TW505956B - Method and subsystem for determining a sequence in which microstructures are to be processed at a laser-processing site - Google Patents

Abstract

Method and subsystem are provided for determining a sequence in which microstructures are to be processed at a laser-processing site by taking into account microstructures located near travel limits of a motor-driven stage. The method includes receiving reference data which represent locations of microstructures to be processed at the site and coalescing adjacent groups of microstructures into clusters of microstructures including edge clusters which contain the microstructures located near the travel limits of the motor-driven stage which moves the microstructures relative to a laser beam at the site. The method also includes dividing a cluster fragment from each edge cluster. The cluster fragments contain the microstructures located near the travel limits. The method then includes sorting the clusters and cluster fragments to obtain data which represent a substantially optimum sequence in which the microstructures are to be processed to increase throughput at the site.

Description

Five Λ invention description (A7 B7) The present invention is a method and a subsystem for determining the order in which microstructures are to be processed at a laser processing point. Technical background Memory repair is a processing program that is used in manufacturing Memory body circuits (DRAM or SRAM) to improve manufacturing yield. Memory crystals are manufactured using additional memory cell rows and rows. When testing memory chips (although still in wafer form), any defects found Will be recorded in the database. Wafers with defective wafers can be repaired using pulsed laser links. The system generally uses wafer processing equipment that transports semiconductor wafers to laser 1 laser processing machines, and Get information about the type of database, which indicates where the link should be cut off and carry out the necessary link dissolution of each wafer. Successive generations of DRAM use delicate element geometries to pack more memory in a smaller size The manufacturing impact of this smaller component is assigned to "1 ^ 彳 excess amount of link geometry. When the component is smaller, the link is smaller and smaller 1 The gap (link-to-link interval) is also shortened. The smaller link geometry requires a laser point size larger than '1' in order to smoothly remove the selected link without affecting the initial phase: Fortunately, it hardly (if possible) damages the output. The links are arranged in small groups in the common gap. The general group can be from two ()-50 to about 0. 8 " Μ wide '6 y Μ long And it is composed of 3 with M gap separation links. Groups can be configured parallel to the gap vector of the X or y axis. Often, multiple groups will be found to be collinear, and the end of a group will be cut Then there are only 2 or 3 "gap" between the beginning of the group. The others are 505956

