Dielectric resonator antenna fed by coaxial line-FDTD Matlab

[grendhell]

dielectric resonator coaxial probe

Hi, I'm trying to implement a DRA fed by a coaxial line using FDTD in Matlab.
I'm following the implementation exposed in :
A coaxial probe feed model for FDTD
Microwave and Optical Technology Letters
Volume 34, Issue 3, Date: 5 August 2002, Pages: 193-198
Riku Mäkinen, Ville Kangas, Janne Lahtinen, Markku Kivikoski

The code seems to work but I have problem with stability. I have to choose a smaller time step than the Courant limit one, sometimes 10 times smaller. Obviously this increase the computational time.
I don't know if instability can be removed or it's just a consequence of BCs between cable and DRA.
I have noticed (but I don't know if it's a general rule) that instability arises when the DRA permittivity constant is higher than the coaxial cable one's.
Does anybody know if it's possible to increase stability and how?
Does anybody know another more stable way (closer to the Courant limit) to couple coaxial line and DRA?
Thanks a lot

grendhell

 

[rrumpf]

dielectric resonator antenna fdtd

I don't think it is possible for me to say for sure where your instabilities are coming from, but I can list of few things it may be that you may not have considered.

1. Algorithm
First, it is very important to consider the staggered nature of the Yee grid. This places the field components at physically different locations so they will be out of phase. Field components in the same grid cell may even reside in different materials. Your code could have instabilities if you don't consider this with your source.
Second, the fields are also staggered in time. This has to be considered when formulating your update equations. Each term in the equation has to be defined at the same instant in time. If this is violated, your code will go unstable, especially over long simulations.

2. PML
The PML boundary condition is excellent because you don't have to worry about angle, frequency, polarization, or inhomogeneous boundaries and it will still work pretty well. If you don't choose your PML parameters correctly, this can cause your code to go unstable. You may want to track the maximum field value in your model to try to figure out if the fields are exploding in the boundaries first.

3. Metals
If you are using a Lorentz model or something like that for metals, you can have "faster than vacuum" phase velocity and you should modify your Courant stability condition accordingly (i.e. reduce time step). Try turning off your dispersion models and use a very large value of permeability and see if this fixes your problem.
Another good practice is to assign the metals to points on the Yee grid that place tangential electric fields on the surfaces of the metal regions. I have noticed instability when I do not do this properly, but I don't think I can fully explain why this happens.

4. Grid Resolution
If your grid resolution is insufficient, this can cause your problem to go unstable as errors accumulate during simulation. For long simulations, it is a bit more important to use sufficiently high grid resolution. You mention you simulation is more unstable when you raise a material value so perhaps this is part of your problem. Try computing your grid resolution this way. First determine the highest frequency you are exciting your problem with. Second, calculate the corresponding wavelength in the region with highest dielectric constant. Third, divide that number by something larger than 10. Fourth, you may have to reduce your resolution further to resolve fine structures. Fifth, use this as your grid resolution.

Hope this helps!!
-Tip

 

[grendhell]

courant limit matlab

Hi Tip, thanks for your reply. Let me expose with more details my simulation following in particular your points:

1 Algorithm
I'm using different permittivity constant "inside" each different material and the arithmetic mean on the boundaries (I also tried geometric and harmonic means). Electric fields and voltage are update at time steps n while magnetic fields and current at time steps n+1/2.

2 PML
The simulation environment is a semi-space delimited by an infinite ground plane at z=0. UPML BCs are then used on the other boundaries. I've already tested the UPML BCs by running several simulations even for 100000 time steps and reflection error is negligible.

3 Metals
The only metals present are the infinite ground plane and the internal coaxial conductor used to excite fields inside the DRA. Both are assumed PEC, hence the internal fields are zero.

4 Grid resolution
Resolution is calculated as you stated. I use dx,dy,dz = 1/20 of the shortest wavelength.

All the materials are assumed perfect and loseless.

