set mesh frequcny

时间：03-23 整理：3721RD 点击：

I am using Ansoft ensemble EM simulator.
How can I set the mesh frequency ????
I try vaious mesh frequency the conslucions are different ...
why ?????????????

does anybody teach me ???????????

You can try to set the frequency toward the middle point of the bandwidth; anyway it is probably more convenient choose the frequency sligthly shifted to the upper side of the bandwidth.

Regards

Hi,

In this program you should put mesh frequency higher than the upper bandwidth. That is the requirement. How high? The higher is teoretically better, but the calculation time increases. You can try also adaptive mesh to make program increase the mesh cells until there would be given difference between results of previous simulation and actual one, but it takes time also. Everything depends how much time are you going to spend on your simulation and how fast your computer is.

Greets,
Redi.

Dear sir:
My middle frequency is 5GHz.
Do I have to set my mesh frequency equal to 5GHz ???
I tried two mesh frequencies,5GHz and 20GHz.
The conclusion are not the same.
The most important is the simulation time.
Mesh frequency,5GHz,takes about 10 minutes.
Mesh frequency,20GHz,takes about 7 hours.
Do you have this experience ???

thanks

It's difficult to establish in advance the best compromise between accuracy and time simulation. As a rule of thumb, I think that it's possible to try the adaptive mesh function starting with a frequency sligtly higher than the middle to avoid long time simulation but having a medium accuracy as start.
When the difference between results of previous simulation and actual one is within a small value you have a simulation with a good accuracy.

In the specific case 20 GHz takes 7 hours and that limits the possibility of other simulations. Therefore I suggest to choose a lower frequency, for example 6 or 7 GHz, and then it's possible to set an adaptive mesh to see the convergence of the results with several simulations (probably 5, 6 or more).
If you have doubts about the results, it is possible to do an other adaptive mesh starting from the last simulation.

Regards.

Hi myem,

Have you tried Momentum or IE3D? I think their speed is much faster than Ensemble. Of course, you need to compare the results for your final design.

Mesfet+

Hi mesfet,

Do you have some experience in using Momentum and IE3D in comparisom to Ensemble? I have been using Ensemble and I found it quite accurate, but the others not. What is their accuracy in comparison to measurement?

Greetings,
Redi.

Hi,
I think the emsemble's solution is accurate and if we can modify the simulation model in HFSS and to solve it the solution is more accurate!I had set up a model in emsemble and HFSS,the HFSS's solution is more closed to the test results.

Hi myem,

Talking about the differences, first of all HF$$ is fully 3D program and En$emble is 2.5D. In HF$$ you can simulate 3D structures and in En$emble only layered structures with infiinite dielectrics, finite metalisation and vias. 0.5D means that currents flowing in vertical direction are assumed to not change in vertical direction. The method of calculation in both antennas is different either, HF$$ uses finite element method and En$emble method of moments. Concluding, Ensemble is suitable for microstrip structures, like filters, patch antennas, microstrip lines and you can do this in quite simple way. In HF$$ you can realize everything which En$emble does and of course a lot more, like waveguides, horn and helical antennas, etc., but it is more complicated to setup the solution correctly.

Greets,
Redi.

Hi

just one more thing to add to above post. z directed curent is modelled and the fields that are produced by this current as well in 2.5 D simulators. However, the current that exists in real vertical elements, i.e. the current that flows perpendicular to z is not taken in the account and the fields that this current creates as well. For thin dielectrics with electricalli short vias this works OK, but for electrically thick sublstrates the accuracy detoriates. One of the possible ways to get better approximation is to devide the thick substrate in several layers as the x-y current is accurately taken into account on interfaces. This will of course increase memory requirement and simulation time. For really 3D problems you should use full 3D solver.

flyhigh

Hi,
I think post paper is important for simulation of thick substrate.And in Ensemble there is 3D model simulation.But I do not know about it,could anyone introduce it?
Thanks!
caocao

Fundamentally, the Ensemble doesn't include 3D model.
when you bought Ensemble software you can assign you want to
buy 3D model package.
But I thought that the 3D effect of HFSS is better than Ensemble.

Hi myem,

It is very interesting what you are saying. I never heard about this 3D package. There is nothing about it on An$oft page. How is it called? Does it really exist?

Greetings,
Redi.

Getting back to the question--

I have been told that about 10% above the interested operating band is a good place to start with the mesh freq iterations. You don't ever get a solid answer to this question. One of the reasons is that the answer changes for the complexity of the problem. The bottom line is You have to decide when the mesh is tight enough for your needs. Unfortunately this may mean many iterations.