Digital Equalisation
An introduction to the most common types of equaliser available within software, and equivalent purpose-built hardware units, suitable for live sound reinforcement, recording and mixing.
Introduction
Equalisers are one of the most common and powerful of all audio processors. Familiar to most of us as the Treble and Bass controls on an amplifier, or the graphic equaliser on a hi-fi, the channel controls on a mixing desk, or tone controls on a TV or radio, the bass boost and ‘loudness' controls on a car stereo... the list goes on.
Equalisation (also known as EQ) allows us to alter the relative levels of various frequency bands within the audio spectrum, correcting or removing unwanted artefacts and enhancing those parts of the signal which are more important or desirable. This guide will introduce and explain some of the common types of equalisation available within software, used to manipulate digital signals, and equivalent purpose-built hardware units, suitable for live sound reinforcement, recording and mixing, and will explore how they can be used to best advantage.
We will also look at the features and particular benefits of different approaches to equalisation - specifically graphic and parametric EQ, filtering, dynamic EQ and enhancement. We will move from simple and surprisingly effective tips for the new user to more advanced techniques, and will cover some of the common problems to which equalisation is the answer.
Filters
Most common types of equalisers are made up of banks of filters, which separate the sound into different frequency bands for processing. Filters are generally defined by which frequency bands they allow to pass through: thus a ‘low pass' filter allows only frequencies below the cutoff frequency to pass, and a ‘high pass' filter those above. A ‘band pass' filter combines the two, attenuating frequencies both above and below a specified band, which is defined by both high and low frequency limits. [A ‘band reject' does the opposite, attenuating within the band's limits, rather than outside them, but this is a fairly uncommon type of filter]. Low-pass Filters (LPF) and High-pass Filters (HPF) are by far the two most common types, and the combination of these two core filters, in both parallel and series, allows the construction of much more complex filters.
Highpass and Lowpass filters are moveable throughout (and indeed beyond) the audible range. The degree to which the levels of frequencies beyond the cutoff are attenuated is measured by how many decibels the levels of frequencies beyond this limit are reduced for every octave the pitch exceeds the filter's cutoff point.
Lowpass filter (1kHz, 24dB/octave)
Highpass filter (1kHz, 24dB/octave)
Bandpass filter (500Hz-2kHz, 24dB/octave)
In these examples, the level of a sounds falling 1 octave beyond the cutoff frequency is cut by 24dB (decibels) for every octave of pitch change. Decibels are a logarithmic unit, but an easy (and quite accurate) approximation is that the energy output halves for every 3dB drop in level. By this estimation a one octave change in pitch is 1/2^8, or 1/256 as loud as an unfiltered frequency. This constitutes a 'steep' filter.
Types of Equaliser
Graphic Equalisers
Graphic Equalisers (or 'graphics' for short) use banks of filters to divide the audio spectrum into a number of fixed ‘bands' of equal width, which can range from a 5-band ‘hi-fi style' graphic to a 31-band professional unit (as pictured), where each band is only 1/3 of an octave wide (think of the tune to ‘mary had a little lamb' - that's the bandwidth of each slider). Because of the clinical precision afforded by their many narrow bands, they are generally used in a live sound environment and Public Address (PA) as corrective tools. They are not usually used for shaping a recorded signal in the recording studio, in whose more sensitive environment their overlapping banks of filters can have a detrimental effect, and we concentrate here on their use in tackling feedback and other live sound problems. Studio Equalisers are usually of the parametric type (see below).
Audient ASP231 - Stereo 31-band Graphic Equaliser [image used with permission]
Because of their design, it is quite easy for the user to visualise the effect of the levels of the various bands, as the layout of the sliders is analagous to an audio spectrogram - the reason that they are known as "graphic" equalisers in the first place!
Parametric Equalisers
Parametric EQs generally only offer between one and six bands [though digital versions now sometimes offer more]. In contrast to a Graphic EQ, each band can be moved up and down the audio spectrum, either in steps or with a continuous control, to cover a desired area, and all but the most basic models also offer the ability to to alter the bandwidth [or ‘Q'] over which the cut or boost will be applied. ‘Q' - standing for 'quality' - is an alternative term to refer to the width of each EQ band, where high Q = narrow bandwidth, and low Q = wide bandwidth. Some more advanced Parametric EQs also allow the user to change the behaviour of the chosen band , and switch it between a shelving and peaking type, and the slope of the shelf.
