New features:

Bugfixes

Documentation

Internal

New features:

• Sounddevice inputs now allow for user-settable sample rates and precision types

• Block-wise frequency domain processing

  • Introduces new acoular.base module with abstract base classes Generator, SpectraGenerator, InOut, TimeOut, SpectraOut

  • Introduces new module acoular.fprocess including new frequency domain processing classes RFFT, IRFFT, AutoPowerSpectra and CrossPowerSpectra

  • Introduces new module acoular.process including the following classes for general purpose (domain invariant) processing: Cache, Average, SampleSplitter

  • Rename MaskedTimeInOut to MaskedTimeOut

  • Deprecates: TimeInOut, FFTSpectra, TimeAverage, MaskedTimeInOut, TimeCache

  • adds unittests test_process.py, test_fprocess.py

  • adds documentation example example_fft.py

    • allow more platforms to build the docs files including Linux, MacOS, and Windows

Bugfixes

• flush file in TimeCache to prevent data loss / corruption

• fixes use of already deprecated traits

• fixes import order to avoid the deactivation of parallelization of numba

• fixes numpy.matlib deprecation warning

Documentation

• updates literature and adds sphinxcontrib-bibtex extension to support bibtex citation in the documentation.

• cleans-up and structures the conf.py file and fixes missing acoular logo in documentation

• fixes missing and wrong docstrings for result methods in acoular.tbeamform

• improves the explanation of the installation of dependencies

• add explanation on how to test specific Python version only

• test documentation build on all latest platforms

Internal

• refactoring of classes in acoular.tbeamform (moves buffer logic away from Beamformer classes)

  • adds new SamplesBuffer class located in utils

• replaces the markdown-link-check by lychee

  • faster CI (written in RUST)

  • allows more files to be checked, including the .rst files in the documentation

  • adds a cron job that runs daily

• sets final version for several deprecated traits. (Will be removed in version 25.01)

• use scipy.linalg consistently over numpy.linalg

• drops support for Python 3.8 and 3.9

• enable Numpy version > 2.0

• run unittests on the latest MacOS

New features:

• Implement a lazy result array for acoular.fbeamform.BeamformerBase derived classes

• Adds flow establishment length traits

• Updates acoular demo with ASCII map and gets rid of writing file

Bugfixes:

• temporary fix to PyTables - NumPy 2.0 incompatibility problem

• Fixes acoular.fbeamform.BeamformerGridlessOrth shgo params and eva_list initial value bug

• makes acoular.sdinput.SoundDeviceSamplesGenerator always importable

Internal:

• use hatch for publishing the Acoular package

• add issue templates

• adds Conda CI

• CI for TestPyPI and PyPI

• remove plot_example.py

• add autolabel rule for fix and linting

• fix linting rules

• add zenodo release to CITAITON.cff

• adds support for Python version 3.12 on Linux, MacOS, Windows

• drops official support for Python version 3.7

• provides new tools to import data in AIAA array benchmark format

• Bugfixes:

  • changes to UMA-16 microphone array arrangement

• Internal:

  • formatting and linting with ruff

  • introduce hatch

  • measure test coverage

  • replace zenodo.json by CITATION.cff

  • Bugfixes CI

  • update LICENSE

  • adds code of conduct

  • allow workflow dispatch for testing on different branches using GitHub

  • improve documentation

  • refine package structure

    • move test directory outside of the source directory

    • remove outdated submodules fileimport and nidaqimport

    • introduce new submodule acoular/tools

• Improve test coverage for BeamformerCMF

• Changes to BeamformerSODIX:

  • correction of wrong cost-function

  • speedups through the use of numpy.einsum_path together with numpy.einsum

  • changed start value pgtol for the optimization with scipy.optimize.fmin_l_bfgs_b solver

• Bugfixes:

  • fixes unrecognized sector arguments in MetricEvaluator

  • handles version-dependent default values for normalize attribute in sklearn solvers (relevant for BeamformerCMF )

  • fixes bug in BeamformerOrth: assigned strongest source at grid index 0 when instead of eva_list the trait n was given

  • fixes broken SpatialInterpolator

  • minor bugfix for single microphone transfer functions calculated with SteeringVector

  • fixes broken NNLS method in BeamformerCMF (wrong keyword argument normalize)

• Internal:

  • new GitHub workflow for CI of the documentation

  • added Zenodo metadata file

  • changes to author name in pyproject.toml

• New class MetricEvaluator to evaluate the performance of source mapping methods according to Herold and Sarradj (2017)

• New class PointSourceConvolve to blockwise convolve an arbitrary source signal with a spatial room impulse response

• All filter classes derived from Filter use SOS filters now

• No more version restrictions for scikit-learn

• Speedups for numba jitted functions by enforcing C-contiguous arguments and the efficient use SIMD processor instructions

• BeamformerOrth now reimplements orthogonal deconvolution to be even faster and has a slightly different interface

• Simple benchmark suite to compare the performance of Acoular core routines on different computers

• Some internal rework in grid and sector classes

• Test coverage is improved

• Bugfixes:

  • minor bugfix for convective amplification in BeamformerCleantTraj and BeamformerCleantTrajSq

  • bugfix in some attributes of ImportGrid

• Supports Python 3.7, 3.8, 3.9, 3.10, 3.11 on Linux, MacOS, Windows

• New build system using hatch, purge setuptools

• New base class BeamformerAdaptiveGrid for gridless algorithms

• New class BeamformerGridlessOrth for gridless orthogonal beamforming

• New class RectSector3D

• Improved ray casting implementation for GeneralFlowEnvironment

• Improved handling of spectra calculation:

