1.1 Graphic layout of the mask set using a CAD editor such as kic, vem, or ledit. kic creates an output file describing the layout in KIC format. vem creates an output file describing the layout in OCT format. ledit creates an output file describing the layout in GDS format.

1.2 Translation of the layout file into a CIF (Caltech Intermediate Format) file using the kictocif, octtocif, and gdstocif commands (respectively). Verification of the layout with an actual printed plot (hardcopy) is available through the cifplot command.

1.3 Translation of the CIF file into a MANN file using the ciftomann command. MANN is the format understood by the Microlab's GCA MANN 3600 Pattern Generator.

1.4 Writing the MANN file into the proper format for the pattern generator using the m36gen command. This creates two EBCDIC files, which must then be transferred to the pattern generator PC (gcapg.eecs.berkeley.edu).

1.5 Generation of masks on the GCA Mann 3600 Pattern Generator.

1.6 Development of masks in the APT automatic mask developer. This process is summarized in the following flowchart:

This document reviews the software aspects of mask generation, specifically, the programs you will need to get your design from layout graphic to mask. For information on operation of the pattern generator and the automatic mask developers, refer to manual entries on the wand.

2.0 CAD Software for Mask Generation

To get started, you will need:

► A good, basic, working knowledge of the UNIX operating system. Without it, learning to use the software involved in mask making is difficult. If you have no UNIX experience, try to find a student who can get you started. Commands you should be familiar with include: ls, cd, mv, cp, vi, and rcp.

► A working knowledge of kic, vem, or ledit. Documentation on these layout editors is available in the Microlab lobby.

Before you begin, please take note of the following important information. It may save you a lot of time and trouble.

► Use the following conventions when naming your files to prevent confusion:

.kic suffix on all KIC format files

.oct suffix on all OCT format files

.gds suffix on all GDS format files

.cif suffix on all CIF format files

.mann suffix on all MANN format files

.tap suffix on all TAP/TIX tape data files

.tix suffix on all TAP/TIX tape index files

As an example, if you have a layout, which consists of an array of MOSFETs, call the KIC format file mosfet.kic, the CIF format file mosfet.cif, etc.

► It is strongly recommended that each directory of CAD layouts contain a file called file.info which specifies the following information:

o the layer names used in the layout;

o the scale of your layout in lambda/µm (this is used with kictocif);

o the scale factor specifying the layout to mask ratio (this is found in the command file used with ciftomann);

o the polarity of your masks;

o whether the layout is centered about the origin (this determines whether you will need to use the '-c 0 0' flag with m36gen). Name this file for the KIC file it applies to, e.g., mosfet.info for mosfet.kic

► The layout must not exceed the usable area of the mask to be made. The following table lists usable ranges for the masks available in the Microlab:

Actual

Plate Size

English

Mask Size

for the gcapg

Metric

Loss

from Borders

Usable Range

in X-Y

2.5 inch

3.0 inch

4.0 inch

5.0 inch

60.5 mm

74.0 mm

98.6 mm

125.0 mm

2 × 6 mm

and corner areas

51.5 mm

64.2 mm

89.6 mm

115.0 mm

5.15 × 10⁴µm

6.42 × 10⁴µm

8.96 × 10⁴µm

11.5 × 10⁴µm

For exposure area on the GCA 10 X Reduction Stepper (gcaws, 5" masks), please refer to lab manual Chapter 4.04, GCA 8500 Wafer Stepper

Be careful to work within the usable range. Consider whether you will be using a scale factor with ciftomann (see Section 2.3.2), since this will affect your usable range coordinates.

You have the option of centering your layout later during tape generation (see Section 2.3.11, m36gen: -c option), but this will change the coordinates of your alignment marks, which must be specified if you intend to use 5" masks with the GCA 10X Reduction wafer stepper.

► The pattern generator cannot generate any geometry < 2 µm on the mask.

► If you are inverting in ciftomann, make sure that there are no "spaces" < 2 µm.

► Layer names used with kic must be between 2 and 4 characters, all of which MUST be upper case.

► Layer names in any kic file must exactly match the layer names in the .KIC file found in the same directory, or kic will not run.

► If you are going to generate a dark field emulsion mask, be sure to place small geometries (such as 2 µm boxes) at the outer corners of the useable range of the layout (see above) to ensure that the field is completely blocked out.

► If you need a small, single open geometry (dark field), use chrome.

► If you need a small, single dark geometry (clear field), use emulsion. If emulsion edge resolution is not good enough, the job cannot be done here.

► If you are making a set of aligning masks, make sure that GCA alignment marks are present.

► Before running ciftomann, make sure your scale factor is correct.

► If you must make polygons, see staff.

► Try to place the physical center or your layout on or close to (X, Y) = (0, 0).

► Alignment mark offsets translate as follows:

(+100 X, +100 Y) (+0.1 X, -0.1 Y)

in µm in mm

on layout (kic) entered into stepper job spec

(Assumes mask label placed to the left in stepper reticle holder)

2.1 Graphic Layout

This section describes the graphic editors, CAD programs, and layout utilities available in the Microlab. You may use other utilities (e.g., AutoCAD) if they are able to produce layout files in one of the supported formats (KIC, CIF, or GDS).

2.1.1 kic

Kic is a menu-driven, interactive, color graphics program for the layout of integrated circuits. Unlike other programs at Berkeley, such as magic, kic does not have design rule checking and other such features. On the other hand, kic is very versatile, and can also be used to lay out designs for sensors, probes, etc. Several programs have been written by students that add to the versatility of kic; these programs can be found on argon in /cad/bin. Examples of kic layouts can be found on argon in /cad/examples. Kic should not be run on argon; it slows down the operation of all lab-related software. Please be sure to log in to a fast workstation (e.g., Gold, Boron, Iridium and Radium) and run kic from there.

Refer to the following documents for further information:

► Manual page for kic available on-line on all computers which support kic by typing man kic.

► The kic Manual by Giles Billingsley and Ken Keller is now available on line (Click here).

► You can also obtain a hard copy by typing tbl ~cad/lib/kic/kic.me | eqn | psroff –me at any Sun station that support kic. The output will appear in the Microlab Office.

These, along with other useful documents, are also available from staff, 406 Cory, at a cost of $5.

