Issue link: https://resources.system-analysis.cadence.com/i/1326178
Best Practices for Efficient and Effective Planar EM Simulation 18 www.cadence.com/go/awr a subset of the frequencies. For example, in Best Practice Tip #7 a power amplifier is set up to have AFS work only on the range of frequencies between 2.0 and 2.2GHz, and 20 frequencies could be asked for in this range. The frequencies of DC, 4.0, 6.0, 8.0, and 10.0GHz are not included in the AFS sweep. IF AFS is having issues converging, it is possible to turn it off and see if the results change. Usually the designer chooses fewer frequency points if this is the case. Sometimes an error or warning message will be issued if AFS is having difficulty. Note that the error criterion of -30dB is a relative error. 30dB means the S-parameters will not change to three digits, not that the result is only valid to an S-parameter value above -30dB. (It is very rare for the default value of -30dB to be changed.) Turning off AFS leads to many more simulations: An easier way to see if this is causing problems is to include a trace in a graph that shows at what frequencies the S-parameters have been calculated using AFS. An example is shown in Figure 18, where a hairpin planar filter is simulated from 9 to 16GHz. Figure 18: Hairpin filter on a PC board (top left), graph showing the frequencies used for AFS (top right), and menu showing FSAMP frequencies used by AFS (bottom) The upper right picture shows the results for the insertion loss, and the brown points are the frequencies that were simulated using AFS. The trace was created by using the FSAMP frequency points for the X-axis sweep instead of the FPROJ frequencies. The built-in variable FSAMP lists the AFS points, while FPROJ lists the simulation points for the circuit simulator. The trace is then set to show only discrete points with no connecting line in the trace options. This method can help in deter- mining if AFS is having issues when a simulation result is incorrect. If the results at the simulated points are accurate, but the interpolated results are suspect, perhaps AFS is the problem. For example, if the circuit is exhibiting non-passive behavior (to be discussed in more detail in Best Practice Tip #10) but at the calculated points appears well behaved, perhaps AFS is failing. On the other hand, if the passivity issue is still there for the calculated points, AFS is probably not the issue.