Underground IoT Connectivity: Antennas, RF and the Alternatives

Why underground IoT connectivity is so hard

Every radio link has a budget: how much signal the transmitter puts out, how much the receiver needs, and everything that gets lost in between. Above ground in open air you have plenty of headroom. Underground, every layer between the antenna and the sky takes a slice of it: soil, reinforced concrete, steel, standing water and, in the case of a utility chamber, a lump of cast iron sitting directly on top of the antenna.

Vodafone's own NB-IoT and LTE-M white paper puts a single brick wall at roughly 10 dB of loss. That's a useful mental yardstick, because 10 dB means only a tenth of the signal survives. Two or three of those and an ordinary 4G device is in trouble.

Frequency matters enormously. A published field study of NB-IoT and LTE-M coverage in underground water system manholes measured around 10 dB of loss inside a typical manhole at 800 MHz, against around 25 dB at 1800 MHz. Same hole, same lid, 15 dB difference purely because of the band. The same study found that a layer of snow on the cover added 3.4 dB and wet leaves bunched on top added 5.9 dB at 800 MHz. Your link budget in July is not your link budget in November.

This is why the low-power cellular technologies matter so much here. They were designed with coverage enhancement built in, trading data rate for reach by repeating transmissions until they get through. Vodafone quotes these maximum coupling loss (MCL) figures, which is the total path loss a link can survive:

Vodafone says NB-IoT can get through two to three double brick walls, which is how it claims coverage in underground car parks and basements. That extra reach comes at a price, which we'll get to in the battery section. For a deeper comparison of the two, see our NB-IoT vs LTE-M guide.

Four environments, four different problems

"Underground" covers very different places, and the problem in each is different enough that one answer won't fit them all.

Utility chambers and manholes

Water and gas meters, valve chambers, sewer level sensors, district heating, telecoms joints. Usually battery powered, no mains anywhere nearby, a metal or concrete lid directly overhead and a high chance of flooding. The device has to send small amounts of data reliably for years without a visit. This is NB-IoT and LTE-M country, and the antenna position decides everything.

Basements and plant rooms

Building management systems, boiler and chiller controls, fire panels, lift controllers, energy meters, CCTV. Mains power is usually available, which opens up more options, but you're typically several floors and a lot of reinforced concrete away from the nearest window. The big risk here is the installer who tests the signal at the top of the stairs, where it's fine, and then mounts the router in the far corner of the plant room.

Underground car parks

EV chargers, ANPR cameras, barrier and payment systems, ventilation and CO monitoring, help points. Long, low, open spaces built from heavy concrete, often with good signal near the ramps and almost none at the back of the lowest deck. Kit is mounted at ceiling height where vans and roof boxes hit it, it gets jet-washed, and it lives in exhaust fumes. EV charging is the growth area, and a charger that can't talk to its back office can't take payment.

Data centres

The interesting one, because the building is often designed to keep radio out. Windowless halls, metal racks, metal cladding and sometimes deliberate shielding. Cellular matters here mainly for out-of-band management: a 4G connection to a console server or router that lets engineers reach equipment when the primary network is down. That's exactly the moment you need it to work, and it's often installed in the one place in the building with no signal at all.

Antenna placement and RF behaviour

A good antenna in the wrong place is a bad antenna. Antennas are tuned on the assumption of a particular environment, and putting them close to metal, water or other electronics changes their behaviour. This is detuning: the resonant frequency shifts, the match to the radio gets worse, and energy that should be radiated gets reflected back into the cable or absorbed by whatever is nearby.

The usual culprits are pipework, the meter body itself, battery packs, steel cable trays, rebar in the chamber wall and the underside of the lid. Many small IoT antennas are designed to work with a ground plane of a particular size. Mount one on a metal plate twice that size, or with no ground plane at all, and you've changed the antenna.

In the manhole study above, the best position was the centre of the chamber, half a metre below ground level, clear of the metal cover. That's a useful starting point rather than a rule, but the principle holds everywhere: get the antenna away from metal and as close to the open sky as the environment allows.

A few other things catch people out:

MIMO needs two antennas and two cables. A 4G Cat 4 router or a 5G device expects at least two antenna ports. Connect only one and throughput drops, and on some devices the connection becomes noticeably less stable. For low-data devices this may not matter, but it's a decision, not an accident.

Coax tails are not free. Coiling up surplus cable in the chamber looks tidy and adds loss and unpredictability. Cut it to length or use the right length to start with.

Directional antennas help, and hinder. A directional antenna pointed at the serving mast can recover a lot of gain, but underground you rarely have a clear view of anything, and if the operator re-plans its network your carefully aimed antenna is pointing at nothing.

