Lesson 14: Final Words

Final Words

The Role of Simulation:

The first key to simulation is good models. It is hardly even a good idea to design with lumped element parts when real models (generally, vendor data) are available. It isn't a great idea to design with ideal transmission lines when microstrip models are available. The agreement between what is designed and what is built is often what was neglected in the models.

Even with good models there is always the chance that some parasitic has been neglected- a connector's inductance, a neglected transmission line length, a stray coupling. Part of the role in simulation is to go back and try to add details back into the model and try to figure out why the measured data for your device does not agree with the simulation data of your model. Those who do not learn from their mistakes.

Even if all of the stray capacitances and inductances have been included in the model, it is still possible for their to be major discrepancies due to the statistical nature of the world. All parts have some tolerance, with parts commonly found with as small as 0.1% variation to as large as 20% variation. Performing yield runs on your circuit can help set your expectations. An even better way to use statistical analysis is to provide another metric for evaluating circuits. For any given matching problem there are several topologies that could work. Statistical analysis can help select which is the best circuit.


When I match at my job, using real components and realistic transmission lines, I generally have to use a mix of lumped elements and transmission lines. Sometimes I can use filter techniques for broadband matching, but often I can't use ideal models for my lumped elements because of the parasitics- I have to use measured data provided by vendors (like the ATC or Coilcraft libraries). In my experience I have found a better solution adding the “real model” matching elements one at a time, moving towards the desired match while keeping below some Q-curve, than I have from designing a matching network assuming ideal components and then trying to transition to real models.

Copyright 2010, Gregory Kiesel