input-protection
A connector feeds a power-input pin directly with no fuse or TVS in the path.
Remedy ¶
Put a fuse and a clamp in the path between the connector and the rail: a fuse for the sustained fault, a TVS for the transient.
What it means ¶
A connector net that reaches a power-input pin, directly or through up to three series parts, must have a protection device on the way, a fuse crossed or a TVS on the path.
Why engineers want it ¶
Every external fault arrives on one net, the power entry. Review checklists ask for a fuse (overcurrent) and a clamp (transient) between the jack and the first regulator; forgetting them is invisible until the wrong charger shows up.
Impact ¶
Shorted loads burn traces or supplies instead of opening a fuse; hot-plug and ESD transients reach the regulator input unclamped.
Scope note ¶
Since WS3-011 the rule walks from each connector net, following series pass elements
(R/L/ferrite/fuse, the reach primitive, 3 hops) to find a power-input pin, and the path is protected
if a FUSE was crossed to get there or a fuse or TVS hangs on any walked net up to it. Before the
walk, a series element split the net and the rule saw nothing, so a fuse-protected board passed by
accident, and an unprotected connector-bead-regulator path passed too. This upgrade closes that
false negative. Fuse OR TVS satisfies the rule; a per-design “which protections are required”
policy is rule configuration (WS3-006). Ground-named nets are skipped, and an unresolved external
net reports not-considered. If the board drives a net itself, that net also reports
not-considered. Its connector is then an outlet, powered by the board, rather than a power entry
(agni issue 1016). reverse-blocking-absent asks
the same question through check.BoardSource, and its page lists the shapes that count.
A current-limiting power switch also counts when it is the entry’s only load. An eFuse or a USB
power switch wired between the connector and the board is one. Its current limit gives the
overcurrent protection a fuse would give (agni issue 1248). Glasgow feeds its 5 V through a TPD3S014
that way. Only the part’s datasheet can say so, through a device_class that names a
current-limiting switch, because a regulator limits its current too and is no protection against its
own failure. The pass names the switch and cites the current limit printed in its datasheet. A
switch beside a regulator on the same net excuses nothing, since the regulator is still fed straight
from the connector. Without the datasheet the entry still fails, and the verdict names the IC it
could not read.
Query structure ¶
select connector nets whose reach finds an unprotected power input.
select N in nets where exists C in N.connections where class(C) == connector
and exists M in reach(N, 3) where exists P in M.connections, pin_dir(P) == power_in
and not (fuse crossed on path(N..M) or exists T on path nets, class(T) in {fuse, tvs})
Reads: component.class, net.attributes (external), net.names (the ground-name skip), on_net, pin.electrical_type, reach. Tier R.
For software readers ¶
A schematic is a graph whose components are nodes with named pins and whose nets are hyperedges joining pins that are wired together. The rule uses these concepts:
- A series (“pass”) element is a two-terminal part wired INLINE, like middleware in a request pipeline. Because a net is “everything directly touching”, an inline part SPLITS one logical connection into two nets, leaving a per-net rule like a function that can only see its own local scope.
- A fuse is a sacrificial circuit breaker, wired inline. “Is there a fuse between the wall plug and the machine” is a PATH question, like “does any middleware in the chain do auth”.
- A TVS diode is a surge protector hanging OFF the path to ground like a pressure-relief valve; it is protection on a path net, not a crossing.
- A ferrite bead is an inline noise filter, electrically transparent here, and the classic innocent reason a power path is split into two nets.
- A rail (VCC, GND) is a shared supply net like a global singleton; the walk never crosses into one.
The two conformance fixtures, drawn:

The walk’s crossing and stopping rules at a glance:
