JPH0955348A - Compression method of graphic data and graphic pattern generation apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the amount of graphic data by a method wherein the graphic pattern of a semiconductor integrated circuit is described more efficiently. SOLUTION: The position of a rectangle 101 is indicated by #x1, #y1, by a designation 'IP', as recognition numbers indicating a position. In the same manner, the position of a rectangle 102 is indicated by #x2, #y2 as recognition numbers, and the position of a rectangle 103 is indicated by #x3, #y3.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention manufactures a graphic pattern necessary for designing and manufacturing a semiconductor integrated circuit such as an LSI, and more particularly, a mask used as an original plate when a graphic pattern of a circuit is transferred onto a wafer. The present invention relates to a method for compressing graphic data and a graphic pattern generator in a graphic pattern at this time.

[0002]

2. Description of the Related Art A graphic pattern forming a semiconductor integrated circuit is generally represented as a polygonal graphic. Actually, for example, when forming a figure pattern on a photomask (original drawing substrate), these polygonal figures are converted into simple figures such as rectangles, trapezoids, and parallelograms and then drawn. In the circuit pattern design stage, even if a figure is represented by a polygon with 10 vertices, in a drawing apparatus using an electron beam or a laser, the polygon is a collection of simple figures such as rectangles. Is treated as.

This is because, in actually drawing a pattern, a simple figure such as a rectangle or a trapezoid is easier to generate, and even if the figure to be drawn is complicated, these simple figures are simple. Because it can be expressed by combining figures. Then, the drawing pattern cannot be generated by the drawing apparatus with the mask pattern as it is after the circuit design is completed, and a process of converting the figure pattern into the drawing data for the drawing apparatus is required.

FIG. 9 is an explanatory view showing a concrete example of the configuration of the graphic data of the above-mentioned drawing apparatus. FIG. 9A shows an example of data description when drawing rectangular data, and FIG. 9B shows a horizontal trapezoid. An example of data description when drawing is shown. As shown in FIG. 3A, the graphic data in the drawing apparatus is used for pattern formation using an electron beam and data K (Kind) indicating a graphic type indicating that a rectangle r (Rectangle) is drawn. When performing, there is data d indicating the dose D (Dose). Then, in order to indicate the position of the rectangle, a position X indicating the x coordinate and a position Y indicating the y coordinate of the lower left apex of the rectangle, and a value w of the width W and a value h of the height H as the size of the rectangle. Have and.

Further, as shown in FIG. 9B, in the case of a trapezoid, the figure type K is z (tranpeZoid). Then, data indicating the length w of the lower base W and the length u of the upper base U instead of the width, and the difference s between the x coordinates of the lower left apex and the upper left apex are provided as the parameter S of the hypotenuse. Although a horizontal trapezoid is shown in FIG. 9B, a parallelogram or a triangle may be used as a special case of this trapezoid. Also, even if all sides are quadrilaterals that are not parallel to the coordinate axes,
It can be expressed by a combination of horizontal and vertical trapezoids.

Further, in the above description, the coordinates of each vertex of the figure may be directly expressed. It is also possible to write the figures continuously. For example, in the description for a rectangle, 5 words are used, and the description order is: dose D, position X, position Y, width W, height H. You just have to decide the rule that it is becoming.

In the above description, the type of numerical data representing the attribute for describing a graphic has been described.
The value and range of these numerical data will be described. In a semiconductor integrated circuit such as a large-scale integrated circuit, the minimum formation pattern dimension is said to be achieved up to about 0.1 μm. In the description of such a fine pattern, for example, a fine grating with an interval of 20 nm is used. The attributes such as the position, width, and height of the graphic pattern are all determined in correspondence with the points on the grid. The value is then the grid spacing, in this case 2
It is expressed in units of 0 nm. This unit is called an address unit here.

Assuming that the chip has a side length of 20 mm, there are 1,000,000 20 nm gratings. That is, when describing a figure pattern in units of 20 nm, it is necessary to describe a 7-digit numerical value in order to specify the figure position and size. In practice, the chip area is, for example, 1
The area is divided into a small area of about mm or a smaller area of, for example, about 50 μm, in which graphic description is performed. This is because instead of drawing the entire chip from the side of the device that draws the pattern,
It corresponds to drawing a part of the divided area in order.

The above will be described in detail by taking a vector scan type electron beam exposure apparatus as an example. In pattern drawing using an electron beam, in order to position the irradiation position of the electron beam with high accuracy, the area in which the electron beam is deflected is limited to, for example, 1 mm square. This area is called a field. To draw an area beyond the field, a sample such as a mask substrate is mechanically moved to join the fields.

