Issue link: https://iconnect007.uberflip.com/i/509724
28 The PCB Design Magazine • May 2015 curate calculations. All solvers obey the laws of physics and for the same data in should give the same impedance out. When choosing a solver you should look for one that permits you to model the layered substructure of the PCB, take into account the trapezoidal nature of the traces and where possible account for resin rich layers or areas of resin richness between differential pairs. The ability to run sensitivity analysis is a useful enhancement too. Modelled vs. Measured The aim of modelling is to get you as close to the desired result in the shortest number of iterations, ideally, right first time; it is worth keeping an open mind when looking at cor- relation of measured vs modelled impedance. Modelling predicts the uniform impedance of a uniform trace in a homogenous medium. It's good to ask yourself a few questions when faced with less correlation than you expect. First, nev- er assume anything. Look at both the measured and modelled and see if there are areas which need attention; often, a 5 or 10 ohm mismatch between modelled and measured can be com- prised of an ohm here an and ohm there. And you can suddenly find 5 or 6 ohms hiding from your initial assumptions. First, did you assume you made what you think you made? A microsection is a wise step IMPEDANCE CONTROL, REVISITED continues to confirm actual geometries. Second, look at the impedance trace—does it slope? This could be from trace taper or, on thin traces, trace re- sistance. Or does it dip at the start or the end, implying that the coupon design or the setting of test limits is not optimised? On the model- ling side, has the trapezoidal shape of the trace been considered? Have you run sensitivity analysis to see which trace parameter has the largest contribution to any error? Attention to the above makes you better prepared to opti- mise correlation on the next spin. I mentioned sloping traces: if not down to taper, this could be because you are using sub 75 micron lines on thin copper—and here you would need to consider measuring instantaneous impedance using a technique such as launch point extrap- olation. Occasionally Expect the unexpected From time to time you might experience re- sults that seem too consistent. This was the ex- perience of one of our fabricator customers who accidentally set a TDR to measure the end of the cable and not the coupon. Unsurprisingly, the TDR reported very repeatable results all within a fraction of 50 ohms. If your measurements are too consistent, you might be measuring the wrong thing; for example, another fabricator customer etched the PCB part number in the ground plane Figure 2: Impedance coupon generation software. the pulse