8.4 km 10GBASE-LR Acceptance: Passing Test, Failing Design
A consultant is asked to accept a newly installed 8.4 km single-mode link between two campus buildings, operating 10GBASE-LR at 1310 nm. The as-built shows 8.4 km of OS2 outdoor cable, five fusion splices in two handhole closures, and three patch panels (one at each end plus one mid-span in a building), each with a mated LC pair at both sides of the panel, for six mated connector pairs in the channel. The contractor's OLTS result at 1310 nm is 5.2 dB, and the project manager says it passes because the optics have a 6.2 dB budget. The OTDR trace shows a splice at 3.1 km with -0.18 dB loss and a reflective event at 5.6 km with 0.9 dB loss. Two connector endfaces, when inspected, show a 40 micrometre pit in the core zone of one and a large scratch across the cladding of the other.
1. Compute the worst-case channel loss and decide whether the link is acceptable.
Read the worked answer
Build the number from components rather than from the single measured value. Using common worst-case planning values: fiber 8.4 km x 0.4 dB/km = 3.36 dB at 1310 nm; five splices x 0.3 dB = 1.5 dB; six mated pairs x 0.75 dB = 4.5 dB; total 9.36 dB. Even with typical rather than worst-case values, say 0.1 dB per splice and 0.5 dB per mated pair, you get 3.36 + 0.5 + 3.0 = 6.86 dB, which still exceeds the roughly 6.2 dB 10GBASE-LR budget. So the design fails on paper no matter which set of numbers you use, and the 5.2 dB measured result only tells you the installed components happened to come in better than worst case. A 1.0 dB margin against a 6.2 dB budget is not acceptable for a 25-year plant: repairs, re-polishing, and aging will consume it. The right fix is to change the design, not to argue about the measurement. Eliminate the mid-span panel by splicing through, saving two mated pairs and up to 1.5 dB, and set acceptance limits at 0.50 dB per mated pair and 0.15 dB per splice, then recompute and re-accept.
2. Interpret the -0.18 dB splice and the 0.9 dB reflective event. What action does each require?
Read the worked answer
The negative splice is almost certainly a measurement artifact, not a gain. Splice loss measured from one direction mixes the actual splice loss with the difference in backscatter capture between the two fibers, and where the downstream fiber has a higher backscatter coefficient the trace can show a gainer. The correct response is to shoot the link from both ends and average the splice loss in each direction, which will normally show a small positive number. Do not re-splice on the basis of a single-direction trace, and do not record the -0.18 dB as a result. The 0.9 dB reflective event at 5.6 km is different: a mated LC pair should be well under 0.5 to 0.75 dB, and a 0.9 dB event with strong reflectance points to contamination, a damaged endface, an incorrect polish mating such as APC to UPC, or a failed alignment sleeve. Inspect and clean both sides, verify the polish type and the adapter, and re-measure. If it remains high after cleaning, replace the patch cords or the panel adapter. This event is also the likely explanation for why the total is as high as it is.
3. What do you do with the two defective endfaces, and why is a passing loss test not enough?
Read the worked answer
Reject both endfaces and require replacement before acceptance. A 40 micrometre pit in the core zone and a long cladding scratch are outside any reasonable acceptance criteria under IEC 61300-3-35 grading. The process is clean first with a lint-free wipe and suitable solvent, dry, and re-inspect; if the defect remains, replace the connector or patch cord, because polishing out a pit in the field is not a reliable repair. The reason this matters even though the link tested at 5.2 dB is that a contaminated or damaged endface can pass loss today and still fail in service: contamination migrates, a pit grows and scatters, and a damaged connector can damage the mating connector the next time someone plugs in. This is exactly the kind of latent defect a loss-only acceptance misses. Record the inspection results, keep the rejected cords as evidence, and require 100% endface inspection as part of the acceptance package so the owner has a baseline for future maintenance.