psf-parser is a lightweight, dependency-free Python parser for Cadence's proprietary PSF (Parameter Storage Format) files.
It supports both the ASCII (psfascii) and binary (psfbin) formats.
Note: While Spectre uses psfbin as the default output format for most simulations, transient simulations typically use the psfxl format, which this parser does not support. To avoid issues, set the output format explicitly to psfbin.
Install psf-parser directly via pip:
pip install psf-parserThe psf-parser package provides a simple API for parsing and accessing PSF file contents.
It automatically detects the file format (ASCII or binary), but you can also specify it manually.
from psf_parser import PsfFile
psf = PsfFile("path/to/psf")
print(psf.header, psf.sweeps, psf.traces, psf.values)If you prefer lower-level access to the raw registry data, you can work directly with the parser classes:
from psf_parser import PsfParser
parser = PsfParser("path/to/ascii/psf", format="ascii").parse()
print(parser.header)
print(parser.registry.types)Notes:
PsfParseracts as a factory and dispatches to the correct format-specific parser.PsfAsciiParserandPsfBinParserhandle the ASCII and binary formats respectively (for internal or advanced usage).
This project is licensed under the MIT License.
This project was made possible by the excellent prior work of:
psf_utils— a psfascii parser built with PLY.libpsf— the original reverse-engineering project for the binary PSF format.
This project is still in its early stages. Since Cadence has never published a specification for the PSF format, the parser is built entirely by reverse-engineering test files and observing patterns.
Currently, the parser is verified against output from:
- Oppoint analyses
- DC sweeps
- AC simulations
- Transient simulations
- Monte Carlo
- Nested sweeps
- Various info analysis statements
If you encounter a PSF or PSFBIN file that fails to parse or produces incorrect output, please contribute by:
- Adding the problematic file to the test suite.
- Creating a minimal test case that demonstrates the issue.
This way, we can ensure continued support for more variations and edge cases over time. PRs, bug reports, and examples are all welcome!