How to calculate attenuation in a waveguide in Feko?

Vishakha Pandharpure_20595

Hi 

I want to estimate attenuation in a circular waveguide. I have modelled the waveguide as a hollow PEC cylinder, with waveguide ports defined at the two circular faces. A waveguide source is defined at one of the ports and the other port is defined as passive. I am requesting S parameter configuration to look at the S21. The operating frequency range is from 10 MHz to 110 MHz.  The cut-off freq. for the w/g is 87 GHz, which is much higher than the operating freq. Hence I expect a large value of attenuation. 

I am attaching 2 images. One of them is POSTFEKO S21 vs freq plot for different waveguide lengths (10 mm, 20mm, 30mm, 40 mm,100mm). According to literature the attenuation in a waveguide at freq. much less than the cut-off freq should linearly increase with length of the w/g. But it is not seen in the S21 plots. 

The other image is an overlay of

1. Feko S21 at 100 MHz for the above mentioned lengths of w/g. (green)

2. Expected S21 at 100 MHz for the above mentioned lengths of w/g. (red)           

I am also attaching the cadfeko model file for 100 mm waveguide. 

Is this the correct way of simulating a w/g to estimate the attenuation ? If not, what's the better way? 

 

Thanks 

Vishakha 

                          

Torben Voigt

Hi Vishakha Pandharpure,

The first thing I notice is the mesh of your model. Due to the size of the mesh elements (because of the low frequency range), the waveguide can no longer really be called round:

Try the attached model. Are the results then closer to expectations?

Best regards,
Torben

BrianWoods

I believe the problems here are related to:

- The mesh that is being used for the simulation does not represent the cylinder well
- One of the waveguide ports has the incorrect orientation.
- Particularly at the lower frequencies it is a good idea to use double precision (you will see warnings suggesting this in the solver output)

Ports:
The WaveguidePort2 in the model is oriented such that the port direction is facing to the outside of the cylinder. The orientation can be seen by zooming into the port and looking at the blue/red arrow. The correct orientation is shown below. This is fixed/adjusted using the "propagation direction opposite to normal" checkbox in the Port definition dialog.

Mesh:
Below is an image of the mesh that I see when opening the model you shared. Clearly this mesh does not represent the curvature of the cylindrical wall of the waveguide well. It is a good idea to look at the mesh in CADFEKO or in POSTFEKO to confirm that it look as expected!

I see you are using Coarse meshing and simulating at frequencies <200MHz in the model you shared. 

I would suggest using Standard meshing - and you can use the Advanced settings on the mesher to control how well the meshing follows curvature - but this does not work well for the model you shared at the frequencies you are simulating.

A better approach is to construct the cylinder in such a way that the mesher is forced to create at least a certain number of triangles around the circumference of the cylinder. One way to do this is by creating circle using a number of arc's and then sweeping these to form the cylindrical structure. I have quickly done this in the model I will attach. This gives a much more consistent mesh and captures the cylinder's curvature much better. In my model a variable (arc_angle) can be used to control the angle of the arcs used to generate the cylindrical face.

Mesh in initial model:



Mesh using a ruled cylinder:

 

Some side comments:

- For this S-Parameter calculation, it is not necessary to have the Standard configuration in the model or a waveguide excitation. Only the S-Parameter configuration is needed and only solving this configuration will reduce runtime.
- You can use adaptive frequency sampling to get a continuous frequency result rather than only individual frequency points. This is not always useful - but for wideband simulations, it helps a lot!
- You could leverage Symmetry to reduce the simulation time for this structure.

Vishakha Pandharpure_20595

@Brian Woods and @Torben Voigt 

Thanks a lot for your replies. 

I made the following changes in the model and now I am getting results as expected from the theory.

1. Instead of constructing a cylinder, I constructed an arc and using transforms built a cylinder.

2. Modified mesh at the two circular faces to finer size.

3. Used only S parameter config. 

4. Didn't use any w/g source.

5. Used double precision. 

Once again, thanks a lot for the valuable inputs.

 

Vishakha