2.1.2 vem

vem is an interactive graphics shell/editor for IC designs represented using the OCT Data Manager. The primary purpose of vem is to provide a means of looking at the graphics representation of OCT views and invoking various CAD tools on these views. vem also provides standard graphics editing capabilities for physical layout design and schematic capture.

Refer to the following documents for further information:

► Manual page for vem available on-line on all computers which support vem by typing man vem.

► VEM Tutorial and Customization documents by David Harrison, available from staff.

2.1.3 ledit

L-Edit is a commercial layout editor licensed from Tanner Research Incorporated. It is a screen-oriented editor with simple menus for drawing various shapes on different layers. A PC version of the L-Edit software is currently available on a dedicated PC in the Microlab lobby. This new version of L-Edit has replaced the old Unix-based version of the software. L-Edit can be used with circuit simulation and analysis modules available from Tanner Research, but the Microlab has not yet purchased licensing for these products. L-Edit supports Tanner's own proprietary file format (TDB) as well as the popular GDS standard. Unlike kic, L-Edit supports rotated boxes and polygons.

Refer to the following documents for further information:

► L-Edit User Guide from Tanner Research, available in the Microlab lobby.

2.2 Translation Scripts

This section describes programs and shell scripts that will attempt to guide you through a multi-step file conversion process, usually from a graphic layout (KIC or GDS file) through to the final mask files. These scripts are generally more user-friendly but less flexible than the conversion utilities described below.

2.2.1 k2m

k2m is a shell script to convert KIC-format cells into MANN format for use with the pattern generator. For files without rotated boxes, we usually run kictocif and ciftomann. For files with rotated boxes, we can use k2m, which goes from kic directly to mann format. The KIC layout must first be flattened. Boxes must, of course, be rectangles, which can be rotated if desired. The layout should be centered.

k2m may be called with the following arguments:

-i inputfile

Name of KIC input file

-k layerfile

Name of layer file (instead of default layers.cmd)

-l lambdasize

Specifies microns-per-lambda scale of KIC file

-o outputfile

Name of file to contain output (MANN data)

-s scale

Specifies optical reduction factor

Refer to the following documents for further information: UNIX manual page for k2m (man k2m)

2.2.2 make_masks

make_masks is a shell script to convert KIC-format cells into MANN format for use with a pattern generator. It is similar to k2m but more user-friendly. Users will be prompted for information and stepped through the conversion process. make_masks also checks for many common errors such as the presence of wires or polygons that cannot be translated accurately.

make_masks may be called with the following arguments:

inputfile

Name of KIC input file

Refer to the following documents for further information: UNIX manual page for make_masks (man make_masks)

2.2.3 gds2tap

gds2tap is a shell script to guide users through the process of translating a GDS layout file to the TAP/TIX format used for pattern generation. Users will be prompted for information and stepped through the conversion process. gds2tap does a small amount of error checking as each conversion step is completed.

gds2tap may be called with the following arguments:

gdsfile

Name of the GDS layout file to be converted.

layerlist

List of GDS layer numbers to be converted, separated with commas or spaces

reduction

Optical reduction expected: 1x, 4x, 5x, or 10x

top_structure

Name of the top-most feature to be converted.

centered

yes if mask be centered at [0,0], otherwise no

units

Specifies microns or mils as the unit of measurement for this layout

Refer to the following documents for further information: UNIX manual page for gds2tap (man gds2tap)

2.3 Conversion Utilities

This section describes the utilities available to convert files from one CAD format to another. These utilities are called by translation scripts (described above) during the translation process. These utilities can be cumbersome to use but they offer processing options and flexibility that may not be available from the translation scripts.

2.3.1 kictocif

kictocif flattens and translates the KIC file created by kic into a CIF file.

Flattening is the process by which sub-cells (called instances) are incorporated into the layout. Otherwise, sub-cells are only referred to or called up in the KIC file. That is, the KIC file does not contain the information in the instance; only the name and position of the instance in the present cell.

Translation of the KIC file to the CIF file is necessary before generating the MANN file, which is in a format that is understood by the GCA MANN 3600 Pattern Generator. Also, hardcopies of a layout (plots) are obtained from CIF files.

kictocif is best invoked in an interactive mode, by typing kictocif.

The program will then prompt with:

Microns per lambda?

For ease of scaling, layout on kic is done in dimensionless units of lambda. The units are later determined in the kictocif translation. Specify scale of layout in response.

For example,

Microns per lambda? .25

means the layout was done in the scale of 0.25 microns per lambda.

The next prompt is:

Hierarchy's root cell?

Reply with the name of the KIC file you laid out using kic, file.kic.

The next prompt is:

Select layers you want to be visible -

s symbolic only

d detailed only

a all layers

Layers in kic are defined as one of two types, symbolic or detailed.

Symbolic layers are always displayed on kic, i.e., features in symbolic layers are always displayed, even when instances are included in the layout and kic is not in expand or peek mode. Detailed layers are displayed in instances only when kic is in expand or peek mode. (See kic Tutorial.)

The separation of layers into the two different types allows selective translation and generation of one mask layer of a given layout.

For further information, consult the manual page for kictocif, available on-line on all computers which support CAD tools by typing man kictocif

2.3.2 ciftomann

ciftomann fractures and translates a CIF file into a MANN file, which is then suitable for mask generation by a pattern generator.

Note: ciftomann can now be run on Gold, Boron, Iridium and Radium (Solaris).

Fracturing is the process by which large areas are broken into small areas suitable for exposure of photosensitive mask plates by the GCA MANN 3600 Pattern Generator.

In addition to the input and output files, ciftomann requires a command file which specifies such information as input layer names, output layer names, scale factor, and layer inversions.

ciftomann is invoked by typing:

ciftomann -i input_file -o output_file -l log_file -f command_file -p PG3600

-i input_file

Specifies the name of the input file, file.cif.

-o output_file

Specifies the name of the output file, file.mann.

-l log_file

Specifies the name of the log file to which diagnostic and error messages are written, file.err.

-f command_file

Specifies the name of the command file, file.comm.

