Model What You Measure and Measure What You Model

Over the last couple of posts, I’ve chatted about the inevitable comparisons that are made between competing simulation products, but there is a far more common comparison that is made to simulation; that is to Test & Measurement. The overwhelming majority of designs conceived in the imagination of engineers eventually make it to the prototype phase and physical test. Once there, the comparisons can get dangerous.?

Countless times over my career, I have heard the following or similar phrase: “HFSS is wrong. It doesn’t match my measurements.” I always reply with the title of this post or ask the question if their model matches their measured prototype. Inevitably, there are some discrepancies between the two which, when corrected, show exceptional agreement between HFSS simulated data and measured data.?

One simple example of this disagreement is the definition of material properties. Many times, simple data sheet values are accepted for dielectric or conductor properties, when frequency dependent models or surface roughness models need to be considered. This is especially true for broadband applications. Another example is the characterization of a microstrip or stripline circuit. In HFSS or planar engines, it is widespread practice to simply place a port on the end of a transmission line and assume that the energy is injected perfectly. In practice, there are always coax-to-transmission line transitions that will always, and sometimes drastically, impact impedance matching to the circuit.??

After twenty-plus years of using electromagnetic tools, my #1 rule for success is to model what you measure, and measure what you model. The closer you can make these, the better your correlation will be. For an example of how to use ports to better match broadband performance, check out the following webinar.?