A direct-write lithography alignment system, alignment method, and signal transmission delay calibration method

By employing adaptive uniform linear motion fuzzy restoration processing and closed-loop delay calibration methods, the positioning accuracy problem caused by motion fuzziness and trigger delay in direct-write lithography technology was solved, achieving efficient and accurate lithography alignment and improving the positioning accuracy and consistency of the equipment.

CN123113251APending Publication Date: 2026-11-17SUZHOU YUANZHUO OPTOELECTRONICS TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611154277.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing direct-write lithography technology suffers from motion ambiguity during uniform motion alignment, leading to decreased positioning accuracy and systematic errors introduced by trigger delay. This makes it difficult to meet high-precision alignment requirements. Furthermore, the hardware delay varies between different devices, affecting the consistency of alignment accuracy across batches of equipment.

Method used

An adaptive uniform linear motion blur restoration process is adopted. Gradient magnitude maps are extracted using Wiener filtering and the Sobel operator to construct a referenceless image quality score. The mean-variance ratio of the gradient magnitude is used as an evaluation index to achieve sub-pixel level pixel coordinate extraction. At the same time, by splitting the trigger signal link into two stages, namely the inherent hardware delay and the software compensation delay, closed-loop calibration is performed using the symmetrical error characteristics of positive and negative motion to eliminate systematic position deviation.

Benefits of technology

It improves restoration accuracy and anti-interference capability, suppresses ringing artifacts during the restoration process, enhances pixel-level positioning accuracy, realizes full-link accuracy control, and ensures consistent alignment accuracy and high efficiency for batch devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113251A_ABST
    Figure CN123113251A_ABST
Patent Text Reader

Abstract

This invention discloses a direct-write lithography alignment system, alignment method, and signal transmission delay calibration method, relating to the field of lithography equipment technology. The system includes a fixed reference stage, a platform position control device, a loading platform, and an image acquisition device. It also includes an image processing module and a coordinate transformation module. The image processing module constructs a search space for motion blur point spread function parameters, generates candidate restored images using Wiener filtering, extracts gradient magnitudes using the Sobel operator, constructs a referenceless quality score, and adaptively selects the optimal parameters before extracting the sub-pixel coordinates. The coordinate transformation module combines synchronously latched platform coordinates to complete multi-coordinate system mapping. The system supports both direct triggering and delayed triggering acquisition modes, and the accompanying delay calibration method can calibrate the inherent hardware delay in a closed-loop manner. This invention significantly improves alignment accuracy and anti-interference capability in high-speed uniform motion acquisition scenarios, while balancing detection efficiency and stability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of photolithography equipment technology, specifically to a direct-write photolithography alignment system, alignment method, and signal transmission delay calibration method. Background Technology

[0002] Direct-write lithography is one of the core processes in the manufacturing of printed circuit boards, semiconductor devices, and micro / nano structures. Before the exposure process, an alignment system is needed to identify alignment marks on the substrate and obtain the precise position coordinates of these marks. This is used to calibrate the exposure coordinate system, and the alignment accuracy directly determines the final product's process yield and manufacturing quality. Traditional stepper alignment solutions require the platform to stop and stabilize above each alignment mark before taking an image. When there are many marks on the substrate, frequent acceleration and deceleration of the platform significantly increases alignment time, limiting the overall capacity of the equipment. To improve efficiency, the industry is gradually adopting uniform motion alignment technology, which completes the mark acquisition by triggering the acquisition during the uniform motion of the platform, eliminating platform downtime.

[0003] However, as the platform's movement speed continues to increase, existing solutions have gradually revealed two major accuracy bottlenecks: First, during camera exposure, the marker continuously shifts with the platform, generating directional motion blur on the image plane, resulting in blurred marker edge features and directly reducing coordinate extraction accuracy. Existing motion blur restoration algorithms mostly use point spread functions with fixed parameters, which cannot adapt to on-site speed fluctuations and low signal-to-noise ratio environments; some adaptive restoration solutions use general evaluation metrics, resulting in poor parameter convergence stability, and the restoration accuracy is insufficient to meet the high-precision requirements of lithography alignment. Second, there is an inherent hardware delay from the generation of the trigger signal to the start of the camera shutter. Under high-speed motion, a microsecond-level delay corresponds to a micrometer-level positional deviation, forming a systematic alignment error. Existing solutions mostly rely on empirical values ​​to compensate for the delay, lacking closed-loop calibration methods, resulting in limited compensation accuracy. Furthermore, the hardware delays of different devices vary individually, making it difficult to guarantee the consistency of alignment accuracy for batch devices.