A command file contains information pertaining to the scale of the mask vs. the layout, layers that are to be translated, how they are to be named, and mask polarity specifications. A typical command file, which, by convention should be names file.comm, looks like this:

scale 10

CONT invert CONT

METL METI

PII PII

NII invert NII

scale 10 specifies that the layout be magnified by a factor of ten. The scale factors to use are as follows:

Projection System

Reproduction Size

Scale Factor

Layout Size

Quintel Contact Printer (quintel)

Canon Mask Aligner (canon)

I-Line GCA Wafer Stepper (gcaws)

G-Line GCA Wafer Stepper (gcaws2)

1:1

4:1

10:1

4”

10 mm × 10 mm

8 mm × 8 mm

10 mm × 10 mm

CONT, METL, and PII are input layer names corresponding to layer names in the KIC file. If you only wish to translate certain layers, leave the extraneous ones out of this file.

CONT, METI, PII, and NII are the output layer names. Output layer names must be unique, and should be no longer than 4 characters. Input layer names need not be unique, as shown above.

Mask polarity is specified using the "invert" command between input and output layer names. Two types of masks are available, which are of opposite polarity. Chrome offers greater accuracy when dealing with line widths smaller than 2 µm. To decide whether inversion is needed based on the mask polarity you desire, please refer to the following figure.

-p pattern_generator

specifies for which pattern generator the MANN file is being generated. For the Microlab, the pattern generator available is the GCA MANN 3600. This is designated as PG3600.

For further information, consult the manual page for ciftomann, available on-line on all Microlab computers by typing man ciftomann.

2.3.3 cif2gds: Converts a layout from CIF format to GDS format. It is usually invoked by typing:

cif2gds file.cif file.gds

where file.cif is the name of the CIF file to be converted and file.gds is the name of the GDS file to be created.

For further information, consult the manual page for cif2gds, available on-line on all Microlab computers by typing man cif2gds or visit their website at: http://www.artwork.com/gdsii/cif2gds.htm. The Microlab has a license for cif2gds.

2.3.4 gds2cif: Converts a layout from GDS format to CIF format. It is usually invoked by typing:

gds2cif file.gds file.cif

where file.gds is the name of the GDS file to be converted and file.cif is the name of the CIF file to be created.

For further information, consult the manual page for gds2cif, available on-line on all Microlab computers by typing man gds2cif.

2.3.5 dxf2gds: Converts a layout from DXF format to GDS format. It is usually invoked by typing:

dxf2gds file.dxf

where file.dxf is the name of the DXF file to be converted.

2.3.6 gds2dxf: Converts a layout from GDS format to DXF format. It is usually invoked by typing:

gds2dxf file.gds

where file.gds is the name of the GDS file to be converted.

2.3.7 gds2pg: Converts a CAD layout from GDS format to an intermediate format of fractured shapes for pattern generation. This is the first step of a two-step process to create pattern generator data. The second step (mmask) converts the fractured data into the exact command syntax required for the pattern generator.

gds2pg is usually invoked by typing:

gds2pg file.gds feature_size layer_list =

file.gds is the name of the GDS file to be converted. feature_size is the size (in GDS units) of the smallest expected gap or line width in the layout. layer_list is list of layer numbers to be converted, separated by commas without spaces. = indicates that the top-most structure in the file should be converted.

Converted structures are written into output files named structNN.INT, where struct is the name of the structure being converted and NN indicates the layer number of these shapes. The .INT suffix indicates that the file contains data in INTermediate PG format.

Conversion results are summarized in a log file input_file.log. This file may contain useful information about errors that occurred during conversion.

For further information, consult the manual page for gds2pg, available on-line on all Microlab computers by typing man gds2pg.

2.3.8 mmask

mmask converts a CAD layout from intermediate PG format (.INT file) to an ASCII representation of a GCA Mann 3600 command file (Davis-Wilder Mann format, a .DWM file).

mmask is usually invoked by typing:

mmask -u:microns -s:1 -p file.INT

-u:microns indicates that units of the mask are microns, not mils.

-s:1 indicates that data should be converted at a scale of 1:1, i.e. no scaling.

-p indicates that mmask should run without user pauses.

file.INT is the name of the intermediate data file to be converted.

Converted intermediate data from file.INT will be stored in file.DWM. A DWM file must be sorted with pgsort2 and converted to TAP/TIX format with m36gen before it can be downloaded to the pattern generator.

For further information, consult the manual page for mmask, available on-line on all Microlab computers by typing man mmask.

2.3.9 pgsort

This program will sort the MANN file in such a way as to minimize the stage movement on the pattern generator. It must be run on every mann file before making a tap/tix file. It is invoked by typing:

pgsort file.mann file.mann.s

where the '.s' indicates that this is the sorted version of file.mann. After sorting, file.mann.s is ready for conversion to TAP/TIX format with m36gen.

For further information, consult the manual page for pgsort, available on-line on all Microlab computers by typing man pgsort.

2.3.10 pgsort2

This program is an alternate version of pgsort above, licensed from Artwork Conversion Software. The name pgsort2 is used to distinguish Artwork's version from the Microlab version. pgsort2 will sort and optimize the flashes in a Davis-Wilder Mann file (.DWM format) to reduce the number of movements and rotations required during pattern generation. pgsort2 is usually invoked by typing:

pgsort2 -uM -t3600 file.DWM

-uM indicates units of the mask are in microns.

-t3600 indicates this data is for a Mann 3600 pattern generator.

file.DWM is the name of the Davis-Wilder Mann file to be optimized.

Optimized data from file.DWM will be stored in file.SOR. This file is also in Davis-Wilder Mann format; the .SOR suffix is used to indicate that this data has already been sorted. The sorted file must be converted to TAP/TIX format with m36gen before it can be transferred to the pattern generator.

For further information, consult the manual page for pgsort2, available on-line on all Microlab computers by typing man pgsort2.

2.3.11 m36gen

m36gen takes a .mann file and generates two EBCDIC files suitable for use on the GCA MANN 3600 Pattern Generator.

m36gen is invoked by typing:

m36gen [-c x y] [-t] [-n] -T newfile file.mann.s

-c x y

Center the masks about the point (x , y), where x and y are integers. Units for GCA MANN 3600 Pattern Generator are 0.1 µm. -c 0 0 will center the mask around the origin, coordinates (0, 0). Be aware that if you use this option to center your layout, the coordinates of your alignment marks will change!

-t dx dy

Translate the masks by dx in the x direction, dy in the y direction. dx and dy are integers with units of 0.1 µm for GCA MANN 3600 Pattern Generator.

-n name

The label on each mask will be of the form name layer_name. If not specified, the name defaults to your login name, in capital letters.