The drawing data is described for each field, and the pattern data in the field is described, for example, in a coordinate system whose origin is the lower left corner of the field. The position of the field is specified by the coordinate value with the lower left origin as the origin. When the address unit is 20 nm and the field size is 1 mm, it is necessary to specify a numerical value up to 50,000 for the graphic description in the field. Since 2 16 is about 64000, 16 bits (2 bytes) are required to describe this numerical value.

A total of 8 bytes is required to represent the position X, position Y, width W, and height H of the rectangle. In the trapezoid,
Furthermore, the length U and the parameter S of the hypotenuse are required. On the other hand, the figure type needs only 2 bits to specify a rectangle and two types of trapezoids, and about 8 bits is sufficient to specify the irradiation amount. It is possible with 2 bytes. FIG. 10 is a plan view showing a certain figure and an explanatory view showing the data describing the figure as described above. In FIG. 10A, 20
1, 203 is a rectangle and 202 is a trapezoid. And FIG.
Data of these figures are described in 0 (b).

In FIG. 10B, coordinate values are described in units of fields. On the other hand, as shown below, there is also a method of dividing a field into smaller regions called smaller subfields and describing a graphic for each subfield. In this description in units of subfields,
For example, the lower left corner is used as the origin of the subfield, where the origin is located in the coordinate system of the field, and the position of each figure is described as the distance from the origin of the subfield.

If the subfield size is 60 μm square and the address unit is 20 nm, the position description is 30
Numerical expressions up to 00 are sufficient, and can be handled with 12 bits. If 16 bits are given to represent each attribute, 4 bits remain, and it becomes possible to indicate what attribute the 12-bit numerical data represents by these 4 bits. As described above, if the description using the subfield is used, FIG.
As shown in FIG. 10B, it is not necessary to fix the description order, and as shown in FIG. 10C, when the attribute value of the previous figure is the same, omit it and inherit the previous value. Is possible. However, in this case, it is necessary to indicate that the description of the graphic has been completed, and it is necessary to allocate 1 bit of 4 bits to it.

In FIG. 10 (c), the symbol at the left end of each 2 bytes represents the content indicated by 4 bits, K indicates the figure type and irradiation amount, X indicates the x coordinate, and XE indicates the x coordinate. The designation of a value and the end E (End) of one graphic description are shown. In the case of FIG. 10A, a rectangle 201 and a rectangle 203
Has the same attributes except for the x coordinate, the rectangle 203
Can be described with only 2 bytes indicated by XE. In this way, the method to omit when the attributes are the same is
An efficient compression effect can be expected.

The following two methods have already been proposed as efficient methods for describing graphic data. One is to express a figure as an array when the figure is regularly arranged in a two-dimensional array, and will be called compression using an array instruction here. In the compression using the array instruction, when one figure or a set of figures consisting of a plurality of figures are regularly arranged at equal intervals, one or a set of figures which is a repeating unit,
This is a method of expressing by specifying the interval and number of repetitions in each of xy directions.

This enables efficient representation of a graphic pattern having a periodicity such as a memory. The other is the registration and reference of the figure group.If there is the same figure group that appears in multiple places in the chip, this figure group is registered in advance and the figure group is registered at the necessary place. It is a method of calling. It has a function similar to that of a program subroutine. Here, it is called registration and reference of the graphic group.

In this method, for example, in the vector scan type electron beam drawing apparatus, the position of the subfield can be arbitrarily designated, so that the subfield can be registered as a unit. The function of calling the data of the same subfield at different locations is effective in the large-capacity memory having a large-scale array or the pattern description of the underlying layer of the gate array.

[0018]

As described above, various measures have heretofore been made in order to efficiently describe graphic data. However, the degree of integration of LSI is reduced by the minimum size. With the increase in chip size, the amount of graphic data is increasing. In particular, OPC (Op
mechanical Proximity Correctio
It is becoming important to use a method of correcting the pattern distortion, which is called n), which is generated at the time of optical transfer on a photomask in advance. To make this correction, it is necessary to add an auxiliary pattern around the pattern to be formed. Therefore, it is inevitable that the number of patterns will increase dramatically.

In 1995, United States SIA (Se
According to the roadmap announced by the Miconductor Industry Association), it is expected that the amount of data in a 1-gigabit DRAM will be 8 gigabytes. this is,
The case where the data is compressed by the above-described conventional method is still a very large amount of data.

The present invention has been made to solve the above problems, and it is possible to reduce the amount of graphic data by more efficiently describing the graphic pattern of a semiconductor integrated circuit. To aim. This reduction in the amount of data not only makes it easier to handle the data, but it also reduces the amount of high-speed memory that the drawing device must have and saves the required computer resources. It is possible to obtain various effects, which leads to improvement in reliability.

