Re-Routed 1 Gbps Backbone: Link Budget and Reach
A campus backbone link at 1 Gbps uses 1000BASE-SX (850 nm) over 50/125 um OM2. As built: 480 m from the main cross-connect, through one intermediate cross-connect, to the telecom room, with 4 connectors total and 2 fusion splices. Design inputs: transmitter minimum output -9.5 dBm, receiver sensitivity -17 dBm, worst-case fiber attenuation 3.0 dB/km at 850 nm, 0.5 dB per connector, 0.15 dB per fusion splice, and a company policy of at least 3.0 dB end-to-end margin. The customer's change order adds a second intermediate cross-connect (2 more connectors) and 70 m of additional route. Patch cords at each end total 6 m. The fiber is confirmed good and the existing link currently runs error-free.
1. Run the numbers for the change order and state whether you sign off on the design staying at 1000BASE-SX. What is your recommendation?
Read the worked answer
Budget
- Allowed loss = Tx min - Rx sensitivity = -9.5 - (-17) = 7.5 dB
- With the 3.0 dB policy margin, usable loss = 4.5 dB
Baseline (480 m)
- Fiber: 0.480 km x 3.0 = 1.44 dB
- Connectors: 4 x 0.5 = 2.00 dB
- Splices: 2 x 0.15 = 0.30 dB
- Total = 3.74 dB, margin = 7.5 - 3.74 = 3.76 dB (passes)
With the change order (550 m route + 6 m patch cords = 556 m channel)
- Fiber: 0.550 km x 3.0 = 1.65 dB
- Connectors: 6 x 0.5 = 3.00 dB
- Splices: 2 x 0.15 = 0.30 dB
- Total = 4.95 dB, margin = 7.5 - 4.95 = 2.55 dB (0.45 dB short of policy)
Loss is not the deciding issue - distance is. 1000BASE-SX is specified to 550 m over 500 MHz*km 50 um fiber. The channel is 556 m, so it is out of specification even if the loss worked. Modal bandwidth is fine (500/0.556 = ~899 MHz available against ~500 MHz needed for 1 Gbps); the reach limit, not bandwidth or loss, is what fails.
Recommendation: do not sign off as SX. Options, best first:
- Shorten the route back under 550 m including patch cords.
- Convert the backbone to single-mode with 1000BASE-LX10 (or 10GBASE-LR if upgrades are planned).
- If multimode must stay, splice through the new cross-connect rather than patching it (2 splices at 0.15 dB instead of 2 connectors at 1.0 dB, saving 0.7 dB) - but this fixes only the loss, not the 556 m reach violation. Note that 1000BASE-LX over multimode is also a 550 m specification, so moving to LX on the same glass does not solve it.
2. After installation, an OLTS reads 9.8 dB end-to-end. The OTDR shows no discrete high-loss event anywhere on the span. What do you check first, and why?
Read the worked answer
The predicted loss is roughly 5 dB, so there is about 4.8 dB of excess that the OTDR cannot localize. A single connector or splice defect would show as a step on the trace; a distributed-looking 4.8 dB with no event points first at the test method, not the cable.
Check in this order:
- Reference/zeroing - verify one-jumper vs two-jumper reference and re-reference the meter. An incorrectly referenced OLTS can add or drop exactly the connector losses at each end.
- Launch condition - multimode 850 nm measurements require the launch condition the meter was calibrated for (typically a mandrel wrap per the manufacturer's instructions). An overfilled launch inflates measured loss by several dB.
- Endface condition - inspect both ends with a scope. Two dirty UPC endfaces at 1-2 dB each explains the gap cheaply.
- Settings - wavelength, fiber type, and calibration factor matched to the meter and to the fiber under test.
- Only then look for distributed loss: tight bends, cable crushed or kinked, cable pulled over a sharp support, or a splice tray radius violation.
Re-test, then cross-check the OLTS number against the OTDR two-point span loss plus the two launch/receive connectors. If those agree, the 9.8 dB is real and you segment the link (test each section between cross-connects) to localize it. Document the corrected method in the test report; a rejected link caused by an unverified meter is an avoidable re-pull.