ExpertQuestion 154 of 161Source: Ansys RedHawk User Manual: Defining Pad and Package Parameters, Package and Board Analysis; Synopsys ICC2 Implementation User Guide: RedHawk and RedHawk-SC Fusion (Creating Taps)

Why can the package make or break IR signoff?

From PDVerse PnR Interview Handbook · pdVerse Mentor Guide

By Tabish Iqbal · 9+ Years VLSI Physical Design Experience

Short Answer

The die grid is only part of the supply path. Package and bump resistance add DC drop that static analysis must include, and package inductance adds L di/dt droop during current steps that only dynamic analysis with a package model shows. A run with ideal taps assumes a perfect supply at the bumps and can look clean while the real chip droops.

A schematic from the board supply through package resistance and inductance, bump or wirebond RLC and pad parasitics to the on-die grid and its decap, with arrows marking where DC drop and L di/dt droop arise. — Technical Reference DiagramWhy can the package make or break IR signoff?

Technical Explanation

Static analysis sees only resistance. Package and bump resistance multiply the average current, so 5 mOhm effective at 2 A is 10 mV before any on-die drop.
Dynamic analysis sees inductance too. When current rises quickly, the package inductance resists the change and the die supply dips by roughly L times di/dt until on-die decap and the package settle.
Package inductance and on-die capacitance form a resonant loop. Activity near that frequency can build a larger droop than a single step, which is why package models matter for dynamic signoff.
In standalone RedHawk, setup package (RH) sets lumped package R, L and C in ohms, picohenrys and picofarads, each defaulting to 0. setup pad (RH) and setup wirebond (RH) add pad and bond or bump parasitics.
For more accuracy, setup pss (RH) points at a PLOC file and a SPICE package subcircuit, and RedHawk also supports S-parameter package models.
Fusion taps are ideal voltage sources. Simple package RLC through rail.pad_files (ICC2) requires rail.allow_redhawk_license_checkout (ICC2), and Fusion does not support dynamic analysis with lumped or SPICE packages by default.
Bump placement changes the result as much as the package values: a region far from any VDD bump carries current through more grid and sees more drop.

What To Check

Package R, L and C used in signoff, and who supplied them.
Tap or bump locations matching the final bump map.
Static drop split into package and on-die portions.
Dynamic droop with and without the package model on the same window.

Command Checks & Actions

ICC2 (icc2_shell)set_app_options -name rail.pad_files -value {bumps.ploc}

Uses real pad locations and simple RLC data instead of hand-placed ideal taps.

ICC2 (icc2_shell)set_app_options -name rail.allow_redhawk_license_checkout -value true

Required before creating taps with simple package RLC models.

ICC2 (icc2_shell)analyze_rail -voltage_drop dynamic -nets {VDD VSS}

Dynamic Fusion run; compare against the standalone packaged run.

RedHawk (redhawk)setup package -r 0.005 -l 20 -c 5

Standalone lumped package model in ohms, pH and pF; R and L are the example's effective 5 mOhm and 20 pH, C is illustrative.

RedHawk (redhawk)setup pss -pad_file pkg.ploc -subckt pkg_model.sp

Uses a SPICE package subcircuit for a more accurate model.

RedHawk (redhawk)perform analysis -vectorless

Dynamic run with the package in the loop.

Healthy, Suspicious & Hard-stop Results

Healthy (illustrative): Static drop 37 mV total, 10 mV of it in the package, and dynamic droop rising from 41 to 71 mV with the package model, still under 72 mV.
Suspicious (illustrative): Package values are placeholders from an older product, or the bump map changed after the taps were set.
Hard stop: Dynamic signoff with ideal taps only, or droop with the package model over budget.

Common Mistake

The Trap: Treating package effects as the package team's problem and signing off the die with ideal taps. The package team in turn assumes the die met its budget, so the combined droop is never analysed until the board shows it.

What The Interviewer Is Testing

Can the candidate separate resistive DC drop from L di/dt droop?
Knowledge of how RedHawk models the package.
Awareness that Fusion taps are ideal unless a package model is enabled.

Follow-up Question & Model Response

"Why does adding decap sometimes make package resonance worse?"

Candidate Model Response: More on-die capacitance lowers the resonant frequency of the package inductance with the die capacitance. If that moves the resonance toward a frequency the workload excites, such as a periodic burst at a submultiple of the clock, the droop can grow. The fix is to check droop across the relevant activity patterns with the package model, not to add decap blindly. Package decap or damping may be the better lever.

Practical Example

Tapeout Scenario: A die draws 2 A average at 0.8 V with a 5% (40 mV) static budget (illustrative). With ideal taps static drop is 27 mV; with 5 mOhm effective package and bump resistance it becomes 37 mV. For dynamic, a 1.5 A step in 1 ns through 20 pH effective inductance adds about 30 mV (20 pH x 1.5 A/ns), taking the droop from 41 mV with ideal taps to 71 mV against a 72 mV budget.

PnR Flow Mentor Guide

Master the Physical Design Implementation Flow

Read the complete 8-chapter PnR Flow Mentor Guide free on the web — library setup through placement, clock tree synthesis, routing, chip finishing, hierarchical implementation, and ECO, all the way to stream-out.

Explore PnR Flow Mentor Guide →