[0004] Therefore, how to retain the high efficiency advantage of uniform motion alignment while simultaneously solving the problems of decreased positioning accuracy caused by motion ambiguity and systematic errors introduced by trigger delay is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution: A direct-write lithography alignment system includes a fixed reference stage, a gantry mechanism, a platform position control device, a loading platform, and an image acquisition device. The gantry mechanism is fixedly mounted on the fixed reference stage. The loading platform is driven by the platform position control device to move at a constant speed along a first direction of main motion, and is used to carry a substrate with alignment marks. The platform position control device is equipped with a position feedback unit for synchronously latching the world coordinate system coordinates of the loading platform at the time of image acquisition. The image acquisition device is fixed to the gantry mechanism for acquiring images of the alignment marks on the substrate. The system also includes an image processing module and a coordinate transformation module. The image processing module is configured to perform adaptive uniform linear motion blur restoration processing: a parameter search space for the motion blur point spread function is defined; each set of candidate parameters is traversed within the parameter search space; a corresponding motion blur point spread function is constructed for each set of candidate parameters; candidate restored images are generated using Wiener filtering; a gradient magnitude map is extracted for each candidate restored image using the Sobel operator; a no-reference image quality score is constructed based on the gradient magnitude; the parameters corresponding to the candidate restored image with the highest score are selected as the optimal spread function parameters; the optimal restored image is output, and sub-pixel coordinates of the alignment markers are extracted. The coordinate transformation module is configured to: based on the preset mapping relationship between the world coordinate system, the platform coordinate system and the camera coordinate system, and combined with the world coordinate system coordinates of the loading platform synchronously latched at the acquisition time and the pre-calibrated world coordinates of the image acquisition device, convert the sub-pixel level pixel coordinates of the alignment mark into physical coordinates under the platform coordinate system, wherein the platform coordinate system is a follow-up reference coordinate system fixed to the loading platform.

[0006] As a further aspect of the present invention: the no-reference image quality score is constructed based on the ratio of the mean to the variance of the gradient magnitude, and the calculation formula is as follows:

[0007] in, The mean gradient magnitude of the candidate restored image. denoted as the gradient magnitude variance of the candidate restored image.

[0008] As a further aspect of the present invention: the noise power spectrum of the Wiener filter is estimated by the image background region: a uniform background region without alignment marks is selected in the image, the variance of the gray value of the region is calculated, and the variance is used as the estimated value of the noise power spectrum.

[0009] As a further aspect of the present invention, the parameter search space is initialized as follows: the initial search range of the blur length is estimated based on the movement speed of the platform and the exposure time of the image acquisition device; the initial search range of the blur angle is determined based on the movement direction of the platform; and the corresponding search step size is set according to the parameter search accuracy requirements.

[0010] As a further aspect of the present invention: the coordinate transformation module performs coordinate calculations based on vector mapping relationships, and the calculation formula is as follows:

[0011] in, To align the physical coordinates of the markers in the table coordinate system, The pre-calibrated coordinates of the image acquisition device in the world coordinate system. The physical coordinates of the alignment mark in the camera coordinate system after pixel equivalent transformation. The coordinates of the platform in the world coordinate system are collected and locked synchronously at all times.

[0012] As a further embodiment of the present invention: the image acquisition device includes an imaging lens, a global shutter camera and a strobe illumination unit; the system also includes a trigger timing control module, which is configured to control the strobe illumination unit to be lit earlier than the camera shutter start time, and the continuous lighting duration of the strobe illumination unit is not shorter than the camera shutter duration.

[0013] As a further aspect of the present invention: the system adopts a direct trigger acquisition mode, and the platform position control device includes a motion controller and a signal conversion module; The host computer calculates the preset platform position P1 corresponding to the alignment mark reaching directly below the image acquisition device based on the CAM drawing of the substrate, and sends it to the motion controller. The motion controller receives the platform coordinates from the position feedback unit in real time. When the platform position matches the preset position P1, it outputs a hardware trigger pulse, which is transmitted to the image acquisition device after being matched by the signal conversion module to trigger image acquisition.

[0014] As a further aspect of the present invention: the system adopts a delayed trigger acquisition mode and also includes a target device, a triggering device and a delay control module; The target device is fixedly installed on the side of the platform and moves synchronously with it, and is a dedicated trigger target independent of the substrate; The triggering device is fixedly installed on the side of the fixed reference platform and is used to output an initial trigger signal when the target device enters its sensing area. The delay control module is connected to the triggering device and the image acquisition device respectively, and is configured to model the transmission link of the trigger signal as a two-level delay: the first level is the inherent hardware delay. The first stage is determined by the physical characteristics of the circuit; the second stage is a software-adjustable compensation delay. ,satisfy Where d is the physical distance between the sensing position of the triggering device and the target acquisition center of the marker to be acquired in the first direction, and the target acquisition center is the fixed imaging position calibrated by the system. The uniform motion speed of the platform is used; the delay control module is used to process the initial trigger signal with a delay and then transmit it to the image acquisition device, so that the shutter start time of the image acquisition device corresponds to the time when the alignment mark arrives at the target acquisition center.

