How Does an Antenna Work? From Wired Signal to Radio Wave and Back

The whole link in one picture

Before the detail, the chain. A signal is never "sent through the air" directly: something has to convert it, twice.

1Before the antenna: a signal that is still guided

Everything up to the antenna port is a voltage travelling inside metal.

In a base station the data path starts as bits in the baseband unit. Those bits are mapped onto a modulation scheme, up-converted to a carrier in the operating band, and amplified. At the output of the radio unit you no longer have "data": you have an alternating voltage whose amplitude and phase carry the information, oscillating several billion times per second.

That voltage travels along a coaxial feeder. A coaxial cable is not a wire in the ordinary sense; it is a transmission line, a structure that guides an electromagnetic wave along its length. The wave exists in the dielectric between the inner conductor and the shield, not in the copper itself.

For the energy to enter the antenna instead of reflecting back towards the amplifier, one condition must hold: the antenna's input impedance must equal the characteristic impedance of the feeder. In radio systems that number is nearly always 50 ohm. When it does not match, part of the wave turns around at the connector and travels backwards. That reflected wave is measured as VSWR and return loss:

• Return loss of 14 dB corresponds to VSWR 1.5, and means about 4 percent of the incident power is reflected.
• Return loss of 10 dB corresponds to VSWR 1.9, and means about 10 percent is reflected.
• A cellular antenna datasheet usually specifies VSWR of 1.5 or better across the band, which is why the 14 dB figure appears so often.

2The conversion: how a guided wave becomes a radiated wave

This is the step most explanations skip. It is also the only step that is genuinely "antenna physics".

Inside the feeder, the electric and magnetic fields are confined: they are trapped between two conductors. An antenna is built so that the two conductors stop being parallel. Once the structure opens out, the field is no longer confined and the wave detaches from the metal.

The mechanism is accelerating charge. Electrons in the antenna element are pushed back and forth by the applied voltage. A charge moving at constant velocity carries a field with it, but a charge that is accelerating radiates: the field lines cannot keep up, they pinch off, and a closed loop of electric field leaves the conductor and travels away at the speed of light. Alternate the current fast enough and you get a continuous train of these loops: a radio wave.

Why the element has to be about half a wavelength

At the resonant length the current distribution forms a standing wave with a maximum at the feed point and a null at the tips, and the input impedance becomes resistive and easy to match. For a centre-fed dipole that length is close to half a wavelength, shortened by a few percent because of the thickness of the conductor.

Wavelength follows directly from frequency:

λ = c / f — 299,792,458 m/s divided by the frequency in hertz

This single relationship explains the shape of every antenna you have ever seen: a 433 MHz LoRa antenna needs a half-wave element of roughly 33 cm, while a 3.5 GHz 5G element needs only about 4.3 cm. The lower the band, the larger the structure, and the more of the mechanical design is driven by physics rather than by choice.

Near field and far field

The region immediately around a radiating antenna does not behave like a travelling wave; energy sloshes back and forth between the antenna and its surroundings. Only beyond the Fraunhofer distance does the structure settle into a wave whose field components are perpendicular to each other and to the direction of travel. That boundary scales with the square of the antenna size:

R = 2D² / λ

For a 2 m tall panel antenna at 3.5 GHz the far field starts at roughly 93 m. This is the reason pattern measurements are made on a long outdoor range or inside a large anechoic chamber, and it is also why the pattern printed on a datasheet describes the antenna in free space, not the pattern you will measure two metres away from it on a rooftop.

3The air: what actually happens between the two antennas

Free space is not a lossless pipe. It is a geometric spreading medium with obstacles in it.

Spreading loss, and why it is not really "absorption"

There is nothing in a vacuum to absorb a radio wave. The loss comes from geometry: the same power is spread over the surface of a sphere that grows with distance, so the power density falls with the square of distance. Written in the units engineers actually use:

FSPL (dB) = 32.44 + 20 log₁₀(d in km) + 20 log₁₀(f in MHz)

At 3.5 GHz over one kilometre this gives 103.3 dB. Two consequences follow immediately, and both drive real design decisions. First, doubling the distance costs 6 dB, which is a factor of four in power. Second, doubling the frequency also costs 6 dB at the same distance, which is why millimetre-wave cells are dense.

