What this chapter covers

When two nets run near each other, a switching aggressor injects charge into a victim through coupling capacitance. That either shifts the victim’s own edge (crosstalk delay) or produces a glitch on a quiet victim (crosstalk noise). The chapter covers Miller effect, aggressor/victim timing windows, SI-aware delay calculation, and physical plus constraint-level mitigations.

This is first-order signoff at advanced nodes, where coupling often dominates net capacitance. Engineers hit it in PrimeTime SI (or equivalent) runs, crosstalk-aware closure, glitch checks before tapeout, and spacing/shielding decisions on clocks and other sensitive nets.

Worked example

An opposite-direction aggressor switching in the victim’s timing window roughly doubles the effective coupling (Miller), slowing a victim rise and adding delay to a setup path. A same-direction aggressor can speed the victim (delta-delay credit) that you must not treat as free margin unless the window overlap is real. If the victim is quiet, the same injection is a noise bump that may be filtered or may cause a functional failure.

A common misconception

Treating every SI delta as a must-fix without checking timing windows, or disabling SI to “clean” slack. Window-filtered aggressors may be legal; unconstrained SI analysis can also be wildly pessimistic. Another error is confusing delay push-out with a functional glitch — they use the same physics and different checks.

Sample interview answers

How does crosstalk delay differ from crosstalk noise? Delay: the victim is switching, so injected current moves the threshold crossing. Noise: the victim is quiet, so injected current is a glitch.

Why do timing windows matter? An aggressor only hurts if it switches while the victim is sensitive. Non-overlapping windows can be filtered; overlapping opposite-direction windows are the usual setup hit.

What is a physical fix versus a constraint fix? Spacing, shielding, layer promotion, and buffering are physical. False-pathing an SI victim without proof is how a real coupling failure is hidden.

The paid PDF develops aggressor-window diagrams, delta-delay versus glitch reports, filtering options, and the 80-question interview set for this chapter.

Previous: Delay calculation · Next: Configuring the STA environment · Related: Crosstalk delay vs noise, Expert STA questions

💡 Key Technical Topics Covered

  • Victim & Aggressor Net Coupling Capacitance
  • Crosstalk Delta Delay (Setup vs Hold Impact)
  • Crosstalk Glitch & Noise Immunity Thresholds
  • Shielding & Net Spacing Fixing Techniques

🎯 VLSI Physical Design Interview Questions Addressed

  • How does crosstalk coupling capacitance affect setup and hold timing?
  • What is crosstalk glitch and how can it cause functional failure?
  • What physical design techniques are used to fix crosstalk noise?