Limitations of AM

Adding Up the Bill

Amplitude modulation works, and the last two lessons showed exactly how: a message riding on the envelope of a carrier, recovered by three components and a good choice of time constant. AM broadcasting reached most of the planet, and a hundred years later it is still on the air. So this lesson has an awkward job — to explain why a technology that plainly works was so unsatisfying that engineers spent two decades looking for something else.

The answer is not one flaw but four, and they are not equally deep. Two of them — the wasted power and the redundant bandwidth — are accounting problems, and AM's own variants fix them. The third is different in kind: noise arrives as an amplitude disturbance, and AM has put the message in the amplitude. That one cannot be fixed inside AM at all, and it is what eventually forced a different modulation scheme. The fourth, selective fading, is the same defect wearing a channel's clothes.

Two Thirds of the Power, None of the Information

Start with the arithmetic, because it is the easiest to make concrete. M3-L2 gave the total power of a standard AM signal:

Divide the sideband power by the total and you get the fraction of the transmitter's output that actually carries information — the modulation efficiency:

At maximum legal modulation depth, then, one third of the radiated power carries the message and two thirds is a carrier that conveys nothing whatever. Put a real transmitter behind that. A 50 kW medium-wave station at m = 1 radiates 33.3 kW of carrier and 16.7 kW of sidebands — and since the two sidebands are identical, only 8.3 kW of that is non-redundant. One sixth of the electricity bill is doing the job.

And m = 1 is the best case, not the typical one. Broadcasters keep peak modulation below 100% for the reasons M3-L2 and M3-L3 each gave, and the average depth over real speech and music is far lower still — programme material spends most of its time well below its peaks. At m = 0.5 the efficiency is 1/9, about 11%. At an average depth of m = 0.3 it is 0.09/2.09 = 4.3%. In the units of M2-L4 that is a handicap of 10 log₁₀ 0.043 = −13.7 dB before the signal has travelled one metre.

This is why AM broadcasting is a business of very large transmitters. To reach the same distance as an FM or SSB station of a given power, an AM station needs several times that power, and it pays for the difference every hour it is on the air.

Noise Arrives as Amplitude, and AM's Message Is Amplitude

Here is the defect that actually matters, and it is not an accounting problem. Almost every disturbance a radio signal meets on its way to the receiver adds to it: thermal noise in the receiver's own front end, atmospheric crashes from lightning, the broadband hash of a switching power supply, the spark of a brush motor. Adding one waveform to another changes the resulting amplitude. So noise, whatever its source, arrives as an amplitude disturbance.

And AM has put the message in the amplitude. The envelope detector of M3-L3 does its job perfectly: it reports the envelope of what arrives. It has no way, even in principle, to distinguish an envelope change that the transmitter put there from one that a lightning strike put there. Both are envelope changes. Both are the signal, as far as the detector is concerned.

Compare that with the alternative the next module will build. An FM receiver can afford to be brutal about amplitude: it passes the incoming signal through a limiter, a circuit that clips it to a constant height and throws every amplitude variation away, because none of the message was ever in the amplitude. The noise goes out with it. There is no AM equivalent of that circuit, and there cannot be — a limiter placed in front of an envelope detector would erase the message and leave the noise.

The practical consequence is that an AM link has no way to trade anything for noise performance. If the signal-to-noise ratio at the receiver is too low, the only remedies are more transmitter power, a better antenna, or a quieter site. You cannot spend bandwidth to buy noise immunity, because widening an AM channel adds nothing but room for more noise. That inability to trade is the deepest limitation in this lesson, and Module 4 opens by breaking it.

Selective Fading: The Same Defect, Delivered by the Channel

The medium- and short-wave bands where AM lives are ionospheric bands. A signal arrives by several paths at once — along the ground and reflected from one or more ionospheric layers — and those paths have different lengths that change through the day. When they add at the receiver they sometimes reinforce and sometimes cancel, which is the slow rise and fall every night-time listener knows as fading.

Fading alone would only make the station quiet, and the automatic gain control of M3-L3 handles that. The problem is that fading is frequency-selective. Multipath delay spreads of a millisecond or so are ordinary at HF, and a delay spread of 1 ms puts the coherence bandwidth of the channel at roughly 1/(2π × 1 ms) ≈ 160 Hz. The carrier and the sidebands of a 5 kHz-wide AM signal are further apart than that, so they fade independently.

Now the arithmetic turns hostile. Suppose the carrier happens to fade 6 dB while the sidebands do not. The sidebands still carry A_m, but the carrier amplitude has halved, so the effective modulation index the receiver sees is 2m. A station transmitting a well-behaved m = 0.6 now presents the envelope detector with m = 1.2 — over-modulation, envelope reversal, and the gross distortion M3-L2 warned about — without anything at all having changed at the transmitter. This is the hollow, watery, swirling sound of a distant night-time AM station, and it is not the receiver's fault and not the transmitter's.

No Capture Effect: Sharing a Channel Badly

Two AM signals on the same frequency simply add, because addition is what amplitudes do. The receiver's envelope detector produces the envelope of the sum, so you hear both stations at once, plus a whistle at the difference between the two carrier frequencies. Nothing in the receiver can pick one and reject the other; there is no threshold beyond which the stronger station wins.

