In short
A good fusion splice is barely measurable: the low fractions of a decibel commonly specified in industry standards. Loss well above that level, or loss that drifts over time, is not normal wear. It points to preparation, cleave quality, fibre mismatch or mechanical stress, and it is worth investigating before it becomes an outage.
A splice schedule lands on your desk and one number stands out. It is higher than the others, nobody has flagged it, and you are left wondering whether it matters. Or a link that tested clean at handover has started dropping packets, and the finger is pointing at a joint somewhere in the route.
Splice loss is one of the few honest measurements in a fibre network. It does not care what the installer promised. This article explains what the number means, what acceptable looks like, and what a high or unstable reading is actually telling you about the workmanship in your network.
It is written for the people responsible for commercial and infrastructure networks: data centres, campus estates, rail and industrial sites. If you are dealing with a home broadband fault, your internet provider is the right call, not a splicing contractor.
What is splice loss?
When two fibres are joined by fibre optic splicing, a tiny amount of the light passing through the joint is lost. That loss is measured in decibels, and every splice in a route adds its own contribution to the total loss budget of the link.
A fusion splice made properly, on prepared, clean, well matched fibre, loses almost nothing. The joint is glass welded to glass. The measurement exists precisely because the difference between a good splice and a poor one cannot be seen by eye: it only shows up on the instruments.
That is why splice loss figures belong in the handover pack. They are the evidence that each joint in your network was made with care, and the baseline you will compare against when something changes later.
What level of splice loss is acceptable?
Industry standards commonly specify maximum values for fusion splices in the low fractions of a decibel, and a well made splice usually sits comfortably inside them. The exact figure your project works to should come from its specification: the loss budget for a long external route, a high-count data centre backbone and a short campus tie cable are not the same.
Two things matter more than any single number. First, consistency: a schedule full of similar, low readings tells you the work was controlled and repeatable. Second, the total: each joint spends part of the link's overall loss budget, so a handful of mediocre splices on a long route can add up to a link that passes at every joint and still fails end to end.
Be wary of any contractor who cannot tell you what values they work to, or who tests only the joints that look suspect. Every splice should be measured, recorded and handed over.
What is high splice loss telling you?
A high or unstable reading is a symptom. These are the usual causes, and each one points at a different fix.
Poor preparation or contamination
Dust, moisture or residue on the fibre end faces scatters light at the joint. Preparation discipline is the single biggest factor in splice quality, which is why a rushed job so often shows up first in the loss figures.
A poor cleave
The fibre ends must be cut cleanly and squarely before fusion. A chipped or angled cleave leaves a joint the splicer's arc cannot fully correct, and the loss reading records the result.
Fibre mismatch
Joining fibres with different core geometries or from mixed cable types can produce loss even when the splice itself is well made. It matters on routes that have been extended or repaired over the years.
Stress and bending near the joint
A splice that measured well on the bench can read badly once it is closed up. Tight bends, pinched tubes and poorly seated splice trays add loss after the event, and can worsen as the enclosure ages.
Damaged or drifting protection
Water ingress, movement or heat-shrink problems in a joint enclosure show up as loss that rises over time. A reading that was stable at handover and is climbing now is a maintenance flag, not a mystery.
Equipment condition
Splicing machines need calibrated arcs, clean V-grooves and serviced cleavers. A run of high readings from one day's work often says more about the kit than the fibre.
What should you do about a high reading?
Start by confirming it. OTDR testing from both ends of the link shows whether the loss really sits at the splice, because a single-direction trace can misread a joint between dissimilar fibres. If the joint is confirmed, the fix is usually simple: open the enclosure, inspect, and re-splice with proper preparation.
If the network is live and the symptom is intermittent, treat it as a fault rather than a tidy-up. Loss that moves with temperature, weather or vibration usually means stress somewhere in the joint, and structured fibre fault finding will locate it faster than opening enclosures on instinct.
Above all, keep the records. A splice schedule from handover turns every future investigation from guesswork into comparison.
FAQs
Is zero splice loss possible?
No joint is truly lossless, but a good fusion splice comes close enough that the reading is barely measurable. If a schedule shows every splice at exactly zero, be suspicious of the testing rather than impressed by the splicing.
Are mechanical splices worse than fusion splices?
Mechanical splices typically carry higher loss and are generally treated as a temporary or restoration measure in commercial networks. For permanent infrastructure, fusion splicing is the expected standard.
Can splice loss get worse over time?
Yes. Water ingress, movement, temperature cycling and stressed fibre in enclosures can all push a reading up over the years. That is why handover records matter: a drifting splice is easy to spot when you have the original figure to compare against.
Do I need every splice tested, or just a sample?
Every splice. Sampling tells you about the sample, and the joint that takes your network down will usually be one nobody measured. Full test records are the difference between evidence and hope.
Splice readings you are not sure about?
Send us the schedule or tell us about the link and we will give you a straight answer on what it means.