Once the cable design and construction anomalies have been identified and solved, there is the issue of the test equipment. In a “what came first, the chicken or the egg” conundrum, there is always the nagging doubt about the test setup. Are the test leads good? Is the setup that was used successfully a couple hours ago still in calibration? Someone used the VNA to test at Ka-Band and you changed it back to W-Band; is the recall calibration still valid? The luxury of the time-honored diagnostic method of swapping items when there is a performance anomaly is not an option when there is only one of everything.
CONNECTORS
Choosing the cable design, by definition, narrows the choice of qualified connectors and what connectors to offer, i.e., only those designed in-house, standard connectors offered by outside vendors or a combination. Since this is still a small, evolving market, the choice is to work with all applicable connector suppliers to have the broadest offering.
The size of the component parts and the tolerances needed—while avoiding a skewed tolerance stack-up that results in interference among the component parts—is the shorthand tale of why working at these frequencies is difficult. Depending on the cable dielectric, the wavelength at 110 GHz ranges from 0.083 to 0.100 in. (2.1 to 2.55 mm) and the tolerance of the pin contact below the reference plane is 0.002 in. (0.051 mm). What did not matter at Ka-Band is now the difference between passing and failing.
Are you pursuing a domestic customer base or a global one? If global, then make everything RoHS (Restriction of Hazardous Substances) compliant—a European Union directive—because major customers will demand it, and there is no upside to maintaining two different versions in inventory. Using RoHS solders and processes requires a higher level of assembler attention and skill. When terminating the connector, heat might be a concern with the ferrule/housing if excessive heat transfer causes the dielectric to change or grow. If the design of the connector is point-and-shoot, the length of the center conductor that is inserted into the center contact may change the response curve of the connector. Here, real-time X-ray is invaluable.
There are some common problems when designing a connector in-house or troubleshooting with a trusted outside supplier. Initially, when making the transition from simulated data provided by CST, HFSS or home-brew software to empirical data, the cable or connector will not perfectly match the model. Be prepared for many iterations of design improvement, as the empirical data informs the design software. Whether using in-house design expertise or relying on a trusted external vendor, the ability to precisely machine component parts (e.g., a ferrule or a dielectric bead) is quite useful and cuts down the development time dramatically, as shown below.
The difference in equipment (e.g., VNA and related adapters) can make a difference in the shape of insertion loss and the VSWR or return loss curves. What you see may differ from what your customer or supplier sees due to the age of the calibration, age and condition of the adapters or settings used on the VNA. The importance of listening, graciously receiving feedback, providing samples and trying to re-create a problem are opportunities to learn.