A Ministry of Economy ¾ Printed by the Consumer Property Cooperative of the Hui Property Bureau A7 B7 Invention Description (/) When the line group is separated by several hundred microns. A typical DRAM chip can have hundreds of link groups that extend across the chip. The above-mentioned utility patent application discloses a high-performance x-y stage for placing a group of semiconductor wafers under a stationary laser beam. The x-y stage is required to achieve the many high-speed movements required to process one area of the wafer. The χ-y stage is usually limited by position, velocity, acceleration, and voltage conditions. The position limit is the limit of the stage travel. The rate limitation may be due to the limits placed on the sensing elements used in the stage and / or the limits imposed by the type of bearing used. Acceleration limits are usually current limits that can be used to drive the stage due to considerations of the motor, amplifier, or power supply. Voltage limiting usually occurs less frequently. Movements that require excessive voltage must be avoided to prevent amplifier saturation. Voltage limits are typically encountered during brief high-acceleration movements. This stage can perform movement using a considerable number of accelerations (such as 3G) and can stop quickly at the end of the movement. The stage must also be operated with very high accuracy (nano-level accuracy). The combination of very high performance and the need for extreme accuracy creates additional constraints on the stage. The stage is susceptible to the power level consumed by the motor coil. If the power consumed causes the coil temperature to rise above 2 ° C, the accuracy of the stage will be compromised. At the same time, if the spectral content of the force applied to the stage contains energy beyond the level that is higher than the frequency of the first mechanical mode of the system, it will cause resonance of the mechanical elements of the stage. These resonances will impair accuracy or require additional security at the end of the move before the necessary level of accuracy is reached. 5 Paper sizes are in accordance with Chinese National Standard (CNS) A4 (210X29 < 7 mm)-Approved ------, 玎 ------ 0 (Please read the notes on the back before filling out this page) 505956 A7 B7 V. Description of the invention (order) 汜 Time. Summary of the Invention An object of the present invention is to provide an improved method and subsystem for determining the order in which microstructures are processed at a laser processing point. To accomplish the above and other objects of the present invention, a method for determining the order in which microstructures are processed at a laser processing point is provided. The method includes receiving reference data representing the location of the microstructures to be processed at that point. The method also includes uniting groups of adjacent microstructures into aggregates that include edge-gathered microstructures, which include being placed on A microstructure close to the travel limit of the motor-driven stage. The motor-driven stage moves the microstructure relative to the laser beam at this point. The method also includes clustering and segmenting aggregate fragments from each edge, where the aggregate fragments include microstructures that are placed near the boundaries of the ir-range. The method further includes classifying the aggregates and aggregate fragments to obtain data representing a roughly optimal sequence, wherein the microstructures are processed with the sequence to increase the yield at that point. The classification step is based on the energy consumed in at least one set of motors: at least one set of coils in response to the motor command. Each clustering and clustering fragment may have a plurality of possible processing directions, and the classification step may include a step of determining an approximately optimal direction for processing microstructures. The classification step may include the step of selecting the approximate U which is initially processed at that point as an aggregate or aggregate fragment, and then determining the majority of possible sequences for processing the remaining aggregate 崚 aggregate fragments, and selecting approximately Best sequence. 6 This paper size applies to China National Standard (CNS) A4 (210X297 mm) (Please read the notes on the back before filling out this page) ¥ Order printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs ☆ Employees of Economic Ben Intellectual Property Bureau Printed by the Consumer Cooperative 505956 A7 __B7 V. Description of the Invention (ψ) The microstructure can be placed on a crystal wafer. Further, in order to accomplish the above and other objects of the present invention, a subsystem K is provided for determining the order in which microstructures are processed at the laser processing point. The subsystem includes a reference device that receives the location of the microstructure being processed at that point. This subsystem also includes a device that unites groups of adjacent microstructures into aggregates containing edge-gathered microstructures. The edge-gathering includes microconstructions that are placed close to the stroke limit of the motor-driven stage. The motor drives The stage moves the microstructure relative to the laser beam at this point. The subsystem further includes means for clustering and segmenting an aggregated fragment from each edge, wherein the aggregated fragment includes a microstructure placed near a stroke limit. Further, the subsystem includes means for classifying aggregates and aggregate fragments to obtain information representing a roughly optimal sequence, where the microstructures will be processed with the sequence to increase the yield at that point. The classification device may be classified based on the energy consumed in at least one set of coils of at least one set of motors in response to a motor command. Each clustering and clustering fragment may have a plurality of possible processing directions, and the classification device in t includes the classification device that determines the approximate optimal direction for processing the microstructures. The classification device may include: selecting approximately U which is initially processed at this point. The aggregate or aggregate fragment determines the device that processes the majority of possible sequences of the remaining aggregate and aggregate fragments and selects the approximately optimal sequence from the plurality of possible sequences. The microstructure may be a chip wire and the wire may be a gold rhyme line ° ____ 7 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 × 297 mm) 1¾ clothing ------, 玎- ---- Wire-(Please read the precautions on the back before filling this page) 505956 A7 ___ B7 V. Description of the invention (ζ) The chip can be a semiconductor memory device and the wire is ablated at this point to repair the device Of defective memory cells. The microstructure may be part of a semiconductor element and wherein the semiconductor element may be a microelectromechanical element. This semiconductor element may be a silicon semiconductor element. The semiconductor element may further be a semiconductor memory. The microstructure may be part of a microelectronic element. The microstructures in each group may have approximately common gaps. The stage may be a χ-y stage and the classification device therein may be classified according to the consumed energy of the plurality of coils of the plurality of motors in response to the motor command. The above objects, other objects, features, and advantages of the present invention can be easily and clearly understood from the following detailed description of the best modes for carrying out the present invention and with reference to the drawings. Description of the figure: Figure 1 is a detailed exploded block diagram of a memory repair system showing the main subsystems according to the present invention; Figure 2 is a system structure diagram with a trajectory data flow; employee consumption of the Intellectual Property Bureau of the Ministry of Economy Figure 3 printed by a cooperative is a set of example diagrams showing a wafer processing location containing a number of wafers, wafer locations, link groups, and reference areas positioned to form a relevant area of a reference surface; Figure 4 is An exploded view of a row of wafer links to be processed within the wafer position: Figure 5 is an exploded view of various moving segment types; ___8 This paper size applies to the Chinese National Standard (CNS) A4 specification (210X297 mm) 505956 Printed by the Ministry of Economic Affairs, Intellectual Property Bureau Employee Cooperatives A7 B7 V. Description of Invention (G) Figure 6 shows the amplitude with respect to time and the pulse / stroke curve of the moving segment; Figure 7 is relative to time The amplitude and the graph of the pulse / pulse curve showing the moving segment; and Figure 8 shows the amplitude with respect to time and the moving segment. Stop / skip / beginning of the curve pattern. Detailed description of the preferred embodiment A laser processing system, generally indicated as Π0, is shown in Figure 1. The wafer 4 is placed within the laser processing system 110 and database information from the user interface 11 is provided to confirm the link on the wafer 4 (Figure 3, which is ablated to repair the defective memory cell). 33). The database information is used by the trajectory planner 12 and the DSP-based controllers 15 and 16 and combined with the mobile stage 6 and 7 to calibrate to define the mobile segment of the trajectory generator, which uses laser, The operations of the focusing optics and the χ-y stage 6 and 7 are performed and coordinated to ablate the links, as shown in the utility application above. These operations include controlling the x-y movement with better high-speed accuracy stages 6 and 7 and simultaneously placing optical elements when the laser generates pulses so that the middle narrow portion of the processing laser beam is co-located with the 33 coordinate of the link. All memory repair systems include dynamic mechanisms to provide relative movement between the wafer surface and the plane of the focal point. In some cases, this may involve utilizing wafer Z movement to control the wafer height of the wafer 4 relative to a fixed height optical channel. In addition, this action can use the "step-by-step" method to move the focusing lens to focus on the wafer from the scene. 9 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) ------- -— ^! ---- 1T ------ 0 (Please read the notes on the back before filling out this page) 505956 A 7 __B7 V. Description of the invention (7) The depth position derived from the material is in the same position. With this system, the entire height of the wafer remains fixed and the lens height of the final objective lens is controlled by the linear servo mechanism and the controllers 14 and 17 of the optical box 18. The positioning of the better configured lens or optical element using the precision placement system for z-axis movement provides a Z-axis resolution of approximately 0.1 / zm or more subtle and has approximately 3dB at 150Hz " small signal " bandwidth. The “trajectory planner” 1 2 is adopted a controller mainly based on the mobile system 6, 7, 17 and the related DSP 16 to design the channel 4 of the wafer 4 and the position 5 of the middle narrow portion of the light beam. The trajectory planner combines information from the user interface 1 1 and the alignment system (the latter is typically installed on an accuracy level stage, such as a wafer level stage). This information is used to define information relative to the coordinate system. Target's laser processing point. This information is derived from the database, generating a set of "link diagrams", selections, and other data related to memory repair operations. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economics. Technical personnel familiar with movement control and estimation will understand the estimated demand for the precise three-dimensional space placement in high-speed memory repair processing systems. With a maximum range of about 300 mm or more, a few tenths of a micrometer is equivalent to the entire area of the latest wafer. Processing above 50 mm / sec or " cutting rate " is advantageous. Meanwhile, U.S. Patent No. 6,144,118, incorporated herein by reference, describes a preferred wafer placement system in detail. Referring next to Figure 3, a system with link fusing will generally need to process (ie, laser fuse) a subset 33 of a large number of links 34 on wafer 4. Information defining the chip link to be processed is provided to the control program. The program will set __ 10 this paper size in accordance with the Chinese National Standard (CNS) A4 specifications (210X297 mm) 505956 A7 B7 V. Description of the invention (g) Define a set of reference positions around the wafer 35 to be processed 32 'Is a set of chip positions. The location will generally contain a sufficient number of points to accurately define the trajectory of the wafer and lens system according to the command generated by the mobile system control, as described in the practical application mentioned above. The pulsed-wave laser-treated beam is generally provided by a Q-switched YAG laser, which has a predetermined pulse width, repetition rate, and wavelength, as disclosed in US Patent No. 5,998,759. A set of control signals 20 are supplied to the laser to generate a set of pulse waves that cooperate with the continuous placement of the wafer and lens. Those familiar with this technology will recognize that laser parity and movement will likely be compounded by transient or cumulative position errors. In the preferred embodiment, a set of adjustable laser delays of the previously "arranged" laser pulses is included to compensate for such positional errors. The time resolution for this correction is preferably 25 parts per billion or Less. Preferably, the complete error correction is exactly defined by a set of " trajectory vectors " which converts the total position error into a delay. This trajectory vector can be contained in a transformation matrix which is operatively connected Go to the controller to dynamically connect the coordinate system. Then referring to Figure 2, the trajectory generation is a procedure that specifies, represents, and implements a continuous movement channel. For the system described here, the stages in motion and The laser is fired so that each link will be fired at the center. The trajectory generation structure is shown in Figure 2. The beam movement is calculated by calculating the segments from the application data; the set points from the segments; and the servo commands from the set points. The most important functional items in this institution are planner, generator, and 11 paper standards that apply Chinese National Standard (CNS) Α4 specifications (210 297 mm) ._ batch of clothes - (Please read the Notes on the back to fill out this page)

