You are here

Arturia PolyBrute

Polyphonic Synthesizer By Gordon Reid
Published October 2020

Arturia PolyBrute

Arturia step into the big league with their most ambitious instrument yet.

Unsubstantiated speculation about an Arturia polysynth based upon the 'Brute' architecture has been flying around since the MiniBrute appeared in 2012, and with increasing intensity after the launch of the MatrixBrute in 2016. But while the company said nothing on the matter and continued to flood the world with soft synths, monosynths, plug‑ins, MIDI controllers, drum machines and audio interfaces, their developers were busy behind the scenes. They recently told me that they had been working on their first analogue polysynth for more than three years, so when they asked me whether I would like to test it prior to its release, it's not hard to guess how I replied.

What I received a couple of days later was a six-voice, bi-timbral, dual-VCO per voice polysynth, with two VCFs, three digital contour generators and three digital LFOs per voice, a digital effects section, an arpeggiator and a polyphonic sequencer, housed in a case with a 61-note velocity- and pressure-sensitive keyboard offering selectable responses, a pressure-sensitive touchpad, plus pitchbend and modulation wheels. In addition, what looks like a decorative groove in the wooden bar behind the keyboard turns out to be a long and responsive ribbon controller. Weighing in at a chunky 20kg, it's a substantial bit of kit with reassuringly solid hardware, although it still uses Arturia's usual faders, which are slightly light and wobbly.

The basic unit of a PolyBrute sound appears to be the patch, but this actually contains two largely independent sounds named A and B. When you select Single (mono-timbral) mode, each voice can play A, or B, or anywhere between, and the sound can morph within the extremes under the control of, well, lots of things. In Split mode, the Lower zone plays A while the Upper can again play A, or B, or anywhere between, and morph between them. Layer mode is similar, with one Layer always playing A while the second can yet again play A, or B, or anything between, or morph between them. (There's a pattern emerging here!) In addition, the Polyphony parameter determines the voice mode for each sound: six-voice polyphonic, unison or mono. polyphonic unison, dynamically assigned unison, voice cycling options, selectable mono/unison note priorities, legato with or without various modes of portamento, note-stealing options and more can be determined independently for the two sounds.

The Signal Path

Arturia's monosynths tend to mix four waveforms to create the output from each oscillator: a sawtooth wave with Ultrasaw, a triangle wave with Metalizer (wave-folding), a variable pulse wave with PWM and a sub-oscillator with square and sine-wave options. In contrast, the PolyBrute's VCO1 loses the Ultrasaw and the sub-oscillator, although tucked away in the menus is an option to apply the Metalizer to all three remaining waves. This makes it possible to create an unusual set of initial timbres, especially when you do things such as sweep the pulse width with the Metalizer turned up high. Some of the resulting sounds can be aggressive and raw, while others are so glassy that they could almost be coming from an early wavetable synth. Nonetheless, I miss the lush Ultrasaw, so I quizzed Arturia's engineers, who explained that they could have kept it if they had removed either pulse-width modulation or the Metalizer. They chose to keep those because, as they put it, "the Ultrasaw effect can be approximated with detuned oscillators". I think that (no pun intended) that's sound thinking.

The PolyBrute's second oscillator loses the Metalizer in addition to the Ultrasaw function, although it retains a sub-oscillator one octave below the fundamental, albeit without the square-wave option. You can sync VCO1 to VCO2, and a knob allows you to determine the amount of sync from nothing through various degrees of 'soft' to 'hard' sync. You don't obtain a smooth sweep of sync possibilities as you rotate this knob but rather (depending upon tuning) a handful of states as VCO1 is locked in different ways to VCO2. You can also use VCO2 as the modulator for 2-op FM synthesis (often called 'cross-mod' on an analogue synth) but bear in mind that, even if you tune the PolyBrute frequently, you're not going to obtain perfect consistency across all voices. Nevertheless, with low FM amounts, you can obtain some very usable new sounds that would not otherwise be possible. All of my best 'plucked' sounds were crafted with the FM knob somewhere around the 10 o'clock position. With a bit of chorus, the Clavinets were wonderful!

