采用FDTD算法的软件，如FDTD solutions, Rsoft Fullwave。用FEM算法的软件，如COMSOL。此外CST也比较常用，但是采用的算法好像是有限积分法，这个和前两个又有什么不同？
本人学生，目前主要在做一点计算电磁场的算法，不过研究的方法并不是题主说的FDTD和FEM，是从积分方法中的PEEC。不知道题主这么问主要是想做计算电磁场的研究还是从事电磁仿真的应用，如果是软件应用的话可以先看看下面这个表，以前给小组写报告的时候总结的（转载麻烦知会一下，谢谢）。其实如果只是应用层面的话，这些仿真软件都基本可以满足要求，主要目前做电磁仿真主流的不外乎也就是电机设计，或者是电磁兼容分析。ANSYS在国内算是比较流行，个人觉得也和他们的推广营销有关，在日本更多会使用JMAG，在欧洲其实有个叫FLUX的软件使用更为广泛（目前本人就在一个和FLUX所属公司合作的实验室里头做研究）。其实如果只是应用层面的话，这些仿真软件都基本可以满足要求，主要目前做电磁仿真主流的不外乎也就是电机设计，或者是电磁兼容分析。ANSYS在国内算是比较流行，个人觉得也和他们的推广营销有关，在日本更多会使用JMAG，在欧洲其实有个叫FLUX的软件使用更为广泛（目前本人就在一个和FLUX所属公司合作的实验室里头做研究）。上述提到的ANSYS、JMAG和FLUX都是基于有限元算法的仿真软件，主要也就是对求解域进行剖分然后求解麦克斯韦方程组。至于求解的精度，其实个人觉得还是可以作为工程预设计使用的，很少会出现量级以上的错误。
艾玛，竟然超过一百赞了，我再来好好修改修改&br&----------------------------------------------------------分割线--------------------------------------------------------------&br&敬仰一下大神：&br&&img src=&/0a7ce9b26bb85dd604ac_b.jpg& data-rawwidth=&192& data-rawheight=&232& class=&content_image& width=&192&&评论区里面有人说这头像不是麦克斯韦，这我就不知道了，也没人真见过麦克斯韦长啥样是不？&br&&img src=&/faaaace855bbba2_b.jpg& data-rawwidth=&941& data-rawheight=&572& class=&origin_image zh-lightbox-thumb& width=&941& data-original=&/faaaace855bbba2_r.jpg&&&br&上图基本上涵盖了所有的仿真算法。对于时域算法，分为微分方程类和积分方程类。对于频域算法，分为低频方法和高频方法。除了高频方法是针对于电大尺寸问题，其它的都是精确的全波仿真算法。至于左下角的混合算法，是灌水的绝佳地方。下面说说最主要的三种精确算法：FDTD，FEM，MOM。比较见下图&br&&img src=&/6d884a0a30a2e3a3bf2eb3e8da2ad2ee_b.jpg& data-rawwidth=&935& data-rawheight=&483& class=&origin_image zh-lightbox-thumb& width=&935& data-original=&/6d884a0a30a2e3a3bf2eb3e8da2ad2ee_r.jpg&&&img src=&/316de6f66d61e95b60cad05c42e88208_b.jpg& data-rawwidth=&932& data-rawheight=&519& class=&origin_image zh-lightbox-thumb& width=&932& data-original=&/316de6f66d61e95b60cad05c42e88208_r.jpg&&&br&关于使用哪种仿真软件，哪种算法，可以参见下面的简介。（纯引用）&br&It really depends on what you' both in terms of the technology
domain (antennas, filters, optics, ... ) and if you're doing pure research or not.&br& If the subject you're working in is well treated, you're probably safe with the a 3d solver which claims to be suited for the task. e.g. some of the first things you'll be looking for is:&br& - is it a full-wave solver? What assumptions are made about the wavelength size relative to the structure?&br& - does it solve in a bounded or unbounded domain? what type of boundary conditions are possible?&br& - is it a time- or frequency domain solver?&br& - what kind of excitations can be applied to the structure (e.g. lumped ports, waveguide ports, plane waves?)&br& - is the interface intuitive and clear? is there access to good tutorials/examples/help files?&br& To see if a solver is good for your topic it is probably best to take some published results with measurement data and try to reproduce these.&br& If you're doing research in unexplored areas of science or engineering, then you'll probably need at least 2 different solvers that are based on different numerical techniques, to compare results between them.&br&Some well know solvers are CST Microwave Studio (FIT method), Ansoft HFSS (FE method), Ansoft Designer 2D (MoM method), Zeland IE3D (MoM Method)&br& I really agree with Tony (tboloney).
The &best em solver& depends completely on the application.
There is no one EM software application or approach that solves all problems equally well.
They all have strengths, they all have weaknesses.&br& HFSS is great for many applications, but certainly not for all.
I wouldn't touch it for planar RFIC applications that involve, for instance, 0.2 the mesher goes crazy with stuff with ultra thin dielectrics if I am also trying to solver large matching circuits (spiral inductors, MIM caps, transmission lines, etc.)&br& On the other hands, I would prefer HFSS over IE3D for BGA or detailed packaging design because I think IE3D has to make too many compromises in mesh selection (for speed) to provide the consistently accurate results that I need, where HFSS more accurately models the complex field behavior around the interconnects (just my opinion).&br& I like HFSS for smaller, complex problems where the analysis space doesn't get too large.
I also like it when I am looking at relatively narrowband simulations.&br& I like the time domain 3D codes (CST Microwave Studio, REMCOM XFDTD, Vector Fields, Flomerics Micro-Stripes) for larger space problems, antenna array analysis, or for when I need very broadband simulations.
