A filter is a tone control. It acts on the harmonic content of sound. When you move the treble and bass knobs on a hi-fi, you use a filter; when you fiddle with the EQ on a mixer, you use a filter. You use a filter when you dial down on that cut-off knob to find the gooiest, resonance-enriched groove. We use filters to kill hi-hats or drop out the kick drum, we use them to give a perceived movement in pitch or to fit sounds between sounds, and we use them because we love the pssswwhooarshh sound it makes.
Filters are everywhere, from effects and signal processing to mixing and instrument controls, but for this article, I will focus on their usage in synthesizers. It’s the one control we all reach for when approaching a new synth. We will look at what a filter is and what it’s doing to the waveforms. And we will look at what makes a Moog filter different from a Roland filter, an Oberheim filter, a Wasp filter, and so on.
What does a filter do?
A filter changes the tone of a sound by removing certain frequencies. You could also see it as a circuit that isolates a range of frequencies for special attention. How a filter affects a sound’s tonal colour depends on the filter type and what it’s designed to do. While there are many types of filter there are four common ones that you see time and again on synthesizers and in modular. These are the low pass, high pass, band pass and notch. Their names do an excellent job of describing what they do, but let’s dig a little deeper.
The low-pass filter is probably the one circuit that has had the most significant impact on electronic music since the invention of the oscillator. It’s the most gregarious of filter modes, the one that gets the party started and, amongst other things, gave us Acid House through its presence on the Roland TB-303. There’s something about its ability to throb that enthrals our internal organs.
Simply put, a low-pass filter allows frequencies below a threshold, which we call the “cut-off”, to pass through. As you dial the cut-off down, you start to lose the sizzle and hiss of high frequencies. Then presence is carved away, and mid-range boxiness becomes softened until you are left with the body, bloom and girth of your fundamental waveform. In other words, it’s particularly good at removing harshness and building warmth. The fundamental frequency is retained all the time, hooking into a solid sound and a strong sense of melody.
The High-pass filter does the opposite. It allows frequencies above a threshold to pass through. As you wind it up, it slashes the bass, kills the fundamental and murders the body of the sound. High-pass is a psycho. They are very helpful in cutting out rumble and hum in audio processing, but in a synthesizer context, they make us feel the absence of something and can be superb at building tension before a drop. Or it can be very helpful in leaving the bottom end to the bass line and keeping your bigger pad sounds away from muddying the water around them.
However, it’s not all about utility and mixing clarity. For tonal changes, the high-pass filter offers an interesting thinness and squeezing effect as the cut-off travels around our synth sound. Something delightful happens when you start pushing the resonance, but we’ll come to that momentarily.
So, what do you get if you combine a low-pass filter with a high-pass filter? You get a bandpass filter. A bandpass filter looks like a graph of normal distribution. It’s a pimple of frequencies, an upturned bucket of tones, that rejects the lows and the highs and leaves us with the mid-tones of a cheap transistor radio.
Of course, there’s much more to it than that, and putting a bandpass filter into motion sends it into highlighting highs and lows while creating space around itself. It’s very good at placing sounds in the frequency spectrum and hints towards vocal sounds and formants.
The Notch filter notches out a very narrow band. It’s designed to eliminate prominent frequencies in a mix, like when your snare is too loud. But in terms of synthesizer use, it brings a sense of phasing and a more subtle movement.
Cutoff and beyond
Filters have additional qualities that dictate the severeness of their edges, the chaos around the cut-off and how they respond to the notes you play. These are called Slope, Resonance and Key-tracking.
You can often imagine that the cut-off frequency is a brick wall or cliff face, after which nothing passes. That is not true. The cut-off frequency denotes the point at which the filter starts to pull the level of the target frequencies towards zero. The severeness of the “slope” from cut-off to zero is measured in decibels per octave. Commonly in synthesizer filters, the gentlest slope is described as being 6dB/Oct, then 12 dB/Oct for a mid-level slope and 24 dB/Oct for the most severe. This corresponds to another way of describing the slope as 1 pole, 2 pole and 4 pole, respectively. Each pole offers 6dB of level drop per octave.
Many fascinating details are available for people who want to understand the relationship between slope, pole and filter design. I can recommend Gordon Reid’s excellent 1999 article “Further with Filters” from Sound On Sound. But in more lay-person terms, a 6dB slope is commonly found as a roll-off in EQ, 12dB 2-pole is your warm and comforting ARP and Oberheim filter, and the 4-pole 24dB is what we find in the juicier Moog and Roland filters.