Test with the lid shut. It sounds obvious. It's the most common mistake in the field. Measure with the cover on, the plant room door closed and, if possible, after rain.

How tough does an underground antenna need to be?

This is where datasheets go quiet and real installations get expensive. An antenna in a chamber, under a lid or on a car park soffit lives a hard life, and the RF performance is irrelevant if it's cracked, flooded or stolen.

Load: it may get driven over

If the antenna is in or through the lid, it has to survive whatever the lid survives. In the UK, access covers are rated under BS EN 124, and the classes tell you what's going to roll over the top:

A through-lid or in-lid antenna fitted to a D400 cover in a road needs to be designed and tested for that duty, flush or near-flush so it doesn't become a trip hazard or get sheared off by a snowplough, and replaceable without digging up the road. A radome that's fine in a footway will not last a winter in a carriageway.

Water: it will get wet

Chambers flood. Assume total submersion and specify IP68 for anything that lives below the lid, with the depth and duration stated, because IP68 means whatever the manufacturer declares it means. Connectors and cable entries are the weak points, so seal them properly and don't leave a coax tail sitting in standing water. Condensation is a slower killer: a sealed enclosure that breathes through temperature cycles will pump moist air in and fill itself with water over a few seasons.

Impact: it will get hit

Impact resistance is rated on the IK scale under IEC 62262. IK08 is 5 joules, IK09 is 10 joules and IK10 is 20 joules. For car park ceilings, ramps and anywhere within reach of the public, IK10 is the sensible minimum. Think about height clearances too: an antenna that hangs below the posted headroom will meet a van roof sooner or later.

Chemistry and atmosphere

Car parks mean exhaust fumes, fuel, oil and de-icing salt brought in on wheel arches. Sewers mean hydrogen sulphide, which is corrosive to copper and brass connectors and is a good reason to look closely at connector plating. Outdoor kit needs UV-stable plastics. And some sewers, wet wells and gas installations contain flammable gas such as methane. In Great Britain the Dangerous Substances and Explosive Atmospheres Regulations 2002 (DSEAR) require a hazardous area classification, and if a chamber is zoned, every piece of electrical equipment in it, antenna and battery included, has to be suitable for that zone. That rules out most off-the-shelf IoT kit immediately.

People and animals

External antennas get stolen, snapped off and painted over. Low-profile, anti-vandal designs with tamper-resistant fixings are worth the money in public places. Below ground, rodents chew cables, so route them in conduit where you can. And if a maintenance contractor can't tell what your antenna is, it will eventually be removed as rubbish.

Getting the cable from where the signal is to where the device is

Often the signal exists somewhere nearby, at street level, by a window or on the roof, and the question is how to get it to the device. The obvious answer is a longer antenna cable. The problem is that coax loses signal with every metre, and loses more as frequency rises. Using LMR-400, a common low-loss cable, as an example:

That's for thick, stiff cable that's awkward to route. Thinner cables such as the RG58 and RG174 tails supplied with most antennas lose far more, and every connector and adaptor adds a little. Thirty metres of the wrong cable can easily throw away everything a good antenna gained. For MIMO, double the cable, the cost and the routing headache.

Then there are the building issues nobody thinks about until the installer is standing there with a drill:

Fire stopping. Running cable between floors or through compartment walls means penetrating fire-rated construction, and every penetration has to be properly fire-stopped to maintain compartmentation. In a managed building this needs sign-off, and in practice it often means using existing risers.

Waterproofing. Basements below the water table are often tanked or built with a waterproof membrane. Drilling through the wall to reach an external antenna can breach it, and a leak caused by your antenna cable is not a conversation you want with the building owner.

Permissions. Landlords, freeholders, listed building consent and data centre operators all have opinions about holes in their walls and things on their roofs. Allow time for it.

The battery cost of a marginal link

Coverage enhancement is not free reach. When an NB-IoT or LTE-M device is at the edge of coverage, it gets that extra 15 to 20 dB by repeating its transmissions, sometimes many times over. Each repetition is radio-on time, and radio-on time is battery.

A device that was sized for ten years in good coverage may last a fraction of that in a chamber where it's working at the limit every time it reports. Worse, a marginal link means failed attempts, retries and re-registrations, each of which costs more energy and can leave gaps in the data. Power Saving Mode and eDRX help by keeping the radio asleep between reports, but they can't fix a bad link.

So the RF decision is also a battery decision. A few decibels gained by moving the antenna can be worth years of battery life and a lot of avoided site visits. Our guide to powering remote IoT covers the energy budgeting side, and LTE-M: three IoT problems it was built to solve goes into the power saving features.

