You are here

Pittsburgh Modular Microvolt 3900

Semi-modular Synthesizer By Gordon Reid
Published February 2019

Pittsburgh Modular Microvolt 3900

Pittsburgh's desktop semi-modular has its eye set firmly on the experimental.

The Microvolt 3900 is a monophonic synthesizer in a chunky steel desktop box that also sports a lump of wood at the front to reassure you of its analogue credentials. It comprises a single 1V/oct oscillator, a low-pass filter and an unusual VCA section, all controlled by an LFO, a random CV source, a looping AD 'Function Generator' and an ADSR contour generator. It also includes an arpeggiator that offers a limited 16-step sequencing capability. All of this is sensibly pre-patched within the synth, but a 39-socket patchbay elevates it to semi-modular status, particularly since you can interrupt the signal path to create new audio routes as well as new modulation possibilities. The sockets themselves are all 3.5mm, which makes it physically compatible with most Eurorack equipment, and the CVs and gate voltages are chosen so that you should have no problem hooking it up to such as system. For clarity, inputs are shown by a box around the text, while outputs are shown using plain text.

The Signal Path

The single VCO generates four underlying waveforms: sine, sawtooth, pulse and 'fold'. The unadulterated sine wave is a little brighter than a pure tone, while switching on the Harmonic Sine function creates a considerably brighter waveform that, on the 'scope, looks much like an unrectified half-wave. The sawtooth is closer to the ideal, and is bright and raspy. The pulse wave has a minimum static duty cycle of around 6 percent, and a maximum of around 94 percent, but you can drive it down to 0 percent and up to 100 percent using pulse-width modulation. Fold is interesting because this starts with a sine wave and (as the name suggests) increases the amplitude and folds the waveform into more complex shapes as you increase the Fold Timbre level. Without resorting to patch cables, you can modulate the oscillator's pitch, the pulse width, and the amount of fold using the output from the Modulation Sources section, which we'll address in a moment. But if you want to go further, there are CV inputs in the patchbay for pitch, FM, pulse width and the Fold Timbre, and you can use these to create subtle or extreme modulation effects.

The manual states that the oscillator's frequency range is variable from sub-audio to ultrasonic, and my tests showed that I could drive it down to frequencies that are best describes in terms of seconds/cycle, and up to 20kHz and beyond, which is a phenomenal range. However, this was also when I discovered that the 3900 has no coarse tuning control. The fine-tuning fader offers a total range of 13 semitones but, if you want to sweep the frequency over a wider range, you'll have to patch a suitable CV into the oscillator FM input.

You can mix any combinations and amounts of the four waves within the Waveform Mixer, which also allows you to add noise (which is, strictly speaking, generated within the Modulation Sources section) and external audio. The input for the external audio is in the patchbay, and the Mixer allows you to set its preamp gain as well as the level injected into the signal path. This makes it possible to obtain overdriven sounds at sensible levels, whether derived from external audio or (with suitable patching) from the internally generated waves.

Following the Waveform Mixer, the signal reaches the so-called Binary Filter. This is a 12dB/oct low-pass filter that offers the expected cutoff and resonance controls as well as a cutoff frequency modulation CV input. I measured its cutoff frequency to range from 12Hz to more than 21kHz, which means that, with patching and careful setting up, you can also use it as an audio oscillator.

The strange name for the filter seems to relate to the fact that it offers two modes. The first is called Stable and produces the usual resonant low-pass filter response, although without the filter ever reaching self-oscillation. With the resonance set to a high value in this mode, the sound was slightly different from what I expected, and viewing this on the 'scope showed that at least two bumps are created in the spectrum, albeit with the second having a much lower amplitude than the main resonant peak.

The second mode is called Unstable Resonance, and this almost always oscillates. I say 'almost' because you have to kick the oscillation into life; it doesn't start of its own accord. But once you've done so, it tends to be self-sustaining unless you push the resonance slider close to its maximum, at which point the oscillation disappears. Yes... I said 'close to its maximum' not 'close to its minimum', which indicates that the control is inverted in this mode. I've not seen this quirk before, and I can't say that it appeals to me. Also, the amount of resonance can affect the filter cutoff frequency, which is odd. This mode is definitely designed for experimentation rather than deterministic sound design and you'll either like its weirdness and unpredictability or you won't.

This then brings us to the VCA section, the front end of which is a two-channel mixer that allows you to determine the balance between the signals presented to the amplifier by the filter and by the patchable VCA input. This also offers two modes: VCA and Low‑pass Gate, the latter of which the manual describes as a "modern interpretation of Don Buchla's famous low‑pass gate". That sounds reasonable, but when I read (and I quote) "Where a standard VCA works in two dimensions, the Microvolt 3900 dynamic VCA technology adds a third dimension" I have to wonder about the new clothes hanging in the emperor's wardrobe. Putting the hyperbole to one side, the VCA mode acts as you would expect, making the signals passing through the amplifier louder and quieter as determined by the panel settings and any CV applied to the VCA gain control input. In contrast, the Low‑pass Gate mode seems to add compression and a low-pass filter to the signal chain so that it allows more high-frequency content to pass when a sound is loud than it does when it's quiet. I like this a lot. Two further controls determine the action of the so-called Pluck generator which, when on (and if the attack of the ADSR is fast enough) generates a simple AD contour whose attack is always instantaneous and whose decay time is determined by the Dynamics Response knob. There's also an Initial Gain control called Dynamics (exactly the wrong name, in my view) that allows you to create drones and other effects.

