An always-on buffer ended up powered from the switched rail. How do you find it and fix it?

Short Answer

Power-aware simulation shows an always-on signal going to X when the domain shuts down, and check_mv_design (ICC2) reports an isolation violation on its buffer tree, because the buffer supply is less always-on than its loads. fix_mv_design -buffer (ICC2) repairs it, either by swapping to a dual-rail buffer whose backup pin sits on the always-on rail, or by moving the single-rail buffer into a nearby always-on voltage area. Then legalize and re-check.

Technical Explanation

• Cause: a single-rail buffer placed in a shutdown VA takes the switched rail as its supply, often from an ECO buffer insertion.
• Symptom: isolation enables, retention controls or wake requests die when the domain turns off, so the domain cannot wake or restore.
• Detection: the buffer supply is less always-on than its load supply, which check_mv_design (ICC2) reports as an isolation violation.
• Swap fix: fix_mv_design -buffer (ICC2) converts the buffer to dual-rail; it needs usable dual-rail cells and a reachable secondary strap.
• Move fix: -move_within_x / -move_within_y (ICC2) move it into a nearby VA; that VA needs logical or physical feedthrough enabled, or ICC2 issues MV-1096.
• After a move: the tool does not check space, congestion or driver location, so run legalize_placement (ICC2) and re-check.

# [ICC2]  icc2_shell
check_mv_design
fix_mv_design -buffer -from U_PC/iso_en_cop -report_only -verbose
fix_mv_design -buffer -from U_PC/iso_en_cop -move_within_x 10 -move_within_y 10 -verbose
legalize_placement
check_mv_design
# [VC LP]  vc_static_shell
check_lp -stage pg
report_violations -app LP

What To Check

• Every buffer on an always-on net is dual-rail or sits in an always-on VA.
• Dual-rail buffers have their backup pin tied to the always-on rail.
• Target VAs for moved buffers allow feedthroughs.
• Power-aware simulation keeps the net valid through shutdown.

Healthy, Suspicious & Hard-stop Results

• Healthy (illustrative): The fix_mv_design Before/After table shows Has Isolation Violation Yes then No, with zero mismatches.
• Suspicious (illustrative): Violations are fixed, but 6 buffers were moved and legalization pushed two of them back toward VA_COP.
• Hard stop: The After column still shows an isolation violation, or check_lp -stage pg reports a floating backup pin such as PG_PIN_UNCONN.

What The Interviewer Is Testing

• Can you explain why a buffer takes the switched rail by default?
• Do you know both ICC2 fixes and their preconditions?

Follow-up Question & Model Response

"Why prefer moving the buffer over swapping to dual-rail?"

Candidate Model Response: A single-rail buffer is smaller and needs no secondary PG connection, so moving it into an always-on area saves area and routing. It only works when a suitable VA is within the move distance and allows feedthroughs. If the library has no usable dual-rail buffers, the move is the only automatic option. When neither works, reroute the net around the shutdown area.