Link Training

What this topic teaches

Link Training converts analog-digital assumptions into a measurable engineering contract. Training protocols establish lane polarity, timing, equalization, and protocol states; digital state machines must synchronize with PHY indications and timeout policies. The hard part is proving which boundary broke first, under which mode, and with which evidence.

The senior-engineer question

When training pass rate, time-to-L0, fallback mode frequency moves, can you isolate the first failing boundary, identify owner, and define the smallest reversible change that proves root cause?

AMS CLOSURE FLOW — Link Training

spec + handoff assumptions
        |
        v
boundary implementation (wrapper/cells/reset)
        |
        v
physical context (floorplan/power/package)
        |
        v
metrics + artifacts (jitter/BER/noise/validity)
        |
        v
root-cause classification -> bounded fix -> regression

Debug rule: always name mode, boundary, evidence tag, and owner.

Picture the boundary behavior

Start every study session by drawing the contract before opening logs or reports. The diagrams below are what to reproduce on whiteboard.

Training state path

detect -> poll -> config EQ -> align -> active/L0
      ^                                  |
      +------------- retry/fallback -----+

Boundary sequence

AMS BOUNDARY SEQUENCE — Link Training

analog macro -> wrapper / boundary cell -> synchronizer or sampler -> digital consumer
      |                 |                         |                     |
  analog assumptions    legal voltage/state       clock/reset contract   protocol/data validity

metric under watch: training pass rate, time-to-L0, fallback mode frequency

Who owns which layer

AMS OWNERSHIP LAYERS — Link Training

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

Evidence to collect

• Primary metric: training pass rate, time-to-L0, fallback mode frequency.

• Primary artifact: training state trace, firmware log, protocol analyzer capture.

• Owners to bring into review: firmware owner, PHY owner, system validation lead.

• One tagged reproduction and one reduced reproducer.

• Cross-check from at least two evidence planes: functional and physical/signoff.

Ownership map

OWNERSHIP MAP — Link Training

artifact                  owner
----------------------    -----------------------------
integration artifact    firmware owner
implementation artifact PHY owner
signoff artifact        system validation lead

Every boundary issue needs a named owner before fixes start.

Subpages in this topic

Each topic includes mechanism, contract I/O, reports, debug, worked example, pitfalls, interview, checklist, theory, design-space, extended case study, walkthrough, comparison matrix, software view, and silicon PPA impact.

Key takeaways

• Carry boundary and mode context with every metric.

• Prove mechanism with tagged evidence before changing silicon-facing logic.

• Close with explicit owner-aligned regression criteria.

Common pitfalls

• Assuming lock/ready/valid means healthy behavior.

• Ignoring package and power contributors in jitter/SerDes issues.

• Waivers without expiry, owner, or mitigation plan.

AMS deep dive

SerDes closure requires protocol, training, and SI evidence together.

Concept diagram

SERDES FLOW

training -> equalization -> lane margin -> protocol stability

Metric graph

BER VS EQ

BER
 ^
 |  high  low  high
 +-----------------> EQ setting

Reports and artifacts

• lane BER

• training state transitions

• EQ sweep report

• link retry counters

Mini case study

Link looked protocol-clean but lane margin collapsed under thermal sweep.

Debug branches

• Correlate LTSSM and lane metrics

• Check SI margins

• Review firmware timeout assumptions

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.