0
\$\begingroup\$

I'm simulating a single microstrip half-wave resonator (1.8 GHz) in Sonnet and using ADS LineCalc tool for calculating the microstrip length but I'm getting a wrong number.

Substrate parameters (LineCalc):

  • Er = 10
  • Mur = 1
  • H = 0.4 mm
  • Hu = 3.9e+34 mm
  • T = 0.001 mm
  • Cond = 5.8e7
  • TanD = 0.002

With Z0 = 50 and E_Eff = 180º I get: W = 0.379400 mm , L = 32.239500 mm.

Substrate parameters (Sonnet)

  • Bottom dielectric layer

    • Thickness = 0.4 mm
    • Erel = 10
    • Dielectric Loss Tan = 0.002
    • Diel Cond = 0 S/m
    • Mrel = 1
    • Mag Loss Tan = 0
  • Conductor(copper)

    • Thickness = 0.001 mm
    • Conductivity 58000000.0 S/m
    • Current Ratio = 0.0

But I've found empirically that the length should be approximately 30.4 mm instead:

enter image description here

Why is this happening?

\$\endgroup\$
3
  • \$\begingroup\$ When you say you "found empirically" that the length should be 30.4 mm, does that mean you built a physical sample and tested it, or that you simulated it with a different tool? \$\endgroup\$
    – The Photon
    Commented Mar 20, 2015 at 0:52
  • \$\begingroup\$ And if you built it physically, what was the actual substrate material? \$\endgroup\$
    – The Photon
    Commented Mar 20, 2015 at 0:53
  • \$\begingroup\$ I mean that I changed the length of the microstrip until I saw the expected result in the simulation without any calculation. \$\endgroup\$ Commented Mar 20, 2015 at 1:58

1 Answer 1

1
\$\begingroup\$

I don't know the details of how Sonnet works, but some typical things that could make the EM simulation differ from the idealized transmission line model are

  • Fringing fields at the end of the stub. A microstrip stub doesn't act like a lumped model of a perfect open circuit. There is some equivalent capacitive or inductive termination. This termination will change the phase of the reflected wave, and so change the line length that gives a resonant response. The 3D solver will model this more accurately than LineCalc.

  • Launch conditions. Depending how Sonnet models the input port, which is essentially a transition from a lumped-element model to the 3D geometry, there could be a discontinuity at the input to the line. This would again produce a reflection which would change the length of the stub needed to see resonance.

    Similarly, if you were to build a real sample of this circuit, any reflection due to a discontinuity at the input would change the stub length needed to get an equivalent to a half-wave resonance.

  • Meshing. If you have chosen too sparse a mesh for the 3D simulation, its results will be inaccurate.

  • Boundary conditions. The choice of how the borders of the simulation region are modeled can affect the simulated result.

    Similarly, if you built a real circuit other conductive features like a lid on a package or an unrelated trace on the same substrate could slightly perturb the behavior of the stub.

As I've mentioned, there's several effects that will also make a real stub behave differently from your model, beyond which there's also the simple inaccuracy in manufacturing that means your W, H, and T parameters will never be exactly what you designed them to be. With this in mind, it's really best practice to make a design that is insensitive to small errors in any parameter; or, if necessary, to provide a tuning mechanism to allow the circuit to be adjusted individually to optimize the performance.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.