[0021]

A method of compressing graphic data according to the present invention is a method for compressing graphic data necessary for expressing a graphic,
Equipped with a reference part that is replaced with an identification number that consists of data with the number of bits that can distinguish a figure.When expressing a figure with data, describe it by replacing it with an identification number, and when actually expressing the figure, refer to the reference part. Then, the identification number of the figure is replaced with the figure data. That is, the graphic data is represented by being replaced with an identification number having a small data amount consisting of the number of bits for distinguishing this kind of number. The graphic pattern generation device of the present invention is provided with a reference unit in which the graphic data is replaced with an identification number composed of data having a bit number capable of distinguishing the graphic. For this reason, the data provided to the figure pattern generator may be represented by being replaced with an identification number having a small amount of data, which is composed of the number of bits for distinguishing the generated figure pattern.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. The present invention realizes efficient compression by utilizing the feature of the graphic pattern of the integrated circuit. First, Tables 1 and 2 below show the features of basic graphic patterns in the concept of the present invention.

[0023]

[0024]

Table 1 above shows the frequency of appearance of the same figure. For example, the first wiring layer has a total of about 6.39 million graphics, and the types of graphics used here are 10,191. The type of figure referred to here includes not only the figure shape, but also its size and dose designation value, and if any of them is different, the type is different. When distinguishing the types of figures including such sizes and specified doses, the notation is distinguished from unique figures or figure types that handle only shapes such as rectangles and trapezoids as unique figure types.

About 10,000 kinds of these unique figures are
Table 1 shows the appearance frequencies. The most frequently appearing unique figure is used about 650,000 times, which corresponds to 10% of the total. Considering the top 64 types of unique figures that frequently appear, these alone alone account for 70% or more of the total figures (72% cumulative percentage).

That is, if these 64 types of figures are registered in advance together with their identification numbers, 70% or more of the whole can be described only by describing the identification numbers. This also applies to the element isolation layer. If only 64 types of identification are used, the amount of data may be 6 bits, which enables extremely efficient description of a unique figure.

Table 2 above shows the distribution of the X position coordinates of the figure when the subfield is used. Explaining in the first wiring layer, the x coordinate is specified about 6.39 million times in the entire chip, but it is about 6 when viewed as the coordinate value in the subfield.
It can be clearly seen that the 40 coordinate values are only used repeatedly. Top 6 most commonly used coordinate values
Four kinds can cover about 30% of the whole.

This also applies to the y-coordinate, and the x- and y-coordinate values that are frequently used are registered together with their identification numbers so that the coordinate values can be used with a smaller bit length. It becomes possible. As a result, in the conventional description, as described above, it takes 2 bytes to describe the coordinate value of either x or y.
If the identification number table is used, x and y can be specified simultaneously with 2 bytes.

The point of the present invention is to register frequently used data in advance and refer to the actual data by using an identification number having a shorter length than the actual data length. Therefore, it is desirable that the data to be registered requires a long data length to describe the actual data and is frequently used. As shown in Tables 1 and 2, since the unique data has a long actual data length and a part of the actual data is intensively used, it is effective to replace it with a short identification number.

Further, the array instruction is also expected to be effective because it is necessary to specify the number and interval in both the xy directions and it is expected that the number of types is small. Further, with respect to the xy coordinate values, x and y can be specified simultaneously with 2 bytes, so replacement with a shorter identification number is considered effective for reducing the amount of data.

As a unit for registering / referring to these data, there are an entire chip, a field unit, and a subfield unit. Hereinafter, description will be given using an example in which data is registered for each field. FIG. 1 is an explanatory diagram showing a data format in the embodiment of the present invention, and shows an example in which 128 unique figures, 64 array instructions, and 64 xy coordinate values are registered in a field. FIG. 1 schematically shows a specific arrangement of graphic data description. FIG.
In the above description, the bit sequence is actually described. However, here, those which only make the figure complicated and are not related to the description are indicated by "-".

The description of the graphic data is in units of 2 bytes (16 bits). In addition, 20 address units
nm, the subfield size was 75 μm, and the field size was 2.4 mm. When the address unit is used as a unit, the subfield can be covered with 12 bits and the field can be covered with 17 bits. Numerical data in the subfield can be specified with 12 bits or less such as coordinate values. Therefore, as shown in Table 3 below, the rule of using the upper 4 bits of 16 bits as a code of what the lower 12 bits represent Systematized.

