Modulation · Entry 05.2
Frequency modulation
Edwin Armstrong's alternative, the capture effect, and why FM sounds clean and then vanishes rather than degrading.
Encode the signal in the carrier's frequency, not its envelope, and noise — which rides the amplitude — becomes almost irrelevant. Almost.
The mechanism Armstrong built
Amplitude modulation is vulnerable by design: anything that disturbs the strength of the carrier — atmospheric static, ignition interference, the slow fading of a skywave signal — writes itself directly into the audio. Edwin Armstrong spent years looking for the exit. He found it in the 1930s by encoding the programme in the rate of the carrier's frequency swing rather than in its amplitude. A louder sound swings the frequency wider; a higher-pitched sound swings it faster. The carrier's envelope stays constant. A receiver built to track frequency deviation and ignore amplitude variation is, in principle, deaf to amplitude noise.
The principle was not new — Carson's 1922 analysis had explored narrowband FM and dismissed it — but Armstrong's key insight was that wideband FM, using far more spectrum than a comparable AM channel, bought a dramatic improvement in signal-to-noise ratio. He demonstrated this publicly in 1935, and later built his Alpine, New Jersey station in 1938, and the results were striking enough that the physics couldn't be argued away: at sufficient signal levels, FM is genuinely quieter than AM by a margin no amount of AM filtering can close.
How FM encodes audio
- Frequency deviation — how far the carrier swings; wider swing = louder audio
- Deviation rate — how fast it swings; faster rate = higher pitch
- Envelope — held constant; this is what blocks amplitude noise
The Federal Communications Commission eventually allocated 88–108 MHz for FM broadcasting in the United States, a band high enough in frequency that ground-wave coverage is essentially limited to line-of-sight, and skywave propagation — which would allow interference across vast distances — is absent under normal conditions.
The capture effect and the cliff edge
FM has a property AM entirely lacks: the capture effect. When two signals arrive on the same frequency, an FM receiver does not blend them into a mess of co-channel interference the way an AM receiver does. Instead it locks onto whichever signal is stronger — usually the one that leads by 6 dB or so — and suppresses the weaker one almost completely. This is why FM co-channel interference sounds like brief, occasional breakthrough rather than the constant blended hash of AM co-channel. The dominant station wins cleanly.
The same physics that makes FM robust against noise gives it a sharp coverage boundary rather than a gentle roll-off. An AM station weakens gradually as distance increases, the audio degrading through hiss and static in ways a listener can tolerate for surprisingly long stretches. FM does the opposite: within its service area the signal is clean and quiet; outside it, the signal-to-noise ratio drops below the FM threshold and the receiver's quieting collapses. Audio goes from pristine to noisy within a very short distance — the cliff edge. There is no gradual twilight, no analogue of the AM fringe where something listenable persists.
This cliff is partly mitigated by pre-emphasis and de-emphasis. FM transmitters boost high audio frequencies before modulation (pre-emphasis, standardised at a 75-microsecond time constant in North America and 50 microseconds in much of the rest of the world); receivers apply a complementary roll-off on output. Because high-frequency noise also gets rolled off, the audible noise floor drops further. It is a system-level trick, not a property of FM itself, and the two time constants are not interchangeable — a broadcast recorded off-air in one region sounds tonally wrong when played back through equipment calibrated for the other.
Key numbers and standards
Multipath — signals arriving at the receiver via reflections off buildings or terrain — is a real limitation. Where AM multipath causes selective fading, FM multipath distorts the instantaneous frequency deviation: two copies of the signal arriving slightly out of time interfere in the FM demodulator and produce a characteristic crinkling distortion in the high frequencies. It is worst at low vehicle speeds, where the geometry changes slowly, and it is the dominant reason FM in cities sounds worse than FM in open country, even when signal strength looks adequate.
Armstrong demonstrated something real. The trade-off he made — spectrum width for noise immunity — remains the foundation of analogue FM broadcasting worldwide, and the capture effect, the cliff-edge boundary and the pre-emphasis curve are all direct consequences of the same underlying decision to put the signal in the frequency, not the amplitude.