Your Internet Drops Out Several Times a Day: Is It Your Router, Your Line, or Your Provider?

It is one of the most exhausting faults in a house, precisely because it never quite becomes an emergency. The internet works. Then for a minute or two it does not — the call drops, the stream buffers, the thing you were uploading dies — and then it comes back on its own before you have finished being annoyed about it. Three times a day, maybe eight. Nothing is ever broken long enough to prove.

Search for it and you will be handed the same list everywhere: move closer to the router, change the Wi-Fi channel, your neighbours are on the same frequency, you have too many devices, buy this app. Some of that is genuinely useful advice for a different problem. If the fault you actually have is that every device in the house loses the internet at the same moment, none of it applies, because none of it is about the thing that failed. Wi-Fi is the last few metres. What you are describing is something further up the line.

The good news is that this is one of the few home-network faults that leaves a paper trail. The box your service plugs into keeps its own record of every time it lost the connection and why, that record survives the drop, and almost nobody looks at it. It is the difference between telling your provider that the internet is bad and showing them a timestamped list of upstream ranging failures. One of those gets a technician sent out.

Before anything else: is it the whole house, or one device?

This single question decides which half of the internet you should be reading, and it takes thirty seconds to answer. Next time it happens, look at something else — a phone, a tablet, a smart TV, somebody else's laptop — while the drop is still going.

If everything is out at the same time, you are in the right guide. The fault is your router, the line coming into the house, or your provider, and nothing you change on any individual computer will touch it.

If one device drops while everything else carries on happily, stop here. That is a device problem, and it has its own causes and its own fixes — on a Windows laptop it is very often a power-saving setting that switches the Wi-Fi radio off behind your back, which we cover in detail in our guide to a laptop that keeps disconnecting from Wi-Fi. A device that stays connected to the Wi-Fi but loses the internet specifically is a third thing again.

One honest complication worth naming, because it sends people down the wrong path constantly: a Wi-Fi problem at the router can also take out the whole house, and it looks identical from the sofa. The way to separate them is a wire. If you have any device that can plug into the router with an Ethernet cable — a desktop, a laptop with an adapter, a games console — leave it plugged in and watch it during the next drop. If the wired device rides straight through the outage while the wireless ones stumble, your internet is fine and your Wi-Fi is the problem. If the wired device goes down too, the trouble is the router or the line, and that is the rest of this article.

The two-window test that splits the problem in half

This is the most useful ten minutes you can spend on an intermittent fault, and it costs nothing. You are going to leave two continuous connection tests running so that when the next drop happens, you are not guessing about it — you can see exactly which side of your own router died.

On a Windows PC, open two Command Prompt windows. In the first, type ping -t followed by your router's address, which is usually 192.168.1.1 or 192.168.0.1 and is generally printed on the router's sticker. In the second, type ping -t 8.8.8.8, which is a public address out on the internet. On a Mac, open two Terminal windows and use ping without the -t. Both windows will now print a line a second, forever, until you stop them with Ctrl+C. Leave them running and go about your day.

When the drop happens, look at what each window was doing. If the router window kept replying the whole time while the internet window filled with "Request timed out", your equipment is alive and talking — what failed is everything beyond your own front door. That is a line or provider fault, and it is the strongest single piece of evidence you can walk into a support call with.

If both windows timed out together, the failure is at or before your router: the router itself, its connection to the modem, or the Wi-Fi link between it and the computer you are testing from. That is a much more local hunt, and the later sections here are the ones you want.

And if the internet window never actually times out during what everyone else experienced as an outage, you have learned something valuable too — the connection is not dropping at all, and what you are chasing is a slowdown severe enough to feel like a drop. That is a different article, and usually a different cause.

Your modem keeps its own status page — and it works when the internet does not

Here is the part that the whole first page of search results leaves out. If your service arrives on a coaxial cable — the round screw-on connector, which around here generally means Spectrum or Cox — the modem is a small computer with a web page of its own, served by the modem itself. It has nothing to do with your internet connection, which means it still loads perfectly during an outage. That is exactly when you want to be looking at it.

On most cable modems you reach it by typing 192.168.100.1 into a browser on a device connected to that network. That address is not a guess or a trick; NETGEAR publishes it as the way into its own cable modems, along with the default sign-in of admin and password, and tells you to open the page it calls Cable Connection to see your levels. Other manufacturers use the same address with their own default password. If you are renting a combined gateway from your provider, it will have its own admin address and password printed on the sticker on the side of it.

The page you are looking for lists your downstream and upstream channels with a power figure for each. Two numbers matter. NETGEAR states its guidance plainly: good downstream power levels are within minus seven to plus seven dBmV, and good upstream power levels are within thirty-eight to forty-eight dBmV. Most modems also show a signal-to-noise figure per channel, and there what matters is less the absolute number than the consistency — the channels should broadly look like one another. One or two channels sitting well away from the rest is a real signal, not a rounding error.

