Optical Fiber Dispersion Eye Closure Case Study

This case study follows a fiber-optic service that looked healthy by received optical power but failed after a maintenance team replaced a transceiver with a lower-cost spare. The route loss, connector inspection and OTDR record were acceptable. The hidden problem was not attenuation. The replacement optic had insufficient chromatic-dispersion tolerance and a wider source spectrum, so the received waveform lost timing margin and the eye closed.

The case teaches a practical telecommunications lesson: optical power margin is necessary but not sufficient. A high-speed fiber service must also preserve waveform quality, dispersion margin, receiver timing margin and error performance.

Case Summary

Item	Engineering relevance
Service	10 Gbit/s single-mode Ethernet backhaul
Fiber route	Existing OS2 route between an aggregation node and an industrial technical building
Trigger	Field replacement of a failed transceiver with a spare that matched wavelength and power but not dispersion tolerance
Hidden weakness	Accumulated chromatic dispersion was close to the spare optic limit
Main symptom	Receiver power stayed acceptable, but FEC corrections and burst errors increased
Useful evidence	Power budget, accumulated dispersion screen, pulse-broadening estimate, eye measurement, BER soak and rollback comparison

The central engineering question was:

Why does the link fail when received optical power is comfortably above sensitivity?

The answer was that the service had a waveform impairment, not a weak-light impairment.

Baseline Data

Use the following simplified values. A real investigation should replace them with the route record, transceiver data sheet, optical test reports and service counters.

Parameter	Value
Data rate	10\ \text{Gbit/s}
Line format for the screening calculation	NRZ equivalent timing screen
Operating wavelength	1550\ \text{nm}
Fiber type	OS2 single-mode
Route length	42\ \text{km}
Fiber attenuation coefficient	0.22\ \text{dB/km}
Chromatic dispersion coefficient	17\ \text{ps/(nm km)}
Mated connector pairs	6
Average connector loss used for budget	0.30\ \text{dB} per pair
Fusion splices	14
Average splice loss used for budget	0.06\ \text{dB}
WDM/filter allowance	0.70\ \text{dB}
Bend and routing allowance	0.50\ \text{dB}
Minimum transmitter power	+1.5\ \text{dBm}
Receiver sensitivity	-18.0\ \text{dBm}
Reserved design margin	3.0\ \text{dB}
Original optic dispersion tolerance	800\ \text{ps/nm}
Replacement optic dispersion tolerance	400\ \text{ps/nm}
Original source spectral width for screen	0.03\ \text{nm}
Replacement source spectral width for screen	0.25\ \text{nm}

The replacement spare had the same connector form factor, nominal wavelength and power class. That made it look interchangeable during maintenance. It was not interchangeable for this route.

Observed Failure Evidence

The service did not fail like a cut fiber or a dirty connector. It failed like a marginal high-speed receiver.

Evidence	Observation
Receiver optical power	Stayed around -11.6\ \text{dBm} before and after the replacement
OLTS insertion loss	Consistent with the commissioning record
OTDR trace	No new localized loss event or reflection spike
Connector inspection	End faces clean after normal inspection and cleaning
FEC and error counters	Corrected blocks increased sharply; burst errors appeared during warm periods
Eye or equivalent receiver margin	Failed the internal eye-margin screen with the replacement optic
Rollback test	Original dispersion-rated optic restored stable service without route work

This pattern matters. If the investigation had stopped at “received power is above sensitivity,” the service would have been released with a latent impairment.

Step 1: Prove That Attenuation Is Not the Root Cause

The path loss budget is:

Fiber loss:

Connector loss:

Splice loss:

Total path loss:

Minimum received optical power is:

Raw sensitivity margin is:

Guarded margin after the reserved design allowance is:

Engineering Comment

The link passes the optical power screen with about 3.4\ \text{dB} after the reserved allowance. That is not unlimited margin, but it is enough to make a pure attenuation diagnosis unlikely. The measured DOM power around -11.6\ \text{dBm} also matches the budget, so the calculation and observation are consistent.