FM behaves completely differently here — its limiter and discriminator suppress the weaker of two co-channel signals almost completely once it is a few decibels down, which M4-L1 introduces as the capture effect. For broadcasting, capture is plainly better: one clean station instead of two ruined ones. AM's answer has to be regulatory rather than technical, which is why medium-wave allocation is a matter of licensed geographic spacing, protected contours, and power reductions at sunset.

The same weakness applies to noise that is not another station. Man-made electrical noise — commutator motors, thyristor dimmers, switching supplies, overhead power lines, poorly filtered LED drivers — is broadband and impulsive, and it is strongest at exactly the low frequencies where AM broadcasting lives. An impulse is a large, brief amplitude event, which is the worst possible shape of interference for a detector that reads amplitude. Every one of them arrives as a click in the audio. A modern domestic electrical environment is far noisier at 1 MHz than the one AM was designed for, which is a large part of why the band sounds worse now than it did in 1960 even though the transmitters have not changed.

The Bandwidth Argument, Stated Honestly

AM is often called bandwidth-inefficient, and that claim needs care, because M3-L2 showed the opposite: at BW = 2f_m, AM is narrower than anything in Module 4. Both statements are true, and the way to hold them together is to be precise about what is being wasted.

The waste is redundancy, not width. The upper and lower sidebands are mirror images of one another and carry exactly the same information, so double-sideband transmission spends 2f_m of spectrum to deliver f_m worth of message. Half of the occupied band is a duplicate. That is a real inefficiency, and it has a real cost in broadcasting: a 10 kHz channel divided by two sidebands leaves about 5 kHz of audio, which is why AM sounds dull. The redundancy is paid for in fidelity.

And AM's own variants already fix it. SSB transmits one sideband and nothing else: half the bandwidth, no wasted carrier, all the power in the information. It is the most spectrally and energetically efficient analogue voice scheme ever standardised, and it is what long-distance HF voice actually uses. But look at what it costs — with no carrier and only one sideband, the receiver has no phase reference at all, so it needs the synchronous detection of M3-L3 with a stable, accurately tuned local oscillator. Every AM variant that fixes the accounting does so by taking away the thing that made AM attractive.

What a Fix Would Have to Do

Stack the four complaints up and the specification for a replacement writes itself. Anything that genuinely improves on AM must:

• Put the information somewhere amplitude noise cannot reach. Not in the envelope. The obvious remaining properties of a sine wave, from M2-L1, are its frequency and its phase.
• Let the receiver throw amplitude away entirely. That requires a constant-envelope transmission, so that a limiter can flatten the incoming signal without destroying anything.
• Spend all of the transmitted power on the message. A constant-envelope signal has no idle carrier term sitting beside the information; the whole waveform is the information.
• Offer a trade. Give the engineer a knob that converts bandwidth into noise immunity, so that a link with spectrum to spare can buy quality with it. AM has no such knob.
• Survive two signals on one channel. Degrade to one usable station rather than two unusable ones.

A wave whose amplitude never changes and whose frequency carries the message satisfies every line of that list. That is frequency modulation, and Module 4 is about to build it. It is worth knowing that the idea was not obviously right at the time: in 1922 John Carson — the same Carson whose bandwidth rule M4-L2 uses — published an analysis concluding that frequency modulation offered no reduction in noise and merely widened the band. He was correct for the narrowband case he examined, and it took Edwin Armstrong's demonstration in 1933 that going deliberately wide converts bandwidth into a large noise advantage. The fourth item on the list above is the one nobody expected.

Where AM Is Still the Right Answer

None of this makes AM obsolete, and a lesson that ended here would be teaching a caricature. AM survives wherever its weaknesses are irrelevant or, in one striking case, are the point.

Civil aviation voice is AM, in the 118–137 MHz band, and it is AM partly because there is no capture effect. If two aircraft transmit at once on a control frequency, both are heard — garbled, with a distinctive heterodyne whistle, but audibly two. Under FM the stronger station would capture the channel and the weaker aircraft would simply vanish with no indication that it had ever spoken. On a safety-critical channel, "both transmissions are audible and obviously overlapping" is worth far more than "one transmission is clean". The defect became a feature.

Medium-wave broadcasting is AM because coverage and receiver cost dominate everything else. A single high-power transmitter reaches an entire country, further still after dark when the ionosphere returns the signal to earth, and the receiver can be built for pennies — which still matters a great deal in emergencies and in places where it matters most. When a storm takes out the mobile network, the transmitter with the largest coverage footprint and the receivers that run for a month on two dry cells are both AM.

And amplitude modulation itself never went away; it went digital. Module 5 will show that the dense constellations carrying modern data rates — QAM, and the OFDM systems built from it — encode information in amplitude and phase together. What changed is not the physics but the surrounding machinery: with error-correcting codes, equalisers and digital synchronisation, amplitude becomes usable again because the receiver can measure, model and correct what the channel did to it. AM's fatal flaw was never that amplitude is a bad place to put information. It was that an analogue receiver had no way to tell what had happened to it.