An EM violation on a power strap: what are the fix options, and what does each cost?

Short Answer

EM fails when current density in a wire or via exceeds the limit, so you either spread the current over more metal or send less current through that segment. Widening the strap, adding a parallel strap, enlarging the via array, and moving current sources or taps all work, but each costs tracks, capacitance, or placement change. Pick by whether the violation is in the wire or in the vias, and by what the surrounding routing can give up.

Technical Explanation

• Current density is current divided by cross-section. Doubling width roughly halves density for the same current, but the tech file EM limit can depend on width, so check the actual limit at the new width rather than assuming twice the capacity.
• Widening costs routing tracks next to the strap and may hit maximum width or slotting rules. It is the simplest fix when the neighbouring tracks are free.
• A parallel strap on the same or another layer splits the current. It needs tracks too, plus vias to tie it in, and its benefit depends on how evenly the current divides between the paths.
• Via arrays fail per cut. If the violation is on the via, adding cuts or a larger array fixes it without touching the wire. Check that the landing metal has room for the enclosure.
• Spreading current sources means moving high-current cells or adding taps or bumps closer to the load so less current flows through the stressed segment. It costs placement change or package change, but it can fix several violations at once.
• Know which mode failed. perform emcheck (RH) checks AVG, RMS and PEAK, default all. PG straps usually fail average-current limits, while RMS limits relate to self-heating. RedHawk does not run EM during static or dynamic analysis by default unless ENABLE_AUTO_EM (RH) is set to 1.
• Static violations are listed in the .em.worst (RH) report under adsRpt/Static in decreasing order above the 100% default limit, top 1000 by default. In ICC2, analyze_rail -voltage_drop static -electromigration (ICC2), with the Apache tech file named in rail.tech_file (ICC2), produces the rail EM map, typically after chip finishing.

What To Check

• Whether the violation is on the wire or on a via, and in which EM mode.
• EM ratio of the worst segments and how many share the same current path.
• Free tracks beside the strap and room for extra cuts.
• IR drop after the fix, since the same change lowers resistance.

Healthy, Suspicious & Hard-stop Results

• Healthy (illustrative): Worst EM ratio 87% after widening the strap and growing the via array to 3 x 4, with margin kept for temperature uncertainty.
• Suspicious (illustrative): All segments pass but three sit at 97 to 99%, one hot-corner re-analysis away from failing.
• Hard stop: Any segment above 100% in the signoff mode and temperature, or a fix that moved the violation to the next segment down the same path.

What The Interviewer Is Testing

• Does the candidate distinguish wire EM from via EM?
• Weighing each fix by its routing and placement cost.
• Knowledge of AVG, RMS and PEAK modes and which applies to PG straps.

Follow-up Question & Model Response

"Why can fixing one EM violation create another further down the strap?"

Candidate Model Response: Current collects as it flows toward the tap, so the segments nearest the tap carry the most. If you widen one segment or add a parallel path that ends partway, the current returns to the original strap at the join and the next segment now carries the full load with no help. The fix has to extend along the whole overloaded stretch, or relieve the current at the source.