The most characterful feature of a filter is found in the resonance or “Q”. Resonance is a boosting of harmonics around the bleeding edge of the cut-off frequency. It causes everything from a thickening to abundant chaos, and with enough encouragement, it can reach self-oscillation. It’s achieved by feeding the output of the filter back into the input. It would be more accurately described as the filters “feedback amount” but it sounds and feels like you’ve picked up the room’s resonant frequency and so we’ve ended up calling it “resonance”.
One thing that can happen as a result of pushing the resonance up on a low-pass filter is that the overall volume of the audio drops. It’s like the boosting of the resonance is also pushing down on the shape of the filter, shaving dBs off the level, particularly at the bottom end. If you look at a waveform on a scope being filtered with high resonance, you will see high-frequency sine waves emerging from the original. These can take on life of their own and start to oscillate at the cut-off frequency. With enough Q your original audio will become overwhelmed, and you’re left with a pure sine wave that’s controllable and playable from the cut-off. Filter designers can build resonance compensation to make up for the loss of gain. This gives it a more even response when playing with the resonance but can also dampen its ability to self-oscillate. It depends on what you’re looking for.
When applying lots of Q to a high-pass filter, you get a boost to the bass before it reaches self-oscillation. The resonance is acting on or near to the fundamental frequency, adding back elements of the bass that were filtered out. It gives a really good underpinning and breadth to the thinness of high-passed waveforms.
You’ll often find a Key-tracking or “KBD” knob on a filter. Key-tracking moves the cut-off frequency relative to the note that’s been pressed on a keyboard. If it’s a low-pass filter, then playing higher notes opens it up, whereas playing lower notes closes the filter. It allows you to find a brighter tone as you play up the keyboard which tends to feel more natural than a static filter where higher notes can be lost.
Filters come alive when animated. This is often a behind-the-scenes function on a synthesizer but can become very front and centre in modular synthesis. Filter modulation usually comes in the form of an LFO, Envelope, or even a combination of the two. But you can use whatever you like.
A smooth sine or triangle LFO can gently glide a tone-crafting filter across your pads in a deliciously dreamy way. Or, if you crank up the rate, you can get a rather nice tremolo or vibrato. As a filter deals with frequencies as well as amplitude, you can get interesting mixes of amplitude (AM) and frequency (FM) modulation depending on the amount of emphasis you put on the resonance.
A multi-stage envelope often accompanies a filter to shift the cut-off on the striking of every note. This is what gives us the classic filtered bass sound. But envelopes are also useful on much slower movements to travel on a tonal journey through each note.
It’s the movement that seems to be important. In working heavily with modular synthesis, I find that the filter cut-off is always in play, from multiple modulation sources and manually through my fingers. It affects the feel, changes the atmosphere, and generates interest and points of emphasis from subtle to chaotic. I would say that it’s a bit overused, except that I can never get enough of it.
We sometimes talk about filters as if they are a generic function that behaves in the same predictable fashion. While they may achieve similar things, the designs behind filter circuits can be very different. The brilliant thing about that is that filters in different synthesizers have unique characters. Arguably, in analogue synthesizers where many components are common, the filter stands out as a defining attribute.
One term you may hear when talking about filter design is “Topology”. This refers to how the components are arranged in a circuit without talking about the individual values. So, when designing a filter, you might draw on the topology of an SEM filter or a Ladder filter etc. Without using the same exact components.
Here are some of the most commonly used filter designs that are still sculpting waveforms today.
The Ladder filter is so-called because the circuit design layout on paper looks a bit like a ladder. Moog used a transistor-based design to build the 904A low-pass filter for the Model 12, and later the sound became famous in the Minimoog. The same design has found its way into every Moog synth since and is generally what people mean when they refer to the “Moog sound”. It uses four transistor poles giving it 24dB of slope and has a resonance that can take it from sharp to squelchy and deliciously soggy.
Ladder filter designs are also found in synths such as the EMS VCS-3 and Roland TB-303. But this design uses a diode-based ladder, giving it a more rubbery and edgy sound. The transistor design in the Moog ladder isolated the individual poles, whereas, in the diode ladder, the poles interact, making it much less predictable.
The ladder filter was one of the few things Moog patented in the 1960s. When ARP introduced the 2600 synthesizer in 1971, the 4012 low-pass filter sounded an awful lot like the Moog one. After a conversation between Bob Moog and Alan R. Pearlman, it was replaced with the redesigned 4072 filter for the orange and black Model 2601 in 1977. In modern recreations, we tend to find both the 4012 and 4072 filter versions included.