The important limitation is that receiver sensitivity is specified under defined waveform conditions. It assumes the receiver is being fed a signal with acceptable dispersion, extinction ratio, jitter, noise and eye opening. A receiver can have enough average optical power and still make wrong decisions at the sampling instant.

Step 2: Check Accumulated Chromatic Dispersion

For a first-pass single-mode screen, accumulated chromatic dispersion is:

Substituting the route data:

The original optic tolerance was:

Original dispersion margin:

The original installation was not generous, but it passed the optic limit.

The replacement optic tolerance was:

Replacement dispersion margin:

Engineering Comment

The replacement optic fails the route dispersion screen by 314\ \text{ps/nm}. This explains why the link can show good received power and poor error performance at the same time. The fiber delivers enough photons, but it does not deliver the waveform inside the timing and dispersion limits assumed by the spare transceiver.

The calculation also explains why a maintenance substitution can create a new failure without any civil work, connector damage or fiber break. The physical plant did not change. The compatibility boundary changed.

Step 3: Estimate Pulse Broadening From Source Spectral Width

Chromatic-dispersion pulse spreading can be screened with:

For the original optic:

For the replacement optic:

The bit period for 10\ \text{Gbit/s} is:

The original broadening fraction is:

The replacement broadening fraction is:

Engineering Comment

The original optic spreads the signal by about 0.21 unit interval in this simplified screen. That is not zero, but it leaves timing margin for receiver bandwidth, jitter, extinction ratio, noise and sampling uncertainty.

The replacement optic spreads the signal by about 1.8 unit intervals. Adjacent bits no longer remain cleanly separated in time. That is eye closure. The receiver may still report normal average optical power because the total optical energy arrived, but the energy arrived at the wrong times for reliable decisions.

This is a first-pass engineering screen, not a full transmitter-receiver model. Real performance also depends on chirp, equalization, extinction ratio, receiver bandwidth, clock recovery, coding, pattern dependence, temperature and vendor implementation. The screen is still strong enough to reject the spare optic for this route unless a deeper system test proves otherwise.

Step 4: Interpret Eye and Error Evidence

The team compared the original optic and the replacement optic under the same route, patching and traffic conditions.

Metric	Original optic	Replacement optic
Receiver optical power	-11.5\ \text{dBm}	-11.6\ \text{dBm}
Dispersion margin by data sheet	+86\ \text{ps/nm}	-314\ \text{ps/nm}
Estimated dispersion broadening	21.4\ \text{ps}	178.5\ \text{ps}
Broadening as fraction of bit period	0.21 UI	1.79 UI
Eye-margin screen	pass	fail
Error behavior during four-hour soak	stable	corrected blocks grew and burst errors appeared

The power rows are almost identical. The waveform rows are not. That separation is the diagnostic signature of a dispersion-limited fault.

Engineering Comment

The receiver does not decide bits from average power alone. It samples a time-varying electrical signal after photodetection, amplification, filtering and clock recovery. Chromatic dispersion smears transitions and creates intersymbol interference. Jitter and noise then consume what remains of the eye opening.

This is conceptually similar to multipath intersymbol interference in radio links, even though the physical mechanism is different. In both cases, energy from one symbol leaks into the decision interval of another symbol.

Step 5: Rule Out Competing Failure Modes

A good case study should not force the preferred explanation. It should show why other explanations were rejected.

Candidate cause	Check	Result
Dirty connector	Inspection, cleaning, OLTS before/after comparison	No loss recovery after cleaning; connector event absent
Fiber break or macro-bend	OTDR trace and power trend	No new high-loss event; route loss matched baseline
Receiver overload	Maximum received-power check	-11.6\ \text{dBm} was far below overload
Wrong wavelength	Transceiver inventory and DOM wavelength	Nominal wavelength matched, but dispersion tolerance did not
Packet congestion	Physical-layer counters and local loopback comparison	Errors followed the optic, not offered load
Temperature-only fault	Temperature trend	Temperature worsened the failure but did not create the underlying negative dispersion margin

Engineering Comment

Connector contamination and dispersion can both create intermittent service trouble, but they leave different evidence. Connector contamination usually changes insertion loss, return loss or an OTDR event. Dispersion-limited operation can preserve power while degrading eye opening, BER, FEC counters and temperature margin.