Alternatives when radio won't reach

Sometimes the honest answer is that no antenna will make it work. That's not a failure, it's a design decision, and there are plenty of options.

Outdoor router, Ethernet back in

Rather than bringing the radio signal down to the device, take the radio up to the signal. Mount an outdoor-rated cellular router where the coverage is good, on the roof, an external wall or a post at the top of the ramp, and run Ethernet back to the equipment. With Power over Ethernet (PoE), one cable carries both power and data, so there's no mains needed at the router.

This swaps a lossy coax run for a cable that doesn't care about RF. Standard copper Ethernet runs up to 100 metres; beyond that, use fibre or a PoE extender. For a basement plant room or the lowest deck of a car park, this is very often the most reliable answer. Remember the outdoor end needs surge protection and proper earthing, because you've just connected an outdoor box on a roof to equipment deep inside the building, and that box is now part of your network's security boundary. Our cellular router security guide covers locking it down.

Mobile repeaters

A repeater picks up the outdoor signal and rebroadcasts it inside. In the UK these are regulated by Ofcom, and only certain types are licence-exempt. Ofcom's rules allow static indoor repeaters that amplify a single operator's signals, with gain limits and other technical conditions, and from 29 February 2024 those repeaters can be 4G-only rather than also having to carry 2G or 3G. Anything outside the rules is illegal to use and can interfere with the network, so stick to compliant, operator-specific products. For larger buildings, the operators and neutral hosts offer in-building solutions such as distributed antenna systems and small cells, which is a commercial conversation rather than a box you buy.

Leaky feeder

For tunnels, long car parks and linear spaces, leaky feeder is the CERN answer: slotted coaxial cable that radiates along its length. It's an infrastructure project rather than a product, but for a long underground space it's the only way to get continuous coverage end to end.

A different radio

Lower frequencies penetrate further. LTE450, which runs LTE at 450 MHz, has been adopted by utilities in parts of Europe precisely because it reaches into basements and chambers that higher bands can't. Our sister site lte450.co.uk covers it in depth. Private networks are another route: a LoRaWAN or Wi-Fi HaLow gateway at street level or on a roof can collect data from devices below and backhaul it over cellular or fibre, which moves the cellular problem to one well-placed box instead of hundreds of difficult ones.

Just use a wire

If there's an existing data network in the building, the cheapest, most reliable connection may be a patch lead. It's not exciting, but it doesn't drop out when it rains. The same goes for fibre already in a duct. Cellular is brilliant where nothing else is available, but it's not obliged to be the answer to everything.

Two special cases worth knowing about

Lift emergency lines

Lift emergency autodiallers have traditionally used analogue phone lines, and the UK industry is targeting 31 January 2027 for withdrawing the PSTN. Many lift owners are moving to cellular autodiallers, often mounted in the lift shaft or machine room, which are exactly the kind of enclosed, concrete-lined spaces this guide is about. Don't swap one obsolete technology for another: EE has confirmed a 2G switch-off in May 2029, Vodafone in 2030, and the UK operators have said no 2G will run beyond 2033. New installations should be 4G with voice over LTE. Our VoLTE explainer explains why that matters for voice.

Out-of-band management in data centres

A 4G out-of-band connection is only useful if it works when everything else has failed. That means the antenna can't be stuck on the back of a rack in a shielded hall with no signal. Plan for an external antenna, a cable route through an approved penetration, or an outdoor router on the roof linked back over Ethernet. And test the out-of-band path regularly, because the first time anyone finds out it doesn't work should not be during an outage.

Survey first, then specify

Every problem above is cheaper to find before installation than after. A proper survey doesn't need to be elaborate, but it does need to be done where the device will live. A practical checklist:

As a rough rule of thumb, many engineers treat an RSRP better than about -100 dBm as comfortable, -100 to -110 dBm as workable with care, and below -110 dBm as marginal for standard LTE, though NB-IoT and LTE-M can work well below that. Use the readings to decide between the options above, not to justify the option already chosen.

Peter's view

After twenty-odd years of this, I'm convinced most underground connectivity failures aren't technology failures. They're survey failures. Somebody tested in the wrong place, or didn't test at all, or tested the antenna and forgot the lid, the cable, the flood, the van roof or the building manager. The kit is usually fine. It's been asked to do something physics won't allow.

The good news is that once you think of it as a path from the device to the sky, with every metre of soil, concrete, cable and cast iron taking its share, the answer usually becomes obvious. Sometimes it's a better antenna in a better place. Sometimes it's NB-IoT instead of LTE. And quite often it's admitting the radio belongs on the roof, and running a bit of Ethernet down to where the work gets done.

Frequently asked questions