I hope this can help to figure out what might be wrong....
Thanks a lot

grendhell

 

[grendhell]

matlab rf circuit

Hi, I've done more tests, and instability arises when the antenna is present: if I use the same permittivity everywhere then the code is stable. Probably this lead to a kind of interface BCs problem.
The point is that I'm already using permittivity average on the interfaces. I will try to increase the resolution, but if this is the solution, I don't see any improvement in computational time terms....
Any idea?

cheers,
grendhell

 

[pappumaity]

Re: dielectric resonator antenna fdtd

Hi, I'm trying to implement a DRA fed by a coaxial line using FDTD in Matlab.
I'm following the implementation exposed in :
A coaxial probe feed model for FDTD
Microwave and Optical Technology Letters
Volume 34, Issue 3, Date: 5 August 2002, Pages: 193-198
Riku Mäkinen, Ville Kangas, Janne Lahtinen, Markku Kivikoski

The code seems to work but I have problem with stability. I have to choose a smaller time step than the Courant limit one, sometimes 10 times smaller. Obviously this increase the computational time.
I don't know if instability can be removed or it's just a consequence of BCs between cable and DRA.
I have noticed (but I don't know if it's a general rule) that instability arises when the DRA permittivity constant is higher than the coaxial cable one's.
Does anybody know if it's possible to increase stability and how?
Does anybody know another more stable way (closer to the Courant limit) to couple coaxial line and DRA?
Thanks a lot

grendhell

can u send me this paper which u refer it here---i am getting that at all-----please------------


can u help me--i need this paper also---------------

R. K. Mongia and P. Bhartia, “Dielectric resonator antennas—A review
and general design relations for resonant frequency and bandwidth,” Int.
J. Microwave Millimeter-Wave Eng., vol. 4, pp. 230–247, July 1994.


can u send me it--------at-sdpt123_maity@yahoo.co.in

or u can upload it here also-----please-i need it-----------help me------

Added after 54 seconds:

I don't think it is possible for me to say for sure where your instabilities are coming from, but I can list of few things it may be that you may not have considered.

1. Algorithm
First, it is very important to consider the staggered nature of the Yee grid. This places the field components at physically different locations so they will be out of phase. Field components in the same grid cell may even reside in different materials. Your code could have instabilities if you don't consider this with your source.
Second, the fields are also staggered in time. This has to be considered when formulating your update equations. Each term in the equation has to be defined at the same instant in time. If this is violated, your code will go unstable, especially over long simulations.

2. PML
The PML boundary condition is excellent because you don't have to worry about angle, frequency, polarization, or inhomogeneous boundaries and it will still work pretty well. If you don't choose your PML parameters correctly, this can cause your code to go unstable. You may want to track the maximum field value in your model to try to figure out if the fields are exploding in the boundaries first.

3. Metals
If you are using a Lorentz model or something like that for metals, you can have "faster than vacuum" phase velocity and you should modify your Courant stability condition accordingly (i.e. reduce time step). Try turning off your dispersion models and use a very large value of permeability and see if this fixes your problem.
Another good practice is to assign the metals to points on the Yee grid that place tangential electric fields on the surfaces of the metal regions. I have noticed instability when I do not do this properly, but I don't think I can fully explain why this happens.

4. Grid Resolution
If your grid resolution is insufficient, this can cause your problem to go unstable as errors accumulate during simulation. For long simulations, it is a bit more important to use sufficiently high grid resolution. You mention you simulation is more unstable when you raise a material value so perhaps this is part of your problem. Try computing your grid resolution this way. First determine the highest frequency you are exciting your problem with. Second, calculate the corresponding wavelength in the region with highest dielectric constant. Third, divide that number by something larger than 10. Fourth, you may have to reduce your resolution further to resolve fine structures. Fifth, use this as your grid resolution.

Hope this helps!!
-Tip

can u help me--i need this paper---------------

R. K. Mongia and P. Bhartia, “Dielectric resonator antennas—A review
and general design relations for resonant frequency and bandwidth,” Int.
J. Microwave Millimeter-Wave Eng., vol. 4, pp. 230–247, July 1994.


can u send me it--------at-sdpt123_maity@yahoo.co.in

or u can upload it here also-----please-i need it-----------help me------

 

[pappumaity]

can u help me--i need this paper---------------

R. K. Mongia and P. Bhartia, “Dielectric resonator antennas—A review
and general design relations for resonant frequency and bandwidth,” Int.
J. Microwave Millimeter-Wave Eng., vol. 4, pp. 230–247, July 1994.


can u send me it--------at-sdpt123_maity(at)yahoo.co.in

or u can upload it here also-----please-i need it-----------help me------