Cosine Taper
How Cosine Illumination Shapes the Pattern
A cosine taper describes the amplitude distribution applied across the length of a continuous aperture or, in discrete form, across the elements of an array. Along a one-dimensional aperture of length L centered at the origin, the illumination is A(x) = cos(πx/L) for |x| ≤ L/2. The weighting equals 1.0 at the aperture center and falls smoothly to zero at both edges. Because the radiated far-field pattern is the Fourier transform of this aperture distribution, tapering the edges down suppresses the abrupt amplitude discontinuity that a uniform aperture presents, and that discontinuity is precisely what generates the high -13.3 dB sidelobes of a uniformly illuminated aperture.
The cosine distribution has a closed-form transform, so its pattern properties are exact rather than numerical. The first sidelobe settles at about -23 dB below the main-beam peak, nearly 10 dB lower than a uniform aperture, and the sidelobe envelope rolls off as 1/n3 with the angular index n, roughly 12 dB per octave faster than the uniform 1/n decay. This rapid far-sidelobe roll-off is why cosine illumination is favored where distant clutter returns or out-of-beam interferers must be rejected. The penalty appears in the main beam: the half-power beamwidth widens by a factor near 1.34, and the taper (illumination) efficiency drops to about 0.81, equivalent to roughly 0.91 dB of directivity loss versus a uniform aperture of identical size.
In practical array design the continuous cosine law is sampled at the element positions, so each element feed network or attenuator setting realizes the cosine weight for that location. Quantization of those weights into discrete attenuator steps, plus element-to-element amplitude and phase errors, raises the achieved sidelobe floor above the ideal -23 dB; 5 to 8 percent RMS amplitude error typically limits real arrays to the -18 to -21 dB range. Designers who need a controllable sidelobe floor often move to a cosine-on-a-pedestal or a Taylor distribution, both of which start from the cosine family but add a pedestal term to set the sidelobe level explicitly.
Cosine Aperture Equations
A(x) = cos(πx / L), |x| ≤ L/2
Far-field pattern (u = (πL/λ)·sinθ):
F(u) ≈ (π/2)² × cos(u) / [(π/2)² − u²]
Half-power beamwidth:
θ3dB ≈ 1.19 × (λ / L) rad
Taper (aperture) efficiency:
ηt = |∫A(x)dx|² / (L × ∫|A(x)|²dx) ≈ 0.81
Where x = position across the aperture, L = aperture length, λ = wavelength, θ = angle from boresight, u = normalized pattern variable. First sidelobe ≈ -23 dB; far sidelobes decay as 1/n3.
Aperture Taper Comparison
| Taper | Illumination A(x) | First Sidelobe | Beam Broadening | Taper Efficiency | Far Roll-off |
|---|---|---|---|---|---|
| Uniform | 1 | -13.3 dB | 1.00x | 1.00 | 1/n |
| Cosine | cos(πx/L) | -23 dB | 1.34x | 0.81 | 1/n3 |
| Cosine² (Hann) | cos²(πx/L) | -32 dB | 1.63x | 0.67 | 1/n4 |
| Hamming | 0.54 + 0.46cos(2πx/L) | -43 dB | 1.47x | 0.73 | 1/n |
| Taylor (n̄=5) | set by SLL | -25 to -40 dB | 1.10 to 1.30x | 0.85 to 0.93 | level floor |
Frequently Asked Questions
What first sidelobe level does a cosine taper produce?
A pure cosine illumination yields a first sidelobe near -23 dB versus -13.3 dB for a uniform aperture, with the far-sidelobe envelope decaying as 1/n3, about 12 dB per octave faster than uniform. That is why it is a common base for low-clutter radar and low-interference comms antennas. For deeper nulls, cosine-squared reaches about -32 dB, and cosine-on-a-pedestal or Taylor distributions let you set the sidelobe floor directly.
How much does a cosine taper widen the beam and reduce gain?
Versus a uniform aperture of the same size, the half-power beamwidth broadens by about 1.34x and taper efficiency falls to about 0.81, a gain loss near 0.91 dB. A 1 m aperture at 10 GHz goes from roughly 1.5° uniform to about 2.0° with a cosine taper. The 3 dB beamwidth constant rises from 0.886 to about 1.19 (λ/L, radians), and 1.19 / 0.886 gives the 1.34x broadening. This beam broadening and gain loss buy the much lower sidelobes.
What is the difference between a cosine taper and a cosine-squared (Hann) taper?
Cosine weights as cos(πx/L), zero at the edges with nonzero slope, giving about -23 dB sidelobes, 1.34x broadening, and 0.81 efficiency. Cosine-squared (the Hann window) weights as cos²(πx/L), reaching zero with zero slope for about -32 dB first sidelobes and 1/n4 roll-off, but it widens the beam to about 1.63x and drops efficiency to about 0.67. Cosine squared trades roughly 9 dB more suppression for more broadening and about 0.8 dB more gain loss.