Diagnosing a Link That Fails Only at 10G
A data center technician reports that a 300 m multimode link between two switches passes a basic insertion-loss test at 850 nm (measured 1.8 dB against a 3.5 dB budget) and links up fine at 1G. When the same link is moved to 10GBASE-SR optics, it fails to establish a link or drops intermittently. The link uses OM3 fiber, two LC connectors at each end (patch panel plus equipment), and was installed three years ago. The technician's loss tester is a legacy unit with an LED source at 850 nm and 1300 nm. The customer wants to know whether to replace the cable or the optics.
1. What measurement is missing that would explain the 10G-only failure, and how should it be performed?
Read the worked answer
The missing measurement is modal bandwidth / differential mode delay (DMD) characterization, or at minimum an ** encircled flux (EF)-compliant launch condition** on the loss test.
Reasoning:
- 10GBASE-SR over OM3 has a reach limit of ~300 m only when the fiber meets OM3 bandwidth specs and the launch is EF-compliant. At the edge of that reach, marginal bandwidth or a non-compliant launch causes exactly this symptom: 1G works, 10G is marginal.
- The legacy LED tester with an overfilled launch can under-report loss and says nothing about bandwidth. A mode-conditioned (EF) launch or an OTDR with a mandrel-wrapped launch fiber is needed to see the real loss and any reflective events.
- Recommended test: OTDR at 850 nm with a proper launch cable to locate high-loss/high-reflectance events, plus a loss test with EF-compliant source to confirm against the 10G budget.
2. The OTDR shows a 1.2 dB event at one patch panel and clean fiber elsewhere. What is the most likely cause and the recommended fix?
Read the worked answer
A 1.2 dB localized event at a connector pair is far above a good LC mated pair (typically <0.3 dB). Likely causes, in order of probability:
- Contaminated or damaged endface - the most common cause of a high-loss connector event. Inspect with a scope; clean and re-test.
- Poorly mated pair / misaligned key - check that the connectors are fully seated and keys are aligned.
- Bad polish or cracked ferrule - replace the patch cord.
Fix path: inspect and clean first, re-measure. If the event persists after cleaning and re-mating, replace the patch cord at that panel. Do not replace the 300 m cable - the OTDR shows the cable itself is clean, so the problem is at the connector, not the fiber.
3. The customer asks whether to just swap in 10GBASE-LR optics instead. Advise.
Read the worked answer
No - that is the wrong fix and may not even work.
- 10GBASE-LR is a single-mode PMD; it will not operate over OM3 multimode fiber. Mixing SM optics onto MMF is a common field error and will not establish a reliable link.
- The correct fix is to restore the multimode link to spec: clean/replace the bad connector, verify with an EF-compliant loss test and OTDR, and confirm the link is within the 10GBASE-SR budget for OM3 at 300 m.
- If the link still won't run at 10G after the connector is fixed, the remaining option is to shorten the reach (e.g., add a media converter/switch closer) or upgrade the fiber plant - but only after the connector fault is eliminated, since the OTDR already localized the problem.