Line 1T Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 505956 A7 _B7_ V. Description of the Invention (Clever) Inserter. The planner uses the analysis of data in specific areas, such as link maps and patching data to form channels for the stage. The output of this module is a set of trajectories; a list of moving clips, each of which is selectively matched with the control clip. When the laser beam and the three-dimensional coordinates of the link to be processed are determined, a set of motion control programs uses the trajectory planner or generator 12 of Fig. 1 to efficiently process the target structure. Referring next to Figures 4 and 5, in the preferred system, the acceleration and velocity curves are related to the following " moving segment " pattern: 1. PVT (position / rate / time) segment 110. It is used to accelerate to the desired position and rate. The time required to travel through this segment is optional; if not specified, the minimum time is calculated. 2. CVD (fixed-rate / distance) segment 111. This type has only a single scalar specification: the channel length of the fragment. The beam moves at a fixed rate over a specified distance. This rate is specified by the end point of the previous segment. Program control is usually performed during the CVD segment. 3. CVT (fixed-rate / time) segment. This is the same as a CVD segment, but the duration of the segment is specified rather than its length. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 4. Stop clip Π3. There is no stipulation in this segment that it should be as fast as possible. "Ablation" means firing a laser pulse to cut link 114. Furthermore, "stop " segment termination movement, preferably as fast as possible. Program control and link blowout are usually related to fixed rate segments. In a better system, the acceleration and rate curves are based on the DSP-based 12 books. Paper size applies Chinese national standard (CNS> A4 specification (210X297mm) Printed by A7, B7, Consumer Cooperative of Intellectual Property Bureau, Ministry of Economic Affairs, i. Description of invention) Servo controller 16 is combined to generate wafer xy movement. Along the optical axis The lens transition is coordinated with the X, y movement, so that when the laser generates a pulse, the narrow part of the beam will be placed at the wafer target position. The Z coordinate of the narrow part of the beam can be between any two sets of structures on the wafer. Dynamically adjusted, including adjacent structures arranged in the same row (along X or y direction) of a single chip. For example, the z-axis incremental resolution (minimum height difference) of the link fusing, preferably It is about and has a limit of about 0.05 # m. The problems involved in channel planning within the system are as follows: given the initial and final position and velocity (vector) regulations, find The "best" trajectory that meets the conditions of the termination point and the restrictions on the position packet, the maximum rate, the maximum acceleration, and the maximum voltage. This optimal trajectory minimizes the trajectory " cost ". For the given The "cost" of the trajectory refers to the weighted sum of the trajectory time and the energy consumed in the motor coil. The weighting factor for time is always one unit (1.0). The weighting coefficient of energy is a variable. Many trajectories can meet the required termination point conditions and meet system constraints. However, the costs associated with multiple solutions can vary widely. The algorithm of the present invention finds the best solution (the lowest " cost ") and uses the trajectory to perform the movement. Consider a simple set of examples. The trajectory that is often encountered when repairing billions of bodies is called " combined ". Laser processing needs to follow two adjacent sets of short-line segments. The segments are collinear and the segments are processed at the same fixed rate. A set of gaps exists between collinear segments. A set of "CV" segments can be used to cross the gap (that is, a fixed rate at the same rate as the previous segment is used. 13 This paper size applies Chinese National Standard (CNS) A4 Specifications (210X 297mm) -------- | batch clothes ------ '玎 ------- 0 (Please read the precautions on the back before filling this page) 505956 A7 ____B7_ V. Description of the invention (L () rate segment). In addition, the gap can be crossed in a little less time using a short acceleration / deceleration curve. The cost of the CV solution is only the time required to cross the gap (distance / rate). Because there is no energy consumption in the CV segment. However, the acceleration / deceleration curve will take less time but greater energy consumption. In this case, the best choice depends on the relative weighting of energy consumption relative to the time of the cost function. Moving curve A line segment contains one or more intervals. A set of intervals is defined as a set of trajectory subsets, where acceleration is limited to a predetermined function of two sets of variables: APK (peak-to-peak acceleration amplitude) and Tint (the duration of the interval) ). The type of acceleration curve is trochoidal or harmonic. Because the type is predetermined, only the amplitude and duration are needed to fully specify the movement during the trajectory interval. The trajectory is a sequence of continuous (time, position , Velocity, acceleration, and shaking) intervals, each interval has a variable duration and acceleration amplitude (may be zero). It should be noted that a set of CV (fixed rate) trajectory segments is a set of single intervals with APK = 0 The system printed by Figure 1 of the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs reached a set of continuous trajectories to form a set of single moving curves. A single moving curve can last for many seconds and contain hundreds of trajectory segments. Multiple intervals are generally required In order to meet the termination point of the trajectory segment, the interval can generally be described as corresponding to "quotation" or "pulse". " C " or "P ". A set of C-fragments has the feature of APK = 0. A set of P-fragments has a non-zero APK. In the combined example given earlier, the navigation solution contains a single set of" C " ”____ 14 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 505956 Ministry of Economy ¾ Ahui B7 printed by the Consumers’ Cooperative of Consumer Property Co., Ltd. A7 B7 5. Inventory (^) interval, and the acceleration / deceleration solution is A set of "P · P" solutions. The other fragment combinations are required to handle the various trajectory requirements required by the system in Figure 1: (C, CP, PC, PP, PCP, PPPP, PPCPP). Not all solution types are applicable to the given trajectory segment. For example, trajectory segments containing different start and final velocities (Vi < > Vf) cannot be satisfied in the c interval. Solution CVT or CVD segment solution is simple (specified time or zero acceleration of distance). Because time was specified and energy consumption was zero, no optimization was performed. CV fragments will not be discussed further. The solution to the PVT trajectory segment is more complicated. The solution consists of the following steps = 1. The decomposition vector specification becomes the X and y axis specifications. 2. Find the minimum cost solution for each axis separately. 3. Get the slowest axis time and find the "quick" axis solution, which depends on the same minimum "slow" cost of elapsed time. The minimum cost solution for a single axis consists of the following steps: 1. Find the best solution for each type of solution (ie, C, P, PC, CP, etc.). 