In addition to the prominent features mentioned above, the following features have been implemented in the FEKO solver for HyperWorks 2017.
The RL-GO solver includes the following improvements since FEKO 14.0:
| • | Added support for automatic equivalent source generation for aperture sources that are defined close to geometry in RL-GO models. This allows equivalent aperture sources to be used relatively close to geometry structures while keeping run times low. (14.0.431) |
| • | Export ray field components (real and imaginary values) at the ray intersection points for RL-GO and uniform theory of diffraction (UTD) to the *.ray file. (14.0.430) |
| • | Improved memory usage (14.0.410) and a significant reduction in memory for typical uncoupled examples (14.0.420). |
| • | The algorithms used to select the initial ray launching increment are extended to increase accuracy in many cases. The improved automatic selection settings often result in increased performance. (14.0.410) |
| • | Improved RL-GO accuracy for certain model configurations. (14.0.410) |
| • | Improved geometry checks for curvilinear triangles prevent possible numerical errors for curvilinear RL-GO models. (14.0.431) |
| • | The run time is reported for equivalent source transformations. (2017) |
Resolved RL-GO issues:
| • | Allow more subdivisions for the equivalent source generation for RL-GO to improve run time. (2017) |
| • | Check that elements solved using the method of moments (MoM) do not touch elements solved using the RL-GO method. (2017) |
| • | Fixed a bug that caused RL-GO rays to pass through geometry in rare circumstances when using incident fields in multiple directions, multiple frequencies loops or multiple configurations. (2017) |
| • | Resolved an issue that caused FEKO to stop with an undefined error for certain RL-GO models that consume a lot of memory. (14.0.431) |
| • | Resolved an issue that caused RL-GO and PO examples with high loss dielectric sheets or coatings to be computed incorrectly. (14.0.431) |
| • | Resolved an issue that caused very poor memory usage for some RL-GO models that requested far fields. (14.0.431) |
| • | Resolved an issue that caused a slowdown for RL-GO models where a large number of far field request points were used. (14.0.430) |
| • | Corrected a check for the applicability of the RL-GO for media with high losses. (14.0.430) |
| • | Resolved various internal error states for specific RL-GO cases. Resolved RL-GO errors include those given when using certain sources, when using certain sources together with planar multilayer structures, when spherical modes are requested and when the model contains a plane wave source and a receiving antenna. (14.0.421) |
| • | Resolved an issue that could have caused incorrect results to be computed when trying to read a corrupt cache file using RL-GO. (14.0.421) |
| • | Improved RL-GO performance when no near field or far field requests are present in the model. (14.0.421) |
| • | Added support for default incident sources when none are specified for monostatic RCS requests solved with RL-GO. (14.0.421) |
| • | Corrected problems with exporting rays to file when using RL-GO. (14.0.411) |
| • | Resolved various internal error states for specific RL-GO problems. In some models the error state was given when using high accuracy mode and in others when requesting the radar cross section. (14.0.401) |
|
The most prominent extension is the support of OpenMP and MPI parallelisation for the FDTD solver. (2017)
Other extensions to the FDTD solver include:
| • | Support of biaxial anisotropic media for the FDTD. (2017) |
| • | Support of RLC loads for the FDTD method. In the past for the FDTD only a single discrete resistance, inductance or capacitance was supported per port. (2017) |
| • | Support for voltage/other sources together with plane wave excitation in the FDTD solver. (14.0.420) |
Resolved FDTD issues:
| • | Resolved a segmentation violation given when multiple plane wave sources are used for an FDTD model. (2017) |
| • | Resolved an issue that caused incorrect results for transverse H-fields on FDTD boundaries (PEC and PMC). (14.0.421) |
| • | Resolved issues that caused incorrect results close to perfect magnetic (PMC) boundaries when using the FDTD. (14.0.411, 14.0.420) |
| • | Resolved an issue that could have caused an internal FEKO error when using empty ports with the FDTD. (14.0.411) |
| • | Changed the phase reference for far fields calculated using the FDTD to be in agreement with the reference used by other supported methods. (14.0.401) |
| • | Improved FDTD pre-processing speed. (14.0.401) |
| • | Improved stability of small field value variation for the FDTD. (14.0.401) |
| • | Resolved an error state given for certain FDTD port configurations. (14.0.401) |
| • | Improved memory usage for the FDTD when using a GPU. (14.0.411) |