[0015] A direct-write lithography alignment method includes the following steps: S1. Uniform speed acquisition: Control the platform to move at a uniform speed along the first direction of the main motion, trigger the image acquisition device to acquire the original blurred image of the alignment mark on the substrate, and synchronously lock the world coordinate system coordinates of the platform at the acquisition time; S2. Adaptive Restoration: Define the parameter search space of the motion blur point spread function, traverse each group of candidate parameters and construct the corresponding motion blur point spread function, generate candidate restored images through Wiener filtering; extract the gradient magnitude map for each candidate restored image using the Sobel operator, construct a no-reference image quality score based on the gradient magnitude, and select the image with the highest score as the optimal restored image. S3. Coordinate Extraction: Extract sub-pixel level pixel coordinates of the alignment markers from the optimal restored image, and convert them into physical coordinates in the camera coordinate system after pixel equivalent calibration; S4. Coordinate Transformation: Based on the preset mapping relationship between the world coordinate system, the platform coordinate system, and the camera coordinate system, and combined with the real-time coordinates of the platform at the time of acquisition, the physical coordinates of the alignment mark are transformed into physical coordinates under the platform coordinate system.

[0016] A method for calibrating the signal transmission delay of an alignment system, applied to a direct-write lithography alignment system in delay-triggered acquisition mode, includes the following steps: P1. Initial Configuration: Obtain the physical distance d between the trigger device's sensing position and the target acquisition center of the marker to be acquired in the first direction; place a calibration reference marker on the platform and set the platform's uniform motion speed to v; initially set the inherent hardware delay. Software-compensated latency ; P2. Forward Acquisition: Control the platform to move at a constant speed in the forward direction, and acquire reference marker images multiple times. Perform adaptive motion blur restoration, marker coordinate extraction, and coordinate transformation on each acquired image to obtain the platform coordinates of the reference marker. Take the average of the multiple measurement results to obtain the forward coordinates. ; P3. Negative Acquisition: Control the platform to move at a constant speed along the negative direction, perform the same image processing and coordinate transformation process as the positive acquisition, and obtain the negative coordinates by taking the average of multiple measurement results. ; P4. Delay Calculation: According to the formula Calculate the inherent latency of hardware And update the software compensation delay value simultaneously. ; P5. Iterative verification: Repeat steps P2 to P4, gradually adjusting the delay parameters until the difference between the positive and negative coordinates is less than the preset alignment accuracy threshold, and the calibration is complete.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. High restoration accuracy and strong anti-interference capability. The ratio of the mean to variance of the gradient magnitude is used as a no-reference quality evaluation index, taking into account both edge sharpness and noise suppression level. Compared with single general indicators such as gradient energy and image entropy, it exhibits stronger convergence stability in parameter optimization under low signal-to-noise ratio environments in industrial settings. It can effectively suppress ringing artifacts during the restoration process, improving sub-pixel coordinate extraction accuracy from the source.

[0018] 2. Systematic delay errors can be calibrated in a closed loop. The triggering link is split into two levels: inherent hardware delay and software-compensated delay. Combined with the symmetrical error characteristics of positive and negative reciprocating motion, accurate calibration of hardware delay can be completed without the introduction of additional precision measuring equipment. This eliminates the systematic positional deviation caused by trigger delay at its source and improves the accuracy consistency of batch equipment.

[0019] 3. Strong architectural compatibility and flexible engineering implementation. It supports both direct triggering and delayed triggering hardware architectures, and the core algorithm is reusable. It can be adapted to the built-in triggering schemes of mainstream motion controllers, as well as meet the special installation scenarios of external sensors, and can flexibly match different device configuration requirements.

[0020] 4. Balancing efficiency and accuracy. The entire process requires no platform stoppage, preserving the high efficiency of uniform motion alignment; simultaneously, algorithmic restoration improves pixel-level positioning accuracy, and delay calibration eliminates system-level position errors, achieving end-to-end accuracy control. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall structure of the direct-write lithography alignment system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the three-level coordinate system vector mapping relationship provided in an embodiment of the present invention; Figure 3 This is a flowchart of the adaptive motion blur restoration algorithm provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the direct trigger acquisition mode link principle provided in an embodiment of the present invention; Figure 5 This is a comparative diagram of motion blur restoration effects provided in an embodiment of the present invention; Figure 6This is a schematic flowchart of the signal transmission delay calibration method provided in an embodiment of the present invention.

[0022] The components in the diagram are labeled as follows: 1-Fixed reference platform, 2-Platform position control device, 3-Loading platform, 4-Triggering device, 5-Target surface device, 6-Gantry mechanism, 7-Image acquisition device. Detailed Implementation

[0023] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. The following embodiments are implemented based on the technical solution of the present invention, and detailed implementation methods and specific operation procedures are given, but the protection scope of the present invention is not limited to the following embodiments.

[0024] Example 1 Reference Figure 1 The direct-write lithography alignment system provided in this embodiment includes a fixed reference stage 1, a platform position control device 2, a loading platform 3, a gantry mechanism 6, an image acquisition device 7, and a computing unit.

