Ground

Two things at once, and a design needs both. Ground is the reference, so saying a rail is at 3.3 V is shorthand for saying it is 3.3 V above ground, and there is no other sense in which a single point has a voltage. Ground is also the return. Current leaves a regulator through a rail, and ground is the path it takes back.

So when you read a schematic, ground touches nearly everything. A rail feeds the parts that want its voltage; ground is on almost every part on the board. So it is drawn as a scatter of little triangles rather than as a wire. On a real board it is not a wire at all but a copper plane, and the drawing’s twenty separate symbols are one net.

The engine treats it as its own thing rather than as a rail at zero volts, in two places worth knowing. Ground names come from a vocabulary of their own, separate from the rail vocabulary, so a project that spells it AGND or 0V declares that alongside its rail patterns. And ground stops the series-reach walk. Imagine a walk tracing connectivity through passives. It would arrive at ground on its first hop and from there reach the entire design. That is true and useless, so ground counts as shared distribution rather than as a path.

Most rules that read it ask whether it is claimed correctly. dl-power-pin-mistyped catches a pin named GND whose symbol never typed it as a supply, sitting alone on a stub because nothing merged it into the plane. power-tap-conflict fires when one net carries two design-wide names. For ground that usually means two spellings. A human reads them as obviously the same, and a tool must not. unconnected-pin and single-pin-net pick up the part that never made it there at all.

The course teaches it in chapter 1, which uses it as one end of the two-terminal decision procedure, and in chapter 5, which explains why an output pin deciding a net’s voltage means connecting it either to a rail or to ground.