What Is Multiband Compression & How Do MB Compressors Work?

Multiband compression is a powerful audio processing tool that can truly improve your mixing when understood and used correctly.

In this article, we'll discuss multiband compression in great detail and develop a solid understanding of when, why and how to use it in our own audio projects.

---

Let's Begin With A Definition Of Compression

Multiband compression is simply a series of independent frequency bands, each with its own independent compressor. Understanding how a single compressor works will allow us to effectively understand how a multiband compressor works (at least in terms of the compression portion of the design).

In other words, if we understand how compression works in general, we can understand how multiple compressors can work together. So then, let's define compression.

Dynamic range compression, in audio, is the process of reducing or “compressing” the dynamic range of the audio signal (hence the name). The main purpose of the process is to reduce the difference between the highest and lowest amplitudes of the audio signal.

In general, this means attenuating only the loudest parts of the signal while leaving the quieter parts unaffected.

In order to attenuate the “loudest parts” of an audio signal, two key questions must be answered:

• What are the “loudest parts”?
• By how much should the “loudest parts” be attenuated?

A compressor answers these questions with the threshold and ratio parameters, respectively.

What is the threshold of a compressor? The threshold of a compressor is a set amplitude limit that dictates when the compressor will engage and disengage. As the input exceeds the threshold, the compressor kicks in (with its given attack time). As the input drops back down below the threshold, the compressor disengages (according to its release time).

What is the ratio of a compressor? The ratio of a compressor compares the number of decibels the input signal is above the threshold to the number of decibels the output signal is above the threshold. In other words, it is the relative amount of attenuation the compressor will apply to the signal.

Other compressor parameters worth mentioning are the following (I’ve added links to in-depth articles on each parameter):

• Attack Time: the amount of time it takes for a compressor to engage/react once the input signal amplitude surpasses the threshold.
• Release Time: the amount of time it takes for the compressor to disengage (to stop attenuating the signal) once the input signal drops below the threshold.
• Knee: the transition point around the threshold of the compressor where the output becomes attenuated versus the input.
• Makeup Gain: the gain applied to the signal after the compression takes place (typically used to bring the peaks of the compressed signal up to the same level as the peaks pre-compression).

In order to attenuate the audio signal, a compressor must have a gain reduction circuit. The types of gain reduction circuits vary from compressor to compressor and largely define the “type” of the compressor in question. The most common compressor types include (I’ve added links to in-depth articles on each type):

• Diode bridge compressor: uses a diode bridge as the gain reduction element.
• FET compressor: uses a field-effect transistor as the gain reduction element.
• Optical compressor: uses an optical assembly (a voltage-controlled light source and light-dependent resistor) as the gain reduction element.
• PWM compressor: uses a pulse width modulator as the gain reduction element.
• Variable-mu/tube compressor: uses a remote cut-off vacuum tube as the gain reduction element.
• VCA compressor: uses a voltage-controlled amplifier as the gain reduction element.

A control signal must effectively control these gain reduction circuits and cause them to attenuate the audio/program signal appropriately. The control signal for these elements, in any compressor type, is referred to as the sidechain signal.

Generally, the control signal/sidechain is based on the input/program audio signal though it may be taken from an external source. Either way, the sidechain signal will go through some sort of level detector/rectifier (peak level, RMS level or otherwise) and come out as a variable DC control signal.

These sidechain signal paths will also alter the signal in some cases to achieve some or all of the aforementioned parameters (threshold, ratio, time controls, etc.).

The overall circuit block diagram could like something like this (with feed-forward topology):

Or, alternatively, something like this (with feedback topology):

---

What Is A Multiband Compressor?

Okay, so now we understand the basics of what constitutes a compressor. Now let's look at how multiple compressors can be used together in a multiband compressor.

A multiband compressor can be thought of as several compressors in one, with each compressor acting on its own defined band of frequencies. Each band will generally have its own set of parameters, including threshold, ratio, attack, release and make-up gain.

Usually, each of the compressors in a multiband compressor will be of the same type (FET, VCA, optical, etc.) in order to maintain some amount of congruency across the circuitry and sonic performance of the unit.

With that being said, the pressing question is, “how are the bands defined in a multiband compressor?”

The answer to that question is via a crossover network.

A crossover network is effectively a collection of filters that work together to split a full audio signal into multiple frequency bands without having an effect on the output when comparing the original sound to the sound if all bands are outputted at unity gain.

The number of bands can vary from compressor to compressor. 3-band, 4-band, and 5-band multiband compressors seem to be the most common, and certain plugins can even alter the number of bands in the compressor.

The lowest band is generally filtered with a low-pass filter that allows all frequencies below a certain cutoff frequency (f_C) to pass through unaffected.

Here's a visual representation of a first-order (6 dB/octave) low-pass filter (LPF). Note the amplitude-frequency graph on top and the phase-frequency graph on the bottom (this will be referenced shortly).

