Jump to content

Wilco Strydom

  • Content Count

  • Joined

  • Last visited

Everything posted by Wilco Strydom

  1. Hi Constantine, Performing CMA on a model with Surface Equivalence results in a combination of real and fictitious modes. Fictitious modes are identified and filtered out by verifying the modal radiation condition, since these modes are non radiating. Internally we increase the number of modes calculated, but for some models there will be very many fictitious modes near resonance and the solver will terminate with an error if at any frequency it is unable to give at least one radiating mode. You can simply increase the number of modes requested and rerun the solver.
  2. Hi Christian, Interesting problem to look at for a beginner in CMA. The first thing I did with your model was to run a simulation with 20 samples in [1.05, 1.07] to see if there was any interaction between the modes. Since there was none, I wont now go into details of cross avoidance and whether or not that is desirable. Let me answer some of your questions directly: Yes, but similarly to current density correlations, only between immediately adjacent frequencies. There is no tracking error. While this might sound strange, Characteristic Modes can and do change their characteristics over frequency, but they do so gradually. Looking at far field patterns, or at instantaneous currents* (particularly insightful of how the mode changes) you can see how each step in frequency is only a gradual change. In your case it so happens that Mode 1 at lower frequencies has the same characteristics of Mode 4 at higher frequencies. Looking at Modal weighting coefficients (always valuable when antennas are involved) shows that at any one point it is the dipole mode that is most excited, hence the total field value observations. *If you switch on arrows for instantaneous currents, note that the current is prone to 180 degree phase jumps from one frequency to the next, so the arrows might flip on you. This is because the current is on both sides of the eigen equation: X*J = lambda*R*J
  3. Note that this bug is only in parallel runs, so if your university is not willing to upgrade at this stage you can simply run your model in sequential as a workaround.
  4. Hi Lydia. I see you are a University User, are you able to update your Feko version? Your outfile reports FEKO.CSV Version 2018.1-424 from 2018-07-20 I suspect this error is triggered by the dielectric-metallic junctions shown in your first image. If that is the case then updating should resolve your problems as that error has been fixed. Good news is that this error is only triggered in parallel runs, so if you cannot update the solver you could always just perform the run sequentially.
  5. Hi Sumithra, I would consider the following approach: Split the band into two simulations with the first ending at 5.6GHz. Use a similar amount of frequency samples for both simulations, thus giving the lower band finer sampling. Request around 10 modes for the lower band and around 20 for the higher band. Review mode tracking and total captured power and adjust simulations as needed to remedy any problems. To answer some of your other questions: Any MoM supported excitation can be used with CMA. 3 modes are likely insufficient. You can verify the total amount of power captured by your requested modes in the *.out file. Modes are sorted according the lowest frequency of the simulation. All the best.
  6. First step would be to see if the Feko calculated source data would provide what you are looking for. Add a 1V voltage source to your CMA model and plot the modal admittance. If this is not what you are looking for you will have to look into writing your own scripts. Have a look around on the forum for LUA scripting in POSTFEKO.
  7. If you are running a model with a source data you will also be able to plot modal source data such as impedance and reflection coefficients.
  8. Hi, Feko *.mat files contain matrix entries and have nothing to do with Matlab m-files. Your antenna was solved in parallel, so those are the matrix files calculated on each processor. If you want a single matrix file you can uncheck "Parallel" in the Run/Launch options in CADFEKO.
  9. Hi Shane, The str2ascii utility is still being supported and maintained and would be the approach to take if you are looking to access the solution vector. The MoM matrix equation has a current vector (not matrix) that can be written to a *.str file and reused for further simulations where the model is not changed but more requests are added for instance. This *.str file is thus the coefficients of the expansion functions. The *.os file contains samples of the expanded surface current. This also represents the values that you are able to plot in POSTFEKO. Note that you are on a very old version of Feko, and this might complicate certain support questions a bit.
  10. Hi RF Engineer, To do this you simply create two 3D Views in POSTFEKO and add the desired modal current to each. In order to view a (scaled) combination of modes you will have to resort to some scripting to combine the modes. Have a look at other threads discussing scripting, or work through the intro section and the example in the Example guide. From what I remember from this article and what is indicated in the images you attached, this is not what you want to do. For one, the excitation is at a point where the currents are equal but neither is at maximum. I also expect that you will want to model the complete patch setup with the feeding included, since this will affect the modes, and I would then also look at Modal Weighting Coefficients in order to ensure that the two modes can be excited equally. I thus expect you would set your script up to look at the difference between the two currents (scaled with MWC) and then look where this plot goes through zero. @Peter Futter can possibly confirm, but I don't think I've seen a case study where a CMA based design for a circularly polarized antenna has been illustrated in full. It would be interesting to see your results.
