How Radar Actually Works: Pulses, Doppler, and Phased Arrays

Point a flashlight at a wall in the dark and you see a spot of light. Now imagine the flashlight could measure how long the light took to bounce back, and how the wall’s motion stretched or squeezed the returning wave. That, in essence, is how radar works: it shouts a radio pulse into empty space, listens for the faint echo, and reads two numbers out of it — the round-trip time, which gives distance, and the frequency shift, which gives speed. Everything else is engineering built on those two ideas.

Radar feels like magic because the numbers are extreme. A pulse leaves the antenna, travels tens of kilometres, scatters off a metal target, and comes back so weakened that the received power can be a million-billion times smaller than what was sent. And yet an air-traffic radar reliably picks an airliner out of that whisper, tells you it is 74 km away and closing at 250 metres per second, all before the next pulse goes out microseconds later.

What this covers: the physics of pulse timing and range, the radar equation and why range falls off as the fourth power, range resolution and ambiguity, the Doppler shift for velocity, FMCW and pulse compression, phased-array beam steering, radar cross section and stealth, and how signal processing pulls targets out of clutter.

Context and Background

Radar — RAdio Detection And Ranging — grew out of the late-1930s race to detect aircraft before they were visible or audible. Britain’s Chain Home network of tall transmitter masts gave defenders minutes of warning during the Battle of Britain, and the wartime invention of the cavity magnetron let radars shrink to microwave wavelengths small enough to fit on aircraft and ships. Eight decades later the same physics runs on a silicon chip behind your car’s bumper.

The reason radar endures is that radio waves do what light cannot: they pass through fog, cloud, rain, dust, and darkness, and they carry their own illumination. A camera needs the sun; a radar brings its own flashlight and measures the return with clock-grade precision. That independence from ambient conditions is why radar guides aircraft, tracks storms, maps the surface of Venus, keeps ships off rocks, catches speeders, and now lets a driver-assistance system brake for a stopped truck in heavy rain when the cameras are blind.

Radar sits alongside two cousins that trade radio for other signals. Lidar swaps radio for laser light, buying far finer resolution at the cost of weather sensitivity — the physics of that trade is covered in how lidar actually works. GPS runs the timing trick in reverse, listening to signals from satellites rather than sending its own, and it leans on relativity to keep its clocks honest, as explained in how GPS actually works. Radar is the active member of the family: it both talks and listens. For the standards-level definition of the frequency bands radars use, the IEEE Standard 521 letter designations are the canonical reference.

The Core Idea: Time of Flight and the Radar Equation

At heart a radar is a stopwatch and a very sensitive ear. It emits a short burst of radio energy, then measures the delay until the echo returns. Because radio waves travel at the speed of light, c ≈ 3 × 10⁸ metres per second, and because the echo makes a round trip out to the target and back, the range is R = c · t / 2. The factor of two is the whole point: the pulse covers twice the distance to the target, so you halve the measured time to get one-way range.

The numbers are quick to work. Light travels roughly 300 metres per microsecond, so a target at 15 km is 30 km round trip, and the echo arrives 30,000 m ÷ (3 × 10⁸ m/s) = 100 microseconds after the pulse left. A handy rule of thumb: every microsecond of round-trip delay corresponds to 150 metres of range. If your receiver clocks a 493-microsecond delay, the target sits at 493 × 150 = 73,950 metres, or about 74 km. Radar is, at bottom, an exercise in measuring time extremely well.

Figure 1: The signal chain of a monostatic radar, where one antenna both transmits and receives.

The waveform generator produces the pulse, a power amplifier boosts it, and a duplexer switch connects the amplifier to the antenna during transmission, then flips to route the faint echo into the low-noise amplifier during listening. The receiver downconverts, digitises, and matched-filters the echo before detection logic extracts range and velocity.

Why received power falls off as 1/R⁴

The cruelty of radar is buried in one exponent. When you double the distance to a target, the echo does not get four times weaker — it gets sixteen times weaker. The received power obeys the radar equation:

P_r = (P_t · G² · λ² · σ) / ((4π)³ · R⁴)

Here P_t is transmitted power, G is antenna gain, λ is wavelength, σ is the target’s radar cross section, and R is range. The 1/R⁴ term is the killer, and it comes from applying the inverse-square law twice. The transmitted wave spreads over a sphere on the way out, so the power density at the target falls as 1/R². The target scatters some of that energy, and the scattered wave spreads over another sphere on the way back, falling as 1/R² again. Two inverse-square legs multiply into 1/R⁴.