2. Choose the best solution between multiple forms (if multiple solutions exist). ____15 This paper size is applicable to Chinese National Standard (CNS) A4 (210X297 mm) -------- | Doing clothing ------ 1T ------ ^ (Please read the note on the back first Please fill in this page again for matters) 505956 A7 B7 V. Description of the invention (θ) The relative cost of energy printed by the employee's cooperative in the Intellectual Property Bureau of the Ministry of Economic Affairs is controlled by the variable κρ. Kρ is used in the cost formula as follows: Cost = T + Ε / Κρ It should be noted that Kρ has the dimension of power. This provides an intuitive guideline for choosing Kp. If Kρ is set to be large, the cost of energy consumed is small. The resulting solution will be minimal. Conversely, setting Kρ to be small will make the energy "expensive", and therefore the solution will often be to minimize power at the cost of time. Because Kρ is a variable, it can be set dynamically. The system of Figure 1 uses this capability to further control power consumption (and therefore improve accuracy). During production operations, a sequence of track segments must be provided to the channel planning algorithm. The channel planning algorithm determines the energy that will be consumed during each moving segment. Using a simple first-order model of the thermal response of the coil, the channel planning algorithm will anticipate the temperature rise caused by the energy consumption history of the trajectory. In a trajectory segment containing many short acceleration segments, the estimated temperature may rise more than required. The channel planning algorithm will then reduce Kρ 値 to bias the mobile optimization solution towards lower power regions. Conversely, when the estimated temperature rise is very small, the channel planning algorithm will increase κρ. The simulation revealing time based on the real recording data of the billionth body is relatively ___ 16 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 505956 Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 _B7_ Description of the Invention (^ 4) The dynamic adjustment of the relative cost of the bear volume reduces the change in the motor temperature. The accuracy of the system in Figure 1 is quite sensitive to the rate of change of temperature (but not particularly sensitive to any particular temperature). The main source of stage temperature change is the change in motor power consumption caused by trajectory. Dynamically adjusting Kp as a function of estimated temperature significantly reduces variation and produces higher accuracy. When performing general memory repair trajectory, the power consumption level in the motor will be strongly affected by κρ. However, the overall output of the system is a weak function of Kρ. Therefore, the components of the stage error caused by the motor consumption can be effectively controlled with a small influence on the overall output. Since the χ-y stage is significantly faster than competitive provision, this form of precise control can be achieved without much impact on market acceptance. Solution Advantages The optimized multi-pulse solution provides the following advantages: Continuous and smooth movement The channel planning algorithm allows continuous movement on the χ-y plane with minimal stops. The movement curve is not limited to any position interruption and the derivative (derivative of acceleration) caused by all shaking. This therefore limits the spectral content of the forces applied to the stage, thereby reducing the mechanical resonance of the stage and supporting structure and reducing power consumption. The trajectory is optimized (based on the relative weight of power consumption, essential). -Consumption minimization and utilization control κρ as a function of the expected motor temperature, a type of temperature reduction ___17 This paper size applies the Chinese National Standard (CNS) Α4 specification (210X297 mm) --------- | batch ------ 1Τ ------- ^ (Please read the notes on the back before filling out this page) 505956 A7 B7 V. Description of the invention (6) A simple device that changes and thus increases accuracy is realized. Link Optimization As mentioned earlier, the specified link order to be cut is rarely the best order to guide the link cut procedure. The algorithm described here rearranges the links and thus achieves reductions: 1) execution time; 2) motor power consumption; and 3) the spectrum inclusion of the power applied to the system. Reduced execution time leads to higher productivity of the system. Reduce motor power consumption and improve system accuracy. By avoiding the excitation of structural resonances that may lead to incorrect placement, the reduced power spectrum contains improved accuracy and yield. The algorithm is modeled based on the conventional "travel salesman" problem. However, in this case, it is of interest to minimize the cost of inspecting each link group rather than just inspecting each group. Minimize time. The classification algorithm also conveniently handles some special situations and conditions that limited the performance of the previous memory repair system. The system printed by Figure 1 of the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs is generally a single "processing point" " Handle multiple wafers. The system's range size allows up to six or eight 64MDRAM wafers to be processed simultaneously. This process is called multi-wafer alignment (MDA). The use of MDA provides significant yield improvements by reducing the number of alignment operations required to process a wafer from one wafer at a time to one processing point at a time. Alignment operations may take approximately the same amount of time as a single-chip link cut. MDA introduces another opportunity to take advantage of link rearrangement to improve yield. Link groups of adjacent chips may be collinear. Processing collinear groups without intervening PVT moving segments (ie, acceleration / deceleration segments) is usually the best solution. The link classification algorithm considers all groups within the processing point. 18 This paper size applies the Chinese National Standard (CNS) A4 specification (210X 29 < 7 mm). 505956 A7 B7 Printed by employees of the Intellectual Property Bureau of the Ministry of Economic Affairs and Consumer Cooperatives 5. Description of Invention (Vb) in order to find the best sequence. Algorithm The algorithm is as follows: 1. Gather the link groups that need to be processed within all processing points. This often includes reading a "repair file" with repair data, as shown in Figure 2, and may convert the customer data format into an internal representation of the link that will be cut. 2. Join adjacent groups. When reading When fetching the list of link groups to be cut, "join" adjacent groups to become " aggregate ". Adjacent groups are combined if they are collinear. Groups can be processed at the same rate and the gap between the end point of a group and the start point of a subsequent group is larger than the minimum scale and smaller than the maximum scale. 3. Identify the " edge " group. If the end point of an aggregation is too close to the actual limit of the stage packet, the stage cannot have enough acceleration to process the group at full speed and stop before the "strike" limit of the stage travel. The aggregation uses all Links adjacent to the "edge" are scheduled to the new cluster and the remainder remains in the original cluster and cut into two clusters. The new lower rate is assigned to the edge cluster so that the stage can have enough space The new cutting rate is reached within the distance between the last link of the new group and the actual limit of the stage travel. 