1. introduction
2. bachus nauer forms
3. GDSII BNF
4. Record header
5. Data types
6. record types overview
7. record types description
8. example file

1. text presentation of GDSII file in  KEYformat
2. hex presentation of same file
3. GDSII file

1. Bit Array:

A bit array is a word which uses the value of a particular bit or group of bits to represent data. A bit array allows oneword to represent a number of simple pieces of information.
 
2. Two-Byte Signed Integer:

2-byte integer = 1 word 2s-complement representation. The range of two-byte signed integers is -32,768 to 32,767.

The following is a representation of a two-byte integer, where S is the sign and M is the magnitude.

smmmmmmm mmmmmmmm

The following are examples of two-byte integers:

00000000 00000001 = 1
00000000 00000010 = 2
00000000 10001001 = 137
11111111 11111111 = -1
11111111 11111110 = -2
11111111 01110111 = -137
 

3. Four-Byte Signed Integer:

4-byte integer = 2 word 2s-complement representation

The range of four-byte signed integers is -2,147,483,648 to 2,147,483,647.

The following is a representation of a four-byte integer, where S is the sign and M is the magnitude.

smmmmmmm mmmmmmmm mmmmmmmm mmmmmmmm

The following are examples of four-byte integers:

00000000 00000000 00000000 00000001 = 1
00000000 00000000 00000000 00000010 = 2
00000000 00000000 00000000 10001001 = 137
11111111 11111111 11111111 11111111 = -1
11111111 11111111 11111111 11111110 = -2
11111111 11111111 11111111 01110111 = -137
 

4. Four-Byte Real

4-byte real = 2-word floating point representation

(See 5.)
 

5. Eight-Byte Real

8-byte real = 4-word floating point representation

For all non-zero values:

• A floating point number has three parts: the sign, the exponent, and the mantissa.
• The value of a floating point number is defined as:
• (Mantissa) x (16 raised to the true value of the exponent field).
• The exponent field (bits 1-7) is in Excess-64 representation.
• The 7-bit field shows a number that is 64 greater than the actual exponent.
• The mantissa is always a positive fraction >=1/16 and <1. For a 4-byte real, the mantissa is bits 8-31. For an 8-byte real, the mantissa is bits 8-63.
• The binary point is just to the left of bit 8.
• Bit 8 represents the value 1/2, bit 9 represents 1/4, etc.
• In order to keep the mantissa in the range of 1/16 to 1, the results of floating point arithmetic are normalized. Normalization is a process where by the mantissa is shifted left one hex digit at a time until its left FOUR bits represent a non-zero quantity. For every hex digit shifted, the exponent is decreased by one. Since the mantissa is shifted four bits at a time, it is possible for the left three bits of the normalized mantissa to be zero. A zero value, also called true zero, is represented by a number with all bits zero.
The following are representations of 4-byte and 8-byte reals, where S is the sign, E is the exponent, and M is the magnitude. Examples of 4-byte reals are included in the following pages, but 4-byte reals are not used currently. The representation of the negative values of real numbers is exactly the same as the positive, except that the highest order bit is 1, not 0. In the eight-byte real representation, the first four bytes are exactly the same as in the four-byte real representation. The last four bytes contain additional binary places for more resolution.

4-byte real:

SEEEEEEE MMMMMMMM MMMMMMMM MMMMMMMM

8-byte real:

SEEEEEEE MMMMMMMM MMMMMMMM MMMMMMMM MMMMMMMM MMMMMMMM MMMMMMMM

Examples of 4-byte real:

Note: In the first six lines of the following example, the 7-bit exponent field = 65. The actual exponent is 65-64=1.

