CN115268232A - An Efficient Data Processing Method for DMD Oblique Scanning - Google Patents

Abstract

The invention relates to the technical field of laser direct writing exposure, in particular to an efficient data processing method for DMD oblique scanning. The invention adopts the combination of host server, SPIE data processing module, high-speed transmission module and data processing module to complete the efficient processing of image data, and utilizes the advantages of the host server's vector diagram synchronous calculation to complete the image simulation transformation, inclination and the striping required by multiple DMDs. With cutting, multiple strip vector images are sent to SPIE data processing module for rasterization synchronization processing, the rasterized data is subjected to contour compression, and transmitted from the high-speed transmission module to the data processing module, and the data processing module receives the data. Time-sharing synchronization operation is carried out from time to time to realize data zero-filling, contour filling, snapshot merging, burst storage zero-filling, mixing and inserting, etc., and finally do frame processing, display the data on the DMD, and realize the high-speed oblique scanning function. The present invention can process graphic data faster, so that high resolution matches high productivity.

Description

An Efficient Data Processing Method for DMD Oblique Scanning

technical field

The invention relates to the technical field of laser direct writing exposure, in particular to a high-efficiency data processing method for DMD oblique scanning.

Background technique

Printed circuit board is the mother of current electronic products. One of the most important processes in its manufacturing process is line exposure, which determines the yield rate and line width accuracy of the circuit board. As consumer electronics products become lighter and lighter, the line width and spacing of circuit boards become thinner and thinner. Especially in the field of flexible circuit boards, the film expansion and contraction in the traditional exposure process has seriously restricted the quality of flexible circuit boards, resulting in the rapid approval of the laser direct imaging process.

The current laser direct imaging equipment mainly uses TI's DLP solution, which is similar to a traditional projector, directly imaging the required lines on the target surface, and can realize simultaneous expansion and contraction with the substrate. In order to pursue a higher resolution of the process, the DMD is usually tilted so that a single pixel is partially overlapped in the scanning vertical direction. Oblique scanning reduces grid precision in a disguised form (improves analytical precision), but it will bring a larger amount of data, which affects the efficiency of laser direct imaging exposure, and is not conducive to the industrialization of this technology in the field of circuit boards.

Contents of the invention

The present invention proposes an efficient processing method for DMD oblique scanning, utilizes the advantages of the host multi-core chip and SPIE extension, and makes the data into a synchronous processing mode as much as possible, so that the underlying FPGA driving the DMD can process data simply and reliably, and realize DMD high-speed scanning exposure function.

For achieving the above object, the present invention provides following technical scheme: a kind of efficient data processing method of DMD oblique scanning, comprises host server, SPIE data processing module, high-speed transmission module and data processing module, is characterized in that:

The host server reads the vector graphics into the memory, borrows the advantages of the host for floating-point operations, and performs affine transformation, tilting, and strip segmentation operations on the vector graphics. The affine transformation can realize operations such as expansion and contraction, rotation, etc. , the strip division is to meet the requirements of synchronous scanning of multiple DMDs;

Described SPIE data processing module is inserted in the SPIE slot of mainframe server, and each SPIE data processing module can carry a plurality of DMD terminals, is connected by high-speed transmission module, and SPIE data processing module carries out sub-area segmentation, grid Operations such as formatting, contour compression, and data transmission can be synchronized in time;

The rasterization processing is divided into scanning direction and step direction analysis, and the analysis dpi of the scanning direction is determined by the minimum resolution multiple of the scanning axis motion feedback; the analysis dpi of the step direction is determined by the lens magnification, and at the same time by the subdivision factor n Decide;

The data processing module performs synchronous filling and double-sided zero padding after receiving the data compressed by the outline. The filling process is that each line of data starts from 0 filling, and the outline compression defines the flipped position information, which can be transformed according to the flipped position information. 0 or 1, followed by operations such as sampling, mixed insertion, burst storage and zero padding, so that the time-sharing synchronization of the data chain is formed from data rasterization to the final data storage in DDR, which meets the data volume of existing oblique scanning need;

Said sampling merging and mixing operation means that under the decision of the subdivision factor n, the DMD single point in the step direction is subdivided into n parts, and it is necessary to extract a value for each DMD single point, and combine these values in Together, it is convenient for post-framing processing;

The zero padding in the burst storage means that the length of each row of data matches the single storage length of DDR, so as to achieve the purpose of high-speed storage/reading;

Described framing processing refers to utilizing FPGA to carry out continuous address data reading to DDR according to the row number N that DMD needs, and this row number N must be the multiple of subdivision factor n, and each row of data corresponds to different DMD row numbers, needs For different starting addresses of intercepted data, the offset changes by 1 bit every n rows, and all rows of data are intercepted to form a frame of graphics, which is sent to the DMD driver chip, and the framing instruction is determined according to the position feedback of the scanning axis, that is Start framing every other multiple of the rasterization accuracy in the scanning direction to complete the entire high-speed scanning process.

