star grounding

If everyone in a building agrees that the lobby floor is the zero mark and measures their height from that one spot, no two people will ever argue about who stands higher. But if each measures from a different doorstep, the small differences pile up. Star grounding applies that idea to circuits: instead of letting ground wander from board to board, you run every ground connection back to one single common point, the centre of a star, so that every part shares exactly the same zero reference.

It helps because current flowing through a shared ground wire creates a voltage along it (Ohm's law acting on the wire's small resistance and inductance), and that voltage contaminates anything else referenced downstream of it. With a star, a noisy, current-hungry section, say a motor driver or a fast digital block, gets its own ground wire running back to the star point, so its heavy return current never flows through the ground a sensitive analog section is using. The big, dirty currents and the small, delicate signals are kept on separate spokes that meet only at the hub.

In practice you keep power ground and signal ground separate all the way back to the star point, often right where the power supply's reservoir capacitor sits. This is why mixed analog and digital boards split AGND and DGND and join them at exactly one place. Honest caveat: pure single-point grounding works beautifully at low frequencies, but at high frequencies the inductance of those long spokes becomes a problem of its own, and a solid ground plane, with many paths and very low inductance, wins instead. Real designs blend the two: star at low frequency, plane at high.