Complete Integrated Circuit (IC) Layout Analysis from GDSII Design File to Parasitics Extraction
This layout analyzer is written in C++ as part of a wider API for the electromagnetic design validation of VLSI IC, package, and board designs. The code design follows a philosophy intended to have the most extensibility and highest level of automation possible for increasingly complex integrated circuits, packages, and boards. Design information from the Graphic Database Stream II (GDSII) file is parsed and stored alongside simulation input information.
A nonstandard mode of operation allows an interconnect modeling platform (IMP) file to be translated to a GDSII file design. Another nonstandard control mode translates a GDSII file design into a Planar Straight Line Graph which may be used as input to Delaunay triangulation software.
Once a design is loaded, the information can be written to another GDSII file or analyzed with a full-wave simulation to extract parasitics between user-specified ports. Parasitics are preferably reported through a Xyce subcircuit file (SPICE derivative from Sandia National Laboratories), though a Standard Parasitic Exchange Format (SPEF) file may be output alternatively. Please carefully read this README.md file for an overview, dependencies, and usage instructions.
| Packages / Library | Languages | Headers | Description |
|---|---|---|---|
| C Standard Library | C | CCtype, Cerrno, Cmath, Cstdio, Cstdlib, Cstring, Ctime | Utilities for testing character categorization, C language error numbers, basic arithmetic operations, C language input/output, manipulating the char* data type, and dates and timers |
| C++ Standard Library: Algorithms | C++ | Algorithm | Useful algorithms including sorting |
| C++ Standard Library: Containers | C++ | Queue, Set, Stack, Unordered_map, Unordered_set, Vector | Utilities for handling queues and priority queues, sets and multisets, LIFO stacks, associative data types, dictionary collection types, and collection data types from existing C++ data types |
| C++ Standard Library: I/O | C++ | FStream, IOStream | Plain text file input/output and command line input/output |
| C++ Standard Library: Numerics | C++ | Complex | Complex floating point number arithmetic |
| C++ Standard Library: Strings | C++ | String | Utilities for easy string handling |
| C++ Standard Library: Utilities | C++ | Utility | General C++ utilities such as pairs and swapping |
| Eigen | C++ | Sparse | API for sparse matrix storage |
| HYPRE | Fortran/C/C++ | HYPRE, HYPRE_krylov, HYPRE_parcsr_ls | LLNL library of preconditioners and multigrid solvers |
| Limbo | C/C++ | GdsReader, GdsWriter | API for GDSII file parsing and writing |
| MakeUtils | Makefile | n/a | Makefile utilities that help find build dependencies |
| Math Kernel Library | Fortran/C/C++ | MKL, Mkl_spblas | Intel libraries for parallel mathematics include BLAS, LAPACK, and Pardiso |
| Open Message Passing Interface | Fortran/C | n/a | Open MPI compilation and runtime tools for parallel computing |
| Parser-SPEF | C++ | parser-spef | API for SPEF file processing |
| Preprocessor | C/C++ | n/a | Inclusion, definitions, and macros such as assertion with message |
Note that modern compilers that support the following language standards are needed to make the package dependencies: GNU Fortran (superset of F95) for Fortran, C99 for C language, and C++17 for C++.