For example,

m36gen -c 0 0 -n CMOS2 -T cmos2 cmos2.mann

generates a tape, whose masks are centered at the origin, and labeled CMOS2 layer name.

m36gen -t 500 500 -T cmos2 cmos2.mann

when executed, will cause the masks to be generated to be translated by 50 microns, in both x and y directions, from its present position.

For further information, consult the manual page for m36gen, available on-line on all computers which support CAD tools by typing man m36gen.

2.4 Layout Display Software

This section describes software that may be used to inspect a layout on screen or on paper. Visual inspection may be useful to track down problems in a layout or converted file.

2.4.1 pgcam

pgcam may be used to display a PG file on an X Windows graphics terminal. pgcam supports common PG data formats (including Mann 3600 and INTermediate files) and provides menu options for browsing and simple editing of fractured layout structures. pgcam is useful for inspection and manual fix-up of fractured layout files.

For further information, consult the manuals from Artwork Conversion Software, available in the Microlab lobby.

2.4.2 pgview

pgview is helper program for pgcam above. pgview speeds up the inspection process by automatically creating a job file for pgcam. The command pgview file1.tap file2.tap will create a job file containing the two named TAP files and then invoke the pgcam viewer.

For further information, consult the online manual page for pgview.

2.4.3 cifplot

cifplot may be used to convert a CIF file into HPGL-2 with Raster Transfer Language extensions which is suitable for plotting on the color wide HP Designjet T1100 plotter in the lobby of the Microlab.

cifplot requires a Colormap file. The Colormap file contains the description of which colors will be used for the individual mask layers (from kic) on the actual plot.

If you have never run cifplot before, it is recommended that you run cifplothp first. cifplothp will set up a ~/.cadrc file for you, and it will run cifplot -H for you.

The -H flag to cifplot creates the HPGL file.

By default, cifplothp will create a ~/.cadrc file which will use /cad/share/CIF/msucolors as the Colormap file. One may modify their ~/.cadrc file to refer to their own Colormap file. On the west wall of the lobby, is a chart, which specifies what color is referenced by which number.

In general, cifplot -H will create a file suitable for the plotter. At the present time, cifplot creates files that use TIFF compression (which the HP plotter understands).

The color plotter is known as plotter. One may queue a HPGL file created by cifplot with the following command:

lp -dplotter filename

Or starting from the cif file one may directly convert the file and plot it with the following command:

cifplot -H cif-filename | lp –dplotter

Reminder: cifplot is a very CPU intensive program. Please, do not run cifplot directly on Silicon2 or 3; run it on Boron or Gold.

Normally, cifplot returns with the expected size of the plot, and the user is then prompted for a decision whether or not to continue the plot. One may also re-scale the plot.

    ***COLOR CIFPLOT v2.0***
Scale: 1 micron is 0.009464 inches (240x)
The plot will be 3.27 feet
Do you want a plot?
y       Proceed with the plot.
n       Abort the plot.
r       Rotate plot.
s       New scale factor.
(y,n,r,s): y

3.0 Mask Layout Checklist

Here is a final checklist you should go through before you attempt to generate your masks:

3.1 Polarity and the Pattern Generator

► Polarities specified in ciftomann correct?

► Dark field emulsion? Geometries needed in corners to force pg to fill field.

► Small, single open geometry (dark field)? Use chrome - do not invert layer in ciftomann to generate an emulsion mask.

► Small, single dark geometry (clear field)? Use emulsion - do not invert layer in ciftomann to generate a chrome mask. If emulsion edge resolution is not good enough, send the job out.

3.2 Alignment Marks

► More than one mask? Alignment marks needed.

► Offset Coordinates: (assumes mask label placed on left)

Example:

Position On Layout (kic) As Entered In Stepper Job Spec

(x,y) (x,y)

in µm ▬► in mm

(+100, +100) (+ 0.1, - 0.1)

**Note sign change in y!

3.3 Scale

► Scale factor used in kictocif correct (µm/lambda)?

► Scale factor used in ciftomann correct (10, 4, 1)?

► All geometries < 2 µm on all edges?

► Inverting? Will there be filled "spaces" < 2 µm?

3.4 Centering

► Is the layout centered properly? i.e., is the physical center or your layout on (x, y) = (0, 0)?

► -c option used with m36gen? Alignment mark offsets will change!

3.5 Non-Standard Geometries

► Polygons? Use several layers of overlapping boxes if steps needed are < 2 µm. Use box2mann instead of ciftomann.

► Non-Manhattan geometries? Use a non-Manhattan layout tool (e.g. L-Edit or Cadence), save your layout as a GDSII file, and then convert with gds2tap.

3.6 Summary

Record the following in the directory containing your layout in the file filename.info where filename is the same as filename.kic.

► Layer names used in layout

► µm/lambda scale in kic

► Alignment mark offsets from origin in µm

► Whether the -c option is used with m36gen

► Scale factor in command file

► Size of masks

► Mask type (chrome vs. emulsion) for each layer

► Polarity of each layer (dark field, light field)

Appendix A

Frequently Asked Questions

Here is a summary of questions dealt with in this FAQ:

1. How do I lay out and make a mask?

2. How do I start up kic?

3. Once I am in kic, how do I call up my file?

4. Should I use boxes or wires to create rectangular features?

5. Do I need to overlap my boxes if I want to make sure that the pattern generator won't leave spaces between them?

6. How do I repeat a design in my layout?

7. How do I make circles?

8. How do I get all of my layers to show up in my kic window?

9. Can the pattern generator make polygons?

10. What are the minimum and maximum box sizes , which the pattern generator can create?

11. Can I use MAGIC to create my layout?

12. Is it true that I should not have designs near the border of my layout?

13. Do I need an alignment mark on every layer of my layout?

14. Is the pattern generator software sophisticated enough to always correctly invert the data for clear field chrome masks?

15. Can I use flash36 to make circles?

16. When using Kic, sometimes the SELec option does not work. What do I do?

17. Can I use subdirectories in kic?

18. How do I make an arc?

19. What is the easiest way to convert a kic file into the format necessary for the pattern generator? What are the tradeoffs between the different techniques?