01000001 00010000 00000000 00000000 = 1
01000001 00100000 00000000 00000000 = 2
01000001 00110000 00000000 00000000 = 3
11000001 00010000 00000000 00000000 = -1
11000001 00100000 00000000 00000000 = -2
11000001 00110000 00000000 00000000 = -3
01000000 10000000 00000000 0000000 = 0 .5
01000000 10011001 10011001 1001100 = 1 .6
01000000 10110011 00110011 0011001 = 1 .7
01000001 00011000 00000000 00000000 = 1.5
01000001 00011001 10011001 10011001 = 1.6
01000001 00011011 00110011 00110011 = 1.7
00000000 00000000 00000000 00000000 = 0
01000001 00010000 00000000 00000000 = 1
01000001 10100000 00000000 00000000 = 10
01000010 01100100 00000000 00000000 = 100
01000011 00111110 00000001 00000000 = 1000
01000100 00100111 00010000 00000000 = 10000
01000101 00011000 01101010 00000000 = 100000
 

6. ASCII String

A collection of ASCII characters, where each character is represented by one byte. All odd length strings must be padded with a null character (the number zero), and the byte count for the record containing the ASCII string must include this null character. Stream read-in programs must look for the null character and decrease the length of the string by one if the null character is present.

• A through Z
• a through z
• 0 through 9
• Underscore (_)
• Question mark (?)
• Dollar sign ($)

• Contains an array of XY coordinates in database units. Each X or Y coordinate is four bytes long. Path elements may have a minimum of 2 and a maximum of 200 coordinates. Boundary and border elements may have a minimum of 4 and a maximum of 200 coordinates. The first and last coordinates of a boundary or border must coincide.
• A text, or Sref element may have only one coordinate.
• An Aref has exactly three coordinates. In an Aref, the first coordinate is the array reference point (origin point). The other two coordinates are already rotated, reflected as specified in the STRANS record (if specified). So in order to calculate the intercolomn and interrow spacing, the coordinates must be mapped back to their original position, or the vector lenght (x1,y1-> x3,y3) must be divided by the number of row etc. . The second coordinate locates a position which is displaced from the reference point by the inter-column spacing times the number of columns. The third coordinate locates a position which is displaced from the reference point by the inter-row spacing times the number of rows. For an example of an array lattice see the next picture.

Aref rotated -30 degrees.
• A node may have from one to 50 coordinates.
• A box must have five coordinates, with the first and last coordinates being the same.

• 0 for square-ended paths that endflush with their endpoints
• 1 for round-ended paths
• 2 for square-ended paths that extend a half-width beyond their endpoints

The following picture shows the pathtypes
 

	Pathtype 0 produces a square-ended path, ending flush with thedigitized endpoints. This is the de-fault pathtype if none is specified
	Pathtype 1 produces a round-ended path. The two ends aresemicircular with center at thedigitized endpoints.
	Pathtype 2 produces a square-ended path. The ends of the pathextend beyond the digitized end-points by one-half the path width.

 

34

GENERATIONS

2202

Two-Byte Signed Integer

This record contains a value to indicate the number of copies of deleted or back-up structures to retain. This numbermust be at least 2 and not more than 99. If the GENERATION record is omitted, a value of 3 is assumed.
 

35

ATTRTABLE

2306

Two-Byte Signed Integer

Contains the name of the attribute definition file. This record is present only if an attribute definition file is bound to the library. The attribute defenition filename may include directory specifiers (separated with "/") and an extension (separated with "."). Maximum record size is 44 bytes.
 

36

STYPTABLE

2502

Two-Byte Signed Integer

Unrelesed Feature
 

37

STRTYPE

2502

Two-Byte Signed Integer

Unrelesed Feature
 

38

ELFLAGS

2601

Bit Array

Contains two bytes of bit flags. Bit 15 (the rightmost bit)specifies Template data. Bit 14 specifies External data(also referred to as Exterior data). All other bits are currently unused and must be cleared to 0. If this record isomitted, all bits are assumed to be 0. The following shows an ELFLAGS record.

For additional information on Template data, consult the GDSII Reference Manual. For additional information on External data, consult the CustomPlus User's Manual.
 

