HFSS copper loss error

I am making 400MHz small loops and HFSS is not predicting the same copper loss I am measuring. My hardware shows 2 to 3 times the loss that HFSS predicts. I am not solving inside the copper. It seems like the skin depth approximation should be adequate, but is there something I am not setting up properly? I tried assigning a boundary condition to the copper surface and got the same answer, but do I need to assign a thickness or roughness to it? Does anyone have experience getting their modeled loss to match the predicted loss with small printed loops?

Measure return Loss vs F instead.. That will indicate if it is skin, transmission line effects or some other load effects.

You definitely should assign a roughness to it. See the attached for some guidelines.

pstuckey said:

You definitely should assign a roughness to it. See the attached for some guidelines.

Click to expand...

I would never have thought that would affect me at 400MHz, but since I see a worst case of 3u roughness in the Rogers' document, and skin depth is also about 3u, it now seems very plausible. I am presently using the outer layer of an FR4 substrate (plated up to about 3mils over original 1oz copper on outer prepreg layer I think). I will certainly experiment with some roughness values in the model as you suggest to see when it starts to matter, but do you have any idea what a typical value might be for this scenario?

Thanks!

For 4350 I've used 2.8 microns and it's worked pretty well.

Surface roughness is interesting at RF. For microstrip the lowest impedance path for the current is the bottom layer of copper (because it's closest to the ground layer), which is also where most of the roughness is. So even though the surface roughness increases resistance more of the current will still flow along that path.

pstuckey said:

For 4350 I've used 2.8 microns and it's worked pretty well.

Surface roughness is interesting at RF. For microstrip the lowest impedance path for the current is the bottom layer of copper (because it's closest to the ground layer), which is also where most of the roughness is. So even though the surface roughness increases resistance more of the current will still flow along that path.

Click to expand...

Well I tried a 3 micron roughness on all surfaces of the loop and the copper loss went up by 60%. It seems pretty clear this is the root of my problem and I will probably just tweak the value until my model matches the data unless I am able to find parameters for my copper.

By the way, there is no ground plane for this since it's just a loop antenna and the current flows on both sides equally. I remember that Sonnet allows the user to specify if the current is on both sides or just one side but I don't know how HFSS handles it. I guess if I am just fudging the value for the roughness it would be a moot point. I don't think the documentation for the skin depth mentioned anything about that issue.

Thanks again for the great advice.

tkb said:

I remember that Sonnet allows the user to specify if the current is on both sides or just one side but I don't know how HFSS handles it.

Click to expand...

That "current ratio" is used for thin metal simulation only, where one subsection represents the conductor loss of top AND bottom side. In this case, the software can not find out the skin current ratio between the surfaces, and the current ratio value is entered by the user.

For thick metal simulation in Sonnet, or for the surface currents in HSS, each conductor surface has its own subsection and there is no need for a "current ratio" factor.