How do you size power-switch cells against an IR-drop target?

Short Answer

Run a voltage-drop analysis with RedHawk Fusion, give each switch cell its on-resistance with set_power_switch_resistance (ICC2), then run size_power_switches -max_irdrop (ICC2) with your drop target. The tool swaps switches for same-footprint cells of a different Vt to meet the target, and skips any switch it cannot improve. You trade drop against switch leakage and area, so check both after the swap.

Technical Explanation

• What sets the drop: switches work in parallel, so drop is roughly domain current times single-switch Ron divided by switch count, plus the grid drop after them.
• Trade-off: lower Ron cuts drop but costs area and off-state leakage; a higher-Vt switch leaks less but conducts worse.
• Step 1: run analyze_rail -voltage_drop (ICC2) to get the worst drop on cells powered through each switch.
• Step 2: set each candidate cell Ron from the switch model file with set_power_switch_resistance (ICC2).
• Step 3: size_power_switches -max_irdrop 0.05 (ICC2) swaps cells to meet the target; the guide example prints -max_irdop, so check the man page for your release.
• Limits: swaps only between cells with the same footprint and pin shapes; if no cell meets the target and Ron, that switch is left alone.
• What breaks: too much drop across the switches slows every cell in the domain, so timing signed off at nominal voltage fails in silicon.

# [ICC2]  icc2_shell
analyze_rail -voltage_drop static -nets {VDD1p0_SW VSS}
set_power_switch_resistance HEADER_HVT_X4 0.015
set_power_switch_resistance HEADER_SVT_X4 0.010
report_power_switch_resistance -verbose
size_power_switches -max_irdrop 0.05
analyze_rail -voltage_drop static -nets {VDD1p0_SW VSS}

What To Check

• Worst drop behind the PD_COP switches before and after sizing, against the 50 mV target.
• Every switch cell in use has a Ron in report_power_switch_resistance -verbose (ICC2).
• How many switches the tool left alone because no same-footprint cell fit.
• Switch-off leakage after any swap to a lower-Vt cell.

Healthy, Suspicious & Hard-stop Results

• Healthy (illustrative): Worst drop behind PD_COP falls from 68 mV to 44 mV against a 50 mV target.
• Suspicious (illustrative): Target met, but 60 switches moved to a lower-Vt cell and switch-off leakage rose 30%; check the standby budget.
• Hard stop: Drop stays at 68 mV because most switches were left alone: no same-footprint cell meets the target, so the array itself must change.

What The Interviewer Is Testing

• Do you know the three-step flow and which step needs RedHawk Fusion?
• Can you explain the Ron versus leakage trade-off?

Follow-up Question & Model Response

"What if sizing cannot meet the target?"

Candidate Model Response: Then swapping Vt within one footprint is not enough, and you change the array itself. Add switch columns, tighten the pitch, or move to a larger drive footprint, then rerun the drop analysis. Check wake-up rush current too, because more or stronger switches turn on harder. The daisy chain may need more stages to keep the rush inside budget.