DWDM System Commissioning: The 5-Phase Checklist

A poorly commissioned DWDM span does not fail on day one — it fails months later, after dust, drift and aging push an already-marginal link over the edge. MapYourTech's commissioning checklist structures the entire turn-up as five sequential phases, each with measurable acceptance thresholds: fiber characterization, amplifier setup, channel provisioning, OSNR verification, and protection testing. This article condenses that guide into an actionable reference with the key formulas and thresholds you need in the field, from OTDR discipline to cascaded OSNR budgets.

Background: Why the Order of the Five Phases Matters

A DWDM terminal is a chain of dependent measurements. Every later phase assumes the previous one is already inside tolerance, so skipping ahead quietly destroys your ability to interpret results. If you light a transponder before the span loss is confirmed, an unplanned 2 dB of connector dirt shows up as a mysterious OSNR shortfall and you spend the night chasing the amplifier instead of the patch panel. If you verify OSNR before the band is fully loaded with ASE noise, the amplifier is running at the wrong gain point and the numbers you record as a baseline are meaningless.

The rule of thumb used by most transport teams is simple: loss first, gain second, wavelengths third, quality fourth, survivability last. Each phase produces a signed number that becomes an input to the next. Anything outside tolerance is corrected before moving on — not "noted for later".

The phases in this checklist are:

• Phase 1 — Fiber characterization: bidirectional OTDR, splice and connector quality, absolute end-to-end span loss.
• Phase 2 — Amplifier setup: EDFA gain mode and gain value, noise figure check, per-channel flatness, Raman decisions, laser safety.
• Phase 3 — Channel provisioning: ITU grid placement, wavelength accuracy, modulation and FEC match, ASE loading.
• Phase 4 — OSNR verification: received OSNR versus FEC threshold, tilt, pre-FEC and post-FEC BER.
• Phase 5 — Protection testing: switch time, detection thresholds, control-plane behaviour under load.

Prerequisites: Tools and Data Before You Touch the Span

Commissioning fails more often from missing inputs than from bad hardware. Collect the following before the maintenance window opens:

• Design pack: span loss budget, amplifier positions, expected per-channel launch power, target OSNR, and the transponder reach table for the modulation formats in the plan.
• Fiber documentation: fiber type (G.652.D, G.655, G.654.E), splice records, connector inventory, and any known high-loss sections from the as-built OTDR traces.
• Test equipment: calibrated OTDR with launch and receive reference fibers (≥500 m for the launch side, ≥500 m receive), optical power source and power meter, optical spectrum analyzer (OSA) with 0.1 nm resolution bandwidth, and a polarization or BERT test set for BER measurement.
• Cleanliness kit: one-click cleaners, lint-free wipes, and 99% isopropyl alcohol. In field studies, contaminated APC ferrules are the single most common cause of an unexpected 1–3 dB of loss.
• Laser safety: verify automatic laser shutdown (ALS) and automatic power reduction (APR) are enabled before any Raman pumps or high-power amplifiers are energized. Class 1M/3B hazard is real the moment the equipment is unkeyed.

Agree on the acceptance template in advance: a single spreadsheet with one row per fiber pair and one column per phase. If the commissioning report is assembled after the fact from memory, it will not survive the first incident review.

Phase 1 — Fiber Characterization

Run bidirectional OTDR at 1310/1550 nm on every fiber pair with launch/receive reference fibers (≥500 m) to eliminate dead zones. Key thresholds: span loss ≈ 0.20–0.22 dB/km at 1550 nm (G.652); fusion splice loss ≤ 0.1 dB (re-splice above 0.15 dB); connector return loss ≥ 45 dB (APC). Then verify end-to-end span loss with a power source and meter; a discrepancy over 1 dB warrants investigation.

Bidirectional testing is not optional. A single-direction OTDR trace attributes the whole event loss to one side of a splice or connector, because the measurement includes the backscatter difference between the two fibers. Averaging the two directions cancels that bias and gives the true event loss. When the two directions disagree by more than 0.15 dB on a splice, treat it as a suspect even if the average looks acceptable.