a professional hardware Parametric Equaliser: Focusrite Red 2 (image used with permission)
a Software Plug-in EQ: Logic Studio Channel EQ
Peaking bands add or subtract frequencies around a centre frequency (see fig.3 and fig.4), whereas shelving bands cut or boost all frequencies either above or below the shelving frequency (depending on whether they are high or low shelving type), creating a 'shelf' in the spectrum (see fig.1 and fig.2)
Examples - 10 decibel (dB) boost applied at 1kHz with different types of parametric band:
fig.1 High shelving band
fig.2 Low shelving band
fig.3 Peaking band - wide bandwidth (low Q)
fig.4 Peaking band - narrow bandwidth (high Q)
Premium parametric equalisers are totally essential in creating, mixing and mastering audio to a professional level, and are one of the professional studio engineer's sine qua non. Truly professional equalisers have long been beyond the budget of all but top-end recording facilities, and many analogue units remain so.
However, digital technology now allows the simulation of vintage and classic EQ devices, and their recreation as software plug-in effects for use within digital audio workstations, for a fraction of the cost of equivalent hardware. These computer models, while still surpassed by the ‘real thing' in terms of sound quality (at least in this writer's opinion), offer an accessibility and ease of use within the digital environment - not to mention price advantage - which put them within reach of a far wider group of users. Digital modelling also enables the creation of new and powerful tools for the digital audio engineer, impossible or impractical to create with electronic circuitry.
Here we look at some of the more common audio applications available, and - in addition to general advice and guidance on good practice - give specific examples of when and how to use EQ processing to improve your recordings and eliminate problems.
Feedback - Search and Destroy
What is feedback ? We all know the screetching or whining sound of a microphone feeding back during a concert or a speech, but what causes it and how can it be eliminated?
Feedback - as the name would imply - is caused by the signal emitted from a speaker being picked up by a microphone, and fed back to the same speaker, whose signal is again picked up by the microphone, whose signal is fed back to the speaker, whose signal is picked up by the microphone... etc etc - in a continuous loop, which results in a sine wave generated at the resonant frequency of the feedback system. The reason that the initial sound generated is centred on a single frequency is down to the combination of speaker, microphone and room acoustics resonating most strongly at this frequency,
The process of finding and eliminating feedback frequencies is often called ‘ringing out the room', and is a surprisingly simple and intuitive process, which will make a vast difference to the quality of the amplified sound, and any recording made from it. Your tool will be a graphic equaliser (the more bands the better) - if you are having problems with feedback, this is the only way to tackle it without turning down the volume or repositioning your equipment. Alternatively, there are relatively new digital units called ‘feedback destroyers' which can carry out this process automatically. The manual method, if correctly done, will give a better result, but the automatic is quicker and simpler for the inexperienced user. Obviously your available equipment will also dictate which approaches are available to you.
The Equaliser or feedback destroyer should be placed last in the audio chain, directly before the amplifier, so that its effect is not in any way diluted or distorted by additional processing.
Manual Feedback Suppression
The first thing you will need to do when tackling feedback is establish where it is occurring, and to do this you will have to induce it. Set up your microphone(s) to your satisfaction [see our Microphone Technique document for guidance] then gently start turning up the volume of the amplifier of the PA (starting from zero). There will come a point where the system starts to feed back, and a whining or humming noise will quickly fade in. This is your first feedback frequency, which you will have to identify and then attenuate, so turn the volume gently back down until it disappears, and then try to judge the point where the feedback starts, so you can fade it back in to an acceptable level, while you work on its elimination.
There follows an iterative process of turning levels up and down. Identifying the offending frequencies is really a matter of practice, experience, and listening, but the following process should allow even the user with no previous experience to improve the response and sound of their system by following the correct method. Be warned, it is time-consuming, and occasionally frustrating, so allow yourself time, and be patient.
An unfortunate part of the process is the necessity of inducing feedback into the system - never a pleasant sound - so always have one hand on the main volume control for the speakers. Feedback can be very unpredictable, so you should always be ready to reach for this control if it starts to get out of hand!
Step 1: Set up your equipment, following steps from our Microphone Technique guide on metering and setting levels. Leave the output volume at or near zero throughout this process, or better still use headphones to monitor the signal. Get the unamplified signal sounding as good as you can
Step 2: Turn up the output volume of the Public Address system gently until feedback begins to occur. Turn it back down, then ride the feedback point to allow you to control it. Try to leave the system feeding back a consistent, but not too loud tone.