4. Classify the aggregation. The input to the classification is an aggregation with (arbitrary) direction assigned to each aggregation. (Arbitrarily) a sorted list. (Direction refers to the direction in which aggregated links will be traversed, forward or backward.) The output of the classification process is a set of identical chains assigned to each group in a new direction. The new paper size of Lu Group is applicable to the Chinese National Standard (CNS) Α4 specification U10X297 mm. Approved clothing_1ΤI I-^ (Please read the precautions on the back before filling this page) 505956 A7 B7 V. Description of the invention (f |) sequence. The reordered list is optimized. 5. Process the group. The sorted list of the group is entered into the channel planning algorithm. The group is crossed by the aforementioned use The order of a single continuous-moving trajectory for the entire set of links is processed. ^ Edge aggregation processing occurs as a group close to the stroke limit is a challenge for any memory patching system. The distance required to stop is a function of the square of the rate And the reciprocal of acceleration. In the system of Figure 1, a compromise is used in the design of the xy stage to limit the number of " overtravel " available between the end of the user's travel range and the actual limit of the travel. The limit is Select as part of the optimization of the motor design of the χ-y stage. If the stage travels at the edge of the user packet at the maximum cutting rate (150 mm / s), then the edge of the user packet and the stroke The limit distance between practical limits (about 200 μM) is not enough to stop the stage. One set of solutions is to reduce the maximum cutting rate to ensure that the stage can always stop. This It is not necessary because there is no need to increase the processing time of the internal link group. Another solution is to use a lower rate for the entire aggregation. This also unnecessarily increases the processing time. The Intellectual Property Bureau, Ministry of Economic Affairs, Consumer Consumption Cooperative The solution adopted here is to make the minimum number of links accept the lower rate limit. The original aggregation occurs on the first link that did not accept the reduced rate limit, regardless of the original group boundary. Location. Furthermore, the rate assigned to the " edge " link is selected as an integer component (1 / η) of the nominal rate. A set of examples will help explain the advantages of this mechanism. Assume the original group The group has a link gap of 4 // M. Furthermore, assuming 20 paper standards are applicable to the Chinese National Standard (CNS) A4 specification (210 × 297 mm) 505956, printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the invention The maximum Q-rate allowed for a particular laser in use is 20,000 pulses per second. The maximum rate during processing will be 80 mm / s. For a simple harmonic acceleration curve that accepts a maximum rate of 30 meters per second squared, the stage will need -200 // M to stop (or reach the cutting rate from stopping). Assuming that some links in the group are sufficiently close to the edge of the user packet, the link will require a lower cut rate. The actual rate selected for the original aggregated fragment will be chosen based on the minimum 値 η, so that the new rate will be equal to the original rate divided by n. The η 値 is obtained interactively using 2 starting from 2 and increasing until the generation rate is small enough to handle the last link (closest to the edge) in the aggregated fragment and the stage maintains a sufficient stopping distance. Processing this aggregated fragment only at V / n will cause the laser emission rate to be reduced to Q / n (in this example, Q = 20,000). This may result in poor changes in link cut quality for links in segments that accept different laser Q rates. This change is eliminated in the algorithm that introduces the concept of "ablation-per-gap" (bpp). In general, bpp = 1 and the laser is fired once at each stage of the mobile gap. When the fragment is shot at V / n During processing, bpp is set to η. The system in Figure 1 recognizes this parameter and emits a laser at a fixed rate equal to n * (V / n) / p = V / p = Q. The system " imaginary " has η -1 groups of "ghost" links are also distributed between each group of natural links. This forces the laser Q-rate to be the same as the Q-rate used on all other links and ensures that all links are consistent The algorithm further achieves an optimization that affects links close to the edge of the user packet. The link classification algorithm includes a set of parameters (T_Settle), which specifies the stage before the first link in the group. Should continue to cut more than 21 This paper size applies Chinese National Standard (CNS) A4 specifications (210X297 mm) ------- 1-clothing -------- 1T ------ 0 (Please read the notes on the back before filling out this page) 505956 A7 B7 V. The description of the invention (L ° |) is long. Generally speaking, PVT fragments (Ie, the stage where the acceleration / deceleration of the stage occurs) will be before the CV segment (ie, the stage where the stage rate is fixed and laser cutting occurs). If T_settle is set to zero, the first stage will be cut A link will occur exactly at the end of the PVT segment and at the beginning of the CV segment. In some cases, the stage cannot fully settle to the required accuracy level without any settling time. This may Causes the laser misplacement of the first few links in the group. The T_settle parameter can be used to pre-expand the nominal CV segment to a distance equal to V * T_Settle. This gives the stage some settling time and a small amount of production cost Improve system accuracy. This parameter is generally a few milliseconds. A non-zero T_settle 値 affects the processing of the "edge" link. Extends the CV segment to accommodate Tjettle only applied to the beginning of the segment. The algorithm can optionally cut a group The preferred direction of the link group. If a group of edge groups is processed starting from the edge and moving towards the inside of the packet, the distance (D_settle) required to accommodate T_settle will be increased. Steps reduce the distance required to accelerate from 0 to V / n. Because the first link position is fixed, this means that a lower set of rates will need to be applied to the group if it is " inward " processed Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. If the group is processed "outward" (the stage stops just before the actual limit of the trip), the D_settle distance will extend the link towards the inside of the stage packet, without Forcing to further reduce the cutting rate of the fragment. Obviously, processing edge-gathering fragments in an "outward" direction seems to maximize the cutting rate. This is not always the case. The function of the classification algorithm is to find the best direction. Categorize and classify 22 paper standards to Chinese National Standards (CNS) A4 specifications (210X297 mm) 505956 A7 B7 Printed by Jade Consumer Cooperative, member of the Bureau of Intellectual Property of the Ministry of Economic Affairs • Classification of inventions (> o) from the beginning (Pi) Start. This position corresponds to the last position (usually near the corner of the position) reached at the end of the alignment operation. The starting rate (Vi) is assumed to be zero. The classification algorithm decides to deal with all " available "aggregated costs in two possible directions. The cost is calculated as follows: Cost = Tpvt + Epvt / Kp + Tcvt where:

Tpvt is the time from (Pi; Vi) to the start of the group;

Epvt is the energy consumption in the motor coil when moving;

Kp is the weighting factor (power unit) applied to the motor energy; and

Tcvt is the time required to complete the C V segment. For each aggregation, the movement is assumed to start at (Pi; Vi). Each aggregation has two sets of cost figures-one for each direction. Costs in different directions can be significantly different. A parameter called N_level is used to control the classification algorithm. If NJevel is set to 1, the aggregation / direction combination that exhibits the lowest cost is selected as the first processing aggregation. It is removed from the " Available " aggregate list and placed at the beginning of the " program list ". The starting position (Pi; Vi) is updated to reflect the position reached at the end of processing the first aggregation and the (non-zero) rate. The classification algorithm recalculates the cost of all available aggregations / directions. It should be noted that due to the significantly different starting conditions, there is no correlation between the cost determined by the second stage and the cost obtained by the first stage. The impact of non-zero start and end rates on costs may not be significant. The procedure of selecting the lowest cost aggregation / direction combination from the available groups is performed until no aggregation is available. 23 This paper size applies Chinese National Standard (CNS) A4 specification (210 × 297 mm) I batch of clothing I-1T ^^ (Please read the precautions on the back before filling this page) 505956 A7 ___B7________ 5. Description of the Invention (yl) Parameters N_Uvel controls the search range for the best aggregation order. In the example given above, N_level is set to 1, and the best aggregation / direction is selected so that the processing result is not considered. Choosing the best aggregation is likely to result in a less efficient order of sequential aggregation. In addition to finding the best solution, if all the results of that choice can be detected, a truly best choice can be made. Then ’considering all the results, the best solution is chosen. For example, suppose at the beginning there are ten groups (10) of aggregates that need to be classified. The cost of the initial set of twenty (20) clusters / directions will be considered. We can select the best five groups from this list of twenty groups (for example). These five sets of lists are then examined further. For each of the five groups, eighteen groups of untreated (18 = 2 * 9) clusters will be considered. From this list, select the best five 絍 and consider each of the five groups for a group of sixteen groups (16 = 2 * 8). Then * preferably five groups are selected from this list and the remaining aggregations are considered. This process continues until there are no unclassified aggregates. Then calculate the total cost of each group of the first five groups. The sum of the costs is obtained using the sum of the costs of the best solution selected from all the basic aggregation options. After that, one set of aggregates is selected and all processing procedures can be repeated again in the remaining set. Obviously, for a large number of aggregates, this thorough search to find the best aggregate is computationally very expensive. Patch data with multiple wafer locations may range from ten clusters to hundreds of clusters. For locations with many clusters, a thorough search can be time consuming. Fortunately, the experience of truly patching data indicates that exhaustive searches have little benefit when searching at a 1-level depth (N_level = l). If the NJevel 24 paper size applies Chinese National Standard (CNS) A4 (210X 297 mm) (please read the precautions on the back before this page) Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs Consumer Cooperatives 505956 A7 B7 5. DESCRIPTION OF THE INVENTION (> y) is set to 3 (for example), the search will be performed at the 3-level. It should be noted that the continuous search level is reached only after the first level at the optimal M-aggregation (M = 5, as described above). Although this may lead to the next best chance of being selected, the additional costs associated with " inappropriate " first choice cannot be compensated by beneficial results (found at sequential search levels). There is a set of qualitative arguments that support experimental observations where the taxonomy does not benefit when M or N_level is a large number. For general repair data, the stage will spend between 80% and 90% of the total time firing lasers in the CV segment. Rescheduling did not affect CV processing time. Rescheduling optimizes the PVT time (in small amounts) and can have a significant impact on the energy directors consumed. However, once the available aggregated " appropriate " channels are found (N_level = l), setting N_level to 2 or 3 will only lead to a small percentage of further time savings. Performing classification to the maximum level results in minimal self-advancement when N —level = 3. Although NJevel's large numbers have only a small improvement, classification with N_level in the range of 1 to 3 has a significant impact on energy consumption and spectrum content. The classification rule looks for opportunities to save momentum (because if you save momentum, no energy is consumed). For example, when two sets of clusters of adjacent wafers can be processed without intervening PVT segments, stage momentum is saved. Adjacent link groups often appear. Classification will find a set of smooth moving curves that shift from the end point of the link group in one set to the start point of the link group in the other set. This smooth curve is generally transferred to three groups of stages than the traditional one: stop, jump, start curve is faster, consumes less energy and has lower spectrum content. 25 This paper size applies the Chinese National Standard (CNS) A4 specification (21〇X: 297mm) —IB— In mi ί I in ^ i nn tn m_i n (Please read the precautions on the back before filling this page) Order Printed by the Intellectual Property Bureau's Consumer Cooperatives of the Ministry of Online Economy 505956 A7 B7 V. Description of the Invention (> Yes) Finally, a set of examples will show why the direction of choosing to deal with edge link groups is not always obvious. Consider clustering at the end of the + X edge position. Suppose that the original aggregate must be divided into two groups to produce aggregate fragments. If the fragment is processed in the + X direction, it can be processed at a higher rate than if it was processed in the -X direction due to the impact of T_Settle. In cluster classification (assume N_level = 1), it is assumed that the original cluster is selected to be processed in the + X direction. In the following classification rule, the cost of processing the fragment is determined using the assumption + X processing and -X processing. Cost will take into account the time difference in the CV segment depending on the achievable speed as a function of direction. However, fragments cannot be processed at the same rate as previously gathered. The end point of the previous gather (the original gather, now cut) and the fragments often do not have enough distance between them to allow the stage to decelerate to + X fragment rate. The associated costs of attempting to process the fragment in the + X direction of the original aggregate and the + X traverse result in a rapid deceleration (from the original aggregate) to the -X rate, and then increase to a lower + X rate to cross the position . The cost of crossing the fragment in the -X direction is often less, because the stage can make a smoother transfer (that is, lower energy consumption) to stop near the end of the packet and then increase to a lower rate (-X direction) To reach the outermost link of the aggregate segment in an instant. Weighted energy savings (E / Kp) are often more significant than increased Tcv time. Benefits of the solution Aggregation classification shows the following benefits in comparison with previous implementations: 1. The algorithm considers links that span the entire location. This resulted in optimizations that could not be achieved when the classification was achieved by the wafer-to-wafer guidelines (savings of 26 paper sizes are applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm)) (Please read the precautions on the back before you go to ' (This page) Ordering line _ Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 505956: · Α7. Β7 V. The opportunity of the invention description (> ψ) was realized. 2 The algorithm explicitly considers groups near the edge of the packet and treats them in the most efficient way that only accepts the limits of the actuating mechanism. 3. The algorithm takes into account the start and end velocities and the distance between the end point of a group and the start point of a subsequent group. This makes the combination of nearby groups clear. Previous algorithms combined nearby groups based on preset distance criteria only. Choosing an appropriate distance threshold to combine them becomes a very important issue. In the algorithm, the only motivation for merging any nearby groups is to reduce the number of aggregates that need to be analyzed. This leads to the choice of shorter combined distances (i.e., groups that are not separated by more than X) than previous algorithms. The classification will be continuously placed " uncombined " aggregates, and may replace CV fragments with PVT fragments between groups. That is, if the initial " aggregation " stage has not merged the aggregates, they can still be combined by the classification step. 4. Conversely, "uncombined" aggregation may be easier to process outside the sequence. When the trajectory being viewed is due to the same data starting from a classification with four different start positions corresponding to the four sets of corners of the processing point This is obvious when it occurs. 5. The algorithm inherently reduces the spectrum content and energy consumption associated with the processing point. Because power consumption is included in the cost formula, optimizing the processing cost results in lower energy consumption. Less significant may be that the same optimization results in longer lasting pulses with less acceleration. The segmentation-cutting multiple pulse wave curve is overly restricted-the curve has more parameters than the boundary conditions. In many cases, you can use More than one set of pulses and strokes of 27 This paper size applies to the Chinese National Standard (CNS) Α4 specification (21〇 × 297 mm) ---------- ^ II (Please read the precautions on the back before (Fill in this page) Printed by the Intellectual Property Bureau Staff Consumer Cooperative of the Ministry of Economic Affairs 505956 A7 B7 V. Invention Description (yi) Combination to adjust to boundary conditions; for a single combination, 'maybe there is an unlimited number of Briefly refer to a specific interval combination, pulse waves are represented by p, and strokes are represented by C; for example, pulse wave / navigation / pulse wave segment will be called PCP. When relevant, we will The difference between being calculated (" M ") and time being specified (" T "). For example, there can be PCP-M and PCP-T solutions, but in fact, they are completely different. · Many different combinations of intervals, or structural patterns, will be made. If we do n’t have a general solution, some boundary condition settings will have an infinite number of solutions. Sometimes, more than one set of patterns will solve the boundary condition. In those cases In this case, the lowest cost solution will be selected. For each type, both -M and -T solutions will be made. The currently solved types are: * Pulse / stroke of Figure 6: This produces C, P, CP and PC segments. Line * Pulse / Pulse in Figure 7: This produces PP and PCP segments. Pulses have the same acceleration amplitude but opposite signs; pulse time is independent. If the first pulse is over Point rate super At the maximum speed, the highest speed voyage interval is used to form a PCP fragment. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs * Stop / Jump / Start of Figure 8: This is also called a Z-movement. A set of PPCPP fragments and the -T solution can simultaneously generate a set of PPCPCP fragments. If no other solution is appropriate * this is the final solution. It includes immediate stop, PCP point-to-point movement, and stop to "level-level point". , And then the last pulse to increase to the final position and velocity. This fragment type generally takes more time than other types and _____ 28 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 505956 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A 7 B7 V. Description of Invention (> b) More energy. However, observing position constraints can make it important for a small, preferably infrequent set of boundary conditions. The strategy for solving the given multi-axis pvt fragments is as follows: 1. Get the minimum time movement of each axis (ie, solve the axis_M problem) ^ Pick the largest from these times. Use the specified time (axis · T) solution to solve other axes. The following sequence is used to solve the problem of both the minimum time and the specified time: 1. Try to design a set of PC and PP fragments; keep a list of the best feasible solutions of each type. 2. If a feasible solution is available, select the group with the lowest cost. 3. If no solution is available so far, design a set of stop / jump / start movements. The feasibility check includes a number of checks, for example, the interval time is non-negative, and actual checks, for example, voltage, rate, and position constraints are not violated. The following chart is a summary of the derived formulas for cycloid and harmonic multipulse acceleration. ; ____ 29 This paper size applies to China National Standard (CNS) A4 (210X29 * 7mm) --------- Batch ----- ^ ------ ^ (Please read first Note on the back, please fill out this page again) 505956 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 __B7 V. Description of Invention (> 1)

