Last month, I commented on one of Bill Schweber's blogs. My remarks were about the (lack of) models for RF connectors.
At that time I remarked:
It is interesting that in the world of passive and active components, it is expected to provide SPICE models and other tools so circuit designers can include them in a higher level simulation. On the other hand, for RF connectors this is not the case. In general, RF connectors are assumed to be 'perfect,' but they frequently are not.
Here I want to expand on the point of what design information you can derive from modeling connectors.
In my past roles in engineering leadership at a few companies, I was a big proponent of simulation in RF systems. There have been steady and ongoing advances in RF simulation software. Companies and products like AWR — Microwave Office, Comsol — RF Module, and Ansys — HFSS, among others, have opened up simulation to many more engineers.
With that, I'm seeing many more problems today than at the start of my career. Over those 30 odd years, the frequency of all systems has increased, both for analog and digital. Today, simulating a 25 Gitb/s digital signal on a backplane requires treating the signal path as a transmission line to understand losses, crosstalk, immunity, and other design factors. On top of that, practically all systems have interconnect, and it isn't remotely possible to treat the connectors as anything but extensions of those transmission lines. That means that connectors can introduce insertion loss from mismatch, and, more worrisome, reflect significant power back up the line into the system.
Most of my work with simulation tools has been in the antenna design domain. Use of RF simulation tools is becoming fairly common in the antenna industry, but even there connectors tend to be overlooked. Extending simulation to include the connectors requires very precise mechanical models, and very accurate material properties for the dielectrics. Many antenna designers ignore connectors in their designs, assuming they are “perfect” in terms of impedance match and discontinuities. At higher frequencies, those assumptions can break down.
I contacted Randy Bancroft of Randwulf Technologies to get some examples for this blog. Bancroft is an IEEE member and professional engineer, as well as past colleague of mine (the latter not being much of a resume point). Bancroft said, “In the past generally connectors were not simulated. The connector was omitted because modeling of physical connectors takes up memory. This often worked well if a connector and its transition to a transmission line had very small impedance discontinuities.”
Of course, the interesting cases for me were the ones that didn't work. Bancroft described one design which used an SMA connector (see an example here) to inject the RF signal to a circuit board which also had an antenna implemented in the PCB metal traces. Later, the customer had an application requiring an N connector (see an example here), and the larger impedance discontinuities caused enough mismatch to reduce the bandwidth of the antenna/connector system. A panic of trial and error ensued to find a board design that would work. Accurate simulation could have identified the problem in advance, as well as been used to find a solution.
A more recent case encountered by Randwulf similarly started with an SMA connector on a PCB, and the customer needed to change to a TNC connector configuration (an example of a TNC is shown here). The customer planned to use a custom-made connector on the board side. Their customer (the end user) selected the mating (TNC male) connector. When the board-launch connector was evaluated in HFSS, it was found to contain impedance discontinuities that were larger than desired at the frequency of the system. Figure 1 shows the analysis and results.
(Source: Adapted from material provided by Randwulf Technologies)
Concerned about the performance of the design, Randwulf obtained the mating connector from their customer and added that into the model. Figure 2 shows the physical model of the board, TNC female edge-launch connector, and TNC male connector.
(Source: Randwulf Technologies)
Not surprisingly, the return loss results showed that at 3GHz, the return loss was -13.3dB (about 1.55:1 VSWR [voltage standing wave ratio]). In the particular application, this was not good enough, so an improved connector design was developed by Randwulf. Figure 3 shows the return-loss simulation for the original and improved designs.
(Source: Randwulf Technologies)
Once the connector manufacturer made the changes, it was verified that the product could be assembled with either the original SMA connector or the new TNC connector and meet requirements. Although the “sledgehammer” approach could have been used to change the board geometry to get a better match with the TNC connectors, this would have required two boards, and the TNC board would have been useless with other connectors. By using simulation as part of the analysis, it was possible to both identify the source of the problems as well as design a better solution. Have you been involved with designs where simulating the entire structure might have saved time or avoided future problems?