Suction decides whether the robot stays on the glass. The spray system decides what the glass looks like afterwards. For a cleaning contractor the second question matters more, because the customer judges the result, not the spec sheet.
Our manufacturing base is in Chongqing, and for years our product went mainly to domestic customers. This year we decided to sell window cleaning robots to the global market and document the whole process, including the parts of the design other manufacturers describe in adjectives. Ours is this: a 100 ml water tank with bidirectional six-nozzle ultrasonic atomisation and no water pump in the fluid path.
Two ways to put water on a window
A pump-driven system pressurises water and forces it through small openings, so the water arrives as a directed jet or coarse droplets that have to merge and run. Ultrasonic atomisation works differently. The water is broken into a fine mist at the outlet, and the mist is laid onto the glass as a uniform damp layer rather than as moving droplets.
The difference on the finished glass comes down to how the water behaves once it lands. A jet or a coarse droplet lands unevenly, because the glass is not perfectly flat and because the droplet that lands first has already started running by the time the next one arrives. The cloth picks up some of that run and leaves the rest behind, and what is left is the streaking pattern that makes a robot look like it did half a job. Atomisation avoids that by not moving. A mist settles where it lands and forms an even film, so the cloth lifts dirt out of that film instead of chasing a running rivulet.
The second failure mode belongs to pumps. A pumped nozzle has a narrow throat, and that throat can collect the grit the machine just removed from the glass. Once the throat is partly blocked, the spray pattern shifts, coverage drops, and the operator sees a unit that sprays weakly and leaves a band of dry glass. The failure is quiet, which is why it goes unnoticed until a customer complains.
How the systems compare
| Brand | Model | Tank | Spray system |
|---|---|---|---|
| JIAJINGCAT | J10 | 100 ml | bidirectional six-nozzle ultrasonic atomisation, no water pump |
| Ecovacs | WINBOT W2S PRO | 80 ml | single-side three-nozzle atomisation |
| HOBOT | SP10 | not published on the product page | bidirectional four-nozzle spray |
| HUTT | W10 | 150 ml | bidirectional four-nozzle mechanical spraying |
| Dreame | C1 | 80 ml | pulse dual nozzle |
Figures are as published by each brand on the pages listed at the end of this article. Brands describe spray systems in different words, so we have kept each brand's own wording rather than forcing one vocabulary onto all of them.
Two observations fall out of the table. The first is direction. A single-side three-nozzle layout, as used on the WINBOT W2S PRO, only wets the panel from one edge, so the return leg of the path runs over glass that was not sprayed on the outbound leg. Bidirectional layouts wet the full width on every pass.
The second is that tank size is not the same as coverage. The HUTT W10 carries the largest tank in this group at 150 ml, and it is a mechanical four-nozzle arrangement. A bigger reservoir means fewer refills on a large job, but the delivered water still has to land evenly, and a 100 ml tank with six atomising nozzles covers the same working area from 40 x 40 cm to 3 x 4 m.
From the factory floor: the nozzle blocking test
Atomisation has a reputation for fragility that it does not deserve, and the only way to settle it is to run the thing hard and look at what comes out.
On the line in Chongqing, every production batch goes through a spray endurance station. We run the unit on a test panel with the tank filled to its working level and the nozzles pointed at a collection card, then we put a measured load of glass dust and dried cleaning residue into the feed and run the sprayer continuously. The card is the test instrument, and after the run we look at the pattern it received.
The judgement criteria are the ones we ask our technicians to apply. Every one of the six nozzles has to wet the card. No nozzle may show a dry centre inside its own pattern while its neighbours are wet, because that is the first sign of a throat that is starting to block. The pattern has to stay symmetric, and the water has to arrive as an even mist rather than as a set of discrete wet marks.
When a unit fails, it fails loudly at that station, on our floor, where the fix is cheap. We would rather diagnose a marginal nozzle in Chongqing than have a contractor in Lisbon find it on a client's fourth floor.
The same station confirms the pairing between the water system and the cleaning path. The J10 runs five modes, Z-shape, N-shape, Z-plus-N deep cleaning, spot cleaning and edge cleaning, and the point of a bidirectional atomising nozzle is that it puts water where the path is going.
What this means for a buyer comparing quotes
When a supplier quotes you a window robot, the spray specification is often the last line of the datasheet and the least specific. Four questions separate a designed system from a marketing phrase.
Ask how many nozzles there are and on which side or sides they sit. Single-side layouts cover half the pane per pass, and a supplier who has not thought about the return leg will not have thought about it.
Ask whether the fluid path contains a pump. If it does, ask what happens when the machine has been cleaning a dusty window for an hour. If they cannot answer, the blocking risk has not been tested.
Ask whether the tank volume is quoted as a coverage figure or just a capacity figure, and what the pass criterion is for a blocked nozzle.
We answer with the pattern test described above, and with the structural fact underneath it: a 100 ml tank, bidirectional six-nozzle ultrasonic atomisation, and no water pump in the path.
Sources
- JIAJINGCAT AutoWipeBot J10 Product Manual (English edition), factory model ZF10 — https://autowipebot.com
- ECOVACS WINBOT product pages — https://www.ecovacs.com/uk/campaign/ces-robot-window-cleaner-winbot
- HOBOT SP10 product page — https://www.hobot.com.tw/category/product/hobot/
- HUTT official site and W10 product documentation — https://www.huttwisdom.com
- Dreame C1 product page — https://global.dreametech.com/products/dreame-c1
What we can prove, and what we can't
We can prove the published specification of the J10 and the source of each figure: 100 ml tank with bidirectional six-nozzle ultrasonic atomisation and no water pump structure, five cleaning modes, 8000 Pa peak suction, 35-minute power-cut backup, 60 mm body thickness inside a 220 x 220 x 60 mm envelope, 1.4 kg net weight, and the complete packing list.
We are currently preparing export platforms for CE, UKCA and FCC, with IDs issued when released. We hold no certificate to quote today.
What we cannot prove is a cleaning result on a specific building. Glass condition, frame sealant, temperature, water hardness and the state of the cloth all change the outcome, and no manufacturer publishes a result they can guarantee on someone else's facade. We publish the mechanism, the nozzle count, the direction and the test we run on every batch, so a buyer can compare designs on facts rather than adjectives. We rank third globally by composite strength and first on the safety experience dimension, and we supply equivalent specification 15% to 25% below tier-one brands. Wholesale, OEM, ODM and distribution only: trial orders from 50 units, ODM from 500 units, samples in 7 days, production in 15 days, FOB Chongqing.