be a color-developing substance in the medium undergoing a color reaction and changing color under pressure.

[0065] Specifically, the calibration medium can be any one of pressure-sensitive developing paper, developing colorant coated on a physical surface, toner coating, or a dark metal plate that has undergone hard anodizing.

[0066] Among them, the pressure-sensitive developing paper develops color at the pressure area, forming clear red or blue pin dots; the developing colorant coated on the physical surface shows color spots at the contact point after the detection probe is pressed; the toner coating is compacted or transferred at the contact point of the detection probe, leaving traces; the dark metal plate with hard anodizing treatment has a hard surface, and when the detection probe is pressed down, it will press out light-colored bright spots on the dark oxide layer, forming a high contrast with the dark background, which is easy to identify. Moreover, the metal plate can be reused repeatedly and is not easily deformed, making it suitable for occasions with high requirements for calibration accuracy and reusability.

[0067] On the other hand, to meet the mass production requirements of simultaneous testing at multiple workstations, the multiple test points on the testing machine are arranged in an array, such as in a 2×4, 4×4, or other row and column configuration, to accommodate multiple semiconductor chips under test in a single test. Since the detection probes corresponding to each test point have their own different processing and assembly deviations, their actual alignment coordinates are independent and different from each other.

[0068] Therefore, the actual alignment coordinates of the detection probes for each of the multiple test acupoints are calculated and stored to establish a coordinate compensation data table for all the test acupoints. During the calibration phase, the actual alignment coordinates of the detection probes for each test acupoint are obtained and stored in the coordinate compensation data table using the test acupoint number as an index.

[0069] During the testing phase, when the robotic arm places a semiconductor chip into a certain test acupoint, it directly retrieves the corresponding actual alignment coordinates from the coordinate compensation data table using the acupoint number, and then calculates the target motion coordinates by combining the actual slip deviation measured that time with the known design position deviation.

[0070] In another preferred embodiment of the present invention, a semiconductor testing system for implementing the aforementioned semiconductor testing alignment method is also provided, which includes a testing platform, a probe group, a robotic arm, and a controller.

[0071] The testing machine has multiple test points for receiving semiconductor chips under test. These test points are arranged in an array to support multiple semiconductor chips under test during the same test.

[0072] The probe group is located above the testing machine and includes multiple detection probes respectively corresponding to each of the test acupoints. Each test acupoint corresponds to a group or one detection probe, which is used to press down on the calibration medium covering the test acupoint to leave an indentation during the calibration phase, and to contact the test pads of the semiconductor chip during the testing phase to complete the electrical performance test.

[0073] The robotic arm is used to grasp and transport the calibration medium and the semiconductor chip to be tested. During the calibration phase, the robotic arm is responsible for holding the calibration medium and reciprocating between the first image acquisition device, the test acupoint, and the calibration platform, thus completing the acquisition of medium and needle mark images. During the testing phase, the robotic arm is responsible for grasping the semiconductor chip to be tested from the tray and placing the semiconductor chip onto the corresponding test acupoint according to the calculated target motion coordinates. The robotic arm can have translational and rotational degrees of freedom to support the rotational correction described in Embodiment 3.

[0074] The controller is connected to the probe group and the robot arm. The controller performs semiconductor chip loading control based on the semiconductor testing alignment method described in the foregoing embodiments. Specifically, the controller receives image recognition results from the image acquisition device, calculates the actual alignment coordinates, slip deviation, angle deviation, and target motion coordinates, and outputs motion commands to the robot arm and pressing commands to the probe group accordingly, thereby coordinating the orderly operation of each component according to the aforementioned method flow.

[0075] This semiconductor testing system unifies the image acquisition and coordinate calculation required for the calibration stage with the grasping, placement, and marking actions required for the testing stage under a unified controller. In terms of hardware, there is no need to configure an independent micro-motion displacement mechanism for each test point. The calibration and alignment compensation of all test points can be completed by a single set of robotic arms and image acquisition devices. This simplifies the mechanical structure of multi-station simultaneous testing equipment while ensuring micron-level alignment accuracy, and reduces equipment costs and maintenance complexity.

[0076] In summary, this invention discloses a semiconductor testing alignment method and system based on development calibration. The method provides a testing machine with multiple testing points, controlling multiple detection probes to press down onto a calibration medium covering the corresponding testing points to form physical needle marks. The needle mark images are acquired and processed by visual recognition to obtain the actual alignment coordinates of each detection probe in the global coordinate system. During the testing phase, a robotic arm is controlled to grasp the semiconductor chip under test. Image recognition is used to obtain the sliding deviation of the semiconductor chip relative to the robotic arm, and the design position deviation of the test pads relative to the geometric center of the semiconductor chip is also obtained. For the target testing point, the target motion coordinates of the robotic arm are calculated based on the sliding deviation, the design position deviation, and the actual alignment coordinates of the corresponding detection probes, and the semiconductor chip is placed. This invention calibrates the actual landing point of the detection probes to known coordinates in one step, completing alignment compensation only by controlling the placement action of the robotic arm. It eliminates the need for an independent micro-motion displacement mechanism for each testing point, eliminating backlash, friction, and tolerance accumulation in mechanical transmission, and achieving micron-level high-precision, high-stability multi-station simultaneous testing alignment.