A caveat to carry into any conversation about those numbers: your provider publishes its own acceptable range for your plan and your equipment, and theirs is the one that decides whether they act. The manufacturer's figures are a sanity check for you, not an argument you can win with. What they are genuinely good for is the pattern. Write the numbers down when everything is working, then look again during a bad spell. Levels that drift a long way between the two readings tell a story that a single snapshot never will.

The event log is the part that settles the argument

Buried a little deeper on that same modem page is an event log, and this is the single most underused diagnostic in home networking. On a NETGEAR cable modem it lives under Advanced, then Administration, then Event Log; other makes put it under Status or Troubleshooting. It is a timestamped list of everything that has gone wrong on the line, and it has been filling itself in without being asked.

You will not need to understand most of it. Three entries are worth recognising. A T3 timeout means the modem sent sixteen ranging requests to the provider's equipment and got no reply to any of them — typically caused, in the standard DOCSIS description of it, by noise on the upstream swallowing those messages, or by the modem being unable to shout loudly enough to be heard. A T4 timeout is the more serious sibling, logged when the modem gets no maintenance opportunity at all inside a thirty to thirty-five second window, and it is usually followed by the modem resetting its connection entirely — which is the minute or two your household experienced as the internet going out. Uncorrectable codewords are the third: your modem repairs a certain amount of damaged data automatically, and this counter records what was too damaged to repair. A count that climbs while you sit and watch means interference happening now.

What you do with that is simple and disproportionately effective. Note the times of a handful of those entries. If they line up with the moments the house lost the internet, you no longer have a vague complaint — you have a fault with a mechanism and a timestamp, and it is on the provider's side of the modem.

One thing to be careful of: a modem reboot usually wipes the log and resets the counters. Read it before you power-cycle anything, or you will be throwing away the evidence you came for.

If you are on fibre, none of that applies — check the ONT instead

If your service arrives on fibre, which around here mostly means AT&T Fiber or Frontier, there is no modem and no signal-level page to read. The box on your wall is an ONT, an optical network terminal, and it belongs to your provider. It has status lights rather than a dashboard, and what they mean varies by model, so your provider's own help page for your specific unit is the thing to look at rather than a generic chart.

There is one gotcha here that catches people constantly, and it is worth knowing before you spend an evening on it. Fibre installations frequently include a battery backup unit so the ONT keeps working in a power cut. That means unplugging the ONT from the wall does not necessarily turn it off — the battery calmly keeps it running, and the reboot you thought you performed never happened. Frontier spells the procedure out on its own support page: if you have a battery backup unit, remove one of the battery leads so it stops powering the ONT, unplug the ONT and wait at least five minutes, plug it back in, reconnect the battery lead after thirty seconds, then wait at least five minutes again for service to come back. Those waits are not padding. Rushing them is why the reboot "did not work" the first three times.

The other fibre-specific thing worth checking is physical and unglamorous. The fibre itself is glass, it does not like being bent tightly, and the patch lead between the wall and the ONT is often the most abused cable in the house — coiled behind furniture, pinched under a desk leg, tucked into a cupboard door that closes on it every day. Intermittent fibre faults are sometimes exactly that mundane.

The things inside your own house that cause this

On cable service, a surprising share of intermittent drops are caused within a few metres of the modem, and they are cheap to rule out before anyone gets sent.

Splitters are the first suspect. Every time the coax is split, both legs lose signal — roughly three and a half decibels per output on an ordinary two-way splitter, and considerably more on a three-way. A house that has been added to over the years accumulates these: one for the TV that is not there any more, one from a satellite installation, one inside a wall plate. Take the modem back to the simplest possible path you can manage, ideally straight off the first split, and see whether the drops stop. If a splitter is genuinely needed, the cheap loose ones from a big-box shelf are the ones that fail.

Then the connectors themselves. Every screw-on coax fitting should be finger-tight and then a nudge more with a wrench, and the little centre pin should be straight and proud. A connector that is merely spun on by hand, or that has a bent pin, or that has been sitting in an outdoor box weathering for fifteen years, produces exactly this fault — fine most of the time, terrible when something moves or heats up.

Anything actively powered on the line deserves suspicion too. An old signal amplifier or distribution amp installed for a cable-TV setup that no longer exists can cause more problems than it solves on a modern two-way service, and its power supply is a common failure point. So is the modem's own power adapter, which is the single most overlooked cause of a device that reboots at random. If you have a spare of the right rating, swapping it is a five-second test.

Finally, where the equipment lives. Passively cooled boxes stacked in a closed cabinet, on top of one another, in a garage or a utility closet, will misbehave when the room gets hot — and they will do it in the afternoon and be fine by morning, which reads as a mysterious pattern rather than a temperature problem.

The pattern tells you more than the symptom does

Two households can describe an identical fault and have completely different causes, and the thing that separates them is when it happens. It is worth keeping a note for a few days — just the times, on a phone or a bit of paper by the router.

Drops clustered in the evening, roughly the same hours each day, and especially if a neighbour on the same street reports the same thing, point away from your house entirely. That is the shape of a congested or faulty segment serving the whole neighbourhood, and it is a category of problem that no amount of rebooting your own equipment will improve.