  • New base class BaseSpectra

  • New class FFTSpectra for time-frequency analysis

  • New class PowerSpectraImport for cross spectral matrix import

• MicGeom now has an aperture trait

• Tests are improved

• Bugfixes:

  • broken numpy.int import

  • one off bug in LineGrid

• New class BeamformerSodix

• New SplitBregman and FISTA solvers in BeamformerCMF

• IDW is now available for virtual rotation

• different steering vector formulations are now available for time beamformers as well

• Speedups:

  • time domain beamformers and CleanT deconvolution now share a common core codebase and all do blockwise processing

• Bugfixes:

  • broken digest in RectGrid3D repaired

  • BeamformerCleant and derived classes now never miss samples

• Supports Python 3.6, 3.7, 3.8, 3.9 on Linux, MacOS, Windows

• New class FiltWNoiseGenerator

• New classes SphericalHarmonicSource, Linesource, MovingPointSourceDipole, MovingLineSource

• New class TimeConvolve

• Speedups:

  • CSM works now in parallel and is faster

  • frequency domain beamformers are abaout 30% faster

  • time domain beamformers and CLEAN-T is now about 10 x faster

• Unittests for all major features, much better coverage

• Bugfixes:

  • integration sectors

  • new grid classes from 20.10 are now imported into module namespace

• Supports Python 3.6, 3.7, 3.8

• New base classes for time signal processing: Filter, FilterBank

  • New filter classes: TimeExpAverage, FiltFreqWeight, OctaveFilterBank

  • Demo script is now part of module (see Getting Started for usage)

• New class for processing sound card input: SoundDeviceSamplesGenerator

• New class for cumulative averaging: TimeCumAverage

• New grid classes ImportGrid, LineGrid, MergeGrid

• New versatile integration sector classes RectSector, CircSector, PolySector, ConvexSector, MultiSector

• New class for mixing multiple channels ChannelMixer

• New class for using arbitrary sources as signal: GenericSignalGenerator

• New time-domain beamforming classes for CLEAN-T method: BeamformerCleant, BeamformerCleantSq, BeamformerCleantTraj, BeamformerCleantSqTraj

• Adds possibility to store/load metadata to/from HDF files with WriteH5/TimeSamples classes

• New submodule tools containing several helper functions, e.g. for data aggregation

• WriteWAV: user can optionally set name of output wav file

• Bugfix: PowerSpectra.freq_range / .ind_low / .ind_high are now correctly updated depending on changed attributes

• Bugfix: SineGenerator amplitude is now set via amplitude attribute

• Some minor fixes

• Adds gradient-based solver for BeamformerCMF (L_BFGS_B)

• Adds possibilty to take into account convective amplification of moving sources in BeamformerTimeTraj

• TraitsUI support of the classes has to be activated via config

• Bugfix: use left-orientated coordinate system in rotating flow environment

• Bugfix: correct wrong angle shift in spline interpolation

• Removes several Python 2.7-specific implementations

• Adds new classes for handling rotating data, including detection of trigger signals and interpolation of sensor data for virtual array emulation (Trigger, AngleTracker, SpatialInterpolator, SpatialInterpolatorRotation, SpatialInterpolatorConstantRotation)

• Introduces new SampleSplitter class, which allows distribution of data streams

• Adds new (global) caching options for more flexible cache file handling (e.g. never cache results, always cache, cache read-only). See config for information on how to use this.

• User can choose whether to use h5py or pytables package for handling hdf files. See config for information on how to use this.

• Change: BeamformerGIB behaviour (not calculating sources with eigenvalue of zero)

• Bugfix: BeamformerTime interpolation

• Bugfix: Integer division in PNoiseGenerator

• Test suite and CI updates

• Supports Python 3.5, 3.6, 3.7

• This will be the last version to officially support Python 2.7

• Cache and data directories are now always created in current directory (Linux and Windows)

• Bugfix: Steering vector backwards compatibility

• Bugfix: Ambiguous caching (changes in an object’s class name as well as changes in the data file content are now monitored for caching)

• PowerSpectra: Frequency range to be evaluated can be set directly

• Some code clean-up

• Renamed examples

• Adds support for Python 3.7

• Introduces new SteeringVector class (see Getting Started and ../examples/index for usage). With this, some of the Beamformer and PointSource traits are deprecated and should no longer be used. While the current version is intended to be fully compatible with older scripts, deprecation warnings will be raised if necessary.

• Introduces optional use of reference distance for SPL evaluation (current default: reference position at (x,y,z)=(0,0,0) )

• Introduces some basic Unit tests to evaluate the beamformer results

• Bugfix: CLEAN algorithm now uses correct PSFs

• some minor Bugfixes

• Added new DAMAS solving strategies (BeamformerDamasPlus)

• Added Generalized Inverse Beamforming

• Floating point precision of CSM, PSF and beamformer customizable (default: float64) – affects cache file size

• PowerSpectra class now includes EigSpectra functionality (EigSpectra still callable for backwards compatibility)

• Inverse methods: unit of sound pressure for internal calculation customizable (default: nPa) for better numeric stability with sklearn solvers. Still returns all values in Pa.

• Bugfix: BeamformerFunctional works now with steering vector formulation II (inverse) and III (true level) which produced incorrect results in the past.

• Bugfix: BeamformerFunctional can only be called when the diagonal of the CSM is included

• Bugfix: Corrected calculation of PSF for steering vector formulation IV

• Bugfix: Behaviour of normalizing PSF at assumed source location (psf=1) is removed

• Added support for Python 3.4, 3.5 and 3.6

• Implementation of fast/parallelized code now with Numba (instead of C++ and SciPy.weave)

• cross spectral matrix (CSM) orientation changed (was transposed in earlier versions). Please do not use the cache files from earlier versions in version 17.11!