RF PCB design gurus: Which 2.5D solver can handle thin dielectrics?

[rf-fil]

I'm trying to extract s-parameters from a PCB. I am evaluating Sonnet right now. What an awesome tool. It's very user friendly. It's super accurate, as far as I can measure, with a thick PCB. Say 1mm.

However, when I try to model a thin PCB, say 0.1mm, the results are way off to the point of being unusable. Does any other EM tool / 2.5D solver handle thin dielectrics?

Btw, I work at UHF, at few hundred MHz.

(Just to be clear, a tool based on circuit theory works well too, but has limitations. Circuit theory tools use manual schematic entry and the PCB has to be modelled as sections of transmission lines or microstrips. What I'm trying to achieve is to get something less labour intensive and more accurate than a circuit theory tool.)

[hendorog]

I'm no RF guru, but for my 2cents, could this be due to the copper not being modelled accurately enough?

It seems logical to me that copper thickness and roughness will have more of an effect on the result as the dielectric layer becomes very thin.

I seem to remember that Sonnet lite (free version) has limitations on the modelling of copper layers.

[rf-fil]

It turns out I have forgotten to model the solder mask  . The Sonnet output now matches what I measured within 1%.

Btw, the actual PCB will be so thin (for EM purposes) because it's a multi-layer board. So this is just the distance of the top layer to the ground plane beneath it. The PCB is much thicker overall.