Knowledge base · Measurement
Three lengths of one OTDR trace: acquisition, span and fibre
A single trace has three different lengths, and the file keeps them in three different places. Until they are told apart everything looks consistent: the figures sit next to each other and resemble one another. They diverge quietly.
This is not pedantry. Loss per kilometre is loss divided by a length, and which of the three goes into the denominator changes the answer several times over.
What the three are
Different questions, different figures — they coincide only on a short link shot carefully.
Acquisition length
How far the recording went: the number of samples times the sampling step. A property of the instrument's settings, not of the cable — the operator set a range and the instrument wrote up to it. The fibre may well have ended much earlier, and then the whole tail of the recording is noise past the end of the link.
Span length
The distance between the two markers the operator placed: from here to there is what we hand over. The near marker is not always at zero — a launch cable or a splice-in point may sit before it — and this is usually the figure that gets copied into the record.
Fibre length
Where the fibre physically ended: the last event, past which the curve drops into noise. It is the same figure as the span only when the section handed over is the entire fibre; on a live network the fibre often carries on beyond the span being accepted.
What it looks like in a signed record
The example is not invented — it comes from an acceptance package for eight fibres. The engineer copied two figures from the «Measurement result» block of the vendor report into the protocol: total loss and attenuation. Attenuation there is computed over the fibre length, and the report says so plainly.
Only the fibres turned out to be of different lengths. Five of the eight terminate at a closure, and for them fibre length and span length agree — 1.024 km — while the rest run on into the existing network: 1.839 and 3.954 km. For those the instrument honestly divided the loss by its own length, and 0.405 and 0.282 dB/km went into the protocol. Recomputed over the span actually handed over, they are 0.727 and 1.088 dB/km.
So the paper carries a figure four times too low, and it sits on the same line as a length it was not computed over: the neighbouring column of the protocol says 1.024 km for all eight fibres. Spotting this on paper is next to impossible — the cable's specified figure lives in another volume, and the rows look consistent. Neither the instrument nor the engineer made an arithmetic error: the lengths diverged, and nobody said so out loud.
How this shows up here
Two chips sit above the curve — Optical length and End of fibre — and they stand side by side when the figures differ. When there is only one figure there is only one chip: the same value under two names would read as a discrepancy.
Acquisition length is deliberately kept out of the summary — its place is in the acquisition conditions line above the graph and in the Trace length field in the parameters. It is the longest of the three, and it is exactly the one that can end up copied into a record.
Where the instrument recorded no length, we stay silent too: an empty field is not filled in with our estimate, because an acceptance package is signed against what is on the screen, and our end-of-fibre estimate is an estimate, not a measurement. It can be seen by turning on the experimental mode, and it is labelled as ours.
Next
A simple check for somebody else's protocol: take the total loss and divide it by the length the protocol itself names. If that does not match the printed attenuation, it was computed over a different length — and it is worth finding out which.