v0.6.0: Improved transient simulations y[n]

This release deals mainly with the "y[n]" tab for simulating impulse and transient response and a better integration of the FFT pop-up widget.

The internals of fixpoint and floating point simulations have been rewritten as a frame-based simulation. This is the base for allowing to interrupt a long running simulation and to load stimuli from file (both coming soon).

In comparison to the v0.6.0b1 prerelease, PR #206 has been merged, a regression bug has been fixed that prevented locking "f_S". Logger noise has been minimized and some small UI fixes have been applied.

For details see CHANGELOG.md.

v0.6.0: Improved transient simulations y[n]

This release deals mainly with the "y[n]" tab for simulating impulse and transient response and a better integration of the FFT pop-up widget.

The internals of fixpoint and floating point simulations have been rewritten as a frame-based simulation. This is the base for allowing to interrupt a long running simulation and to load stimuli from file (both coming soon).

For details see CHANGELOG.md

Transient simulations y[n]

This release deals mainly with the "y[n]" tab for simulating impulse and transient response and a better integration of the FFT pop-up widget.

The internals of fixpoint and floating point simulations have been rewritten as a frame-based simulation. This is the base for allowing to interrupt a long running simulation (coming soon).

For details see CHANGELOG.md

Second update for v0.5.0 "Lock Frequencies"

There is no release v0.5.2 due an error that occurred during the release process.

Bug fixes

• Fix inaccessible fields "F_C" for manual design in elliptic filter
• requirement.txtnow includes matplotlib 3.3

New features

• Duty cycle of rect pulses now can be varied

• Absolute frequency specifications now can be locked i.e. they dont change when f_S is updated

There is a new binary release for Ubuntu 20.10 that might or might not work with other distros as well - please report!

For details see CHANGELOG.md

First bugfix release for v0.5.0

• Fix some bugs related to spectrum display in impulse response tab

• New stimulus: "sinc"

For details see CHANGELOG.md

Release v0.5.0 "Spectrogram and help function"

Highlights

• Improved functionality of transient response tab (spectrogram plot, complex stimuli, real and imaginary part of spectra, ...)
• Cycle through fine, coarse and no grid
• Help button in plot widgets links to (currently outdated) help pages in ReadTheDocs
• Taskbar icon now works under Windows as well

For details see CHANGELOG.md

Release v0.4.0 "Matplotlib 3.3 and numexpr"

Highlights

• compatibility to matplotlib 3.3
• numexpr for calculations in the textboxes and for providing formula based stimuli in the transient response tab
• license information is now included and presented more clearly
• the binary for windows builds upon the non-mkl versions of numpy / scipy, reducing size dramatically

For details see CHANGELOG.md

First bugfix release for v0.3.0

New Features

• Binary (generated with pyInstaller) available for Windows. This is quite a large file because I haven't managed yet to replace numpy-mkl by a nomkl version (linked against OpenBLAS). The spec file for pyInstaller is part of the repo, so you can try for yourself.
• overlay ideal frequency response in impulse response window and scale it with the number of FFT points for impulse responses. This allows a direct comparison of the fixpoint frequency response (calculated from the transient response) and the ideal response (calculated from the filter coefficients).

Improvements

• limit minimum of pass band frequency of differentiator to prevent crashs
• replace deprecated parameter 'Hz' by 'fs' in remez routines
• improve legend display for h[n] frequency window (scatter plots were missing)
• include license file in distribution

Bug Fixes

• fix path and file names for configuration template files were wrong
• catch exceptions raised in the central fil_save function to avoid crashes caused by badly conditioned filters
• Pop-up FFT windows were constructed with sym = False for filter design and with sym = True for spectral analysis, the correct settings are the other way round. This bug affected only the window display, not the actual calculations.
• The units for coherent gain were wrong in several places

Feature and bugfix release "FFT Window"

New Features

• display pop-up window for showing time and frequency properties of FFT windows for filter design and spectral analysis of transient signals.
• improve parameter entry and tooltipps for windowed FIR filter design
• add more windows and use same widget for filter design and spectral analysis
• new stimuli AM, FM, PM for transient response
• first successful generation of a binary for Windows using pyinstall. "spec" - files are included for own experiments

Improvements

• improve robustness when importing csv Files (still room for improvements)
• suppress "divide by zero in log10" warnings
• reduce number of debug logging messages, especially from matplotlib
• eliminate false logging warnings and errors
• refactor common libraries, all libraries now are within the folder "libs"
• simplify clipboard instantiation
• improve robustness by enforcing inputs <> 0 ('poszero', 'negzero') in input fields
• rename pole/residue export format suffix to *.txt_rpk

Bug Fixes

• only install PyQt5 when it cannot be imported to prevent double installation under conda.
• don't try to import PyQt4 (pyfda isn't compatible anymore) as this gives strange error messages on some systems
• stop installation when matplotlib version 3.1.0 is detected as it isn't compatible
• fix crash with matplotlib version 3.2.1 due resulting from changed colormap list
• fix crash when selecting Fixponpoint Tab for a filter design without fixpoint widget
• fix error when trying to load *.npz files: numpy.load() requires allow_pickle = True since version 1.16.3 for security reasons
• fix error when trying to load pickled files
  [ ERROR] [pyfda.input_widgets.input_files:187] Unexpected error: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()
• fixed buggy export of coefficients in Xilinx and Microsemi format
• Changing the CSV option box parameters from 'file' to 'clipboard' or the other way round in input_coeffs or input_pz would not update the other tab

Release v0.2.1

• Installation choked on systems with default encoding ASCII, now utf-8 format is explicitly required when reading version.py
• pyqt5 now is a requirement so that it is installed automatically. Hopefully this fixes the problem that pyqt4 was tried to install on some systems
• the usage of codecs.open() instead of standard open() somehow messed up reading README.md for PyPI. This also works now.
• an updated meta.yamlis stored in the recipy subdirectory for documentation purposes