The practical consequence is brutal. To detect a target at twice the range, all else equal, you need sixteen times more transmitted power — or you must claw back that factor through antenna gain, integration time, or a larger effective aperture. This is why long-range radars have enormous dishes or arrays and megawatt-class transmitters, and why doubling a radar’s reach is a far bigger engineering leap than it sounds. It also explains why a stealthy target that halves its radar cross section only cuts detection range by about 16%, not by half — range scales as the fourth root of the cross section.

Antenna gain and aperture

Gain is the radar’s way of not shouting in every direction at once. An antenna with gain G concentrates energy into a narrow beam, and that same directivity focuses the receiver’s sensitivity on the same patch of sky. Gain appears squared in the radar equation precisely because the antenna helps twice — once on transmit, once on receive. A larger physical aperture yields a narrower beam and higher gain for a given wavelength, which is why big antennas both reach farther and locate targets more precisely in angle.

Pulses, Resolution, and Ambiguity

A radar’s pulse is not just a trigger for the stopwatch; its shape sets what the radar can and cannot distinguish. Two properties matter most: the pulse width, which controls how finely the radar separates targets in range, and the pulse repetition frequency, which sets how far the radar can see before echoes start to alias.

Range resolution — the minimum spacing at which two targets show up as two blips rather than one smear — is set by the pulse width τ. Two echoes overlap unless the second target is far enough behind the first that its echo arrives after the first echo has finished. The resolution is ΔR = c · τ / 2. A one-microsecond pulse gives 3 × 10⁸ × 1 × 10⁻⁶ ÷ 2 = 150 metres of resolution: two aircraft closer than 150 metres in range merge into one return. Want to resolve targets one metre apart? You need a pulse about 6.7 nanoseconds long.

Figure 2: A single pulse’s life — leave, scatter, return after delay Δt, and the receiver’s window closes before the next pulse.

The figure traces one pulse from transmission to echo and shows the timing window. The receiver must catch the echo before the next pulse fires, and any echo that arrives after that window is misassigned to the following pulse — the origin of range ambiguity.

The resolution paradox and pulse compression

Here is the tension. Short pulses give fine range resolution but carry little energy, so they detect poorly — a 6.7-nanosecond pulse simply does not put enough joules on a distant target. Long pulses carry lots of energy and detect well but blur range resolution to hundreds of metres. For decades this looked like an unavoidable trade: you could have reach or resolution, not both.

Pulse compression breaks the deadlock. Instead of a plain pulse, the radar transmits a long pulse whose frequency sweeps across a band — a chirp, like the rising whistle of a swept tone. The pulse is long, so it carries plenty of energy. On receive, a matched filter delays the low-frequency start of the chirp more than the high-frequency end, compressing the long echo into a sharp spike. The effective resolution is now set by the swept bandwidth B, not the physical pulse length: ΔR = c / (2B). A chirp sweeping 150 MHz yields one metre of resolution while the transmitted pulse might be tens of microseconds long. Modern radars almost universally use pulse compression; a plain unmodulated pulse is now the exception.

PRF, maximum unambiguous range, and aliasing

The pulse repetition frequency (PRF) is how many pulses the radar fires per second, and it sets a hard ceiling on range. After sending a pulse, the radar must wait for the echo before it can trust the next one. If it fires again too soon, an echo from a distant target can arrive after the second pulse went out, and the radar will mistakenly time it against that second pulse — reporting a far target as if it were close. This is range aliasing, exactly analogous to how a wagon wheel appears to spin backwards in a film.

The maximum unambiguous range is R_max = c / (2 · PRF). A radar running at 1,000 pulses per second has R_max = 3 × 10⁸ ÷ (2 × 1,000) = 150 km: any target beyond 150 km folds back and appears at a false, shorter range. Lower the PRF and you see farther but fire fewer pulses, weakening detection and, as we will see, worsening velocity measurement. Radars often stagger the PRF between values so that the true range stays consistent while the aliased ghosts jump around, letting the processor identify and discard them.

The Doppler Shift: Measuring Velocity

Range tells you where a target is; Doppler tells you how fast it is coming or going. When a target moves toward the radar, each successive wave crest of the returning echo is emitted from slightly closer, bunching the crests together and raising the frequency. A receding target stretches the crests and lowers the frequency. This is the same effect that makes an ambulance siren drop in pitch as it passes you — the physics is identical, only the wave is radio rather than sound.

