Link fails the loss budget

Likely causes

• Contaminated endface at one of the mated pairs

  most common

  Loss shifts by a tenth or more every time you unmate, clean and remate the same connector. Scope shows debris on the core or ferrule. Usually worse in one direction.

• Connector not fully seated or latched

  common

  Loss drops the moment you unmate and push it home firmly. LC latch that clicks when it should already have been clicked, SC not bottomed, a connector backed part-way out when somebody dressed the bundle. Often intermittent: the number moves when the cabinet door shuts or the bundle is lifted. Cheapest fix in the trade and the one techs skip on the way to re-terminating.

• Bad or stale reference, or worn reference cords

  common

  Every fiber in the panel is over by about the same amount: a flat offset, not a spread. Re-reference with fresh cords and the whole set moves together. A negative result bigger than your tester's uncertainty proves the reference path was lossier than the link.

• Unaccounted connections in the path

  less common

  OTDR shows more events than the as-built shows connections: a patch-through at an intermediate frame, or a splice nobody documented. The overage is a few tenths per extra connection, so it stacks.

• Bend in the run or in a slack coil

  less common

  1550 nm loss is higher than 1310 nm on the same fiber. On a healthy singlemode link 1550 should never read meaningfully higher.

• A genuinely out-of-spec splice or mated pair

  occasional

  One discrete OTDR event sits well above the rest, at a bulkhead or closure you can name. Other fibers in the same sheath are clean.

• Core or mode mismatch in the path

  rare

  Several dB of difference between the two test directions on the same fiber, and the same signature on every fiber through one jumper or cassette.

What to bring

• Tier 1 light source and power meter (OLTS)
• fiber inspection scope, 200x or 400x, with bulkhead tips
• cleaning cassette, 1.25 mm and 2.5 mm sticks, 99% IPA, lint-free wipes
• known-good reference and patch cords in the correct fiber type
• OTDR with launch and receive cords
• VFL
• cable print and as-built

Steps

1. 1

   Step 1: Read the result, the margin and the wavelength before you touch anything

   Write down measured loss against the calculated budget at each wavelength. A few tenths over points at connectors, a loose mate, the reference, or one connection nobody documented. More than 1 to 2 dB over points at a bend, a bad splice, a fiber-type mismatch, or several extra connections. And check which way you are out: if measured loss is lower than the budget and the link still misbehaves, you may have too much light, not too little, which is the receiver-overload sequence.

   If that doesn’t do it

   If nobody ever calculated a budget, build one from the print (length, connection count, splice count) before you re-terminate anything. Otherwise you are chasing a number no one agreed to.

2. 2

   Step 2: Reseat every connection in the path

   Unmate and remate each one firmly with the meter still reading. Listen for the LC latch, bottom the SC, push the MPO home. Watch the number as you do it. Free, one minute, and it is the whole fault often enough that it belongs before the cleaning kit comes out.

   If that doesn’t do it

   No change on any of them: scope and clean next.

3. 3

   Step 3: Scope and clean every endface in the path, both sides of every adapter

   Scope, clean, scope again. Wet-then-dry on anything showing a smear. A clean mated pair in the field usually lands around 0.1 to 0.3 dB. That is what good workmanship looks like, not the limit. The allowed per-connection loss comes from the standard edition and the job spec your contract names and it is looser than that, so read it before you condemn anything. Use 0.1 to 0.3 to judge your own work, use the spec to judge pass or fail. Include the tester's own cords and the far side of every bulkhead. That is the one people skip.

   If that doesn’t do it

   If a face won't come clean in three cycles, or shows dark pits inside the core zone, replace the jumper or re-terminate that end. If cleaning buys you a tenth and then loses it again on the next mate, work the endface sequence.

4. 4

   Step 4: Swap both patch cords for known-good ones

   Free and it takes a minute. Correct fiber type and connector type. If loss drops into spec the cord was the fault. Bin it, don't put it back in the box.

   If that doesn’t do it

   Loss unchanged means the fault is in the permanent link. If one specific port stays high with every known-good cord you try in it, the adapter-sleeve sequence covers that. Otherwise move to the reference.

5. 5

   Step 5: Re-reference with fresh cords, then retest and audit for negatives

   Use the reference method the spec calls out and record which one you used. Scope the reference cords before setting the reference. Then scan the data set for negative values. A negative bigger than your tester's stated uncertainty (for most field meters that is around a tenth, and the figure is on the instrument's spec sheet) is proof the reference path was lossier than the link, and everything taken against it needs retesting. A couple of hundredths negative is noise, not evidence. Check the source was warmed up and the temperature hasn't moved before you blame the cords.

   If that doesn’t do it

   Still failing: test in both directions.

6. 6

   Step 6: Test bidirectionally

   Several dB of asymmetry means a core or mode mismatch, or a large mode-field step: the fiber-type sequence. A few tenths of asymmetry is normal connector variation, not a fault.

   If that doesn’t do it

   Symmetrical and still high: put an OTDR on it, next step.

7. 7

   Step 7: Shoot the OTDR from both ends with launch and receive cords

   Cords must match the link's fiber type and be long enough for the pulse width, or the end connectors won't be measurable at all. Locate the event, then average the two directions before you call any single event bad.

   If that doesn’t do it

   No single event stands out but total loss is high: suspect the fiber itself. Wrong grade, wrong type, or a long stressed section. Check the cable print against the submittal.

8. 8

   Step 8: Walk the route to the event distance

   Convert trace distance to a location on the print and allow for slack coils: coils eat tens of metres and will send you to the wrong room. First suspects: cinched ties, a crushed spot, a jumper pinched in a door, a coil wound tighter than radius.

   If that doesn’t do it

   Nothing visible at that distance: recheck your slack allowance and shoot from the far end to bracket it.

9. 9

   Step 9: Re-terminate or re-splice, retest, and save the passing result

   One change at a time with a retest after each. Keep the passing result in the job file with the reference method recorded on its face.

   If that doesn’t do it

   Still failing after re-termination: escalate to replacing the section. The cable is damaged.

Check it with a calculator

• Fiber Budget Calculate optical loss

More Fiber Optic guides

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

All 54 guides