applets.

Radar Principles

← All apps

About Radar Principles

How a radar works, from the inside: the everyday sums of a radar engineer, with pictures of what they mean on the screen. Type into any box and everything that depends on it follows; each pair of boxes works both ways. Numbers can be typed with commas, as 2.5k or as 1e6, and every box has its own units.

PULSES: how often the radar sends a pulse (PRF), the gap between pulses (PRI), how long each pulse lasts and how strong it is, the radar's frequency and aerial width, and a target (pick one under TARGET) at some range: the page works out whether the radar can see it and where it shows. POWER: dBm, watts and dBW; dB and plain ratios; and a chain of gains and losses to add up. WAVES: frequency and wavelength, the NATO radar band (with the IEEE letters beside it for comparison), and the wave against a ruler. HORIZON: how far a radar can see over the curve of the Earth, and what the air does to that (pick the atmosphere under AIR; it feeds PULSES too). RESOLVE: two contacts close together, and whether the radar shows them as two or as one blip; drag the second contact on the close-up, or use the sliders, and FIX IT shows what would split each example. The scope on the left puts them together. RANGE − and RANGE + (or the − and + keys) pick its range scale, as on a real set: 0.25 to 96 nm with the usual rings for each, and 120 and 240 nm for air search; CLUTTER offers 1.5 to 12 nm. An echo beyond the scale is off the screen. Past the maximum unambiguous range (red ring, shaded beyond) nothing can show at its true range, because the next pulse has gone out, but a late echo still folds in closer as a ghost. There is no horizon ring: whether the target shows is decided by its echo, horizon and all. The dotted green ring is the detection range for the chosen target, and the blip is your target, where the radar would show it, brighter the more it has to spare. RUN turns the aerial at its own rate (24 times a minute for the ship's set, 6 for the long-range one) until PAUSE (RESUME carries on); left alone for 5 minutes it pauses by itself. (The 1937 station's aerial didn't turn; its picture turns 6 times a minute just to be drawn.) SILENT switches the transmitter off, as a ship may when it doesn't want to be found: a radar that transmits can be heard much further away than it can see. COPY LINK copies a link to the numbers you have set.

CLUTTER: a small boat (A) in the sea clutter close in and a fishing boat (B) under a rain shower further out, with the three controls a ship's radar operator has: GAIN, SEA and RAIN. Set the sea state, the wind and the rain, and pick I or F band; the two small screens show the same scene on each band (click one to use it). RUN turns the aerial for 40 turns (a turn every 2.5 seconds, as a ship's radar turns 24 times a minute) while the shower drifts downwind. NEW CHALLENGE hides the small boat somewhere new: set the controls, then click where you think it is (or focus the scope, move the cross with the arrow keys and press Enter).

Using a radar at sea (plotting other ships for their closest approach, the collision rules, a traffic separation scheme and a coast as the radar sees it) has its own app, Radar Navigator, built on the same sums. How a radar turns its plots into steady tracks, with alpha-beta and Kalman filters, is in Radar Tracker.

THE STORY

The idea of finding ships by bouncing radio waves off them is old: Christian Hülsmeyer patented his "telemobiloscope" in 1904 and showed it spotting ships on the Rhine, but nobody bought it. In the 1930s several countries worked on radar in secret, each on its own. In Britain, Robert Watson-Watt and Arnold Wilkins showed in February 1935 that a bomber flying near the BBC's short-wave transmitter at Daventry gave a clear echo, and a chain of coastal stations, Chain Home, followed. The first stations were working in 1937 and about twenty were ready when war came in 1939. They worked at about 20 to 30 MHz, waves 10 to 15 metres long, with huge wire aerials on masts over 100 metres tall, and sent 25 pulses a second, kept in step with the 50-cycle mains so that neighbouring stations didn't confuse each other.

The cavity magnetron, built by John Randall and Harry Boot in Birmingham in 1940, made powerful pulses on waves only 10 centimetres long, and with it radar sets small enough for aircraft and ships. The round map-like screen used here, the plan position indicator, came into use in the same years. Its amber glow is that of the long-afterglow phosphors many radar screens used: each echo has to stay lit until the trace comes round again. The word radar, short for "radio detection and ranging", was coined by the US Navy in 1940; Britain called it RDF until 1943. The letter bands began as wartime code names. Two sets are in use today: the IEEE's (Std 521: L, S, C, X…), common in civil and marine use, and the NATO (EU/ECM) bands A to M, used in military and electronic-warfare work. This page uses the NATO bands throughout.

