How a device decides which signal to use

A device joins one access point and stays with it until its own rules say the link has become bad enough to look for another. Those rules are conservative, because moving costs time and risks a gap, and they differ from one maker of phones and laptops to the next. An extender cannot pull a device across; it can only be there when the device decides to look.

The wireless standard puts the choice of access point in the hands of the station, the device joining. When a phone turns its radio on, it scans, builds a list of what it can hear, and picks one door into each known network, normally the strongest. From then on it is associated with that one access point and sends everything through it.

Nothing in the standard obliges the phone to reconsider while the link still works. The router does not know a better door exists; the extender does not know the phone is nearby. Each access point only knows which devices are attached to it.

Moving from one access point to another has a cost. The device must scan, which interrupts traffic; it must authenticate and associate again, which takes a fraction of a second; and if the new link turns out worse, it has traded a working connection for a poorer one. Makers therefore tune their devices to move late rather than early. The behaviour is so consistent that it has a name in the trade: the sticky client.

The result is the scene every extender owner recognises. A phone joined the router in the hall, was carried to the far bedroom, and is now holding a weak link to the router at a slow rate while sitting a metre from an extender offering a strong one. From the phone’s point of view nothing is wrong: the link still works, so there is no reason to look elsewhere.

Each maker sets its own conditions, and they are rarely published in full, but the triggers fall into a few recognisable classes.

None of these is something the extender controls. Under the two-name arrangement described in the second network name, even a lost link does not lead to the extender, because a device does not switch between differently named networks unless a person has told it to join both.

Later additions to the wireless standard let access points advise devices rather than leave them entirely to their own judgment. One provides a list of neighbouring access points so the device need not scan blind; another lets an access point suggest that a device move, and to where; a third speeds up the re-authentication when it does. Mesh systems built as a set lean on these, and on a shared controller that can see every device from every node.

A stand-alone extender generally implements little of this, and a device that does not understand the advice ignores it anyway. Some access points can also refuse to serve a device whose signal is too weak, forcing it to look elsewhere; that is a blunt tool, and a device with nowhere better to go is simply disconnected.

A dual-band device has a second decision to make: 2.4 GHz or 5 GHz. Most current phones prefer 5 GHz when its signal is adequate, because it is usually faster and less crowded, and will stay on it even as it weakens through walls. Where a router and extender publish both bands under one name, the device may therefore be on the far router’s 5 GHz signal rather than the near extender’s 2.4 GHz one, and be behaving exactly as designed.