3. A Hybrid MM/FE Technique to Analyze Horn having Discontinuities
This chapter presents the solution of Helmholtz equation using 2-D finite element method (FEM) to calculate the fields and their cutoff wave-numbers for non-regular surfaces. Delaunay triangulations are used to develop the FEM solution. The related theory for accurate prediction of eigenvalue and eigenvector for the solution of Helmholtz equation is discussed. The issues related to the calculation of reaction matrix for non-regular junction employing MM technique are elaborately discussed. The performance of this hybrid technique which is a combination of MM and 2-D FEM is evaluated in case of stepped cylindrical horn containing inner posts.
3.2 Helmholtz equation solution using 2-D FEM technique
as circular, rectangular, elliptical etc. When we need to model waveguide having non-regular geometry or having discontinuity, numerical techniques become more useful. Several numerical techniques such as FDM, FEM, FDTD are available for analysis of such geometry. From literature [53,54], we find that FEM provides very good accuracy. FEM technique for solving the wave equation in guided structure are reported in this chapter.
The wave equation in a homogenous medium as mentioned in Section 2.1 is given by∇2φ+kc2φ= 0.
Here,φis the scaler potential of longitudinal magnetic field for TE, longitudinal electric field for TM mode and electric field for TEM mode, respectively. Solution using variation method for φ can be obtained by optimizing the integral equation as given below:
F(φ) = 1
2 |∇φ|2−k2cφ2
dS (3.1)
The natural boundary conditions i.e. Dirichlet or Neumann boundary conditions must be satisfied.
Now, the total surface is split into small sized meshes like triangular, rectangular etc. This small sized mesh is called an element and every element has some nodal points. The scalar potential and nodal point potential for nth element can be written as φne(x, y) and φnei, where iis the local node number of the nth mesh element. Let the scalar potential in an element vary as:
φne(x, y) = [
2ndorder
$ %& '
1storder
$ %& '
1 x y x2 y2 xy x3 y3 x2y xy2
& '$ %
3rdorder
...]
(
an bn cn dn...
)T
& '$ %
Constant
(3.2)
The choice of number of node points (np) depends upon the order p of the scalar potential function.
The relation between number of node points and order of polynomial isnp = (p+1)(2p+2). The constant values of scalar potential function can be evaluated using node points, as specified using cartesian
3. A Hybrid MM/FE Technique to Analyze Horn having Discontinuities
coordinate values for equation (3.2). The constant values using equation (3.2) can be written as:
⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎣ an bn cn ... ...
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎦
=
⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎣
1 x1 y1 x21 y12 x1y1 · · · · 1 x2 y2 x22 y22 x2y2 · · · · 1 x3 y3 x23 y32 x3y3 · · · ·
· · · · · · · · · · ·
· · · · · · · · · · ·
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎦
−1⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎣ φne1 φne2 φne3 ... ...
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎦
(3.3)
Combining equations (3.2) and (3.3), the scalar potential function can be written as
φne(x, y) = [ 1 x y x2 y2 xy · · · · ]
⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎣
1 x1 y1 x21 y21 x1y1 · · · · 1 x2 y2 x22 y22 x2y2 · · · · 1 x3 y3 x23 y23 x3y3 · · · ·
· · · · · · · · · · ·
· · · · · · · · · · ·
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎦
−1⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎣ φne1 φne2 φne3 ... ...
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎦
(3.4)
Equation (3.4) can also be written as
φne(x, y) = (
α1 α2 α3 · · · · ) (
φne1 φne2 φne3 · · · · )T
(3.5)
where α is called the shape function. Substituting the equation (3.5) into equation (3.1), we obtain
F(φne) = 1 2
(
φne1 φne2 · · · · )
⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎣
∇α1· ∇α1ds
∇α1· ∇α2ds · · · · ∇α2· ∇α1ds
∇α2· ∇α2ds · · · ·
· · · · · · · · · ·
· · · · · · · · · ·
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎦
⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎣ φne1 φne2 ... ...
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎦
−k2c 2
(
φne1 φne2 · · · · )
⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎣
α1α1ds
α1α2ds · · · · α2α1ds
α2α2ds · · · ·
· · · · · · · · · ·
· · · · · · · · · ·
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎦
⎡
⎢⎢
⎢⎢
⎢⎢
⎢⎣ φne1 φne2 ... ...
