Issue link: https://iconnect007.uberflip.com/i/396835
October 2014 • The PCB Design Magazine 15 Step 1: S-parameters measurement and pre-qualification The first step in the systematic validation process is to make S-parameter measurements with VNA up to 50 GHz for 28–32 Gbps data rates and pre-qualify them for further analysis. For the reference, the CMP-28/32 kit includes S-parameter in Touchstone format measured for all structures by an expert with SOLT cali- bration up to the SMA connector. We will start with the example of formal pre-qualification of these measured data. The process is described in details in [1, 6] . We start with preliminary estima- tion of the passivity, reciprocity and causality metrics computed for discrete and bandwidth limited datasets. All metrics conveniently range from 0 (bad) to 100 (excellent) and further sub- divided into bad, questionable, acceptable and good intervals as described in [6] . If all metrics fall into acceptable (blue) or good (green) inter- vals, we proceed and estimate the final quality metrics with the rational approximation or ra- tional compact model. Models measured with high quality allow rational approximation with high accuracy—the root mean square error of such approximation can be used to construct the final quality metric [1, 6] . The end result of the measured S-parameters quality evaluation in Simbeor software is shown in Figure 2. All models measured by the expert pass the final quality test (Quality column). Note that the frequency-continuous approx- imation of the discrete Touchstone models with the rational compact models is 100% causal by definition, because of the passivity is ensured by the software from DC to infinite frequency in this process. Such models are usable not only for the original Touchstone model quality eval- uation, but also for further validation in time domain—fast and accurate computations of TDR/TDT and eye diagrams. Step 2: Broadband material model identification After S-parameters are measured and pre- qualified, the next step is to identify broadband dielectric and conductor roughness models. The model identification with generalized modal S- parameters (or GMS-parameters) is the simplest and most accurate way to do it [7-9] . It requires S-parameters measured for two line segments with different length. Line of any type with any impedance can be used. It also does not require modelling of the connectors and launches. The SINk OR SWIM AT 28 GBPS continues feature Figure 1: CMP-28/32 channel modeling platform with 27 structures to benchmark software with stack- up and broadband dielectric and conductor roughness models identified in Simbeor software [5] .