Test Results Phase Shifter
Schiffman Phase Shifter Test Results
The Schiffman phase shifter designed first was the model used in the systems for manufacture. A separate component of the phase shifter and the reference line was manufactured as well in order to compare the design to the manufactured outcome. The manufactured phase shifter also incorporated the SMA launch that was designed for this project, as well as for future uses. A modification had to be made post-manufacturing to solder one of the legs of the connector to the metallic surface it was sitting on. This decision was made based on initial data that showed significant loss in the component in transmission and reflection, and most importantly in phase. Two measurements were taken of the component: gated and un-gated. Gating allows the network analyzer to bypass the connector launches and take the data of just the component, much like the simulation. In essence, the start and stop points of the measurement are being set in time to a point beyond the launch. The launch was designed well, but proved to inject a lot of noise into the circuit, as indicated by Figures 1 and 2 below.
Figure 1: Un-gated phase (left) and reflection and transmission (right) of the Schiffman phase shifter and reference line.
Figure 2: Gated phase (left) and reflection and transmission (right) of the Schiffman phase shifter and reference line.
The gated results are the best ones to analyze since they do not include the lossyness and noise introduced by the connectors. Table 3 below shows a summary of the results.
Table 1: Summary and comparison of the simulated and manufactured results for the Schiffman phase shifter, and the re-design.
As can be seen from Table 1 above, the reflection and transmission of the simulated and manufactured circuit matches well, with an average error of 3.27% for reflection. The phase difference, however, shows a significant difference of almost 3deg at each point. The shape of the phase is different as well, comparing the graphs of the measured phase and the simulated phase. From the tolerance analysis in Table 1, it seems that phase is most drastically affected by dielectric thickness and mode suppression via location. The shift upward of 3deg could be attributed to a slightly thicker Rogers dielectric, and even an offset of the holes. Overall, the results show much promise in the design. Even though the phase is not exactly what was simulated, the shape of the phase matches well with the simulated result. The next step would be to analyze the characteristics of the actual model compared to the HFSS result, and try to have the HFSS simulation match the actual component. The variations in the HFSS model to make it look like the manufactured model can be inversely applied to create an offset and a more desired result upon manufacture.