[0034]

In Table 3, Nos. 1 and 2 are used for other purposes, No. 3 indicates the start of a subfield, No. 4 indicates the start of registration data, No. 5 indicates the x coordinate value, and No. 6 Indicates x coordinate and drawing, No7
Indicates the y coordinate, No8 indicates the y coordinate and drawing, and No9
Indicates the width of the rectangle or the length of the trapezoid lower base, and No. 10 indicates the height of the rectangle or the distance between the trapezoid lower base and the upper base. No11 shows the length of the trapezoidal upper base, and No12
Indicates a trapezoidal hypotenuse parameter, No. 13 indicates a figure type and dose designation, No. 14 indicates the start of a repeat command using 3 words, No. 15 designates the identification number of x and y coordinates, and No. 16 is unique. Specify the figure identification number, No
Reference numeral 17 designates the identification number of the repetitive instruction.

Table 3 above shows the data description rules for the description of FIG. Table 3 shows rules of data description for embodying the present invention, and symbols such as "SF" indicate what attribute value is indicated by data of generally 2 bytes (1 word) length. . The reason why it is described as “generally” is that there is also data composed of a plurality of words such as an array instruction. These attributes are indicated by the first 4 bits of the first word or the like. Here, the reason for using 4 bits or the like is that, as will be shown later, the reference of the recognition number of the figure or the repetitive instruction (array instruction) is determined by the first 5 bits.

The method of attribute designation will be briefly described below. SF means the start of a subfield and gives a unique number to the data of that subfield by the value of n having a 12-bit length, and in some cases, it is possible to repeatedly use the data of that subfield. There is. X, XE, YE, W, H, U, S and K are as described above. Here, only X or Y can be described with E, that is, as the last attribute value of the graphic description. The reason is that the positions of figures are different in all figures, except for very special drawing such as multiple drawing. That is, even if another attribute value such as W or H changes, W or the like is described first, and X or Y is described last.

In K, 4 bits (16 types can be designated) are assigned to the figure type, and 8 bits (256 types) are assigned to the irradiation amount designation. All of the above is described in one word,
The R array instruction is the number of repetitions (nx, n
y) and the number of figures to be iterated (# of fig
s) is 4 bits each and x, y in the second and third words
It is described by 3 words indicating the direction repeat interval. The registration header indicated by G is used for registration of the attribute value using the identification number. Unique attributes, array commands, and
There are four types of coordinate value X and coordinate value Y.

In the example shown in Table 3, the designation of the type of registration attribute is indicated by 3 bits, and the number of attribute items to be registered by the attribute is indicated by 9 bits (up to 512 types can be identified). The main purpose of specifying the number of items is to make the registration area variable length when the number of items to be registered is small. K shown in FIG.
ind = F means registering a unique figure (Figure), and the number of registered items is 128. The data of the unique graphic to be registered has the same description rule as that of the normal graphic data designation. With this notation, there is no item such as XY indicating the end of the graphic description in the attributes of the unique graphic, so the item cannot be omitted.

For example, in the case of a rectangle, one item can be registered by designating three attribute values K, W and H. Moreover, since the number of attribute items to be specified increases in the trapezoid, the description length differs depending on the type of figure. Although 128 kinds of unique figures can be defined in this way, the identification number when referencing is not specified,
It was decided to assign numbers from 0 to 127 in the order of appearance. "IF" in Table 3 refers to the registered unique figure
Then, the identification number can be specified by the lower 11 bits. However, the maximum value of the unique figure type that can be actually registered in the field is 512 because it is determined by the item number description at the time of registration.

In FIG. 1, the array instruction is registered next.
One array instruction is composed of three words, R'indicating the number of repetitions and dx and dy indicating the repetition intervals. The reason why R'is used here is that the number of figures to be targeted is defined in the normal repeat instruction shown in Table 3, but the number of figures is not specified in the array instruction to be registered. This makes it possible to reduce the number of array instructions and increase the frequency of reference. When referencing, as shown in "IR" in Table 3, it is necessary to also specify the number of target figures,
Here, the identification number is 6 bits (64 types) and the number of figures is 5
Bits are allocated. The type of array instruction to register is 6
4 is due to this limitation.

In FIG. 1, the coordinate values X and Y are continuously registered. The formats of x and y shown in Table 3 are also used for registration. Table 3 shows that the number of items registered in FIG.
As shown in “IP” of 6), 6 bits (up to 64 types can be identified) are assigned to the identification numbers at the time of reference.

As described above, after the registration process is finished, the description of the drawing figure is started for each subfield. In FIG. 1, after the start of the 0th subfield is indicated by "SF", "I"
There is a reference designation of the array instruction by "R". Since n = 6, it can be seen that it is effective for 6 figures. These six pieces are composed of two kinds of unique figures and are repeated.

