Some questions about HFSS

[baby_1]

Hello
I have designed a mono helical antenna in HFSS with these settings:
Solution typeriven
a Helix with these settings

and "analysis" option:


and frequency sweep:

and here is my design


but it takes more than 3 hours to analyze it.(a core2quad cpu 2.5Ghz and 4Gig RAM)

1)why did it too long take?
2)how can decrease the analysis's time?
3)if our design doesn't convergence ,what does it cause?

 

[Felis_Silvestris]

Nothing seems wrong with your setup that I can see. My guess would be the model size is large compared to the wavelength and that the model uses a lot of tetrahedra. Right-click on Analysis --> Setup1 and select convergence to see how many tetrahedra points were used for your model.

What are you using for a radiation surface?

 

[baby_1]

Thanks Felis_Silverstris for your replay
Here is the Convergence:

I use a cylindrical shape for radiation.And as I see it sweeps other frequencies that aren't listed in my sweep frequency window or I didn't defined them(such as .225 .375)

 

[Felis_Silvestris]

Yes, that's a lot of points for a regular PC and that will take time to finish. I use a server farm for jobs of that size and larger. Usually, I'll use 8 cores and 16GB to 64GB of memory for jobs requiring hundreds of thousands of tetrahedra.

One thing I noticed is that it says "N/A" at the 6th adaptive mesh. Did you stop and start this model? You may want to run it again starting with no previous mesh. Of course that could take hours - so don't delete your data - copy the model (just the *.hfss file) and run that in a separate directory. You can also delete a mesh in a solved project - but like I said - you may not want to delete hours of simulation time at this moment.

- - - Updated - - -

One more thing: PML may be better and may even reduce required mesh size. The radiation boundary needs to be 1/4 lambda (or further way) and fields should strike the radiation boundary perpendicular. I think most people agree that the PML is better for antenna work.

Also, the interpolation sweep picks its own frequencies - you may have observed that.

 

[baby_1]

Thank you so much for your complete explanation.
if we set a small value for "Maximum number of passes" and it doesn't converge , could you tell me what it is the result of our simulation? can we trust the result?

 

[Felis_Silvestris]

Well, I would say it boils down to this: how can you achieve usable data from HFSS with a minimum number of points or tetrahedra? You can take your existing model, reduce the mesh size and compare the data to previous HFSS data with a tighter convergence. You can do that by reducing the maximum number of passes or increasing the delta-s criteria (yours is currently set to s=0.02 - try using 0.03). Try it to see if it works for you. But, at some point you'll just lose accuracy - and that will turn into wasted time. But, it's still worth a try. Compare models using less points to those using more points. I wouldn't use a single comparison, but a few to get an idea. If you feel you can live with the error (or can't tell the difference) - the continue with less points.

Also, there are steps that could reduce the model size. Although, I don't see symmetry right off the bat from your model - people have cut their physical model space by half using symmetry. If you're solving the inside of metal then you'll get a warning that this will use a lot of meshing points. For your project don't solve inside metal (that's the default setting anyway). It's physically large enough that you won't need to solve inside metals or other highly conductive objects. A PML boundary might be better for reducing the number of tetrahedra. Occasionally, imported objects can create unnecessary meshing points due to imperfections in those imported models (such as small hard to see gaps or slight misalignment). You can view the mesh in HFSS and ~sometimes~ you'll see something obvious. But, I can see you've generated the helical using one of the scrips within HFSS - and did not import anything - so I really wouldn't expect that to be a problem.

What is the size of the air cylinder with the radiation boundary? Can you change that to PML and reduce the volume?

Also, I've heard of software packages designed for antenna simulation only - they may be far more efficient compared to a more generalized finite element method.