Because the wave makes a round trip, the shift is doubled compared with a one-way source. The Doppler frequency is f_d = 2 · v_r / λ, where v_r is the target’s radial velocity (its speed component along the line of sight) and λ is the radar wavelength. Only the radial component counts: a target crossing directly in front of the radar, neither approaching nor receding, produces zero Doppler even at high speed — a real blind spot that pure Doppler systems inherit.

Put in numbers. An X-band radar at 10 GHz has λ = c/f = 0.03 metres. A target closing at 300 m/s produces f_d = 2 × 300 ÷ 0.03 = 20,000 Hz — a 20 kHz shift on a 10 GHz carrier, a fractional change of two parts in a million, yet trivially measurable because the radar compares many pulses over time. A police radar reads your speed the same way, just at a shorter range and a different band.

MTI and pulse-Doppler processing

The reason Doppler matters operationally is that it separates movers from the stationary world. Ground, buildings, sea, and hills reflect enormous echoes — collectively clutter — that can dwarf the return from an aircraft. But clutter is stationary, so its Doppler shift is (nearly) zero, while a moving target is not.

Moving Target Indication (MTI) exploits this by subtracting successive pulse returns from the same range. Whatever did not move cancels; whatever did move survives. Pulse-Doppler radar goes further, running a bank of filters (a Fourier transform across pulses) that sorts returns by their Doppler frequency, so the processor sees not just “something moved” but how fast. This is how an airborne radar spots a low-flying target against the ground beneath it — the target lives in a different Doppler bin from the terrain.

There is a catch that mirrors range ambiguity. Just as range aliases beyond R_max, velocity aliases beyond a maximum unambiguous velocity set by the PRF: v_max = λ · PRF / 4. Here the trade-off bites hard. A high PRF gives wide unambiguous velocity but short unambiguous range; a low PRF does the reverse. You cannot have both, which is why radars switch PRF modes or interleave several to resolve range and velocity together.

Waveforms: Pulsed, CW, and FMCW

Not every radar pulses. The choice of waveform is a design decision driven by what the radar needs to measure and how cheap it must be.

A pure continuous-wave (CW) radar transmits an unbroken tone and never stops listening. It cannot measure range at all — with no timing mark on the wave, there is no round-trip delay to clock — but it reads Doppler beautifully, because comparing the returned frequency with the transmitted frequency directly yields velocity. This is exactly what a simple police speed gun and a Doppler weather sensor do: they only care about speed, so they skip range entirely.

Frequency-Modulated Continuous-Wave (FMCW) is the clever hybrid, and it is why your car can have radar. An FMCW radar transmits continuously but sweeps its frequency up in a ramp — a chirp again. At any instant, the echo coming back was transmitted a moment ago, when the frequency was lower, so mixing the echo with the currently transmitted frequency produces a beat frequency proportional to the round-trip delay, and hence to range. FMCW measures range and Doppler at low peak power, because energy is spread continuously rather than crammed into a brief pulse. Low peak power means cheap, compact solid-state transmitters — no magnetron, no high-voltage pulse hardware.

That is precisely the trade automotive radar wants. Modern car radars operate in the 76–81 GHz millimetre-wave band, with the 76–77 GHz slice harmonised globally and the 77–81 GHz slice used for high-resolution short-range sensing, per single-chip transceiver datasheets such as Texas Instruments’ AWR-series FMCW radar-on-chip. At 77 GHz the wavelength is under 4 mm, so a whole antenna array fits behind a plastic bumper, and the wide 4 GHz sweep bandwidth available in the 77–81 GHz band delivers centimetre-class range resolution. FMCW plus mmWave gives a car cheap, all-weather, high-resolution sensing that no camera matches in fog.

Phased Arrays: Steering the Beam Without Moving the Dish

A classic radar points its beam by physically rotating a dish. That is slow, mechanically fragile, and can only look one way at a time. A phased array radar replaces the single dish with hundreds or thousands of small radiating elements and steers the beam electronically, in microseconds, with no moving parts.

A phased array steers by exploiting interference. Each element radiates the same signal, but the radar delays — or phase-shifts — the signal to each element by a controlled amount. If every element fires in step, the combined wavefront goes straight out the front. If the radar applies a progressive phase delay across the array, the wavefront tilts, and the beam points off-axis. Change the phase pattern and the beam swings to a new angle before the next pulse, letting one array track dozens of targets in different directions on a time-shared basis.

Figure 3: Progressive per-element phase shifts tilt the combined wavefront, steering the beam electronically without moving the antenna.

The figure shows a common feed splitting the signal to four elements, each with its own phase shifter. A linearly increasing phase across the elements tilts the wavefront and points the beam; the steering angle θ satisfies sin θ = (phase step × λ) / (2π × element spacing).