Tom Oberheim took a different route with the Synthesizer Expander Module (SEM). This had the gentler slope of a 2-pole 12dB/Oct filter but also pulled in multiple modes giving it some unique qualities. Along with low-pass, it also had high-pass, bandpass and notch, all of which could be morphed between. It lacked the imposing power or self-oscillation of the 24dB filters in the largely monophonic synths of the time. Instead, it proposed a more nuanced form of tonal shaping, perhaps suited to the emerging range of polyphonic synthesizers.
A filter based on the Operational Transconductance Amplifier (OTA) is generally regarded to have a warm and well-behaved tone unless it’s the Korg MS-20 which we’ll return to in a minute. OTA designs were behind the CEM3320, which found its way into the Sequential Prophet-5, Oberheim OBXa, Roland SH-101 and Fairlight, amongst others. It’s very smooth and has much less internal distortion making it fantastic for large polyphonic sounds. The Roland Jupiters also used an OTA design but with different chips.
Korg MS-20 Filter
The MS-20 was initially based on a proprietary Korg-35 design but later ended up using LM13600 OTA chips. The Korg-35 was a bit noisy, pushing the relatively calm 12dB low-pass filter into screaming fits of rage at high resonance levels. It was placed in series with a 6dB high-pass filter giving it useful bandpass capabilities. It’s aggressively distinct and is a favourite for high-energy and distorted sounds.
Formanta produced the Polivoks synthesizer in the Soviet Union in the 1980s. It was the USSR’s answer to the Minimoog and had an unusual switchable low pass and band pass filter. The filter is based on a programmable op-amp for a 2-pole 12dB slope. The defining characteristic is that it tends to fall apart around the edges. It’s delightfully unstable and eager to pile on the resonance.
You don’t usually come across a filter called “Sallen-Key”, but it’s the design behind some other filters that you have heard of. Most notable are the Korg MS-20 and the Steiner Parker Synthacon. Arturia uses the Steiner Parker implementation in its various Brute synthesizers. The Sallen-Key design is simple, and capable of high resonance and multimode.
The Wasp filter has a similar vibe to the Korg MS-20. It was built in the 1970s by UK-based EDP and is loved for being unruly and aggressive. It’s CMOS-based and abuses digital inverters and other components to produce its distorted sound. It has a control that mixes from low-pass to high-pass, giving us bandpass in the middle.
These filters are being constantly re-engineered, reimagined and reanimated for use in current synthesizers and modular systems. Many virtual analogue synths and software synthesizers give you a list of dozens of filters to choose from so you can explore the differences depending on how good the modelling is. In analogue recreations of vintage synths, it’s common to include all versions of filters used in the life of that synthesizer. Miniaturisation and Surface mountable components have helped get these alternative circuits into the confined space of an instrument.
In modular synthesis, the filters are more overt. You can choose a filter module because it is based on the exact topology and audio response that pleases you. It also leaves room for development, experimentation and new filter-based directions.
AMSynths has put much effort into replicating the filters from Roland synths down to the finest detail. Whereas System80 uses the Jupiter-6 filter as the inspiration for its 860 filter using modern components. Ladder filters are everywhere, from the AJH Synth Minimod Transistor Ladder’s faithful reproduction to the beautifully made Traigh from Instruo to the complex kaleidoscope of Mannequins Three Sisters. Erica Synths out of Latvia seem to have a handle on the perfect Polivoks clone, and Doepfer has the most enjoyable Wasp clone you’ll find anywhere.
So, how do different filters sound in the hardware reality of a modular synthesizer? I put together five Eurorack low-pass filters for comparison, and I’ll do my best to describe how they sounded. But do go and check out my filter comparison video to see for yourself.
First, I ran through a static sawtooth wave and observed what happened as I moved the cut-off and resonance knobs. Then I patched an envelope to the cut-off and routed a sequence to the same sawtooth oscillator. The filters in question were the Instruo Traigh ladder filter, the System80 Jove Jupiter-6 filter, the Wavefonix 2140 OTA filter, the Erica Synths Polivoks filter and the Doepfer A-124 Wasp filter.
The Traigh is so smooth; it feels like a sunrise, filling your ears with warmth as you turn the cut-off. As you bring in the resonance, there’s an immediate drop in beefiness, but overtones, like sunbeams, start to tickle their way through. The Traigh offers a lot of gain at both the input and output stages, so there’s scope for compensating for the drop. It’s the gentlest of transforms through the frequency range as the resonance sweetens to tonal changes. You can go all the way up on the resonance until the last fraction, when a threshold is past, and it leaps into self-oscillation. There’s a tiny amount of grit behind it when you still have the input patched. Remove the input, and you have a pure and very playable sine wave.