Item Cycloid Harmonic Acceleracion n A (2-r 1 α (/) «cos ~ fj Velocity AT AT mv (〇« v # +-cos— V (f) = ν · ) 3 ^ ^ vf + i-silty Jerk ,, Ax nt y (/) =-jrc〇s— aA. 2ge X〇 * ~ sm ~ Single Pulse A 2 AT 厶 V s 丨 丨 丨 丨 丨 丨 丨 " 丨 丨 丨 丨 丨 丨 丨 3C τ AT: Δρ a TV, ♦ —7 " r # »Ty fm 丨 丨 丨 丨 丨 丨 丨 丨 丨 丨 'X αρ dp * Tv, ♦ --j- -ΛΤ % * Γν / -— Scan or Stop △ v2 άρ9Τ Point-to-Pomt Mocioa 2 AT2 厶 P * 1111111111111-1 11 丨 1111111 KAT2 Δρ «j-1111111 Pulse Energy y 〔f)-! (F) (please first (Please read the notes on the back page)

、 1T line 30_ This paper size applies to China National Standard (CNS) A4 specification (210X297 mm) 505956 A7 B7 V. Description of invention

Item

Cycloid

Harmooic

Mimmum-Cosi Pulse

Puise-Oniy Solution Δν A = one (2v / ^ Δ v) ΔΡ

Pulse / Cmisc, Time Specified ΙΓάν · Α · 4 (Δ /?-TvJ rAv2, Cruise! Pulse A: 6v: ——, Cruise I PulseTy, 4 (Γν7-Lp), Pulse / Cruise IV Δν5 —-, Pulse t Cruise Γν7-Δρ

Pulse / Cruise, Minimum Cost

KpKsLv, Cruise / Pulse A1

H ΚβΚΑ ^, Cruise / Pulse [K.KAv J-, Pulse t Cruise V v / Κ ¥ Κ, άι Pulse I Cruise

Pulse / Puise. Time Specified

Solve for A; T1 jfA V * — ^ ♦ (Γξν ^ ν7) -2Δ ^ Μ -—— a 0 rl

Solve for A: ♦ Printed by Γ (κ * v,) · 2Δ /?) W · d s o Economy 『Ministry of Intellectual Property, Bureau of Consumer Affairs, X Consumer Service Co., Ltd.