| Structure / Class Name | Header | Description |
|---|---|---|
| AsciiDataBase | limboint | Full GDSII design being read/constructed with file name, metadata, units, data of all geometric cells, and conductor information |
| GeoCell | limboint | GDSII geometric cell with name, metadata, and all GDSII elements |
| boundary | limboint | GDSII element for a polygonal boundary |
| path | limboint | GDSII element for a path |
| node | limboint | GDSII element for an electrical node |
| box | limboint | GDSII element for a box outline |
| textbox | limboint | GDSII element for a text box |
| sref | limboint | GDSII element for a structure reference |
| aref | limboint | GDSII element for an array reference |
| strans | limboint | Linear transformation applied to some GDSII elements |
| pslg | limboint | Lists of vertices, segment connections, and regions for a PSLG |
| SolverDataBase | solnoutclass | Design name, output files, information other than layout, and all results of the simulation necessary for parameter extraction |
| SimSettings | solnoutclass | Simulation settings including units, design limits, and frequency sweep parameters |
| Layer | solnoutclass | Layer information from physical stack-up including name, GDSII layer number, z-coordinates, and material properties |
| Waveforms | solnoutclass | Needed information for time-domain plots (unimplemented) |
| Parasitics | solnoutclass | Collection of ports and sparse matrices for admittance parameters |
| Port | solnoutclass | Port information including name, reference direction, supply and return coordinates, and containing GDSII layer |
| Aperture | solnoutclass | Standard template details for a Gerber photoplotter aperture |
| fdtdMesh | fdtd | Master class for all discretization, conductor, and solver data |
| fdtdPatch | fdtd | Information of a single discretized patch |
| fdtdBound | fdtd | Information of a single conductor or dielectric boundary region |
| fdtdCdt | fdtd | Solver information common to all conductors |
| fdtdOneCondct | fdtd | Solver information regarding discretization and excitation of a single conductor region |
| fdtdPort | fdtd | Solver representation of a port |
All software is under active development without any notices regarding the timing of nature of updates. Contributions are welcome through forking and pull requests. Please direct all questions, bug reports, and issues relating to this repository to the primary maintainer:
- Dan Jiao, Synopsys Professor of ECE, Elmore Family School of ECE, Purdue University
Developers:
- Michael Hayashi, Former Graduate Research Assistant, Elmore Family School of ECE, Purdue University
- Dr. Li Xue, Former Graduate Research Assistant, Elmore Family School of ECE, Purdue University
- Dr. Shuzhan Sun, Former Graduate Research Assistant, Elmore Family School of ECE, Purdue University
Many thanks are owed to the project sponsors for making the development of this software possible:
 |
 |
|---|
Follow the instructions given in INSTALL.md to install this software and run it from the command line. Users from Purdue University should read purdue_install.md instead for specific steps unique to their environment.
Every GDSII file needs to have a simulation input file with the .sim_input extension created for it. The custom syntax of the simulation input file is given at the bottom of this section. It is important that the information in the blocks underneath the headers in all capital letters appear in the order given. Comments start with '#' and continue for the rest of the line. Comments may appear almost anywhere in the file so long as in-line comments are preceded by a space. The following are placeholders, angle brackets and words in between, which must populated depending on the design:
- <six design extents>: The six coordinates for the smallest rectangular prism enclosing the design with implied units in the order of xmin (the smallest x-coordinate in the design), xmax, ymin, ymax, zmin, and zmax (the largest z-coordinate in the design).
- A lowercase 'e' may be used for scientific notation.
- For example,
30.00 +149. -47.0 +120. 0.00 +7.53.
- <lengthUnit>: The unit by which all given and reported lengths (e.g., total size of design, layer coordinates and heights, and port coordinates) must be multiplied in order to convert to meters. A lowercase 'e' may be used for scientific notation. For example,
1e-6. - <freqUnit>: The unit by which all given and reported temporal frequencies must be multiplied in order to convert to hertz. A lowercase 'e' may be used for scientific notation. For example,
1.0e+6. - <freqStart>: The smallest frequency in the frequency sweep with implied units. For example,
1000.0. - <freqEnd>: The largest frequency in the frequency sweep with implied units. For example,
3000.0. - <nfreq>: The total number of frequencies to include in the sweep. A value of
1means that <freqEnd> must equal <freqStart>. A value of2means that only <freqStart> and <freqEnd> are included. Any larger integer requires interpolation. - <freqScale>: Integer representing the scale for frequency interpolation. The recommended value is
0for logarithmic spacing. The value1is for linear spacing. - <numLayer>: The total number of layers in this design. The integer here must match the number of active <layer entry> lines following this line in the file.
- <layer entry>: An entire line representing a single layer.
