Why one antenna isn't enough
Picture a single loudspeaker in a field, playing a steady tone. The sound spreads out in a broad, soft dome — loud near the cone, weak everywhere far away. A single radio antenna does the same thing with radio waves: its radiation pattern is wide and gentle, and most of its energy goes places nobody is listening. For a 5G base station at 28 GHz that's a disaster, because at those frequencies the air itself eats your signal — recall from the link-budget rung how brutal free-space path loss becomes as frequency climbs. You need to concentrate energy, not spray it.
One classic fix is to build a big dish — a parabolic reflector that focuses energy like a flashlight reflector focuses light. It works, and it delivers enormous gain. But a dish points one way. To track a moving target you must physically swing the whole heavy structure on motors, and you can only look in one direction at a time. A radar that takes two seconds to mechanically sweep the sky is useless against an incoming missile. We need a way to make a beam and aim it instantly, electronically.
Interference is the engine: how an array forms a beam
Drop two stones into a still pond at the same instant. Where the ripples meet crest-to-crest they pile up into a tall wave; where crest meets trough they cancel into flat water. Radio waves do exactly this. An antenna array is just many small radiators driven by the same signal, spaced (typically) half a wavelength apart. In any given direction, each element's wave arrives having travelled a slightly different distance. Add them up: where they arrive in phase they reinforce — that's the beam. Where they arrive out of phase they cancel — that's a null.
Here is the key move. If you feed every element the same signal at the same time, the waves add up in phase straight ahead (broadside), and the beam points perpendicular to the array. But if you delay each element a little more than its neighbour — a progressive phase shift across the array — then the direction in which all the waves line up tilts off to the side. Change the phase slope, and the beam swings. No motor, no movement: just numbers fed to phase shifters. That is beamforming, and steering it is electronic scanning.
Plane wave leaving an 8-element array, spacing d = λ/2
Progressive phase shift Δφ between adjacent elements steers the beam to angle θ
elem: 0 1 2 3 4 5 6 7
phase: 0 Δφ 2Δφ 3Δφ 4Δφ 5Δφ 6Δφ 7Δφ
| | | | | | | |
v v v v v v v v
~~~~ ~~~~ ~~~~ ~~~~ ~~~~ ~~~~ ~~~~ ~~~~
\ \ \ \ \ \ \ \
\ \ \ \ \ \ \ \ wavefronts add
\ \ \ \ \ \ \ \ up in phase ONLY
\_____\_____\_____\_____\_____\_____\_____\ along angle θ
beam -> θ
Steering law: sin(θ) = Δφ / (k·d) = (Δφ · λ) / (2π · d)
with d = λ/2 : sin(θ) = Δφ / π ( Δφ in radians )Gain that grows with the crowd
Here is the part that feels like getting something for nothing — but isn't. When N elements all aim their energy in the same direction, two things happen at once. On transmit, the powers add and the coherent fields add, so the peak power density in the beam direction grows like N² relative to a single element radiating the same per-element power. On receive, the array pulls in N times the signal energy while the noise in each receiver — its thermal noise — is independent and adds incoherently. Both effects mean array gain rises with element count.
The clean way to state it: an ideal N-element array adds 10·log₁₀(N) dB of array gain on top of the gain of one element. Double the elements, gain rises 3 dB. And the beam gets sharper as it gets stronger — the more elements (the larger the aperture in wavelengths), the narrower the main lobe. A pencil beam from a thousand elements can be a fraction of a degree wide, which is exactly what a radar needs to tell two close targets apart.