| • | Resolved an internal error state given for certain FDTD models when solved using a GPU. (14.0.401) |
|
The FEM solver changes include:
| • | Support parallel and series RLC loads for the FEM line port. In the past, only complex impedances were supported. (2017) |
| • | Support anisotropic materials for the FEM. This allows non-reciprocal ferrite materials to be modelled with the FEM. (2017) |
| • | Added support to re-use the LU-decomposition (when possible) for models where the FEM/MoM method is used. (14.0.410) |
| • | Improved the pre-processing speed for certain FEM models. (14.0.410) |
| • | Resolved an issue that caused an internal error state for certain FEM models computed in parallel. (14.0.430) |
| • | Resolved an issue that could have caused incorrect results for FEM models with rotationally symmetric modal ports. (14.0.421) |
Resolved FEM issues:
| • | Resolved an issue with incorrectly impressed FEM modal port eigenmodes on the general 3D FEM mesh in certain instances. (2017) |
| • | Only give an error stating that active materials are not supported when these defined materials are used in the model. (14.0.431) |
| • | Resolved an issue that could cause geometry checking to fail for certain FEM models. (14.0.411) |
| • | Resolved an issue that could have led to an internal FEKO error for FEM-only examples. (14.0.410) |
| • | Fixed a bug that prevented a non-perfect electric conductor (non-PEC) from being used in FEM modal ports. (14.0.401) |
|
CMA changes include:
| • | Improved accuracy and stability of higher order characteristic mode computation. (14.0.430) |
| • | Improved the speed of tracking degenerate characteristic modes. (14.0.430) |
| • | Improved the characteristic mode tracking algorithms. (14.0.410) |
|
Cable changes include:
| • | Speed improvements for certain cable modelling scenarios. (14.0.420) |
| • | Cable harnesses now allow loading with 1-port Touchstone files. (14.0.420) |
Resolved cable issues:
| • | Resolved an internal error state that occurred for nearly parallel cable paths. (14.0.431) |
| • | General improvements to multiconductor transmission line (MTL) cable modelling that should result in more accurate results for certain sensitive models. (14.0.401) |
| • | Resolved an issue that caused an error to be raised for certain MoM/MTL models where different paths are physically connected. (14.0.431) |
| • | Resolved an issue that caused an internal error state for certain multiple configuration cable examples. (14.0.430) |
| • | Resolved an issue that caused an internal error state for cable models where component values tend to be close to internal thresholds of the SPICE circuit solver. (14.0.430) |
| • | Improved performance for cables. (14.0.421) |
| • | Significantly reduced solution time when using the default SPICE solver. (14.0.420) |
| • | Fixed a bug that caused and error state for certain cable models during the Delaunay triangulation phase. (14.0.401) |
| • | Fixed a bug that caused in an internal error state for certain cable problems that have intersecting cables. (14.0.401) |
| • | Fixed a bug that might cause incorrect results when using multiple cable configurations with irradiating multiple transmission line solutions. (14.0.401) |
|
| • | Upgrade to Intel MPI Library 5.1 Update 3. (2017) |
| • | Upgrade to MPICH 3.2 for Linux systems. (2017) |
| • | Upgrade to SuperLU 5.2.1 (May 22, 2016), a sparse matrix solver not used as the default solver. (2017) |
| • | Upgrade to SuperLU_DIST 5.1.0 (May 15, 2016), a sparse parallel matrix solver not used as the default solver. (2017) |
| • | Upgrades of various parallel libraries and compilers to improve the general stability of iterative processes when run in parallel. (14.0.401) |
| • | Updated an external library that may improve memory usage for certain parallel sparse methods (FEM, MLFMM). (14.0.401) |
|
| • | Improved memory estimation and reporting for the fast far field calculation. (2017) |
| • | Included the global response surface method (GRSM) and adaptive response surface method (ARSM) as new optimisation methods. These methods internally build a response surface that updates as more sample points are added. The ARSM terminates when one of the convergence criteria is met, while the GRSM continues to test different areas of the design space. (14.0.430) |
| • | Added support to import near field scan data (*.NFS) with irregular grid spacing as source data in FEKO. (14.0.430) |
| • | The domain Green's function method (used for array modelling) now supports dielectric media modelled using surface equivalence. (14.0.430) |
| • | Support for importing electric and/or magnetic near fields on all faces of a cuboid simultaneously from an *.efe/*.hfe file. (14.0.421) |
| • | SAR can now be calculated for non-magnetic materials, even when the model contains magnetic materials. (14.0.421) |
| • | Added the option to allow additional stabilisation for the MLFMM to address models with severe convergence problems. (14.0.420) |