This app grew from the Python scripts EW Converter.py (dBm, watts, dB and wavelength) and PRF PRI Converter.py. They use 3 × 108 m/s for the speed of light; this page uses the exact value, 299,792,458 m/s, which makes wavelengths and ranges 0.07% shorter.

HOW IT WORKS

Pulses. PRI = 1 ÷ PRF. A radar times each echo from the most recent pulse; radio waves travel at c, out and back, so an echo after time t means a range of c·t ÷ 2. An echo has to be home before the next pulse goes out, so the furthest range the radar can be sure of, the maximum unambiguous range, is c·PRI ÷ 2. An echo from further away arrives after the next pulse and is timed from that one instead: it shows at a range that is too short by a whole number of those ranges, a second-time-around echo. Radars catch these by changing their PRF now and then: a real target stays put, a ghost jumps.

The duty cycle is the share of the time the radar is transmitting, pulse width ÷ PRI, and the average power is the peak power times the duty cycle. Range resolution is c·τ ÷ 2 for a pulse τ long: two targets closer together than that send back echoes that overlap into one. It is also roughly the blind range close in, since the receiver is shut off while the pulse goes out; an echo arriving while a later pulse is being sent is lost too (eclipsed). A plain pulse τ long needs a bandwidth of about 1 ÷ τ.

Seeing a target. The radar range equation gives the echo's strength against the receiver's own noise, for one pulse: SNR = Pt·G²·λ²·σ ÷ ((4π)³·R⁴·k·T₀·B·F·L). Pt is the peak power; G the aerial's gain, here from the rule of thumb G ≈ 26,000 ÷ (horizontal beamwidth × vertical beamwidth, both in degrees), with the horizontal beam from 70λ ÷ D as in RESOLVE and the vertical beam set by each kind of set; λ the wavelength; σ the target's radar cross-section, the area of a perfect reflector that would send back as much; k·T₀ = 4 × 10⁻²¹ W per hertz, the noise of a receiver at room temperature; B the receiver's bandwidth, about 1 ÷ τ for a plain pulse, so a longer pulse carries more energy and reaches further; F the noise figure (3 to 5 dB for a microwave set); and L every other loss lumped together (8 to 12 dB here: waveguide, the beam's shape, processing). The echo fades as the fourth power of the range: twice as far is 12 dB weaker. As the beam sweeps past, a target gets a number of hits (beamwidth × PRF ÷ how fast the aerial turns), and adding them up means each one can be weaker. How strong each must be comes from Albersheim's equation, here for an 80% chance of seeing it with a one in a million chance of a false alarm (the IMO's radar standard asks for 8 scans in 10 with a false alarm chance of 1 in 10,000, so this page is a little stricter): 12.5 dB for a single steady pulse, about 4.5 dB each for ten. Real echoes fluctuate from sweep to sweep as the target rolls and turns (the Swerling case 1 model), which needs about 5.3 dB more.

Lobing. Over the sea the radar's waves reach the target two ways: straight, and bounced off the water, which turns the wave over. Where the two arrive in step the echo is up to 16 times (12 dB) what it would be in free space; where they arrive out of step it cancels: lobes and nulls. Both ways the factor is F⁴, with F = |2 sin(2π·h₁·h₂ ÷ (λ·R))| on a smooth sea (h₁ the aerial's height, h₂ the target's). Past the last lobe (its peak is at 4·h₁·h₂ ÷ λ), from about 12·h₁·h₂ ÷ λ out, where the echo is back to its free-space strength, the two cancel more and more and the echo falls as R⁻⁸ instead of R⁻⁴: that, not the horizon, is what limits low targets such as boats. The page cuts the heights down for the curve of the Earth (so the echo vanishes at the horizon) and weakens the bounce for rough water (waves 0.2 m rms, a slight sea) or rough land (about 1 m, which kills most of it, as under the airfield and long-range sets). The detection range is the outer edge of reliable detection: searching in from the horizon, the first range where the echo is strong enough, counting inside the lobes only the free-space echo; inside that edge the echo still fades in the nulls (the scope says "faded"), and on a lobe it can be seen further. For the navigation set this gives a 1 m² small boat at about 3.3 nm and a large ship at about 16 nm, and it does a little better than the minimum ranges in the IMO's radar standard for a 15 m aerial, as a 25 kW set should: a 10 m boat with no reflector (about 2.5 m², 2 m up) at 4.8 nm against the standard's 3.4, a navigation buoy at 6.0 against 4.6, a ship over 5,000 gross tonnage (10 m up) at 13.8 against 11.

