Antenna Effect Repair
The idea
The antenna effect is a manufacturing hazard, not an operating one. Layers are fabricated bottom up. When a metal layer is plasma-etched, its wires collect charge; a wire already connected downward to a transistor gate, but not yet connected to the driver that will eventually discharge it (the driver's connection arrives on a higher layer, later), can only discharge through the gate's thin oxide, and enough charge ruptures it. The check is a ratio, evaluated per layer for every net:
collecting metal areathe area of the net's wire on the layer being etched (some processes use perimeter or side-wall area instead)connected gate areatotal gate oxide area the wire is connected to through lower layers at that point in fabrication- the ratio must stay below a process-specified limit; violations are per net, per layer
The recipe
Step 1 (check). After detailed routing, compute the ratio for every net on every layer using the fabrication order (only metal connected to the gate through lower layers counts).
Step 2 (repair by diode). For a violating net, insert a small reverse-biased diode connected to the net near the gate; during etch it drains the charge into the substrate, in operation it is off and adds only a little capacitance. Requires a free placement site.
Step 3 (repair by layer hopping). Alternatively, break the long run by jumping to a higher layer and back near the gate: only the short lower piece touches the gate during the lower layer's etch, and the long upper piece is etched after the gate is already connected to the driver's diffusion. Costs vias and a track on another layer.
Step 4 (iterate). Re-check; repeat until no violations remain.
Watch it work: one long wire, one gate
A driver, a long M2 run, and a gate. The walkthrough shows the etch charge, the ratio arithmetic against the limit, and both repairs.
Pros, cons, and the trade-off
Strengths
- Cheap check: an area ratio per net per layer, computed in one pass over the routing.
- Diodes are local and simple: one cell near the gate, no re-routing.
- Layer hopping needs no area: the repair happens in the routing itself.
Weaknesses
- Diodes need sites: in a dense region there may be no free placement site near the gate.
- Diodes add load: extra capacitance on the net, which matters on critical paths.
- Hopping consumes routing resources: vias and upper-layer tracks in a possibly congested region, and it can perturb timing.
The trade-off
Antenna repair is a small algorithm with an outsized place in the flow because its failure mode is silent: a violated ratio does not fail simulation, it fails on the wafer. The two repairs trade area against routing resources, and every sign-off flow runs the check because no other stage will catch it.
Is it actually used today?
The evidence for antenna checking and repair comes from open-source tools, the process design kits that define the limits, and the sign-off checklist.
- Open source: OpenROAD includes an antenna checker and repair module that computes ratios per net and layer and inserts diodes; the flow scripts run it after detailed routing.
- The process: every PDK (Process Design Kit) specifies antenna rules and ratio limits, and foundry sign-off requires a clean antenna check.
- Commercial: documented. Antenna checking and repair (diode insertion, layer hopping) are standard sign-off steps in every production flow.