20. What is a 'lambda'? Why do I care?

21. What is the file .KIC (or .KICLayers) used for? Why don't I have one?

22. What does 'Manhattan geometry' mean?

23. What other programs exist to help me with my layout?

24. What use is a 'symbolic' layer?

25. How does the alignment happen using the GCA steppers?

26. How can I check that mann file is okay?

27. How do I transfer my mask files to the GCA pattern generator?

1. Q: How do I lay out and make a mask?

A: kic is the most common mask layout software used at Berkeley. The easy way to get started is with the online tutorial. On argon, the text file describing the tutorial is in ~cad/doc/kic.me; type tbl kic.me |psroff –me for a printout; type tbl /cad/doc/kic.me | nroff -me | lookat for viewing on-line. Type man kic for more information.

Most users now work at workstations running X-windows and use versions of kic called xkic or xkic3. Some users do their layout with other layout software such as vem or ledit. There are online manual pages and additional documentation for all of these programs on argon.

2. Q: How do I start up kic?

A: xkic is available on all Microlab workstations, though it may only be displayed on color consoles. It is recommended that you run it on Solaris computers for speed (Gold, Boron, Iridium, Radium). Please do not run kic on Silicon or Argon, as it will slow it down considerably.

You type xkic or xkic3 on the command line.

3. Q: Once I am in kic, how do I call up my file?

A: You click on Edit on the left hand side of your kic window. Kic will prompt you with, Cell? at the bottom left corner of your kic window. Now you may type in the filename of the file, which you'd like to edit or look at.

You may also include the file name on the command line, for example xkic3 myfile.k will start an xkic session and load in the file myfile.k automatically.

4. Q: Should I use boxes or wires to create rectangular features?

A: Boxes are recommended, but with the proper care, wires can be used in layout. Specifically, if your wires are Manhattan geometry and you are using ciftomann as your conversion program then there shouldn't be any problems.

5. Q: Do I need to overlap my boxes if I want to make sure that the pattern generator won't leave spaces between them?

A: This should never be required, the pattern generator is normally maintained so that adjacent boxes leave no space or line between them. Furthermore, if you use ciftomann any overlap is deleted before the mask is made anyway.

6. Q: How do I repeat a design in my layout?

A: Designs are repeated in kic using what are called instances. These are simply kic files that you call into other kic files. A partial analogy might be the Unix file system --- directories contain files and subdirectories just like kic files contain boxes and instances.

To use an instance you first create the instance file just like any normal kic file. Let’s assume you called it myinst.k (the name is arbitrary, including the file extension). Next you open the kic file where you wish to use the instance. Once here, first click on Insta on the left hand side of your window to move to the instance submenu. Click on the MASte (master --- the name of the instance) and type the filename at the Master? prompt at the bottom of the window (in this case type myinst.k). A copy of this file may now be placed anywhere in your layout simply by pointing with the left mouse button. This may be repeated as often as necessary, and you may also rotate or mirror the instance before including it into the file (play with the menu buttons on your own, or read the kic manual for details). This "point and include" feature is active as long as you remain in the instance submenu.

If you want several regularly spaced copies of the instance, you can use the array features of kic. Click on #X and #Y and enter how many copies of the design you want in the x and y directions, respectively. Next click on DX and DY to set the amount of space to place between copies of the instance. Finally, point to the coordinate where you'd like the array to be positioned. I don't know of any limit to the nesting depth (i.e., your instances may contain further instances, and so on), but you may not have a loop of instances (in other words, the nesting must not be recursive---hopefully it is obvious why this is necessary).

7. Q: How do I make circles?

A: Using kic, the flash command generates polygons that cannot currently be converted into the proper format for making masks here at Berkeley. Instead, use Richard Moroney's program round (see #23).

If you are using ledit, circle shapes should be converted into polygons and fractured into rectangular mask flashes automatically.

8. Q: How do I get all of my layers to show up in my Kic window?

A: There are two reasons why you might not see your layers in kic: 1) they have the hidden attribute, or 2) boxes in that layer are contained in instances.

Check to see that the layer has the visible attribute by looking for a colored box next to the layer name at the bottom of the kic window. If there is none, then move to the ATtri (attributes) submenu and select VISib (visible) to choose which layers should be visible.

Boxes in instances may be made visible using either the EXpnd (expand) or Peek (peek) buttons. Expand makes boxes in instances visible in both the main window and the magnified window, peek only makes boxes in instances visible in the magnified window. Note that "symbolic" layers are always expanded in instances, but expanded up one instance level only.

9. Q: Can the pattern generator make polygons?

A: The pattern generator can only make rectangular boxes with a limited range of aperture size and rotation angle. Kic is capable of laying out many fancy objects like arcs, circles or polygons, but these must be converted to simple rotated boxes before a mask can be made on the pattern generator.

10. Q: What are the minimum and maximum box sizes , which the pattern generator can create?

A: The minimum box size is 2 µm; maximum is 950 µm. These result in a minimum feature size of 0.2 µm when using the GCA wafer steppers (10X reduction). The maximum aperture size of the PG is limited by exposure considerations (preventing lines or spaces from occurring between adjacent boxes), there is no limit to the maximum size of a box in kic. The resolution of a box on the PG is 0.1 µm (0.01 µm with the GCA steppers).

The maximum position for the center of a box on the PG is ± 57.5 mm from the center of the plate. Note that only features within ± 50.0 mm will print on your wafer, the extra area is typically used for fiducials and mask labels.

11. Q: Can I use MAGIC to create my layout?

A: Yes, as long as you have not used rotated boxes, and your file can be put into cif format. (vem is capable of handling rotated boxes.) If you have the entire MAGIC package, then GSDII can be put into cif as well.

12. Q: Is it true that I should not have designs near the border of my layout?

A: No, you may place your designs anywhere in the exposed area of your wafer. Realize that the GCAWS does not expose the entire 10 mm field of the die, it chops off the corners. A kic file marking the printing range of this stepper is available from most kic users who use the GCA.

13. Q: Do I need an alignment mark on every layer of my layout?

A: No. You only need an alignment mark on the first layer of your wafer, you may then align every later layer to this mark. It is generally prudent to include a mark on every layer so you have the option of aligning to any previous masking step, but this is not necessary at all.

14. Q: Is the pattern generator software sophisticated enough to always correctly invert the data for clear field chrome masks?

A: Not always. Have you tried making an emulsion master and then copying to chrome? However, if your features are smaller than 5um, emulsion is not the answer either. The pg is really not the best tool for making 1:1 contact masks but it is more than adequate for making the 10X masks for the stepper. If you need better masks for the contact printer, Marilyn has made a copy of the information on outside mask making services and she can send that on to you if you wish. To the best of my knowledge, if you have Manhattan-only geometry you can invert your mask without difficulty using ciftomann. If you use rotated boxes you cannot invert your mask.

