About Aerial Patterns
A radar aerial sends its energy out in a beam and listens back through the same beam. How narrow the beam is, how much leaks out to the sides, and how many pulses land on a target as the aerial turns all come down to the aerial's size, measured in wavelengths, and to how its length is lit. This page is the recorder of an aerial test range: the aerial turns slowly in front of a distant probe and a pen draws what it hears against angle.
APERTURE: one straight aerial, such as the slotted waveguide of a ship's navigation radar, of the length you choose, on the band you choose. The taper is how strongly each part of the aerial is lit: UNIFORM lights it evenly; COSINE and COSINE² fade to nothing at the ends; TAYLOR holds the nearest sidelobes at the level you set; HAMMING is a raised cosine borrowed from signal processing. The upper chart is the pattern in decibels against angle, the polar chart the same in the half-space in front of the aerial. The test sheet compares the real 3 dB beamwidth with the rule of thumb 70 λ/D degrees.
ARRAY: a row of separate elements, each fed through a phase shifter. Giving each element a little more phase than its neighbour tilts the wavefront, so the beam points off to the side with nothing moving. The pattern is the array factor (blue dashes) times the pattern of one element (green dots). Spread the elements more than about half a wavelength apart and, when steered, copies of the main beam called grating lobes appear.
HITS PER SCAN: how many pulses strike a target each time the beam sweeps past it: the beamwidth in degrees × the PRF ÷ (6 × the rpm), since an aerial turning at 1 rpm turns 6° a second. SWEEP turns the aerial on the display, at true speed or slowed down so you can watch the pulses. The sheet works out, with Albersheim's equation, what adding the hits together is worth.
SIDELOBES: a large ship close aboard sends back so much that its echo also gets in through the aerial's sidelobes, and the display paints it on every bearing where a sidelobe points at it: arcs at the ship's range. Change the aerial, its taper, the target and the SEA control and watch the arcs. SLB adds a sidelobe blanker: a second, wide aerial whose signal is compared with the main one.
THE STORY
The link between an aerial's size and its beam is the same diffraction that limits a telescope: the narrowest beam an aperture can make is about one wavelength divided by its width, in radians. That is why the radars of the late 1930s, working at wavelengths of metres, needed huge aerials or broad beams, and why the cavity magnetron of 1940, giving real power at 10 cm and then 3 cm, let aircraft and small ships carry radars with narrow beams.
Tapering an aerial's lighting to trade a slightly wider beam for much lower sidelobes is an old idea, borrowed by aerial designers from the same mathematics as optics and filters. In 1946 C. L. Dolph found the distribution for an array that gives the narrowest beam for a chosen sidelobe level, with every sidelobe equal. In 1955 T. T. Taylor, at Hughes Aircraft, published the form for a continuous aperture that can actually be built: the first few sidelobes near the chosen level and the rest falling away. Taylor's distribution is still the starting point for most low-sidelobe radar aerials. Richard Hamming's window came from signal processing, and is used for aerials too.
Slotted waveguide aerials, rows of slots cut in the wall of a waveguide, each letting out a little of the wave travelling along it, became the usual horizontal aerial of ships' navigation radars: light, cheap, and the slot sizes set the taper.
Electronically steered arrays came into service in the 1960s and 1970s; the US Navy's AEGIS SPY-1, at sea from 1983, was among the first at sea in large numbers.
HOW IT WORKS
The pattern. For an aperture of length D lit by a(x), the far-field pattern is the Fourier transform of a(x) in u = (D/λ) sin θ. Each taper here is a sum of cosines across the aerial, so each pattern is a sum of shifted sinc functions, sin(πu)/(πu), worked out exactly. Even lighting gives the sinc itself: a 3 dB beamwidth of 0.886 λ/D radians (50.8 λ/D degrees) and a first sidelobe 13.26 dB down. A cosine taper widens the beam to 1.19 λ/D (68.1 λ/D degrees) and drops the sidelobes to 23 dB down; that is close to the 70 λ/D that radar people use, which is why the rule works for real, tapered aerials. The cost of a taper is gain: the aperture efficiency |∫a|² ÷ ∫|a|², 81% (0.91 dB) for the cosine.
Taylor. The pattern's nulls are moved so the first n̄ − 1 sidelobes sit near the chosen level; past n̄ the nulls are the uniform aperture's again. Too large an n̄ for a shallow sidelobe level and the lighting rises again towards the ends, which real aerials avoid; too small an n̄ for a deep level and the sidelobes just past the n̄-th stand above the level. The sheet says when either happens.
Arrays. With elements d apart and a phase step of 360° × (d/λ) × sin θ0 between neighbours, the waves from all elements add up in the direction θ0. They also add up wherever sin θ differs from sin θ0 by a whole number of λ/d: the grating lobes. They stay out of view while d/λ is below 1 ÷ (1 + |sin θ0|). The aerial looks shorter from the side, D cos θ0, so the steered beam is wider by about 1 ÷ cos θ0. The element pattern cos θ is that of an ideal element in a large array (it sees its own share of the aperture foreshortened); cos1.5 θ is often used to allow for the mismatch real arrays suffer at wide angles.
Hits and integration. Albersheim's 1981 equation gives the signal-to-noise ratio each pulse needs so that n pulses added together after the detector give a 90% chance of detection with a one-in-a-million chance of a false alarm; for one pulse it is about 13 dB. The green line is the ideal of adding the pulses in phase (coherently, as pulse-Doppler radars do), where n pulses are worth n times one. The range a radar sees a target at goes as the fourth root of its sensitivity, so a gain of 12 dB doubles it. The pulses near the edges of the beam come back weaker; radar range sums usually allow about 1.6 dB for this, and the page leaves it out.
Sidelobe echoes. The page uses a typical small ship's I band navigation radar (25 kW, a 0.08 µs short pulse, receiver noise figure 6 dB, 6 dB of losses) with the radar range equation. The echo comes in through the aerial twice, out and back, so two-way sidelobes are twice as many decibels down: a −30 dB sidelobe gives −60 dB. Far out and behind, small errors in any real aerial stop the sidelobes falling much below about −40 dB; the page holds them there. The display paints anything at least 8 dB above the receiver noise. SEA (sensitivity time control) turns the receiver's gain down close in. A sidelobe blanker compares the main aerial's signal with a wide guard aerial set a few dB above the main aerial's highest sidelobe, and blanks the display where the guard's is stronger; it is found mainly on military radars, against jamming as much as strong echoes, and it can blank a small wanted echo close to a strong one. A navigation radar relies on a low-sidelobe aerial and careful use of the gain and sea controls.
WHAT IT LEAVES OUT
Patterns are one-dimensional cuts through the beam, in the far field (at least 2D²/λ away; for an 8-foot I band aerial that is about 360 m). Elements are points (isotropic) unless you choose otherwise; there is no coupling between elements, no edge effects, no errors in the lighting except the fixed floor on the SIDELOBES tab, and no radome. Targets are steady (no fluctuation), the sea is flat and calm, and the ships, buoys and radar figures are typical, not any real vessel or set. Not for navigation or design.
FURTHER WATCHING AND READING