Span Loss (dB) = α×L + splices + connectors + margin
Example (80 km): 16.80 + 0.64 + 1.40 + 2.00 = 20.84 dB

Record the following per fiber: measured span loss at 1550 nm, worst splice event, total connector count, and the position of any event more than 0.3 dB. Mark high-loss events on the as-built drawing — when the link degrades two years later, this trace is your only baseline for comparison.

Phase 2 — Amplifier Setup

For inline EDFAs use AGC mode tied to measured span loss (gain = span loss ± 0.5 dB), confirm noise figure ≤ 6 dB (typical 4–6 dB), and verify per-channel power flatness ≤ ±1 dB. For spans over 25–28 dB loss, consider Raman: confirm the fiber type (G.652/G.655/G.654) is correct in the NMS before enabling pumps — and always verify ALS/APR laser-safety functions first (Raman pumps run 500 mW–2 W).

AGC (automatic gain control) is preferred over APC (automatic power control) during commissioning because it holds gain constant and therefore keeps the amplifier noise figure predictable. In APC the amplifier chases a target output power, which means a partially loaded band drives the gain up and the noise figure degrades. Only switch to APC or a power-equalized mode after the full channel plan is provisioned.

Single-amp OSNR = 58 + P_in − NF   (0.1 nm RBW @1550 nm)
Cascaded: 1/OSNR_total = Σ 1/OSNR_i

The cascaded formula is the one that catches people out. A single amplifier delivering 30 dB OSNR looks healthy, but ten of them in series deliver roughly 20 dB — the harmonic sum punishes the weakest stage hardest. Calculate the cascade in the design phase and then confirm it with a live OSA measurement; if the measured value is more than 2 dB below prediction, re-measure span losses and verify the noise figure reported in the amplifier's alarm/threshold report (ATR).

For Raman-assisted spans, confirm the fiber type in the NMS first: Raman gain depends on the effective area of the fiber, and enabling a G.655 profile on a G.652 span produces the wrong gain shape. Verify the pump wavelengths, the resulting gain profile, and the safety interlocks in one session.

Phase 3 — Channel Provisioning

Provision on the ITU-T G.694.1 grid (12.5/25/50/100 GHz). Verify transponder wavelength with an OSA (±0.01 nm tolerance), start with one mid-band test channel, then edge channels, then the full plan — using ASE noise loading to fill unused spectrum and keep EDFAs stable. Confirm modulation format (QPSK/8QAM/16QAM) and FEC mode match the design.

The sequencing — mid-band first, edges next, full plan last — is deliberate. The middle of the C-band is where the EDFA has the flattest gain and the lowest noise figure, so a failure there points at the transponder or the provisioning, not the amplifier. Edge channels (roughly 1530–1535 nm and 1560–1565 nm at the band limits) are the first to show gain ripple and the first to miss OSNR. Provisioning them after the mid-band baseline is established isolates amplifier tilt problems from transponder problems.

• Confirm grid: 50 GHz spacing is the common default; 100 GHz for legacy, 75 GHz for some 400G/800G line systems, 37.5 GHz and 12.5 GHz for flex-grid.
• Confirm wavelength accuracy with an OSA and log the measured value, not the nominal value from the plan.
• Confirm modulation and FEC versus the design table — a 16QAM transponder provisioned as QPSK will produce an unusable OSNR margin and mislead the next phase.
• Enable ASE loading on unused channels so the amplifier sees a constant total power. Without it, gain per channel creeps up as you add channels, and later channels are provisioned against the wrong amplifier state.

Phase 4 — OSNR Verification

Measure OSNR at the receiver pre-amplifier input. Typical requirements with 3 dB margin above FEC threshold: 100G DP-QPSK ≥ 12 dB; 200G DP-16QAM ≥ 18 dB; 400G DP-16QAM ≥ 20 dB. Check flatness ≤ 3 dB across the band, record pre-FEC BER below threshold, and post-FEC BER must be error-free over ≥ 15 minutes.

