AWR eBooks

Best Practices for Efficient and Effective Planar EM Simulation 21 www.cadence.com/go/awr the designer that the AWR AXIEM simulator is having a problem that needs to be understood, and that the answer as it exists can't be trusted. There are occasional reasons to use the enforce passivity feature, but they are rare. It certainly shouldn't be used as a mechanism to avoid resolving problems with the EM simulator. The correct way to resolve the problem is to try to understand from where it is coming. The problem is usually not the mesh. The mesh approximates the current: a mesh that is too coarse won't give an accurate answer, but it usually won't violate passivity. Meshing is a concern in two situations. First, a finer mesh is needed if the results are inaccurate. Second, a too large mesh count can lead to computational issues and large memory usage. The most common place to look for passivity problems is the port definitions. Every port has a ground definition (this was discussed in Best Practice Tip #1 regarding autoports). Implicit ports have their ground at infinity, which can result in passivity problems. Explicit ports with ground straps can have de-embedding issues, which means calibration problems. The easiest way to check this is to disable de-embedding for the problem and rerun it. De-embedding can be globally turned off under the AWR AXIEM tab Options – Enable De-embedding Global, and, when unchecked, no ports are calibrated. If the answer is now passive, it appears there is a bad calibration. Unfortunately, de-embedding (calibration) is required for explicitly grounded ports because of the parasitic effects of the ground reference straps. Turning off the de-embedding is only useful for checking passivity violations. The next thing to check is AFS sweeping. Best Practice Tip #8 explains how a trace can be added to the passivity graph, which only shows the simulated points, not the interpolated ones (the _FSAMP frequencies are plotted). If the results are passive for these frequencies but have problems at the interpolated ones, AFS might be the problem. Energy Conservation Measurement The other measurement that can help analyze the situation is the energy or sum of powers measurement. This measurement is used to see how much power is absorbed in the circuit. Figure 20 shows the results for the filter in Figure 18. Imagine 1W of power is incident from Port 1. Several things can happen to it if the S-parameters represent a physical structure. First, it can reflect: |S11| 2. Second, it can exit out to Port 2: |S21|2. Third, it can be absorbed by the structure: 1 - |S11|2 - |S21|2, because of conservation of energy. The built-in measurement SUMPWR, located under linear measurements, provides the sum of the squares of any row of the S-parameter matrix, in this case |S11|2 + |S21|2. Figure 20 shows the results for the each of the two ports being excited. The vertical axis is the amount of power not absorbed by the structure, so a value of 1 means all power is either reflected or transmitted, and a value of 0 means all power is absorbed as either conductor loss, dielectric loss, or radiation. A value greater than 1 is non-physical. Figure 20: Results for the filter in Figure 18 showing the calculated power absorbed by the structure when a given port is excited The main difference between SUMPWR and PASSIVITY is that now one port at a time can be excited and energy violations could be missed if the wrong port is excited. PASSIVITY examines the overall S-matrix, and it does not depend on exciting a certain port. SUMPWR enables designers to probe the various ports individually, if they are trying to localize the problem. In Figure 20, the passivity problem in the measurements at the lower frequencies occurs when Port 2 is excited, not Port 1. The passivity and energy conservation measurements will not expose all S-parameter problems, but if there are issues, they can help point them out and can give the designer more confidence that at least the laws of physics aren't being violated. Finally, these measurements are very quick to run and can be applied to any S-parameter file.

• Cover