How 5G actually works

One big mast does not work

Put a single enormous transmitter on a hill and everyone can hear it. The problem is the other direction: every phone in the county is now sharing one lump of spectrum, and spectrum is finite and expensive. Double the users and everyone gets half as much. The answer is to do the opposite of what instinct suggests — turn the power down, and build far more of them.

The picture in the textbook is out of date

Classic diagrams colour the honeycomb in, because early networks had to give neighbouring cells different frequencies to stop them interfering — often a repeating pattern of seven. That threw away six sevenths of the spectrum in every cell.

Modern networks, 4G and 5G alike, do the opposite: every cell uses every frequency, all the time, and the interference that creates is managed rather than avoided. Base stations coordinate with each other, schedule around the worst clashes, and use the aiming tricks in step 06 to keep signals off each other. It is far harder, and it is worth roughly seven times as much spectrum.

Why masts appear where they do

A cell is planned around demand, not around neatness — which is why there is a mast on a retail park and none along a valley with twenty houses in it. Each site costs real money to build and to keep running, so the map of coverage is, in the end, a map of where enough people stand still.

Listening in the dark

The phone does not know which frequency to try, so it sweeps the bands it supports looking for a very distinctive repeating pattern. That pattern does two jobs at once: it identifies a cell, and it hands the phone the network's timing — the shared clock that everything after this depends on. Get the timing wrong by a few microseconds and your transmission lands in somebody else's slot.

A phone that has been in this place before cheats: it remembers the last frequencies that worked and tries those first, which is why switching on at home takes a second and switching on after a long-haul flight takes considerably longer.

Choosing, and being chosen

System information tells the phone which operator this cell belongs to. Your SIM knows which network is home and holds a ranked list of others to accept when it is not — which is exactly what roaming is. It also carries rules the phone must obey: how much power it may use, which frequencies are permitted here, and whether this cell is even accepting new connections.

Emergency calls are the exception

All of the above works without a SIM, and without a subscription. A phone with no account, or one that has been barred, can still complete this sequence far enough to place an emergency call on any network with coverage — which is why an old handset in a drawer with a dead contract will still dial 999.

The SIM is a computer, not a memory card

It has a processor, storage and its own small operating system, and it is built so that physically attacking it destroys what is inside. The important thing it holds is a key that was written at manufacture and has never been read out since. When the network asks a question, the SIM does the arithmetic internally and returns only the answer — the phone it is plugged into never sees the key either.

Both directions, since the 2G lesson

Early digital networks only checked the phone. Nothing checked the network, so anyone with the right equipment could put up a mast, have nearby phones attach to it, and listen. From 3G onwards the check runs both ways: the network must prove it also knows the key before your SIM will answer.

What 5G genuinely improved

On 4G, the very first message a phone sent could contain its permanent identity in the clear. That is what surveillance devices exploited — sit in a car, listen, and collect a list of every subscriber who walked past. 5G fixes it properly: the permanent identity is encrypted with the home operator's public key before it is ever transmitted, so only your own operator can read it. Everything after that uses temporary identities that are rotated regularly.

This is one of the few 5G improvements that is genuinely, structurally better rather than simply faster — and almost nobody advertises it.

eSIM changes the plastic, not the logic

An eSIM is the same secure chip, soldered in, with the profile downloaded rather than posted. The keys are generated and stored in exactly the same way. What changes is commercial: switching network becomes a two-minute job instead of a trip to a shop, which is precisely why it took so long to arrive.

What this buys

• It degrades gracefully. A crowded Wi-Fi channel collapses, because the collisions themselves consume the airtime. A crowded cell just gives everyone a smaller share.
• It can prioritise. An emergency call, a voice packet and a background sync are not equal, and the scheduler knows it.
• It can exploit good luck. Phones constantly report how clean their signal is; the scheduler favours whoever can currently use the spectrum most efficiently, while keeping everyone fed.
• It can be sold. Guaranteed service levels are possible when someone controls the allocation, which is the basis of the entire enterprise 5G pitch.

Why uploads are slower than downloads

Mid-band 5G usually shares one block of frequency between both directions, alternating in time. The split is deliberately lopsided — typically three or four downlink slots for every uplink one — because that matches what people actually do. It is also why your upload speed does not improve when the cell is quiet: the pattern is fixed by configuration, not by demand, and changing it would interfere with neighbouring cells that are still using the old pattern.

Where the low latency comes from

4G divided time into one-millisecond chunks; you waited for the next one before anything could happen. 5G can use chunks as short as an eighth of a millisecond, by spacing its subcarriers eight times further apart. That is where the headline latency figures come from — though as step 08 explains, the radio is rarely the part that is slow.

The part that is not like Wi-Fi at all

Wi-Fi runs in unlicensed spectrum: free, shared, first come first served. Mobile spectrum is licensed — auctioned by the regulator, exclusive to one operator in one country, and extraordinarily expensive. UK operators have collectively spent billions on slices of it.

That cost is why a mobile network can promise things Wi-Fi cannot. Nobody else is legally allowed to transmit in your operator's band, so interference comes only from their own cells, which they control. It is also why the coverage map looks the way it does: having paid for the spectrum, an operator has to build enough sites to make it earn its keep.