Measure at the right point. OSNR referenced at the transmitter is a marketing number; OSNR at the receive pre-amplifier input is what the DSP actually sees. Use interpolation, not a channel-passing filter measurement, and be aware that dense 50 GHz systems require the OSA to resolve the channel adequately — a too-wide resolution bandwidth reads artificially low OSNR.

Post-FEC verification is the final gate: run the channel error-free for at least 15 minutes with live or tester traffic and confirm zero errored seconds. A link that clears pre-FEC BER but shows sporadic post-FEC corrections is not commissioned — it is a ticket waiting to be raised.

Phase 5 — Protection Testing

Validate protection switching completes within 50 ms: verify protection group configuration, hardware detection thresholds, and control-plane loading. Document every result as the baseline for future troubleshooting.

Test both directions and both failure modes: forced switch from the management system, and a hard fiber pull on the working path. A control-plane-forced switch can complete in a few milliseconds while the real hardware-triggered switch takes longer, so record them separately. For optical protection (OLP) confirm the switching threshold, the hold-off time, and that the standby path actually carries traffic rather than merely showing "ready". For OTN or packet-level protection confirm the APS protocol state transitions, and for any G.8032 ring confirm the guard timer does not cause a second switch during a flapping event.

Finally, load the control plane: run protection tests while the NMS is polling and while any GMPLS/ASON or segment-routing control plane is converged. A protection switch that meets 50 ms on an idle system can miss it when the CPU is busy.

Commissioning Record: What to Capture for the Future

The deliverable of commissioning is a baseline, not a green light. Capture and archive:

• Bidirectional OTDR traces in the native .sor format, one file per fiber per direction, plus a PDF export for reviewers.
• Measured span loss and the calculated budget side by side, with any discrepancy over 1 dB explicitly explained.
• Amplifier settings: mode, gain, output power, tilt, and the reported noise figure per amplifier.
• OSA trace of the full loaded band, plus per-channel OSNR and power values in a spreadsheet.
• Pre-FEC and post-FEC BER for every channel, with the test duration.
• Protection switch times for forced and hard-failure scenarios, both directions.

Store this alongside the as-built documentation. During the first real outage, the difference between "this link was always marginal" and "this link regressed" is exactly this file.

Common Commissioning Issues

FAQ

How long should the whole turn-up take?

For a single span with two terminals and one amplifier site, a competent team budgets a full shift: 3–4 hours for bidirectional OTDR and span loss on all pairs, 1–2 hours for amplifier setup, and 2–3 hours for provisioning plus OSNR and protection testing. Cascaded multi-span links with Raman scale roughly linearly with span count.

Can I commission a link that is not fully loaded?

You can commission the equipment, but not the link. OSNR and power measurements taken on a partially loaded band are not representative. Use ASE loading to emulate a full band, and re-verify OSNR once the real channels are lit.

What OSNR margin is enough?

Three dB above the FEC threshold minimum, five dB preferred for spans that will be in service for a decade. Aging, repairs and added splices all consume margin, and a link commissioned at zero margin will fail at the first re-splice.

Why do my two OTDR directions disagree?

Backscatter coefficient differences between the two fibers on either side of an event. Always average the directions for event loss; a persistent large difference points at a mismatched or dirty connector.

Related: fiber-level measurement discipline is covered in OTDR testing basics, the hardware side in DWDM system components: MUX, EDFA and DEMUX, the receive-quality math in coherent optics: pre-FEC BER, OSNR and DSP checks, and the design budget in optical power budget calculation. The Infinera DWDM resource guide and the H3C 光模块故障排查手册 cover vendor-specific module issues.

原文链接：https://mapyourtech.com/dwdm-system-commissioning-interactive-checklist