PSF Utilities — Read Spectre Data Files ======================================= .. image:: https://github.com/KenKundert/psf_utils/actions/workflows/build.yaml/badge.svg :target: https://github.com/KenKundert/psf_utils/actions/workflows/build.yaml .. image:: https://pepy.tech/badge/psf_utils/month :target: https://pepy.tech/project/psf_utils .. image:: https://img.shields.io/pypi/v/psf_utils.svg :target: https://pypi.python.org/pypi/psf_utils .. image:: https://img.shields.io/pypi/pyversions/psf_utils.svg :target: https://pypi.python.org/pypi/psf_utils/ :Author: Ken Kundert :Version: 1.8rc1 :Released: 2024-01-15 What? ----- *psf_utils* is a library that allows you to read data from a Spectre PSF ASCII file. Spectre is a commercial circuit simulator produced by Cadence Design Systems. PSF files contain signals generated by Spectre. This package also contains two programs that are useful in their own right, but also act as demonstrators as to how to use the library. They are *list-psf* and *show-psf*. The first lists the available signals in a file, and the other displays them. Accessing the Results --------------------- You can use the PSF class to read ASCII Parameter Storage Format files. When instantiating the class you pass in the path to the file and then the resulting PSF object contains the signals. For example, the following lists the signals present in a ASCII PSF file:: from psf_utils import PSF from inform import Error, display kinds = { 'float double': 'real', 'float complex': 'complex', } try: psf = PSF('adc.raw/tran.tran') for signal in psf.all_signals(): kind = signal.type.kind kind = kinds.get(kind, kind) display(f'{signal.name:<15} {signal.units:<12} {kind}') except Error as e: e.terminate() This example plots the output signal:: from psf_utils import PSF from inform import Error, display import matplotlib.pyplot as plt try: psf = PSF('adc.raw/tran.tran') sweep = psf.get_sweep() out = psf.get_signal('out') figure = plt.figure() axes = figure.add_subplot(1,1,1) axes.plot(sweep.abscissa, out.ordinate, linewidth=2, label=out.name) axes.set_title('ADC Output') axes.set_xlabel(f'{sweep.name} ({PSF.units_to_unicode(sweep.units)})') axes.set_ylabel(f'{out.name} ({PSF.units_to_unicode(out.units)})') plt.show() except Error as e: e.terminate() *abscissa* and *ordinate* are NumPy arrays. As such, you can perform computation with them:: out = out_p.ordinate - out_n.ordinate from numpy import sin sine = sin(sweep.abscissa) Reading large ASCII data files is slow, so *psf_utils* reads the PSF file once, then pickles the data and writes it to disk. On subsequent runs the pickled data is used if the pickle file is newer that the corresponding PSF file. Things are a bit different for DC operating point results. In this case, *sweep* is None and the results are scalar `quantities `_:: from psf_utils import PSF, Quantity psf = PSF('opamp.raw/op.dc') with Quantity.prefs(map_sf=Quantity.map_sf_to_greek): for signal in sorted(psf.all_signals(), key=lambda s: s.name): name = f'{signal.access}({signal.name})' print(f'{name:>20} = {signal.ordinate}') Utility Programs ---------------- Two utility programs are installed along with the *psf_utils* library: *list-psf* and *show-psf*. The first lists the signals available from a PSF file, and the second displays them. They both employ caching to speed up access to the data. They also cache the name of the PSF file so that it need not be given every time. *show-psf* also caches its arguments, so if you run it again with no arguments it will simply repeat what it did last time. For example, here is a typical session:: # list the signals in noise PSF file > list-psf -f resistor.raw/pnoise.pnoise Using resistor.raw/pnoise.pnoise. R1:flicker R1:total R2:fn out R1:thermal R2:rn R2:total # list them again, this time in long form > list-psf -l Using resistor.raw/pnoise.pnoise. R1:flicker A²/Hz real (12042 points) R1:thermal A²/Hz real (12042 points) R1:total A²/Hz real (12042 points) R2:fn A²/Hz real (12042 points) R2:rn A²/Hz real (12042 points) R2:total A²/Hz real (12042 points) out A/√Hz real (12042 points) # list only those that match R1:* (assumes nonomatch variable is set