Kiran R S-FDTD - 1D Date: April 21, 2020

Major Works done

• Time domain RBF- Modify the present 1D RBF code to represent a free space domain with a
dielectric slab of relative permitivity 4, to find the Electric field intensity at some point.
• Time domain S-RBF- Frame the equivalent stochastic RBF code to estimate the distribution
of Electric field intensity at some point.
• Time domain RBF- Modify the 1D RBF code above to find the reflection and transmission
coefficients at some points.
• Time domain S-RBF- Frame the equivalent stochastic RBF code to estimate the distribution
of reflection and transmission coefficients at some points.
• FEM- Modify the old 1D-FEM code to include a dielectric slab of relative permitivity 4, and
estimate the reflection and transmission coefficients at some point. A Monte Carlo simulation is
run on this with the dielectric slab permitivity realised from a Gaussian distribution.
• Comparison- Compare the results of S-RBF and MC-FEM

1
Kiran R S-FDTD - 1D Date: April 21, 2020

1. Time domain RBF- Modify the present 1D RBF code to represent a free space domain with a
dielectric slab of relative permitivity 4, to find the Electric field intensity at some point. This is done
by changing the C1 vector permitivity according to the domain. The domain is a 4m long line along
the x axis from ’0’ to ’4’. a source is there at 1.45m. There is a dielectric slab after 1.95m upto 2.10m.
We measure the electric field at 2.15m. This domain setup is repeated for all the works done above.

2. Time domain S-RBF- Frame the equivalent stochastic RBF code to estimate the distribution of
Electric field intensity at some point.
The stochastic equations are framed following the PCE expansion with Galerkin procedure. Final form
of the electric field equations in PCE coefficient form is as follows.

   
Ψ2k
 
n+1/2 < > if i = k n+1/2 < Ψk > if i = k
hi1 J11
0 otherwise 0 otherwise
 
   
m  m  
∆tP (x) X 
n+1/2 Ψi (ξ)Ψk (ξ)
 ∆tG(x) X 
n+1/2 Ψk (ξ)

en+1 = enk − h − J
  
k i2 12
< Ψ2k > 0 i=1  r (ξ)  < Ψ2k > 0 r (ξ)
  
 
i=1
 n+1/2 < Ψ2k > if i = k
 
 n+1/2 < Ψk > if i = k
   
hi3 J13
 
0 otherwise 0 otherwise
(1)

2
Kiran R S-FDTD - 1D Date: April 21, 2020

In the above expression,

n+1/2 n+1/2 n+1/2 n+1/2 n+1/2 n+1/2

where length1 and length2 are representing the boundaries of the dielectric slab and length3, the
end point of the domain.
Magnetic field equations do not have any change. This result is comapared with equivalent Monte Carlo
simulation with the dielectric slab permitivity realised from a Gaussian distribution and following is
the result.

3. In order to find the reflection coefficient from the time domain fields, we have to convert this
into its frequency domain equivalent. We utilize the fft() command in matlab for this. For getting
better ferquency resolution, the time domain signals need to be appended with enough zeros and thus
getting N point FFT using the fft(). Frequency shift is achieved using the matlab command fftshift().
Following is the S11 and S21 estimated at the reflection point ( @ 1.90m ) and the transmission point
( @ 2.15m ) respectively. We consider 0.4 GHz as the point to estimate the stochastic effect on S11
and S21.

3
Kiran R S-FDTD - 1D Date: April 21, 2020

4
Figure 1: Distribution of E
Kiran R S-FDTD - 1D Date: April 21, 2020

Figure 2: S11 and S21

In the figure 2 the bandwidth is truncated at 3σ (Only upto 1GHz even though 3σ is 1.35 GHz). This
is evident from the follwing derivation which is the fourier transform of the Gaussian pulse.

4. Time domain S-RBF- Frame the equivalent stochastic RBF code to estimate the distribution of
reflection and transmission coefficients at some points. In the above code PCE is applied and through
the Galerkin procedure the S-RBF is formulated. This is same as the code in (2) but suitable post
processing is applied. The results are compared with its equivalent Monte Carlo simulation. Result is
as in figure 3. This result is not matching exactly.. There is a small difference in the distribu-
tion in figure 1. This difference may be due to that. First we need to fix that.

5. Modify the old 1D-FEM code to include a dielectric slab of relative permitivity 4, and estimate the
reflection and transmission coefficients at some point. A Monte Carlo simulation is run on this with
the dielectric slab permitivity realised from a Gaussian distribution. FEM is run for a frequency in the
range 0.01GHz to 1GHz and transmission coefficient (at point 2.15m) is as shown in figure 4 below.
A comparison of the distribution of the of transmission coefficient at the point 2.15m with three different
diatribution profiles N (4, 0.1), N (4, 0.2) and N (4, 0.3) are shown below.

6
Kiran R S-FDTD - 1D Date: April 21, 2020

7
Kiran R S-FDTD - 1D Date: April 21, 2020

Figure 3: Distribution of Reflection and Transmission coeff

Figure 4: Transmission coeff comparison between FDTD-RBF and FEM

9
Kiran R S-FDTD - 1D Date: April 21, 2020

(b) Zoomed (c) Zoomed

Figure 5: Comparison by changing the variance

10