What is the exact mechanism by which crosstalk delay differs from crosstalk noise, and why does the Miller effect make an opposite-switching aggressor roughly twice as impactful as a quiet one?

Short Answer

Crosstalk delay happens when the victim net is already switching, and a neighboring aggressor net's coupling current speeds up or slows down that transition — this shifts when the victim crosses its switching threshold, changing timing. Crosstalk noise happens when the victim is quiet, and coupling current alone produces a spurious glitch bump. The Miller effect is why an aggressor switching in the opposite direction from the victim roughly doubles the effective coupling capacitance compared to a quiet aggressor, since the voltage across the coupling capacitor changes by twice as much.

Technical Explanation

• Two adjacent nets share a coupling capacitance formed by their physical proximity. When one net (the aggressor) switches, it injects current into the other net (the victim) through that coupling capacitance, in addition to whatever the victim's own driver is doing.
• If the victim net is quiet (not switching) when the aggressor switches, the injected current has nothing to add to or subtract from — it simply produces a temporary voltage bump, or glitch, on the otherwise steady victim net. This is crosstalk noise, and the concern is whether the bump is large enough or long enough to be mistaken for a real logic transition downstream.
• If the victim net is itself switching at roughly the same time, the injected current adds to or subtracts from the victim's own charging or discharging current, which changes how fast the victim's voltage moves and therefore shifts exactly when it crosses its logic threshold. This is crosstalk delay — a timing shift on a signal that was already going to transition, not a new spurious event.
• The delta delay (the net change in the victim's delay caused by aggressor coupling) depends heavily on the relative switching direction: an aggressor switching in the same direction as the victim tends to speed the victim up (delay decreases), while an aggressor switching in the opposite direction tends to slow it down (delay increases) — and it is this opposite-direction case that dominates worst-case crosstalk delay analysis.
• The Miller effect explains why opposite-direction switching is roughly twice as impactful: the current through a coupling capacitor is proportional to the rate of change of the voltage difference across it. When the victim and aggressor switch in the same direction, that voltage difference barely changes. When they switch in opposite directions, the voltage difference swings through twice the normal single-ended voltage range, so the effective coupling capacitance the victim sees is roughly doubled compared to a quiet-aggressor case.
• The tool computes crosstalk delay using SPEF-extracted coupling capacitance values between adjacent nets combined with Liberty or CCS (composite current source) driver models, and reports the resulting delta delay per timing arc so it can be added into the normal timing calculation for that specific corner and switching scenario.
• What breaks: analyzing only same-direction aggressor scenarios, or ignoring the coupling capacitance entirely on a net believed to be "far enough" from its neighbors, can miss the opposite-direction, Miller-doubled worst case that actually determines whether a path meets setup or hold.

Follow-up Question & Model Response

"For a long, minimally-shielded bus running next to a fast switching clock buffer, would you expect the dominant risk to be crosstalk delay or crosstalk noise, and why?"

Candidate Model Response: I would expect crosstalk delay to be the dominant risk if the bus lines are themselves actively toggling near the same time as the clock buffer switches, since that is exactly the opposite-direction, Miller-effect scenario that produces the largest delta delay and directly threatens setup or hold on the bus's own timing paths. If the bus lines are instead quiet at the moment the aggressor switches — for example, a control signal that only changes rarely — the greater risk becomes crosstalk noise, since a large glitch on an otherwise steady signal can be misread downstream even without any real transition intended. I would check both, but I would prioritize whichever switching scenario actually matches how that specific bus is used in the design.