Similarly, the highest band is generally filtered with a high-pass filter that allows all frequencies above a certain cutoff frequency (f_C) to pass through unaffected.

Here's a visual representation of a first-order (6 dB/octave) high-pass filter (HPF). Note, once again, that the amplitude-frequency graph on top and the phase-frequency graph on the bottom:

All other bands will be filtered with band-pass filters (BPF), which, as the name suggests, pass a band of frequencies between two cut-off points. Here is an amplitude-frequency and a phase-frequency graph for a simple second-order (6 dB/octave) band-pass filter:

Note that, in all filters, the cut-off frequency takes place at the 3 dB attenuation point (half-power point) from the original level (power). When splitting audio into separate bands with a crossover network, adjacent bands crossover (overlap) at their cutoff frequencies.

Let's consider a visual representation, then, of a 4-band crossover. Note that, in this case, the filter slopes are at -24 dB/octave, making the HPF and LPF fourth-order and the two BPFs eighth-order. The crossover frequencies are as follows:

• 100 Hz: red LPF to orange BPF
• 1,000 Hz: orange BPF to green BPF
• 5,000 Hz: green BPF to blue HPF

Increasing the filter order will effectively increase the slope of the filter by an additional 6 dB/octave per order increase but will also increase the phase shift by an additional 90º per order increase.

So, generally speaking, as we increase the order, we have greater independence of each band from its adjacent bands at the expense of phase distortion in the overall output at and near the crossover frequencies (where the cutoff frequencies of two crossover filters overlap).

Of course, linear phase EQs could be used in more advanced plugins to eliminate phase distortion, but this would require a lot of CPU and is not often the most practical approach.

Once the bands are split with defined roll-offs/slopes/filter orders and crossover frequencies, a compressor is assigned to each and every band. Each band can then be compressed by varying amounts, largely independent of every other band (except for the small region at/near the crossover points, where there may be some interplay between compressors).

In fully functional compressors, each band can be compressed with a wide variety of parameters. These parameters may include:

• Input Gain: the gain applied to the signal within the band before the compression takes place (typically used to drive the compressor harder or as pseudo-EQ).
• Threshold: the set amplitude limit that dictates when the compressor will engage and disengage. As the input exceeds the threshold, the compressor kicks in (with its given attack time). As the input drops back down below the threshold, the compressor disengages (according to its release time).
• Ratio: the amount of decibels the input signal is above the threshold to the amount of decibels the output signal is above the threshold. In other words, it is the relative amount of attenuation the compressor will apply to the signal.
• Attack Time: the amount of time it takes for a compressor to engage/react once the input signal amplitude surpasses the threshold.
• Release Time: the amount of time it takes for the compressor to disengage (to stop attenuating the signal) once the input signal drops below the threshold.
• Knee: the transition point around the threshold of the compressor where the output becomes attenuated versus the input.
• Makeup Gain: the gain applied to the signal within the band after the compression takes place (typically used to bring the peaks of the compressed signal up to the same level as the peaks pre-compression or as pseudo-EQ).
• Lookahead: a delay technique that allows the sidechain to compress the program signal earlier than otherwise possible for improved accuracy.

A multiband compressor is indeed a powerful tool for mixing with any or all of that functionality on a per-band basis.

Multiband compression is a fantastic tool and helps avoid over-compression in signals that are particularly well-represented in certain frequency bands. By adjusting a multiband compressor, we can achieve a smoother gain reduction across all bands even if one of those bands is over-represented in the frequency response.

Multiband compression also works great as a de-esser to help eliminate problem frequencies (particularly sibilance) without entirely EQing them out of the signal. By narrowing a band around a particularly offensive frequency and compressing it relatively hard, we can effectively reduce the offensiveness of such a frequency band. In addition to this, we can use the gain (input and/or makeup) to turn down that band in a sort of pseudo-EQ kind of way.

Suppose a vocalist is particularly active in the vocal booth, moving closer and further from the mic. In that case, the proximity effect could cause an accentuation of bass frequency as the singer is close that fades as the signer moves further away. A multiband compressor can help smooth this out, bringing the low-end down during the close times and leaving it be during the far times, all without affecting the mid-range and high frequencies.

The practical uses for multiband compression are virtually limitless. Any time a specific band could use a different compression style versus the other bands, a good multiband compressor is a superb tool.

Whether it's an active vocalist, an offensive frequency, an instrument/source that's weighted in one band, or even a mix bus, a multiband compressor can really help fine-tune the compression applied to the particular tracks in a mix.

---

Multiband Compression Vs. Dynamic EQ

Before we get into the examples, I'd like to make the note that dynamic equalization performs very similarly to multiband compression and can be used in many of the same applications to better or worse effect.

To learn more about the differences between dynamic EQ and multiband compression, check out my YouTube video on the matter:

---

Multiband Compressor Examples

Before we wrap things up, it’s always a great idea to consider some examples. Let’s look at 4 different multiband compressors to help solidify our understanding of this type of compression.