Pulsc / Pulse, Acceleration Specified

Solve for T:

Solve for T: ： Tl 2 (ν · * νΓ) ·-Factory 雠 丨 丨 丨 丨 丨 丨 丨 __ Λ / IV Λ • 2d / > j > 31 This paper size applies to Chinese National Standard (CNS ) Α4 size (21〇297mm)

505956 A7 B7 V. Description of the invention (/ |) Although the best mode for completing the present invention has been described in detail, those skilled in the art will be aware of various designs and embodiments of the present invention, such as the following application Defined by patent scope. Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs, Consumer Cooperatives 2 3 This paper size applies to the Chinese National Standard (CNS) A4 specification (2 丨 0X297 mm) 505956 A7 B7 V. Description of the invention (> c) Component reference table 4. .... Wafer 6 > 7 ... Mobile Stage 11 ... User Interface 12 " ... Track Planner 14, 17 ... Controller 15, 16 ... Controller 18 "... Optical box 32 " ... reference position 33 ... link subset 34 "... link 35 ... Japanese and Japanese film 110. · • ... laser processing system 110 ...... PVT (position / ¾ 111 ... &Quot; .CVD (Fixed 113 •••… Stop segment 114 •••… links -------- 丨 ^ 丨 ^ -----, 玎 ---------- 0 (Please read the notes on the back before filling out this page) „Printed by the Ministry of Economic Affairs and the Intellectual Property Bureau's Consumer Cooperatives 33 This paper size applies to China National Standard (CNS) A4 (210X297 mm)

Claims (1)

Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs, Consumer Cooperatives 505956 as Β8 C8 D8 VI. Application for Patent Scope 1. A method to determine the order in which microstructures are to be processed at the laser processing point. The method includes: References to the location of the microstructures to be processed; Unite adjacent groups of microstructures to form a cluster of microstructures that contain edge clusters that contain microchips that are placed close to the travel limits of the motor-driven stage The structure, the motor-driven stage moves the microstructure relative to the laser beam at the processing point; gathers and divides an aggregate fragment from each edge, wherein the aggregate fragment contains the microstructure placed near the stroke limit ; And. Classify the aggregates and aggregate fragments to obtain information representing a roughly optimal sequence, the microstructures will be processed according to the sequence to increase the yield at the processing point. 2. The method of claim 1 in which the stage is classified according to the energy consumed in at least one set of coils of at least one set of horses and sinuses in response to a motor command. 3. The method according to item 1 of the scope of the patent application, in which each of the aggregated and polymerized fragments can have a plurality of possible processing directions, and wherein the classification step includes a step of determining an approximately optimal direction for processing microstructures. 4. The method according to item 1 of the patent application scope, wherein the classification step includes a step of selecting a roughly optimal aggregate or aggregate fragment that is initially processed at the processing point, and then determining a majority to process the remaining aggregate and aggregate fragments Possible sequences and select approximately the most ife sequence from the plurality of possible sequences. 34 This paper size is in accordance with China National Standard (CNS) A4 (210 X 297 mm) (Please read the notes on the back before filling this page) * ϋ I ϋ mmmmmmm KiB_i n ϋ I · nn I-Line I Ministry of Economy Printed by the Intellectual Property Bureau's Consumer Cooperative 505956. 'A8 B8 CS. D8 VI. Application for patent scope 5. The method according to item 1 of the patent scope, where the microstructure is placed on a wafer. 6. A subsystem for determining a sequence of microstructures to be processed at a laser processing point, the subsystem comprising: means for receiving reference data representing the position of the microstructures to be processed at the processing point; I unites adjacent groups of microstructures to become a device containing aggregates of edge-gathered microstructures that contain microstructures that are placed close to the travel limit of a motor-driven stage, which is relatively The microstructure is moved by the laser beam at the processing point; a device that gathers and divides an aggregated fragment from each edge, wherein the aggregated fragment includes a microstructure that is placed near the stroke limit; and the aggregated A device that sorts and aggregates fragments to obtain information representing a roughly optimal sequence, wherein the microstructures will be processed according to the sequence to increase the yield at the processing point. ^ 7 The subsystem according to item 6 of the patent application scope, wherein the classification device is classified according to the energy consumed in at least one set of coils of at least one set of motors in response to a motor command. 8. The subsystem of item 6 of the scope of the patent application, wherein each of the clustering and clustering segments has a plurality of possible processing directions and wherein the classification device includes a device that determines the approximate optimal direction for processing microstructures. 9 · If the subsystem of the scope of patent application No. 6, wherein the classification device includes a selection of the approximately optimal aggregate or aggregate fragment that is initially processed at the processing point, and decides to process the majority of the remaining aggregate and aggregate fragments 35 This paper size applies to China National Standard (CNS) A4 specification (21 × 297 mm) ------------- Installation -------- Order ------- --- line (please read the notes on the back before filling this page) 505956 Cb 0 > > VI. Application for Patent Scope Intellectual Property Bureau, Ministry of Economic Affairs, Employee Consumption Cooperatives prints possible sequences and chooses from these many possible sequences Roughly optimal sequence of devices. 10. The subsystem of claim 6 in which the microstructure is placed on a wafer. 11. The subsystem according to item 10 of the patent application, wherein the microstructure is a chip wire β 12. The subsystem according to item n of the patent application, wherein the wire is a metal wire. 1 3 · The subsystem of item i of the patent application scope, wherein the wafer is a semiconductor memory element and wherein the wire is ablated at the processing point to repair the defective memory cell of the element. 14. The subsystem of claim 6 in which the microstructure is part of a semiconductor element. 15. The subsystem according to item 14 of the patent application scope, wherein the semiconductor element is a microelectromechanical element. 16. The subsystem according to item I4 of the patent application scope, wherein the semiconductor element is a silicon semiconductor element. 17. The subsystem according to item 14 of the patent application scope, wherein the semiconductor element is a semiconductor memory. 18. The subsystem of item 6 of the patent application, wherein the microstructure is part of a microelectronic component. 19. The subsystem of item 6 as claimed in the patent application, wherein the microstructures in each group have approximately a common gap. 20 · If the scope of patent application is the highest? Subsystem of item, in which the grade is 36 paper standards, applicable to China National Standard (CNS) A4 specification (210 X 297 public love) ί ', the meaning of α-? Please fill in this page) 505956' / vb ARCD Sixth, the scope of patent application-a kind of x-y stage and wherein the classification device is classified according to the energy consumed by the majority of the coils of the majority of motors in response to the motor command. i -----— Installation · —None of the above; 1-Please fill in this page urgently) Order: -line., printed by the Economic I Intellectual Property Bureau employee sales cooperatives 37 This paper size applies to Chinese National Standards (CNS) A4 size (210 X 297 mm)