⎤
⎥⎥
⎥⎥
⎥⎥
⎥⎦
= 1 2[φne]T
Ce(n)
[φne]−k2c 2 [φne]T
Te(n)
[φne]
(3.6)
3.2 Helmholtz equation solution using 2-D FEM technique
Equation 3.6 is derived for a single element. Solution for the entire region is given by
F(φ) =
*N n=1
F(φne) (3.7)
From the equations (3.6) and (3.7), F(φ) can be expressed in matrix from as
F(φ) = 1
2[φ]T [C] [φ]−kc2
2 [φ]T[T] [φ] (3.8)
whereCandTare the global matrices consisting of local C(en) andT(en) matrices coefficients, respec- tively and [φ] represents the scalar value of all the node points over the global surface. N presents the total number of elements of entire surface.
3.2.1 Solution of TEM and TE modes
TEM and TE modes must satisfy the Neumann boundary condition. The solution of TEM and TE mode will be optimization of equation 3.8 with respect to [φ]. So, ∂F∂[Φ](φ) = 0.
[C−k2cT][φ] = 0 or
[T−1C−kc2I][φ] = 0 (3.9) where I is the unit matrix. k2c is the eigenvalues of T−1C matrix and it should be real quantity as C and T are symmetric matrices. In case of TEM mode, kc is equal to zero and the solution of the eigenvector will be the scalar potential.
3.2.2 Solution of TM modes
In case of TM mode, the boundary condition should support the Dirichlet boundary condition.
So, the scalar potential at the boundary nodes (φp) should be zero. From the equation (3.8), we can write for the prescribed nodes [φp] and the free nodes [φf] as:
F(φ) = 1 2
(
φf φp )⎡
⎢⎣ Cf f Cf p Cpf Cpp
⎤
⎥⎦
⎡
⎢⎣ φf φp
⎤
⎥⎦−kc2 2
(
φf φp )⎡
⎢⎣ Tf f Tf p Tpf Tpp
⎤
⎥⎦
⎡
⎢⎣ φf φp
⎤
⎥⎦ (3.10)
Here ∂F∂[Φ(φ)
f] = 0. So, TM mode solution should be [Tf f−1Cf f −k2cI][φf] = 0.
3. A Hybrid MM/FE Technique to Analyze Horn having Discontinuities
3.2.3 Surface Meshing
Surface mesh plays an important role in performance of the numerical electromagnetic calculations.
Several types of mesh such as triangular, rectangular, hexagonal or their mixed types are generally used for surface meshing. Mainly, triangular mesh having the ability to handle any arbitrary surface are widely used. One of the triangular mesh is Delaunay triangle. It is used for numerical electromagnetic calculation due to the availability of fast algorithm for Delaunay triangulation and in such schemes the absence of similar type of triangle improves the error performance in electromagnetic calculations done numerically [55]. To develop the FEM code, PDE Toolbox in MATLAB for Delaunay triangulation has been used. Some codes for graphics interface are developed to draw the desired surface in PDE Toolbox and collect the information as required. Basically, the information of triangulation are presented as position vector of node points (it specifies the node number and their position in cartesian coordinate) and a vector which contains nodes’ number for every triangle and its sub-area information. To close the PDE tool box after it is used, a code has been developed. Any type of planer objects in PDE Toolbox can be drawn using circle, rectangular, elliptical, polygon and using the operations of substraction, intersection and union etc. Also, a MATLAB code is written to identify the boundary nodes and interior nodes of the surface. Specifically, interior nodes are used for TM mode solution and sometimes boundary nodes are also used for triangulation of 3D objects. The boundary edge size can be controlled using step size for triangulation. In Figure 3.1-(a), we represent the property of Delaunay triangular mesh in which the nodes of a triangle cannot be located inside the circum-circle. Also, different objects (rectangle R1,R2,R3 and circle C1) for triangulation using the operation of intersection, union and substraction, respectively are drawn in Figure 3.1-(b). The sub-area of its surfaces (1, 2, 3 and 4) are also presented in same Figure.
3.2.4 Evaluation of local coefficient matrix in closed form
2D-FEM code has been implemented based on scalar potential as a 1st order polynomial using the Delaunay triangular mesh. Closed form matrix of local Ce(n) and Te(n) are formulated on the basis of Appendix 8.2 as given below:
C(en)= 1
4An[pipj+qiqj]3×3 where i= 1,2,3 &j= 1,2,3 (3.11)