If the oscillators have another oddity, it's in their tuning and how they respond to modulation. VCO1 is always quantised to semitones, but VCO2 has two fine-tuning ranges as well as two wider chromatic ranges. When you look at how their tuning can be modulated things get even weirder, with major, minor, Phrygian and other scale modes in addition to portamento and glissando. Many interesting results are possible: for example, making one oscillator step in fifth and octaves while the other plays a major ninth scale as you modulate them. It's weird, but wonderful.

The oscillator mixer offers three inputs: VCO1, VCO2 and variable-colour noise ranging from red to white. You can direct the output from each of these to one or both of the filters, either in series (the output from VCF1 feeds the input of VCF2) or in parallel, or anywhere between. However, you can't choose the filter slope: VCF1 is a resonant 12dB/oct Steiner filter while VCF2 is a resonant 24dB/oct low-pass ladder filter. I enquired why the filter architecture had been constrained in this way and the chaps at Arturia explained that, had both slopes from the MatrixBrute been included on both filters, it would have been necessary to find ways to morph between them. This would have been both complex and expensive, so they reverted to what they view as the essence of each. Again, this made sense to me.

Arturia PolyBrute VCO and Filter section.PolyBrute's VCO and Filter section.

Despite the lack of cutoff slope options, there's a lot more here than meets the eye. In addition to its cutoff frequency, resonance, 'Brute factor' (filter output feedback), bipolar contour amount and output level controls, VCF1 has a knob marked LP/BP. However, this should be annotated LP/BR/HP/BP because between the low-pass and band-pass responses at either extreme lie notch and high-pass modes, and you can sweep from each to the next. This makes VCF1 a true multimode filter. Likewise, VCF2 (which offers overdrive in place of feedback) has a powerful trick up its sleeve. Most low-pass ladder filters attenuate any low frequencies passing through them when you increase the amount of resonance, making it impossible for them to go 'zeooow' while retaining a big, warm bottom. But a few vintage synths had the ability to do both simultaneously and so it is here, which is fab.

A large Master Cutoff knob sits between the two filters allowing you to adjust their cutoff frequencies simultaneously, but I fear that this will most often be used for creating manual filter sweeps because, for reasons that I can't quite fathom, adolescent males never seem to get tired of doing this. Next to this you'll find the key-tracking knob. At its maximum, the Ladder filter tracks the keyboard at 100 percent over a small range, whereas the Steiner filter is less accurate in this respect. However, if you inject some signal from the oscillators, the filters will lock and you can even create five-pitched sounds with VCO1, VCO2, the VCO2 sub, and both filters oscillating. From the panel, you can modulate the filters' cutoff frequencies using the noise generator (which is hard-wired to the cutoff frequency of VCF2) and VCO2 (which is hard-wired to the cutoff frequency of VCF1).

Once the signal has passed through the filters, the PolyBrute positions each voice within the stereo field. There are numerous options here. When Voice Pan and Distrib Centered are selected and the Stereo knob is at its maximum, Voices 1 and 4 are centred, Voice 2 is panned hard left, Voice 3 is panned hard right, and Voices 5 and 6 are half-left and half-right respectively. These values are also sent as percentages in the range ±100 percent to the modulation matrix, appearing as a source in row C. With Voice Pan and Distrib Gradual chosen, a different set of values with more evenly distributed panning is generated. When Voice+Filter panning is chosen, the same panning schemes are used when the filters are in series, but the outputs from the filters are panned hard left (Ladder) and hard right (Steiner) when they are in parallel. Since the filters' Series/Parallel parameter sweeps from one extreme to the other, you can position the voices anywhere between these modes. Whichever panning mode you choose, the Stereo knob determines the width of the result. Oh yes... and all of this can be modulated, so you're not going to be short of spatial effects using the PolyBrute.

The audio now reaches the stereo effects, and there are three routing options through these. Insert passes the signal through all three in series with the Modulation effect followed by the Delay and the Reverb. Send passes the signal through the Modulation effect and then the Delay and Reverb in parallel. Bypass, as its name implies, bypasses the effects and thereby ensures a pure analogue signal path from the oscillators to the outside world.

