showstack

The word

What analogue actually means

Analogue means one physical quantity standing in for another, continuously and proportionally. A microphone diaphragm moves the way the air moved; the voltage it produces varies the way the diaphragm moved. The voltage is an analogue of the sound, and at no point is it a number.

That is the whole difference, and both consequences follow from it directly. An analogue signal has infinite resolution in principle — there is no step size, no smallest change — and it has no error correction whatever, because every imperfection the wire adds is indistinguishable from signal. Noise accumulates at every stage and never comes back off. A digital signal has a step size you chose and a floor you designed, and it either arrives intact or it does not.

Which is better depends entirely on what you are protecting. Copy an analogue tape ten times and you have a recognisably worse tape. Copy a file ten times and you have the same file. But an analogue system degrades gracefully — it gets noisy, then noisier, and stays usable a long way down — where a digital one works perfectly and then stops, which is the cliff every digital link on this site has in it somewhere.

The parts

Five components, and what each one is actually for

Resistor, capacitor, inductor, diode, transistor. Almost everything else is a combination of those, and each one does one thing.

One resistor and one capacitor

The most useful circuit in the trade

Put a resistor and a capacitor together and you have a filter, a delay, a de-bounce, and the tone control on every analogue console ever built. Two equations describe it and they are the same fact seen twice.

In the time domain the capacitor charges through the resistor on a curve, reaching 63% of the way in one time constant — τ = RC — then 95% in three and 99% in five. That is how long a mute ramp takes, how long a de-bounce holds off a bouncing switch contact, and how long a power supply holds up when the mains dips.

In the frequency domain the same two components have a corner at f = 1 / 2πRC, where the capacitor’s reactance equals the resistance. Below it the signal passes, above it the slope is 6 dB per octave, and stacking another stage adds another 6. Every analogue EQ on every console is networks of these, and every filter slope you have ever specified in a processor is counting poles.

The console

How a rig was controlled with no memory and no processor

An analogue lighting console has no computer in it, and the way it worked is genuinely clever rather than merely primitive.

Each fader is a potentiometer — a resistor with a sliding tap — producing a control voltage, usually 0 to 10 volts, and that voltage runs down its own wire to its own dimmer. One conductor per channel, all the way. A 48-way rig meant a 48-core cable, and the reason multicore looms were as thick as an arm is that there was no other way to do it.

Memory is where it gets interesting, because there is none. What a two-preset console has instead is two complete banks of faders and a crossfader between them. The operator sets the next state on the inactive bank while the current one is live, then crossfades. The crossfader is a dual-gang potentiometer wired so one half fades up as the other fades down — the cue stack is physical, the memory is the operator’s hands, and the fade is a resistance changing.

Then the cable became the problem, and the answer was multiplexing: send the channels one after another down one pair instead of all at once down many. Analogue multiplexing came first (AMX192, D54), sampling each channel in turn as a voltage. Digital multiplexing followed and won, because once you are sending them in sequence anyway you may as well send numbers, which do not degrade. DMX512 is that idea and nothing more: 512 numbers in a row, over and over, on one pair.

The dimmer

Three ways to make a lamp half bright, two of them obsolete

The console sent a control voltage. Something had to turn that into less light, and the three answers in order are a good short history of the whole trade, because each one solves the previous one’s worst problem and introduces its own.

The resistance dimmer is a variable resistor in series with the lamp. It works, and it is terrible: the power you take away from the lamp does not vanish, it comes out of the dimmer as heat. A 1 kW lamp at half brightness means several hundred watts being burned in a box backstage, which is why resistance boards lived in their own room and why the operator worked in the heat. Worse, it is load dependent. A resistance dimmer built for a 1000 W lamp given a 500 W one will not black out, because the two resistances now divide the voltage differently. You matched the dimmer to the lamp or you did not get a fade. The portable version, a rack of these on a frame with a row of levers along the top, is the piano board, and the name is purely about the shape of the thing you stood at.

The autotransformer dimmer — a Variac, a tapped winding with a sliding brush — fixed both faults at once. It transforms rather than burns, so it is efficient and runs cool, and because it varies the voltage rather than dividing it, it is load independent: any lamp within its rating fades properly. What it is instead is enormous. One iron-cored transformer per channel, each the size of a paint tin, which is why a 60-way autotransformer board is a piece of furniture and why they were operated by hand, sometimes with mechanical linkages so one lever could move several.

The thyristor dimmer is what replaced both and what almost every dimmer still is. It does not reduce the voltage at all — it switches the mains on partway through each half cycle and lets the rest through, so the lamp sees a chopped waveform and averages it. Cheap, small, load independent, and controllable by a tiny signal, which is what made a remote console with hundreds of channels practical.