Why millimetre wave did not take over

It was the headline of the early 5G marketing, and the physics were never on its side. At those frequencies the signal is absorbed by foliage, blocked by ordinary glass and attenuated measurably by rain — and by the hand holding the phone. Getting useful coverage means a small cell roughly every couple of hundred metres, each needing power and fibre. It works beautifully in a stadium, an airport or a dense city block, and it makes no economic sense anywhere else.

The reason some 5G is barely faster

An operator with no spare mid-band in an area can run 5G in its existing 4G frequencies, splitting each one between the two standards from moment to moment. Your phone displays 5G — accurately — and delivers 4G speeds, because it is the same spectrum doing the same work with a slightly more modern encoding. It is not dishonest, but it is the single biggest reason people conclude 5G was overhyped.

Why this arrived with mid-band, and not before

An array needs its elements spaced about half a wavelength apart. At 3.5 GHz a wavelength is roughly 8.5 cm, so sixty-four elements fit into a panel two people can carry up a tower. At 700 MHz a wavelength is 43 cm, and the same array would be the size of a garage door.

That is the whole reason massive MIMO and mid-band 5G appeared together. It is not that nobody thought of beamforming earlier — it is that at the frequencies mobile networks previously used, the hardware would not have fitted on the mast.

The same frequencies, twice over

Aiming does more than improve one person's signal. If two phones are in genuinely different directions, the base station can send completely different data to both of them at the same moment on the same frequency, because each beam is quiet where the other is loud. That is real multiplication of capacity out of no extra spectrum, and it is the single largest reason a mid-band 5G cell carries so much more than the 4G cell it replaced.

Finding you in the first place

A narrow beam is only useful if it is pointed correctly. The base station sweeps a set of candidate beams across its sector, the phone reports which it heard most clearly, and the pair settle on one — then keep checking, because turning your body through ninety degrees can be enough to change the answer. On millimetre wave, where the beams are narrowest, a phone can switch beams several times as you walk a few steps.

Not too eager, not too late

Handover is tuned with deliberate stubbornness. A target has to be better than the current cell by a margin, and stay better for a set period, before anything happens. Without that, a phone sitting exactly between two masts would bounce between them continuously — each handover costing a small interruption, the sum of which is a call that breaks up while standing perfectly still. Getting these thresholds right is a large part of what radio planners actually do.

Two completely different states

A phone doing nothing is not connected in any meaningful sense. It has no allocation, it is not being scheduled, and the network does not know precisely which cell it is in — only roughly which area, which is enough. It wakes periodically to listen for its own name being called.

When a call or a message arrives, the network broadcasts a page across every cell in that area; your phone hears it, connects properly, and only then does anything happen. This is why the first second of receiving a call feels different from the rest, and why a phone idling all day uses so little battery despite being reachable the whole time.

Why trains are the hardest place

Everything is against you at once. The train is moving fast enough that handovers arrive every few seconds. It is a metal tube, which is an excellent radio shield. The line runs through cuttings and tunnels chosen by Victorian engineers with no interest in line-of-sight. And several hundred people are all doing this simultaneously in one small volume, which concentrates demand into whichever cell the train happens to be passing.

The two 5Gs nobody explains

Most 5G deployed so far is non-standalone: a 5G radio bolted onto the existing 4G core, using the old network for all the control and signalling. It was quick to deploy and it delivered the headline speeds, which is exactly why operators did it.

Standalone 5G replaces the core as well, and it is where nearly all the interesting claims live — slicing, the very low latency figures, the machine-to-machine features. Until a network is standalone, a phone showing 5G is using a 5G radio and a 4G brain. This is the single biggest gap between what 5G was promised to do and what most people have actually been using.

Where the latency really goes

The radio can genuinely get a packet across in a millisecond or two. Then it travels on fibre to a core site, which may be a hundred miles away, then out to a server that could be anywhere on Earth. A ten-millisecond radio and a thirty-millisecond journey to a data centre in Dublin gives you forty milliseconds, and no amount of 5G improves the second number.

Edge computing is the answer to this and it is not a radio technology at all: move the thing you are talking to physically closer — into the operator's own regional sites. The headline latency figures for 5G assume this has been done, and for most services it has not.

Backhaul is the quiet bottleneck

Every site needs a fibre connection sized for its busiest moment. Upgrading the radio at a site that is already limited by its backhaul achieves precisely nothing, and it is a common and expensive mistake. A surprising amount of the money and time in a 5G rollout has gone into digging, not into antennas.

Symptom to cause

The one thing actually worth trying

Toggle flight mode. It forces the phone to release its connection and repeat steps 02 and 03 from the beginning, which genuinely fixes a phone that is clinging to a distant cell it attached to hours ago somewhere else. Almost every other piece of folk advice — restarting, removing the SIM, resetting network settings — is a slower version of the same thing.

If it makes no difference, the problem is not on your phone. It is capacity, coverage or backhaul, and all three belong to somebody else.

What 5G was actually for

The speed was never the interesting part, and a decade on it is clear the marketing oversold it badly. What the standard genuinely delivers is capacity in dense places, far better identity privacy, and a network that can be divided up and guaranteed in software. Those are unglamorous, they are real, and you will never see any of them on a billboard.

Conspiracy theorists

Some people have been destroying 5G equipment and spreading rumours that quite frankly do not make any sense if you understand the technology.