You are on page 1of 11

TEMA II.

Seminario: Técnicas numéricas en sistemas radiantes

This is what you will find here:

1. General considerations (*)


2. Classification and selection criteria
3. Finite Difference Time Domain
4. Method of Moments
5. Finite Element Method

*: Many slides taken from R. Rumpf, “Course on Computational Electromagnetics”, University of Texas at El Paso
https://www.youtube.com/watch?v=l_CZFyJd4WI&list=PLLYQF5WvJdJVmCm4cDrKmek6cDJZWVomk) Fig. 2.1

Fig. 1.2
Fig. 1.3

Try this…

clear all
tic
M=10000; N=10000;
z=linspace(1,2,M); t=linspace(3,4,N);
U=zeros(M,N);
for k=1:M
for l=1:N
a(k)=log10((z(k))^(1/2));
b(l)=exp(-3*t(l)/2);
U(k,l)=log10(pi^(1/2))*(sqrt(a(k))+b(l));
end
end
EllapsedTime=toc

Recommendation: Don’t try to remove “U=zeros(M,N);”! Fig. 1.4


Try this… And compare to this…
clear all clear all
tic tic
M=10000; N=10000; M=10000; N=10000;
z=linspace(1,2,M); t=linspace(3,4,N); z=linspace(1,2,M); t=linspace(3,4,N);
U=zeros(M,N); U=zeros(M,N);
for k=1:M Ct=0.5*log10(pi);
for l=1:N b=exp(-1.5*t);
a(k)=log10((z(k))^(1/2)); sqrta=sqrt(0.5*log10(z));
b(l)=exp(-3*t(l)/2); for k=1:M
U(k,l)=log10(pi^(1/2))*(sqrt(a(k))+b(l)); for l=1:N
end U(k,l)=Ct*(sqrta(k)+b(l));
end end
EllapsedTime=toc end
EllapsedTime=toc

Fig. 1.5

Fig. 1.6
Fig. 1.7

Fig. 1.8
Techniques in Computational Electromagnetics
 FVTD (Finite-Volume Time-Domain)
 MOM (Method of Moments, also called Boundary Element Method)
 FDTD (Time Domain Finite Difference)
 FEM (Finite Element Method)
 GO/UTD (Geometrical Optics/Uniform Theory of Diffraction)
 PO/PTD (Physical Optics/ Physical Theory of Diffraction)

Full wave: based on ‘mesh’ concept.


Popular in antenna design and field
analysis of RF circuits. Highly precise:
Errors only come from discretisation.

High Frequency: based on physical


aproximations (Optics). Demand less
computing time. Popular on indoor and We will focus on
outdoor propagation models (ray tracing) these three

Source: Stutzmann, “Antenna Theory and Design” Fig. 2.9

Hybrid methods

MLFMM: Multilevel Fast Multipole Method


https://www.feko.info/product-detail/numerical_methods Fig. 2.10
Selection criteria

https://www.feko.info/product-detail/numerical_methods Fig. 1.11

Selection criteria (II)

https://www.feko.info/product-detail/numerical_methods Fig. 1.12


Full wave methods: selection criteria

PEC: Perfect Electric Conductor (Imprenetrable objects)


Nota: A diferencia de las ecuaciones diferenciales, las integrales garantizan el cumplimiento de la condicìón
de radiación, es decir el correcto comportamiento del campo lejano (proporcional a 1/r). Importante para
problemas de radiación y scattering

Source: Computational Electromagnetics for RF and Microwave Engineering, D. B. Davidson Fig. 2.13

Finite difference concept


Taylor series reminder…

Thanks to this we can convert a derivative into a quotient of differences:


Error

Also for second derivative…


Fig. 2.14
Extension to 2D and 3D. The Poission equation example /

2-D    1   2  2
3 1  0    h   2  h  0(h )
3

 x 0 2  x 0

h 0    1   2  2
2 1 2  0    h   2  h  0(h )
3

 x 0 2  x 0
   1   2  2
3  0    h   2  h  0(h )
3

4  y 0 2  y 0
   1   2  2
4  0    h   2  h  0(h )
3

 y 0 2  y 0
1  2  3  4  h 2 0 / 
1 0 Poisson
 2 1  2  3  4  40  4
h2 1  2  3  4
0 Laplace (0  0)
4
3-D

3
0 6 1  2  3  4  5  6  h 2 0 / 
h 0 Poisson
2 1 6
1  2  3  4  5  6
5 0 Laplace
6
4 Fig. 2.15

FDTD
FDTD can model a device over a large band of frequencies in a single simulation ->
well suited to broadband and transient analysis

http://emlab.utep.edu/pdfs/Poster_FDTD.pdf Fig. 1.16


Method of Moments (MoM)
 Converts an integral equation into a matricial one and solves this one
 To do so the unknown function is considered as a linear combination of a set of base functions.
 One only needs to calculate the function at the surfaces (contours) instead of the entire volume -
-> very suitable for problems with small surface/volume relations.

http://web.mst.edu/~jfan/slides/Li2.pdf Fig. 1.17

http://web.mst.edu/~jfan/slides/Li2.pdf Fig. 1.18


FEM
Base concept: A given function might have a sophisticated dependence over a large
domain, but if we divide the domain in small pieces the calculation of this function is
easier. FEM subdivides a large domain into smaller, simpler parts, the so called “finite
elements
Major advantage: Able to address complex geometries.

http://web.mit.edu/16.810/www/16.810_L4_CAE.pdf Fig. 2.19

2-D FEM Method: MATLAB pdeModeler (formerly “pdetool”)

https://es.mathworks.com/help/pde/ug/pdemodeler-app.html;jsessionid=b9d36cc4e42677b9f74346997315?s_tid=gn_loc_drop Fig. 2.20


Boundary conditions
   
D2 n  D1n  nˆ2 ·( D2  D1 )   B2 n  B1n  nˆ2 ·( B2  B1 )  0
    
E2t  E1t  nˆ2  ( E2  E1 )  0  
nˆ2  H 2  H1  J S

Unitary vector points from 1 to 2

Potentials always continous ,


Two types
 Dirichlet: You specify potential
 Neumann: You specify the derivative of the potential (i.e. Electric Field) in
the normal direction of the contour. Excelent option for surfaces of
conductors, in this case this is just to specify the charge:


D2 n  D1n  D2 n   
n

You can also specify mixed conditions: Dirichlet in one part


of the interface, Neumann in the other.
Fig. 2.21

1. Define geometry… 2. Boundary conditions…

3. Mesh… 4. And solution:

Fig. 2.22

You might also like