PESA versus AESA

Phased arrays come in two families. A Passive Electronically Scanned Array (PESA) uses one central transmitter feeding all elements through phase shifters — simpler and cheaper, but with a single point of failure and higher losses. An Active Electronically Scanned Array (AESA) puts a tiny transmit/receive (T/R) module behind every element, so each element has its own amplifier and phase control. AESA is the modern gold standard: it is more sensitive (thousands of small amplifiers beat one big one), fails gracefully (losing a few T/R modules only slightly degrades performance), can form multiple beams at once, and is harder to jam because it spreads energy across many frequencies and directions. Fighter radars, weather radars, and 5G base stations all lean on AESA-style beamforming.

Grating lobes: the sampling trap

Phased arrays have a failure mode borrowed straight from sampling theory. If the elements are spaced too far apart — more than about half a wavelength for wide steering — the array produces grating lobes: full-strength copies of the main beam pointing in unwanted directions. These are spatial aliases, the exact analogue of a Nyquist-violation ghost in a digital signal, and they let the radar transmit energy where it does not intend and receive echoes it cannot place. Keeping element spacing at or below λ/2 suppresses them, which is one reason mmWave arrays need so many closely packed elements — the spacing shrinks with the wavelength.

Radar Cross Section and Stealth

The target gets a vote in whether it is seen. Radar cross section (RCS), symbol σ, measured in square metres, quantifies how much of the incident radar energy a target scatters back toward the receiver. It is not the physical size of the object — a smooth metal sphere and a jagged corner reflector of the same frontal area can differ in RCS by orders of magnitude. RCS depends on shape, material, orientation, and the radar’s wavelength.

Stealth is the art of shrinking σ, and it attacks the radar equation directly. Because received power scales with σ, halving RCS cuts detection range by only the fourth root — about 16% — so real stealth demands dropping RCS by factors of thousands, not two, to meaningfully shorten the range at which a target is detected. Designers use two main tools.

Shaping angles every flat surface and edge so that incident radar energy reflects away from the transmitter rather than straight back. The faceted look of early stealth aircraft and the smooth blended chines of later ones both serve this: a monostatic radar, which receives where it transmits, sees almost nothing if the geometry throws the specular reflection off to the side. Corners and right angles are the enemy, because a right-angle corner (a corner reflector) bounces energy straight back regardless of angle — which is why some targets want high RCS and bolt on corner reflectors, and stealth designs obsessively avoid them.

Radar-absorbent material (RAM) coats the surface with lossy substances that convert radio energy into heat instead of reflecting it. RAM is tuned to specific bands, so a shape optimised against X-band fighter radars may be more visible to a long-wavelength L-band or VHF radar, whose wavelength is comparable to whole airframe features and can excite resonances that shaping cannot fully suppress. Stealth is therefore always relative to a threat band, never absolute.

Trade-offs, Gotchas, and What Goes Wrong

Radar’s elegance hides a thicket of compromises, and most operational failures trace back to one of them. The range-velocity ambiguity trade is the deepest: a PRF that gives unambiguous velocity blinds you to range, and vice versa, so no single PRF is right for everything. Multi-PRF schemes resolve the conflict at the cost of processing complexity and dwell time.

Clutter is the everyday nemesis. Ground, sea, rain, and even flocks of birds throw back echoes that can bury a real target. The tangential blind spot is nastier still: any target moving exactly across the line of sight has zero radial velocity and therefore zero Doppler, so an MTI or pulse-Doppler radar that rejects zero-Doppler returns can lose it entirely at the worst moment.

Figure 4: The detection pipeline — cancel stationary clutter, sort by Doppler, then set an adaptive CFAR threshold before declaring a target.

The figure shows how raw echoes flow through clutter cancellation, a Doppler filter bank, and a Constant False Alarm Rate (CFAR) detector that raises or lowers its threshold based on the noise in neighbouring range cells — so a target must stand out from its local background, not a fixed level. A fixed threshold would either drown in clutter or miss weak targets; CFAR adapts.

Multipath is another trap. Energy reflecting off the ground or sea can arrive at the target and receiver by two paths, creating interference nulls where a target vanishes and false elevations where it appears to hover. And weather cuts both ways: rain that a weather radar wants to see is attenuation and clutter to an air-defence radar, especially at high bands like Ka and W where atmospheric absorption climbs steeply. The band choice — L and S for long-range search, X for tracking and fire control, Ka and W for short-range high-resolution sensing — is always a compromise between reach, resolution, antenna size, and weather penetration.

