Only Wi-Fi devices drop
The broadband line may be healthy. Focus on channels, band steering, mesh, placement or router radio stability.
If your internet keeps dropping, separate brief Wi‑Fi failures from router or mesh crashes, FTTP optical faults, FTTC retrains, electrical interference and load-triggered queueing. The pattern at the moment of the drop is the key evidence.
Dropout troubleshooting guide
If your internet keeps dropping, watch what changes at the exact moment of failure, compare Ethernet with Wi-Fi and record timing or load triggers before changing hardware.
Start here
A brief self-recovering failure needs different evidence from a total outage. Use the pattern at the moment of the drop.
The broadband line may be healthy. Focus on channels, band steering, mesh, placement or router radio stability.
The router session, ONT, copper line, power or provider path is the likely boundary.
Queue saturation or bufferbloat can make the connection appear dead even without a physical line loss.
Quick reference
The visible symptom, likely cause and next test are combined so you can collect evidence while the problem is happening.
The router radio, wireless channel, band steering, mesh roaming or local signal path is failing while the broadband line remains available.
Next: use the Wi-Fi stability guide.A router reboot, WAN or PPP session loss, DSL retrain, ONT fault or provider event is more likely than one device.
Next: record the exact light and time before restarting.The full-fibre optical path has lost signal or the ONT connection is unstable.
Next: avoid repeated reboots and report the LOS evidence if it persists.A microfilter, faceplate, extension wiring or copper line fault can disturb the DSL signal.
Next: test through a known-good filter at the master test socket.The upstream queue may be saturated, delaying DNS, calls, games and acknowledgements until the connection appears stalled.
Next: run the bufferbloat test, pause the trigger upload and use the bufferbloat fix guide if loaded latency is confirmed.The device adapter, driver, power-saving mode, saved network or security compatibility is more likely than the broadband line.
Next: compare another device in the same place and test the affected device over Ethernet.Test in order
Do not rely on a test run after the connection has already recovered. Record the device scope, lights and timing during the failure.
One device points to local hardware or software. Whole-house drops move the investigation to Wi-Fi infrastructure, router, ONT, access line or provider.
If Wi-Fi bars stay present while browsing stops, the device still reaches the router. If the Wi-Fi icon disappears, focus on the wireless path.
Note Internet, Broadband or DSL, PON, LOS and LAN/PORT behaviour at the exact moment of the drop.
A stable wired connection proves the broadband service is still present and isolates the problem to wireless radio, mesh or the device.
On FTTP, compare ONT and router WAN state. On FTTC or copper, test from the master socket with a known-good filter.
Log whether the drop follows uploads, evening usage, weather, telephone ringing, heating equipment or another repeatable event.
Record time, duration, affected devices, light state, Ethernet result and trigger. This is more useful than repeated factory resets.
Catch a brief dropout with continuous ping: open two Command Prompt or Terminal windows. Run one continuous ping to your router/default gateway and a second to a public target. The highlighted commands below test 1.1.1.1; for the router test, use the same command but replace 1.1.1.1 with your router/default gateway address.
ping -t 1.1.1.1ping 1.1.1.1If both targets stop replying together, investigate the local Wi-Fi/Ethernet/router path first. If the router keeps replying but the public target fails, the break is more likely beyond the router. Stop the test with Ctrl+C. A single missed ping is not proof of a line fault, so look for a repeatable cluster that matches the visible dropout.
Evidence matrix
Use this table while the fault is live. It turns a vague complaint into a specific next test.
| Observed behaviour | Likely boundary | Best next action |
|---|---|---|
| Wi-Fi icon disappears | Router radio crash, channel interference, mesh roaming or local signal loss. | Split bands, test near the main router, remove mesh nodes temporarily and compare Ethernet. |
| Router Internet light turns red or amber | WAN or PPP session loss, provider outage, cable fault or router instability. | Record the time and light, check the WAN lead and provider status, and avoid repeated reboots. |
| ONT LOS flashes red | Loss of optical signal or unstable fibre path. | Check the lead is not loose or sharply bent and contact the provider if LOS persists. |
| Drop follows a landline ring | Microfilter, faceplate, extension wiring or copper line fault. | Use a known-good filter at the hidden master test socket and retest. |
| Drop follows upload or backup | Bufferbloat or saturated upstream queue. | Pause the upload, run the loaded-latency test and use the bufferbloat fix guide if the result confirms queue saturation. |
| Only one device drops | Device driver, adapter, power saving, saved network or security mismatch. | Update the device, forget and rejoin the network, and compare Ethernet or another Wi-Fi network. |
Do not confuse a stall with a line drop. If service lights remain stable and the fault only appears under load, test queueing before reporting an optical or copper fault.
