Short answer: the skylight robot rarely fails because of its brush. It fails because the roof has no sensible way to deliver power and water to it. In 2026 the workable field setup is a 230V supply taken from a dedicated roof socket or a small inverter generator, a hose run kept under about 60 metres with a pump that holds at least 2 bar at the machine, and a water tank you can lift. Everything beyond that — long extension leads, thin hoses, borrowed building sockets — is where deployments stall.
Where should skylight robot power come from on a roof?
Three sources cover almost every job: a permanent roof socket installed for facade access, a portable generator, or a battery pack built into the machine.
A dedicated roof socket is the best answer where the building has one, and most high-rise buildings with facade access equipment already do. Ask the building engineer what the socket is rated for and whether it is on a circuit that trips when the lifts start. A socket sharing a circuit with lift machinery will drop the robot mid-pass.
Battery machines suit short cycles. A machine carrying four to six hours of runtime will clean a small atrium or a run of ten panels without any cable at all, which removes the biggest tripping and tangling hazard on a glass roof. The trade is weight and cycle length — batteries do not like being stored half-charged in a plant room over winter.
Generators are the fallback and they carry two problems on roofs: exhaust and noise. Never run a petrol generator in an enclosed plant room or next to an air intake. On an open roof, place it downwind of the work area and out of the drip line. Our notes on high-level facade access planning cover how access routes and work zones are normally set before equipment lands on the roof.
How long can extension leads run on a roof?
Shorter than most people assume. Voltage drop, not cable length in the catalogue, decides this. A 1.5mm² lead at 50 metres is fine for a machine drawing 6A; the same lead at 80 metres with the pump running is not. The machine will not fail outright — it will run slow, the pump will lose pressure, and the brush will stop loading properly, which reads as a cleaning failure rather than a power problem.
Practical limits worth carrying into a survey:
- Keep total lead length under 40–50 metres on 1.5mm² cable, or move to 2.5mm² beyond that.
- Use one continuous lead rather than three joined ones. Every joint is a water ingress point on a wet roof.
- Run the cable on the dry side of the roof, clipped or weighted, never across a walkway or the drip line under the glass.
- Fit an RCD at the source, not at the machine. On a wet roof the protection has to be upstream.
If the roof geometry forces a long run, the honest answer is to move the water and power to a fixed point mid-roof and work outward from there in both directions. That costs a day of planning and saves a season of voltage-drop complaints.
What pump pressure does a skylight cleaning robot need at height?
The number that matters is pressure at the machine, not at the tap. Every metre of vertical lift costs roughly 0.1 bar, and every ten metres of hose costs friction losses on top of that. A ground-level pump delivering 4 bar through 80 metres of hose and 20 metres of lift may deliver barely 1.5 bar at the brush, which is not enough to wet and flush a dirty panel.
Working figures from field setups: 2 bar minimum at the machine for a wet-wash pass on normal dust load, 3 bar for heavy construction dust or pollen season, and a flow rate around 4–6 litres per minute so the brush stays wet across a full panel pass. Below that the machine smears rather than cleans, and smearing on backlit glass is worse than leaving it dirty.
The tank decision follows from the flow rate. At 5 litres per minute, a 200-litre tank gives forty minutes of continuous work — enough for one decent roof session, not a full day. That means either a refill plan with a second tank, a fixed water point on the roof, or accepting that the machine works in short blocks.
Filtration is the part everyone skips. Take unfiltered tap water onto a hot roof and the machine sprays mineral-loaded water onto glass that dries in seconds. A simple inline filter and, on hard water, a small DI cartridge solves it for a few hundred in parts. Our page on crossing frames, mullions and edges is worth reading alongside this, because the hose run and the machine path have to be planned together.
What does a realistic day-one roof setup look like?
Plan the site before the machine arrives: a power point with RCD protection on the dry side, a water source with a filter, a 20–30 metre hose to a mid-roof staging point, a machine tether or safety line independent of the hose, and a defined drip zone below the working area.
Then test it empty. Run the pump for five minutes with the machine off and check pressure at the brush head, not at the pump. Walk the full cable and hose route once carrying nothing. Most first-day problems are route problems, and they are cheap to find on foot and expensive to find mid-pass.
Budget the consumables too: filters, hose connectors, cable clips and RCD test time. Owners who track them find power and water plumbing, not brushes, is the recurring line. Our field reviews of skylight robot setups show the same pattern across buildings — the installations that work are the ones where the roof was prepared before the machine arrived.
Key Takeaways
- Power and water plumbing, not the brush, decides whether a skylight robot deployment survives its first month.
- Prefer a dedicated RCD-protected roof socket; batteries suit short cycles, generators are a fallback with noise and exhaust constraints.
- Keep extension leads under 40–50 metres on 1.5mm² cable, single continuous run, dry side of the roof.
- Measure pressure at the machine: 2 bar minimum, 3 bar under heavy dust, with 4–6 litres per minute of flow.
- Plan a mid-roof hose and water staging point and filter the water — this beats any upgrade to the machine itself.