The figures are typical, not those of any real set, and PULSES leaves out rain and clutter (see CLUTTER), which cut ranges further. The horizon and the air are part of the same sums (see Horizon). The targets' radar cross-sections are rough typical microwave values: a small boat 1 m², a yacht 3 m², a fishing boat 30 m², a merchant ship 10,000 m², a fighter 2 m², an airliner 30 m². A real echo swings by ten times or more as the target turns, and differs from band to band. Each target also sets a typical height for HORIZON and for lobing: 1, 4, 5 and 18 m for the boats and the ship, 6,000 m for the fighter and 10,000 m for the airliner. The 1937 station's figures are approximate: its aerial floodlit a wide arc (taken here as a beam about 100° wide), so the gain rule is rough there, and it listened on separate aerials, while the sums assume one aerial sends and listens.

Power. Decibels compare two powers: dB = 10 × log10(ratio), so every 10 dB is ×10 and every 3 dB is about ×2. dBm is decibels compared with 1 milliwatt (0 dBm = 1 mW, 30 dBm = 1 W), dBW compared with 1 watt. Gains and losses in a chain just add in dB. For voltages, which go as the square root of power, the ratio is 10dB ÷ 20.

Bands. NATO bands: A 0 to 250 MHz, B 250 to 500 MHz, C 0.5 to 1 GHz, D 1 to 2, E 2 to 3, F 3 to 4, G 4 to 6, H 6 to 8, I 8 to 10, J 10 to 20, K 20 to 40, L 40 to 60 and M 60 to 100 GHz. The IEEE letters cover the same spectrum differently: a ship's 9.41 GHz radar is IEEE X band but NATO I band, a 3.05 GHz one IEEE S but NATO F, an airfield radar at 2.8 GHz IEEE S but NATO E, and 1.3 GHz IEEE L but NATO D. The same letter can mean different things: NATO L (40 to 60 GHz) is nowhere near IEEE L (1 to 2 GHz), and NATO K (20 to 40 GHz) only partly overlaps IEEE K (18 to 27 GHz). WAVES shows the IEEE letters under the NATO bar for comparison; everywhere else the letters are NATO's. Picking a band jumps to a typical radar frequency in it.

Waves. Wavelength = c ÷ frequency. The ruler is drawn at real size using the browser's own idea of a centimetre, which is close on most desktop screens but not exact on all of them.

Horizon. Over a smooth Earth of radius R a height h sees to √(2Rh). In a normal atmosphere radio waves bend gently downwards, following the Earth a little; engineers allow for it by drawing the waves straight over an Earth 4/3 the real size. With R = 6,371 km that gives d ≈ 2.23 × √h nautical miles for h in metres (4.12 × √h km), and the radar and the target each add their own: d ≈ 2.23 (√h1 + √h2). A nautical mile is 1,852 metres; this page gives distances in nautical miles and speeds in knots, as ships and aircraft do.

The radio horizon is not a wall. Past it the waves creep on round the Earth by diffraction, getting weaker quickly and smoothly (the smooth-Earth diffraction of ITU-R P.526, its first term): about 2 dB a kilometre one way at I band, less at longer waves. So a target just past the horizon may still be seen if it is big; the sums decide, not a line. The air changes all this. SUB-REFRACTION (an Earth only 0.9 times its size) bends the waves less and shortens ranges; SUPER-REFRACTION (3 times) bends them more and stretches them. A duct is a layer where the air bends the waves back down faster than the Earth curves away, trapping waves shorter than a cut-off of about 2.5·d·√(ΔM × 10⁻⁶) for a duct d deep with a refractivity deficit ΔM. The EVAPORATION DUCT forms in the damp air just over the sea, typically 10 to 20 m deep (15 m here); it traps waves above about 1.8 GHz, best at I band, so a ship's radar low down in it sees ships and coast far past its usual range, while aircraft above it are unaffected. The SURFACE DUCT here comes from a trapping layer 150 to 250 m up (warm dry air over cool sea); it traps almost every radar band and gives very long, anomalous ranges low down, with a radar hole just above it where less energy reaches. Inside a duct the trapped waves spread in one dimension, not two, against free space gaining as range ÷ a reference (12 km for the evaporation duct, 25 km for the surface one) while leaking a little (0.04 and 0.06 dB/km one way, more near the cut-off). This is an honest simplification, not a propagation forecast: real ducts vary hour by hour and place by place. The side view is a coverage diagram: height above the sea (on a square-root scale, so low layers show) against range, with the Earth's curve taken out, each spot shaded by how strong the signal is there against free space, a duct's layer tinted, and a few rays traced through the air: over a flat sea they seem to curve up as the Earth falls away; in a duct they bend back down and are trapped.

