Issue link: https://iconnect007.uberflip.com/i/1359517
56 DESIGN007 MAGAZINE I APRIL 2021 designers, and EDA tools rounding nominal copper thicknesses for half-ounce copper to 0.7 mils (18 μm), 1.4 mils (36 μm), and 2.8 mils (71 μm). I'm not normally against rounding, but when you're rounding in the wrong direc- tion, it needs to be questioned. Table 2 lists actual thicknesses and conversions. Board thickness, too, will be affected. On a four-layer design, the difference may not be significant, but on a 20-layer design using one-ounce copper throughout and the wrong assumptions, your board thickness will be off by as much as 4 mils. I'm pretty sure the mechanical engineers, if no one else, would appreciate it if PCB designers worked with a sharper pencil. Conclusion Whenever I talk in person with SI con- sultants—people who do SI consulting for a living—I ask them this question: "Of all the smoke-jumping projects you've been brought in for where there were serious SI problems, how many of those projects had stackup issues?" So far, the only answer I've gotten back has been "100%." All told, the factors I've noted that impact impedance can put your designs near the edge of your tolerance targets. So, it's not just about having an expensive simulator, it's about feeding it the right parameter values, and that's much of what I've been focused on for the last few years with my company, Z-zero. Shoot me an e-mail if you have any war sto- ries on the subject. I always like hearing from designers on the front lines. DESIGN007 References 1. Happy Holden, e-mail correspondence, Dec. 2020. Bill Hargin is a veteran signal integrity engineer and the founder of Z-zero. Table 2: Copper foils are available in several thicknesses, measured by weight. The most common thicknesses used in multilayer PCBs are shown here, including nominal thicknesses, per IPC-4562A.