They are especially well suited to broadband simulation, and they can also simulation and show direct transient field behaviour, which the frequency domain codes (like HFSS, etc.) can't do.&br& However, when you get to planar problems (microstrip, stripline, etc.) you find that you have to push the 3D meshing tools pretty hard to get consistent agreement between measured and calculated, and at some point you find that the planar tools give you a better value in terms of simulation time, accuracy, and general grief required to get a model together and simulate it.
The planar EM tools pretty much break down into open domain (IE3D, Ansoft 2D which used to be called Ensemble, EMpicasso, and Agilent Momentum) and shielded domain (Sonnet, AWR EMSight).&br& The open domain planar tools handle antenna systems with ease, while the shielded domain planar tools usually show better error convergence for highly tuned circuits or for error sensitive Q analyses for on-chip spiral inductors.&br& The planar tools are usually Method of Moments, though there are starting to appear a few tools based on Partial-Element Equivalent Circuit (PEEC) method.
I don't think the PEEC code developers have figured out how to de-embed ports well yet, so their frequency range for accurate simulation seems to be restricted to the low microwave frequencies.&br& There are also very large-space problems (like analysis of radiation scattering off of a car or a ship, or analysis of a very large reflector antenna system) that none of the above will handle with any degree of what could be called efficiency.
For these cases, you probably need a wire-plate MoM code like FEKO, or you need a hybrid approach that combines a local fine analysis where current or field activity is detailed, and an optical approach for &large space& propagation.&br& So you can see that the answer to your question is &it depends.&&br& I also find that the answer also depends a lot on your error requirements.
If you are working in a commercial company, you will find that the ability to understand the sources of error in your simulations will drive your choice a lot of the time.
If you are in an academic setting, this may not be as important to you as you may be looking more for trends.
Bew with some tools they can pop up and bite you without warning if you don't investigate the limits of the tool that you are using.
And this can affect your career in a big way!&br&好像有点长，简单总结一下：&br&（1） 因为HFSS采用有限元，四面体剖分，建模准确。因而适合尺寸不大，结构复杂，窄带的问题&br&（2）CST采用的有限体积分，六面体剖分。适合时域分析，宽带分析，另外可以处理稍微大一点的物体。&br&（3）至于其它的不是主流自己慢慢看吧。&br&最后不要迷信仿真软件啊，他们都是不准的，特别是CST。要做一个有物理图像的人，下面是不同软件计算结果的比较。&br&&img src=&/7f7a592debbb709e930414_b.jpg& data-rawwidth=&548& data-rawheight=&415& class=&origin_image zh-lightbox-thumb& width=&548& data-original=&/7f7a592debbb709e930414_r.jpg&&&br&在结束之际，再来看看最酷炫的电磁场仿真图：&br&&img src=&/96fba9dfdca_b.jpg& data-rawwidth=&542& data-rawheight=&271& class=&origin_image zh-lightbox-thumb& width=&542& data-original=&/96fba9dfdca_r.jpg&&&img src=&/274b968d_b.jpg& data-rawwidth=&347& data-rawheight=&234& class=&content_image& width=&347&&&img src=&/97cb3ed765ba28e83d60a42b7f301cc7_b.jpg& data-rawwidth=&1092& data-rawheight=&808& class=&origin_image zh-lightbox-thumb& width=&1092& data-original=&/97cb3ed765ba28e83d60a42b7f301cc7_r.jpg&&&img src=&/763ad6a8d023b20a22b892a8b7a56c51_b.jpg& data-rawwidth=&903& data-rawheight=&803& class=&origin_image zh-lightbox-thumb& width=&903& data-original=&/763ad6a8d023b20a22b892a8b7a56c51_r.jpg&&
艾玛，竟然超过一百赞了，我再来好好修改修改----------------------------------------------------------分割线--------------------------------------------------------------敬仰一下大神：评论区里面有人说这头像不是麦克斯韦，这我就不知道了，也没…
白冰冰我还爱你&br&FDTD作为一种有限差分时间域数值计算方法，有一次性脉冲分析（宽频），自然结合非线性谐波分析，自然结合Lorentz , 导体等电介质的优势。缺点是处理高Q震荡时候需要花费相当长的时间（高Q震荡需要很久才会消失），此外他的空间差分方式单一，只能分割成长方体。对于非常规边界和材料形状的仿真有些力不从心。&br&FEM则基本克服了高Q仿真，长方体差分的缺陷，可以对空间进行任意分割（不均匀的四面体，例如）。但问题是，他一般是对单一频率频域进行仿真，因此对于脉冲源的仿真比较复杂，并且难以分析系统响应的波形，也难以分析非线性系统出现的谐频。对于宽频源的仿真需要更长时间。但对于高Q震荡的计算不在话下，频谱分辨率也可以达到任意高。
白冰冰我还爱你FDTD作为一种有限差分时间域数值计算方法，有一次性脉冲分析（宽频），自然结合非线性谐波分析，自然结合Lorentz , 导体等电介质的优势。缺点是处理高Q震荡时候需要花费相当长的时间（高Q震荡需要很久才会消失），此外他的空间差分方式单一，…
已有帐号？
无法登录？
社交帐号登录天津大学电磁场与电磁波(数值算法)_图文_百度文库
两大类热门资源免费畅读
续费一年阅读会员，立省24元！
评价文档：
天津大学电磁场与电磁波(数值算法)
上传于||文档简介
&&天​津​大​学​电​磁​场​与​电​磁​波​教​学​课​件​ ​ ​金​杰
大小：954.00KB
登录百度文库，专享文档复制特权，财富值每天免费拿！
你可能喜欢}