And it introduced the problems we still have. Chopping the waveform generates harmonics, the third among them, which is a triplen — and triplens do not cancel in a three-phase neutral, they add up in it. It also makes the filament move: the sudden switch-on each half cycle is a mechanical shock, which is filament sing, and the choke in every dimmer exists to slow that edge down. A dimmer that buzzes has a choke doing its job; a lamp that buzzes has one that is not big enough.

The VCA

A fader that carries no audio at all

The audio console has its own version of the same problem: how one hand controls many things without the signal having to go through that hand.

A voltage-controlled amplifier is a gain stage whose gain is set by a control voltage rather than by a knob in the signal path. Once you have that, the fader stops being part of the audio circuit and becomes a source of DC. It is a control surface in the literal sense, decades before that phrase existed.

That is what a VCA group is, and it is why it is not a subgroup. A subgroup sums the audio of several channels into one bus and gives you a fader on the sum. A VCA group sums nothing: the channels stay on their own paths and the group fader simply adds an offset to each channel’s control voltage. Pull a VCA group down and each channel’s own fader moves in effect but not in fact, all their sends and inserts scale with them, and the audio never met an extra amplifier.

The practical difference shows up the moment you use a post-fade send. On a subgroup, a reverb send taken from the channel is unaffected by the group fader, because the audio has already left. On a VCA, it follows, because the channel gain itself moved. Every digital console still models this distinction and still calls them by these names, which is why the vocabulary of a 1975 console is worth knowing on a 2025 one.

The amplifier

One idea in three sets of clothes

An amplifier is a small signal controlling a much larger current from the power supply. The signal does not get bigger; it steers something bigger. What separates the classes is how the output devices are driven, and it is entirely a bargain about heat.

Class A keeps the output devices conducting all the time, over the whole waveform. Beautifully linear and appallingly inefficient — well under a third of the power drawn reaches the loudspeaker and the rest is heat, at idle as much as at full output.

Class AB splits the waveform between two devices, one handling the positive half and one the negative, with a small overlap so neither switches off exactly at the crossing point. That overlap is the point: without it you get crossover distortion right where the signal is quietest and the ear is most sensitive. Efficiency around 50 to 60%, and it was the standard for touring amplifiers for decades.

Class D is not analogue in the middle at all. The output devices are switched fully on or fully off at a high frequency, with the proportion of on-time following the signal, and a passive filter at the output turns that back into a waveform. A device that is fully on or fully off dissipates very little, so efficiency runs past 90% — which is why modern amplifiers are a fraction of the weight and why an amp rack no longer needs its own air conditioning.

All three still put a fixed voltage across a load through some output impedance, which is why the speaker impedance arithmetic is the same whatever the class, and why halving the load impedance still asks the supply for twice the current.

Inside the output stage

Push-pull, damping factor, slew rate, and which number on the sheet is lying to you

“Two devices, one per half of the waveform” has a name: a push-pull output stage. One device sources current into the load on the positive half, the other sinks it on the negative. They are almost always the complementary pair the components section named — an NPN and a PNP, or their MOSFET equivalents — and the bias current through both at the crossing point is the single adjustment that separates a clean amplifier from a harsh one.

Which device type matters more than the marketing suggests. A bipolar output device is driven by current and gets more conductive as it heats, so it will run away and destroy itself unless something watches its temperature and pulls the bias back. That something is a bias servo, usually a transistor bolted to the same heatsink, and it is a real failure point on old amplifiers. A MOSFET is driven by voltage and, in the region amplifiers use, gets less conductive as it heats, so it self-limits. That is why MOSFET outputs are more thermally forgiving and why they took over the touring market.

Damping factor is the number most often quoted and most often meaningless. It is just the load impedance divided by the amplifier’s output impedance, and it describes how firmly the amplifier can resist the voltage a moving cone generates back into it. A figure of 500 sounds authoritative. Then you put 30 m of cable between the two, and the cable’s own resistance is in series with the amplifier’s, so the damping factor the driver actually experiences is 8 divided by (the output impedance plus the cable), which for a couple of tenths of an ohm collapses 500 to something in the tens. The cable sets the damping factor, not the amplifier. That is the whole argument for short speaker runs and fat conductors, and the voltage drop calculator is the same arithmetic.

Slew rate is how many volts the output can move in a microsecond. Too low and a fast transient comes out as a ramp rather than an edge, which is not the same distortion as clipping and sounds like smearing rather than hardness. It is rarely the limit on a modern amplifier and it is worth knowing the word, because it is the one specification that describes a failure in time rather than in level.

