Plenty of throughput, short spans, and the rain rules.

Millimetre wave works in bands much higher than the ordinary radio link, and that has three consequences that come together. There is a great deal of room in those bands, so a good deal more throughput can be carried. The beam is a pencil and not a cone, so there is almost no interference from anyone: they would have to be exactly on your line. And those waves drop away very quickly along the path, so the spans are short.

The usual format is a small box with the antenna built in, the kind that bolts to a parapet or a post and is powered by the network cable itself, mounted in pairs and facing each other. In security it turns up in one specific case that is more and more frequent: a lot of cameras have to be taken from one building to another, the distance is short, there is a rooftop-to-rooftop view, and between the two there is a street, a railway line or somebody else's plot that makes the civil work impossible.

In these bands heavy rain absorbs the signal. It is not a nuance: it is the factor that rules the design. Drizzle is not noticeable; a hard storm can bring the link down to a fraction of its capacity, or drop it. And the longer the span, the more rain it has inside it and the more it falls.

That is why the design is not done with the distance the equipment «reaches», but the other way round: you start from how many hours a year you accept the link running degraded, you look at what kind of rain falls in that place, and out of that comes how long the span can be. The same pair of units gives a considerably longer span in a dry place than in one with big storms. Anyone who proposes a span to you without that sum is selling you a catalogue distance.

And out of this comes an architecture decision worth taking at the same moment: what happens to the recording on the afternoon that storm falls. The answer we give almost always is that the video should be recorded at the far end as well, so that the link carries what is being watched and the alerts, and the gaps are recovered when the rain passes. With that, the worst storm of the year is an hour without live view and not an hour without recording. Without it, the weather decides what gets kept.

The advantage: because the beam is so narrow, the air is practically clean. There is no fighting with the industrial estate next door or with the link somebody put up in November, which is the plague of the conventional free bands.

The problem is the same fact seen from the other side: aiming is far less tolerant and moving is not an option. Tenths of a degree matter, and that turns the mounting into part of the link: a mast that flexes in the wind, a half-tightened clamp, a parapet that moves with the sun, a post shared with a sign. All of that, which a low band forgives, here turns into a link that comes and goes depending on the day and on the hour.

And the clearance is just as strict. In a low band you talk about leaving room around the line; here, plainly, there can be nothing: not a branch, not a crane working in the middle for two weeks, not a tall lorry if the span passes low. A leaf in the path does not attenuate a little: it cuts. It is worth looking as well at what is going to be built in between, because a new building is not negotiable.

Not all millimetre-wave bands work the same way from the administrative point of view: some are used freely and others are subject to coordination or to some registration of the link, and that depends on the country and on the particular band. It is not an obstacle, it is predictable paperwork, but it has a lead time.

We treat it like any other condition of the project: it is asked about at the start, we check what applies at that location and the lead time goes into the site calendar. Discovering it at the end is what turns a one-week installation into a month of waiting with the material already bought.

Each end has its browser console —link status, signal quality, throughput passing and a log of how it has gone—, alignment tools in the unit itself, which is what is used on the roof, and central management to see several links.

What really has to be left set up is the watching of the signal over time. A millimetre-wave link does not degrade little by little: either it works or it does not. But the margin does drop —because a mounting gives, because something grows, because the radome gets dirty—, and the margin that drops today is the storm that drops your link in March. With the history saved that can be seen coming; without it, it is discovered on the day of the storm.

And one piece of the installation that looks minor: the power goes up and the data comes down the same network cable, and that cable goes to a roof. The sealing of the connector, the cable support and the surge protection where it enters the building are not accessories: they are half the faults on these links.

To join two buildings with a rooftop-to-rooftop view when a street, a railway line, a canal or a plot that is not yours has to be crossed, and when what has to pass is the traffic of many cameras and not that of an office. As a replacement for a leased line between two nearby sites. As a temporary link while the trench is being dug, which is a very rewarding use: it goes up in a day and the material is recovered afterwards. And as a second route for a fibre.

And in a city case that comes up more and more: several cameras in a square, a market or a car park that have to be brought to the council's control centre, with roofs in between and no permission to dig up the street.

On long spans: this technology is for short distances and everything you stretch is paid for in availability on the day it rains. And where there is no total and permanent clearance there is no gentle version of this: if there is vegetation, if there is a crane, if there is going to be building work, it is not the place. Nor where the only possible mounting is a mast that moves.

In places with very heavy rain and a span that was already at the limit. There the link exists eleven months of the year, and the missing month is the one somebody is going to ask about the recording.

As the only path for the video with no recording at the far end. It is the design mistake we have seen most often with this technology: the link is magnificent and the storm takes away an hour of recording that is not recovered.

And where there is a viable fibre route. If fibre can be laid, fibre gets laid: rain does not affect it, wind does not affect it, it does not have to be aligned and it lasts decades. Radio comes in when the fibre cannot, not when the fibre is dearer this quarter.

The day to day is handled by any IT team: looking at the status, seeing the throughput, checking the signal history, updating, restarting. There is not much to configure on a link like this once it is aimed, and that is an advantage. The administrator account is handed over in the customer's name, the configuration saved and the commissioning report with the signal and the margin measured, which is the document everything will be compared against for years.

What does not stay in the house is the roof. Aligning one of these links is instrument work and patience, and a badly aimed link looks correct: it locks on, it gives throughput and it goes down with the first storm. The same with the design of the span: lengthening it, moving an antenna or changing band is not a console adjustment, it is doing the rain sum again.

And if your IT team already has its radio brand, we adapt without arguing. In millimetre wave the brand decides little: what decides is the length of the span, the rain at that place, the clearance and the rigidity of the mounting. What we do not sign, with any brand, is a span without that sum done, because the one who pays for it is the one left without recording in March.