One convenience feature cost me nine devices

One evening two complaints arrived about the same Wi-Fi:

1. “The new camera cannot be set up.” Its light went red every time.
2. “In the garden the Wi-Fi keeps dropping.”

Two different devices, two different symptoms. One cause — and it is a convenience feature the router ships with.

What band steering does

A modern Wi-Fi transmits on two bands: 2.4 GHz carries far and gets round corners, 5 GHz is fast and does not reach as far.

Band steering — TP-Link, who make my router, call it “Smart Connect” — runs both bands under one name and decides for itself where a device ends up. One network, one password, no thinking required.

It does two things while at it, and that matters later: it shares the name, and it steers the bands.

First complaint: the camera

Red on this camera means no network or no internet. Three suspects were obvious, and all three were checked first:

Suspect	Measurement	Result
DHCP range full	sweep over the 50 addresses	8 taken — no
wrong encryption	router interface	correct — no
network hidden	same page	not hidden — no

The first suspect was especially tempting because I had narrowed the DHCP range that same day. It would have been convenient to take that as the cause. It was wrong.

The cause was band steering. This camera can do 2.4 GHz only — like almost everything cheap and battery-powered. The router was pushing it onto the 5 GHz band.

A device pushed onto a band it cannot even receive never registers. It says “no network”. What it means is “no network that I can hear.”

The error in fixing it was mine

My first thought was to rename the bands as well, so they could be told apart. On opening the dialog it turned out that after switching the feature off, both bands keep the same name.

So I wrote that without renaming, switching it off would achieve “nothing at all”.

That was too absolute. The feature does two things — and the second part, the steering, falls away even when the names stay the same. That alone was enough.

So only one thing was saved: band steering off. Same name, same password, same encryption. Not a single device in the house had to be reconnected. After that the camera set up on the first attempt.

A feature that does two things can be switched off halfway. Often that is the whole intervention.

Second complaint: the garden

Here it would have been obvious to change the operating mode of the repeaters. Instead I first looked at who out there was attached to what:

Satellite	Device	Band
outside	grill	2.4 GHz
outside	phone	5 GHz
outside	speaker	2.4 GHz
garden	solar metering	2.4 GHz
garden	camera	2.4 GHz
garden	two speakers	2.4 / 5 GHz

The phone is the only device outside on 5 GHz — and the only one that moves.

The mechanism behind it: near a satellite, 5 GHz is the better choice and gets chosen. Walk further away and the device clings to that connection until it breaks. Through hedges and round corners, 5 GHz carries far worse.

Every stationary outdoor device sits on 2.4 GHz and runs without trouble.

The problem does not hit the network. It hits whatever moves.

What does not help

This is the most useful part, because both are obvious and both are wrong.

Turning up the router’s transmit power. The satellites are already at maximum, and the router is deliberately weaker so that devices prefer the satellites. A stronger router makes the handover worse, not better.

Switching the repeaters’ operating mode. It changes nothing about the band choice. And TP-Link’s own dialog states literally that after switching, the Wi-Fi initially stands at “Open”. With three devices that means: switch three times, reconfigure three times immediately — and in between, each time, an open Wi-Fi in the garden.

What probably helps: separate names per band. Then you can deliberately use 2.4 GHz outside. The price is two networks and devices that no longer switch on their own.

What I have not checked yet: whether the router offers a threshold that only assigns 5 GHz above a certain signal strength. That would be the more elegant route and belongs checked before any rebuild of network names.

The side finding is bigger than the occasion

The same network sweep, exactly the same parameters, before and after:

	before	after
reachable devices	24	33
lost	—	none

What came back: the solar metering, the stove switch, the robot mower, two outdoor lamps, one satellite and three more.

That hits precisely the old complaint that the network is “not always stable”. Band steering pushes devices back and forth between bands; for a device that can only do one of them, that is not an offer but an interference.

And the caveat belongs with it

Saving restarts the radio modules. Every device registers afresh — part of the gain may simply come from that, as a dead connection cleared out.

A gain that follows a restart is not yet a gain from the change.

So the documentation carries an open item: measure again over the coming days, with the same parameters. If it stays at 33, it was the band steering. If it falls back to 24, it was the restart.

What I’d suggest

1. First check what your device can actually do. “2.4 GHz only” is rarely on the box and explains half of all setup problems.
2. Do not suspect today’s change. Mine was obvious and innocent. Measuring cost five minutes.
3. Look at who is on which band before you rebuild anything. The answer was in the router’s device list, not in a guess.
4. Do not turn up the transmit power. In a mesh of router and satellites, a weaker router is intentional.
5. Read the warnings in your manufacturer’s dialog. “Initially open after switching” is an imposition you want to know about in advance.
6. Measure before and after with identical parameters — and write down that a restart happened in between.

So what?

Both complaints sounded like a weak network. Neither had anything to do with range.

Band steering is not a bad feature. It is a feature that assumes devices can do both bands and switch by themselves. In a household with cameras, sockets, a robot mower and solar metering, neither assumption holds for the majority of devices.

What occupies me most is not the feature but how long it could do quiet damage: nine devices were intermittently unreachable, and nobody noticed — because a device that is sporadically missing looks exactly like a device that happens to be asleep.