Storage

Tape, vinyl, and three discs that look identical

Recording is the other half of the analogue story, and each medium is a different physical trick.

Tape is magnetic domains in an oxide coating, aligned by a recording head as the tape passes it. The signal is continuous, so tape has the analogue signature exactly: lovely graceful saturation when pushed, and a noise floor that gets worse with every generation of copying. Capacity is linear density times track count times length, which is why a wider tape running faster sounded better and cost more of everything.

Vinyl is a groove whose wiggle is a direct mechanical analogue of the waveform — the most literal analogue medium there is. You can see the signal with a microscope.

Optical discs are where it gets calculable, and where the answer to “why does a Blu-ray hold more than a CD” is genuinely satisfying. All three are the same 120 mm polycarbonate disc, spun the same way, read by a laser that follows a spiral of pits. What changed is how small a spot the optics can focus, and that cannot beat diffraction: spot ≈ 1.22 λ / NA.

Infrared to red to blue-violet takes the wavelength from 780 to 650 to 405 nm, and the numerical aperture rose from 0.45 to 0.60 to 0.85. Both help, and together they take the spot from about 2.1 µm to 0.58 µm. Density goes with the square of that, because a disc is a surface — about 13×. But capacity actually rose about 36×, and the missing factor of 2.7 is not physics at all. It is coding: better modulation, better error correction, more of the disc used. The optics gave one factor and the mathematics gave another.

Why an old tape sounds wrong

Three faults that are not the recording, and one that is not EQ

If you are ever handed a reel and asked to get the show audio off it, four things account for nearly every complaint, and only one of them is the tape being old.

Azimuth is the angle of the head gap relative to the tape’s direction of travel, and it should be exactly perpendicular. If the machine that recorded and the machine playing back disagree even slightly, the top of the track is read a fraction of a moment before the bottom, so high frequencies partly cancel across the width of the track. It presents as a dull recording. It is not dull — it is a phase error, and no amount of top-end EQ fixes it, because you are boosting something that is cancelling itself. Aligning the playback head to the tape recovers it completely, and this is the single most common reason an archive tape sounds worse than it is.

Wow and flutter are the same fault at two speeds: the tape is not moving at a constant rate. Slow variation, below roughly 4 Hz, is heard as pitch drifting and is called wow; faster variation is heard as a roughness or warble on sustained notes and is called flutter. Wow is usually a capstan or a slipping pinch roller; flutter is often a bearing or tape dragging on a guide.

Bias is the one that is not a fault at all. Magnetic tape does not respond linearly to a small signal, so recorders add a strong high-frequency tone — far above hearing, typically 100 kHz or so — to push the signal into the part of the curve that behaves. Bias level is a real alignment setting: too little and the recording is distorted and thin, too much and the top end is gone. A machine biased for one tape formulation and fed another is mis-set by definition.

Noise reduction is where a tape can be actively misread. Dolby A, B, C and SR all work by compressing certain bands on the way in and expanding them by the exact inverse on the way out. Play a Dolby-encoded tape without decoding and it is bright and forward; decode a tape that was never encoded and it is dull and lifeless. Worse, the expansion tracks against a reference level, so if the machine’s alignment does not match the one that recorded it, the decode mistracks and the sound pumps. Getting an encoded tape back needs the type and the reference level, and the box is often the only place either was written down.

What was lost

And what genuinely was not

It is worth being unsentimental. Digital control did not win because it sounded better; it won because it was cheaper, thinner, recallable and did not degrade. A show that took two operators and a 48-core loom takes one console and a pair now, and the cues survive being written down.

What was genuinely lost is smaller than nostalgia claims but it is not nothing. Graceful degradation, so a fault produced a worse signal rather than none. One wire per channel, so a fault was localised by construction. And an interface where the state of the show was visible as physical positions across a surface, which is a real cognitive advantage that no amount of screen has fully replaced.

What was not lost: the analogue layer never went anywhere. Every input has a preamplifier, every output has a reconstruction filter, every dimmer still switches a real current, and every one of the five components above is still in every box on the rig. The digital part sits in the middle of an analogue sandwich, and the two ends are where the difficult faults live.

Where this goes next

The calculators on this page

The optical spot calculator does the diffraction arithmetic and separates what the physics gave from what the coding did. RC time constant gives the corner and the settling times together, because they are one fact. Transformer ratios shows why impedance follows the square. And power, earth and the things that trip is the same components at rig scale, where getting them wrong hurts somebody.