Everything in the course so far has happened inside a wire. We chose a carrier, we modulated it, we shaped its pulses, we counted its bits per hertz. But a radio link is not a wire — at some point the signal has to leave, cross a room or a continent as a free-space wave, and then be picked up again. The component that performs both halves of that trick is the antenna, and it is the most physical, least abstract part of any radio.
Start with what does not radiate. A charge sitting still has an electric field around it, but that field just sits there too — nothing travels. Push that charge along a wire at a constant rate and you get a steady current, which adds a static magnetic field looping around the wire. Still nothing travels. A DC circuit, no matter how much current it carries, radiates no power at all: this is why the wiring in your walls is not a transmitter.
The one thing that makes a charge radiate is acceleration. When a charge changes velocity, the field lines attached to it cannot rearrange instantly — information about the new position propagates outward at c, so a kink appears in the field and travels away from the charge forever. That kink, self-sustaining because a changing E field makes a B field and vice versa (Module 1), is the electromagnetic wave. Radiated power scales with the square of the acceleration:
Now put an alternating current on a straight wire. Electrons are driven toward one end, decelerate, reverse, and race back the other way, f million times a second. They are accelerating almost constantly, in step, all along the wire — and their contributions add up in phase in some directions. That is an antenna. Nothing more exotic is required: a conductor carrying an oscillating current radiates, whether you wanted it to or not, which is also why unintended radiation from clock traces is an entire engineering discipline.
The cleanest way to think of an antenna is as a transducer, in the same sense as a loudspeaker. A loudspeaker converts an electrical signal into a pressure wave in air; an antenna converts a guided electromagnetic wave — the one travelling along a coaxial cable or a circuit-board trace, bound to its conductors — into an unguided wave that propagates on its own with no conductor to follow. It changes the form of the energy, not the information the energy carries.
Like a loudspeaker, it also has to be matched to what it is driven by. A feedline presents a characteristic impedance (50 Ω almost everywhere in radio); an antenna presents its own input impedance, and any mismatch reflects power back down the line instead of launching it. A half-wave dipole in free space sits near 73 Ω, which is a large part of why 50 Ω and 75 Ω became the standards they are. The parameters that quantify how well the transducer performs — gain, radiation pattern, polarization, bandwidth — are the whole of M7-L2; here we only need the picture.
An antenna does not amplify. This trips up nearly everyone the first time. A passive antenna adds no energy, so its “gain” is not amplification — it is redistribution. Concentrating radiation into a narrow beam makes the signal stronger in that direction and correspondingly weaker everywhere else. Total radiated power is fixed by the transmitter; the antenna only decides where it goes.
For the charges along a wire to add up constructively rather than fight each other, the wire has to be a sensible fraction of the wavelength of the current on it. Make it much shorter and the current distribution is squeezed into a tiny region, radiation resistance collapses, and almost all your power warms the conductor instead of leaving it. The natural size, where the wire resonates and current flows freely, is half a wavelength — and via c = fλ (M1-L2) that fixes a physical length from a frequency:
| Band | Frequency | λ = c/f | Half-wave dipole |
|---|---|---|---|
| AM broadcast | 1 MHz | 300 m | 150 m (in practice a 75 m λ/4 mast over ground) |
| FM broadcast | 100 MHz | 3 m | 1.5 m |
| Cellular (GSM) | 900 MHz | 33.3 cm | 16.7 cm |
| WiFi / Bluetooth | 2.4 GHz | 12.5 cm | 6.25 cm |
| Satellite / radar | 10 GHz | 3 cm | 1.5 cm |
| 5G mmWave | 28 GHz | 1.07 cm | 5.4 mm |
The table explains a great deal of the physical world at a glance. AM stations need land, not equipment racks, because their quarter-wave masts are 75 m tall. A car’s old whip antenna was cut for the 100 MHz FM band. And a 28 GHz 5G array can put dozens of elements in the space of a fingernail — which is exactly why beamforming became practical at millimetre wave and not at VHF.
