Two OTDR Directions Disagree on One Splice
A 1.8 km single-mode link between two buildings has one fusion splice at about 0.9 km and a patch panel at each end. The acceptance specification is a mean splice loss of 0.15 dB or less at 1550 nm. Test file A, taken from end A, reports the splice at 0.42 dB. Test file B, taken from end B, reports the same splice at 0.06 dB. The technician averages the two figures, gets 0.24 dB, books the splice as failed and orders a re-splice. Before the crew arrives you review the files. File A used a 30 ns pulse at 1550 nm with 30 s averaging. File B used a 100 ns pulse at 1310 nm with 10 s averaging, and in that file the splice appears only 6 m after the launch connector. File B end-to-end loss reads 0.62 dB against a calculated 0.58 dB.
1. Is the 0.24 dB average a valid basis for rejecting the splice? Justify your answer.
Read the worked answer
No. An average is only meaningful if both measurements were made on the same optical conditions, and here they were not.
Three defects in the comparison:
- Different wavelengths. File A is 1550 nm and file B is 1310 nm. Splice loss and Rayleigh backscatter level both vary with wavelength, and the backscatter-coefficient mismatch between the two fibres differs between the bands. Averaging across wavelengths does not cancel anything.
- Different pulse widths. A 100 ns pulse is roughly 10 m wide; a 30 ns pulse is roughly 3 m wide. The event resolution is not comparable, and the wider pulse smears the splice along the trace.
- The splice sits inside a dead zone in file B. At 6 m after the launch connector, the splice is within the attenuation dead zone of that connector's reflection, so the reported 0.06 dB is not a resolved measurement. It may be inflated or masked.
Cross-check: the end-to-end loss of 0.62 dB against a calculated 0.58 dB is a sensible link, which does not support a 0.42 dB single splice. The measurement is suspect before the fibre is.
So the correct conclusion is not 'splice fails' but 'measurement invalid, re-test'.
2. Describe exactly how the splice should be re-measured so the result is defensible.
Read the worked answer
- Same wavelength: 1550 nm in both directions, as the specification requires.
- Same pulse width and range: for example 10 to 30 ns with the same range and resolution in both files, so the event is resolved to a comparable length.
- Same averaging: at least 30 s, and note the laser and the fibre type used for the launch and receive leads.
- Get out of the dead zones: add a launch fibre of 500 m to 1 km at each end so the splice is far from the launch connector reflection, and a receive fibre so the far end is also clear.
- Use a consistent measurement method: the least-squares or two-point method over the same span either side of the splice, in both directions, and check that the splice is not inside a ghost or a reflection.
- Calculate: splice loss equals the mean of the two directional values, (L_AB + L_BA) / 2, because backscatter-coefficient mismatch adds to one direction and subtracts from the other.
- Sanity-check against the link budget: the end-to-end loss should agree with the sum of fibre, splice and connector losses to within the OTDR's uncertainty.
- Retain both traces with the file names, instrument serial number, wavelength, pulse width, averaging time and date.
3. After re-measurement the values are 0.19 dB from A and 0.11 dB from B at 1550 nm. Decide whether to accept or re-splice, and state what you record.
Read the worked answer
Calculation: mean = (0.19 + 0.11) / 2 = 0.15 dB.
Decision. The splice is exactly on the limit, not over it, and the test uncertainty of a good OTDR is around plus or minus 0.05 dB. It should be accepted only after considering two things:
- The end-to-end link should have adequate margin. If the link loss budget leaves comfortable headroom, a 0.15 dB splice is normal for a field fusion splice and a re-splice is unlikely to improve it much.
- The visual record of the splice, and the proof test, should show no bubble, bulge or offset. If either is poor, re-splice regardless of the number.
Rejecting a splice on a single direction or on a mixed-wavelength average is the wrong call, and it costs a truck roll and a service interruption.
Record: both directional values, the mean, the wavelength, pulse width and averaging used, the launch and receive fibre lengths, the instrument identification, the calculated end-to-end loss, the as-found and as-left figures, and a note that the result is at the limit and accepted against the specification with the reason. If the result had been above 0.15 dB, the correct entry would be a fail with both traces attached, and a re-splice booked.