Bend loss: loss climbs at the longer wavelength

Likely causes

• Tie-wrap or hook-and-loop cinched down on the cable

  most common

  Non-reflective loss event that grows at 1550 and grows again at 1625. Cut the tie and the loss goes away on the retest. Look for the tie at exactly the event distance. Strictly this is a microbend: lateral crush, not a radius violation, though on a thin jumper the same tie can kink the cable into a genuine macrobend. Either way it is found the same way and cutting the restraint fixes both.

• Slack coil wound tighter than the cable's minimum bend radius

  common

  The true macrobend. Event distance lands on a slack store, a splice tray or a service loop, and several fibers in the same sheath show it at the same distance.

• Cable pulled around a sharp corner or over a conduit edge

  common

  Event at a known bend, sheath often visibly deformed or shiny. Shows in both OTDR directions, non-reflective, and wavelength-dependent.

• Jumper kinked behind a panel or pinched in a cabinet door

  less common

  Intermittent. Loss moves while you open the door or lift the bundle: the only failure mode you can watch change on the meter in real time.

• Radius violated inside a furcation or a tight termination transition

  occasional

  Event within the first few metres of a termination, and the loss only appears after the panel is buttoned up and the jumpers dressed.

• Bend-insensitive fiber masking the bend

  occasional

  Loss looks acceptable at 1310 and 1550 on a G.657 route, and the bend only shows at 1625. G.657 hides the bend optically; the glass is still sitting at a radius the cable was not designed for. The loss goes away when you release it. That part is fully reversible. The mechanical risk does not: a tight coil held under long-term tension is a future break, not a future loss reading. Recoil it anyway.

What to bring

• OLTS with 1310 and 1550, plus 1625 if available
• OTDR
• cable spec sheet for minimum bend radius
• side cutters
• hook-and-loop wrap to replace the ties you cut
• as-built

Steps

1. 1

   Step 1: Compare loss at both wavelengths on the same fiber

   Singlemode: 1550 should never read meaningfully higher than 1310, because the fiber's own attenuation is lower there. If 1550 is higher, you have bend loss. That alone is the diagnosis. On a short premises link the normal difference is small, so look at the direction of the difference, not its size. Multimode: compare 850 to 1300, same reasoning, smaller and harder to read. Worth knowing which bend you are hunting: a tie cinched onto the jacket is a microbend, a coil wound too tight is a macrobend. Both get worse as wavelength gets longer, so this comparison finds either, but the fix is different. Cut the restraint for one, recoil larger for the other.

   If that doesn’t do it

   Both wavelengths equally high: it isn't a bend. Work the loss-budget sequence.

2. 2

   Step 2: Add 1625 nm if your kit has it

   Bend loss gets worse as wavelength gets longer. 1625 exaggerates it and will find bends 1550 barely shows, which is exactly why it's the out-of-band monitoring wavelength on live systems.

   If that doesn’t do it

   No 1625 capability: work the OTDR at 1550 and hunt by distance instead.

3. 3

   Step 3: OTDR at the longest wavelength you have and note the event distance

   You want a non-reflective loss step. Reflective means a connector or a crack. Non-reflective and wavelength-dependent means a bend. Shoot both directions and average.

   If that doesn’t do it

   No discrete event but total loss is high at the long wavelength: suspect a long stressed section rather than one point: overtensioned pull, crushed innerduct, a run that was walked on.

4. 4

   Step 4: Walk to the distance and look before you cut anything

   Convert trace distance to a real location and allow for slack coils at each end: they routinely account for tens of metres and will put you in the wrong room. First suspects: ties, hangers, tray corners, a ceiling coil, a bundle crushed under a cable manager.

   If that doesn’t do it

   Nothing visible at that distance: recheck the slack allowance and shoot from the other end to bracket the event between two measurements.

5. 5

   Step 5: Release the suspect restraint with the meter still connected

   Leave the meter reading and cut the tie or unwind the coil while you watch. The number drops the instant the stress comes off. That is your confirmation and it costs nothing.

   If that doesn’t do it

   No change: reseat the fiber and move to the next suspect at the same distance.

6. 6

   Step 6: Re-dress to radius and retest at every wavelength

   Recoil the slack no tighter than the cable's own minimum bend radius, and leave yourself margin past it: that figure is a floor, not a target, and the next tech to dress that bundle will disturb your coil. The radius varies by cable construction, so read the cable spec sheet rather than working off a rule of thumb, and use the loaded figure anywhere the cable is under tension, since loaded and unloaded are different numbers.

   If that doesn’t do it

   Loss persists with all stress relieved: the glass is permanently damaged. Replace the section. If instead the loss comes and goes with the weather rather than with the stress, the water-ingress sequence is the better fit.

Check it with a calculator

• Bend & Tension Check bend limits
• Fiber Budget Calculate optical loss

More Fiber Optic guides

• Link fails the loss budget
• Dirty endface that won't clean up
• Duplex link dark: polarity reversed
• Link errors because there is too much light
• Hard break or cable cut

All 54 guides