15. Q: Can I use flash36 to make circles?

A: No - there is no translation for circles made by flash36, which the pattern generator understands. Please see #7 above.

16. Q: When using Kic, sometimes the SELec option does not work. What do I do?

A: Note that polygons are most easily selected using the "AREa" method, instead of "SELec." You only need to point to an area including any part of the polygon to select the entire polygon (or wire, for that matter).

17. Q: Can I use subdirectories in kic?

A: By including the command Path? in your .KICLayers file you can list a path of directories where kic will hunt for your instances. I use Path? ( . inst text/home/bsac/0h/moroney/kic/lib) and then make subdirectories called inst (for instances) and text (for text labels) in each of my layout directories. When asked for the Master? in kic, I just enter myinst.k instead of inst/myinst.k.

18. Q: How do I make an arc?

A: Some versions of xkic (and xkic3) have been properly modified so that the ARC menu option may be used inside kic itself. You can check the version you are using by making an arc, then using AREa to select just a part of the arc. You should see the outline of a few rotated boxes that make up the arc. If instead you see the entire arc highlighted then you have an uncorrected version of kic and may not use this feature.

You may also make arcs by creating your layout in L-Edit by using Richard Moroney's program arc (see #23).

19. Q: What is the easiest way to convert a layout file into the format necessary for the pattern generator? What are the tradeoffs between the different techniques?

A: This is a matter of taste. There are four general ways to solve this problem.

(1) Use kictocif and ciftomann (system utilities). Works fine if you have no rotated boxes and all Manhattan geometry. Ciftomann does some optimizations on the file, including eliminating over-lapping boxes. It also can be annoying when this algorithm produces unexpected results (very narrow boxes that cause errors), or rounding problems on the output file. Mark made an updated version of ciftomann, called ciftomann_new that solves this problem - it is installed on the BSAC cluster, are converted using poly2mann. Resulting files are merged together.

(2) Use k2m (Richard Moroney's shell script using his box2mann program). This is essentially a version of poly2mann that will work on Manhattan geometry also. This program converts a flattened kic file to a mann file in one step and is quicker and (in my opinion) more efficient than using options (1) or (2).

(3) Use make_masks (Jack Judy's shell script). This is essentially identical to Richard's k2m script except that it prompts the user for more information and generally assumes you need help understanding what is going on. Functionally it is identical to k2m.

(4) If you are using L-Edit or Cadence to create GDS files, use gds2tap to guide you through the conversion process. gds2tap will prompt you for necessary information and generally assumes you need help understanding what is going on.

20. Q: What is a 'lambda'? Why do I care?

A: Lambdas are the units used in kic files. You must decide how many lambdas represent a micron, typical choices are 0.1microns-per-lambda and 1.0 microns-per-lambda. Note that kic files are written with a resolution of 0.01 lambdas, but the minimum box size you may enter in the kic window is 1 lambda. Boxes smaller than this are allowed, you just have to create them manually (either by software or hand editing of your kic file).

21. Q: What is the file .KIC (or .KICLayers) used for? Why don't I have one?

A: This file (.KIC for xkic, .KICLayers for xkic3) defines options for kic such as layer names, colors and a search path for instances. You may use the system default, or selecting UPdat (update) in the ATtri (attributes) submenu will create a file in your current directory.

22. Q: What does 'Manhattan geometry' mean?

A: Manhattan geometry means all lines are at right angles to each other, you can't have a box turned 45 degrees, for example.

23. Q: What other programs exist to help me with my layout?

A: Richard Moroney has written a number of useful utilities for layout and file conversion, they may be found (along with some help documentation) in the directory /cad/bin on argon. "Man" pages also exist for many of Richard's utilities, but not all. Mail moroney@eecs.berkeley.edu for more information or help.

24. Q: What use is a 'symbolic' layer?

A: Layers with the attribute "symbolic" are always visible in instances, even when you have not selected peak or expand. They are useful for marking connection points to your instances, making labels, and such.

25. Q: How does the alignment happen using the GCA steppers?

A: First you align the mask to the GCA column using the fiducials on the mask. Next you align the wafer to the machine using the joysticks and your entered key offset. The mask itself is never directly aligned with the wafer, the marks you align to on the monitor are generated by the computer, not your mask.

26. Q: How can I check that mann file is okay?

A: Though a mann file contains strings identifying boxes, one cannot run xkic on a mann file. To check that a mann file is okay, one could make a mask, or one can try converting the mann file back to a kic file (with mann2kic).

27. Q: How do I transfer my mask files to the GCA pattern generator?

A: Mask files may be transferred to the GCA PG via network file sharing from your home directory on argon. For more information, see the procedures described in lab manual Chapter 3.3 (GCA 3600F Pattern Generator), Section 3 (GCA Pattern Generator PC).

Appendix B

L-Edit Mask Making Manual

The purpose of this document is to give some helpful hints on using L-Edit to generate mask layout patterns and successfully convert the layouts using the Microlab’s conversion utilities (gds2tap).

Advantages of this layout software

► Easy to use - This is a straightforward layout editor, which is easy to learn and is recommended for lab members with little mask making experience.

► Any angle features - L-Edit allows for any angle lines and polygons, arcs, circles, pie slices and with the new user programmable interface (UPI) can create a wide variety of additional features such as spirals and lettering.

► Quick turnaround - Because the program is easy to use, it has speed up the mask making procedure significantly. Masks that used to take a week to design and layout now takes a day.

► Advanced features - L-Edit has many advanced features such as process modeling which allows you to visualize the processing steps of the individual layers.

L-Edit is a user-friendly layout editor. The program allows you to create multi-layered mask patterns using wires, boxes, circles or arcs of any angle or orientation. The program is straightforward but there are a few beginner points that will make it easier to get started. There is a comprehensive manual for the program, which is located above the computers in the Microlab.

Technology File and Settings

The program will initially open into a preset technology file that defines the layers, wire widths for the individual layers, internal units, grid spacing etc.

You may use these default settings or create one specifically for your masks. Some of the settings defined in the technology file will be important when you try to convert the mask and it is best to get these issues straight before you begin.