[0025] The fixed reference stage 1 adopts a high-rigidity marble base, which provides an absolute installation reference for the entire system and suppresses the influence of external vibrations on imaging accuracy.

[0026] The platform position control device 2 includes a linear guide rail assembly, a servo drive unit, a motion controller, and a position feedback unit. The loading platform 3 is mounted on the slider of the linear guide rail and is driven by the servo drive unit to perform uniform linear motion along the first direction of the main motion. The position feedback unit collects the physical position of the loading platform 3 in real time and feeds it back to the motion controller. The motion controller can latch the platform's world coordinate system coordinates at the current moment while outputting the trigger signal, ensuring strict synchronization between the trigger moment and the coordinate moment.

[0027] Optionally, the position feedback unit may use precision position detection devices such as optical grating rulers or magnetic grating rulers; the motion controller may use an ACS motion controller, which has high-precision position comparison and hardware trigger output capabilities.

[0028] The bearing surface of the carrier platform 3 is used to fix the substrate to be exposed. The substrate surface has pre-made alignment marks of a specific shape as an identification reference for position alignment.

[0029] The gantry mechanism 6 spans above the movement path of the loading platform 3, with its bottom fixedly mounted on the fixed reference platform 1. The image acquisition device 7 is fixedly mounted on the crossbeam of the gantry mechanism 6, with its shooting direction perpendicular to the bearing surface of the loading platform 3. The image acquisition device 7 includes an imaging lens, a global shutter camera, and a strobe illumination unit; the imaging lens is preferably a telecentric lens, which can ensure consistent imaging magnification at different object distances and eliminate perspective distortion; the strobe illumination unit is preferably an LED strobe light source, synchronized with the camera triggering sequence, providing high brightness and high uniformity illumination during the exposure cycle.

[0030] The computing unit is an industrial control computer that runs dedicated alignment software. It has a built-in image processing module and coordinate transformation module, which are responsible for image restoration, mark positioning and coordinate transformation calculation, respectively, and finally output the alignment mark coordinates in the table coordinate system.

[0031] When the alignment marks on the substrate are arranged in a multi-row, multi-column array, the platform 3 first moves laterally along the second direction perpendicular to the first direction, so that the entire column of marks is aligned with the image acquisition device 7 in the second direction; then it moves at a constant speed along the first direction, and the acquisition is triggered sequentially according to the spacing of adjacent marks and the movement speed. The batch acquisition of the entire column of marks is completed in a single constant speed movement, which can reduce the number of platform starts and stops and improve alignment efficiency.

[0032] Example 2 This embodiment details the principle and implementation process of adaptive motion blur restoration in the image processing module, corresponding to... Figure 3 Motion blur caused by uniform linear motion can be modeled as a convolution process between the original sharp image and a point spread function. The shape of the point spread function is determined by two core parameters: blur length and blur angle. The blur length is positively correlated with the platform's motion speed and the camera's exposure time, while the blur angle corresponds to the platform's motion direction.

[0033] This invention employs Wiener filtering as the core algorithm for image restoration. Wiener filtering is the optimal restoration algorithm in the sense of minimum mean square error, achieving a balance between image restoration and noise suppression. Its frequency domain expression is:

[0034] in, The spectrum of the original blurred image. The spectrum of the point spread function. For its conjugate, The noise power spectrum, The power spectrum of the original image. This is the spectrum of the restored image.

[0035] To address the problem of unknown noise characteristics in industrial settings, this invention estimates the noise power spectrum through the background region of an image: A uniform background region without alignment markers is selected from the image, and it is assumed that this region has a total noise power spectrum. 1 pixel, grayscale value First, calculate the average grayscale value of the background. Then, the variance of the grayscale values ​​is used as an estimate of the noise power, calculated using the following formula:

[0036] This method does not require prior knowledge of noise characteristics and can dynamically adjust noise parameters based on real-time acquired images to adapt to different field environments.

[0037] To achieve adaptive optimization of the point spread function parameters, this invention constructs a referenceless image quality scoring mechanism, the specific implementation of which is as follows: First, the Sobel operator is used to extract gradients from the candidate restored image. The Sobel operator contains two 3×3 convolution kernels in the horizontal and vertical directions, respectively:

[0038] The candidate restored image is convolved with two convolution kernels to obtain the horizontal and vertical gradient components of each pixel. The gradient magnitude of a single pixel is then synthesized using the following formula:

[0039] After traversing the entire graph, a complete gradient magnitude map is obtained.

[0040] Subsequently, the mean of the gradient magnitude plot was calculated. With variance The ratio of the two is used as the quality score for the no-reference image, i.e.:

[0041] The technical principle behind this scoring system is as follows: When the point spread function parameters match the actual motion parameters, the restored image edges are sharp and clear, and the gradient response is concentrated at the true edge locations, resulting in a high mean gradient magnitude and low dispersion (i.e., low variance). When the point spread function parameters mismatch, the restoration process amplifies noise and produces ringing artifacts, forming a large number of randomly distributed false edges. This leads to a decrease in the mean gradient magnitude and a significant increase in dispersion (i.e., increased variance). Therefore, a higher score indicates better edge quality and lower noise levels in the restored image, corresponding to a higher degree of point spread function matching.