In addition to the chorusing and phasing annotated on the panel, the Modulation effect offers three types of flanging, a second phasing option, ring modulation, bit-crushing, and a down-sampler. Similarly, the Delay offers seven additional modes through its menus, and the Reverb offers five. Nonetheless, the controls are basic: just Intensity for the Modulation effects; Level, Time (with optional sync) and Regeneration (with eight high-pass and low-pass filter options) for the Delay; and Level, Time and Damping (with four menu options) for the Reverb. Happily, all of these are possible destinations in the matrix. My only problem with the effects was a line in the draft manual that suggested that users can design their own effects within the PolyBrute Editor. You can't. I've pointed this out, and I hope that the manual will be corrected before the synth is shipped.


The PolyBrute offers three contour generators. VCF ENV is a velocity-sensitive ADSR hard-wired to the filters' cutoff frequencies and it's also available as a source in the modulation matrix, while VCA ENV is hard-wired to the audio amplifiers but isn't available as a matrix source. The third, MOD ENV, is a five-stage DADSR whose output is only accessible via the matrix. Again, there's more here than is apparent, with quick and percussive modes that alter the curves of the Attack and Decay phases for each of the generators, and three ways in which velocity can affect the response times of the VCF and VCA contours as well as the amount by which they're applied. Natural emulates acoustic instruments by shortening the Attack while extending the Decay and Release of the chosen contour, Shorten shortens all of these phases, and Extend extends all of them. In addition, each contour can be looped, either twice, thrice, or indefinitely to create the famous Trapezoid response or more complex shapes for use as additional LFOs.

Arturia PolyBrute's Envelopes and LFOs section.Arturia PolyBrute's Envelopes and LFOs section.

This then brings us to the three visible LFOs, any of which can be chosen to be unipolar or bipolar, and all of which must be routed through the matrix to have an effect. LFO1 offers seven waveforms, rate (from 0.02Hz to 100Hz), arpeggiator/sequencer/clock synchronisation, control over the start phase when you trigger it, and three triggering modes: mono (which turns it into a single free-running LFO applied to all voices equally); poly (which is free-running but with an independent LFO for each voice); and poly-retrig (an independent LFO for each voice, with each being initialised when its voice is triggered). LFO2 is similar but replaces the phase parameter with a fade-in parameter that applies a delay as well as a fade. There are independent retriggering parameters for each of these, and you can also determine independent clock sync ratios for each.

LFO3 is rather different. Instead of offering predefined waveforms, it generates a triangle wave that you can shape in two fashions. Curve allows you to bend the phases to make it 'pointy' at the top and rounded at the bottom or vice versa. Simultaneously, Symmetry allows to slew the underlying shape from a sawtooth to a triangle to a ramp wave, or anywhere between. The Rate and Sync controls then do as you would expect, although only two Retrig options are offered: Mono and Poly-Retrig. These come into their own when you find that LFO3 has a mode that causes it to run through a single cycle and then stop. This turns it into an AD contour that you can shape in many ways to create shapes ranging from gentle piano-style decays to aggressive pings and thumps. Even more unusual is a switch marked xLFO1 that causes LFO1 to modulate the level of LFO3 when LFO3 is free-running. Alternatively, when LFO3 Single mode is selected, LFO1 acts as its trigger. You can combine these modes to generate complex modulation signals that I might expect to obtain from a modular synth or a VCS3, but not from an analogue polysynth. This is powerful stuff.

Although the 64-slot modulation matrix reveals eight destinations when you look at the panel, there are four pages of these, meaning that you can direct any of its 16 predefined sources (four of the rows have dual inputs) to your choice of 32 assignable destinations. At first sight it appears that neither switches nor menu parameters can be matrix destinations, but a small number can be assigned from another menu.

