Power Sequencing in Multi-Rail Systems

Modern boards commonly need three or four rails — a core voltage, an I/O voltage, an analogue rail, perhaps a higher rail for interfaces. Many devices specify the order in which those rails may appear, and both the rise and the fall need thinking about.

Why order matters

Integrated circuits contain protection structures between their supply pins and their inputs. If an input is driven while its supply rail is still low, current can flow through paths never intended to conduct. At best the device latches into a state that only a full power cycle clears; at worst it is damaged. The same applies in reverse on shutdown, where rails collapsing in the wrong order can leave inputs powered from a rail that has already gone.

Common requirements

• Core before I/O, or the reverse — the device datasheet decides, not convention.
• A maximum time between rails, so one is not left waiting for another.
• A limit on how far rails may separate in voltage during the transition.
• A defined shutdown order, which is easy to overlook because it is harder to observe.

Achieving it

Where rails come from separate supplies, use the inhibit or enable input that many units provide, driven from a supervisor or from a monotonically rising earlier rail. Hold the system in reset until every rail is valid. Then verify it on a scope: capture all rails on one trigger, at switch-on and at switch-off, and at both the fastest and slowest input conditions. Sequencing that is correct at nominal input can invert at the extremes.

Talk to us about supplies with inhibit and power-good signals for sequenced systems.