[0042] The complete process of adaptive restoration is as follows: 1. Parameter space initialization: Estimate the reasonable range of blur length values ​​based on the platform's motion speed and camera exposure time; determine the range of blur angle values ​​based on the platform's motion direction; set the search step size based on the target search accuracy and generate all candidate parameter combinations.

[0043] 2. Candidate Image Generation: Traverse all candidate parameter combinations, construct a corresponding motion blur point spread function for each set of parameters, restore the original blurred image through Wiener filtering, and generate the corresponding candidate restored image.

[0044] 3. Quality score calculation: For each candidate restored image, calculate its no-reference image quality score according to the method described above.

[0045] 4. Optimal parameter selection: Compare the scores of all candidate parameters, select the parameter with the highest score as the optimal diffusion function parameter, and output the corresponding optimal restored image.

[0046] 5. Marker Coordinate Extraction: Position the markers in the optimally restored image and extract their sub-pixel coordinates. Optionally, conventional sub-pixel positioning methods in this field, such as gray-scale weighted centroid method or template matching method, can be used.

[0047] Experimental verification Reference Figure 5 The adaptive point spread function estimation algorithm described above was verified in a set of exemplary simulation tests. The test object was a standard motion-blurred marker image with a blur length of 20 pixels and a blur angle of 45°, and the centroid coordinates of the image features restored by theoretical parameters were used as the baseline ground truth. Figure 5 The left side shows the ideal, clear marked image; the middle side shows the original blurred image acquired during uniform motion; and the right side shows the image restored using this method.

[0048] This experiment uses the degree of centroid coordinate matching as the evaluation criterion for restoration accuracy. The Euclidean distance between the centroid of the image to be restored and the centroid of the reference ground truth image is calculated as the restoration error. The smaller the error value, the higher the restoration accuracy. The error calculation formula is as follows:

[0049] in, Here are the coordinates of the centroid of the image to be restored. The centroid coordinates are used as the reference true values.

[0050] The algorithm was run under three different initial parameter settings. The experimental results show that even with large initial parameter deviations (3 pixels for blur length and 3° for angle), the algorithm can still reliably converge to the optimal parameters. In the three sets of tests, the Euclidean distance between the feature centroid of the restored image and the theoretical true value does not exceed 0.001 pixels, demonstrating good anti-interference ability and stability.

[0051] Example 3 This embodiment details the definition of a three-level coordinate system and the principle of coordinate transformation, corresponding to... Figure 2 .

[0052] This invention defines three parallel coordinate systems, all of which use the left-hand rule, with the X-axis and Y-axis pointing in the same direction and only having translational relationships: 1. World Coordinate System: Fixed to the fixed reference platform, it is the absolute spatial reference of the entire system, with its origin located at the mechanical zero position of the motion platform; the coordinates of the loading platform output by the position feedback unit are the coordinates in the world coordinate system, denoted as [equation missing]. .

[0053] 2. Platform Coordinate System: Fixed to the platform, it moves synchronously with the platform, with its origin located at the reference point on the platform. The coordinates of the alignment mark in the platform coordinate system represent the position of the mark relative to the substrate, and are the final output of the alignment system, denoted as... .

[0054] 3. Camera coordinate system: Fixed to the image acquisition device, with the origin being the projection of the image acquisition device's field of view center onto the supporting surface; the coordinates of the marker pixels extracted from the image, after being converted using pre-calibrated pixel equivalents, yield the physical coordinates in the camera coordinate system, denoted as... .

[0055] Among them, the pixel equivalent is calibrated using a high-precision checkerboard calibration plate: the calibration plate is placed on a platform, multiple sets of calibration plate images at different positions are acquired, and the physical size of a single pixel is calculated by the ratio of the physical distance between the corner points of the checkerboard to the pixel distance, thus achieving high-precision calibration.

[0056] World coordinates of image acquisition device The standard marking method is used for calibration: the platform is moved to the mechanical zero position, and a high-precision standard mark with known platform coordinates is placed at the center of the field of view of the image acquisition device. The world coordinates of the platform at this time are read, which are the pre-calibrated world coordinates of the image acquisition device. After the system is installed, the calibration is stored in the calculation unit once and remains unchanged thereafter.

[0057] Based on vector geometric relationships, the world coordinates of the alignment marker can be derived through two independent paths: Derived from the path of the cargo platform: Marker world coordinates = Cargo platform world coordinates + Marker platform coordinates, that is...

[0058] Derived from the camera path: Marked world coordinates = Pre-calibrated world coordinates of image acquisition device + Marked camera coordinates, i.e.