Internal Units - This defines the precision to which the computer stores your features. This should be set to at least 10 times smaller then your smallest feature size. Problems with conversion will occur if you make a wire that is the same width as the internal units. The computer defines a wire by centering it across the selected coordinates and then would try to place the wire as half an internal unit on either side of the selected coordinate. A good value for the internal units for a contact mask is .1 or .01 microns. So 10 or 100 internal units per micron. Also, make sure you define your spacing in units of microns.

Wire Widths - You can define the default width of a wire for each layer in the layer settings. Individual wires widths may then be edited using ctrl E or by the edit pull down menu when the wire is selected.

Grid and Snap - This feature is very useful when generating masks by eye. It is often helpful to change this often to parallel the feature size that you are working with.

Getting Around

Panning and Zooming is easily done by using the right hand number pad or arrow keys for panning and the + and - for zoom. You can pan and zoom while you are drawing objects or window selecting which makes it easy to freehand draw features such as long thin lines.

Manhattan, 45-degree angles, any angle features

The buttons on the left that designate a wire, box or circle can be clicked multiple times to switch between strictly Manhattan geometry (90 degree angle features), features at 45 degree angles and any angles.

Cells and instances

You may use many different cells that can contain part or all of the final mask feature. It is often useful to create a cell that contains a feature that is repeated many times in your final layout (top most cell). You can then call (instance) this cell multiple times or instance an array of cells onto your top most cell. If you decide to change one feature in a cell that is repeated, the change will be copied into all the instances of that cell. It is important to remember the name of your top most cell for the conversion. The conversion program can take a guess of what your top most cell will be by using the default (=) but it sometimes picks the wrong cell. If your conversion is producing the wrong pattern, make sure you are converting the correct cell. You may also wish to create a unique name for your top most cell because this will be used as part of the final name of your converted files.

Things to keep in mind about the conversion program

Drawing acute angles

The conversion program has difficulties converting acute angles. It will often take a simple acute angle and break it into a dozen or so rectangles. If an acute angle is necessary, it can be easily and quickly created by manually breaking the angle into small rectangles. Remember that the smallest feature that the pattern generator can flash is a 2-micron square so do not make anything that will appear on the mask as smaller than 2 microns. This method is illustrated below.

Also, if you are generating angles that protrude out of a feature, it is often more efficient to define the angle as a rotated box vs. a triangle or part of a polygon. This method helps avoid acute angles in your patterns and is shown below.

Both methods will simplify conversion and yield a much lower flash count.

Gaps in the Conversion

The conversion program will occasionally have gaps when converting random curves. If you have areas with strange features, make sure you view the file using pgcam during the conversion to make sure there are no missing spots. Pgcam has a very useful feature that allows you to place boxes to cover up any gaps.

User Programmable Interface

The new version of L-Edit for the PC and soon for Unix is able to utilize C scripts to generate patterns. This capability is very handy for generating complex features such as spirals, ovals or arrays of objects that differ slightly. The UPI can also generates alphanumeric lettering.

Getting Ready for the Conversion

In order to convert a mask layout, you must save your files in the .gds format. Each layer must have an associated gds number, which you will need for the conversion. This can be changed on either the layers or cif/gds export drop down menus depending on which version you are using. In this same menu, there is an option to define how many rectangles are used to define a circle. Circles are generated with an array of rotated rectangles. The more rectangles you use, the smoother the circle will be. The default is 64 but 20 is sufficient for a well-defined circle and will reduce your flash count. Finally, save your file with the suffix .gds, which can be selected using the save as command. You are now ready for the conversion.

Flattening you files eliminates all the cell structures and puts everything onto a single cell. You do not need to flatten you files in order to convert but you may wish to eliminate any unnecessary cells before conversion.

Conversion

Converting your files is straightforward using the gds2tap conversion script. The conversion programs perform a series of steps and although it is not necessary to understand the details of the conversion program it may help explain what is occurring with your files.

The pattern generator uses a pair of shutters, one for each dimension, which will adjust to define a box or rectangle with a minimum width of 2 microns and a maximum width or a few hundred microns. The shutters can rotate to create angled features and have a minimum rotation angle of 0.5 degrees. The mask is placed on a stage that is translated to expose the entire mask. In order to create the arbitrary features in a mask layout the files must be fractured into a series of boxes and rectangles that the pattern generator can expose. This is the first step in the conversion program. gds2tap allows you to view your files after this step in order to verify that your conversion has performed correctly.

Another procedure the conversion program performs is to sort the individual flashes to increase the speed and reduce the wear on the pattern generator (pg).

The rest of the conversion programs are used to get the information into a form that the pg can understand. Our pg is fairly specialized and accepts files as a modified cif format. Every feature must be defined as a box and the pg will not accept wires or polygons even if they are four sided polygons. The gds2tap script will automatically perform all the functions necessary to make the file pg ready. This conversion can take some time especially for large files and on the slower computer systems.

Things to know before you begin the conversion process.

► Name of the file you wish to convert.

► gds layer numbers for the layers you wish to convert.

► Name of the top most cell. Using the default (=) will allow the program to search for the top most cell.

► Units for the conversion (microns is default).

► Which lithography system you are using to expose your wafers.

► Whether you want the mask centered.

The script will prompt you for the information at the beginning of the program. To initiate the script, simply type gds2tap.

So that's it. Everything should work perfectly. Well in theory at least but we have worked hard to make this process simple and straightforward. If you have any problems or have other suggestions that would be helpful to the Microlab community please discuss it with the Microlab staff so that it can be incorporated into a continuously improving system. Since we have one site license, the conversion can only be performed by one person at a time. If your program stops in the middle, someone else may be using a program.

Printing

L-Edit does a fairly good job of printing the features, however, the resolution is limited. To obtain a high quality image of your mask layout you can print from pgcam during or after the conversion. Pgcam does an excellent job of printing multilayer layouts with good resolution. It can also print the image to several file types including tiff and eps.

Appendix C

Mask Making for ASML Wafer Stepper (6”)

Purpose

This chapter gives some guidelines for 6” mask generation from your layout design.

Applicable Document

Chapter 4.12 (ASML DUV Stepper Model 5500/90)

Reticle Specifications

Nominal edge length: 6-inch square

Recommended thickness: 0.250 inch

Optional thickness: 0.150 inch or 0.120 inch

Recommended glass type: Ultra low thermal expansion quartz

Recommended reticle flatness: 1 micron

Pattern Specifications

The maximum field size is:

Wafer Level

Reticle Level

Diameter

Maximum Y

Maximum X

29.7 mm

21.0 mm

148.5 mm

105.0 mm

For correct operation, in addition to the device layout image, the reticle must have:

► Patterns for pre-alignment the reticle to the reticle table.

