Why British houses are so hard on WiFi

/01 — HOW TO READ THE NUMBERS

What a decibel actually costs you

Signal loss is measured in decibels, and decibels are logarithmic — which is why the numbers below are more brutal than they look.

• 3 dB halves the power. 10 dB leaves you a tenth. 20 dB leaves you a hundredth.
• Every 6 dB is equivalent to doubling the distance from the router. So a 30 dB wall is roughly like moving the device thirty-two times further away.

There is no UK or EU standard test method behind the coverage figure on a box. The only meaningful design target is the signal at the device: the long-standing professional planning figure is −67 dBm at the edge of coverage, with 15–20% overlap between cells. Above about −75 dBm a device will connect but will not deliver anything like its rated speed — which is exactly the state that produces “full bars and nothing works”.

And frequency runs the wrong way for modern equipment: the same wall costs more decibels the higher the band. The 6 GHz band on a Wi-Fi 6E or Wi-Fi 7 system is the worst of the three at getting through old British masonry, even though it is the fastest. This is why expensive new kit sometimes performs worse than the old router in a period property.

Even the international standards bodies now acknowledge this: the ITU building-entry-loss model formally splits buildings into “traditional” and “thermally efficient” populations, and states that where modern thermally-efficient methods are used — metallised glass, foil-backed panels — the loss is significantly higher.

/03 — WALL BY WALL

British housing stock, wall by wall

• Solid-brick Victorian and Edwardian walls are typically 215 mm or 328 mm plus plaster. Plan on roughly 30–40 dB at 5 GHz per wall — an inference from the measured table above, not a measurement of that wall type. Two of those and 5 GHz is finished.
• Party walls in terraces and semis: treat as opaque at 5 and 6 GHz. Never plan on reaching through one.
• Stone cottages — Purbeck and Portland limestone, granite, sandstone — commonly have 450–600 mm rubble-filled walls. There is no clean measured dataset for random rubble stone, and we are not going to invent one. Rule of thumb: one access point per room cluster, wired from the start.
• Flint and cob, typically 450–900 mm, hold moisture — and water is the strongest absorber in this frequency range, so damp walls attenuate more than dry ones and vary with the season. No verified figure exists for cob.
• Chimney breasts are solid masonry columns running the full height of the house. There is a permanent dead zone directly behind them on every floor. Move the node; turning the power up will not help.
• Concrete floors — 1960s and 70s flats, beam-and-block, modern slabs: assume vertical coverage between floors does not work. Steel reinforcement mesh has a spacing far smaller than the wavelength, so it behaves as a solid conductive sheet.
• Expanded metal lath in late-Victorian to post-war ceilings and partitions is effectively a Faraday screen per room. The giveaway symptom: signal dies at exactly the wall line with no gradient, and 2.4 GHz behaves no better than 5 GHz.
• Extensions: the old external wall is now an internal wall — and it is still 215 mm of solid brick. Treat it as an external wall when planning.

/05 — HOW MANY ACCESS POINTS

Node counts, outbuildings and listed buildings

Rules of thumb — every one of these is a rule of thumb, not a measurement:

• Post-2000 timber-frame semi, around 80–100 m²: router plus one node, wireless backhaul acceptable.
• 1930s cavity-brick semi, around 100 m²: router plus one or two.
• Victorian solid-brick terrace over three floors: one access point per floor, wired. The staircase is your only propagation path.
• Stone cottage with 450 mm walls: one per one or two rooms, wired.
• Anything with foil-backed board, metal lath, underfloor heating foil or concrete floors: per floor and per zone, wired, regardless of floor area.
• Bungalows have no vertical problem but a long horizontal run through a spine wall. One central unit often does a 1930s bungalow; a long 1960s one with masonry internal walls needs two, at opposite thirds.

Keep 2.4 GHz switched on in old housing stock. Do not let anyone talk you into disabling it — in a solid-wall house it is often the only band that reaches.

Garden offices and outbuildings. The path usually crosses at least one 215–330 mm masonry wall and one low-E glazed unit — easily 50–70 dB combined at 5 GHz. This is a building-fabric problem, not a product problem, and no mesh system solves it. In order of preference: external-grade or direct-burial Cat6 in ducting to an access point in the outbuilding; fibre for long runs or lightning exposure; a point-to-point link with genuine external line of sight — mounting one on an internal windowsill pointing through coated glass throws away 20–35 dB before it starts; or a separate mobile broadband service.

Garages are usually single-skin with a metal door acting as a reflector — the fix is a cabled access point inside, not a stronger signal from the house. Steel container offices are close to a sealed Faraday cage: the antenna has to be outside the metal skin, cabled through a gland.

A legal point that catches people out: a new electrical circuit supplying a detached outbuilding is notifiable under building regulations in England and Wales and needs a registered electrician or prior notification. Data cabling alone is not notifiable; power is. Any external Ethernet run also needs surge protection.

Listed buildings: the guidance is mitigation, minimisation and reversibility — route through existing voids and service routes, and use skirtings and mouldings as concealed paths. Upgrading wiring does not automatically require consent, but cutting into original plasterwork, stonework or beams alters the fabric and can. Speak to the local conservation officer first: unauthorised works to a listed building are a criminal offence. Anyone who tells you cabling “never” needs consent is not someone to take advice from.

/07 — PLACEMENT & COOLING

Where equipment fails in summer, and what to do

The placements that cause summer faults: airing and meter cupboards (no ventilation, plus a hot cylinder); sunlit windowsills; on top of or underneath a TV, set-top box, console, amplifier or NAS; sealed AV cabinets with a glass door and no rear cut-out; equipment stacked so each unit blocks the next one’s vents; soft surfaces — TP-Link’s own guidance says not to place units on paper, cloth or fabric, and carpet is a common offender; dust blanketing the heatsinks; and anything active left in a loft.

Are fans and cooling pads sensible, or snake oil? Genuinely useful, with a caveat. A cheap cooling pad or small fan blowing across the case does increase heat transfer from a passively-cooled device, and they cost around £10–£20. But it is second best. Moving the device out of the cupboard, off the sofa, out of the sun or off the top of the television is free, permanent, and does more. Fix the placement first, and use a fan only where the cabinet genuinely cannot be relocated.

Snake oil: “signal-boosting” heatsinks and stick-on cooling accessories sold on the basis that they improve throughput. Cooling restores performance that heat took away. It does not add performance the device never had.

The UK-specific one that costs people permanently: on copper and FTTC lines, avoid repeatedly power-cycling the router. Repeated resyncs are read as line instability and can leave you on a permanently lower speed. You can genuinely slow your own line for good by rebooting it every hot afternoon.

For businesses: PoE delivers up to 60 W or 90 W per port on modern standards, and the losses show up as heat in the copper. Industry guidance recommends keeping the temperature rise from power delivery to 15 °C above ambient, leaving cables unbundled where possible, and limiting bundles where not. Do not coil surplus PoE cable — cut it to length. And remember a switch with a 150 W power budget in a sealed cupboard is a 150 W heater.

One thing that is not happening: heat does not meaningfully weaken the radio signal over household distances. The failure is in the electronics, not the propagation. Worth knowing before you spend an afternoon on the phone to your provider.