[0059] Both equations describe the world coordinates of the same physical point, therefore the equations hold true. After simplification, we obtain the vector calculation formula for coordinate transformation:

[0060] In dynamic acquisition scenarios, the platform is in continuous motion. Therefore, the platform coordinates that are synchronously latched at the shutter start time must be used for calculation to ensure the accuracy of the conversion results. This is the core difference between dynamic alignment and static alignment.

[0061] Example 4 This embodiment illustrates the implementation chain of the direct trigger acquisition mode, corresponding to... Figure 4 .

[0062] The complete triggering and data acquisition chain is as follows: 1. Preset position calculation: The host computer imports the CAM drawing of the substrate and obtains the nominal coordinates of the platform for each alignment mark; combined with the pre-calibrated world coordinates of the image acquisition device, the preset position P1 of the loading platform when the alignment mark moves to the direct below the image acquisition device is calculated, and the preset positions of all marks are sent to the motion controller in the order of movement.

[0063] 2. Hardware position comparison: After the platform accelerates to the set speed, it maintains a constant speed. The position feedback unit transmits the real-time coordinates of the platform back to the motion controller at a fixed sampling frequency. The motion controller compares the current coordinates with the preset position P1 in real time through the internal hardware comparator.

[0064] 3. Trigger signal output and conversion: When the real-time coordinates of the platform match the preset position P1, the motion controller immediately outputs a trigger pulse through the hardware IO port and simultaneously latches the platform coordinate value at that moment; after the trigger pulse is level-matched and electrically isolated by the signal conversion module, it is transmitted to the camera and strobe illumination unit of the image acquisition device.

[0065] 4. Image Acquisition and Transmission: After receiving the trigger signal, the camera starts the global shutter to complete the acquisition of one frame of image, and transmits the image data back to the host computer through the data bus.

[0066] 5. Coordinate Calculation Output: The host computer performs adaptive motion blur restoration, marker coordinate extraction and coordinate transformation on the received image, and finally obtains the accurate physical coordinates of the alignment marker in the table coordinate system.

[0067] This mode uses hardware-level position comparison triggering, which has the advantages of lower latency and better consistency compared to software polling triggering, and can ensure the accuracy and repeatability of the trigger position under high-speed movement.

[0068] Example 5 This embodiment illustrates the implementation scheme of the delayed trigger acquisition mode, which is suitable for special application scenarios where it is inconvenient to use the internal triggering function of the motion controller.

[0069] The system is additionally equipped with a target surface device, a triggering device, and a delay control module.

[0070] The target device is fixed to the side of the platform via a detachable structure and moves synchronously with the platform. It is a dedicated trigger target independent of the substrate, and its triggering characteristics are not affected by the substrate material or surface condition, thus ensuring the stability and consistency of the trigger signal.

[0071] The triggering device is a non-contact position sensor, fixedly installed on the side of the fixed reference platform, with the sensing area facing the movement path of the target device. When the target device moves into the sensing area with the platform, the triggering device outputs an initial trigger signal. Optionally, the triggering device can be a laser displacement sensor, photoelectric switch, or other similar device.

[0072] The delay control module is connected to the triggering device and the image acquisition device respectively, and is used to adjust the initial trigger signal by delay before outputting it, so that the camera shutter start time corresponds precisely to the time when the marker arrives at the target acquisition center. This invention models the transmission link of the trigger signal as a two-stage delay: 1. Inherent hardware latency This refers to the total fixed delay from the output signal of the triggering device to the actual start of the camera shutter. It includes sensor response delay, circuit transmission delay, camera trigger response delay, etc., and is determined by the physical characteristics of the hardware and cannot be directly adjusted.

[0073] 2. Software-adjustable compensation delay This refers to the delay amount that the delay control module can adjust through software parameters to offset the effects of inherent hardware delays.

[0074] Two-level delay satisfies the following constraints:

[0075] in, The physical distance between the sensing position of the trigger device and the target acquisition center of the marker to be acquired in the first direction, wherein the target acquisition center is a fixed imaging position calibrated by the system; The constant velocity of the platform. The theoretical time required for the target surface to move from the trigger sensing position to the marker reaching the target acquisition center, minus the inherent hardware delay, is the additional software compensation delay that needs to be added.

[0076] In addition, the trigger timing control module synchronously controls the lighting sequence of the strobe illumination unit, so that the illumination unit lights up earlier than the camera shutter starts, ensuring that the light source brightness has reached a stable value when the shutter opens, and that the total lighting time covers the entire exposure cycle, thus avoiding uneven brightness in the image.

[0077] Example 6 This embodiment illustrates the signal transmission delay calibration method under delayed trigger acquisition mode, corresponding to... Figure 6 It is used to accurately measure the inherent delay of hardware and achieve closed-loop calibration of trigger timing.

[0078] This method utilizes the symmetrical error characteristics of positive and negative reciprocating motion: the inherent hardware delay causes the shutter start time to lag behind the theoretical time. During positive motion, the platform continues to move forward during the lag time, resulting in measured marker coordinates that are larger than the true value; during negative motion, the measured marker coordinates are smaller than the true value. The deviations of the two measurements are equal in magnitude and opposite in direction, therefore the inherent hardware delay can be calculated from the difference between the positive and negative coordinates.

