Selective fading

When the night sky bounces two versions of the same signal back to earth slightly out of step, the interference is not random noise — it eats certain frequencies inside the audio and leaves others alone.

Why some frequencies fall and others survive

Ordinary fading hits the whole signal at once: a cloud moves, the ionosphere shifts, and the received level drops uniformly. Selective fading is different. Two or more paths reach the same receiver — typically a ground wave and a skywave, or two skywaves reflected from different heights in the F layer — and they arrive with a small difference in travel time. Because they started as the same transmission, they interfere with each other. Where they arrive in phase the signal reinforces; where they arrive out of phase it cancels.

So far that is just multipath. The selective part comes from what a time delay does to a wideband signal. A delay of, say, 100 microseconds between two equal-amplitude paths produces cancellation at frequencies spaced by the reciprocal of that delay — roughly 10 kHz apart in that example — and reinforcement at frequencies in between. Across the 10 kHz bandwidth of a standard AM channel, such a comb of peaks and nulls means some audio frequencies are lifted while their neighbours are suppressed. The effect sits inside the audio, which is what makes it perceptually strange. The transmitter has not changed. The receiver is not broken. The distortion is built into the propagation geometry of that moment.

What it sounds like, and why AM is especially vulnerable

Amplitude modulation carries its information in the envelope of the carrier. Intelligibility depends on the relative amplitudes of the carrier and its sidebands reaching the detector in the same ratio they left the transmitter. When selective fading suppresses the carrier while leaving one sideband relatively intact — or attenuates only the upper sideband — the recovered audio is garbled: vowel formants shift, sibilants disappear or pop, and a voice can sound as though it is speaking through a broken telephone. The International Telecommunication Union recognises this mechanism explicitly in its propagation documentation for medium-wave and high-frequency planning, because it sets a practical ceiling on night-time coverage quality that raw field strength alone cannot predict.

The condition worsens when the sky is active. After sunset, the D layer — which absorbs skywave during the day — collapses, and signals that the ground wave reaches reliably at 100 kilometres can now be overridden or joined by skywaves having bounced once or twice. The geometry between transmitter, ionosphere and receiver determines the differential path length, and that geometry shifts continuously as the reflecting layer moves up and down through the night. A receiver sitting at medium distance may cycle through good reception and severe distortion over intervals of seconds to minutes, with no change in the transmitted signal.

Single-sideband transmission largely sidesteps the carrier-versus-sideband problem, which is one reason professional and maritime HF services adopted it. But medium-wave broadcasting has historically used double-sideband AM for receiver simplicity, and at those frequencies selective fading remains an engineering reality every night. When Edward Appleton demonstrated the reflecting layer's height experimentally in 1924, he used the phase relationship between ground wave and skywave to make the measurement — precisely the same physics that produces selective fading. The mechanism that proved the ionosphere exists is the mechanism that degrades AM audio after dark.

1. Two paths, same source → phase difference depends on path-length difference
2. Path-length difference is frequency-independent → phase difference grows with frequency → some audio frequencies cancelled, others reinforced
3. Carrier and sidebands affected unequally → AM envelope distorted → audio intelligibility collapses even when signal strength is adequate

Mitigation without a cure

Engineers can narrow the gap between the two paths by choosing antenna height to suppress high-angle skywave radiation, or by operating directional arrays that reduce the energy reaching the ionosphere at the relevant angles. Receivers with steep skirt selectivity can reject an interfering skywave arriving from a distant co-channel transmitter, but they cannot separate two paths that originate from the same source. Diversity reception — two antennas spaced to sample uncorrelated fades — helps on point-to-point links, but is impractical for broadcast listeners. In the end, selective fading is a geometric fact of night-time propagation, not a fault anyone can fully engineer away.

• Multipath
• The Grimeton exception
• Ground wave