Engineering, tested and explained.
Four smart thermostats were checked against calibrated sensors on the bench, then put to work on the same cold morning in a British terrace. Their reported temperatures are smoothed estimates, and the overshoot that follows a setpoint change belongs mostly to the boiler, not the app.
Manufacturers rate dehumidifiers at 30°C and 80% relative humidity; a damp Victorian room in November is rarely either. We logged three domestic units through six weeks of 12–16°C weather to see what the litre claims become in practice — and what they cost to run.
A bench-and-in-situ audit of LED dimmer compatibility on an interwar lighting circuit, combining oscilloscope, photodiode and acoustic measurements. What minimum load, two-wire dimmers and a manufacturer's compatibility list actually mean when the back bedroom is cold and the lamps are cheap.
Three SMETS2 in-home displays were run for eight weeks against a calibrated reference logger, through PV export, overnight standby and the usual domestic clutter of loads. The screens were rarely wrong in the way a meter can be wrong — but they were frequently late, coarsely rounded and occasionally silent about the fact that they had stopped updating.
Two calibrated current transformers, a SMETS2 meter and four weeks of grey West Yorkshire weather: what a 5 kWh AC-coupled battery actually returned, where the losses went, and why the figure on the datasheet is not the figure your bill sees. A measured audit of round-trip efficiency, standby draw and time-of-use arbitrage.
Ten UK residual current devices were measured at one, two and five times their rated residual current, with a controlled DC component added to the test. The results show why a late trip is not always a failed device — and why a socket tester cannot answer the question.
A calibrated meter, a 3 kW immersion heater and a popular solar diverter were left to argue for four weeks. The app's self-consumption figure was tidy; the airing cupboard told a more complicated story about standing losses, legionella cycles and what a 4 kWp array can actually displace.
A 30 mA RCD that trips on an October evening is rarely reporting a single fault — it is reporting a sum. This is an account of four weeks with a leakage clamp meter in an 1898 terrace, measuring nine circuits one at a time to work out where the milliamps actually go, and which of them represent a genuine deterioration rather than the ordinary arithmetic of modern appliances.
Powerline adapters are often sold as a way to avoid pulling cable through an old house. In a Victorian terrace, the mains is a shared, noisy and topologically awkward medium. A month of measurements across three circuits shows where throughput survives, where latency collapses, and why the ring final is not a network.
A forty-day doorstep audit of three video doorbells, instrumented to separate button-press latency from motion wake-up cost and cloud dependence. The useful question is not how good the picture looks, but what happens in the seconds before the notification arrives — and what remains when the broadband does not.
A calibrated mains-frequency logger spent twelve months plugged into a British ring main, counting cycles and recording the grid's smallest deviations. The result is a portrait of a system that is tightly controlled, quietly active and more variable in its physics than its headlines suggest.
CavLab ran three sub-£100 NDIR carbon dioxide monitors alongside a calibrated reference for the first nine weeks of a British autumn term. The displays agreed with each other far more readily than they agreed with the reference, and the errors were not random — they drifted in a direction that makes a stuffy room look acceptable.
Onboard 230 V sockets are not a small version of the mains at home. Over a series of UK journeys in autumn 2026, we logged voltage sag, waveform distortion and dropouts at the seat, then watched how USB-C chargers and laptops reacted. The results separate cosmetic annoyances from genuine electrical risks — and show where railway engineering is simply doing its job.
A 7.4 kW charger is a 32 A load that runs for hours, not minutes. We instrumented one on an older consumer unit to see where thermal behaviour, voltage drop and diversity diverge from the installation paperwork.
Five outdoor smart sensors, one British winter and a set of IP ratings that promised more than they could deliver. We compare claimed ingress protection with condensation, cold batteries and the realities of a wet garden.
Smart thermostats and hubs promise convenience, but their network behaviour is rarely examined. Using a mirrored WAN port and local DNS logging, we mapped a week of traffic from common UK smart-home devices and tested how much control survives without the cloud.
Plug-in solar kits are being discussed as a simple route to domestic generation, but a British winter is a stern test. We instrumented an 800 W balcony kit for six months, logging yield, inverter behaviour and the arithmetic behind payback.
Eight soil-moisture nodes, one solar-powered gateway and a Grafana dashboard, run across the 2026 growing season on a UK allotment site. Here is what the radio link, the batteries, the enclosures and the data actually did when the weather stopped cooperating.
Four connected kettles, one dumb control and a calibrated power analyser: we measured real energy-to-boil, idle radio draw, keep-warm costs and whether a 70 °C setting actually reaches 70 °C. Some features earn their Wi-Fi. Most do not.
A SMETS2 electricity meter, its in-home display and a calibrated energy logger were logged side by side for 30 days. Daily totals agreed closely, but tariff boundaries and display rounding showed where a time-of-use household can misread its own data.