OM4 Certification with a 62.5 um Launch Cable and No Mandrel
You are certifying 12 fibers of a 240 m OM4 (50/125 um) backbone at 850 nm with an LED source and a power meter, following the cabling standard's multimode field-test method. The kit you were handed contains a 62.5/125 um launch reference cable, and you use it because it is the only one with the right connector. You set the reference, then measure each fiber. Results range from 0.9 to 1.7 dB, which is 0.6-1.1 dB higher than the same fibers measured last year. Fiber 5 reports -0.3 dB. The link budget for the 10GBASE-SR channel allows 2.6 dB of cabling loss, so everything passes, and your colleague says to sign off. The meter datasheet gives +/-0.2 dB accuracy at 850 nm and the source +/-0.05 dB stability over 8 hours; you are working in an unairconditioned telecom room at 31 C.
1. List the practices in this test that do not conform to a standard multimode field measurement, and state the direction and rough size of each error.
Read the worked answer
- 62.5 um launch cable on 50 um fiber. The larger core overfills the fiber under test. The mated pair loses light to the core mismatch, and the fiber is launched with an overfilled mode distribution so high-order modes leak out along the first tens of metres. The measurement reads high, typically 0.3-1.0 dB on a 240 m run at 850 nm - which is the same order as the whole discrepancy.
- No launch-conditioning (mandrel wrap or mode filter). The multimode field method requires a controlled launch, e.g. five turns on a 25 mm mandrel for 50/125 fiber, so that an equilibrium mode distribution is set. Without it the reading depends on how the jumper is draped and on the source, and it cannot be compared either with last year's data or with the budget figure, which was derived assuming a conditioned launch.
- Signing off because it passes. The 2.6 dB allowance must cover the fiber and both mated connections. A method that inflates loss is not evidence of compliance - it just hides margin you may actually need.
- 31 C ambient. Source output and meter response drift; if the room temperature moves during the run, re-set the reference. This is a small effect but it undermines repeatability.
2. Fiber 5 shows -0.3 dB. What does a negative insertion loss physically mean, and what do you do about it?
Read the worked answer
A passive link cannot amplify light, so -0.3 dB is physically impossible and is a measurement artefact. It means the measurement path delivered more light to the detector than the reference did. The usual causes, in order of likelihood:
- The reference connector was dirty or damaged when the reference was taken and was then cleaned or replaced, so the reference was set too low.
- The reference jumper or the meter port was disturbed between the reference and the measurement.
- Source output drifted upward, or the meter's zero drifted.
- A different jumper or adapter was used for the measurement than for the reference.
Action: stop the run. Inspect and clean every endface, re-set the reference with the same launch cable and the same meter port, and do not disconnect the meter end of the reference jumper for the rest of the run. Finish with a reference check - reconnect the reference cable exactly as at the start and confirm the meter still reads the reference value. If a negative number repeats on a known-good path, the meter or source is suspect and should be swapped out.
3. A difference of 0.6 dB from last year's data is quoted in the meeting. Can you defend that as a real change? Support your answer with an uncertainty estimate.
Read the worked answer
No - not from this data set.
Uncertainty contributors (1 sigma):
- Meter accuracy at 850 nm: 0.20 dB
- Source stability over the run: 0.05 dB
- Connector reconnection repeatability: 0.15 dB
- Launch-condition variability (no mandrel, uncontrolled jumper dressing): 0.20 dB
RSS = sqrt(0.20^2 + 0.05^2 + 0.15^2 + 0.20^2) = sqrt(0.0400 + 0.0025 + 0.0225 + 0.0400) = sqrt(0.105) = 0.32 dB, expanded (k=2) = about 0.65 dB.
A 0.6 dB difference sits inside the expanded uncertainty, so it cannot be called a real change in the fiber. Worse, the shift has a known systematic explanation - the launch cable core size changed and the launch condition was not controlled - so it is not random scatter at all, it is a method change.
Conclusion: the predecessor data and today's data are not comparable. Re-test with a 50 um launch cable and the mandated mandrel wrap, then compare. Do not attribute the drop to aging, moisture or damage until the method is identical.