Step 3: Try repositioning mics/speakers to see whether you can eliminate some of the main frequencies by this simple method. Once you have improved things as much as possible, leave them alone - once you start the equalisation process, speakers should be static, and mics as much as is possible. Radio mics will by their nature be mobile, so allow them to move around the space they will be required to cover. Acoustics may lead to pockets of resonance and feedback in different parts of the room.
Step 4: Turn up the output volume of the Public Address system gently again, until feedback returns. Turn it back down, and ride the feedback point to allow you to control it. As before, try to leave the system feeding back a consistent, but not too loud tone.
Step 5: Identify the offending frequency by going along the sliders of your graphic EQ, turning each one down, then returning it to its former level. At some point you will hit on the right one to match your feedback tone, and it will disappear. Again, turn it back up until the feedback starts to build again, then pull it back down so that you are cutting enough signal to kill the feedback, plus a little bit to spare.
Step 6: Return to the main volume control, and turn it up gently again, until the next feedback frequency is found (NB - this may be the same as the last one, depending on how resonant it is. If so, notch its fader down a little more and continue) If it is a new frequency, return to Step 4. Repeat until the output level is a bit higher than you will need for the actual performance. This should give you some breathing space on the day.
Be aware that once a room is filled with people, the acoustic will change. Thanks to the bodies breaking up the regular nature of the room, it will usually be less resonant than when empty, and so will assist in reducing feedback. Still, you should always have an eye on the main volume control, as your ‘panic button', and leave yourself a bit of headroom to spare.
Vocal Enhancement
Equalisers, along with compressors, are one of the two most useful tools for enhancing the quality and intelligibility of spoken word recordings. They can be used to enhance the richness of a vocal recording, by gentle boosting of the resonant qualities of a speaker's voice, and to improve presence and clarity by enhancing the higher frequencies.
In addition to their potential for vocal enhancement, equalisers can also be deployed correctively - in a recording as well as live sound environment. High-pass filters can remove frequencies which fall outside the range of the human voice, and so improve isolation from background noises. Notched bands can be used to ‘tone down' or remove unwanted acoustic resonance generated by the recording space, and in combination with some more advanced compression techniques, they can also be used to remove sibilance (the hissing 'ssss' sound some voices can produce on combination with certain microphones) - a process called ‘de-essing' - and other unwanted vocal artefacts.
Many microphone preamplifiers, and some microphones, feature a ‘low cut' filter (aka high-pass filter), which will cut all signal below a frequency - usually situated at around 80-100Hz. These filters can lend a ‘drier' or crisper sound to the vocal, and are a very simple tool to use. Even in the absence of audible low frequency artefacts, they can still help to improve presence.
Other significant areas where vocal recordings will sometimes benefit from a gentle boost are in the midrange (200-500Hz), to create a richer sound if the recording sounds tinny or scratchy, and in the high frequency range (10kHz+) to create a greater sense of ‘air', if the recording sounds dull or indistinct. The upper midrange (1kHz-5kHz)can be a minefield, and too much boost in this area can lead to a harsh or'honky' vocal sound, but again if the recording lacks clarity or presence, a gentle boost in this range may help. There is no hard-and-fast rule here, and every voice will require different treatment. Use your ears and your judgement.
You may notice that I have repeatedly used the word ‘gently' when referring to EQ boosts, and this is intentional. A general rule when using an equaliser is that it is better to cut than to boost where possible. EQ cuts are often most effective when applied deeper, but over a narrower bandwidth than an equivalent boost, and conversely, boosts should be gentle and over a wider bandwidth, or you risk creating a quite unnatural sound. In this sense the ear is more sensitive to presence than absence.
Mastering
High-end parametric equalisers are often used as a final stage in processing master versions of musical and other audio recordings. They are used to adjust the final overall balance of harmonics and timbres, and to lend a final 'sheen' to the master. Use of mastering tools at a professional level is a specialised skill, and requires exceptional analytical hearing, but even the enthusiast can improve their stereo masters with a little well-judged equalisation, and many audio programs (Logic Audio for example) include preset Mastering EQ settings for different genres and playback destinations. Once again, trust your ears as to whether this treatment suits your material.
