Detecting inside a place where a detector could not survive.

The principle is the same as any aspirating system: a fan draws air from the enclosure through a network of pipes with calculated holes, and takes it to a very sensitive detection chamber that gives the alert long before there is a flame. The first signal arrives when something is heating up, not when it is burning.

The difference is in the family of equipment. This firm also makes detection panels, so the aspirating unit and the head end that watches it are designed to work together. It is not a marketing detail: an aspirating detector has far more information than a point detector —smoke levels, air flows, aspirator status, filter dirt— and when the head end is from the same family, all of that arrives whole where it is looked at. Hung off another manufacturer's panel by a volt-free contact, what arrives is «alarm» and «fault», and half of what has been paid for is lost.

It is Swiss electronics and you come across it where the conditions are hard and the installation has to last: tunnels, transport, infrastructure, industry, technical rooms, heritage buildings. Places where changing the system in ten years' time is not a realistic option.

A point detector has to be inside the place it watches. That means it has to stand whatever is in there: cold-store cold, humidity, condensation, heat, corrosive vapours, vibration, pressure washing, dust. And it means somebody has to go in to maintain it, which in some enclosures is a work permit, a shutdown or a set of protective equipment.

In an aspirating system, inside the enclosure there is only plastic pipe and holes. The electronics, the aspirator and the filter are outside, somewhere ordinary, reachable and at a reasonable temperature. That changes three things: it allows detection in enclosures where a detector would not last, it allows the equipment to be maintained without going in, and it lengthens the life of the system a great deal because what ages is in a corridor and not inside a chamber at twenty degrees below zero.

The typical cases follow on their own: cold stores and blast freezer tunnels, drying chambers, enclosures with aggressive vapours, silos and dusty enclosures, ventilation ducts, closed electrical cabinets, and enclosures where you have to draw air from inside a machine. In all of them the question is not which detector copes: it is how to detect without putting anything in that has to cope.

Aspirator, filter, detection chamber and electronics on one side; a pipe network with holes on the other. The network is not improvised: branch lengths, diameters, and the number and size of every hole are calculated with a tool, because whether all the sampling points draw a similar amount of air depends on that. With the holes done by eye, the first ones draw almost everything and the last ones almost nothing, and that is not visible at handover: it becomes visible the day the outbreak starts at the end of the branch.

In difficult enclosures there are also two things worth allowing for from the design stage. One is the return air: if you are drawing from a cold store or from an enclosure under negative pressure, you have to think about where the air goes after it has been analysed. The other is condensation and ice: drawing cold air towards a warm place creates water inside the pipe, and that is solved with routes, falls and drains that have been thought about, or it is not solved and the system gives air flow faults every winter.

The building work is the big line and the one budgeted worst: running pipework to where the risk is, fixing it, sealing the penetrations between compartments, and doing it in enclosures that often cannot be stopped. And if you also have to go into a chamber or a duct, there are permits and working windows that set the timescale more than the installation does.

And the maintenance is not that of a detector: filters that get changed, an aspirator that is a mechanical part with a limited life, air flows that get checked, and the smoke test with a stopwatch repeated to confirm that the network is still the one that was calculated. A network that has been in a ceiling for years has changed, because somebody has done building work. In cold or corrosive enclosures there is also an inspection of the inside of the pipework that does not exist in an ordinary installation.

It is not for an office or a corridor: there the works and the maintenance are not justified and a point detector does the job. It is not for an enclosure with no air movement where the pipe cannot be taken to where the risk is, because then you are paying for sensitivity to sample air that does not pass the place that matters. And in enclosures with permanent, very dense airborne dust you have to look at whether heat detection is the right answer, even if it detects later.

We say no when the client wants the equipment and does not want the maintenance schedule, and when there is nobody who is going to deal with the first alert. A system that gives the alert very early is only useful if somebody goes in and looks; if not, the alerts are ignored first and the thresholds are raised afterwards, and then what has been bought is a very expensive detector that detects late.

Not the water. Sprinklers, dry risers, hose reels, pressure sets, tanks and their pipework are another trade, with another licence and another signature. In tunnels and in industry that matters especially, because it is very common for the detection to have to give the order to a water installation put in by another company: the order and its supervision are ours, the pipework is not, and somebody has to answer for the point where they meet.

On gas suppression, the usual: detection, release logic, prior warnings, interlocks and the electrical order, yes. Cylinders, pipework and nozzles, no.

What the client has to learn is not how to work the equipment: it is how to respond to the first alert. Somebody goes, looks and searches. In a cold or closed enclosure that can mean a work permit, and that is why the procedure has to be written beforehand and rehearsed once, because improvising a permit at seven in the morning does not go well.

The thresholds are not touched from inside. It is the eternal temptation and it is what kills these systems: nudging the sensitivity down every time it is a nuisance until it no longer detects early. If an alert is a nuisance and is not an outbreak, you analyse why; you do not raise the threshold.