What Is Gain?

We need to talk about gain. Gain is not just one of the most important, and widely misunderstood, concepts in amplification. It’s also, perhaps, the most foundational and useful type of node belonging to the Web Audio API.

Gain is relative

This is the first thing to get your head around. In either an analog or a digital context, gain is how much more you add. That value is necessarily relative to how much you’re getting in the first place. Adding more signal = amplification.

You can add more in multiple different places along your signal chain. These places are sometimes called gain stages. Each stage makes a relative adjustment, in series.

The gain value of a gainNode is set as a number. A gain of 0.75, followed by a gain of 0.5, followed by a gain of 1.5, would give you an outgoing value of 0.75 × 0.5 × 1.5… or 0.5625.

<chain-blam>
  <gain-blam gain="0.75"></gain-blam>
  <gain-blam gain="0.5"></gain-blam>
  <gain-blam gain="1.5"></gain-blam>
</chain-blam>

For a more practical application, consider the feedback loop that creates an echo effect. Inside this loop, a single gain stage is passed through multiple times.

A lowish gain value, like 0.5 gives noticeably diminishing returns. This approximates a true echo, which loses energy as it is partially absorbed by the surfaces it bounces off.

<echo-blam 
  feeback="0.5" 
  cutoff="2000"
>
</echo-blam>

The “Light Lo-Fi Listen With LFOs” example uses an <echo-blam> for the plucked, guitar-like instrument.

Giving the echo effect’s gain (often labeled “feedback” in a commercial unit) a value approaching 1 causes it to go into a state called self-oscillation. In this state, echos compound and get louder, ultimately leading to distortion.

You get distortion when you hit a ceiling

The next part is really important. It’s important because it accounts for the entire history, and enduring popularity, of rock ‘n’ roll over the best part of a century. Which is a pretty big deal.

Sometimes you ask for more and there’s no more to give. That sounds bad, right? And, yeah, theoretically it is. But it’s also the simplest way to create distortion. Sometimes we like distortion. Oftentimes, we love it. We apply gain to create distortion so often, the two terms are often used interchangeably.

When you tell an amplifier (gain stage) to give you 1.5, but all it can manage is 1.25, then 1.25 is its ceiling. There’s nowhere for the remaining 0.25 to go, so it gets decapitated (or “clipped”). The more powerful the amplifier, the less likely decapitation will occur. Hence, powerful amplifiers are said to have a lot of headroom.

Sound decapitation isn’t fatal, though. If it were, rock ‘n’ roll would fall silent. Instead, lopping the top off just changes the shape of the waveform. How abruptly the head is lopped off (how hard the clipping) helps determine the perceived character of the distorted sound.

A high input gain combined with extremely hard clipping can produce something akin to a square shaped wave. This is why some fuzz guitar effects can make your guitar sound like a synthesizer: synthesizers often produce sounds based on basic square waveforms.

Nobody likes unintended distortion, but nobody in revolutionary France was getting decapitated by accident. Distortion is transformation; the creation of a new, more complex and textured sound from the old. To conservatives, transformation is transgression. The bold, distorted sound of rock ‘n’ roll is, therefore, “the devil’s music”.

For some devilishly distorted beats, using HYPERBLAM’s signal-decapitating <saturator-blam> element, listen to the “AutoMayhem” example.

Distortion sounds loud

As I’ve already established, distortion only happens when amplitude hits a ceiling. It’s literally the effect of sound not getting louder. So why does distortion sound loud?

Again, it’s a question of relativity. The more gain you apply, the more of the sound hits the ceiling. First it’s just the louder parts of the sound, then it’s the quieter parts too. The ceiling doesn’t move, but more of it is covered. It’s more maximum at once, which is perceivably louder.

Both distortion and compression add perceived loudness. A compressor just uses a more complex algorithm, flattening the dynamic range without introducing clipping.

The “loudness wars” are where producers competitively add compression, to make their tracks sound louder than their peers’. Unfortunately, this results in popular music sounding, overall, more flat and lifeless. Where the quieter sounds aren’t allowed to be quiet, the loud ones don’t come off as loud. Remember: it’s all relative.

Gain isn’t just for gain

When working with the Web Audio API, gain nodes aren’t just useful for affecting gain. A gain node with a gain value of 1 transparently reproduces any signal sent to it. As such, it can be used as a junction box.

Each of the effects available in HYPERBLAM are based on a primitive Box class that implements wet and dry parallel paths. The mix of the effect is the ratio between the wet (affected) and dry (unaffected) gain nodes.

As in the “Baby’s First Blam” example, the mix can be calibrated using the (aptly named) mix property.

<reverb-blam mix="0.5">
  <sample-blam src="/sounds/IRs/stadium.mp3"></sample-blam>
</reverb-blam>
...
<dial-blam to="reverb-blam" prop="mix">
  <label>
    reverb
    <input type="range" min="0" max="1" step="any">
  </label>
</dial-blam>

Under the hood, any change in the mix apportions gain to the wet and dry signals. The dry signal is always 1 minus the wet signal:

setMix() {
  this.setParam('wet', this.mix);
  this.setParam('dry', 1 - this.mix);
  this.prevMix = this.mix;
}

The this.prevMix value is handy when it comes to bypassing the wet signal. When bypass is removed, the former mix can be restored.