[0079] The complete calibration steps are as follows: P1. Initial Configuration: Pre-acquire the physical distance d between the trigger device's sensing position and the target acquisition center in the first direction; place a high-precision standard reference mark on the platform, the platform coordinates of which are known and their accuracy is much higher than the alignment accuracy; set the uniform motion speed of the platform to v; set the inherent hardware delay in the initial state. Software-compensated latency .

[0080] P2. Forward Acquisition: Control the platform to move at a constant speed v along the forward direction. When the target surface passes the trigger device, trigger the camera to acquire the reference marker image; repeat the acquisition multiple times, performing adaptive motion blur restoration, marker coordinate extraction, and coordinate transformation on each acquired image to obtain the corresponding platform coordinate system coordinates; take the average value of the multiple measurement results and record it as the forward coordinates. .

[0081] P3. Negative Acquisition: Control the platform to move at a constant speed v along the negative direction. Use the exact same trigger parameters, acquisition parameters, and image processing procedures as the positive acquisition. Repeat the acquisition multiple times and calculate the platform coordinates of the reference mark. Take the average value of the multiple measurements and record it as the negative coordinate. .

[0082] P4. Delay Calculation: Calculate the inherent hardware delay based on the difference between positive and negative coordinates. The calculation formula is: [Formula omitted for brevity]. Write the delay control module and synchronously update the software compensation delay value. .

[0083] P5. Iterative verification: Repeat the forward acquisition, negative acquisition and delay calculation steps, and gradually correct the delay parameters until the difference between the positive coordinate and the negative coordinate is less than the preset alignment accuracy threshold, and the calibration is completed.

[0084] This calibration method does not require the introduction of additional precision measuring instruments. It can complete closed-loop calibration solely by relying on the device's own motion system and imaging system. It is easy to operate, has high calibration accuracy, and can effectively eliminate the delay inconsistency caused by individual hardware differences, thereby improving the alignment accuracy consistency of batch equipment.

[0085] Those skilled in the art should understand that the above embodiments are merely illustrative examples, and equivalent substitutions and conventional adjustments to technical details without departing from the core principles of the present invention are all within the scope of protection of the present invention.

Claims

1. A direct-write lithography alignment system, comprising a fixed reference stage, a gantry mechanism, a platform position control device, a loading platform, and an image acquisition device, wherein the gantry mechanism is fixedly mounted on the fixed reference stage, and the loading platform is driven by the platform position control device to move at a constant speed along a first direction of main motion, for carrying a substrate with alignment marks; the platform position control device is equipped with a position feedback unit for synchronously latching the world coordinate system coordinates of the loading platform at the time of image acquisition; the image acquisition device is fixed to the gantry mechanism for acquiring images of the alignment marks on the substrate; characterized in that, It also includes an image processing module and a coordinate transformation module: The image processing module is configured to perform adaptive uniform linear motion blur restoration processing: setting a parameter search space for the motion blur point spread function, traversing each group of candidate parameters within the parameter search space, constructing a corresponding motion blur point spread function for each group of candidate parameters, generating candidate restored images through Wiener filtering; extracting gradient magnitude maps for each candidate restored image using the Sobel operator, and constructing a no-reference image quality score based on the gradient magnitude. The parameters corresponding to the candidate restored image with the highest score are selected as the optimal diffusion function parameters. The optimal restored image is output and the sub-pixel coordinates of the alignment markers are extracted. The coordinate transformation module is configured to: based on the preset mapping relationship between the world coordinate system, the platform coordinate system and the camera coordinate system, and combined with the world coordinate system coordinates of the loading platform synchronously latched at the acquisition time and the pre-calibrated world coordinates of the image acquisition device, convert the sub-pixel level pixel coordinates of the alignment mark into physical coordinates under the platform coordinate system, wherein the platform coordinate system is a follow-up reference coordinate system fixed to the loading platform.

2. The direct-write lithography alignment system according to claim 1, characterized in that, The no-reference image quality score is constructed based on the ratio of the mean to the variance of the gradient magnitude, and the calculation formula is as follows: in, The mean gradient magnitude of the candidate restored image. denoted as the gradient magnitude variance of the candidate restored image.

3. The direct-write lithography alignment system according to claim 1, characterized in that, The noise power spectrum of the Wiener filter is estimated by the image background region: a uniform background region without alignment marks is selected in the image, the variance of the gray value of the region is calculated, and the variance is used as the estimated value of the noise power spectrum.

4. The direct-write lithography alignment system according to claim 1, characterized in that, The parameter search space is initialized as follows: the initial search range of the blur length is estimated based on the movement speed of the platform and the exposure time of the image acquisition device; the initial search range of the blur angle is determined based on the movement direction of the platform; and the corresponding search step size is set according to the parameter search accuracy requirements.

