Spherical coordinates

The formula

Spherical to Cartesian, in the physics convention (ISO 80000-2: θ is the polar angle measured down from the +z axis, φ is the azimuth measured around from +x) ^[7]^[8]:

x = r·sinθ·cosφ,    y = r·sinθ·sinφ,    z = r·cosθ

Reading: tilt a ray θ off vertical, swing it φ around, walk out distance r. The sinθ in front of both horizontal components is the ray's footprint in the xy-plane.

Going back:

r = √(x² + y² + z²),    θ = acos(z/r),    φ = atan2(y, x)

Reading: r is 3-D Pythagoras, θ compares height to total distance, and azimuth uses atan2 — never atan(y/x) — for the same quadrant reasons as in plane polar. Check on the point (1, 1, 1): r = √3 ≈ 1.732, θ ≈ 54.74°, φ = 45°, and the forward formulas return (1, 1, 1) exactly.

The volume element:

dV = r²·sinθ·dr·dθ·dφ

Reading: the Jacobian of the transformation is r²·sinθ (DLMF 1.5.41 ^[9]) — cells grow with range squared and shrink toward the poles. Sanity check: integrating over 0 ≤ r ≤ R gives (4/3)·π·R³, the sphere volume.

The line element — how much actual distance each coordinate step buys:

ds² = dr² + r²·dθ² + r²·sin²θ·dφ²

Reading: a step in azimuth is worth r·sinθ·dφ of real motion. A degree of φ near the pole covers almost no ground; at θ = 90° and 1 km range, 0.1° of azimuth is 1.75 m of cross-range.

Solid angle, the currency of radiation problems:

dΩ = sinθ·dθ·dφ,    full sphere: ∮ dΩ = 4π sr
P_rad = ∮ U(θ, φ)·dΩ

Reading: total radiated power is radiation intensity summed over the whole 4π steradians. A cone of 1° half-angle subtends 2π·(1 − cos 1°) ≈ 0.000957 sr — roughly 1/13,000 of the sphere, which is why pencil-beam antennas have the gain they do.

Where you meet it

On the antenna range. A far-field pattern is gain versus (θ, φ) on a sphere of fixed radius, and the positioner sweeps it out one great-circle cut at a time. The directivity number on the test report comes from integrating the measured pattern with that sinθ weight in place — leave it out and the answer is confidently wrong.
At the tracking radar and on the range safety console. The instrument measures range, azimuth, and elevation — spherical coordinates in field clothing, with elevation measured up from the horizon instead of down from zenith: el = 90° − θ. Every track file conversion to Cartesian runs the formulas above.
In the structures and thermal review. Pressure vessel heads, radomes, nose cones, tank domes — anything with spherical curvature gets meshed and analyzed in r, θ, φ, and the mesh has documented special handling at the pole because the coordinate system degenerates there.
Anywhere GPS meets analysis. Geodetic latitude is measured from the equator; the math convention measures θ from the pole. Huntsville sits near latitude 34.7°, which is colatitude 55.3°. Software that feeds one into a formula expecting the other produces positions that are wrong by exactly the complement — a classic integration-test find.

How it works

Use spherical coordinates when the physics depends on distance and direction from a point: a radiating antenna, a point charge, a blast, a star. The payoff is structural. Laplace's equation separates in these coordinates, the angular part produces spherical harmonics, and a far field falls off as 1/r with a pattern that depends only on (θ, φ) — which is why an antenna pattern is a function of two angles and not three coordinates. If the problem has an axis instead of a center, cylindrical coordinates fit better; if it's a box, stay Cartesian.

The traps, in the order they bite:

The convention swap. Physics texts, antenna literature, and ISO 80000-2 use (r, θ, φ) with θ polar and φ azimuthal. Most American calculus textbooks swap the two Greek letters — their θ is the azimuth and their φ is the polar angle ^[7]^[9]. Same geometry, same formulas, opposite labels. A pattern file, a textbook integral, and a simulation tool can all be individually correct and still disagree. The tell: in the physics convention, sinθ multiplies both cosφ and sinφ in the conversion. Check which angle lives inside the sine before trusting anything imported.
Three vertical-angle conventions in one building. Colatitude θ (from zenith), elevation (from horizon), latitude (from equator). The RF group, the radar crew, and the GPS receiver each use a different one, and all three angles describe the same ray. The conversions are trivial — el = 90° − θ — which is precisely why nobody writes them down and the error survives until a boresight check fails.
The missing sinθ. A uniform grid in θ and φ is not a uniform sampling of the sphere; it piles points up near the poles. Only 13.4% of a sphere's area lies within 30° of the axis, though those two polar caps cover a third of the θ range. Averaging measured pattern points without the sinθ weight overcounts the polar caps, and the resulting directivity can be off by a decibel or more depending on where the beam points.
The pole singularity. At θ = 0 the azimuth is undefined — every φ describes the same point. This is the coordinate-system face of gimbal lock: an az-over-el positioner tracking a target straight through zenith demands an instantaneous 180° azimuth slew. Range software calls it the keyhole. It isn't a physics problem; it's the coordinates, and the standard fixes are a tilted mount, a different gimbal order, or quaternions.

One more habit: the unit vectors r̂, θ̂, φ̂ point in different directions at every point on the sphere. That's not a nuisance — it's the standard polarization basis. E_θ and E_φ components on a pattern measurement are field projections onto the local θ̂ and φ̂ directions, so a co-pol/cross-pol call depends on the coordinate convention too. The angle trap and the polarization trap are the same trap.

History

Engineers used spherical coordinates before mathematicians owned them. The latitude-longitude grid — azimuth and complement-of-polar-angle on a sphere — is generally credited to Eratosthenes in the third century BC, refined by Hipparchus, and deployed at scale in Ptolemy's Geography in the second century AD, which catalogued the known world's major places by latitude and longitude ^[1]^[2]. Positioning by two angles on a sphere is older than almost everything else in this reference.

The analytic machinery came much later, from the same lineage as plane polar coordinates. Alexis Clairaut was the first to think of extending polar coordinates to three dimensions, though he only raised the possibility; Leonhard Euler was the first to actually develop the system ^[3]^[4].

The reason the coordinates matter to RF engineers traces to gravity, not antennas. In January 1783 Adrien-Marie Legendre presented the Académie des Sciences a memoir on the attraction of spheroids, introducing the functions now called Legendre polynomials — published in his 1784 Recherches sur la figure des planètes — and Laplace, who reviewed the work, built the corresponding expansion of the potential in spherical coordinates ^[5]^[6]. Those angular functions grew into spherical harmonics: the natural vibration modes of the sphere, and the same functions inside every modern near-field-to-far-field transform on an antenna range ^[6]. The notation finally got adjudicated in our own era: ISO 80000-2 pins down the physics convention — θ polar, φ azimuthal — as the standard, which is the convention antenna work has always used and calculus classes still often don't ^[7]^[8].

Related tools

/tools/antenna-gain-beamwidth — gain and beamwidth are statements about how a pattern fills solid angle
/tools/link-budget — the 1/r² spreading loss is the far-field radial dependence integrated over the sphere
/tools/radar-horizon — range and elevation geometry on a spherical Earth
/tools/orbital-velocity-period — orbital mechanics runs in r, θ, φ around the Earth's center
/tools/convert-angle — the degrees/radians step every conversion above assumes you got right

Sources

Written by HE in our own words from the cited sources — engineering judgment included, your stamp still required. All entries →