In this section, we’ll discuss:

• 500 series multiband compressor: SM Pro Audio MBC502
• 19″ rack unit multiband compressor: Maselec MLA-4
• Multiband compressor effect pedal: EBS MultiComp Blue Label
• Multiband compressor plugin: Vengeance Sound Multiband Compressor

SM Pro Audio MBC502

The SM Pro Audio MBC502 is a 500 Series optical multiband compressor with two distinct bands separated by a 24dB/octave Linkwitz-Riley 4th-order crossover filter. Unfortunately, this unit has been discontinued.

The centre frequency that separates the two bands can be selected via a 16-stage frequency range selector. One band will be high-passed at the set frequency, and the other will be low-passed at the set frequency.

Each band features its own independent input level, compression amount (threshold), attack time and release time controls. Each band can be muted or bypassed, and the entire unit also has a master bypass switch. The master level can be monitors via the MBC502's VU meter.

Maselec MLA-4

The Maselec MLA-4 is a stereo multiband compressor and expander with 3 independent bands.

Starting with the inputs, the MLA-4 features a left and right input, each with its own level/gain control.

The crossover frequencies can be selected between the low and mid bands (100, 200, 400 or 800 Hz) and between the mid and high bands (1.5k, 3k, 6k or 12k). The crossover filters have slopes of 6 dB/octave.

There a 5 different sidechain links available in the MLA-4:

• L -> M: The low band signal controls the compression of the mid band as well.
• L -> H: The low band signal controls the compression of the high band as well.
• L -> MH: The low band signal controls the compression of all bands.
• Off: Each band controls the compression of itself.
• Linked: Each band is compressed via the same sidechain signal which is made up of all bands.

There is also a master threshold control by which each band's threshold will be referenced to.

Each of the MLA-4's bands has its own ratio (for expansion and compression), threshold, attack time and release time parameters. For expansion purposes, each band also has an 11-position gain control capable of applying ±2.5 dB.

Note that expansion ratios have the larger number second, and compression ratios have the larger number first.

Each band has its gain reduction LED meter, and the output of the MLA-4 can be set to output all three bands, each band individually or a signal that bypasses the compression/expansion circuits completely.

EBS MultiComp Blue Label

The EBS MultiComp Blue Label is a multiband compression pedal designed for pedalboards.

This relatively simple pedal offers three potentiometer/knob controls and 3 modes of operation.

The Comp knob adjusts the compressor's ratio between 1:1 and 5:1; the Sensitivity knob adjusts the compressor's threshold between -25 and +6 dB, and the gain knob offers up to 15 dB of makeup gain.

In Normal mode, the MultiComp Blue Label acts as a normal “one-band” compressor. In Tube Sim mode, the pedal engages its tube simulator circuit that adds additional harmonics and warmth to the sound.

In MB mode, though, the pedal becomes a 2-band compressor where the bass and treble compressors work independently. The band balance trimmer inside the pedal body controls the balance in sensitivity between the higher and lower band so that one set of controls will affect the two bands differently.

Vengeance Sound Multiband Compressor

The Vengeance Sound Multiband Comressor is a multiband compressor plugin. It may look intimidating, but this powerful digital audio tool isn't so difficult to navigate once we understand how it works.

This multiband compressor plugin offers 1 to 4 bands of compression. The slope/roll-off of the band crossover linear-phase filters can be selected from 6, 12, 24, 36, 38 or 96 dB/octave settings. The crossover low-pass, band-pass and high-pass filters will not affect the phase of the signal whatsoever. The crossover points between bands can be adjusted by simply clicking within the multiband display.

Presets can be chosen, input and output levels can be monitored, and parallel processing is made possible with the mix knob.

Each band has its own dedicated and independent controls for compression and limiting. The compressor controls include the typical threshold, ratio, attack, release, boost (makeup gain) and hard/soft knee switch. The limiter parameters are output level and release time. Gain reduction is shown for both effects, and each band also has an additional band input/output scope for extra monitoring.

Each band can be soloed or muted and has its own preset and mix control.

This is just scratching the surface of this powerful plugin. Check it out for yourself for more details!

---

Related Questions

Should you EQ or compress first? There is great debate as to whether an equalizer or a compressor should come first in the audio signal chain. There's no rule stating that either should come first. However, in general, you'll likely get the most out of the EQ and compressor if you follow these standards:

• For tonal shaping, it's often best to compress beforehand to avoid having to alter the compressor settings.
• For audio signals with that require significant filtering, it's best to EQ first so as to not feed the compressor with unwanted frequency content.

What is dynamic audio equalization? Dynamic EQ is a type of equalization where the EQ of certain frequencies is triggered dynamically as those frequencies surpass a set amplitude threshold in the audio signal. Dynamic EQ, like a compressor, will have threshold, attack and release settings to alter the EQ of a signal dynamically.