Every radio signal is an electromagnetic wave, and frequency is the rate at which that wave oscillates. Change the frequency and you change almost everything about how the signal behaves: how far it reaches, how well it bends around obstacles, how much information it can carry, and how big the antenna needs to be to catch it.
None of this is policy or convention. It is physics, and it sets the terms for every network design decision that follows.
Start with free-space path loss. Even in a perfect vacuum with nothing to absorb or scatter the signal, a receiver farther from a transmitter receives less power — the wave spreads over a growing sphere of area as it travels. That spreading loss increases with distance, but it also increases with frequency. Double the frequency and you add roughly 6 dB of path loss at the same distance. This single fact drives much of the architectural difference between, say, a rural 700 MHz deployment and a city-centre millimetre-wave 5G small-cell layer.
01Low frequencies: distance and penetration, at a cost
Below roughly 1 GHz, signals travel far and behave forgivingly. A base station transmitting in the 700–900 MHz range can cover tens of kilometres under open terrain. More usefully for network operators, low-frequency signals diffract — they bend around hills and the corners of buildings with far less loss than higher bands. They also penetrate building materials more readily, because the wavelength is long relative to the typical dimensions of a brick or a concrete wall.
The wavelength of a 900 MHz signal is around 33 centimetres. A signal at 700 MHz has a wavelength close to 43 centimetres. These are large compared with a window frame or a wall cavity, so energy flows around and through rather than being blocked outright. This is why coverage deep inside a building — the basement car park, the stairwell — is almost always driven by low-band spectrum, and why emergency services networks have historically been built on sub-1 GHz allocations.
The cost is channel width. The useful bandwidth available at 700 MHz is measured in tens of megahertz at most; the entire band is narrow in absolute terms. Bandwidth is the pipe, and a narrow pipe has a low ceiling on data throughput regardless of the modulation scheme applied. Shannon's theorem is unambiguous: capacity is proportional to bandwidth. You cannot squeeze a gigabit per second through a 10 MHz channel no matter how cleverly you encode the signal.
Bandwidth is the pipe, and a narrow pipe has a low ceiling on data throughput regardless of the modulation scheme applied.
From this piece
02High frequencies: capacity and its consequences
Above roughly 1 GHz, and especially above 6 GHz, the physics reverses. Path loss climbs steeply with frequency. A millimetre-wave signal at 28 GHz suffers dramatically more free-space spreading loss over the same distance than a 3.5 GHz mid-band signal. Rain, humidity and atmospheric oxygen absorb energy at specific frequencies — oxygen has absorption peaks around 60 GHz that make that band nearly useless for anything beyond a few hundred metres. Diffraction falls sharply too: a 28 GHz wave has a wavelength of about 11 millimetres, small compared with a door frame or a window sill, so it does not bend around edges; it is effectively blocked by them. Materials that low-band signals pass through with modest loss can stop a millimetre-wave signal almost entirely.
What high frequencies offer in return is bandwidth. The spectrum above 24 GHz contains contiguous allocations that can be hundreds of megahertz or even multiple gigahertz wide. Wide channels support high data rates. A 400 MHz channel at 28 GHz, with a capable modulation scheme, can deliver multi-gigabit throughput to a small area — something physically impossible at 700 MHz. This is why 5G NR includes millimetre-wave operation for dense urban scenarios: the capacity is genuine, even though the coverage per site is measured in hundreds of metres rather than tens of kilometres.
Antenna size follows wavelength directly. A half-wavelength dipole at 700 MHz is about 21 centimetres long. The same antenna at 28 GHz is roughly 5 millimetres. This miniaturisation is what enables the large antenna arrays packed into 5G millimetre-wave hardware: hundreds of elements fit into a panel the size of a paperback book, enabling the narrow, steerable beams that partially compensate for the poor propagation with directed gain.
03The trade in practice
Real network design lives in the negotiation between these extremes. Mid-band spectrum — broadly 1 GHz to 6 GHz — occupies the middle ground: better coverage than millimetre wave, more capacity than sub-1 GHz. The 3.5 GHz band that anchors most early 5G deployments sits here precisely because it offers workable cell sizes alongside channel widths that were difficult to find at lower frequencies.
High bands deliver peak capacity in specific dense locations.
From this piece
No single frequency band wins overall. Low bands anchor coverage, particularly outdoors and in-building. Mid-bands carry the bulk traffic load. High bands deliver peak capacity in specific dense locations. A complete network uses all three layers because the physics at each frequency serves a different part of the problem — and the physics is not negotiable.
The engineer choosing spectrum is not choosing a preference. They are choosing which set of constraints to accept.