[0077] The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A semiconductor testing alignment method based on development calibration, characterized in that, include: A testing machine is provided, wherein the testing machine has a plurality of test points for receiving semiconductor chips under test; Multiple detection probes are controlled to press down onto the calibration medium covering the corresponding acupoints to form multiple physical needle marks. Acquire stitch images of each of the physical stitches, and process the stitch images based on visual recognition to obtain the actual alignment coordinates of each detection probe in the global coordinate system; During the testing phase, the robotic arm is controlled to grasp the semiconductor chip to be tested, and the slippage deviation of the semiconductor chip relative to the robotic arm is obtained through image recognition. Obtain the design position deviation of the test pads on the semiconductor chip relative to the geometric center of the semiconductor chip; For any target acupoint, the target motion coordinates of the robotic arm are calculated based on the slip deviation, the design position deviation, and the actual alignment coordinates of the corresponding detection probe. The robotic arm is then controlled to move to the target motion coordinates to place the semiconductor chip.

2. The semiconductor testing alignment method according to claim 1, characterized in that, The methods for obtaining the actual alignment coordinates include: The robotic arm is controlled to adsorb and transport the calibration medium to a position above the first image acquisition device for image acquisition, thereby obtaining an image of the first medium. The first medium image is subjected to image recognition processing to obtain the first coordinates of the geometric center of the calibration medium in this state; The robotic arm is controlled to place the calibration medium at the preset center coordinates of the corresponding test acupoint, and the detection probe is controlled to press down to form the physical needle mark; The robotic arm is controlled to return to the preset center coordinates, re-adsorb the calibration medium, and transport it above the first image acquisition device for image acquisition, so as to obtain a second medium image. The second medium image is subjected to image recognition processing to obtain the second coordinates of the geometric center of the calibration medium in this state; Calculate the amount of physical and mechanical slippage generated by the calibration medium during the pressing of the detection probe based on the first coordinate and the second coordinate. When calculating the actual alignment coordinates based on the stitch image, the physical mechanical slippage is deducted.

3. The semiconductor testing alignment method according to claim 2, characterized in that, The method for obtaining the actual alignment coordinates also includes: After obtaining the second coordinates, the robot arm is controlled to transport and place the calibration medium on a calibration platform, and the second image acquisition device located above the calibration platform is used to acquire an image of the calibration medium to obtain the stitch image. The needle mark image is processed by image recognition to obtain the third coordinate of the geometric center of the calibration medium in this state, and the needle mark coordinate of the physical needle mark; The actual alignment coordinates are calculated based on the first coordinate, the second coordinate, the third coordinate, the needle mark coordinate, and the preset center coordinate.

4. The semiconductor testing alignment method according to claim 3, characterized in that, The formula for calculating the actual alignment coordinates (Xtlp, Ytlp) is as follows: Xtlp = Xt1 + (Xlp1 - Xmu1)*Kup - (Xmd2 - Xmd1)*Kdown Ytlp = Yt1 + (Ylp1 - Ymu1)*Kup - (Ymd2 - Ymd1)*Kdown Wherein, (Xt1, Yt1) are the preset center coordinates, (Xmu1, Ymu1) are the third coordinates, (Xlp1, Ylp1) are the needle mark coordinates, (Xmd1, Ymd1) are the first coordinates, (Xmd2, Ymd2) are the second coordinates, Kup is the pixel resolution coefficient of the second image acquisition device, and Kdown is the pixel resolution coefficient of the first image acquisition device.

5. The semiconductor testing alignment method according to claim 1, characterized in that, During the testing phase, the sliding deviation of the semiconductor chip relative to the robotic arm is obtained through image recognition, and the angular deviation of the semiconductor chip relative to the robotic arm is also obtained. Before controlling the robotic arm to move to the target motion coordinate, the method further includes: controlling the robotic arm to perform rotational correction based on the angle deviation.

6. The semiconductor testing alignment method according to claim 1, characterized in that, The calibration medium includes a thin film that covers the test acupoint and is capable of producing identifiable physical or chemical changes under pressure.

7. The semiconductor testing alignment method according to claim 1, characterized in that, The calibration medium includes any one of pressure-sensitive developing paper, developing colorant coated on a physical surface, toner coating, or a dark metal plate that has undergone hard anodizing.

8. The semiconductor testing alignment method according to claim 1, characterized in that, The multiple test acupoints on the test machine are arranged in an array; The actual alignment coordinates of the detection probes are calculated and stored for each of the multiple test acupoints to establish a coordinate compensation data table for all the test acupoints.

9. A semiconductor testing system, characterized in that, include: The testing machine has multiple test points for receiving semiconductor chips under test; A probe group located above the testing machine, the probe group comprising multiple detection probes respectively corresponding to each of the testing acupoints; A robotic arm is used to grasp and move calibration media and semiconductor chips to be tested; A controller, connected to the probe group and the robotic arm, performs semiconductor chip loading control based on the semiconductor testing alignment method according to any one of claims 1 to 8.

Images