The high-speed transmission module can use USB3.0, Gigabit power grid or 10 Gigabit optical fiber network port, depending on the actual data volume and cost of the system.

Compared with prior art, the beneficial effect of the present invention is:

The invention provides a high-efficiency data processing method for DMD oblique scanning, which uses a combination of a host server, a SPIE data processing module, a high-speed transmission module, and a data processing module to complete the efficient processing of image data, and utilizes the advantages of the host server vector diagram synchronous calculation to complete the image The simulation transformation, tilt and strip cutting required by multiple DMDs, and the multiple strip vector graphics are sent to the SPIE data processing module for rasterization and synchronous processing. The rasterized data is subjected to contour compression, and is transmitted to the data processing module by the high-speed transmission module. The data processing module performs time-sharing synchronous operation when receiving data, realizes data zero padding, contour filling, snapshot merging, burst storage zero padding, mixed interpolation, etc., and finally performs frame processing to display the data on the DMD to realize High-speed tilt scan function.

Description of drawings

Fig. 1 is the overall flowchart of the data skew processing of the present invention;

Fig. 2 is the schematic diagram of oblique scanning of the present invention;

Fig. 3 is the relation figure of DMD inclination and image inclination of the present invention;

Fig. 4 is a schematic diagram of the time-sharing synchronization operation from graphic rasterization to storage in the present invention;

Fig. 5 is a schematic diagram of parallel recombination arrangement of the present invention;

FIG. 6 is a schematic diagram of framing processing in the present invention;

Detailed ways

The technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

A high-efficiency data processing method of DMD oblique scanning proposed by the present invention can decompose the exposure pattern into scanning direction and stepping direction resolution in order to achieve high precision of exposure. The resolution in the scanning direction is determined by the minimum resolution multiple of the motion feedback of the scanning axis; the resolution in the stepping direction is determined by the size of the DMD single point and the lens magnification, and the oblique scanning can subdivide the DMD single point to further improve the resolution in the stepping direction. resolution. The tilt angle A of the tilt scan is related to the subdivision factor n, that is, A=arctan(1/n).

The invention utilizes the advantage of synchronous calculation of the vector diagram of the host server to complete the simulation transformation and tilt of the image and the strip cutting required by multiple DMDs, and sends the multiple strip vector diagrams to the SPIE data processing module for rasterization synchronous processing. The rasterized data is subjected to contour compression, and is transmitted to the data processing module by the high-speed transmission module. The data processing module performs time-sharing synchronous operation when receiving data, realizes data zero padding, contour filling, snapshot merging, burst storage zero padding, mixed interpolation, etc., and finally performs frame processing to display the data on the DMD to realize High-speed tilt scan function.

Plotted with n=4 in the example. After the DMD is tilted, a triangular area will be formed on each side of the stepping direction. Because of the need for multiple DMD exposure splicing, the triangular area needs to be out of light, that is, the DMD is in the off state in the triangular area (the bitmap is 0 in this area) . In order to achieve high-efficiency processing, all zero padding operations here are delegated to the data processing module for execution. Under the influence of the subdivision factor, the rasterized bitmap presents a 4-column arrangement mode. According to the DMD single point position, it is necessary to extract points in the bitmap to obtain values. DMD single point can be understood as the combination of 4 pixels in the stepping direction in the figure, because the tilt effect can be equivalent to all points arranged in a row in the scanning direction. The effective area is 2 rows of dots, and the triangular area is 1 row of dots. In order to meet the requirements of exposure uniformity, the triangular area should not be exposed, that is, 4 columns of zero padding operation.

The effective rank of DMD is H×L, and the subdivided rank is TH×TL after tilting according to the subdivision factor n, TH=(H-1)×n+L, TL=(L-1)×n+H. The number of invalid columns in the triangular area InL=H-n, the total required number of columns for a single DMD can be confirmed according to the total number of columns and the number of invalid columns, and then the width of the strip division can be confirmed.

