Symptom
A fiber port logs link down and link up events, seconds apart or minutes apart, in clusters; devices beyond the link show brief communication losses; the events follow temperature, weather, wind, traffic, or activity in a cabinet.
The short answer
Intermittent Fiber Link
A fiber link that goes up and down has a receive power that is crossing the receiver threshold, or a switch port that is being reset by something other than light. The transceiver diagnostics distinguish them: if the receive power at the moment of a drop is at or below sensitivity, the link has too little margin and something is modulating the loss, which is usually a connector that is not fully seated, a bend that moves with temperature or vibration, water in the plant, or a transceiver that is failing. If the receive power is steady and healthy through the drops, the flapping is a port or protocol event: a duplex or speed negotiation, a loop protection feature, a redundancy protocol reconverging, or a switch that is rebooting. Trend the power and the link state together and the pattern names the cause.
Key points
- Trend receive power with the link state; a drop with power at the threshold is an optical problem, a drop with healthy power is not.
- A link with little margin flaps before it fails; measure the margin and restore it.
- Connectors that are almost seated, and cords that move, are the usual optical causes.
- Temperature cycling exposes a marginal splice, water in the plant, and a transceiver near its limit.
- A port that flaps with good light has a negotiation, loop protection, redundancy, or power problem on the switch.
Possible causes and what to check
| Possible cause | What to check |
|---|---|
| Insufficient margin | Receive power within a decibel or two of sensitivity; the link drops when temperature, dust, or aging takes the last of it. Find and fix the loss. |
| Connector not fully seated or latched | A cord that was pushed in but not clicked; drops when the cabinet door closes or someone leans on the tray. Reseat every connector in the run. |
| Bend or strain that moves | A cord under tension, a cable in a tray that vibrates or moves in the wind at an aerial span; loss varies with movement. Relieve and secure. |
| Water in a splice case or cable, freezing | Drops in cold weather and recovery when it warms; a reflectometer trace with a stretch of elevated loss. Open the case, dry, and reseal or resplice. |
| Transceiver failing or overheating | Transmit power that wanders, a module temperature at its limit in a hot cabinet, errors that rise before each drop. Substitute the module and cool the cabinet. |
| Switch port autonegotiation or duplex mismatch | Link cycles with healthy light; forced settings at one end and negotiation at the other. Set both ends the same. |
| Loop protection or redundancy protocol | A ring protocol or spanning tree that blocks and unblocks the port as topology changes elsewhere; the log shows the reason. Fix the topology event. |
| Switch or media converter power | A supply that browns out under load or a loose power connector; all ports on the device blink together. Check the supply. |
| Link fault pass-through on a media converter | A fault on the copper side takes the fiber side down by design; the copper device is the real flapper. |
Optical or not
The first question decides everything else, and the transceiver diagnostics answer it. Log the receive power at both ends at a short interval, or at least read it during a bad period, and line it up against the port log. Power that dips to the sensitivity floor as the port goes down is an optical fault; power that holds steady while the port cycles is a switch fault. Doing anything to the fiber before this question is answered wastes the visit half the time.
| Pattern | Points at |
|---|---|
| Drops at night or in cold, recovers by day | Water freezing in the plant, a marginal splice, a connector contracting |
| Drops in heat | A transceiver at its temperature limit; a cabinet with no ventilation |
| Drops in wind or with traffic vibration | An aerial span under strain, a cord that moves, a loose connector |
| Drops when a door opens or someone works nearby | A cord not latched, a bend that moves, a tray that flexes |
| Drops on a schedule | A device rebooting, a backup or update job, a redundancy test |
| Drops together with other ports on the device | The switch, its power, or a protocol event, not the fiber |
| Drops with rising errors first | Margin gone: a connector degrading or a transceiver failing |
Chasing an optical flap
- 1
Measure the margin
Receive power at both ends against receiver sensitivity. Less than 3 decibels is the problem regardless of what triggers it.
- 2
Reseat and latch
Every connector in the run, both sides of every coupler, until each clicks; inspect and clean while it is out.
- 3
Move things
With the power reading in view, flex the cords, tap the couplers, open and close the door. A power dip that follows the hand is the fault.
- 4
Trace the run
A reflectometer trace during a bad period compared with the acceptance trace; the event that grew is the target.
- 5
Substitute the transceiver
A module whose transmit power wanders or whose temperature is high; swap for a known-good one and trend again.
- 6
Restore the margin
Fix the event, or fit a longer-reach transceiver pair if the link was designed too close to the edge.
Chasing a port flap
- 7
Read the log
Managed switches log the reason for a port transition: negotiation, a protocol block, an error threshold, a power event.
- 8
Match the settings
Speed and duplex fixed at both ends or negotiated at both ends; never one of each.
- 9
Look for the topology event
A ring or spanning tree that reconverges has a cause elsewhere: another link that flaps, a device that reboots, a new switch that joined with a better priority.
- 10
Check the power
Supply voltage at the switch under load, the redundant supply if fitted, the power connector.
- 11
Isolate the media converter
If one is in the path, its link fault pass-through setting and the health of the copper side.
Frequently asked questions
- The link only drops for a second or two. Does it matter?
- It matters to everything that runs over it: a redundancy protocol reconverges, controllers time out their connections and retry, a historian gets a gap, and an operator sees a communication alarm. A link that drops for a second has a fault that is growing.
- Why does it work fine when the technician is standing there?
- Because the conditions that expose the fault, cold, wind, vibration, a closed door, are not present, and because the technician reseated the connector on arrival. Trend the power and the port state through the conditions that produce the drops.
- Could the fiber be too short?
- Yes: a long-reach transceiver on a short run can overload the receiver, which produces errors and drops that look like a bad link. The receive power above the receiver overload level on the data sheet is the check, and a fixed attenuator is the fix.
- How do I tell a transceiver problem from a fiber problem without spares?
- Transmit power and temperature on the suspect module: transmit power that wanders or sits below the rating with a module temperature near its limit is the module. Fiber problems show as receive power that changes while the far transmit power holds steady.
Related topics
- High Optical LossA link that is up but running near the receiver floor: the loss budget from transmitter to receiver and what ate it. Dirty connectors, bad splices, tight bends, a run too long for the module, or a wrong fiber type. Measuring the loss and getting margin back.
- Dirty or Damaged ConnectorThe most common fiber fault: a connector end face with dust, oil, or a scratch that takes decibels from the link and damages its mate. How to inspect with a scope, what contamination and damage look like, how to clean properly, and when to replace it.
- Fiber Link DownA fiber link with no light: a cut or crushed cable, a failed or unseated transceiver, a reversed patch cord, a dirty connector, a disabled port, a mismatched module pair, or a dead media converter. Reading port state and optical power to find the break.
- Ethernet Device Drops OfflineAn industrial Ethernet device that disappears and comes back. Duplex mismatch, cabling, spanning tree, and the switch counters that identify it in minutes.
- Switch Port Errors IncrementingWhat each counter on a managed switch port means and which fault makes it grow: check errors from a damaged cable or noise, late collisions from a duplex mismatch, runts from a bad transceiver, discards from congestion, link flaps from a marginal connection.
- SCADA RedundancyWhat redundant SCADA servers protect against and what they do not, how failover works for tags, alarms, history, and clients, the network and controller layers beneath it, and why an untested failover is not redundancy.
Direct contact
Have a controls question?
Reach Eric Sullivan directly about anything on this site, a controls or automation topic, or one of his personal projects.