A repeater spends half its airtime repeating, and that half was yours

This looks at mesh networking from the practical end — what holds up once conditions stop being ideal.

A repeater hears, then says the same thing again

A simple repeater receives a frame on one radio and retransmits it on the same channel, which consumes a second slot of airtime. Because the channel can only carry one transmission at any moment, everything relayed uses twice the airtime it otherwise would have needed. The practical result is that devices behind the repeater see roughly half the throughput that the repeater itself is receiving.

Chain a second repeater behind the first and the fraction halves again, which is why long chains of them perform so badly. The halving happens even when the signal indicator looks excellent, because it is a scheduling cost rather than a signal quality problem.

A dedicated backhaul changes the arithmetic

Multi-radio nodes reserve one radio for traffic travelling between nodes and use the remaining radios to talk to your devices. Because those two conversations happen on different frequencies, they no longer take turns, and the halving largely disappears from the path.

The backhaul radio still competes with neighbouring networks on its own channel, so a congested band undermines the whole arrangement anyway. Some systems borrow the backhaul radio for clients when convenient, which quietly reintroduces the cost the system was bought to avoid. Whether a given design truly dedicates a radio matters considerably more than how many nodes come in the box.

Wire is the upgrade nobody advertises

If any node can be reached by a cable, the wireless backhaul problem disappears completely for that branch of the network. Existing television coaxial or telephone wiring can often be pressed into service using adaptors that carry data along it. Powerline adaptors use the mains wiring and work well within one circuit, though performance falls sharply across circuits or through extension leads.

In practice, a wired node behaves like a second independent access point rather than a repeater, so it adds capacity instead of dividing it. This is why a single well-placed cable through one wall frequently outperforms two more wireless nodes placed hopefully around the house.

Roaming is the client's decision again

Nodes in a mesh share a single network name, but nothing compels a device to move from its current node to a closer one. Standards exist that let the network recommend a better node and hand the connection over quickly, and support for them varies between devices.

Without that support a device holds its old node until the link degrades badly, which is usually when a call finally breaks up. Walking between nodes with a call in progress is the honest test, because a stationary speed test never exercises roaming at all.

Reducing transmit power on each node sounds counterproductive but often improves roaming, because leaving a distant node then happens sooner.

Where nodes actually belong

A node placed inside the dead spot has nothing good to relay, because it is receiving the same weak signal that you are. The right position is roughly halfway, where the node still hears the main unit clearly and can also reach the problem area.

Line of sight through a doorway beats a shorter path through two walls, which makes corridors and stairwells unexpectedly useful routes. Stacking nodes directly above one another works badly, because most antennas radiate least in the directions straight up and straight down. Every node should be tested by moving it and measuring again, since predicting indoor propagation from a floor plan is unreliable.

When a mesh is the wrong answer

If the problem is one thick wall, a directional link or a cable driven through that wall is a smaller and better fix. If the problem is the incoming line rather than the coverage, more nodes distribute the same limited bandwidth more evenly and nothing more. If old devices are the bottleneck, they will occupy airtime at slow rates no matter how many nodes are placed around them.

In the datasheet, interference from neighbours is not cured by adding transmitters; it is usually improved by choosing a quieter channel and narrowing it. Working out which of these applies takes an afternoon of measurement and saves the expense of solving the wrong problem entirely.

The takeaway

Count the hops, not the nodes; every wireless hop is paid for in airtime.

Understanding the failure mode tells you more than the feature list does.