FIG. 2 shows the state of the graphic arrangement according to the above description (FIG. 1). The "#" symbol in the figure indicates the identification number when referring to it. Although the description of the array instruction is omitted, the repetition intervals dx and dy are shown in FIG. 2 as a 2 × 2 array.
It was shown to. The identification number of the unique figure is described as # f1 or the like. The rectangles 101 to 103 in FIG.
˜106 correspond to # f2. The position of the rectangle 101 is fixed by the “IP” designation of # x1 and # y1, the position of the rectangle 102 is fixed by # x2 and # y2, and the position of the rectangle 103 is fixed by # x3 and # y3. The position of the figure using # f2 is
The rectangle 104 uses “IP”, but the rectangle 105,
106 uses coordinate values x11 and y12 according to the usual notation. Then, the y coordinate value corresponding to x11 of the rectangle 105 is #y designated by the graphic position of the rectangle 104 immediately before.
The value indicated by 4 is still used.

In the description shown in FIG. 1, thereafter, an example is given in which the positions of individual figures are continuously designated without using a repetitive command. Figure 3 shows how the figure is laid out.
It was shown to. In the figure, a rectangle 107 and a rectangle 109 are designated by "IP". In the rectangle 108, the x coordinate value is designated by x12 while using the y coordinate value of the rectangle 107. In FIG. 3, the reason that x11, y12, etc. are not referred using the identification number is that the number of times these numerical values appear is small, and in this example, they are not in the upper 64 positions.

In the description example shown in FIG. 1, data shared in the entire field is registered, but more efficient data description is possible by registering a graphic attribute suitable for the area in a smaller area. May be FIG.
This is an example in which only xy coordinate values are described, but the state is shown in which 32 types of XY coordinate values are registered in the field and 32 types of XY coordinate values are registered in the subfield. In this case, the identification number is 0
It is determined in advance that fields 1 to 31 are for field registration, and fields 32 to 63 are for subfield registration.

If there are not 32 types of coordinate values registered in the field, the unregistered identification number is not used, and the number that can be referred to again from the number 32 registered in the subfield. If the attribute values can be registered in the subfields as described above, even if a locally characteristic graphic pattern exists, it is possible to refer to the corresponding pattern. The same designation can be made for the array command and the unique figure.

On the contrary, when the data has no locality, for example, in the case of describing a large-capacity memory, it is desirable from the viewpoint of compression to share the attribute value in the entire chip. In this way, the effective registration / reference unit changes depending on the data to be handled, so specify at the beginning of the pattern data description which attribute value, what area unit, and how many items are to be registered / referenced. It is also possible to do so. Of course, these description rules are premised on that the electron beam drawing apparatus, which is a pattern generating apparatus, supports the function.

For example, when forming a 20 mm square chip, the size of this chip area is 100 mm square on a reticle that is 5 times larger than that used in photolithography. If this is filled up with 75 μm square subfields, the number of subfields exceeds 1.7 million. Therefore, the position designation data of the subfield within the field is also important for data compression.

The subfields are regularly laid out in the normal field. In the description of memory etc., from the viewpoint of data compression, even when assigning a figure group to a subfield and aligning the position of the subfield with the pattern layout,
The position becomes highly regular. Therefore, the reference by the identification number of the xy coordinate value has a great effect on the designation of the coordinate value of the subfield.

In FIG. 5A, the subfield 5 in the field 51 when the values assumed in FIG. 1 are used.
2 shows an example of arrangement. There are 32 subfields 52 per side of the field 51. That is, there are 32 × 32 = 1024 subfields 52 in one field 51, but there are only 32 types of xy coordinate values. If the coordinate values are directly described, each word has a 17-bit length, so two words are required to describe the x or y coordinate value. On the other hand, with reference to the identification number registered in advance, it is possible to specify both xy coordinate values with one word.

When registering subfield coordinate values, 64 kinds of coordinate values may be registered in total, but if the subfield area is a square, the same value can be used for xy, and in that case, 32 kinds of coordinate values can be used. Registration will be sufficient. FIG. 5B is an explanatory view showing a concrete description example thereof, and after the code indicating the registration start of the subfield coordinate value, the value is described every two words together with its identification number. The identification number is assigned to 12 bits of the first word, the upper 4 bits of the coordinate value is assigned to 4 bits, and the lower 16 bits of the coordinate value is assigned to the second word. The coordinate value can be described in 20 bits.

The description of the identification number is not always necessary. Further, if the xy tables are separately registered, it is possible to enter the recognition code thereof. The coordinate value corresponding to each subfield is described following the code indicating the start of description of the subfield coordinate value. Here, the description is made every two words, and the first word is n in “SF” in Table 3.
, The X coordinate value is specified by the subfield coordinate value recognition number in the upper 8 bits of the second word, and the Y coordinate value is also specified by the lower 8 bits.

The position designation within a field is common to each field unless the field position and the subfield position are aligned with the pattern layout. Therefore, the description of these subfield coordinate values is made prior to the graphic description of each field. If the positions of subfields are aligned with the pattern layout, the number of types to register increases, but x
If y is 256 or less, each xy coordinate value itself can be specified with one word. The same as the designation of the in-field coordinate value of the subfield can be applied to the designation of the in-chip coordinate value of the field. It is also possible to register field coordinate values in advance and refer to them.