Resolve. A radar can only tell two contacts apart if their echoes don't run into each other. In range, each echo lasts as long as the pulse, so it fills c·τ ÷ 2 of range (150 m for a 1 µs pulse): two contacts on the same bearing closer than that come back as one longer echo. That depth is the same at every range. In bearing, a contact keeps echoing for as long as the beam is sweeping across it, so on the screen each echo is smeared across about one beamwidth; two contacts at the same range closer in bearing than that merge into one. A rule of thumb gives the beamwidth: about 70 × λ ÷ D degrees for an aerial D wide at wavelength λ (the 70 is rough: it runs from about 50 to 75 depending on how the aerial is fed, and the rule only holds when the aerial is many wavelengths wide). So a wider aerial or a shorter wave gives a narrower beam. A beam of fixed angle gets wider in metres the further out it looks, R × the beamwidth in radians, so bearing resolution gets worse with range while range resolution doesn't. Two contacts far enough apart in either range or bearing are seen as two. The close-up is drawn like a B-scope, range up and bearing across, each stretched to fit, and each echo is drawn as a block the size of its smear; real echoes have softer edges, and the display and the operator's eye blur them a little more, so these are best-case figures.

A short pulse carries little energy, so long-range radars use pulse compression: a long pulse swept in frequency or coded in phase is squeezed in the receiver into a short spike, giving the range resolution of a pulse about 1 ÷ B long (c ÷ 2B of range for a bandwidth B) with the energy of the long one. This page doesn't model it beyond the note in the FORMATION example. Radars can also split contacts by their speed (Doppler), which isn't shown here either.

Clutter. The sea sends back echoes of its own: strongest close in, where the beam looks down on the waves more steeply, and stronger looking into the wind, where the radar sees the faces of the waves, than looking down it. Rougher seas send back more, and further out. Sea clutter flickers from one turn to the next, with brief bright sea spikes, while a boat's echo stays put, so watching a few turns helps. Rain echoes too, about 20 dB more at I band (3 cm waves) than at F band (10 cm): the echo from raindrops goes as the fourth power of the frequency. At I band rain also soaks up some of the waves passing through it, so things behind and inside a shower come back weaker. GAIN sets how faint an echo can be and still show: more brings up small echoes, and noise and clutter with them. SEA (sensitivity time control) turns the gain down close in, less and less further out (40 dB less for each tenfold range), to follow the way sea clutter fades with range; too much and the small boat goes with the clutter. RAIN (fast time constant) passes only changes in the echo along each line out from the middle, so a long smooth echo like rain shrinks to its front edge while a short hard echo like a boat stays. No control helps when a boat's echo is no stronger than the clutter in the same place.

The numbers on this tab (dB above the receiver's noise) come from a typical ship's radar worked through the range equation: I band 25 kW with a 1.9 m aerial (a 1.2° beam), F band 30 kW with a 3.7 m aerial (1.9°), both 15 m above the sea on a 0.25 µs pulse. The sea's reflectivity comes from the NRL sea clutter model (Gregers-Hansen and Mital, 2012), a fit to decades of measurements, for the horizontal polarisation ships' radars use; times the patch of sea in each cell, it gives I-band clutter out to about 1½ nm in a slight sea, 3 nm in a moderate one and 4 to 5 nm in a very rough one, about 6 dB stronger looking into the wind than down it (the GIT model's wind factor). At F band the sea reflects about 10 dB less; the bigger aerial and wider beam claw some back, so on the screen F-band clutter is a few dB weaker and stops sooner. Sea clutter is spiky: its strength follows a K-distribution, so a few patches flash up well above the rest (the sea spikes), each smeared across the beam. The shower is a few overlapping rain cells 2 to 3 nm across with a heavier heart, 4 mm an hour (a shower) or 25 mm an hour (heavy rain) at the heart, drifting downwind at 25 knots. Its echo comes from Z = 200·R1.6 (Marshall and Palmer) and the beam's volume, about 20 dB weaker at F band; heavy rain at I band also soaks up about 2 dB for each nautical mile out and back (ITU-R P.838), hardly any at F band. The small boat A is 1 m² and 1 m up, the fishing boat B 30 m² and 5 m up at I band, a little less at F band (radar reflectors and small craft send back less there). Both lose out to lobing past the last lobe, which comes much sooner on F band's longer waves: one reason I band is better for small, low targets. These are typical values, not any particular set: real clutter depends on much more (swell, the set's own processing), and on the day an operator may see less or more.

FURTHER WATCHING AND READING

Tracking contacts → Radar Tracker · Navigating with radar → Radar Navigator

Try
Target
m