Bit	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
word1	6 (hex) # of bytes in record
word2	26 (hex)								01 (hex)
word3	unused														external data	template  data

 

39

ELKEY

2703

Two-Byte Signed Integer

(Unreleased feature)
 

40

LINKTYPE

28

Two-Byte Signed Integer

(Unreleased feature)
 

41

LINKKEYS

29

Two-Byte Signed Integer

(Unreleased feature)
 

42

NODETYPE

2A02

Two-Byte Signed Integer

Contains two bytes which specify nodetype. The value ofthe nodetype must be in the range of 0 to 255.
 

43

PROPATTR

2B02

Two-Byte Signed Integer

Contains two bytes which specify the attribute number. The attribute number is an integer from 1 to 127. Attribute numbers 126 and 127 are reserved for the user integer and userstring (CSD) properties which existed prior to Release 3.0.
 

44

PROPVALUE

2C06

ASCII string

Contains the string value associated with the attribute named in the preceding PROPATTR record. Maximumlength is 126 characters. The attribute-value pairs associated with any one element must all have distinct attribute numbers. Also, the total amount of property data that may be associated with any one element is limited: thetotal length of all the strings, plus twice the number of attribute-value pairs, must not exceed 128 (or 512 if the element is an Sref, Aref, contact, nodeport, or node).

For example, if a boundary element uses property attribute2 with property value "metal," and property attribute 10 with property value "property," the total amount of property data is 18 bytes. This is 6 bytes for "metal" (odd-length strings must be padded with a null) + 8 for "property" + 2 times the 2 attributes (4) = 18.
 

45

BOX

2D00

No Data

Marks the beginning of a box element.
 

46

BOXTYPE

2E02

Two-Byte Signed Integer

Contains two bytes which specify boxtype. The value of the boxtype must be in the range of 0 to 255.
 

47

PLEX

2F03

Two-Byte Signed Integer

A unique positive number which is common to all elementsof the plex to which this element belongs. The head of the plex is flagged by setting the seventh bit; therefore, plexnumbers should be small enough to occupy only the right-most 24 bits. If this record is not present, the element is not a plex member Applies to Pathtype 4.Contains four bytes which specify indatabase units the distance a path outline begins before orafter the last point of the path. Value can be negative.
 

50

TAPENUM

3202

Two-Byte Signed Integer

Contains two bytes which specify the number of the current reel of tape for a multi-reel Stream file. For the first tape, the TAPENUM is 1: for the second tape, the TAPENUM is 2. For each additional tape, increment the TAPENum by one.
 

51

TAPECODE

3302

Two-Byte Signed Integer

Contains 12 bytes. This is a unique 6-integer code which iscommon to all the reels of multi-reel Stream file. It verifies that the correct reels are being read.
 

52

STRCLASS

3401

Two-Byte Signed Integer

Not used
 

53

RESERVED

3503

Two-Byte Signed Integer

This record type was used for NUMTYPES but was not required.

54

FORMAT

3602

Two-Byte Signed Integer

Defines the format of a Stream tape in two bytes. The possible values are:

1. for GDSII Archive format
2. for GDSII Filtered format
3. for EDSM Archive format
4. for EDSHI Filtered forrnat

A Filtered Stream file contains only the elements on the layers and with the datatypes you specify during creation ofthe Stream file. The list of layers and datatypes specified appear in MASK records. At least one MASK record must immediately follow the FORMAT record. The MASK records are terminated with the ENDMASKS record.
 

55

MASK

3706

ASCII string

(Required for and present only in FilteredStream file. )
Contains the list of layers and datatypes specified by the user when creating the file. At least one MASK record must immediately follow the FORMAT record. More than one MASK record may occur. The last MASK record is followed by the ENDMASK record.
In the MASK list, datatypes are separated from the layers with a semicolon. Individual layers or datatypes are sepa-rated with a space. A range of layers or datatypes is specified with a dash.

An example MASK list looks like this: 1 5 -7 10 ; 0- 255

56

ENDMASKS

3800

NoData

(Required for and present only in FilteredStream file.)
Marks the end of the MASK records. The ENDMASKS record must follow the last MASK record. ENDMASKS is immediately followed by the UNITS record.