Knowledge base · Measurement
Optical return loss (ORL): what the number is, and why everyone's differs
ORL is how much of the light sent into a link comes back to the source, all of it at once. One number for the whole trace, arrived at in different ways: two instruments will put different values on the same file, and both will be right on their own terms.
Check the ORL in your own file
The return loss of a link is not the reflectance of a single event: reflectance belongs to one connector or splice and sits in a row of the table, while ORL sums everything that came back, the fibre's own scattering included.
By Ilya (engineer, 6 years in fibre optics) ·
Where the returning light comes from
ORL takes in both reflected and scattered light, and how much of it builds up depends on the length of the link as well.
Reflection off end faces
Fresnel reflection wherever light crosses between media: connectors with an air gap, mechanical splices, an open fibre end, a crack. This is the largest part and the one that can be fixed; it is why angle-polished connectors go into links that care.
Scattering in the fibre
Rayleigh scattering is a property of the fibre itself, not a defect. A fraction of the scattered light heads back towards the source and lands in the ORL, along the whole length of the link.
The length of the link
The longer the fibre, the more scattered light has come back, and that is what keeps the figures apart. ORL is a property of the whole link rather than a datasheet value of a component, and it does not carry from one span to another.
Why the number depends on what measured it
Return loss is measured by two different instruments. An OTDR sorts the returning light by time and can tell where it came from. An ORL meter (OCWR, often part of an optical loss test set) sends a continuous signal and measures only the total. It measures that total directly, but it cannot tell reflection from scattering and needs the far end of the link terminated. These are two different measurements of one quantity, and they are not obliged to agree.
On the OTDR itself, the number is derived from the curve rather than measured. Into that derivation go the fibre's backscatter coefficient, taken from the instrument's settings, the stretch it was computed over, and the point it is referenced to. The file records none of the three, only the result. So different manufacturers land on different values for one span: the arithmetic is the same, and what differs is the convention about what counts as the link.
The backscatter coefficient itself is a setting of the instrument rather than a constant of the fibre: the manufacturer says outright that the operator sets it and that it is kept separately for each wavelength. The files show as much: at 1310 nm instruments write one value, at 1550 nm one a couple of decibels lower. But that knob will not reproduce someone else's number. Across our whole corpus the coefficient sits around −80 dB and varies between manufacturers by a few decibels, while the ORL figures for one span differ by tens.
It makes sense to compare an ORL figure with one obtained the same way: reflectometric with reflectometric, test-set with test-set. A discrepancy between the two methods on one link is not an error of either. And it makes sense to compare links of comparable length: on a short one the dominant contribution is the reflection at the far end, but with length it fades on the way back, and beyond roughly forty kilometres ORL is set by scattering alone, where whether the link is terminated barely matters.
The stretch is chosen by a person
In the instrument's own manual, ORL is a manual measurement like any distance: the operator puts one marker at the start of the fibre, the second at the end of the span, and the instrument prints the return loss between them. So the stretch is decided by whoever measured. Only the result goes into the file, without the boundaries it was taken over.
Next to it in the same manual is the reflectance measurement of a single event, and it works differently: the left marker on the rising edge of the spike, the right one on its peak. Reflectance is a local figure for one joint, ORL is a sum over the whole chosen stretch, and there is nothing to compare between them. Reflectance is written as a negative number, and it does not read the way it first looks: the further the number is from zero, the weaker the reflection. Return loss is written positive in reports, and the rule is the same: further from zero is better.
The same instrument computes return loss two ways at once, so the numbers differ before any arithmetic starts. The first is manual, between the A and B markers. The second takes the whole span, and there the boundary moves: it runs either from the start of the span to its end or over the whole fibre, depending on whether the span ends are included in the span itself. Change that one setting and the same fibre gives a different figure. The same manual says that when the ends are included, their loss and reflectance go into the threshold check as well, so the setting changes the verdict as well as the figure.
On our side no operator picks the stretch. The figure from the instrument's table keeps to the span markers where the instrument wrote them, and the figure from the curve runs from the start of the recording to the end of the fibre. There is one way to move the start: place cursor A, and the return loss is counted from there. The instrument counted over the stretch its operator named, and that alone makes the two figures differ, even where everything else agrees. It is one more reason to show our figure only where the instrument wrote none, and never under one label with the instrument's.
What we show
Where the instrument wrote an ORL into the file, the totals carry its number, credited to the instrument. Where it wrote none, the experimental mode shows our own figure, marked as a calculation: it is our quantity and not a substitute for theirs, and the two never stand under one label. We compute it one of two ways, and the file decides which. Where the instrument wrote an event table with at least two stretches of fibre, the figure comes from that table, by the same formula VeEX firmware uses: the backscatter of each stretch and the reflection of each event, dimmed by the loss on the way out and back. Where the table holds only the start and the end of the line, the instrument did not write its own map into the file, and the figure comes from the curve. Both ways need the backscatter coefficient out of the file; where the file states none, there is no figure: we do not fill in a plausible value for silica.
There are no threshold values in the app, for ORL or for anything else. What an installation requires of its return loss is set by the design and the type of system, and it comes from the site documentation, not from us.
Next
The sign: formulas usually write ORL as a negative number, while instruments and reports print it positive, as "return loss 45 dB". It is the same quantity: the larger the number, the less light came back and the better the link.