Imagine an ideal receiving antenna connected to a power meter. At one orientation, the meter shows a strong response. Turn the antenna through a right angle and its response falls to zero in our mathematical example. Nothing about the incoming wave's frequency has changed. The antenna is responding to a different part of the same electric field. This dependence on direction is called polarization coupling. The National Radio Astronomy Observatory explains it through the response of a linearly polarized receiving antenna. [1]

The useful question is not simply whether a radio wave is present. It is how that particular receiver responds to the wave that reaches it. A receiver can miss a component that another receiver would detect. An ideal zero reading therefore does not mean that the space around the antenna contains no radio energy. Nor does a larger reading automatically describe a faster internet connection.

We will use original calculations with an imaginary, carefully controlled receiver. There is no physical antenna adjustment or radio transmission to perform. First we will separate the wave's travel direction from its electric-field direction. Then we will calculate how a tilted receiver changes voltage and power, and finally ask what a rotation experiment could still leave unknown. That last question matters: different polarization states can produce the same power readings.

Consider an ideal linear receive channel exposed to a monochromatic, fully linearly polarized plane wave. The incident frequency, propagation direction and field amplitude are fixed. Comparing stationary receive-axis orientations changes the channel's response even though the incident wave is unchanged. NRAO gives the voltage dependence as cos θ and the power dependence as cos² θ, where θ measures the relative orientation of the feed and incident electric field. [1]

This is a question about polarization selectivity, not the existence of radiation. A null in one ideal receiver port does not imply zero incident electromagnetic energy. The same distinction prevents a polarization factor from being used as a direct percentage of network throughput, which would require a separate communication-system model.

The original examples below isolate field projection, same-load power detection and the interpretation of angular response. They are not measured antenna data or a software simulation. We will define the geometry before using the cosine law, preserve the sign of voltage before squaring, and compare linear, circular and unpolarized inputs. In particular, a constant reading from a rotating linear analyzer does not uniquely identify an unpolarized input. That ambiguity limits what can be concluded from a power trace alone.

For the simple wave in this article, draw its travel direction straight toward you. The electric field moves side to side in a plane across that direction. In a linearly polarized wave, it oscillates along one fixed line: it points one way, passes through zero, and points the opposite way. Its axis stays the same even though its arrow reverses. NRAO's wave primer distinguishes this electric-field behavior from the direction in which the wave propagates. [3]

Now put two perpendicular reference lines on that transverse plane and call them x and y. Let the incoming electric field lie along x. A receiving axis can also lie along x, along y, or somewhere between them. The angle we need is between the incoming field's axis and the receiving axis in this shared plane. It is not the angle between the receiver and the wave's travel direction.

Frequency answers a different question: how rapidly the oscillation repeats. Polarization describes the electric field's directional behavior. [2] A smaller sideways projection can have exactly the same repetition rate. Words such as horizontal and vertical only help when the reference frame is known; rotating a drawing on a screen does not change the physical relationship between its two axes.

Use a right-handed coordinate system with propagation along +z. The x and y directions span the transverse plane. For a free-space plane wave, both electric and magnetic fields are perpendicular to propagation; the electric field therefore needs only two transverse components. [2][3][5] These assumptions are specific to the plane-wave description, not a universal statement about every field near an antenna.

Let E(z,t)=E0 x̂ cos(kz−ωt), with E0 a positive field amplitude, k the wavenumber and ω the angular frequency. A real unit receive axis in the same basis is u(θ)=x̂ cos θ+ŷ sin θ. Here θ is an in-plane angle measured from x toward y. Every u remains perpendicular to z, so the incoming direction stays broadside to the ideal linear element.

A linear polarization axis is unchanged by the field's half-cycle sign reversal. Its frequency is determined by ω, whereas its polarization is determined by the relative transverse components and their phase behavior. [2][5] Defining both vectors in the same basis is essential: a dot product between coordinates expressed in unrelated local frames has no physical meaning until the frames are related. Horizontal and vertical are labels within that geometry, not universal directions detached from it.

