Riding a Carrier: AM and FM Explained

Why a whisper can't fly

Picture the chain you met on rung 1: a message goes in one end, gets shaped for the channel, radiates across a gap, and is reconstructed at the far end. The very first stumbling block is brutally physical. Human speech and music live below about 3 kHz to 20 kHz. To radiate efficiently, an antenna wants to be a respectable fraction of a wavelength long — roughly a quarter. At 3 kHz the wavelength is 100 km, so a 'good' antenna would be 25 km tall. Nobody is building that to broadcast a love song.

There's a second, equally fatal problem: everyone's whisper occupies the same low band. If two radio stations both pushed raw audio into the air, their signals would pile on top of each other into mush. Audio is born sharing one tiny slice of frequency, and you cannot stack a city's worth of conversations there.

The fix for both problems is the same elegant trick: modulation. Take a fast, pure sine wave — the carrier — at, say, 1 MHz (wavelength 300 m, antenna a friendly 75 m) and let the slow message gently reshape it. Now the message *rides* the carrier. Pick a different carrier frequency for each station and they slot into separate lanes of the spectrum, side by side, never touching. One trick, two wins: things radiate, and everybody gets a lane.

Amplitude modulation: writing on the carrier's height

The most intuitive way to scribble a message onto a carrier is to make the carrier taller and shorter in step with the message. Loud bit of music? Pump up the carrier's amplitude. Quiet moment? Shrink it. The carrier's frequency never changes — only its envelope breathes with the sound. This is amplitude modulation (AM), the scheme behind the AM band on your car radio and the oldest broadcast technology still in daily use.

message m(t)        carrier (1 MHz)        AM signal s(t)
   .-.                |||||||||||||           .-.
  /   \   X  |||||||||||||||||||   =   /''-..'   '..-''\
_/     \_     |||||||||||||||||      _.|||||||||||||||||.._
              (fast pure sine)        ^ envelope = m(t) ^

  s(t) = [ 1 + m * cos(2*pi*f_m*t) ] * A_c * cos(2*pi*f_c*t)
         \_______ envelope _______/   \___ carrier ___/

AM in one picture: the message reshapes the **envelope**; the carrier still oscillates fast underneath. 'm' is the modulation index.

Notice the term `1 + m·cos(...)`. The modulation index m sets how deeply the message swings the amplitude. With m = 0.5 the carrier dips to half and swells to 1.5× — comfortable. Push m past 1 and the envelope tries to go negative; the waveform 'folds over' and the receiver hears ugly distortion. This is overmodulation, and broadcast engineers guard against it religiously.

Why does the receiver bother to find the envelope at all? Because the envelope *is* the message. The simplest AM detector is almost laughably cheap: a diode and a capacitor. The diode chops off the bottom half of the wave; the capacitor smooths the fast carrier ripples away and what's left tracing the tops is your original audio. That dirt-cheap envelope detector is exactly why AM radios could be built for pennies a century ago — and why a 'crystal radio' needs no battery at all.

Sidebands and the 2× bandwidth rule

Here is the single most useful fact about AM, and it falls out of one line of trigonometry. Multiply a carrier at f_c by a single message tone at f_m and the product identity splits it into three pure frequencies: the carrier itself, plus a tone above at (f_c + f_m), plus a tone below at (f_c − f_m). Modulation doesn't keep the signal at one frequency — it *spreads* it into a small band around the carrier.

spectrum of an AM signal (single 5 kHz tone on a 1 MHz carrier):

  amplitude
    |              ###  (carrier, 1.000 MHz)
    |        |||         |||
    |       lower        upper
    |     sideband      sideband
    |   (0.995 MHz)    (1.005 MHz)
    +----+-------+-------+--------> frequency
             |<--- 10 kHz --->|
          bandwidth = 2 x f_m = 2 x 5 kHz

A single tone becomes three lines. A real audio signal smears each sideband into a band of width = the message bandwidth.

Real audio isn't one tone — it's a whole spectrum of tones up to some highest frequency, call it W. Each tone makes its own pair of sidebands, so the upper sideband becomes a mirror copy of the audio sitting just above the carrier, and the lower sideband a flipped copy just below. The signal stretches from (f_c − W) to (f_c + W). That gives AM its golden rule:

  AM transmission bandwidth  =  2 x (message bandwidth)

  Broadcast AM voice/music limited to W = 5 kHz
     -->  bandwidth = 2 x 5 kHz = 10 kHz
     -->  stations spaced 10 kHz apart on the dial
          (540, 550, 560 ... kHz)

Why AM stations sit on a 10 kHz grid: each needs a 10 kHz lane, and you can't let lanes overlap.

