WiFi that survives plaster, wire lath and four hundred neighboring networks

It depends on walls and on how many devices are active at once, and never on square footage. The coverage figure on a box is measured in a space with almost nothing in it. In an open modern office, one access point per 2,000 to 3,000 square feet is a reasonable starting point. In a pre-war building with plaster over wire lath, it can be one per two or three rooms, because the walls behave like partial screens rather than like drywall.

Density changes the answer independently of coverage. Forty people in a conference room all on video is a capacity problem: every device shares airtime on the same channel, so the fix is another access point with a deliberately small cell, not more power on the existing one. We give a number after a predictive model and a walk with a meter, and the number after the walk is the one that goes on the quote.

For a two-bedroom apartment with drywall partitions, mesh is often genuinely fine and we will tell you so. For anything larger or older, the arithmetic works against it. A repeater that talks to your device and to the upstream node on the same radio has to send every frame twice, which roughly halves throughput per hop; two hops leaves about a quarter. Tri-band units with a dedicated backhaul radio avoid the halving, but that radio still has to cross the same wall that blocked the client signal in the first place.

The practical result in a lath-and-plaster building is that the node either sits where it has good coverage and a weak uplink, or where it has a good uplink and does not reach the room you bought it for. Meanwhile a Cat6 drop typically runs $150 to $400 installed, terminated and tested, subject to survey (roughly what three consumer mesh units cost) and it still works in ten years when the access point on the end of it is obsolete. We do use mesh, deliberately, for the one position where cable genuinely cannot go.

Because it is behind more wall than it can afford. At the same distance, 5 GHz is already about 6 to 7 dB down on 2.4 GHz purely from frequency, and it loses far more crossing masonry, terra-cotta or metal lath. In the room next to the access point, 5 GHz will be much faster. Two lath walls away, 2.4 GHz gets through and 5 GHz does not, so the device falls back and appears slower.

The instinct is to raise transmit power, and that makes it worse in a way that is hard to see. The access point can now be heard further away, but your phone has a small antenna and much less power, so it cannot answer from where it now sits. You get a large cell full of devices with one-way audibility, all consuming airtime at very low data rates and slowing down everybody else. The real fix is another access point closer to the problem room, at lower power, on a different channel. Almost every “we need a stronger router” conversation ends there.

Often, yes, and we will test rather than assume. Cat5e carries a gigabit to the full 100 m channel and will carry PoE+ to an access point that needs 25.5 W. If the runs test clean and are the right length, reusing them saves the most expensive part of the job.

Two things push us the other way. First, uplink speed: a current tri-band access point can move more than 1 Gbps across its radios in a busy space, so a gigabit uplink becomes the constraint exactly where you needed the capacity. That argues for Cat6A and a multi-gig switch port at those positions. Second, power: if a unit wants 802.3bt, high-power PoE in a dense bundle is a heat question, and larger conductors handle it better. Our normal recommendation is to reuse existing Cat5e for the ordinary positions and pull new Cat6A to the two or three access points that carry the load.

Almost never, in a pre-war building. Corridor access points have to get through the corridor wall, then a solid apartment entry door, then the unit’s own partitions. Residents at the front of a unit get nothing usable, and because every apartment still runs its own router on the same three crowded 2.4 GHz channels, the building system tends to add interference rather than replace it.

The two designs that do work are per-unit coverage (a cable drop and an access point inside each apartment, centrally managed, which is a genuine riser project with a genuine budget and makes sense mainly where the building intends to provide internet as a service) or amenity coverage, meaning properly engineered wireless in the lobby, laundry, gym, roof deck, package room, community room and garage. When we ask boards what problem they are actually trying to solve, amenity coverage is the answer about nine times in ten, and it costs a small fraction of the alternative.

They can share the physical cabling and switches. They should not share a network segment. Building systems belong on their own segments with rules that let each one reach only what it needs: cameras to the recorder and a time source and nothing else, no route out to the internet at all; access control to its controller and its management platform; guest wireless isolated from every one of them and from your own devices. And nothing production sits on the default VLAN, because that is the first place anybody looks.

The reason is not theoretical. Cameras and building controllers are among the most-exploited categories of connected device, they often run firmware that lags badly, and a flat network means anything that reaches the guest WiFi also reaches the door system. Segmenting it is a configuration decision made at design time and costs essentially nothing; retrofitting it after a building has run flat for five years is a project. This is one of the reasons we prefer to do the wireless, the cabling and the security systems as one scope rather than inheriting three vendors’ assumptions.