05. Filters & EQ
Filters and EQs are probably the most used effects in music production, would it be for mixing/mastering, sound designing or live performances.
Filters do a very simple tasks really: they cut some frequencies of the sound you're designing. So they are at the core of subtractive synthesis.
They can either be used to removed unwanted frequencies from a sound, or to leave more space for other layers or other tracks in a mix.
In this video, We'll see a lot of different types of filter, each allowing you to cut or lower frequencies in a different way, so you can hear how they sound.
But before we begin, here are a couple of things that are important to mention:
- Because filters will remove frequencies from a sound, they work better if you start with a sound that is rich in harmonics.
Like a sound made by FM synthesis or with a lot of distortion for example.
- Filters won't work particularly well on a sine wave for example, as the sine wave is only one frequency, so that's not a lot of content to cut, to shape the sound.
And in Serum, you can chose the type of filter you want to use here, and then chose what you want it to filter : oscillator A, oscillator B, or noise. You also have another filter in the effect section, and another extra filter in the distortion effect.
So filters generally come with 2 main parameters:
- The cutoff, which is the frequency were the filter starts affecting the sound
- And the resonance, also called Q, which amplifies more or less the frequencies around the cutoff point.
The low pass filter, also called high cut filter, is probably the most common. It will cut all the frequencies above the cutoff point, keeping only the lower part of the sound. That generally makes the sound smoother.
With the resonance you can control how much the frequencies are boosted at the cutoff point.
This filter doesn't cut abruptly every frequencies above the cutoff though, it cuts them progressively more and more the farther it goes from the cutoff point. So it cuts them with a slope.
You'll see sometimes a number attached to the low pass filter, like LP12, LP18 or LP24. It refers to the steepness of this slope.
The higher the number, the more abrupt the cut.
For a low pass 12 for instance, the 12 stands for 12 dB per octave. So a frequency 1 octave above the cutoff frequency will be lowered by 12 dB, and a frequency 2 octaves above the cutoff frequency will be lowered by 24dB. This gives you an idea of the slope of the filter.
So here is a couple of examples of low pass filters, with different steepness, and different cutoff and resonance settings.
The high pass filter, also called low cut filter, will cut all the frequencies below the cutoff point and keep only the higher ones. This usually makes the sound thinner, and sometimes harsher.
It is useful to keep only the high end of a cymbal for example, or to keep only the texture of one sound to layer it with another sound.
Similarly to the low pass filter, you can sometimes see a number attached to the high pass filter, it also refers to the abruptness of its slope.
The band-pass filter cuts the frequencies above and below the cutoff point, while boosting the frequencies around it.
The resonance here controls how much these frequencies are boosted.
A common way of using this filter is to make a "telephone effect" when it is put on a voice recording for example
The notch filter, also called band cut filter, will cut a band of frequencies around the cutoff point. The resonance here will set the width of this band. A higher resonance will result in a narrower band being cut, and a lower resonance will result in a wider band.
The notch filter is useful to get rid of a narrow slice of frequencies, to remove a small part of the sound you don't want.
I personally like to move the cutoff point while the note is being played, with a LFO for instance, to add movement to the sound.
There are actually two types of comb filters, there are feedforward and feedback comb filters. Let's see the feedforward form first
The feedforward comb filter is like a combination of several notch filters. It will be a succession of several points that will cut several bands of frequencies.
They are sometimes called flanger comb filters when the frequencies that are cut are multiples of this first frequency. the second notch filter being a twice the frequency of the first one, the third notch being at three times the frequency of the first one, and so on, following the harmonic series.
And other feedforward comb filters are sometimes called phaser comb filters. But we'll see why when we talk about flangers and phasers.
the cutoff would move the series of notches, and the resonance will then set their depth.
So these comb filters can be useful to cut a frequency with its harmonics.
The feedback form of the comb filter works in the opposite way, it will cut all frequencies and let through only the frequencies around the cutoff point and its harmonics, so it's like a series of narrow bandpass filters.
This type of filter can be very cool to use on white noise to let a particular note go through and have kind of a breathy tone.
The resonance will also define how much of these frequencies will be let through
Formant filters are designed to mimic vowel sounds, such as a, e, i, o, or u. It does that with a series of band-pass filters, usually between 2 and 5. They let through only some frequencies, selected depending on the vowel we want to reproduce.