Extend the angle comparison to 180 degrees. The receiving axis now points opposite to the incoming field's chosen positive direction. Its signed projection is minus one instead of plus one. With our fixed electrical polarity reference, that means the voltage waveform is reversed: a positive part becomes negative, and a negative part becomes positive. Squaring either waveform gives the same power. [1][3]

An ordinary power value cannot tell these two signs apart. Both the 0-degree and 180-degree cases give 16 µW in our example. Linear polarization describes an axis, so turning that axis through half a full rotation returns the same polarization alignment. The sign still matters when comparing voltage waveforms using a defined reference; it has simply disappeared from this power measurement.

At 90 degrees, the selected component is zero. This does not erase the incoming field. Another ideal receiver aligned with that field would respond to it. The important phrase is “zero in this channel under these assumptions”. Real equipment can have imperfect polarization rejection or receive waves from several paths, so the exact mathematical zero is not a promised meter reading outside the model.

For a directed receive-axis and terminal-polarity convention, cos(θ+π)=−cos θ. The signed voltage amplitude changes sign, equivalent to a π phase reversal for a nonzero sinusoid, but cos²(θ+π)=cos² θ. Thus the power response is periodic in 180°, even though the signed projection changes. [1][3] No claim about an unspecified device's wiring or phase-sensitive output is needed.

This separates a polarization axis from an oriented vector used to represent terminal voltage. The linear state has the same axis after a half turn. Reversing the chosen voltage reference cannot create negative received power; same-load power is quadratic. At θ=0 and θ=π, the illustrative voltage ratios are +1 and −1, while both power ratios equal one.

At θ=π/2, the ideal channel is orthogonal to the incident electric field and its selected component vanishes. This is a response null, not a statement that the incident field or its energy flux vanishes. The conclusion also relies on perfect linear selectivity and the single specified incident wave. Cross-polar leakage, altered arrival geometry or additional fields fall outside the controlled calculation; they must not be silently folded into a claim of an exact real-world null.

A circularly polarized wave behaves differently from our starting linear wave. At a fixed place, the tip of its electric-field arrow rotates around a circle. Two perpendicular components of equal size, offset by a quarter of an oscillation, produce this behavior. The field remains across the travel direction; the wave is not travelling around that small circle. [2][5]

An ideal linear receiving channel takes one projection of that rotating field. Averaged over full cycles, it receives half the total normalized power, whichever transverse angle is chosen. With the same total-power normalization of 16 µW, its mean reading is 8 µW at every angle. Changing the receiver's orientation changes the timing of that projection, but not its mean power.

Unpolarized radiation also gives half its total mean power to an ideal linear channel at every angle. [2] Yet it is not the same state as a circularly polarized wave: it lacks that stable circular relationship between its components. Therefore a flat 8 µW rotation trace could fit either of these two inputs. Power readings alone have left out information needed to distinguish them. Calling a flat trace “unpolarized” without another suitable measurement would turn an incomplete observation into an unsupported identification.

At a fixed position, write a circular field as E(t)=E0[x̂ cos φ+ŷ sin φ], where φ=ωt plus a constant phase. Its squared magnitude is E0² at every instant. Projection onto u(θ) gives E0 cos(φ−θ), whose cycle-averaged square is E0²/2, independent of θ. This derives the half-power response for an ideal linear analyzer without selecting a convention-sensitive circular-handedness name. [2][5]

The comparison uses equal total mean incident power, not equal peak amplitude between a one-component linear sinusoid and a two-component circular field. Those peak-amplitude conventions have different mean-square totals. With a total response normalization of 16 µW, the circular input gives 8 µW per linear channel; a fully aligned linear input gives 16 µW to one channel and zero to its orthogonal partner.

An unpolarized input has no preferred polarization in its averaged second-order statistics and likewise contributes half its total power to any ideal linear analyzer. [2] Its statistical component relationships differ from the coherent quadrature relation of the circular example. Consequently a one-channel mean-power rotation trace cannot distinguish these states. Resolving that ambiguity requires additional polarization-sensitive information, such as appropriately measured intercomponent phase correlations, rather than merely more angles of the same mean-power observable.