Once you've specified a destination you can direct any source or sources to it by pressing the appropriate buttons in the matrix and then determining the amount of modulation applied. You can route multiple sources to a single destination, and a single source to multiple destinations, each with independent polarities and amounts. You can also direct some modulators to modulate themselves and to control the amounts of other modulation paths. This is far more flexible than a pin matrix, and if you try to achieve the same results using a modular architecture you'll need a lot (and I mean a LOT) of mixers, multiples, inverters and amplifiers. But despite its immense flexibility, using the matrix is considerably simpler than describing it, so you'll soon be whizzing around it.

Arturia PolyBrute's Modulation Matrix controls.Arturia PolyBrute's Modulation Matrix controls.

Morphing, Motion & Performance Controls

The easiest way to hear sounds A and B is to select Single timbrality and sweep the Morph knob from one extreme to the other, which slides almost every parameter from its value in A to its value in B. Almost? Well, there are some parameters, such as polyphony and timbrality, matrix connections, effects routing and types, for which morphing would be either impractical or nonsensical. But if you want A to have one set of matrix connections and B to have a different set, you can set things up so that the second set has zero Amounts when A is selected, and vice versa. It took me a while to get to grips with this but, once mastered, the sound design potential of Morphing became staggering. A number of additional housekeeping functions help you to manage your Morphs, allowing you to import A or B from an existing patch into the edit buffer of the current patch, copy from A to B, and swap A and B. Then, once you've created the perfect Morph, you don't have to save the patch at either of its extremes because the Morph knob position is itself stored as a patch parameter. Of course, the Morph position is also a matrix destination, and the possibilities suggested by this are again enormous.

The Motion Recorder is another welcome facility. This allows you to record the motion of a single control and save it as part of a patch. Making a recording is simple and, when you replay it, you can do so at a speed ranging from 12.5 percent to 800 percent of the recorded speed, either once or looping indefinitely. My experiments showed that replay is per-voice rather than per-patch, which is excellent. The most flexible use for this is to record the Morphée in Matrix mode, because that allows you to modulate and loop up to 32 parameters simultaneously.

The Morphée offers an additional and versatile performance control.The Morphée offers an additional and versatile performance control.The number and flexibility of the performance controls on the PolyBrute is laudable, but that hasn't stopped Arturia from indulging in a bit of hyperbole when describing the Morphée. Nonetheless, a pressure-sensitive pad is an excellent facility, even when it looks like someone has left a wallet on the side of the keyboard. If you select the Matrix as the pad's destination, its X, Y and Z axes are directed to the J, K and L rows respectively. Alternatively, you can use the pad to control aspects of the arpeggiator and sequencer, with the X axis affecting the velocity and the range of octaves played, the Y axis affecting the Gate time, and the Z axis adding ratcheting. But I suspect that you'll most often use it to control morphing.

With this mode selected, the X/Y response is not easy to describe (the manual fails to do so too — see box) although the Z axis is simple to understand; it again sends its value to the L row of the matrix. Whichever destination you choose, you can determine the Morphée's response mode. In Tap mode, the bottom left-hand corner returns a value of (0,0), the top right-hand corner generates (127, 127), and the output returns immediately to (0,0) when you stop touching it. Hold mode is the same except that the last value is retained when you lift your finger. In Scan mode, the first place you touch is defined as (0,0), the rest of the pad is scaled appropriately, and the values slew (rather than jump) back to zero when you lift your finger. In Hold Scan mode, the last value is again retained when you lift your finger.

Not just a handy recess for spare change, the wooden groove above the keyboard is also a ribbon controller.Not just a handy recess for spare change, the wooden groove above the keyboard is also a ribbon controller.The ribbon controller can also be routed to any available parameter via the matrix and it offers the same absolute, absolute/hold, relative and relative/hold modes as the Morphée. These can again be chosen on a per-patch basis which, in my opinion, makes it a very friendly ribbon controller. Similarly, the wheels have per-patch options, and the modulation wheel offers four destinations: the Matrix, the Master Cutoff, the LFO1 amplifier and Vibrato. The last of these directs a dedicated (and otherwise invisible) sine wave LFO with three predefined speeds of approximately 3.4Hz, 5.6Hz and 10Hz to the pitch, releasing the programmable LFOs from this simple duty.

