Cabling: a hundred metres, and not all of them are yours

The article on networks sets the chain rule and sends the first of the two deciding figures down here. Here it is, and it constrains more than it looks.

Cabling is the only purchase in this pillar with a fifteen-year life. A machine gets replaced, a switch gets replaced, a line changes operator. A cable run through a buried conduit stays there until somebody opens a wall.

That shifts the question. Everywhere else in the pillar you choose a product; here you choose a route, and the route is decided before the walls close. Afterwards it gets paid for with a breaker.

This article covers where the cabinet goes, the hundred metres and how they divide, the three ways a cable dies, and the difference between measuring a link and checking that it works.

The international generic cabling standard sets the maximum length of a copper channel at a hundred metres. It is the figure everybody remembers, and it is remembered wrongly, because those hundred metres are not available to the fixed part.

The intended split is sharp: ninety metres at most for the permanent link — the cable running in the wall, from the wall socket to the patch panel — and ten metres in total for the patch leads, the one at the desk and the one in the rack.

That separation is what makes the figure useful. The ninety metres are laid once and never change. The ten get spent after you have gone, by people who do not know the constraint: a five-metre lead replaced by a ten-metre one because it was lying in a drawer, and the channel is out of specification.

The practical consequence is a design rule rather than a compliance rule: size the fixed link at seventy or eighty metres rather than ninety, and keep the margin for the leads other people will choose. A link at the exact limit is a link that will fail on the day somebody does something perfectly normal.

It is also what decides the cabinet location from the previous section. If the furthest machine is ninety-five metres of real cable away — not as the crow flies, but following the conduits, the drops and the rises — the location is wrong, and that is known before a single wall is drilled.

The most important difference between an installation that ages well and one that will have to be redone comes down to one thing: can another cable be pulled without breaking a wall?

Three answers exist and they get decided on site. Surface trunking, the least elegant and the most honest: everything is visible, everything can be changed, and it ages well in a plant room or a service corridor. A suspended ceiling or a raised floor, which is the best answer when the building has one.

And buried conduit, which is neat and acceptable on one non-negotiable condition: that it is oversized and that a draw string stays in it. A conduit filled to capacity is a dead conduit, because the next cable will not go through.

What is never acceptable and what we see regularly: cable buried directly in concrete or plaster with no conduit. It is not replaceable, it does not repair, and on the day it fails the only answer is surface trunking laid over the top — which is to say the solution refused fifteen years earlier on aesthetic grounds.

The question to put to the builder or the electrician, before the works, fits in a sentence: where do the network conduits run, what diameter, and will there be a draw string. Those three answers are worth more than the cable choice.

It is an electrical installation rule before it is a network rule, and it is broken constantly for an understandable reason: the electrical conduit already exists and slipping a network cable into it is tempting.

The technical problem is interference. A power cable running alongside a network cable for several metres induces noise in it, and that noise turns into transmission errors — hence slowness and intermittent losses that nothing in the network’s behaviour explains.

The practical problem is more serious and it is not about computing: mixing circuits at different voltages in one conduit is bad electrical practice, and it becomes a genuine risk on the day somebody works on what they believe is a low-voltage cable.

The rule serious installers apply is physical separation: distinct conduits, and a crossing rather than a parallel run when the two must meet. Where separation is impossible, a screened cable and correct earthing reduce the problem without removing it.

The check you can run yourself is visual: open a trunk and look at what runs in it. If network and power cables are in there together over several metres, you know the likely cause of the slowness nobody can explain.

What we will refuse: installing a fixed link sized to the standard’s limit. The margin exists for the leads other people will choose after us, and a link at ninety metres is one that will fail for a perfectly normal reason.

We will refuse to bury cable without conduit, whatever saving is proposed, and to land wall cables straight into a switch with no patch panel. Those two savings are the only ones on this page that get paid for every year for fifteen years.

We will refuse to strip certification out of a quotation to bring the total down. We will propose other lines to cut, including our own, but not the one that is the only measured proof of what was installed.

What we do: the cabinet point chosen on a plan before any purchase, with the real distance to the furthest machine; the fixed link sized with margin; two sockets per desk and the off-desk sockets counted; oversized conduit with a draw string; the patch panel labelled as the network article requires; and a certification report per socket, handed over with the drawing.

And what you can do this week without us: open a trunk, see whether network cables sit alongside power cables, and look under the desks for the small switches nobody declared. Those two observations explain most of the slowness we get asked to diagnose.