Although the above description has been made on the premise of the existence of subfields, the present invention operates more effectively if the entire area is divided into small subareas when the graphic data of the entire chip is described. This is because The reason will be described below by taking Table 2 above as an example. In the first wiring layer shown in Table 2, X that appears when subfields are used
The type of coordinate value is at most 641 as shown in the section "rank". On the other hand, there are 27824 types of X coordinate values used when the entire chip is formed without introducing a small area called a subfield. By introducing a small area called a subfield, it can be seen that the types of coordinate values used are reduced by nearly two digits.

That is, nearly 30,000 numerical data are replaced by approximately 640 numerical data, and as a result, approximately 640 numerical data are obtained.
Since the frequency of appearance of numerical data of seeds has increased and the number of kinds has decreased, it is possible to shorten the data length that specifies the numerical value. Thus, when the present invention is applied to the description of position coordinates, more effective compression can be achieved by using a method of dividing the entire area into small areas.

Next, how to determine the size of this small area will be described. In the example of Table 2, the side length of the subfield area is 45 μm. Here, if this length is 4
At 5.1 μm, it is expected that the types of numerical data will not decrease as much as shown in Table 2. To give a simple example, the numerical value group described at 1.0 μm intervals is 10.0 μm.
If the small area is described as a unit, 10 kinds of numerical values may be repeatedly used, but if the width of the small area is 10.1 μm, new 10 kinds of numerical values must be used again in the adjacent small area. This means that it is important to determine the size of the small area in consideration of the periodicity of the coordinate values used.

In the design of an LSI, how to arrange the wiring at predetermined intervals is determined in advance in relation to the manufacturing process technology, and the numerical value can be used to easily determine the size of the small area. You can Even if the numerical value cannot be obtained, it is possible to estimate the coordinate period used in the design from the design data describing the mask pattern. For example, in the design pattern data, the data has a hierarchical structure, and upper cells call lower cells, so if you pay attention to the coordinate values to call, estimate the coordinate cycle being used. Is possible.

By the way, in the example shown in FIG. 1, a graphic is described in units of subfields, but the case of graphic description in units of fields will be briefly described below. When 2 bytes (16 bits) are required for the coordinate value description, etc., a mechanism for determining what the numerical attribute is is necessary. The description of data is handled in units of bytes, and units of 2 bytes and 4 bytes are particularly desirable for computer processing.

As an example, there is a method in which two bytes of control code are prepared at the beginning of each graphic description, and the content and length of the data description that follows is described therein. Considering the example corresponding to “IP” in Table 3, it is sufficient to specify that “the position is specified by the identification number using the unique figure used before” by using the 16-bit control code. "IP"
So, compared to the allocation in 4 bits, there is a sufficient margin. In this case, since the next 2 bytes can be used to describe the identification number of the xy coordinate value, 25 bits of 8 bits each are used.
It is possible to refer to six types of coordinate values.

If 2 bytes are assigned to the control code, at least one graphic 4 bytes is required for the description, but a configuration in which the control code is 1 byte is also conceivable. As an example, when it is desired to specify only the x-coordinate value by specifying the identification number, the upper 1 byte indicates that, and the lower 1 byte can describe the identification number. As described above, various designation methods are conceivable, but the main point is to register numerical data having a high appearance frequency together with the identification number in advance, and use the identification number for reference and use. Further, even when the coordinate value designation in the field exceeds 16 bits, the data description length is determined by the bit number of the identification number in the identification number reference. For this reason,
It has the advantage that the specification is short and it is easy to comply with the 2-byte constraint.

The above description shows a great effect when the array instruction compresses data in the LSI pattern. In particular, when the mask magnification is small, the number of figures included in one subfield is large, and the effect of the array instruction is great even in a pattern such as a memory that is clearly not the array itself. Even if you process data one by one without considering the sequence,
In the data compression, it is important to sort by the same unique figure when the drawing data is finally used and to judge the regularity from the position coordinate value and use the array command when the array command can be applied.

In this case, for the group of figures that can be arranged, only the position coordinates of the reference figures that are required when the array representation is performed as the position coordinates for description are necessary, and the coordinate values that can be calculated by expansion are meaningful. There is no. 3 shown in FIG.
Explaining with x4 array data, the reference coordinate value is x31,
At y31, the repetition interval is (x32−x31), (y
32-y31).