Practical Recommendations

If you are reasoning about a radar system — designing, buying, or just interpreting one — start from the physics rather than the spec sheet. The radar equation tells you almost everything about reach: if a vendor claims double the range of a rival at the same power and antenna, be skeptical, because that demands a 16× advantage somewhere. Ask where it comes from — aperture, integration time, or lower noise.

Match the waveform to the job. If you need range and velocity at low cost and short range, FMCW at mmWave is the automotive-proven answer. If you need long reach, pulsed with pulse compression buys resolution without sacrificing energy. If you only care about speed, CW is cheap and simple.

• Pick the band for the mission: L/S for long-range search, X for tracking, Ka/W for short-range high resolution — and remember weather attenuation rises with frequency.
• Choose PRF deliberately: high for velocity, low for range, staggered or multiple to resolve both.
• Use pulse compression to escape the reach-versus-resolution deadlock; think in swept bandwidth, not pulse length.
• Prefer AESA where budget allows, for graceful degradation, multi-beam operation, and jam resistance.
• Keep element spacing ≤ λ/2 in any phased array to avoid grating lobes.
• Design detection around CFAR so thresholds adapt to local clutter, and watch the zero-Doppler blind spot.

Frequently Asked Questions

How does radar measure distance?

Radar measures distance by timing how long its echo takes to return. It transmits a radio pulse, then clocks the delay until the reflected echo comes back. Because radio waves travel at the speed of light and the pulse makes a round trip, range is R = c · t / 2. Every microsecond of round-trip delay corresponds to about 150 metres of range. The whole method is a precise stopwatch: measure the delay accurately and you know the distance accurately.

How does radar measure speed?

Radar measures speed using the Doppler effect. A target moving toward the radar compresses the returning wave to a higher frequency; a receding target stretches it to a lower one. The frequency shift is f_d = 2·v_r/λ, where v_r is the radial velocity and λ the wavelength. Only motion along the line of sight counts, so a target crossing directly in front shows zero Doppler even at high speed — a genuine blind spot for pure Doppler systems.

Why does radar range fall off so fast with distance?

Because the inverse-square law applies twice. The transmitted wave spreads over a sphere on the way out, weakening as 1/R², and the echo scattered by the target spreads over another sphere on the way back, weakening as 1/R² again. Together they give 1/R⁴ in the radar equation. The practical result is that seeing twice as far requires sixteen times more power or an equivalent gain elsewhere, which is why long-range radars are so large and powerful.

What is a phased array radar and why is it better?

A phased array steers its beam electronically by phase-shifting the signal to each of many small antenna elements, tilting the combined wavefront without physically moving the antenna. This lets it switch beam direction in microseconds and track many targets at once. Active arrays (AESA) put an amplifier behind every element, giving high sensitivity, graceful failure, multiple simultaneous beams, and strong resistance to jamming — advantages a single rotating dish cannot match.

How does stealth defeat radar?

Stealth shrinks a target’s radar cross section using shaping and absorbent materials. Shaping angles surfaces so incident energy reflects away from the transmitter rather than back to it, and radar-absorbent material converts radio energy into heat. Because detection range scales as the fourth root of RCS, stealth must cut cross section by huge factors to matter. It is also band-specific: a shape tuned against high-frequency fighter radars may still show up on long-wavelength VHF or L-band search radars.

Why do cars use 77 GHz FMCW radar instead of pulsed radar?

Automotive radar needs range and velocity, all-weather operation, low cost, and a tiny antenna. FMCW delivers range and Doppler at low peak power, so it runs on cheap solid-state chips with no high-voltage pulse hardware. At 77 GHz (within the 76–81 GHz band) the wavelength is under 4 mm, so a full antenna array fits behind a bumper, and the wide sweep bandwidth gives centimetre-class resolution — a combination pulsed low-frequency radar cannot match in that form factor.

Further Reading

• How lidar actually works: the physics of laser ranging — radar’s optical cousin, trading weather robustness for far finer resolution.
• How GPS actually works: timing, trilateration, and relativity — the same time-of-flight idea run in reverse, listening instead of transmitting.
• How OLED displays actually work: the physics of self-emissive pixels — another first-principles teardown of everyday technology.
• IEEE Standard 521 radar band letter designations — the canonical L/S/C/X/Ku/K/Ka/W band definitions.
• Texas Instruments AWR-series FMCW automotive radar-on-chip datasheet — a real 76–81 GHz single-chip transceiver and its parameters.

By Riju — about