Spectral Analysis
Spectral Analysis is the term used for inspection and quantification of the audio 'spectrum', which represents all frequencies of sound as separate pitches, in the same way that the visual spectrum represents all frequencies of light as different colours. In the same way that altering the colour balance of a photograph can dramatically alter our perception of its content, its perceived ‘quality', and the psychological response it elicits, altering the audio spectrum of a signal with an equaliser or filter can transform its effect on the listener, or alter its place within its audio context.
The human ear has an approximate hearing range from 20Hz (hertz = cycles per second) to 20kHz (kilohertz = thousands of cycles per second). Any sound source vibrating slower than 20 times per second will be effectively inaudible (though you might feel it, similarly to the sometimes unsettling effect of a 32ft + organ pipe, whose fundamental frequency can be around 15Hz), as will those oscillating faster than twenty thousand times per second (which you would hear - if at all - as a very very high whining or whistling noise). Different creatures have different hearing ranges - hence ultrasonic dog whistles and mouse deterents. The graphs below are calibrated logarithmically within the average human range (20Hz-20kHz).
All sounds can be broken down into simple elements (sine waves) at different and distinct frequencies and levels, using mathematical analysis of the signal - primarily the Fast Fourier Transform (a function analysis tool applied to the audio waveform). These elements can then be manipulated in the digital domain. Here is an example of an audio spectrum, generated from a note played on a cello :
Each peak represents the level of audio energy at its respective frequency generated by the cello within the audio spectrum. The right hand end is the top of the hearing range, and the left hand end the bottom. To give an idea of scale, the human voice can cover the range of 100Hz - 1kHz (1000Hz), or just over 4 octaves, from the lowest bass to the highest soprano.
A sine wave contains no harmonics other than the fundamental pitch, and is a pure single frequency signal, and therefore a single peak. All of the different frequencies in the cello's spectrogram are made up of complex vibration of the string, physical interaction between the cello's four strings, the bow, the performer, and the resonance of the wooden body of the instrument, not to mention the acoustic reflections from the performer, the room, the floor, the vibration of the floor itself etc etc. Basically, the whole environment contributes to the sound and its harmonic content. This is often referred to as the sound's 'timbre'.
If we take our cello note and apply a quite extreme EQ curve to it (indicated by the shaded grey area, with the ‘0' marking on the left hand scale being the centre point of no cut or boost), we can see the effect on the harmonic spectrum, compared to the unequalised version above:
The same frequency spectrum is present in both examples, but the relative levels of the various harmonic elements are skewed by the EQ curve. Adjusting the levels of audio energy within the spectrum in this way often alters the overall energy contained within the signal. For example, applying a large amount of bass boost will increase the overall level of the signal (albeit all the extra energy being contained within the bass end of the spectrum), which may lead to the signal subsequently overloading a circuit or channel later in its life. For this reason, you should compensate for EQ cuts and boosts with corresponding boost or cut to the overall level immediately afterwards, so that the sound's perceived level remains similar, and within the limits of your system. Many equalisers offer input and output level meters and offsets, to ensure that you are not radically altering the overall signal level (unless of course that is what you intended to do...)
EQ Glossary
Audio Spectrum
The Audio Spectrum is similar in concept to the visual spectrum of colours. Low frequencies equate to the red end of the colour spectrum, and high frequencies to the violet. Similarly, frequencies below the bottom of the human hearing range (20Hz approx) are called ‘infrasound' - analogous to infra-red light - and those above 20,000Hz, ‘ultrasound' (as used in ultrasonic medical imaging etc) - similarly to ultra-violet light.
Frequency
Pitch - measured in Hertz (1Hz = 1 cycle per second). In Western music, the A above Middle C = 440Hz, at concert pitch.
Octave
A term derived from the Western scale, but now more generally meaning an exact doubling or halving of pitch - depending on whether the pitch is being raised or lowered by 1 octave. Thus, if A3 = 440Hz then A4= 880Hz and A2 = 220Hz
Frequency Band
A range of frequencies, usually specified in frequencies (e.g. 250-300Hz), octaves or a fraction thereof. As above, a one octave rise in pitch is an exact doubling of frequency. The width of this frequency band is called - unsurprisingly - its bandwidth.
Shelving EQ
A high (or low) Shelving EQ band defines a frequency above (or below) which levels will be increased or decreased by a certain ratio, expressed in decibels (see fig.1)
Peaking EQ
a Peaking EQ band defines a frequency at which levels will be increased or decreased by a certain ration, expressed in decibels. The degree to which adjacent frequencies are affected will depend on the width of the EQ band (see fig.3 and fig.4)