in shell) > list-psf -l R1:* Using resistor.raw/pnoise.pnoise. R1:flicker A²/Hz real (12042 points) R1:thermal A²/Hz real (12042 points) R1:total A²/Hz real (12042 points) # show a graph containing signals that start with R1: > show-psf R1:* # show the thermal noise of R1, and then the total noise minus the flicker noise > show-psf R1:thermal R1:total-R1:flicker > show-psf out # show a graph containing only out > show-psf # show out again, exactly as in previous run > show-psf -M out # show out again, this time include point markers > show-psf -P out # show out again, this time only show point markers > show-psf -s out.svg out # send graph of out to svg file # list signals in a PSF file from a DC operating point file: > list-psf -f diffamp.raw/tran.dc Using diffamp.raw/pnoise.pnoise. in_n in_p out_n out_p Vdd Vdd:p # show the DC voltages > show-psf \* V(in_n) = 47.678 µV V(in_p) = 47.623 µV V(out_n) = 876.16 µV V(out_p) = 876.16 µV V(Vdd) = 2.5 V I(Vdd:p) = −10.05 µA # list signals in transient PSF file > list-psf -f diffamp.raw/tran.tran Using diffamp.raw/pnoise.pnoise. in_n in_p out_n out_p Vdd Vdd:p # display differential output and differential input > show-psf out_p-out_n in_p-in_n Converting to PSF ASCII ----------------------- *psf_utils* only supports PSF ASCII files. As an alternative, `libpsf `_ is a Python package that can read both ASCII and binary PSF files. Or, you can use the Cadence *psf* program to convert various types of simulation results files into PSF ASCII format. To use it, simply specify the input and output files:: > psf -i adc.raw/tran.tran -o adc.raw/tran.psfascii > list-psf -f adc.raw/tran.psfascii In this example there is nothing special about the 'psfascii' suffix, it is simply mnemonic. Rather, the output is in ASCII format because the ``-b`` (binary) option is not specified. *psf_utils* does not support SST files, which are generated by AMS Designer, Cadence's mixed-signal simulator. You can recognize SST files in that they come in pairs, and the two files have ``.dsn`` and ``.trn`` suffixes. In this case, Cadence's PSF utility cannot help you either. However, you can use the *simvisdbutil* to convert the data to a CSV file, which would allow you to access the data, though not with *psf_utils*. For example, the following converts all waveforms contained in ldo.trn into CSV data:: simvisdbutil ldo.trn -csv -timeunits s -output ldo.csv Examples -------- `Flicker Noise `_ is a simulation script that shows how to write simple Python scripts that run Spectre and use *psf_utils* to extract and display the desired results. Releases -------- Latest development release '''''''''''''''''''''''''' | Version: 1.8rc1 | Released: 2024-01-15 1.8 (2024-01-15) '''''''''''''''' - Fixed issue with *NaN* 1.7 (2023-12-13) '''''''''''''''' - fixed units on terminal currents - add support for *inf* - add support for *PZ* and *STB* files 1.6 (2022-09-18) '''''''''''''''' - changed ``--no-cache`` command line option to ``--refresh-cache``. 1.5 (2021-11-11) '''''''''''''''' - Renamed *plot-psf* to *show-psf*. - Improved support of DC operating points. - Suppress access function names when printing members of PSF structures. - Correct invalid units produced by Spectre on oppoint files (I, R, C). 1.4 (2021-10-21) '''''''''''''''' - Allow signal names to contain backslashes. 1.3 (2021-03-21) - Improve support for DC operating points. 1.2 (2021-01-07) '''''''''''''''' - Support PSF files that contain DC operating points. - Support PSF files where values are given in a group. 1.1 (2021-01-30) '''''''''''''''' - Allow, but ignore, properties on traces. 1.0 (2020-11-03) '''''''''''''''' - Production release .. hide the pre-production releases **0.7 (2020-09-23)**: - Add ability to show individual points - Improve the cursor values display - Increase precision of both cursor values and axis labels **0.6 (2020-04-16)**: - modest refinements **0.5 (2020-01-08)**: - beta release **0.4 (2019-09-26)**: - Allow glob patterns to be passed to both *list-psf* and *show-psf*. **0.3 (2019-09-25)**: - Fix import errors in *show-psf* command. **0.2 (2019-09-25)**: - Fix dependencies. **0.1 (2019-09-25)**: - Initial version