Simulation Improvement Results 180deg
Table of Contents
0/180 Degree Coupler Improvements Page
This page depicts changes that were made to the Knochel coupler post-manufacturing for potential improvements.
The overall goal is to evaluate and improve upon the results obtained by Knochel and Mayer. This cannot be achieved by referring to the implemented circuit because the Z6 tuning stub was removed since its added bandwidth was deemed unnecessary. This investigation was performed with the Z6 stub still attached with some key results. All simulations are summarized below for continuation in future works.
Problems and Concerns
First, with the method of development from Ansoft Designer to HFSS, it was discovered that conversion directly from Designer and into HFSS for stripline components resulted in severe degradation of the signals. For example, when this circuit was tuned to approximately -25dB to -30dB return loss in Designer and exported to HFSS, the return losses immediately jumped to approximately -8dB to -10dB over the full bandwidth.
In addition, by attempting to combat this result and obtain an ideal model in HFSS, all return losses could only be suppressed to approximately -15dB over the full bandwidth, instead of the -20dB over most of the bandwidth achieved without the Z6 stub. An investigation into the fields showed that significant coupling between the Z6 stub and all bordering transmission lines (Z1, Z2 and Z3) had occurred. Different models were tested to determine the ideal shape of the component to minimize coupling, but no notable solution was found.
The next series of simulations sought to optimize the dielectric stackup. Here are a few sample simulations. Be sure to right click on each link and select "Save Target As..." Then save the file as a *.HFSS file. It should be stressed that before downloading any of the attached files, ensure that HFSS Version 11 is on your computer.
Each of the files above represents a possible stackup solution. They do not represent a completely tuned circuit. The purpose behind each of the above circuits is to investigate the potential change in the stackup, not to present a final solution given a permanent change in the stackup.
From all of the simulations, it was determined that a 20 mil stripline stackup of a Rogers 5880 with a dielectric constant of 2.2 yielded the best simulation. The reason for this is that with this stackup combination, the width of all transmission lines in the circuit was reduced significantly in size. For example, the width of Z3 was reduced from 110mil to 41mil. This provided Z2 an extra 30 mil of length that is unaffected by the junction with Z3's width, thus providing greater individual transmission line behavior and improving the quality of the system.
From this limited testing, it was determined that any dielectric stackup with a lower dielectric constant and a thinner composition yielded better results than the current 120mil stackup with Rogers 3003 dielectric.