Close to an antenna the fields are messy. Energy sloshes back and forth between the antenna and the space immediately around it without ever escaping, E and B are not in the neat perpendicular ratio of a travelling wave, and the field shape changes as you move. Far away, all of that settles down: the wave is locally a plane wave, E and B are perpendicular and in phase, power falls as 1/r², and the shape of the pattern stops changing — only its amplitude does. Engineers therefore split the space around an antenna of largest dimension D into three regions:
Take D = 1 m and f = 10 GHz, so λ = 3×10⁸/10¹⁰ = 0.03 m. The reactive boundary is 0.62√(D³/λ) = 0.62√(1/0.03) = 0.62 × 5.774 = 3.6 m. The far-field boundary is 2D²/λ = 2 × 1² / 0.03 = 66.7 m. So:
Sixty-seven metres is an inconvenient number, and that is the point: characterising a large high-frequency antenna needs either a very long outdoor range or a near-field scanner that measures close in and transforms the result mathematically. One honest caveat — the 2D²/λ criterion assumes D is comfortably larger than λ. For a small antenna (a 6.25 cm dipole, say) the formula returns centimetres, which is nonsense; there the binding condition is simply r > a few wavelengths, conventionally r > 3λ. Always apply whichever of the two is larger.
Why this matters outside the lab. Human RF-exposure limits are written in terms of far-field power density, but a phone against your head is deep in the near field of its own antenna — which is exactly why handset compliance is measured as SAR (absorbed power per kilogram of tissue) instead. It is also why an NFC tag or a wireless charger works at all: those are near-field devices, coupling reactively at a few centimetres, and they are not trying to radiate.
Here is the most useful fact in the subject. For any antenna built from ordinary linear, passive materials, the reciprocity theorem guarantees that its behaviour is identical whether it is transmitting or receiving. The same radiation pattern, the same gain, the same input impedance, the same polarization response, the same bandwidth. An antenna that radiates 20 dB more strongly toward the horizon than upward is also 20 dB more sensitive toward the horizon than upward, by the same numbers.
What that buys an engineer is enormous: you measure once and use the result both ways. Antenna ranges almost always test in transmit mode because it is easier to instrument, then quote the figures for receive without a second thought. Link budgets use one gain number per antenna regardless of direction. A base-station sector that covers a given area on the downlink covers the same area on the uplink. And a handset can use one antenna for both, switched or duplexed in time or frequency, rather than carrying two.
Where reciprocity stops. The theorem needs linear, passive, reciprocal materials, so it breaks the moment you add something that is not: a ferrite circulator, a magnetised plasma, or an active element inside the antenna structure. Two practical footnotes as well — reciprocity is about the antenna, not the whole radio, so a transmit chain and a receive chain still differ in their amplifiers and losses; and it says nothing about the channel being symmetric when the two ends use different frequencies, which is why FDD systems cannot simply assume the uplink channel looks like the downlink one (Module 10).
Transmitting is easy to picture — you shake charges and waves come off. Receiving is the same physics run backwards: an arriving wave’s electric field pushes the free charges in the conductor, driving a small current that the receiver amplifies. But it is often more useful to think of a receiving antenna as a collecting area: the incoming wave carries so many watts per square metre, and the antenna intercepts a certain effective patch of it. That patch is the effective aperture, and reciprocity ties it directly to gain:
Read the λ² and you get the single most important consequence in antenna engineering. At fixed gain, effective aperture shrinks as the square of frequency. An isotropic antenna (G = 1) collects λ²/4π: at 100 MHz that is 9/12.57 = 0.72 m², at 2.4 GHz it is 0.015625/12.57 = 12.4 cm², and at 10 GHz it is 9×10⁻⁴/12.57 = 0.72 cm². Over that 100:1 span of frequency the collecting area falls by 10,000:1.
This is why low frequencies need physically big antennas and why high frequencies need gain to survive at all — a millimetre-wave link only closes because a dish or a phased array claws back the aperture that λ² took away. It is also the term that makes free-space path loss appear to depend on frequency, which is a subtler statement than it looks and is the first thing M8-L1 unpacks when it derives the Friis equation. For now, one pointer is enough: gain, aperture and distance combine into a received-power formula, and that formula is Module 8’s job.