AWR eBooks

Best Practices for Efficient and Effective Planar EM Simulation 20 www.cadence.com/go/awr Passivity Measurement Figure 19 shows the passivity results for the filter shown in Figure 18. The passivity measurement is located under the Linear Measurements category and called PASSIVE. The measurement returns a positive number if the structure is positive at a given frequency. If it is negative, the S-parameters are non-passive and therefore the S-parameters are not physical. The figure shows that the measured results are non-passive at lower frequencies (the measurement works on any S-parameter file) and the AWR AXIEM results are passive. Passivity works on an S-parameter file with any number of ports and does not require the user to specify a specific port (as opposed to the power measurement discussed next). The value will change with changing port impedances, but as long as the port impedances have positive resistance, a passive result cannot change into a non-passive one. The algorithm used is technical in detail, but essentially looks at the eigenvalues of the matrix. The figure shows a result that is passive (positive) except for the measured data at lower frequencies. The expectation for a physically meaningful S-parameter file for a typical interconnect structure simulated in AWR AXIEM software is that the measurement will start off at about 0, and then gradually become more positive with increasing frequency. This is because the losses in the system will increase with frequency: conducting loss, dielectric loss, and radiation. If the structure has a resonance, for example the bandpass filter shown in Figure 19, the passivity would be expected to peak at the resonance because of increased conductivity loss due to increased current. For an antenna, peaked passivity would be expected at the radiation frequency of the antenna. The passivity number should never be negative. However, in typical S-parameter files it can happen. Figure 19 shows small negative values for passivity for the measured data; the violations are within the noise floor of the measurements. Figure 19: Passivity results for the filter shown in Figure 18 It should also be noted that a passive result does not mean the result is accurate. If an open stub is simulated and the result looks like a short circuit, the answer is obviously terribly inaccurate, yet it will still be passive. Also, the user should pay attention to the trend in the curve. It should be getting more passive with increasing frequency, except for resonance effects. If the trend is in the wrong direction, the user should be concerned, even if the results are still passive. In addition, users should remember that they are looking at analysis of numerical data. There are always floating-point arith- metic roundoff considerations, so a stub with no conductor or dielectric loss in a metal box (so there is no radiation) should have an S-parameter of magnitude 1. The user should not be surprised if the software gives 1.0001, which is non-passive. Small violations in passivity, especially at higher frequencies, can be relatively harmless. The biggest problems with passivity occur when the S-parameters are used in circuit simulations with HB. A non-passive S-parameter file produces energy, which can be a disaster in a nonlinear circuit simulation. But most circuits have some resistance in the models that can overcome a small passivity issue. The software manual states that passivity violations of less than 0.04 can be neglected. This number is a rule of thumb based on observation and should not be taken too literally. But it does point out that small passivity violations shouldn't be cause for discarding the entire simulation. On the other hand, once the violations are 0.1 or greater, the designer should be worried. Some users find the Enforce Passivity setting under the AWR AXIEM options tab of the EM structure, check it, and think they have solved the problem. This setting changes the S-parameters to keep the passivity at 0 when they would otherwise become non-passive. Unfortunately, there is no physical reason behind this process. S-parameters are supposed to represent the behavior of a physical structure, but this operation removes all physical meaning from the problem. The results are telling

• Cover