When we talk about filters in audio, they are more important than they might seem at first; there are even sound engineers who don't use them in 90% of cases, but it is necessary to know about them, as well as how they work and what they are used for. When we ask a sound engineer about filters, the first thing that comes to mind is the filters found on the input channels of mixing consoles. * Photo: David Bisbal 2017 tour with Meyer Sound
But filters are present in equalizers, whether analog, digital, parametric, graphic,… They are also found in loudspeaker processors, crossovers… and indeed in most existing audio equipment. Obviously, depending on the application they are intended for, they will be of one type or another.
A filter is characterized by its frequency response curve, which shows how the different frequencies at the input are attenuated or amplified. All electronic audio components have their own particular frequency response curve, although ideally this curve would be flat (a horizontal straight line) between 20 and 20,000 Hz, since any deviation means an artificial change in the timbre of the sound.
Filter types according to cutoff frequency
Leaving aside these design problems, which will continue to give electronic component designers headaches, the most common filters can be classified, according to the shape of this response curve, into four large families: low-pass, high-pass, band-pass and band-reject.
The low-pass filter. It lets through frequencies below a certain value, called the cutoff frequency, also known as the cut-off. In an ideal filter, this frequency would be a discontinuity in the response curve, so that every frequency above this value would be completely attenuated and every frequency below it left untouched. In practice this is not possible, and all real filters have a slope and ripple in the area near the cutoff frequency. The steeper this attenuation slope, the more precise the filter, although that is not always the desired goal. On audio equipment, the abbreviation used to indicate the presence of this filter is LPF (Low Pass Filter).
The high-pass filter. It does the opposite of the previous one, letting through only frequencies above the cutoff frequency; on equipment it is labeled HPF (High Pass Filter).
The band-pass filter. It lets a band of frequencies through, eliminating the rest. It is defined by its center or resonant frequency and its bandwidth.
The band-reject filter. It works the opposite way to the band-pass filter. Like the latter, it is characterized by its resonant frequency and bandwidth.
When we spoke earlier about the most common filters found in audio, we were referring to "high-pass" filters, that is, the ones labeled HPF on mixing consoles. These usually have a fixed frequency on low- and mid-range consoles, with a cutoff frequency typically around 100Hz, while on higher-end consoles this cutoff frequency is variable and can range from 20Hz up to 400Hz or 600Hz.
Some consoles also have variable low-pass filters; this type of filter is very practical, since it leaves the whole EQ free for boosting or cutting specific frequencies on the channel.
In filters, as we can see in the drawing above, the center frequency f0 is related to the bandwidth, which is taken as a reference at the -3dB point, that is, the span between f1 and f2. These filters are used in equalizers, whether graphic (you can only change the level of a series of preset frequencies) or parametric (you can change the gain, the frequency and its relationship to bandwidth).
* Photo: DiGiCo on the David Bisbal 2017 tour.
The equalizer is the device that lets us modify the frequency response curve of an audio system. This is done using banks of parallel band-pass filters: by adjusting their controls, we alter the incoming signal and thus modify the sound system's response until we achieve an ideal response for the venue and the type of music in question.
Most equalizers also have variable high-pass or low-pass filters
It is also true that most sound equipment has so-called subsonic filters, that is, filters that block frequencies above or below the audible range of the audio spectrum, from 20Hz to 20,000Hz.
Another important filter parameter is the slope value, that is, its decibels per octave.
Staying in the world of equalizers, slopes are most commonly 6 dB or 12dB per octave.
Let's look at a practical case: if a mixing console channel has an HPF filter with a fixed frequency of 100Hz, and according to the manufacturer's specifications this filter's slope is 12dB/octave, that does not mean we get 12dB of attenuation at 100Hz, but rather that one octave below, that is, at half the frequency, at 50Hz, we will get that 12dB of attenuation.
Digital filters
A common misconception is that analog filters alter the phase relationship while digital ones do not. We must not forget that the filters found in digital equipment, such as equalizers or loudspeaker processors, are nothing more than faithful digital reproductions of analog filters.
The possibilities of digital filtering are many; here are some of them:
Simulating analog devices. Selective filtering can create "downward" emulations, that is, simulate the sonic characteristics of low-quality devices such as telephones, old radios...
