Talk:Finite-difference time-domain method

‹See TfM›

FDTD simulation software packages

Quite a few software simulation packages exist for the calculation of fields using the FDTD method. Two of these are widely used in research and development of active and passive microwave components: CST and the Vector Fields/CONCERTO package. Most researchers will use only one simulation software package, leading to the effect that the different packages aren't often comprehensively compared to one another.

Relevance of publication trend pic?

How useful is that pic in the final section? It seems like it takes up a lot of space for imparting very little information. I would think it sufficient to just state the general trend of publications on FDTD. 68.62.179.96 (talk) 05:51, 29 August 2008 (UTC)

I agree with that assessment. Do feel free to change the article. The button labelled edit at the top is there for a reason! -- Jitse Niesen (talk) 11:22, 29 August 2008 (UTC)

Alright, well I made this change and someone reverted it, saying that it was "essential to understanding the development of this technique." What is to be done here? I really see no reason why the growth of publications related to FDTD deserves more than a few sentences. Nor do I see any similar pics on pages for related techniques like Finite Element Method. I think I am going to delete the pic again and leave a note asking the next person who puts it back to post something on the this talk page. I hope this is not improper Wikipedia'ing!128.148.5.94 (talk) 06:23, 28 October 2008 (UTC)

I will respond. My perspective is one of a person who has been active in FDTD since 1971. Computational electromagnetics at the time involved ONLY phasor-domain techniques. Our early attempts to publish on what I called "FDTD" were all rejected by the major journals such as IEEE TAP. Our conference papers were relegated to the last day, last hour, when hardly anyone was left. There was incredible resistance to anything new coming from grid-based time-domain modeling. You were *not* there to face this resistance. I was, and I battled it one paper at a time -- doing the "homework problem" validations that disproved all of the doubters' claims of inaccuracy. I and a few others greatly expanded the scope of FDTD applications, and published the fundamental papers and books that eventually made FDTD perhaps the most widely used and powerful CEM technique. The near-exponential growth of FDTD publications shown by the graph in question is indicative of the turn of opinion by the worldwide CEM community. This graph vividly depicts when this occurred so that newcomers to the field realize just how recently all of this happened, and gain some perspective on the revolution in CEM brought about by FDTD. I argue strongly that we cannot and must not ignore this history. FDTD did not just come down from the sky. Its current widespread acceptance and use is the result of literally three decades of unceasing hard work against tall odds. Let this graph remain. Sincerely, Allen Taflove, Northwestern University —Preceding unsigned comment added by 24.12.211.91 (talk) 13:29, 29 October 2008 (UTC)

Dr. Taflove, no disrespect was meant and I am not trying to denigrate your contributions; however, I do not feel that the graph really conveys your above point. Perhaps it would be better served by a history section expanded from the brief paragraph in the introductory section of the article? The graph alone does not say much, and I think words would provide better context for the rise in the popularity of the technique.

128.148.155.57 (talk) 20:55, 29 October 2008 (UTC)

As an impartial observer, I'd have to agree with the OP - regardless of his familiarity with the method, his suggestion would seem to improve the article. Additionally I was about to make that same suggestion regarding a new history section, but he posted first. If that section were to grow bulky enough, then an offshoot-page ("History of the finite-difference time-domain method") should be created, and the graph would then be appropriate there. Beefnut (talk) 21:05, 29 October 2008 (UTC)

I have no objection to creating a history section, and would volunteer to draft one. It need not be bulky. Regards, Allen Taflove

Certainly, as you would seem to be uniquely positioned to do so.138.16.102.228 (talk) 20:36, 31 October 2008 (UTC)

As you can see, I have posted a draft history section for everyone to discuss. At this time, I've omitted the additional references that it cites, awaiting your comments. Thanks for the opportunity to provide some input to the FDTD page. Regards to all, Allen Taflove —Preceding unsigned comment added by 24.12.211.91 (talk) 20:33, 1 November 2008 (UTC)