Advanced checks
Use these panels only after the device scope and light evidence are known.
On FTTC or copper lines, Dynamic Line Management reacts to repeated instability. Frequent manual power cycles can look like genuine line drops and may result in a slower or more protected profile.
Leave the router powered while collecting evidence unless one controlled restart is needed.
On copper and FTTC lines, faulty power supplies, central-heating thermostats or controls, pumps, switching equipment or failing street lighting can create REIN (Repetitive Electrical Impulse Noise): repeatable bursts of electrical interference that can force DSL to lose synchronisation or retrain.
Record the exact time and any nearby electrical event that coincides with the dropout. Weather-related drops can also indicate water ingress or cable movement and should be escalated with the timing log.
If PON and LOS remain normal but the router Internet light changes, the router, WAN lead or authentication session is more likely than the optical path.
If LOS changes at the same moment, preserve that evidence and report the optical fault.
If a wired PC, console or the router-to-ONT link drops intermittently, swap that Ethernet lead for a short, known-good cable before changing deeper settings. Also check for loose seating, damaged locking tabs or strain where the cable bends into the router, ONT or wall plate.
If the dropouts stop after replacing the lead, the original cable or its termination is the likely fault. Avoid non-standard CCA (Copper Clad Aluminium) Ethernet sold as category cable; standards-compliant copper cabling is the safer choice, especially where PoE is involved.
For FTTC or copper, the master test socket bypasses extension wiring and many faceplate faults. Use a known-good filter where required and compare stability over the same period.
If the master socket stabilises the line, repair or leave disconnected the faulty internal wiring.
Is the router dropping the internet or rebooting itself? If all status lights go out and then cycle through the normal power-on sequence, the router itself is restarting. If the router remains powered but only the Internet, WAN or DSL state changes, investigate the broadband path rather than assuming a full hardware reboot.
A router can freeze or reboot because of overheating, unstable power, firmware faults, memory pressure or a problematic mesh node. Keep it ventilated, check the power adapter and connection, and test the main router without extenders or secondary nodes.
If only the router or Wi-Fi layer fails while the access line stays stable, hardware replacement may be justified after the controlled test.
Cloud backups, CCTV and file synchronisation can fill a small upstream and stall interactive traffic without changing physical line lights.
Run the loaded-latency test. If confirmed, use the bufferbloat fix guide for SQM or QoS settings.
Apply the result
Apply the narrowest fix and repeat the same controlled test.
Do: separate bands, choose a cleaner channel, move the router and test without mesh or extenders.
Retest: same room and device while Ethernet remains the control.Do: update firmware, improve ventilation and simplify the topology to the main router.
Retest: leave secondary nodes disconnected for several hours.Do: pause the trigger and use the bufferbloat fix guide to configure effective queue management.
Retest: controlled upload and download load.Do: replace the filter or faceplate and keep faulty extension wiring disconnected.
Retest: the same time period from the master socket.Do: preserve the light state and contact the provider rather than repeatedly rebooting.
Retest: after the provider restores the optical path.Do: update drivers, disable aggressive power saving, forget and rejoin the network and compare another adapter.
Retest: the same device over Ethernet and Wi-Fi.Escalate with evidence
Escalate when the evidence shows the failure is beyond one device or the local Wi-Fi radio.
Date, time, duration, weather or load trigger, affected devices, Ethernet result and whether the service recovered by itself.
Router Internet or DSL light, ONT PON and LOS state, master-socket result, cable checks and whether the router rebooted.
Focused answers
These answers match the FAQ structured data in the page head.
Brief self-recovering drops usually point to Wi-Fi roaming, router instability, a provider session drop, upload saturation or an access-line retrain. Check whether all devices drop together and watch the router lights during the failure.
Yes. Weak signal, crowded channels, mesh backhaul, band steering and router radio instability can cause wireless dropouts while the broadband line and Ethernet remain stable.
Dynamic Line Management monitors line stability on copper and FTTC services. Repeated manual router reboots can look like genuine line drops, so avoid power cycling repeatedly while diagnosing the fault.
A red or flashing LOS light usually indicates loss of optical signal. Check for a loose or sharply bent fibre lead and contact the broadband provider if the light does not clear.
Contact the provider if Ethernet drops at the same time as Wi-Fi, router or ONT lights show a service fault, or repeated tests prove the problem lies beyond the home network.