That is, the coordinate values other than x31 and y31 are not necessary as the coordinate values of the graphic description, but x31 and y3.
Even with 1, the appearance frequency is once. In other words, when determining the registered coordinate values, instead of simply paying attention to the figure and evaluating the appearance frequency of the coordinate value, the figures that can be arranged are arranged and the coordinate values necessary for the figure notation appear. desirable.

Hereinafter, using the patterns shown in Tables 1 and 2,
Table 4 shows a comparison between the case where the present invention is applied and the case where the present invention is not applied. The application condition of the present invention is that the number of registrations of repetitive instructions is 32 types at maximum in both fields and subfields, 128 types at maximum in unique graphics, and 3 maximum in x and y coordinate values respectively.
There are two types, and the coordinate value to be registered is determined from the appearance frequency of the coordinate value necessary for the description after the array notation. Also, registration of subfield coordinate values and reference by identification number are used. 57% for the first wiring layer and 51% for the element isolation layer
And the effect of the present invention is shown.

[0066]

Although the application methods of the present invention have been specifically described above, in order to realize them, a program for creating data based on the above-mentioned notations is used at the same time as interpreting these notations to obtain data. With the ability to restore
A graphic pattern generator such as an electron beam exposure device is required. As described above, data has been conventionally compressed, and thus the conventional electron beam drawing apparatus also had a function of restoring the compressed data.

However, in the present invention, the graphic data is compressed differently from the conventional one, and in order to utilize the compressed data of the present invention, it is necessary to be able to restore the compressed data of the present invention. Among the functions for decompressing the data compressed by the present invention, the functions of expanding the arrayed figures from the intervals and the number of repetitions, and referring to the registered figure group by changing the position are the same as the conventional ones. It is the same. However, in the figure pattern generating device according to the present invention described below, a numerical value that is frequently used is compressed and then restored, and is involved in the restoration of the numerical data itself in data compression.

First, a conventional figure pattern generator will be described. FIG. 7 is a configuration diagram showing a partial configuration of a conventional graphic pattern generation device. In the figure, the decompression circuit of the omitted numerical value in the compressed data is shown for each main function.
Note that, here, for simplicity of explanation, only rectangular data is targeted. First, before starting drawing, the drawing data stored in the magnetic disk (not shown) is written in the drawing data storage memory 302. This is because the high-speed writing with the electron beam cannot catch up if the data is read sequentially from the magnetic disk in each process. Here, for example, the target data is written in the drawing data storage memory 302 by using the time for the stage to move the data of the next field.

Drawing proceeds as follows. The address counter 301 sequentially reads the contents of the memory 302 into the register 303. As in the case of FIG. 1, the data is described in units of 2 bytes, and the 12-bit lower 303b indicates a numerical value or the like and the 4-bit upper 303a.
Represents the attribute. Register 303 is the upper 30
The data of 3a is passed to the decoder 304, and the decoder 304
Then, it is determined what the lower 303b indicates by the value. And an X register 310 for storing the position X,
Y register 311 for storing the position Y, W for storing the width W
Register 312, H register 313 for storing the height H,
A signal line 321 is used to instruct the corresponding register of the F register 314 that stores the type of figure and the D register 315 that stores the designated dose value to fetch the value of the lower order 303b.

By this series of processing, the value of the position X is stored in the X register 310, for example. This process is repeated by incrementing the value of the address counter 301 one by one. The decoder 304 uses “XE” and “YE” in Table 3.
When the end of one figure data such as is detected, the numerical data is set in the register and the signal line 32
The value 0 is set in each of the registers from the X register 310 to the D register 315 to notify the drawing control circuit (not shown) that one drawing is ready. In this way, the latest numerical data used last time is
It is possible to always hold in each register of the X register 310 to the D register 315.

Next, the graphic pattern generator according to the present invention will be described. FIG. 8 is a configuration diagram showing a partial configuration of the graphic pattern generation device according to the present invention. The operation of this figure pattern generator will be described below. As in the conventional case shown in FIG. 7, data is read from the magnetic disk (not shown) into the drawing data storage memory 302 prior to drawing.

However, in the figure pattern generator of the present invention, the data to be registered at this time is stored in the position coordinate X registration data storage memory 305 and the position coordinate Y registration data storage memory 3.
06, Stored in the unique figure registration data storage memory 307. Then, the other data is stored in the drawing data storage memory 302 as in the conventional case. The repetitive data are omitted here.

Lower 30 of 12 bits of register 303
3b is connected to each register from the X register 310 to the D register 315 as in the conventional case, and the position coordinate X
The registration data storage memory 305 is also connected to the unique figure registration data storage memory 307. This is configured such that when the reference commands of “IP” and “IF” shown in Table 3 come, the identification number thereof can be given to the storage memory.