Drops that track the weather are their own tell. Something that gets worse in heat is usually a connection or an amplifier expanding; something that gets worse after rain is usually water finding its way into an outdoor fitting or a damaged drop cable — and in this part of the world that means the fault appears once in January and then hides until the following winter, which is a genuinely hard thing to get a technician to catch.

A connection that degrades over days and comes right after a reboot is a different animal again, and it usually means aging equipment rather than a line fault. If the fix in your house has quietly become "restart the router every few days," that is not a workaround, it is a symptom — and if the router is more than about five years old, its age is a reasonable first suspect.

And drops at a precise, repeating time — the same minute every night — are worth taking literally. Some provider gateways run a scheduled overnight reboot or firmware window, and some households discover that the culprit was a timer switch or a smart plug that somebody set up months ago and forgot.

What to say when you call, so you get a technician and not a modem swap

The default response to "my internet keeps dropping" is a remote reset, then a new modem. If the fault is in the line or further out, both of those will fail and each round trip costs you a fortnight. What changes the outcome is walking in with specifics.

Give them three things. The times of the drops, from your own notes. What the modem's event log recorded at those times, by name — say T3 timeouts or T4 timeouts, and say the uncorrectable count is climbing, if that is what you saw. And your downstream and upstream levels as they read during a bad spell. Then ask, in those words, for a line and signal check rather than an equipment replacement.

It also helps to state the thing you already ruled out: that a wired device dropped at the same moment as the wireless ones, so this is not a Wi-Fi complaint. That one sentence skips an entire script.

Check the provider's own app or status page while you are at it, both for a declared outage in your area and for a maintenance window. And if the fault has run for weeks, it is entirely reasonable to ask for a credit for the period — providers do issue them, but generally only when asked.

One last piece of realism about intermittent faults: the technician who arrives will very likely find everything working. That is normal, and it is why the log matters. A visit with timestamps and channel readings in hand is a different visit from one where somebody runs a speed test on the doorstep, gets a good number, and leaves.

When it really is your own equipment

Not every whole-house drop is the provider's, and it is worth being honest about the cases where the call would waste everyone's time.

Routers wear out, and they tend to do it exactly like this — stable for years, then increasingly unreliable under load, needing more frequent restarts. Heat and a failing power supply are usually what is actually happening inside. If yours is old enough that its manufacturer has stopped issuing firmware for it, that is both a reliability problem and a security one.

Two routers running at once is the other common self-inflicted version, and it is very easy to end up with after switching providers or adding your own router behind a provider gateway. It produces odd, selective failures rather than clean outages, and it is worth counting how many boxes in your house are actually doing routing.

Mesh systems have their own variety of this. If the satellite units talk back to the main one over Wi-Fi rather than a wire, that link is a normal Wi-Fi connection with all a Wi-Fi connection's frailties — and when it wobbles, every device attached to that satellite drops while the rest of the house is unaffected. That looks like a whole-house outage to whoever happens to be at the far end of the house.

And extenders, particularly the cheap plug-in kind, are a frequent cause of a fault described as constant dropping when what is really happening is devices being handed back and forth between two networks with the same name.

The Southern California version of this

A few local things shape how this fault shows up in the homes we work in. The housing stock is the biggest one. A great many of the houses here went up between the 1950s and the 1970s, and the coax in them was run for cable television long before anybody needed it to carry a two-way internet service. Splitters in wall voids, a run out to a converted garage or a casita added later, fittings in an outdoor box on a sun-facing wall — that is the classic cable-side profile of an intermittent drop, and it is why the walk-the-cable step is worth doing here before assuming the fault is out in the street.

Heat is the other. In the inland valleys and the desert cities, a modem and router sitting in a garage or a closed cabinet through an August afternoon are operating well outside what they were designed for, and the resulting pattern — solid all morning, flaky from mid-afternoon, fine again after dark — gets misread as a provider problem for months. Moving the equipment somewhere with air around it is free and fixes a genuine share of these.

And the mix of services matters. Fibre has reached a great many more streets here in the last couple of years, so the two houses on either side of a fault can be on completely different technologies, which makes the neighbour comparison less useful than it used to be. If you are asking around the street to work out whether it is just you, ask what service they are on before you read anything into the answer.

How we can help

We do not sell internet service and we get nothing from any provider, so we have no reason to tell you the answer is a new plan. Most of the time the answer is a specific loose fitting, a splitter that should not be there, a router past its useful life, or a line fault that needs to be proved rather than described.

What we do on these is the diagnostic half that is hard to do from a phone script: set up the logging so the drops are captured with times, read the modem's levels and event log, separate a Wi-Fi fault from an internet fault properly, walk the cable path in the house, and hand you a written summary you can take straight to your provider so the visit produces a line check instead of another box.

We look after home and small-business networks across Southern California and the Coachella Valley, in person or by remote support. If the internet has been dropping for weeks and you are on your third replacement modem, that is a good moment to have somebody look at the evidence rather than the equipment.