The Jove feels louder and slightly harsher but is still deliciously smooth on the turn of the cut-off. There’s less drop when the resonance increases and those harmonics really shine. The resonance is much less polite, pulling in edges once past the 12 o’clock position. But there’s also a liquidity that the Traigh doesn’t have that morphs alarmingly into screams as you continue to build the resonance. There’s no self-oscillation at the end of the turn, but you can step through plenty of hugely chaotic resonant peaks.
You’d imagine the 2140 would be similar to the Jove as they share some topology, but that’s not the case at all. The 2140 has a signal gain knob that pushes it nicely into distortion before you even start to filter. It’s smooth enough before resonance is engaged, but once you turn that knob, it explodes into self-oscillation. You get up to about 11 o’clock on the knob of useful, relatively calm filtering, and then the resonant harmonics overwhelm the signal. But it has a superpower in the form of a gain compensation switch. It brings the input signal back in to duel with the exploding resonance and can get very fierce. It has some furious tones, but in that smaller, lower range of resonance, there are some lovely ones too.
The Polivoks isn’t anywhere near as unruly as you’d imagine. It’s smooth and serious, but there’s definitely a straggling something buried in the background. It has a brightness that might be related to the lack of level drop as you add the resonance. It’s substantial and broad sounding, and the resonance adds some lovely tones all the way round to about 4 o’clock when things start to get dicey. The self-oscillation is full-on and fabulous, but if you pull back a little, you get these rich tones that crackle around the edges.
The Wasp is immediately different. The original breadth of the sawtooth has been lost somewhere and replaced with an unexpected crispiness. Maybe it’s because the waveform is inverted. The level knob offers a lot of oompf to drive the filter and you get the sense of heat and dryness rather than the wetness of some of the other filters. It has a literal vocalised “WOW” factor as you manipulate the cut-off and absolutely no drop-off when you pile on the resonance. As with the Jove there’s no self-oscillation, but instead, you get some wonderfully playful distortion that adds to the richness, and powerfully emphasises the harmonics rather than overwhelming the signal.
Once you pump through a sequence and animate the cut-off with an envelope, filters really come alive. Finding and manipulating the sweet spots of a dynamic and organic-sounding filter is, I think, one of the funnest things in modular.
Once again, the Traigh is as smooth as butter. It has a roundness to it that’s soft and cuddly. At no point do you hit distortion or chaos; it has that famous squelch of a rolling harmonic edge being drawn from each note. Even with the resonance in self-oscillation, it has a sweetness to it that’s musical and fabulous. And then, as you ease off, the broadness comes back with such a solid, rounded sound.
With the resonance approaching 12 0’clock, the Jove gives a deliciously liquid sound. It’s juicy and dripping with energy, drawing in a bit of distortion as you increase the Q. You can play with both the cut-off and the amount of envelope coming through to get a wide range of textures. There’s no self-oscillation to worry about, and so instead, you get some very striking zaps as the envelope tugs on the resonant screams.
The 2140 is thicker and creamier than the Jove at low resonance levels. It gets a bit nuts at anything past 12 o’clock on the resonance knob, but goodness, when you turn on the compensation, it sits up and smashes you with thrills.
The Polivoks is fast becoming my favourite. It has a likeable grit to it that sits behind the sumptuous combination of fundamental and resonant overtones. The resonance doesn’t overwhelm; it pours in nicely behind the audio, bringing some pleasing edges and wetness. The self-oscillations are a bit too unruly with the input signal, but staying just back from the point of chaos is a lovely and chewy place to be.
You find the dissonance as soon as you plug the Wasp in; it’s there from the start. It’s different and compact somehow; you can imagine it cutting through a mix without a big bottom end. Then as you turn up the resonance it gets both boxy and drippy. The drips are slightly behind the audio’s main force, making it feel unstable in all the best ways. And as it’s not self-oscillating, you get nicely controllable chaos at high Q with a very likeable distortion.
I hope that’s given you a sense of how the filters may differ.
Whatever filter sound you like can be yours in software or hardware. We have an unprecedented opportunity to sample these fantastic circuits and compare them next to each other in a rack or in our DAWs. It leaves us with a question about whether any new topologies will emerge that can sit alongside the classics or whether we are happiest in the familiar sound of the filters we love.