The identification number is the address itself, and based on the result of judgment of the data of the upper 303a by the decoder 304, the reading from the corresponding registration data storage memory is enabled by the signal line 322, and the registered numerical data is registered. Are passed to the appropriate register in the X register 310 to the D register 315. At this time, the decoder 304 stores the position coordinate X registration data storage memory 305, the position coordinate Y registration data storage memory 306, in the corresponding registers from the X register 310 to the D register 315.
It is necessary to notify by the signal line 321 that the data to be fetched from the unique figure registration data storage memory 307 has arrived. Further, according to the description rule of Table 3, when the "IP" command is received, the signal line 320 notifies that one graphic drawing is ready.

In the above description, the timing is omitted at all. However, for example, it takes time for reading the values from the position coordinate X registration data storage memory 305, the position coordinate Y registration data storage memory 306, and the unique figure registration data storage memory 307. The time must be taken into account to set each register from 310 to D register 315. These functions can be easily realized by using the current digital circuit technology without using special techniques.

The subfield coordinate value or the field coordinate value can be restored in the same manner as the graphic data. In these coordinate value descriptions, it is highly possible that they are regularly arranged in an array. Therefore, if all the coordinate values are registered, all the data become the identification numbers and the actual data description is lost. Therefore, the hardware corresponding to FIG. 8 has a simpler configuration. Further, if the field coordinate values are used, the processing can be performed by software because the processing has a relatively long time.

The present invention can also be applied to display of graphic data. When displaying a large amount of data, the amount of data can be reduced by using the registration and reference using the feature of the graphic described in the present invention. And it is easy to realize. For computers that display graphic data, it is more efficient to use data access using indirect addresses.
It is also suitable for high-speed processing.

[0079]

As described above, according to the present invention, the graphic data necessary for expressing the graphic is replaced with the identification number composed of the data of the number of bits that can distinguish the graphic. Therefore, it becomes possible to describe the figure pattern more efficiently, and it is possible to reduce the amount of the figure data to be described as compared with the conventional case.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a data format in an embodiment of the present invention.

FIG. 2 is a plan view showing a state of graphic arrangement according to the description of FIG.

FIG. 3 is a plan view showing a state of figure arrangement when the positions of individual figures are continuously designated without using an array instruction.

FIG. 4 is an explanatory diagram showing an example in which position coordinates are registered in each of a field and a subfield.

FIG. 5 is an explanatory diagram showing a relationship between a field and a subfield.

FIG. 6 is a plan view showing a case where graphics are arranged in a 3 × 4 array.

FIG. 7 is a configuration diagram showing a partial configuration of a conventional graphic pattern generation device.

FIG. 8 is a configuration diagram showing a partial configuration of a graphic pattern generation device according to the present invention.

FIG. 9 is an explanatory diagram showing a specific configuration example of graphic data in a conventional drawing apparatus.

FIG. 10 is a plan view showing a figure and an explanatory view showing data describing the figure.

[Explanation of symbols]

301 ... Address counter, 302 ... Drawing data storage memory, 303 ... Register, 304 ... Decoder, 305 ...
Position coordinate X registration data storage memory, 306 ... Position coordinate Y
Registration data storage memory, 307 ... Unique figure registration data storage memory, 310 ... X register, 311 ... Y register,
312 ... W register, 313 ... H register, 314 ... F
Registers, 315 ... D registers, 320, 321, 32
2 ... signal line.

Claims (5)

[Claims] 1. A reference unit in which graphic data necessary for expressing a graphic is replaced with an identification number composed of data having a bit number capable of distinguishing the graphic, and when the graphic is represented by data, the identification number is provided. A method for compressing graphic data, characterized in that when a graphic is actually expressed by replacing it with, the recognition number of the graphic is replaced with graphic data by referring to the reference part. 2. The method for compressing graphic data according to claim 1, wherein the area in which the graphic is arranged is divided into small areas, and the position data of the graphic is described in a coordinate system within the small area. A method for compressing graphic data, characterized in that the graphic data is replaced with an identification number composed of data having a bit number by which the position of the graphic in the small area can be identified. 3. The method for compressing graphic data according to claim 1 or 2, wherein the graphic data is arranged in accordance with a predetermined rule, and the graphic data is assigned to the identification number. A method for compressing graphic data characterized by replacement. 4. The method of compressing graphic data according to claim 1, wherein when the same graphic is arranged in an array and the layout of the graphic is a desired form, A method of compressing graphic data, characterized in that, after being expressed, a coordinate number of the graphic having an appearance frequency equal to or higher than a desired threshold value is replaced with an identification number. 5. A graphic pattern generator for generating a graphic pattern indicated by the specified graphic data according to designated graphic data, wherein the graphic data is replaced with an identification number composed of data having a bit number capable of distinguishing the graphic. A graphic pattern generation device characterized by comprising a reference unit storing this.