| • | Support was added for near field potential requests when using the MLFMM. (14.0.410) |
| • | Always compute the near field when using the MLFMM even in the presence of an infinite PEC ground plate. (14.0.401) |
| • | Increased the applicable models for which source optimisations are done. (14.0.420) |
| • | Reduced the required separation distance for aperture equivalent sources. (14.0.420) |
| • | Support for curvilinear wires as part of windscreen antennas. (14.0.420) |
| • | Speed improvements when using aperture sources. (14.0.410) |
| • | Support for scalar potentials was added for planar Green’s function solutions. (14.0.410) |
| • | The medium reported for near field requests when using VEP elements is now based on the medium at the observation point. (14.0.410) |
| • | Support was added to ensure that the required amount of hard disk space is available when writing out large intermediate solution files (*.mat/*.lud/*.ngf). (14.0.420) |
| • | Support negative angles when importing far fields using *.ffe files. (14.0.401) |
| • | Support for importing I-DEAS universal format mesh files is added (14.0.410) and extended to read wire radius values. (14.0.411) |
| • | Configuration names are now written to exported field request files (*.ffe, *.efe, *.hfe, *.os, *.ol). (14.0.410) |
|
| • | Fixed a bug that caused the LU-decomposition to fail for certain MoM type problems where many frequencies are requested. (2017) |
| • | Fixed a race condition that sometimes caused an error when writing UTD rays to disk for some parallel jobs. (2017) |
| • | Fixed a bug that incorrectly extended numerical Green's function (NGF) files when changes are made to the model or solution parameters such as frequency. (2017) |
| • | Resolved an issue that could have caused incorrect results to be calculated when using the numerical Green's function (NGF) for certain models. (14.0.411) |
| • | Fixed a bug that caused an internal error state when using the NGF together with the FEM in parallel. (14.0.401) |
| • | The restriction that the relative permeability and magnetic loss tangent should stay constant for thin isotropic dielectric sheets is removed for the high frequency techniques (RL-GO, PO and UTD). (14.0.431) |
| • | Resolved an issue that caused an error when importing a stored matrix file when using multiple configurations in a parallel execution environment. (14.0.431) |
| • | Resolved an issue that caused an internal error state for certain MLFMM models that use the SPAI preconditioner in parallel. (14.0.431) |
| • | Fixed a bug that caused an internal error state when using MLFMM together with certain near field sources. (14.0.401) |
| • | Improved error handling for certain parallel sparse linear reordering algorithms (used by MLFMM) when the model is of a multi-scale nature. (14.0.401) |
| • | Consistently calculate losses at wire-triangle junctions to ensure the reported radiation efficiency is always positive. (14.0.431) |
| • | Resolved an issue with loads specified by node position and spherical mode receiving antennas where scaling was not as expected when using an SF card. (14.0.431) |
| • | Resolved an issue where incorrect reflections were calculated for magnetic thin dielectric sheets and coatings when solved with physical optics (PO), UTD, or RL-GO. This involves models containing sheets or coatings where any layer has a relative permeability larger than 1. MoM and MLFMM solutions for these models are not affected by this bug and remain correct. (14.0.430) |
| • | Removed the option to calculate radar cross section (RCS) in a different reference frame than the incident source. (14.0.430) |
| • | Improved handling of certain complex spherical mode sources that are close to geometry. (14.0.430) |
| • | Resolved an issue that prevented certain examples from being solved in parallel when using MPICH (not the default) with the Windows operating system. (14.0.430) |
| • | Resolved an issue that caused an internal error state for certain examples that use multiple aperture sources. (14.0.430) |
| • | Resolved an issue that caused an internal error state for double negative materials (DNG). (14.0.430) |
| • | Resolved an issue where an error was raised for certain edge port S-parameter models. (14.0.430) |
| • | Improved the accuracy of windscreen geometry checking. (14.0.430) |
| • | Increased the number of digits displayed for simulation run time. The times for longer runs are now displayed, instead of the asterisks that are printed out in case of an overflow. (14.0.430) |
| • | Added a check to disallow MoM wire connections to large element physical optics (LE-PO) triangles. (14.0.430) |
| • | Added a check to disallow wire segments from being located inside a periodic boundary. (14.0.430) |
| • | Resolved an issue that caused runfeko to hang if the license file is not available. (2017) |
