Foundational theory

Jitter Budgeting is central to PLL & DLL Clocking. System timing margin is consumed by source jitter, transfer jitter, supply noise, and distribution skew; the budget must close from reference clock to endpoint sampler. Senior AMS owners always tie observed failure to boundary assumptions, ownership, and measurable evidence before changing RTL or layout.

Core concepts explained

• System timing margin is consumed by source jitter, transfer jitter, supply noise, and distribution skew; the budget must close from reference clock to endpoint sampler.

• Primary metric: total jitter budget, deterministic/random split, downstream timing margin

• Primary artifact: jitter budget spreadsheet, phase-noise summary, STA uncertainty map

• Owners: timing lead, PLL owner, clock tree owner

• Boundary and mode context are mandatory for any claim.

• Treat lock/ready/valid bits as evidence, not proof of health.

Why this matters at signoff

At tapeout and bring-up, Jitter Budgeting escapes are expensive to fix. Clock quality is a system property, not just a lock bit. Wrong diagnosis burns schedule across analog, digital, and package teams.

Mental model

JITTER PATH
reference + PLL + supply/package + distribution = endpoint uncertainty

Worked intuition

1. Name boundary and product mode where failure appears.

2. Open total jitter budget, deterministic/random split, downstream timing margin and identify worst scenario.

3. Trace clocks/resets/config from analog macro to digital consumer.

4. Verify wrapper and handoff assumptions on the failing path.

5. Collect jitter budget spreadsheet, phase-noise summary, STA uncertainty map and freeze evidence tags.

6. Classify root cause: contract gap, physical coupling, sequencing bug, or tool-view mismatch.

7. Propose minimal bounded change plus cross-domain regression.

Common misconceptions

• Lock high means clock quality is automatically good.

• Boundary cells are one-time checklist items, not runtime risks.

• SerDes training failure is always firmware.

• If average metric is healthy, there is no silicon risk.

Visual reinforcement

Jitter budget stack

TOTAL JITTER BUDGET
  source jitter
  + PLL intrinsic jitter
  + supply/package induced jitter
  + distribution skew/uncertainty
  = endpoint timing margin consumption

Layer responsibilities

AMS OWNERSHIP LAYERS — Jitter Budgeting

layer                owns                                failure mode
------------------   ----------------------------------  --------------------------
spec contract         clocks/resets/interfaces            hidden assumption drift
wrapper logic         synchronizers/framing/flags         silent data corruption
physical integration  floorplan/isolation/power           coupled noise and droop
signoff governance    waivers/checklists/dashboard        release with blind spots
closure               debug order + regression            fix regresses another mode

AMS deep dive

Clock quality evidence must include jitter and sequencing, not lock status alone.

Concept diagram

CLOCKING FLOW

reference -> PLL/DLL -> distribution -> endpoint margin

Metric graph

JITTER TREND

jitter ps
  ^
  |   o baseline
  |      o stress mode
  |         o failure edge

Reports and artifacts

• jitter budget

• lock/unlock counters

• phase-noise snapshot

• STA uncertainty deltas

Mini case study

False-lock condition released reset early; endpoint logic sampled unstable clock edge patterns.

Debug branches

• Lock qualification policy

• Reset sequencing check

• Package/supply contributors

Senior review question

Ask: what boundary condition proves this topic is actually closed?

Key takeaways

• State boundary, mode, and evidence tag with every claim.

• Always align analog, digital, and physical owners before signoff decisions.

Common pitfalls

• Fixing averages while tails still fail.

• Skipping package/supply evidence in jitter or SerDes issues.

• Shipping with waivers that lack owner and expiration criteria.

Theory reinforcement

Clock quality is a system property, not just a lock bit.