A radio link looks deceptively simple: a transmitter sends energy, a receiver catches it. In practice every step of that journey is a negotiation with physics, and understanding the full chain — from power amplifier to demodulator — explains almost everything about why networks are designed the way they are.
01Power Out, Antenna Gain and the Effective Radiated Signal
The transmitter in a handset produces a few hundred milliwatts at most. That power travels along a short feed path to the antenna, losing a small but real fraction along the way — feed loss, measured in decibels, is small but not zero. What the antenna then does is redistribute what remains. It does not create energy; it shapes the pattern of radiation, concentrating output in some directions at the expense of others. That concentration is called gain, also measured in decibels relative to an isotropic radiator (dBi) or a half-wave dipole (dBd).
The combination of transmit power, feed loss and antenna gain gives you a number called the Effective Isotropic Radiated Power, or EIRP. It is the single figure that describes what the transmitter is actually putting into the air in the direction of interest. A handset with 23 dBm of transmit power and a modest antenna might achieve an EIRP not far from that number; a base-station sector antenna with 18 dBi of gain stacked onto a 40 W amplifier produces an EIRP in the kilowatt range — or expressed in dBm, something in the neighbourhood of 60 dBm. The asymmetry matters: the mast shouts; the handset whispers back. Managing that imbalance — the uplink-versus-downlink budget — is a central design problem in every cellular system.
02The Path: Where the Power Goes
Managing that imbalance — the uplink-versus-downlink budget — is a central design problem in every cellular system.
From this piece
Once the signal is in the air, the dominant mechanism eating it is free-space path loss. This is not a property of atmosphere or obstacles; it is a simple geometric fact. Energy radiates outward in a sphere, and the power intercepted by any fixed-area receiving antenna decreases as the square of the distance. Double the distance, lose 6 dB. Move from 1 km to 10 km, lose 20 dB. The relationship also scales with frequency: higher frequencies suffer greater free-space loss at the same distance, which is one of the reasons millimetre-wave 5G NR cells are small while sub-1 GHz LTE cells can span many kilometres.
Free-space loss is the baseline. Over a real propagation path, additional mechanisms pile on top. Terrain and buildings produce diffraction loss — the signal bends around obstacles but loses energy doing so. Dense urban canyons introduce multiple reflections, which at first sounds helpful (and sometimes is, via multipath diversity) but also causes destructive interference between copies of the signal arriving at slightly different times and phases. That phenomenon is fading. Slow fading varies over hundreds of metres and tracks gross changes in the environment — entering a building, moving behind a hill. Fast fading varies over distances of half a wavelength or less and is driven by the interference pattern of multipath arrivals. At 900 MHz, a half-wavelength is about 17 centimetres; at 3.5 GHz it shrinks to around 4 centimetres. For a person walking, fast fades can come and go several times a second.
Atmospheric effects are usually secondary at frequencies below a few gigahertz, but above roughly 10 GHz oxygen and water-vapour absorption become meaningful, and rain attenuation at millimetre-wave frequencies can be severe enough to shape network deployment decisions.
03What Arrives at the Receiver
After the path, the signal lands at the receiving antenna — now with an antenna gain of its own, adding back some decibels. The difference between the EIRP transmitted and the level arriving at the receiver output is called the path loss (or, accounting for both antenna gains, the link loss). The received signal power might be −90 dBm, −100 dBm, or lower still. To put that in perspective, −100 dBm is one tenth of a picowatt (100 femtowatts).
That vanishingly small signal does not arrive alone. The receiver itself generates thermal noise — random electron motion in the input circuitry — whose power depends on temperature, bandwidth and the receiver's noise figure. This is the noise floor. In a 10 MHz channel at room temperature, the thermal noise floor is around −104 dBm; a real receiver with a noise figure of 7 dB raises that floor to roughly −97 dBm. The ratio of signal power to noise power at the receiver input is the received SNR, and it is this number — not raw signal strength alone — that determines whether the signal can be decoded.
Between signal and noise sits a quantity called Eb/N0 (energy per bit relative to noise spectral density), which bridges the physical layer to the modulation and coding scheme in use. A link running a high-order modulation — 256-QAM, say — needs a substantially higher SNR than one running QPSK. Modern cellular systems use adaptive modulation and coding (AMC) to select the most efficient scheme the current SNR supports, shifting automatically as conditions change. When the link is strong, the system extracts the most bits per hertz; when it degrades, it backs off to a more robust scheme and trades throughput for reliability.
04The Link Budget: Counting Every Decibel
Engineers formalise all of this in a link budget — a ledger of every gain and loss between transmitter and receiver, summed in decibels. A typical entry reads:
Transmit power, plus antenna gain, minus feed loss, minus path loss, minus shadowing margin, plus receive antenna gain, minus receiver noise figure — and the result must exceed the required minimum signal level (the receiver sensitivity) by a margin called the fade margin or link margin. If it does not, the link fails.
The fade margin deserves attention. A system designed to the median signal level would fail half the time — every point where instantaneous fading dips below median would produce an outage. Practical networks are designed for coverage probability, not median coverage: planners add a margin (typically 5–10 dB for an outdoor link, more for in-building penetration) to ensure the link holds through the statistical distribution of fades and shadowing. That margin costs range — every decibel added to the margin is a decibel subtracted from how far the transmitter can reach — so every dB is contested.
05Uplink, Downlink and the Real Asymmetry
The link budget must be solved in both directions separately. Downlink (mast to handset) typically benefits from high base-station transmit power and a high-gain directional antenna. Uplink (handset to mast) is constrained by the handset's power limit — around 23 dBm in LTE for a Category 1 device, and governed by power control mechanisms that prevent any one handset from drowning out others on shared spectrum. The base station compensates on its receive side with multiple antennas, advanced receiver algorithms, and in modern systems, massive MIMO — large antenna arrays that can spatially separate signals and effectively add array gain on the uplink to compensate for what the handset cannot transmit.
Power control is worth pausing on. In every CDMA and OFDMA system, the network continuously adjusts each handset's transmit power to arrive at the base station at roughly the target SNR — no more, no less. Transmitting more than necessary wastes battery, heats the handset and raises the noise floor for every other user on the same carrier. The algorithm closes the loop many times per second, and it is one of the reasons modern cellular links are far more efficient than their early-generation predecessors.
Received power −55.6 dBm against an assumed −100 dBm sensitivity: a margin of 44.4 dB. Scrub the live version on the homepage →
06Why the Math Is Never Perfect
The received signal is also corrupted by interference from adjacent cells and co-channel transmissions — a contributor distinct from thermal noise, and one that pure link-budget arithmetic does not fully capture. Real networks operate in a regime defined not by noise alone but by signal-to-interference-plus-noise ratio (SINR), and managing interference through frequency planning, sectorisation, beamforming and coordination is where much of cellular network engineering actually lives. The physics of the path are fixed; the engineering task is to make the most of the budget while keeping interference under control on both sides of every link.
Fast fading varies over distances of half a wavelength or less and is driven by the interference pattern of multipath arrivals.
From this piece
| Figure | What it describes |
|---|---|
| 23 dBm | typical maximum handset uplink transmit power (LTE Category 1) |
| 6 dB | path loss increase for each doubling of distance (free-space) |
| 20 dB | free-space path loss increase for a 10× increase in distance |
| −104 dBm | approximate thermal noise floor in a 10 MHz channel at room temperature |
| 17 cm | half-wavelength at 900 MHz (governs fast-fading scale) |
| 4 cm | half-wavelength at 3.5 GHz |
| 5–10 dB | typical outdoor fade margin in network planning |