For a formant filter with 2 points, one point could take a value between 300 and 1000Hz, and the second point could take values between 1000 and 3000Hz. The combination of the two will define the quality of the vowel, as shown on this graph. Additional point would refine the precision of the vowel
You can see here a more precise chart, giving the frequencies of 5 points for 5 different vowels: a, e, i, o, and u, for 5 different types of voices. From the lower to the higher: bass, tenor, countertenor, alto and soprano.
For the example, let's try to make a formant filter with several band- pass filters. Let's to it with 3 points
I will have an oscillator with a sawtooth waveform, because it's each in harmonics. And then I will make an Effect Rack with 3 channels, with one filter on each channel. This way the sound will go in parallel in each of these channels, and we'll hear the three outputs at the same time. Each output giving one frequency from one filter.
So I'm going to take 3 frequencies from the chart for "a" for as tenor voice and copy each on a different bandpass filter.
There I did the same with different values for different vowels to hear how they sound.
The high shelf filter will boost or lower all the frequencies above the cutoff point equally.
It is not that common as a standalone plugin, it is most often seen in equalisers, but we'll see those in a minute.
The resonance here will then change the steepness of the transition slope between the boosted part and the rest of the signal.
And you'd have a gain knob to set the strength of the boost or the cut.
A high shelf filter can be nice to add a bit of brilliance or to reduce the harshness a bit, if you don't want to cut the high frequencies completely.
The low shelf, similarly to the high shelf, will boost or lower all the frequencies below the cutoff point. It's also pretty rare as a stand alone, but it frequent in an EQ.
A tilt filter, also called tilt EQ, is like a combination of a high shelf and a low shelf filters. It will either boost the trebles and lower the basses or boost the basses and lower the trebles.
The cutoff frequency act like a pivot point, then you'd have a gain knob to set the strength of the effect, and the resonance would set the size of the slope of the transition.
The all pass filter is a bit particular because it has what we call a flat frequency response, which means it doesn't cut or boost any frequency.
Instead it delays the signal a little, by an amount relative to the frequencies it delays.
It means that the bass frequencies can be delayed differently than the higher frequencies.
They are not really used as stand alone filters, but they're used inside phaser effects. So I just wanted to let you know that they exist.
We'll talk more about them when we talk about phasers.
Parametric EQs are like a combination of several filters. It is presented with a screen representing the frequency spectrum, with the low frequencies on the left and the higher ones on the right. And there's a line on which you can add points.
Each point you add is a new filter, and you can chose if it acts like a low pass, high pass, high or low shelf, or a point EQ.
Each point will then have 3 settings:
- The frequency, which is similar to the cutoff point of the filter, it determines where this point is on the frequency spectrum.
- The gain, which determines how much the frequencies around the point are lowered or boosted.
- And the resonance, often noted Q. Which determines how wide or narrow is the frequency range affected by that point. A higher resonance meaning a narrower range.
It is one of the most versatile and most used effect both in the sound design phase and in the mixing phase, to make every sounds of a track sound well together.
A good rule of thumb is to lower frequencies more than you boost them. It is often advised to cut frequencies you don't want more than boosting the ones you like.
Another good rule of thumb is to use high gain values with high resonance and little gain values with small resonance. So if you want to cut and frequency drastically, it's often better to use a high resonance to have a narrow range of action, and if you use a small resonance value to act on a wider range, it's often better to be subtile with the gain knob.
With this in mind, an equalizer is also an excellent tool to dissect a sound and see what it is made of.
When you use an EQ, you can take a point with a rather high resonance and an high gain, and sweep it across the frequency range. It is like a magnifier to hear each part of the sound.
If some region sound too harsh, too loud or too dissonant, you can lower them. Step by step, it is a good way to really carve your sound.
Some EQs even allow you to hear only what is affected by the point you're moving. On Ableton's EQ8, it's the button with a headset icon.
Some EQ also allow you to process differently the center and the sides. Using a high pass filter on the sides is a good way to ensure the bass part of your sound stays in mono, which is often what you want.
Some EQ also allow you to process differently the left channel and the right channel, which it always good to enhance the stereo wideness of your sound.
Graphic EQs are another type of EQ where the frequency spectrum is sliced in several bands. You'd then have a slider on each one to control how much you want to boost or cut the frequencies in that band.
The bands are already set so you don't have control over the frequency of each point, nor over the resonance of each point which is often fixed. But this kind of EQ often allow you to have a lot more points to play with.