- The four pieces of information that must be included in order are the layer name at the start of the line, the z-coordinate of the bottom of the layer (designated with 'z = ' before a length with implied units), the height of the layer (designated with 'h = ' before a length with implied units), and the relative permittivity of the dielectric making up that insulating parts of that layer (designated with 'e = ' before a unitless floating-point number greater than or equal to unity).
- A lowercase 'e' may be used for scientific notation.
- In order to maintain correspondence with the layer numbers used in a GDSII file, the layer names must be a plain integer (recommended) or be an alphanumeric string with a capital 'M' before the layer number.
- Optional specifications for each layer entry include the conductivity of conductors in that layer (designated with 'sigma = ' before a nonnegative floating-point number with units of S/m) and the dielectric loss tangent (designated with 'TanD = ' before a nonnegative, unitless floating-point number).
- It is recommended to add one layer below the bottommost GDSII layer and a second layer above the topmost GDSII layer to serve as planes of perfect electrical conductor (PEC) representing how the design would behave in a testing environment.
- For example,
ILDM2 z = 1.316 h = 0.12 e = 8.0.
- <numPorts>: The total number of ports in this design. The integer here must match the number of active <port entry> lines following this line in the file.
- <port entry>: An entire line representing a single port excitation.
- The eight pieces of information that must be included in order are the port name, the x-coordinate of the supply point of the port, the y-coordinate of the supply point of the port, the z-coordinate of the supply point of the port, the x-coordinate of the return point of the port, the y-coordinate of the return point of the port, the z-coordinate of the return point of the port, and the directionality of the port.
- Port entries sharing the same name will be converted to a single port with multiple excitations.
- A lowercase 'e' may be used for scientific notation.
- The directionality of the port is either '+1' for input ports, '-1' for output ports, or '0' for bidirectional ports or ports with uncertain power flow.
- A port should be added for each input/output pin of the device. Additional ports are needed for every transistor or other active device in the design. For printed circuit boards (PCBs), at least one port is needed for each component on the populated layout.
- For example,
port1 +146. -16.0 4.53 +146. +6.00 4.53 -1.
TOTAL SIZE
<six design extents>
lengthUnit = <lengthUnit>
FREQUENCY
freqUnit = <freqUnit>
freqStart = <freqStart>
freqEnd = <freqEnd>
# Block-interrupting comment
nfreq = <nfreq>
freqScale = <freqScale>
DIELECTRIC STACK
numStack = <numLayer>
<layer entry> # In-line comment
<...>
<layer entry> # In-line comment
# Post-block comment
PORT
numPorts = <numPorts>
<port entry> # In-line comment
<...>
<port entry> # In-line comment
# Post-block comment
- Example Files
- Purdue On-Chip Electromagnetics Laboratory for singleStrip.imp (single stripline)
- Yuan Ze University Design Automation library for inv.gds (inverter logic gate), nand2.gds (2-input NAND gate), and xor.gds (XOR gate)
- NanGate Open Cell Library FreePDK45 for SDFFRS_X2.gds (scan D flip-flop with reset/set)
- Peter Monta and the Tuva Design Intel 4004 Anniversary Project for 4004.gds (Intel 4004 processor from 1971)
- Open Compute Project for original Gerber files for Facebook Wedge 100 that led to WEDGE100_REV01.gds (100G Ethernet Switch board)
- Prof. Dan Jiao of the Purdue On-Chip Electromagnetics Laboratory for her guidance of the project
- The staff of the Defense Advanced Research Projects Agency (DARPA) Microsystems Technology Office (MTO) for their sponsorship, feedback, and support of this endeavor
- Prof. C.J. Richard Shi and the students of the University of Washington RAIL research group for supplying designs and validating solver results against commercial tools
- The other contributors of the Intelligent Design of Electronic Assets (IDEA) team and the Posh Open Source Hardware (POSH) team within the DARPA Electronics Resurgence Initiative (ERI)
GNU Public License v2.0 (see LICENSE)