5. The direct-write lithography alignment system according to claim 1, characterized in that, The coordinate transformation module performs coordinate calculations based on vector mapping relationships, and the calculation formula is as follows: in, To align the physical coordinates of the markers in the table coordinate system, The pre-calibrated coordinates of the image acquisition device in the world coordinate system. The physical coordinates of the alignment mark in the camera coordinate system after pixel equivalent transformation. The coordinates of the platform in the world coordinate system are collected and locked synchronously at all times.

6. The direct-write lithography alignment system according to claim 1, characterized in that, The image acquisition device includes an imaging lens, a global shutter camera, and a strobe illumination unit; the system also includes a trigger timing control module, which is configured to control the strobe illumination unit to illuminate earlier than the camera's shutter start time, and the continuous illumination duration of the strobe illumination unit is not shorter than the camera's shutter duration.

7. The direct-write lithography alignment system according to claim 1, characterized in that, The system adopts a direct trigger acquisition mode, and the platform position control device includes a motion controller and a signal conversion module; The host computer calculates the preset platform position P1 corresponding to the alignment mark reaching directly below the image acquisition device based on the CAM drawing of the substrate, and sends the result to the motion controller. The motion controller receives the platform coordinates from the position feedback unit in real time. When the platform position matches the preset position P1, it outputs a hardware trigger pulse, which is then transmitted to the image acquisition device after being matched by the signal conversion module to trigger image acquisition.

8. The direct-write lithography alignment system according to claim 1, characterized in that, The system adopts a delayed trigger acquisition mode and also includes a target device, a triggering device, and a delay control module; The target device is fixedly installed on the side of the platform and moves synchronously with it, and is a dedicated trigger target independent of the substrate; The triggering device is fixedly installed on the side of the fixed reference platform and is used to output an initial trigger signal when the target device enters its sensing area. The delay control module is connected to the triggering device and the image acquisition device respectively, and is configured to model the transmission link of the trigger signal as a two-level delay: the first level is the inherent hardware delay. The first stage is determined by the physical characteristics of the circuit; the second stage is a software-adjustable compensation delay. ,satisfy Where d is the physical distance between the sensing position of the triggering device and the target acquisition center of the marker to be acquired in the first direction, and the target acquisition center is the fixed imaging position calibrated by the system. The constant velocity of the platform. The delay control module is used to process the initial trigger signal with a delay before transmitting it to the image acquisition device, so that the shutter start time of the image acquisition device corresponds to the time when the alignment mark arrives at the target acquisition center.

9. A direct-write photolithography alignment method, characterized in that, Includes the following steps: S1. Uniform speed acquisition: Control the platform to move at a uniform speed along the first direction of the main motion, trigger the image acquisition device to acquire the original blurred image of the alignment mark on the substrate, and synchronously lock the world coordinate system coordinates of the platform at the acquisition time; S2. Adaptive Restoration: Define the parameter search space of the motion blur point spread function, traverse each group of candidate parameters and construct the corresponding motion blur point spread function, generate candidate restored images through Wiener filtering; extract the gradient magnitude map for each candidate restored image using the Sobel operator, construct a no-reference image quality score based on the gradient magnitude, and select the image with the highest score as the optimal restored image. S3. Coordinate Extraction: Extract sub-pixel level pixel coordinates of the alignment markers from the optimal restored image, and convert them into physical coordinates in the camera coordinate system after pixel equivalent calibration; S4. Coordinate Transformation: Based on the preset mapping relationship between the world coordinate system, the platform coordinate system, and the camera coordinate system, and combined with the real-time coordinates of the platform at the time of acquisition, the physical coordinates of the alignment mark are transformed into physical coordinates under the platform coordinate system.

10. A method for calibrating the signal transmission delay of an alignment system, applied to the direct-write lithography alignment system of the delay-triggered acquisition mode as described in claim 8, characterized in that, Includes the following steps: P1. Initial Configuration: Obtain the physical distance d between the trigger device's sensing position and the target acquisition center of the marker to be acquired in the first direction; place a calibration reference marker on the platform and set the platform's uniform motion speed to v; initially set the inherent hardware delay. Software-compensated latency ; P2. Forward Acquisition: Control the platform to move at a constant speed in the forward direction, and acquire reference marker images multiple times. Perform adaptive motion blur restoration, marker coordinate extraction, and coordinate transformation on each acquired image to obtain the platform coordinates of the reference marker. Take the average of the multiple measurement results to obtain the forward coordinates. ; P3. Negative Acquisition: Control the platform to move at a constant speed along the negative direction, perform the same image processing and coordinate transformation process as the positive acquisition, and obtain the negative coordinates by taking the average of multiple measurement results. ; P4. Delay Calculation: According to the formula Calculate the inherent latency of hardware And update the software compensation delay value simultaneously. ; P5. Iterative verification: Repeat steps P2 to P4, gradually adjusting the delay parameters until the difference between the positive and negative coordinates is less than the preset alignment accuracy threshold, and the calibration is complete.