controlled-impedance routing

Controlled-impedance routing means deliberately building a trace to a specific, exact impedance — most often 50 ohms, or 100 ohms for a differential pair — rather than letting it be whatever the layout happens to make it. For slow signals a trace is just a wire and its impedance does not matter. But when a signal changes fast enough, the trace starts behaving like a transmission line, and if its impedance is not controlled, the signal bounces back off the mismatched ends as reflections that corrupt the data. Controlling the impedance keeps the signal clean.

The impedance of a trace over a ground plane is set by simple geometry: the trace width, the height of FR-4 between trace and plane, and the dielectric constant of that FR-4. Make the trace wider or move it closer to the plane and the impedance drops; make it narrower or further away and it rises. So to hit a target like 50 ohms you compute the right width for your specific stack-up — and crucially, the trace must run over a continuous, unbroken reference plane the whole way, because if the plane underneath has a gap, the impedance jumps right there and you get a reflection.

Why it matters: any fast interface — USB, HDMI, Ethernet, DDR memory, high-speed clocks — specifies a trace impedance you must hit, often within ten percent, or the link simply will not work reliably. You ask the fab to guarantee it (they tweak the stack-up and test coupons), keep differential pairs tightly matched in length and spacing, and never let the reference plane break under the trace. The honest note: this lives at the boundary of RF and transmission-line theory; here we only set the trace up correctly. Below a few hundred MHz on short traces you usually do not need it at all — adding it everywhere is wasted effort.