Accurately characterizing coaxial connectors requires a high-frequency network analyzerespecially at frequencies above 1 GHz. Unfortunately, complications can arise when testing adapters or cable assemblies that have connector ends that do not attach directly to the network analyzer's test ports. When testing non-insertable devices, careful measurement techniques and calibration methods are required to achieve the highest accuracy.

A non-insertable device is a component with interconnecting ends that differ from the analyzer's test-port connectors (Fig. 1). It could have the same gender or connectors with differing families, requiring the use of "between-series" adapters to mate the connectors. An adapter adds its own electrical contributions to the test setup, such as insertion loss and return loss. Because of the inability to remove these errors during calibration, the device under test (DUT) will introduce errors in the measurement plane in both cases. Newer network analyzers, however, come with software/firmware that offers enhanced calibration methods for measuring non-insertable devices.1 In fact, many of today's analyzers have instructions built into the calibration menus. These instructions describe the setup and calibration steps that are needed to include error correction for these types of devices. The menus also include instructions for fixed mechanical calibration standards and electronic-calibration (Ecal) modules.

Methods For Enhanced Calibration
Zero-length thru: The most proficient and fastest VNA calibration method is called the "Zero-Length Thru" method. If testing for an insertable device, the test ports are calibrated individually and then connected directly together during the "transmission" sequence.2 No delay, loss, capacitance, or inductance compensation is required during this process. The calibration plane is now at the test-port connector's electrical reference plane. The DUT is inserted between the test ports. The device's response is displayed on a graphical overlay, which shows the measurement result in logarithmic or linear format. Depending on the VNA used, other formats can be simultaneously added to the display.

Adapter removal: Although this method produces a very accurate measurement, it is labor intensive and requires connecting and disconnecting adapters from the test-port cables. Should any of them come loose, it introduces measurement errors into the calibration plane. Calibration states must also be saved for each adapter on every port in the analyzer's memory. In addition, it is necessary to "recall" the states after the calibration steps are complete and the data is joined together using a complex algorithm generated by the analyzer's software.

Swap equal adapters: This method is by far less labor intensive and has been proven to produce repeatable results. Both test ports are calibrated using the open short and load standards (or Ecal module). One of the port adapters is then removed and a high-quality, phase-equal adapter is attached for the transmission portion of the calibration procedure. Traditionally, this method uses an adapter that matches the DUT's exact phase and delay characteristics. Yet this type of adapter is not always available, which means that it must be custom-built in most cases. To include this type of adapter in the calibration set, the stored set is re-written. In addition, the adapter's delay and other characteristics must be known before re-writing the calibration set. A better approach is to have test-port cables that allow one to change the test-port connectors (Fig. 2). After calibrating both ports, the calibrated adapter is removed from one of the test-port cables. A high-quality phase-equal adapter is then attached in its place for the transmission step. After the through portion of the calibration procedure is complete, the original calibrated port adapter can be re-installed. The loss and reflection from the uncalibrated phase-equal adapter (used as the through) is negligible, adding almost no error into the measurement.

Between series: This example presents the most challenging of all methods when calibrating for repeatable error-corrected measurements. The connectors are usually different at each end of the DUT. With no practical means to calibrate this type of device, the engineer is left to decide which calibration kit to use as the reference standard for one end while ignoring the other.

By carefully following the setup and calibration steps, one can correct for non-insertable or between-series devices. A couple of features to highlight are the intermediate-frequency-bandwidth (IFBW), additional time domain, and automatic-port-extension (APE) settings. The IFBW on PNA models from Agilent Technologies, for example, can be adjusted to 10 Hz or below, which allows the noise floor to be lowered significantly. The time domain menu allows physical length to be set based on the devices' velocity and can also display the one-way response versus the traditional two-way response. The units of length can be selected to display millimeter, inch, or feet. Another feature is the APE function, which allows an adapter or cable jumper to be added onto one or both of the calibrated ports. Applying APE after the initial calibration will automatically compensate for the added length by an extrapolation of the curved-fit algorithm containing loss, delay, and other characteristics of the extension adapter or cable jumper. The adjusted calibration reference plane is shifted to the open end of the adapter. Using a lab-grade short or open is all that is required for this feature to correct the intrinsic errors in the added port extension(s).

The right microwave test equipment and accessories are mandatory for connector manufacturers who want to lead the ever-changing wireless communications industry. Applying these enhanced calibration techniques correctly assures repeatable error-corrected measurements each time.

1 Agilent Application Notes 1291-1B and 1287-3, www.agilent.com
2 Commonly known as the "thru" sequence of the calibration set up