Issue link: https://iconnect007.uberflip.com/i/258943
20 The PCB Design Magazine • February 2014 quency point 5 . This is the simplest way to iden- tify the conductor roughness model. 3) If the dielectric model is not available, identify dielectric and conductor roughness models separately. In addition to two line seg- ments with rough copper, make two or more transmission line segments with flat rolled copper on the same board. First, use segments with flat copper to identify parameters in di- electric model following the procedure in Fig- ure 1. Then use the identified dielectric model for rough segments and identify the conductor roughness model following the same procedure Figure 1, but for the roughness model. This is the simplest way to separate loss and dispersion effects in conductor surface roughness and di- electric models. 4) If the dielectric model is not available, identify dielectric and conductor roughness models simultaneously. It can be done with multiple line pairs with different widths of strips in each pair (narrow, regular and wide strips made of the same rough copper for instance). Dielectric model and conductor roughness model parameters can be optimized simultane- ously following the procedure in Figure 1, until differences of GMS-parameters for segments of all strip widths reach the stopping criteria. The resulting dielectric and roughness models will be usable for a given range of the strip widths. Though the procedure is the most complicated and may lead to multiple possibilities (ambigui- ties). Overall, the material identification proce- dure described here is the simplest possible. It needs measurements for 2 t-lines with any geometry of cross-section and transitions. No extraction of propagation constants (Gamma) from measured data is required. The extraction of Gamma is difficult and error-prone. Also, no de-embedding of connectors and launches is required. De-embedding of PCB structures is usually difficult or even impossible due to in- homogeneity of dielectrics and manufacturing variations. The approach needs the simplest numerical model—only propagation constant has to be computed for a given cross-section and with the material models to identify. No 3D electromagnetic models of the transitions PCB AND PACkAGING DESIGN UP To 50 GHz continues feature Figure 2: A CMP-28 board stackup and view. Dielectric parameter data from manufacturer also shown.