The Arpeggiator, Sequencer & Matrix Arpeggiator

The PolyBrute incorporates a 32-note arpeggiator, a 64-step sequencer and a 16-step 'matrix arpeggiator', and all the data that define each of these are saved on a per-patch basis.

Arpeggios can cover one to four octaves, with six conventional playback modes plus a seventh that generates quasi-random patterns from the depressed notes. The rate ranges from 30 to 240 bpm, and you can define the clock ratio, Gate length and Swing. The other controls are for Hold and Tap Tempo. If you stumble across something that you want to keep, you can convert arpeggios into sequences, either to save them or to edit and then save them, which brings us to...

PolyBrute's Arpeggio and Sequencer controls.PolyBrute's Arpeggio and Sequencer controls.

The PolyBrute's sequencer. In some ways this is similar to a polyphonic version of the MatrixBrute v2 sequencer, but offering six-note polyphony on each step. When recording in real time (with or without the built-in metronome) it stores velocities, durations and the movements of up to three knobs and sliders, and you can overdub if you have too few hands to do everything simultaneously. You can also enter and edit sequence data in step time, adding notes, erasing or replacing notes, adding accents (which set the note velocity to 127 no matter what the recorded velocity is), adding slides and ties, changing the durations of individual notes or all of the notes in a step, muting and unmuting steps, muting updating and erasing modulation tracks, and more. You can also copy and paste short sequences to create longer ones up to the maximum of 64 steps. (Sequences can have any length up to the maximum, even though it appears at first sight that only 8, 16, 32 and 64-step options are available.) Other commands allow you to delete sequences, erase notes while retaining the modulation tracks, erase the modulation tracks while retaining the notes, and to copy and paste sequences (with or without the automation tracks) from one patch to another. You'll make a lot of use of the matrix buttons when programming all of this because each group of three rows displays the on/off, accent and slide status of each note. The colours of the buttons are also key to programming, because they are the means by which you can tell what's active and what it's doing as well as when.

There are four playback modes. Forward, pendulum and random are self-explanatory while the fourth (Walk) is a quasi-random walk through the sequence with each step having a 50 percent chance of moving to the next, a 25 percent chance of repeating, and a 25 percent chance of moving back a step. This is what's called aleatoric music, and the result is not so much a walk as a Pythonesque silly walk. No doubt somebody with a Stockhausenesque bent will love it.

On playback, the keyboard is split, with the Lower zone triggering and transposing the sequence and the Upper zone allowing you to accompany it. However, the sequence replays at the pitch at which you recorded it only when you press C2. When you play above or below this, the sequence is transposed accordingly. This means that, if you recorded a sequence in F#, it will be replayed in F# when you press C. If you then want to transpose it to (say) A#, you'll have to play... umm, let me see, that would be E. If you're playing over the top of this, you're going to get confused. Arturia suggest that you record every sequence in C and then transpose it as required, but better solutions would be to provide a root note parameter, or to make the note that initiates playback the root. Furthermore, any accompaniment played in the Upper zone appears to take precedence over whatever is being played by the sequencer, so note stealing can occur and, if you get carried away, you'll end up with little or nothing emanating from the sequencer itself. But whatever else is happening, the output from the sequencer is always available as a polyphonic source in the matrix, so you can obtain yet more modulation effects on a per-voice basis.

There's also a third SEQ/ARP mode called the Matrix Arpeggiator, which is a hybrid of the arpeggiator and the sequencer that allows the matrix buttons to determine which notes in a held chord are played on each step, at what octave and with or without a slide or accent or both. Again, you can record three modulation tracks, and many of the sequencer's playback facilities are duplicated here. It can create some great little patterns, but be careful if you activate Hold and press new notes while it's playing because the results are not always predictable. The other feature it offers is the ability to shift notes upward or downward by a semitone to create what Arturia call 'approach notes'. These can be musically useful to lead from one repetition to the next.

Published October 2020