The Finite-Difference Time-Domain (FDTD) Method - Part Ii: Numerical Techniques in Electromagnetics
The Finite-Difference Time-Domain (FDTD) Method - Part Ii: Numerical Techniques in Electromagnetics
Nikolova 2004
9. Some excitation functions (sources) The problem of choosing the proper space-time dependence of the excitation g(x,y,z,t) requires special attention. Here, we confine our attention to the time-dependence of the sources and their discretization. Sinusoidal excitation n 2 t , n = 0,1, 2,3, sin( t ) = sin T
g (t ) = e
Nikolova 2004
( t t0 )2
g (t ) = e
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -1 0 1 2
( t t0 )2
t0 = 0
t0 = 2.5
t0 = 5
t
3 4 5 6
Nikolova 2004
f (t ) = e
16( t t0 ) 2
g (t ) = e
( t t0 )2
g (t ) = e
( t 1.5) 2
g(t)
=4 =9
0.5 1
= 16
1.5 2 2.5 3
4
Nikolova 2004
g =e
n
t 2 ( n n0 )2
The discrete Gaussian excitation is controlled by two numerical constants: and b, where b denotes the half pulse width. Recommended b : b > 30 .
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
Nikolova 2004
2b t truncation level
0 0.5 1 1.5 2 2.5 3
9. Some excitation functions (sources) cont. At the truncation level: | n n0 |= b Set truncation level at e 2
Note: n0 b
e t 2b2
e 2
t2
= A = (
/ b) 2
g =e
n
A( n n0 )2
The truncation level should be comparable with the precision of numbers. For single precision (6 significant digits), recommended value is = 4. This corresponds to truncation at exp(-16). For double precision (12 significant digits), = 5. The pulse width in terms of b is determined according to the desired width of the excitation frequency spectrum.
Nikolova 2004 6
9. Some excitation functions (sources) cont. The Fourier transform of the Gaussian pulse is a Gaussian function of frequency: 2 2 g ( f ) e f / We require that the spectral value at the highest frequency of interest is at 0.3 of the maximum: 2 e 2 f max / = 0.3
2 2 2 f max 2 f max t 2 = ln 0.3 A = = ln(10 / 3) b 2
ln(10 / 3) b= f max t
If we set f max t = t / Tmin
Nikolova 2004
32 ln(10 / 3) = 1/ 32 then b =
7
9. Some excitation functions (sources) cont. Typically, the spatial step h is first determined according to the finest detail of the structure. Then, t is computed such that c t = 1/ D h where D denotes the dimensionality of the problem (D=1,2,3). Often, t computed this way is much smaller than Tmin/32. Then, b is determined as
b=
ln(10 / 3) min h
Note that for good accuracy, there should be at least 32 h in one min. Often, however, the finest detail in a microwave structure is much smaller than min/2.
Nikolova 2004 8
9. Some excitation functions (sources) cont. Band-limited excitations (a) sine wave modulated with a Gaussian pulse
Nikolova 2004
9. Some excitation functions (sources) cont. Sine wave modulated with a Gaussian pulse, cont.
Nikolova 2004
10
9. Some excitation functions (sources) cont. (b) sine wave modulated by a Blackman-Harris window g (t ) = B (t )sin( t ),
w = / N ( N = 710)
a0 = 0.35875 a1 = 0.48829 a2 = 0.14128 a3 = 0.01168 N controls the bandwidth of the spectrum.
Nikolova 2004
11
10. Time-domain Maxwell equations for a dispersionfree medium The FDTD algorithm is based on the Maxwell curl equations
H(r, t ) E(r, t ) = + m H(r, t ) + J im (r, t ) t E(r, t ) H (r , t ) = + e E(r, t ) + J ie (r, t ) t In rectangular coordinates, the above is written as E z E y H x i +mHx = J mx t y z H y E x E z J i +mH y = my t x z E y E x H z i +mHz = J mz t x y
Nikolova 2004 13
Reduction to 2-D problems Consider a field and its sources, which do not depend on one of the spatial coordinates, e.g., the x coordinate. Maxwells equation then reduce to two decoupled systems of equations, each of which involves only three field components.
Nikolova 2004 14
=0 x
15
10. Time-domain Maxwell equations for a dispersionfree medium cont. TEx modes E z E y H x i +mHx = J mx t y z TMx modes E x H z H y i + e Ex = J ex t y z H y E x i +mH y = J my t z E x H z i +mHz = J mz t y
Nikolova 2004
H x i J ey t z H x E z i + e Ez = J ez t y + eEy =
E y
16
10. Time-domain Maxwell equations for a dispersionfree medium cont. Important note: The field decomposition into TE and TM modes in 2-D problems is different from the 3-D TE and TM decomposition you know from waveguide theory! Consider the TEz and TMz modes in a waveguide and the TEz and TMz modes in a 2-D problem. (a) In 2-D problems the field is independent of z. (b) In a waveguide the field depends on z. (c) In 2-D problems the 2-D field has only 3 nonzero components (TEz: Ex,Ey,Hz; TMz: Hx,Hy,Ez). (d) In a waveguide the 3-D modal field has 5 nonzero components (TEz: Ez=0; TMz: Hz=0)
Nikolova 2004 17
10. Time-domain Maxwell equations for a dispersionfree medium cont. The 2-D TEz mode E y E x H z + mHz = t x y E H i x + e E x = z J ex t y E y H z i + eEy = J ey t x The 3-D TEz mode F = zF ( x, y , z , t )
E = F H + mH = F t Ez = 0 F F Ex = ; Ey = y x
18
10. Time-domain Maxwell equations for a dispersionfree medium cont. The 2-D TMz mode H y H x E z i + e Ez = J ez t x y H E x +mHx = z t y H y E z +mH y = t x The 3-D TMz mode A = zA( x, y , z, t )
H = A E + eE = A t Hz = 0 A A ; Hy = Hx = y x
19
10. Time-domain Maxwell equations for a dispersionfree medium cont. Take as an example the TEz0m waveguide modes Ez = 0 Ex , H y , H z F F Ex = =0 ; Ey = This is a 2-D TMx mode! y x
H y 2 F +mH y = t yz H x 2 F +mHx = =0 t xz H z 2 F 1 2 F +mHz = 2 2 2 t z c t 2 F 2 F = 2 + 2 y x
Nikolova 2004
b a
10. Time-domain Maxwell equations for a dispersionfree medium cont. There are no TMz0m waveguide modes! The boundary conditions
make only the trivial solution possible: Az ( x, y , z, t ) = 0 Thus, the field is zero, too.
b
x
a
Nikolova 2004 21