hfss out of memory

I am trying to model a transformer using two spiral inductors on two separate chips. This setup is commonly used for magnetic resonance coupling to send data between two substrates. I want to eventually export the results into a spice simulation; however, I am getting some execution errors when I try to simulate the setup.

I get:

[info] Simulation completed with execution error on server: Local Machine. (2:42 PM Mar 22, 2007)
[error] Initial mesh, process mesh3d : Unable to recover all the surface triangles. Additional information may be available in the "model_analysis.html" file in the project results directory.

Now, I imagine you would usually get this due to not enough mesh points or even due to discontinuities on surfaces that are supposed to be united or touching, however I cannot find any such problems with my model.

A note: The reason why I am using inductors of so many turns is because Q is not important in the coupled resonators. The magnitude of L is most important since you need adequate coupling between the two inductors. Furthermore, since the data is typically sent by inducing short current pulses in a positive or negative direction across the tx inductor, high Q inductors are prone to oscillatory behavior.

Can some one please help me out here!!

Hi, tompson,

Don't worry, "[error] Initial mesh, process mesh3d : Unable to recover all the surface triangles." might be caused by the large ratio between the layers' length/width and thickness. If you have very thin layers, please merge it into other layers. In simulation, the simpler substrate definition, the better.

Hope this helps,

Ruri

so, you're saying the substrate dimensions are wrong?

Hi, tompsom,

I did not download your hfss project before, therefore I only guessed that you might have the thin layer problem.

Now I've downloaded your project, I do not quite understand your project ... I'm just curious, does the transformer structure in your HFSS project extractly like the one in reality? Since I found in your project everything is doubled: two substrates,two oxides, two ground ....

Construct structure exactly the same with the one you will make measurement is extremely important for all EM modelings.

Good luck,

Ruri

tompson,

To my experience, HFSS cannot solve such a complex structure. Even if it can succeed in the initial mesh, It cannot reach the accuracy in the adaptive passes.

I have tried a much simpler structure (also a transformer) and HFSS resulted in "out of memory". My computer has 2GB memory, which is the maximum that HFSS can deal with, at least for version 9.2.

Best regard,
Leander

it is. I will also be testing structures with fewer turns and different dimensions. I have the fewer turn models working fine in HFSS. 10 turns is ok, but 12 does not. I'm guessing the adaptive meshing algorithm cannot deal with such a complex structure.

As I have mentioned, high Q and frequency responce characteristics are not imporant for this project.

The reason why everything is doubled is because I drew the base substrate first and then mirrored the model to save on construction time. I have actually written a matlab script that generates the entire structure from scratch depending upon dimensions you give it. There is also a script for inductors which I have also added. See attachment. Feel free to try it out!