Sonnet standard and internal co-calibrated different results

Sonnet standard and internal co-calibrated ports giving different results

Hi,

I am simulating the interconnection between four transistors as explained in the attached pdf file. The transistor is in the form of BEC (Base Emitter Collector). In the file you will find that I made a connection between the collector and emitter through a bridge like structure (Port2, Port3, Port4 and Port5). Under the bridge, there is a connection for the two bases (Port1, Port6, and Port7). I made two different simulations. In the first simulation, I assigned the port6 and port7 as co-calibrated internals ports with sonnet box as the ground plane. While in the second simulation, I extended the two ports (port6 and port7) in bringing them to the sonnet wall to make them standard ports and defined a proper reference plane to make it look similar to the first simulation. However, I got totally different results in both cases. I don't know which solution is correct and ideal for this type of problem. In the second case while de-embedding the reference plane, does the sonnet also remove coupling due to other ports for the required reference plane ?

Thanks a lot. View attachment Connections.pdf

Re: Sonnet standard and internal co-calibrated ports giving different results

The simulated results are attached here.

This is a difficult one, because the "true" effect of the DUT is rather small compared to possible artefacts from port calibration. The difficulty is that coupling between different port groups is not removed. You should contact Sonnet support to discuss the best strategy, they will explain in detail what is removed (and what is not).

volker@muehlhaus said:

This is a difficult one, because the "true" effect of the DUT is rather small compared to possible artefacts from port calibration. The difficulty is that coupling between different port groups is not removed. You should contact Sonnet support to discuss the best strategy, they will explain in detail what is removed (and what is not).

Click to expand...

I simulated the same problem in ADS momentum to see the difference. See the attached pdf file. My frequency of interest is 110 GHz to 250 GHz. And the minimum width of the via is 0.19 um. For me the ads mesh looks less denser for such problem. Could you please comment on the momentum em setup that I have selected for the current problem in the attached file ? The results here are different than sonnet. I will ask the sonnet technical support.View attachment connectionsADSmomentum.pdf

In ADS, you can also define an absolute dimension for mesh size (µm) instead of cells per wavelength. This gives you more control over the mesh for RFIC work, where dimensions are much smaller than wavelength.

From the picture, it is difficult to see if your ports are in the right layers. You can check this with 3D view: switch ports display on in the menu, this will show arrows for the ports.

volker@muehlhaus said:

In ADS, you can also define an absolute dimension for mesh size (µm) instead of cells per wavelength. This gives you more control over the mesh for RFIC work, where dimensions are much smaller than wavelength.

From the picture, it is difficult to see if your ports are in the right layers. You can check this with 3D view: switch ports display on in the menu, this will show arrows for the ports.

Click to expand...

The ports are on the right layer. From where can I define the absolute dimension for mesh size (um). I could not find this option. Can we use this option to set dimensions for each layer separately ? But it will increase the simulation time a lot.

searchforknowledge said:

From where can I define the absolute dimension for mesh size (um). I could not find this option. Can we use this option to set dimensions for each layer separately ? But it will increase the simulation time a lot.

Click to expand...

Here it is ... global or per layer, see marked tabs.