The investigation therefore needed both optical plant evidence and receiver performance evidence.

Step 6: Evaluate Corrective Options

The engineering release team considered three options.

Option A: Restore a Dispersion-Rated Optic

Reinstall an optic with at least:

If the team requires a minimum dispersion allowance of 50\ \text{ps/nm}:

The original 800\ \text{ps/nm} optic meets this rule:

This option restores compatibility without changing the fiber route. It should still be followed by BER, FEC, eye or equivalent receiver-margin evidence.

Option B: Move to a Lower Bit Rate Temporarily

At 2.5\ \text{Gbit/s}:

The replacement optic broadening fraction would become:

Lowering the bit rate improves timing margin, but 0.45 UI is still a serious impairment unless the actual transceiver and receiver are rated for it. This can be a degraded-service workaround only after live validation. It is not a clean permanent release.

Option C: Shift Wavelength Without Checking Attenuation

A move toward 1310\ \text{nm} can reduce chromatic dispersion in standard single-mode fiber, but attenuation may increase. If the attenuation coefficient rises from 0.22 to 0.35\ \text{dB/km}:

The guarded optical margin was only:

The extra attenuation would consume:

more than the guarded margin. A wavelength change therefore cannot be approved just because it reduces dispersion. It must also pass the optical power budget, overload check, receiver specification and service validation.

Engineering Decision

The accepted corrective action was Option A: install a dispersion-rated optic approved for the documented route dispersion and source spectrum, then update the spare-parts list so that field replacement requires dispersion compatibility, not only wavelength, connector type and power class.

Release Validation

The release package should include evidence that directly addresses the failure mechanism.

Validation item	Acceptance intent
Route dispersion calculation	Confirm accumulated dispersion is inside the installed optic tolerance
Optical loss and power check	Confirm attenuation margin remains acceptable
Receiver overload check	Confirm the selected optic is not too strong for the route
Eye, BER or equivalent margin test	Confirm waveform quality, not only average power
FEC and error-counter soak	Confirm no growth under representative traffic and temperature
Connector inspection and OLTS record	Preserve evidence that connector loss is not the hidden cause
Transceiver inventory update	Prevent future substitution with incompatible spares
Operations handover note	Explain the dispersion constraint in maintenance language

The validation boundary should name the transceiver model, firmware where relevant, wavelength, fiber type, route length, dispersion coefficient, accumulated dispersion, receiver sensitivity, overload limit, measured power, test duration, temperature range and acceptance criteria.

Common Engineering Mistakes

Common mistakes in this failure mode include:

1. Treating receiver sensitivity as a universal guarantee instead of a conditionally specified value.
2. Accepting a spare optic because wavelength and power match while ignoring dispersion tolerance.
3. Using DOM optical power as the only physical-layer health metric.
4. Assuming an OLTS pass proves high-speed waveform quality.
5. Applying a lower bit rate as a permanent fix without proving BER and timing margin.
6. Changing wavelength to reduce dispersion without recalculating attenuation and margin.
7. Failing to record transceiver model and dispersion tolerance in the route acceptance file.

Transferable Lessons

A fiber route can be attenuation-limited, connector-limited, reflection-limited, dispersion-limited, receiver-limited or operations-limited. The first healthy-looking number is rarely the whole diagnosis.

For high-speed fiber links, the engineering release question should be:

Does this exact transceiver, at this wavelength and data rate, over this exact route, meet power, dispersion, timing, error-rate and operational replacement requirements?

If the answer is documented only as “link up” or “received power is above sensitivity,” the acceptance evidence is incomplete.