Tilting is generated in the scanning direction. For scanning exposure, the bitmaps before and after tilting can be equal, so the graphic tilting operation is placed in the vector map with less data. According to the transformation matrix acquired by alignment, affine transformation is performed on the vector diagram in the host server, and then the vector diagram is divided into strips according to the effective column width of each DMD, and the operation of the tilt angle A of the scanning direction is done well. Segmented and skewed strip vector graphics are sent to the corresponding SPIE data processing module for rasterization processing.

The time-sharing synchronous model of graphics rasterization and storage into DDR, that is, strip area cutting→rasterization→contour compression→high-speed transmission→parsing and subpackaging→zero padding→snapshot merging→mixing and interpolation→burst storage zero padding → Store in DDR. Each step is in a synchronous operation state after the first execution, that is, the previous area bitmap is being transmitted at high speed, and the latter strip area cutting is already being executed. The design of each step of time-sharing synchronization is consistent in time planning, which can control the time from the end of all rasterization to the last area bitmap stored in DDR within 0.01s, and realize the front end of efficient processing of oblique scanning, as shown in Figure 5.

Finally, according to the position feedback pulse of the scanning axis, the data processing module will perform frame processing, that is, form a frame of graphics required at the moment, and transmit it to the DMD driver chip, as shown in Figure 6. The specific way to read data is as follows:

1) The initial address of the first line of the first frame is D ₀ , and the interception length is L

2) The initial address of row i of frame 1 is D ₀ +floor(i/4)-1

...

3) The initial address of the first row of the jth frame is D ₀ +(j-1)n·N _a , where N _a is the length of each row of graphics after zero padding when it is stored in the DDR, and the interception length is L

4) The initial address of row i of frame j is D ₀ +(j-1)n·N _a +floor(i/4)-1.

Claims (2)

1. A high-efficiency data processing method for DMD oblique scanning comprises a host server, an SPIE data processing module, a high-speed transmission module and a data processing module, and is characterized in that:

the host server reads the vector diagram into the memory, and performs affine transformation, inclination and stripe division on the vector diagram by using the advantages of the host on floating point operation, wherein the affine transformation can realize operations such as expansion and contraction, rotation and the like, and the stripe division meets the requirement of synchronous scanning of a plurality of DMDs;

the SPIE data processing module is inserted into an SPIE slot of a host server, each SPIE data processing module can carry a plurality of DMD terminals and is connected by a high-speed transmission module, the SPIE data processing module carries out operations such as sub-area division, rasterization, outline compression, data transmission and the like on the divided strips, and the operations can be time-sharing synchronization;

the rasterization processing is divided into a scanning direction and a stepping direction for analysis, and the analysis dpi in the scanning direction is determined by the minimum resolution multiple of the motion feedback of the scanning axis; the analytic dpi of the stepping direction is determined by the lens multiplying power and a subdivision factor n;

the data processing module carries out synchronous filling and bilateral zero padding after receiving data of outline compression, wherein the filling processing is started from 0 filling of each line of data, the outline compression defines overturning position information, the data can be filled in 0 or 1 in a convertible way according to the overturning position information, and operations such as point drawing, mixed insertion, burst storage zero padding and the like are carried out next, so that time-sharing synchronization of a data chain is formed from data rasterization to final data storage in a DDR, and the data volume requirement of the existing oblique scanning is met;

the snapshot merging and mixed insertion operation means that the DMD single points in the stepping direction are subdivided by n parts under the determination of a subdivision factor n, snapshot values need to be taken for each DMD single point, and the values are combined together to facilitate the subsequent framing processing;

the burst storage zero padding means that the length of each row of data is matched with the single DDR storage length, so that the aim of high-speed storage/reading is fulfilled;

the framing processing means that the FPGA is used for continuously reading the address data of the DDR according to the number N of lines required by the DMD, the number N of the lines must be multiple of a subdivision factor N, each line of data corresponds to different DMD line numbers, different initial addresses of intercepted data are required, the offset varies by 1 bit every N lines, all the lines of data are intercepted to form a frame of graph and transmitted to a DMD driving chip, a framing instruction is determined according to position feedback of a scanning axis, namely framing is started once every multiple distance of rasterization precision of one scanning direction, and the whole high-speed scanning process is completed.

2. The method for processing data of DMD in slant scan according to claim 1, wherein the step of: the high-speed transmission module can use USB3.0, gigabit power grid or ten-gigabit optical fiber network port, depending on the actual data volume and cost of the system.

Images