Jon Martens of Anritsu and Al Neves of Wild River Technology used DesignCon 2012 to discuss high-confidence S-parameter measurement methodologies for 15 - 28 Gbpsec. The pair presented their findings during a half-day tutorial on Monday, January 30th.

The objective was to assist the signal integrity practitioner who needs to transition from 20 GHz bandwidth measurements to greater than 50 GHz bandwidths for connectors, backplanes, components, 3D EM modeling and correspondence, material extraction, statistical simulations, and SERDES characterization.

This tutorial dealt with improving confidence for >70 GHz measurements. With the heightened bandwidth requirements and fixture complexity, obtaining sufficient accuracy and repeatability for S-parameter measurements for 28 Gbps applications is very difficult. Starting with a macro understanding of uncertainty/repeatability mechanisms and characteristics of well-known calibration and de-embedding methods, this tutorial explored some of the better approaches and how to choose between them. Aside from the raw uncertainties in the S-parameter data and an understanding of how they affect end results (such as compliance results), time domain data and extracted models were also discussed. A series of case studies with practical measurement tips helped to emphasize key points.

Topics included: Relating VNA hardware to measurement error; error boxes and how they relate to practical measurements; overview of advanced calibration methods, including common (SOLT, SOLR, TRL, etc.) and newer (Partial information, Hybrid, Imposed constraints, etc.); error minimization and how error and uncertainty propagate through a system; de-embedding methods including time domain; relative sensitivities of various methods to mistakes/errors; calibration validation; adopting calibration approaches based on fixture design; and advanced practical topics related to TRL/LRM and common mistakes.