These devices are characterized by fairly narrow frequency curves. In this case, the simplest tool is usually the graphic equalizer. If the program has EQ presets, just select the appropriate one. If you can't find the preset you want, once you have set up the equalizer properly, you should save it for future use.
Low-pass filtering when reducing the sample rate. Reducing the sample rate causes aliasing, a phenomenon in which phantom frequencies appear that were not in the original sound. To avoid aliasing, before converting a file to a lower sample rate, you should apply a low-pass filter with a cutoff frequency equal to half the new sample rate. In other words, if we want to convert a file recorded at 44,100 Hz to 22,050 Hz, we must filter the sound, removing its frequency components above 11,025 Hz.
Many programs today simplify this process by offering the option of filtering automatically when converting to a lower sample rate
Sound enhancement. This is especially effective and advisable with vocal recordings. Bear in mind that many of the velvety, seductive voices of today's singers are a direct product of the lab. Try boosting different frequency bands on recordings of your own voice. Remember that the human voice has components between roughly 100 Hz and 5,000 Hz, and that manipulating multiples of these fundamentals, the harmonics, can produce very surprising effects.
* Photo: David Bisbal 2017 tour with Meyer Sound
Spectral analysis. So as not to work blind, spectral analysis is a very useful tool that can give us an idea of the qualities and shortcomings of any sound. Sometimes annoying fixed-frequency noises caused by electrical interference creep into recordings. Spectral analysis lets us detect these pesky frequencies so we can then filter them out. Imagine extracting a particular instrument's sound from a mix; some spectral analysis programs achieve this with some success, the best known and most veteran being “AudioSculpt.”
Creative filtering. Blending science and art, it lets you radically change the nature of any sound, creating "musical" sounds from other sounds that have little or nothing to do with each other; the program mentioned above also does this quite successfully.
Phase relationship in filters. We know that filters are classified by their slope, that is, by their decibels per octave; we have also mentioned that using filters affects the phase relationship of an audio signal.
The filter types are
Butterworth
Bessel
Linkwitz Riley vs Butterworth
They are named in honor of their discoverers; the filters most widely used in the audio industry are Linkwitz Riley filters.
There are other types of filters that, every now and then, some audio manufacturer presents as new in a commercial display of technological innovation, but that is not the case: the filters are the ones they are, and those presented as new are usually combinations of some of the above.
- LPF filter 100Hz 6dB/octave
- LR 12dB/octave filter at 100Hz
- 2nd-order LR HPF and LPF at 100Hz
Phase response of filters
Indeed, if we look at the phase response of each filter, since each one is 2nd order we will have 90º upward on the HPF and 90º downward on the LPF, so when the two are summed the difference will be 180º, with the resulting cancellation at that point. To solve this problem, it is enough to invert the polarity of one of the filters; remember that this shifts the phase by 180º, so there will be summation at the crossover point.
Using filters in crossovers or digital loudspeaker processors
Obviously, explaining and illustrating all the consequences of using each type of filter would take too long, so we will list the characteristics and consequences of each filter. When tuning a sound system, it is essential, when combining the different low, mid and high bands through the crossovers, to know the consequences on the system's phase and frequency response.
-Butterworth filters are characterized by 3dB of attenuation at the crossover point.
-Linkwitz Riley filters are characterized by 6dB of attenuation at the crossover point.
-Bessel filters are characterized by 8dB of attenuation at the crossover point.
If we combine two Butterworth filters, HPF and LPF, the resulting frequency response is +3dB at the crossover point; this is because this type of filter is -3dB at the crossover point and has an identical phase response, so they add up by 6dB, and the result is +3dB. The same happens with each filter type; let's look at the practical case of a Linkwitz Riley filter.
Conclusion
With all this theoretical knowledge well absorbed, more than one sound engineer will probably think twice before touching an equalizer or any other piece of equipment made up of filters. A good grasp of these basic principles is also essential when tuning the different bands of a sound system; for that we need to arm ourselves with a good measurement microphone, a Mac or PC laptop, a transfer function program such as the already very popular MAC FOH or SMAART LIVE, and adjust the digital loudspeaker processors that are essential in any self-respecting sound system.
Because if the system is well tuned and the artist is very good, our work is greatly simplified, letting us enjoy more and more the music we are working with, which in other times kept us too busy solving technical problems rather than doing what really matters: enjoying what we love most, live music.