Frequency modulation: writing on the carrier's pitch

Edwin Armstrong, who already had AM down cold, asked a heretical question in the 1930s: what if we *forbid* the amplitude from carrying the message, and instead wiggle the carrier's frequency? Keep the carrier's height rock-steady, but let it speed up when the music goes loud and slow down when it goes quiet. That is frequency modulation (FM) — and the payoff is a near-magical immunity to noise.

message loud  -->  carrier squeezes faster  -->  higher freq
message quiet -->  carrier stretches slower -->  lower freq
message zero  -->  carrier rests at f_c (center)

   /\  /\  /\        /\/\/\/\        /\  /\  /\
  /  \/  \/  \  ->  /||||||||\  ->  /  \/  \/  \
  |  amplitude is CONSTANT the whole time  |
  freq swings +/- Df (the 'frequency deviation') around f_c

In FM the height never changes; the spacing of the wiggles does. The message lives in the pitch, not the loudness.

Two numbers define an FM signal. The frequency deviation Δf is how far the carrier swings from center at full message volume — for FM broadcast, ±75 kHz. The modulation index β = Δf / W compares that swing to the highest audio frequency W. Unlike AM, FM's bandwidth is not simply twice the message; it depends on β through a beautifully practical shortcut called Carson's rule:

  Carson's rule:   B  ~=  2 x ( Df + W )  =  2 x W x ( beta + 1 )

  FM broadcast:  Df = 75 kHz,  W = 15 kHz  (full hi-fi audio!)
     beta = 75 / 15 = 5
     B  ~= 2 x (75 + 15) = 180 kHz

  Compare AM hi-fi-attempt:  W = 15 kHz -> B = 30 kHz only

FM buys clean 15 kHz audio but spends 180 kHz of spectrum — six times an AM lane. That's the trade in one calculation.

Why FM sounds cleaner — and the trade you just met

Now we cash in Armstrong's heresy. Almost all natural noise — lightning, sparks, thermal hiss in the receiver — corrupts a signal's amplitude. AM carries its message in exactly that amplitude, so the noise lands right on the message: crackle, pop, hiss. FM carries its message in frequency, and the height is constant by design. So the FM receiver can do something AM never can: run the signal through a limiter that brutally clips it to a flat-topped square, *throwing the amplitude away entirely*. Any amplitude noise rides on top — and gets clipped off with it.

There's a second gift. Because FM smears the message across a wide band (that fat 180 kHz), the receiver gathers signal energy from a broad swath of spectrum, and a wider deviation yields a larger jump in signal-to-noise ratio at the output than you 'paid' in input SNR. This wideband FM noise gain is roughly proportional to β² — a real, quantifiable bonus for spending more bandwidth. Trade spectrum, get fidelity. It is the same coin AM refused to spend.

  Worked contrast: same noisy night, same transmit power
  -----------------------------------------------------
                       AM station        FM station
  message bandwidth W   5 kHz             15 kHz (hi-fi)
  spectrum used         10 kHz            180 kHz
  noise hits the...     amplitude=message frequency (limiter kills amp noise)
  thunderstorm crackle  loud pops         barely audible
  result                tinny + noisy     full-range + quiet

Same power, same storm: AM puts the message where the noise is; FM puts it out of the noise's reach.

But pin down what FM actually paid. To get that clean 15 kHz audio and that noise immunity, FM ate 18 lanes of AM spectrum (180 vs 10 kHz). That is the bandwidth-versus-robustness trade, and it is the spine of this entire track. AM is spectrally thrifty but fragile. FM is spectrally lavish but tough. Every scheme you'll meet later — digital ASK/FSK/PSK, higher-order multiplexing schemes, the spread-spectrum tricks behind GPS and Wi-Fi — is a fresh answer to the same question: how much bandwidth will you spend to buy how much robustness?

Take the message — slow audio with bandwidth W, far too sluggish to radiate.
Pick a carrier at f_c, high enough for a sane antenna and its own spectrum lane.
Modulate — write the message onto the carrier's *amplitude* (AM) or its *frequency* (FM).
Pay the bandwidth bill — AM costs 2W; FM costs ~2(Δf + W), much more.
Demodulate — recover the envelope (AM) or track the frequency wiggle (FM); FM's limiter scrubs amplitude noise away.

Where this lands you

AM and FM are *analog* — the carrier's height or pitch varies smoothly and continuously, a direct echo of the sound. The next rungs swap that smooth wiggle for crisp symbols: instead of a continuous envelope you'll flip the carrier between a few discrete states to send bits. The box that does that flipping and un-flipping has a name you already know — a modem (modulator-demodulator). But the AM/FM intuitions carry straight over: switch amplitude in discrete steps and you have ASK; switch frequency and you have FSK. You are not leaving these ideas behind; you're digitizing them.