A skylight cleaning machine should be judged on pad geometry, gasket material, edge sensing count, tether rating, water handling, hatch-compatible weight and spare part availability. Suction figures rank eighth. Everything above that list is marketing.
Skylights are not vertical windows lying down. They collect water, grit and debris differently, and they are usually accessed through a hatch rather than a balcony.
Why does pad geometry beat suction force?
A circular pad cleans a round footprint. When the machine steps sideways by a square-based distance, the corners of each pass never overlap, and you get faint arcs in raking light. A square or rounded-square pad with 40-80 mm overlap solves it.
More suction does not fix a geometry gap. It just holds the machine harder against the streaks it is leaving.
Which gasket material works on hot glass?
| Material | Heat range | Life in roof use | Note |
|---|---|---|---|
| Natural rubber | -10 to 60 °C | 6-12 months | cheap, hardens fast |
| Silicone | -40 to 180 °C | 18-30 months | better, more costly |
| EDPM / EPDM blend | -30 to 120 °C | 12-24 months | good UV tolerance |
| TPE | -20 to 80 °C | 9-15 months | soft, easy to fit |
On a roof in southern China where surface temperature reaches 65 °C in July, natural rubber gaskets harden inside six months. Silicone is worth the premium there. In a northern European atrium, natural rubber is adequate and the cost difference is not justified.
How many edge sensors do you need?
Four. Two is what cheap units ship with, and two means the machine is confident on three sides and uncertain on the fourth. On a rectangular skylight with a drainage lip on the low side, that uncertainty becomes a stall or, worse, a fall.
Look for IR pairs at each corner plus a mechanical bumper. The bumper is the fallback for corners IR misses, and it costs almost nothing to include, which tells you something about the units that omit it.
What else separates a roof machine from a window machine?
- Water tolerance. Roof panels have standing water; window units are not built for a 6 mm film at the low edge.
- Hatch-compatible weight. 12 kg and 400 mm wide gets through a standard 700 mm hatch. 26 kg does not.
- Tether anchor. Rated, with a documented inspection interval, not a nylon strap.
- Dust pre-clearance. Roof glass collects grit that grinds a gasket. A dry pre-pass or brush mode matters.
That water point is the one people forget. A machine designed for vertical glass expects water to run off. On a 10° skylight, water sits, the pad hydroplanes, and tracking degrades within a single pass.
What does the machine cost to own in 2026?
Take a mid-band skylight machine at USD 3,200. Consumables over five years: 10 gaskets at USD 70, 5 pad sets at USD 120, 3 drive belts, a pump service. Roughly USD 2,300 across the period. Add anchor and water treatment in year one.
Total cost of ownership lands near USD 6,000 over five years, or about USD 1,200 a year. Compare that with a rope crew invoice on a 600 m² roof at USD 3-6 per m² per clean.
For a run-down of units at this band and above, the Lingyun Y3 review covers obstacle-crossing behaviour, and the full skylight robot review archive has the side-by-side notes.
Who should not buy a skylight cleaning machine?
Anyone with two small roof windows and a safe ladder. The machine will not pay back and will not reach the awkward corner near the ridge.
Anyone whose roof is above 30° pitch and has no tether anchor. The load case changes and the machine that works on a 10° roof is not the same machine.
Anyone who cannot service it locally. A machine with a two-month parts lead time is a liability on a building where cleaning is a compliance item.
Key Takeaways
- Pad geometry and overlap prevent streaks more than suction force.
- Silicone gaskets last about twice as long as natural rubber on hot roof glass.
- Four edge sensors plus a bumper is the minimum sensible configuration.
- Five-year cost of ownership on a USD 3,200 machine lands near USD 6,000.
- Above 30° pitch, check the load case before buying anything.