Hello

I am trying to investigate the effects of cutouts in the top side ground of a stripline using Momentum. The layer stack used is:

open boundary,

top ground (metal strip),

dielectric,

signal conductor (metal strip),

dielectric,

closed boundary (perfect conductor).

The reason for using explicit metal strip and not closed boundary for the top ground is to be able to draw holes or slots into the ground, which are to be investigated.

If I fill up the top ground (no cutouts), I expect that the resulting (ideal) stripline has parameters as predicted by LineCalc. However, looking at results from Momentum simulation, the characteristic impedance is too high, i.e. 60 Ohms instead of 50 Ohms. (In contrast, if I simulate the same transmission line but with the top ground modeled by a closed boundary, the line parameters obtained from Momentum simulation agree very well with those from LineCalc.)

I found that when increasing the mesh density to beyond 45 cells per wavelength, the line parameters gained from Momentum approach those obtained from LineCalc. I think that the problem lies in the fact that with a coarse mesh, the current density in the top ground plane (metal strip) is not approximated correctly. Going to finer mesh increases the transversal resolution of current density, giving more accurate impedance information. (Simulation settings: Edge mesh 0 um active, thin layer overlap extraction and mesh reduction are also active, always in microwave (not in RF) mode.)

As I read the help to the mesh setup, a feature called thin layer overlap extraction has been introduced to prevent the problem I mentioned above. However, whether it is enabled or not, the cells of the mesh in the top ground are much wider than the width of the signal conductor.

My main question: How can I force Momentum to generate a denser mesh in the top ground in the vicinity where it overlaps with the signal conductor?

An alternative would be to map the top ground layer as metal slot (instead of strip). What could be expected from this approach?

Any help on this is greatly appreciated.

Regards,

Daniel

I am trying to investigate the effects of cutouts in the top side ground of a stripline using Momentum. The layer stack used is:

open boundary,

top ground (metal strip),

dielectric,

signal conductor (metal strip),

dielectric,

closed boundary (perfect conductor).

The reason for using explicit metal strip and not closed boundary for the top ground is to be able to draw holes or slots into the ground, which are to be investigated.

If I fill up the top ground (no cutouts), I expect that the resulting (ideal) stripline has parameters as predicted by LineCalc. However, looking at results from Momentum simulation, the characteristic impedance is too high, i.e. 60 Ohms instead of 50 Ohms. (In contrast, if I simulate the same transmission line but with the top ground modeled by a closed boundary, the line parameters obtained from Momentum simulation agree very well with those from LineCalc.)

I found that when increasing the mesh density to beyond 45 cells per wavelength, the line parameters gained from Momentum approach those obtained from LineCalc. I think that the problem lies in the fact that with a coarse mesh, the current density in the top ground plane (metal strip) is not approximated correctly. Going to finer mesh increases the transversal resolution of current density, giving more accurate impedance information. (Simulation settings: Edge mesh 0 um active, thin layer overlap extraction and mesh reduction are also active, always in microwave (not in RF) mode.)

As I read the help to the mesh setup, a feature called thin layer overlap extraction has been introduced to prevent the problem I mentioned above. However, whether it is enabled or not, the cells of the mesh in the top ground are much wider than the width of the signal conductor.

My main question: How can I force Momentum to generate a denser mesh in the top ground in the vicinity where it overlaps with the signal conductor?

An alternative would be to map the top ground layer as metal slot (instead of strip). What could be expected from this approach?

Any help on this is greatly appreciated.

Regards,

Daniel

1.) Where did you get the impedance from that does not coincide with LineCalc's results? Is it the Z0 in the dataset resulting from your momentum simulation? This one is calculated for the purpose of calibration of the single ports using a static cross section solver. It often differs quite a bit from the full wave simulation result and IMHO should only be regarded as a rough estimate. LineCalc in contrast uses closed form approximations derived in the 1950s. A better way to figure out your actual line impedance is simulating a piece of straight line, terminated with 2 ports of exactly the impedance the line is supposed to be (Z0). Then plot s11 or s22 on a Smith chart. Of the arc starting from the origin (or even circle), determine the impedance Z1 where it intersects the real axis again. Your actual Z is then sqrt(Z0*Z1). If you met your intended Z0 you will only see a dot of course.

2.) I suppose you defined your ports as single ports. Thus you don't have explicit ground references, which works fine with two closed boundaries. With your strip metal layer acting as upper ground you need to connect it to your ideal lower ground plane, e.g. using vias. Otherwise it is left floating which may give quite peculiar results, depending on its physical dimensions.

3.) Talking about dimensions of your (finite) ground plane, of course it has to be sufficiently wide as compared to your narrow signal strip. LineCalc assumes an infinite ground.

4.) The accuracy of impedance calculation is affected by the mesh density perpendicular to the axis of propagation much more than by the density along your line (which is determined by the cells per wavelength). Try to use "transmission line mesh" option for the signal strip and the finite ground, as well as edge mesh for the signal strip.

5.) Defining the upper ground plane as slot metal should lead to a much leaner mesh and generally work well. But you are constrained to zero thickness ideally lossless ground metal then. When defining cutouts, be aware that Momentum seems to give nonphysical results when there are ports defined underneath such a cutout. I tried to simulate such cutouts above a stripline spiral inductor once and ended up with apparent amplification. But there I certainly had some sort of ill-defined geometry. However, I also moved back to a strip metal ground plane and finally succeeded. But again not without some tweaking, vias and the like.

Hope this helps somehow...

Best regards, eltz