| • | It is ensured in RUNFEKO that parallel FEKO runs on Linux internally use the bash shell instead of another default shell the user might have set up. (14.0.430) |
| • | Resolved an issue where importing material parameters into CADFEKO using an XML-file resulted in an error. (14.0.421) |
| • | Resolved an issue that caused an error when importing a stored matrix file using multiple configurations in a parallel execution environment. (14.0.421) |
| • | Improved accuracy for near field calculations using PO. (14.0.421) |
| • | Improved cross-coupling accuracy for wires above infinite high conductivity layers. (14.0.421) |
| • | Resolved an issue that could have caused incorrect results for projects with multiple configurations when using network port sources. (14.0.421) |
| • | Improved parallel scaling for far field calculation using UTD projects. (14.0.421) |
| • | Inform users when current requests are not supported and that no currents will be exported. (14.0.421) |
| • | Improved far field parallelisation for certain model configurations. (14.0.420) |
| • | Resolved an issue that could potentially cause an error state for certain PO models. (14.0.411) |
| • | Resolved an issue that could cause incorrect results when using higher order basis functions (HOBF) in conjunction with the MoM when the model is run out-of-core. (14.0.411) |
| • | Resolved an issue that could cause inaccurate results when using the MoM in a model containing curvilinear as well as flat elements. (14.0.411) |
| • | Resolved an issue that prevented FEKO from running using MPI ABI compatibility on certain native MPI implementations. (14.0.411) |
| • | Improved the power loss calculation for periodic boundary condition (PBC) structures. (14.0.411) |
| • | Resolved an issue where the stored matrix was not read in certain circumstances. (14.0.411) |
| • | Resolved an issue that could have caused an internal FEKO error for certain current requests. (14.0.411) |
| • | Resolved an issue that could have caused incorrectly orientated far fields and request combinations. (14.0.411) |
| • | Improved the handling of components using spherical modes (for example receiving antennas) to allow closer distances between objects. (14.0.411) |
| • | Resolved an issue that the names of near field objects could be imported incorrectly from *.nfs file. (14.0.410) |
| • | Resolved an issue that could have led to incorrect data when requesting the scalar potential gradient. (14.0.410) |
| • | Resolved an issue that could have led to incorrect data for the magnetic frill excitation when power scaling is applied. (14.0.410) |
| • | Resolved an issue that could have led to incorrect data when using finite arrays in conjunction with waveguide ports or microstrip ports in specific model configurations. (14.0.410) |
| • | Resolved an issue that could have led to high memory usage and an internal FEKO error for spherical mode examples. (14.0.410) |
| • | Improved checks for source optimisation in the vicinity of geometry. (14.0.410) |
| • | Improved curvilinear element support. (14.0.410) |
| • | Resolved an issue with PREFEKO that could have led to an error state when importing Sigrity near field data (NFD) files. (14.0.410) |
| • | Fixed a bug that resulted in incorrect power calculation when using multiple dipoles with specific near field requests. (14.0.401) |
| • | Improved the robustness of geometry checking phases. (14.0.401) |
| • | Fixed a bug that might cause an incorrect error condition when checking combined field integral equation (CFIE) geometry consistency. (14.0.401) |
| • | Fixed a bug that caused certain processes to be repeated between different configurations when changing radiating sources in problems that contain networks. (14.0.401) |
| • | Fixed a bug that caused incorrect results for certain periodic boundary problems where the total problem size is large. (14.0.401) |
| • | Improved the algorithm that chooses between standard and fast far field calculations. (14.0.401) |
| • | Fixed a bug that could lead to incorrect results when using far field point sources in certain circumstances. (14.0.401) |
| • | Fixed a bug that prohibited source information from being written to the text output file in certain instances. (14.0.401) |
| • | Support negative angles when importing far fields using *.ffe files. (14.0.401) |
| • | Fixed an internal error state that was caused when running certain models using PO in parallel. (14.0.401) |
| • | Multilayer electrically thin coatings are now supported for the MoM. (14.0.401) |
| • | Fixed a bug that caused output files to be deleted when they are larger than 2 GByte in size. (14.0.401) |
| • | Fixed a bug that caused certain intermediate files to be deleted incorrectly when multiple jobs run on the same system. (14.0.401) |
| • | Fixed a bug that caused OPTFEKO to incorrectly farm jobs on a local machine. (14.0.401) |
|