The additional references are now posted. A few links to the source documents are missing, but will be added as available. Allen T. —Preceding unsigned comment added by 24.12.211.91 (talk) 21:41, 9 November 2008 (UTC)

The changes proposed by 193.147.222.244 have grammatical errors, add nothing to the article, and in fact, reduce the quality of the article. Hence, I've deleted them. Allen T. —Preceding unsigned comment added by 24.12.211.91 (talk) 14:58, 20 November 2008 (UTC)

There are many other possible reasons why publications have grown exponentially, such as the push in the mid ninties in academia to increase publication output. For example Why I am not a Professor. I would delete the whole section as mostly just fluff. If you wanted to make a statistical case it would be better to show that the proportion of papers in some well defined area (say papers published at IEEE_PDE) has increased markedly. --Jaded-view (talk) 23:56, 30 August 2009 (UTC)

Uses of FDTD

I think it is worthwhile to include subject areas where FDTD is used. I've started a list, I'm not qualified to cover all the areas. Thus I would appreciate if someone more qualified than me continues. I think it is very useful to know which subjects can benefit from FDTD.Eranus (talk) 14:28, 2 June 2009 (UTC)

It is pointless to attempt such a list — you'd be hard-pressed to find topics in electrodynamics that are not studied with FDTD except for magneto/electrostatic phenomena (where you only need the electric field or the magnetic field, but not both, since the coupling between the two is negligible in the quasistatic limit). You'd end up listing all topics in non-quasistatic electromagnetism.

(Sorry I reverted with the "rollback" button by mistake; I meant to press "undo" and add an edit summary). —Steven G. Johnson (talk) 17:45, 2 June 2009 (UTC)

I'm a Phd student investigating guided mode resonance and a realtively new user (and admiror) of MEEP which I understand you are an important contributor to. I do not have the wide scope, but at least for guided mode resonance I have confidence about where FDTD allows me to do things I could not do previously with RCWA for example, and where are it's difficulties. I thought it would be helpful for each expert to list in his subject area the advantages and disadvantages of FDTD. Do you think this may be a good idea as a separate page with a link from the main FDTD page? I think it is not a bad idea to have a long list of the vast topics in modern EM research and the advantages of FDTD in analyzing them? So the list will not be complete and only those who are Wiki users would update, but a partial list is better than non? Curious to hear your thoughts.Eranus (talk) 10:42, 3 June 2009 (UTC)

Listing the advantages of each method is an extremely tricky topic, because the advantages and disadvantages depend strongly on how the methods are implemented and on details of the problem. It is completely impossible to make blanket statements about the advantages of FDTD vs. RCWA for guided-mode resonances, for example: it depends on the nature of the guided mode and the waveguide geometries of interest, and it also depends strongly on how RCWA is implemented (RCWA is essentially yet another name for transfer/scattering-matrix methods based on eigenmode expansions, but there are many choices of basis for such things and many solution techniques) and how FDTD is implemented (e.g. some FDTD schemes have adaptive resolution, some (like Meep) are combined with special signal-processing techniques to extract resonances, and so on). Because of these facts, and also because it is hard to find papers making comprehensive comparisons of methods even for specific problems, it seems impractical to include such an article in Wikipedia.

Unfortunately, Wikipedia cannot simply rely on "experts in each area" to contribute their thoughts. Whether or not we can find experts in each area, that would be original research. We have to rely on already-published sources.

We have a brief overview and comparison of a number of different numerical approaches in appendix D of our book, which you can read online at http://ab-initio.mit.edu/book/ if you want. Perhaps this will be helpful to you. —Steven G. Johnson (talk) 12:31, 3 June 2009 (UTC)

Thank u for your detailed reply. BTW, I know the book well (even from it's previous version), unbelievable you guys post it online for free (Thats fantastic, very rare I think). What you say makes sense, though I think there are probably specific problems in modern optics that have been only simulated with FDTD and are published, I think it would be good to list these. I'll wait several more years to gain more perspective and experience as I'm very new in the world of numeric techniques (and Wikipedia). Eranus (talk) 13:35, 3 June 2009 (UTC)