Array gain from element count N (ideal, lossless combining)
N elements array gain over 1 element typical use
---------- ------------------------- ---------------------------
4 10·log10(4) = +6 dB small IoT / WiFi steering
16 10·log10(16) = +12 dB automotive radar (a panel)
64 10·log10(64) = +18 dB 5G mm-wave base station
256 10·log10(256)= +24 dB high-end radar / SATCOM
1024 10·log10(1024)=+30 dB big AESA radar, Starlink-class
Each element may itself have ~5 dBi of patch-antenna gain,
so a 64-element panel: 5 dBi (element) + 18 dB (array) = ~23 dBi EIRP boost
...before you even count the per-element power amplifier.Inside a real phased-array transceiver
Now the whole track comes home. A modern phased array is not one radio — it's a small radio behind every single element, replicated by the hundreds. Trace the receive path of one element and you'll meet old friends. The wave hits a tiny patch antenna etched in copper. The first thing it touches is a low-noise amplifier, placed right at the element because — as the noise-figure rung taught — whatever noise the first stage adds sets the noise figure of the entire chain, and you cannot recover from it downstream.
On transmit, each element drives its own power amplifier. This is profound: instead of one giant PA fighting to be linear and cool, you distribute the burden over hundreds of small PAs, each running comfortably — and their outputs combine in the air, in the beam, rather than in a lossy on-chip combiner. Between antenna and amplifier sits the matching network the whole track has hammered on: every element must present a clean impedance match, verified with S-parameters, or return loss steals your precious mm-wave power back as reflection.
- Per-element front-end. Patch antenna → T/R switch → LNA (receive) or PA (transmit), with a matching network on each side tuned and checked with S-parameters.
- Phase shifter + variable gain. Each path carries a digitally-set phase shift (the steering law's Δφ) and an amplitude trim, so the array can both aim the beam and shape its sidelobes.
- Combining / splitting. On receive the N phased signals are summed into one; on transmit one signal fans out to N. This may be analog (RF combining), digital (one ADC per element), or hybrid — the dominant 5G compromise.
- Beam controller. A processor computes the phase/amplitude weights for the desired beam direction and writes them to every element — re-aiming the entire array in well under a microsecond.
The systems that live or die by it
5G mm-wave. At 28 and 39 GHz there is glorious bandwidth — but the path loss is savage and a single human body can block the link. Beamforming is not a bonus here, it's the only thing that makes mm-wave work: the base station forms a tight beam aimed precisely at your phone, your phone forms a beam back, and as you walk the network re-steers both beams dozens of times a second to keep them locked. The array gain is exactly the term in the link budget that buys back the path loss and closes the link.
Automotive radar. The 77 GHz radar in a car is a phased array on a chip. It must, in a fraction of a second, sweep a fan of beams across the road and resolve a pedestrian from a lamppost a metre away at 150 metres' range — angular resolution that only a wide aperture of many elements can deliver. No motor would survive the vibration or the timing demands; electronic scanning is the only option.
Satellite & Starlink terminals. A low-Earth-orbit satellite crosses your sky in minutes, so a fixed dish can't track it and a motorized dish is bulky and slow. The flat Starlink "dishy" is a phased array of thousands of elements that electronically locks onto one satellite, then re-points to the next as the first sets — all with zero visible movement. The same physics powers AESA radars on fighter jets, which can track many targets while simultaneously searching, because the beam can be in a different place every microsecond.
Closing the track
Step back and look at what a phased array really is, and you'll see this entire track condensed into one object. The matching networks you learned to design now appear hundreds of times over. The S-parameters you measured characterize every element and every interconnect. The LNA sets the receive sensitivity; the PA sets the transmit power; noise figure still rules the front of the chain. The antenna gain and radiation pattern are no longer fixed properties of one part but something you compose, in real time, out of many — and the link budget is the scoreboard that tells you whether the whole orchestra closes the link.
That's the frontier. Beamforming is where the maths of phasors and interference, the craft of low-noise and high-power amplifiers, the discipline of matching and measurement, and the accounting of the link budget all converge — and it's why the same handful of fundamentals you've built up rung by rung underpin the most advanced radios on Earth and in orbit. Build the intuition that a beam is just timing, that gain is just agreement among many waves, and you can read any phased-array datasheet ever written.