  11. Currently a CMA configuration with modal excitation coefficients where the source is a plane wave loop over multiple angles of incidence will only consider the final angle of incidence.
  12. You will always get the most significant modes at every frequency, so the result set is the same with or without tracking. If tracking is disabled those modes will just be numbered 1 to 6 at every frequency. To make untracked plots easier to interpret I usually set all line colours to black.
  13. What is happening here is that the modes are assigned numbers so that we can consistently refer to a single mode across frequencies. At every frequency the *.out file will list the modes of that frequency in order of Modal Significance. The mode tracking algorithm will however determine mode numbering. In your example of getting modes in order 3-2-1, that simply means that at this frequency mode 3 is the most significant followed by mode 2 then lastly mode 1. When you request 6 modes you will only receive results for 6 modes at every frequency, but you will receive the most significant 6 modes at that frequency. This means that over the requested frequency range some modes will fall out of significance and new modes will enter significance. It is then very possible for a mode with the number 'mode 100' to be in the top 6 most significant modes at a later frequency. If you would like the modes numbered 1 to 6 at every frequency you can disable tracking in the *.pre file. The easiest way is to use EDITFEKO. The CMA request will look something like this: ** Characteristic modes OM: 6 : 1 : -1 ** CharacteristicModes1 With the last number indicating that tracking is disabled.
  14. To answer the first part of your question: Characteristic modes are numbered at the lowest frequency, with mode 1 being the most the significant at that frequency. Some modes that are used by the tracking algorithm are also assigned numbers at this frequency, and as modes fall into and out of the modal significant range (determined by the number of modes requested) new mode numbers are assigned dynamically. The user can give no input on the numbering. As Peter said, a screenshot of the modal plots you are having problems with would help to answer your question fully.
  15. The Characteristic Mode formulation for dielectrics solved with the surface equivalence principle (SEP) results in a set of eigen modes that contains both internal and external resonances. The internal modes (sometimes call non-physical or fictitious) do not radiate any power, and are thus removed from the result set. It is advised to increase the number of modes requested when dealing with such models.
  16. Modal Q factors are not currently directly available in FEKO. If you would like to perform your own Q factor calculations with FEKO extracted Z-matrices you can have look at the following paper: http://ieeexplore.ieee.org/document/5996827/
  17. Correct. Are you sorting your modes. FEKO will report eigenvalues with the smallest magnitude (most significant) up until the number of requested modes. To use your syntax from above, with Z = R + jX, and V the source (or driving vector), the MoM matrix problem is: ZI = V I suggest perhaps doing a study of the different CEM methods before delving too deeply into Characteristic Mode theory.
  18. Hi Hamad, Unfortunately at this stage there is no direct way to access the basis function coefficients (vector Jn) of eigencurrents. We understand that this could have some utility for researchers such as yourself and are looking into making this available in the future.
  19. They are both [Nx1] vectors of basis function coefficients that describe a surface current density. An eigencurrent vector is one of N possible solution vectors to an eigenproblem. The solution to the MoM matrix problem is a single current vector that describes the currents as excited by the sources present in the problem.
  20. Assuming you are referring to edges that form the boundary of a surface, what information about these edges are you looking for? The surface current (which is a vector quantity represented as a surface density in Ampere/meter (A/m)) can be exported to the *.out or a *.os file where the vector quantity will be reported on the triangle centre and the current magnitude will be reported for triangle corners. The *.str file reports the weighting coefficients of the matrix solution so that can be used to construct the surface current at any point on the model.
  21. With a request of 6 modes FEKO will give the most significant 6 modes at every frequency. It is not a tracking error when modes fall out of significance and are replaced by others. If you would like the track the 6 most significant modes at the first frequency across the whole frequency range you will need to request more modes (10 should be sufficient) to account for the fact that some of those 6 modes are likely to drop in significance.
  22. Hi BM Kung, I ran your problem with your original proposed 31 frequency points over the 3GHz range and there are no tracking errors, even with 100MHz steps. I think what you are having a problem with is mode 4 and mode 6 not being reported over the whole frequency range? What is happening here is that FEKO solves the CMA problem independently at each frequency point and returns the most significant modes up to the number of requested modes (in this case 6). Tracking is then applied as a post-processing step over frequency samples. If a mode that is initially ranked within the top 6 loses its ranking at a higher frequency FEKO will stop tracking that mode and return the new more significant mode instead.
  23. Syntax: mat2ascii FILENAME [b] Filename: name of the *.mat file (with extension) Options: b block number (e.g. for multiple frequencies) (default is first block, b=1)
  24. Unfortunately if you intend to do your own modal manipulation and further post processing, solving the eigen equation from scratch will be the only way to achieve this.
  25. You should search through the Matlab documentation for the functions you require.
  • Create New...