Why Your 4-20 mA Signal Drifts, and the Cable Is Usually to Blame

A pressure transmitter on a fertiliser plant near Vadodara started reading 0.4 bar high every afternoon. Only in the afternoon. The instrument was swapped, calibrated, swapped again. Somebody suspected the DCS card. Three weeks and a good deal of shouting later, a young engineer walked the cable route and found the answer: for about sixty metres the signal cable shared a tray with the feeder to a bank of cooling fans that came on when the ambient temperature climbed. The fans ran in the afternoon. The reading drifted in the afternoon.

Nobody had bought a bad transmitter. They’d bought a bad cable route, and then an unscreened cable to put in it.

This is the most common failure in plant instrumentation and it almost never gets diagnosed as a cable problem, because cable is the last thing anyone suspects. It’s the cheapest item on the loop and it has no moving parts. Yet a 4-20 mA loop is a low-energy analogue signal travelling a long way through an electrically filthy environment, past motors, contactors, variable frequency drives and switchgear that produce exactly the kind of interference that lands on a signal conductor.

Understanding why it happens makes the fix obvious. There are two ways noise gets into a signal cable, and they need two different defences.

The first is magnetic. Current in a nearby power conductor creates a magnetic field, and an alternating current creates an alternating field, which induces voltage into any conductor sitting inside it. A screen does very little about this. Twisting does. In a twisted pair the two conductors keep swapping position along the length of the cable, so each of them spends roughly the same time nearer the noise source and further from it. The induced voltage ends up almost equal on both, and a differential receiver rejects it. That’s the entire reason instrumentation cable is built as pairs and triads rather than as loose cores, and it’s why pulling two separate single cores through a conduit to make a “pair” is a waste of everybody’s time.

The second is electrostatic, and it comes from voltage rather than current. Here the screen is what earns its money. It intercepts the coupling and takes it to earth before it reaches the conductor. Which is where the second big site mistake lives: the screen is earthed at both ends. Do that and you’ve made a loop between two earth points that are almost never at exactly the same potential, so current flows through the screen and injects noise instead of removing it. Earth it at one end, at the control room, and leave the field end insulated. A drain wire exists precisely so you have something you can clamp under a terminal, and the reason it’s tinned is to slow the galvanic corrosion that would otherwise develop between copper and the aluminium foil it touches.

Then there’s the question of whether one overall screen is enough or whether every pair needs its own. Cost pressure pushes people towards the collective screen, written CAM, and on a cable carrying a single type of signal that’s a reasonable call. It stops being reasonable the moment one multipair carries half a dozen analogue loops from different instruments with different earth references. Now the pairs interfere with each other, and an overall screen can’t help, because it’s outside all of them. That’s what individual screening plus a collective screen is for, written IAM/CAM or ICAT. APAR’s screened and instrumentation cables are made in both, and frankly, on mixed analogue work the extra cost of individual screens is trivial next to three weeks of hunting a phantom transmitter fault.

Which brings us to the specification itself, and to a piece of trade knowledge that saves arguments. BS 5308 was superseded by the harmonised BS EN 50288-7 back in 2005. It’s still quoted on purchase orders every week, because the harmonised standard doesn’t cover the higher voltage ratings and cable dimensions that process plants actually need, and a publicly available version turned up in 2009 as PAS 5308. So “BS 5308” on a document tells you less than it looks like it does. Part 1 is polyethylene insulated, Part 2 is PVC insulated, and each part splits into Type 1 unarmoured and Type 2 with single wire armour over a bedding layer. Four different cables. Pin down which one before anybody prices it. The instrumentation cable selection guide lays out the combinations properly.

The other habit worth breaking is borrowing a spare core from the control cable when a signal needs to get somewhere. It looks free. It isn’t. Control cable isn’t twisted into pairs and isn’t screened, and putting a low-level analogue signal in among switching contacts and indication circuits is asking for the same drift, just from a different direction. Keep power, control and signal in separate cables. APAR’s LV XLPE/PVC control cables run up to 61 cores, so there’s very little excuse for the two functions to share a sheath, and the core-count guide makes the case for buying generously in the first place.

Route discipline matters just as much as cable choice, and it’s cheaper. Signal cables belong on their own tray, and where a signal route has to meet a power route it should cross at ninety degrees rather than run alongside. Coupling builds up along a parallel run; a crossing barely couples at all. Two metres side by side is usually survivable. Sixty metres next to a fan feeder, as our Vadodara friends discovered, is not. The same principles apply in a residential building, where the victim is a camera picture rather than a transmitter, and we’ve written them up in a separation guide.

None of this is exotic engineering. It’s twisting, screening, single-point earthing and keeping cables apart, and all four decisions get made at the design stage for very little money. The alternative is what usually happens instead: an instrument that reads correctly on the bench, drifts in service, gets replaced twice, and eventually teaches somebody an expensive lesson about a cable tray.

If your loop drifts on a schedule, walk the route before you touch the instrument. Something nearby is switching on the same schedule, and it’s talking to your signal.