Beamwidth is a trade, not a quality

An antenna does not create power; it concentrates it. Directivity from a rectangular beam can be estimated from the two half-power beamwidths, and the estimate is close enough for selection work:

D ≈ 41,253 / (θ_H × θ_V)

Set the vertical beamwidth to the 7 degrees typical of a cellular panel and vary the horizontal beamwidth. The number you get back is the price of coverage: every degree you narrow adds gain in one direction and removes it from another.

Three more things the air does to the signal

• Multipath. Buildings and ground reflect the wave, so several copies arrive with different delays. This causes fading and intersymbol interference, but it is also the raw material that makes MIMO spatial multiplexing possible. No multipath, no second data stream.
• Fresnel clearance. A link needs more than a clear line of sight; it needs an elliptical volume around that line to be free of obstructions. At 3.5 GHz over 5 km the first Fresnel radius is about 10 m, so a rooftop edge or a tree crown can cost several dB while the visual path still looks open.
• Polarisation mismatch. A linearly polarised receive antenna that is rotated 90 degrees relative to the transmit antenna sees almost nothing. Cross-polarisation also matters inside MIMO panels, where it is deliberately used to decorrelate the two streams.

4The terminal: the hardest antenna in the link

The base station has metres of aperture and watts of power. The handset has neither.

Everything in the previous section assumed an antenna with room to work. A smartphone has roughly the volume of a small box, contains a metal frame, a display, a battery, and a user's hand. Its antennas are printed on the edge of the PCB or moulded into the frame, and they are compromises on every axis.

• Aperture. Radiation resistance and bandwidth both scale with electrical size. At 700 MHz a handset antenna is a small fraction of a wavelength, so it is inherently narrowband and inefficient.
• Efficiency. Total efficiency of a well-designed handset antenna is typically 40 to 60 percent, and it can halve when the user's hand covers the radiating edge.
• Body loss. Head and hand absorption typically costs 3 to 6 dB, and up to 10 dB in a poor grip. This is a loss the network cannot compensate for except by increasing downlink power.
• MIMO. Two or four receive antennas in a handset do not increase range; they increase throughput and reliability under good signal conditions, by exploiting the decorrelation that multipath provides.

5Back to wire: how the wave becomes data again

Reciprocity guarantees the physics works in reverse. It does not guarantee the signal is strong enough to be understood.

The receiving antenna intercepts a fraction of the incident wave. The electric field pushes electrons along the element, producing a voltage at the terminals measured in microvolts. Because the antenna is reciprocal, its radiation pattern, directivity and impedance are the same in receive as in transmit. What changes is the scale: the received power is set by the power density at that point multiplied by the effective aperture, which for a small antenna is modest.

From the terminals, the signal goes into a low-noise amplifier, then a filter, then the mixer and analogue-to-digital converter, and finally the demodulator. Whether any of this works is decided by one comparison: is the received power above the sensitivity of the receiver?

Sensitivity itself is not a fixed property of a radio. It follows from thermal noise and the required signal-to-noise ratio:

• Thermal noise floor at room temperature is −174 dBm per hertz.
• A 20 MHz channel therefore carries about −101 dBm of noise.
• Add a 7 dB receiver noise figure: about −94 dBm.
• Require 3 dB signal-to-noise ratio to demodulate: sensitivity is about −91 dBm.

That chain of four lines is why a datasheet sensitivity figure of −100 dBm implies a narrower channel, a lower noise figure, or a more robust modulation. The number is never independent of the conditions it was measured under.

What this changes when you specify an antenna

Every parameter in the five stages above maps onto a line in a purchase specification.

Two of the three calculators on this page also exist as standalone tools, with more inputs: the link budget calculator covers both directions including feeder and body loss, and the Fresnel zone calculator checks obstruction clearance for a planned hop. If you are still deciding between coverage shapes, the comparison of sector and panel antennas takes the same physics one step further into selection.

Frequently asked questions