► Reticle alignment marks for aligning the reticle to the wafer.

Location of ASML Mark File

The file/layout is available in .GDS or .CIF format on silicon:/ ~cad/asml/asml_fiducials.

To get started, you will need:

Your layout design scaled to what you expect on the wafer.
File for the ASML fiducials (~cad/asml/asml_fiducials) is also wafer scaled.
A CAD editor such ledit, cadence or xkic3.

How to generate final layout and submit mask request using ledit:

a) First open up both ASML template (asml_fiducial file) and your design layout. Use “cell” command and open asml_fiducial file. Then use “cell” command followed by “instance” to bring up your design layout. Finally merge the two together.

Note: Always open up asml_fiducial first then bring up your layout as an instance, so that positions of the pre- and alignment marks do not change.

b) Check coordinates of the pre- and alignment marks on layout make sure that they are correct at wafer level, as per following table 1.

Pattern	X size (mm)	Y size (mm)	X pos (mm)	Y pos (mm)	Remarks
Pre-alignment 1	0.7	0.7	13.55	13.9	Mark on right top side of the wafer
Pre-alignment 2	0.7	0.7	-13.55	13.9	Mark on left top side of the wafer
Border needed for clear field mask only	1.7	1.7			With chrome border
Reticle alignment mark 1	0.3668	0.3668	12.6	0	Mark on right center side of the wafer
Reticle alignment mark 2	0.3668	0.3668	-12.6	0	Mark on left center side of the wafer
Border needed for clear field mask only	1.6	1.6			With chrome border

Table 1

c) Save the merged file as .gds and use gds2tap shell script to get the .tap/.tix files. Preferred workstations for example: Gold, Boron, Iridium, and Radium. CAD jobs (i.e., gds2tap) on Silicon or Argon will be killed immediately!

usage:

gds2tap may be called with the following arguments:

gdsfile

Name of the GDS layout file to be converted.

layerlist

List of GDS layer numbers to be converted, separated with commas or spaces

reduction

Optical reduction expected: 5X

top_structure

Name of the top-most feature to be converted

centered

Use no, since you want to keep your settings

Warning: Using centering option will individually centered layer and change the coordinates of your alignment marks.

units

Microns

Refer to the following documents for further information: UNIX manual page for gds2tap (man gds2tap)

Note: The standard requirement for the fiducials is the chrome border, so to get a dark filed mask is very strait forehead, while generating the clear field mask is required more concerns.

Clear field mask making for ASML with the Pattern Generator

Some considerations:

► Dark field mask means: the patterns will be clear (exposed), background area remains chrome (i.e. contact holes)

► Clear field mask: the data, pattern should be unexposed (chrome) and the background will be clear.

► The reticle alignment marks must have a dark (chrome) background with an area of at least 8.0 × 8.0 mm² centered around on each alignment mark.

► There is no option for utilizing emulsion or iron mask for ASML (all chrome mask).

► Generating a clear field chrome mask requires higher flash numbers than the dark filed version of the same layout.

► Maximum density allowed in case of the GCA Pattern Generator: 150 thousand flashes.

► If you need a clear field version of your design (i.e. metal layer) on a chrome mask, you should change the original polarity of the design. Changing the polarity can be carried out easily with some of the CAD software (by generating a new layer), also by ciftomann conversion utility (Section 2.3.2).

Hints: The generated mann file can be saved as .cif file, which can be opened with L-edit. This way the inverted design and the marks for ASML could be merged together and you can submit as one layer (even if the polarity was different originally).

Be aware: Inverting a layer could result some very small features on your design. It will be dropped by the Pattern Generator if smaller than 2 micron on the mask.

Appendix D

Mask Making for GCAWS6

Purpose

This appendix gives some guidelines for GCAWS6 mask generation from your layout design.

Applicable Document

Chapter 4.11 (GCAWS6 6” Wafer Stepper)

Reticle Specifications

Nominal edge length: 5-inch square

Reticle thickness: 0.09 inch

Recommended glass type: Soda lime glass (STD)

Pattern Specifications

The maximum field size is:

Wafer Level

Reticle Level

Maximum Y

Maximum X

15.6 mm

78.0 mm

The recommended field size is:

Wafer Level

Reticle Level

Maximum Y

Maximum X

10.06 mm

50.3 mm

For correct operation, in addition to the device layout image, the reticle must have:

· Reticle alignment marks for aligning the reticle to the reticle table (RMS alignment windows).

· Global alignment targets for aligning the reticle to the wafer.

· Optional Micro DFAS alignment targets for finer alignment.

Location of GCAWS6 template file

The template file for GCAWS6 is available in .GDS or .CIF formats on silicon: ~cad/gcaws6/

The template contains the following objects:

- Border for the recommended die size

- RMS alignment windows (for reticle alignment)

- Global alignment targets (in the center of the die)

- Micro DEFAS alignment marks (in the corners of the die)

Positions and locations of the alignment marks

Pattern	X size (mm)	Y size (mm)	X pos (mm)	Y pos (mm)	Remarks
RMS alignment window #1 (Reticle alignment window #1)	0.4	0.4	-2.2	11.5	Upper left part of the pattern
RMS alignment window #2 (Reticle alignment window #2)	0.4	0.4	0.0	-11.5	Lower center part of the pattern
Standard die size*	10.06	10.06	0.0	0.0	With chrome border
Global alignment mark	0.132	0.132	0.0	0.0	Mark can be moved anywhere within the image field. Positive and negative phase targets may be used.
Micro DEFAS marks (Fine alignment marks)	0.117	0.107			1, 2 or 4 marks can be used per field in combination with the global alignment target to optimize the alignment. Mark(s) can be placed anywhere within the image field. For the best results, however, the 2 targets should be placed across from each other in the field as close to the X-axis as possible; 4 marks per field should be located in the 4 corners, diagonally across from each other. Both positive and negative phase marks can be used.

*Note: The die size can freely be chosen between 0 to 15.6 mm. However, if your die size differs from the standard one, a new job is needed to be created on the GCAWS6 machine.