The final module in the pre-patched signal path is an output section that controls the final signal level. It comprises a drive control that can push the signal into clipping, and an output level fader.

The Microvolt's signal path is clearly mapped out on its front panel.The Microvolt's signal path is clearly mapped out on its front panel.


The 3900 offers four modulation generators: the LFO and quasi-random signal generator within the Modulation Sources panel, a looping two-stage contour generator called the Function Generator, and a conventional ADSR contour generator, the last two of which are by default triggered and gated by MIDI Note On and Off.

The LFO generates two families of waveforms. The first is triangular and, above a certain frequency, you can determine its shape from a ramp through to a triangle and all the way to a sawtooth. The second is a pulse wave with a duty cycle that (again, above a certain frequency) varies from about 4 percent to about 96 percent. At low frequencies the wave shaping has no effect, which is strange. The LFO offers two ranges, one that can be specified at its slowest in terms of minutes/cycle(!) and the other of which can reach as high as 136Hz, which is fast enough to generate some interesting AM and FM effects.

The second Modulation Source is called the Complex Random Generator, and this generates two signals — a CV in the range 0-5 V, and a +5V gate. It works a bit like a six-step sequencer that has a quasi-random chance of the CV being changed and a gate being output on each step. The likelihood of these events occurring is set by the Random Complexity slider which, since it controls the chance of a gate being generated, feels a bit like a rate control even though the underlying clock is actually the clock selected in the MIDI/Arpeggiator section. (See box.)

Described incorrectly in the manual as an attack/release contour, the Function Generator is an attack/decay contour generator with optional cycling. With the attack and decay stages set to their minima and cycling on, the waveform obtained is a triangle with a maximum frequency of around 200Hz. Increasing the attack or decay times skews the waveform toward a ramp or sawtooth wave respectively, and the Rise Mod and Fall Mod knobs (which are described in the manual as controlling the symmetry of the modulation signal) determine the amount by which any CV inserted into the Function Mod input affects those times to modify the shape and frequency of the contour or oscillation.

The manual's description of the ADSR contour generator is even more confusing, talking (as it does) about smoothing gates and triggers that pass through each of its four stages. But there's nothing unusual about it. I measured the response and found that the maximum attack time is rather meagre at just 3s, while the decay and release stages have maxima approaching 15s. It was while testing this that I found that the contour on the review unit has a slight voltage offset — a sustain of zero isn't quite equal to 0V, which can be a pain in the arse.

At this point, it's worth noting that, in common with other semi-modular instruments, the 3900 focuses too much on the exciting stuff and too little on the glue — VCAs, mixers and multiples — needed to take full advantage of them. Consequently, I'm grateful for the single Inverting Mixer/Splitter/Offset panel found in front of the patchbay. This comprises two inputs, two outputs and a bipolar level control, and can act as a two-channel mixer, a two-way multiple, an inverter and, with no inputs, a DC source. It's hugely useful, and I wish that the 3900 sported at least half a dozen of these.

The MIDI/Arpeggiator Panel

The final panel comprises six multi-function buttons that control a plethora of functions. Some of these are obvious: you can transpose the pitch in octaves between -3 to +2 octaves (the manual says ±2 octaves, which is another error), hold the current note, switch the arpeggiator on and off, and tap the internal clock tempo. Others are less obvious. For example, the arpeggiator, which has a maximum capacity of 16 notes, has two modes: 'as played' and sequenced. You can enter a sequence of up to 16 steps (notes and rests, but no ties) but the method is clunky because the sequence loops as you enter new notes. What's more, it's not possible to enter notes without using a MIDI keyboard, and committed analogue fanatics are not going to like that. Worst of all, there's no way to store the sequence. This means that it's lost as soon as you release it which, for me, makes it pretty unless.

The Microvolt's connections are on the main panel, with just a power switch and a socket for the external PSU at the rear.The Microvolt's connections are on the main panel, with just a power switch and a socket for the external PSU at the rear.Five further arpeggio functions are accessible using button combinations. The first four are Hold (continue playing the arpeggio after you release the notes), Transpose (transpose the arpeggio when the next note is played rather than resetting it), Random (play the notes in a random order) and Range (play the arpeggio or sequence over a range of one, two or three octaves). The fifth is a random arpeggio generator that, as the name suggests, generates an arpeggio of random pitches when three of the buttons are switched off and then pressed in the correct order. I don't know if you'll find a use for this. I didn't.

Four pitch-bend options are also accessible using button combinations. There are described in the manual as ±2, ±4, ±8 (a fifth) and ±12 semitones. Quite apart from the fact that a musical fifth is seven semitones (yet another daft error), the actual ranges are ±2, ±5, ±7 and ±12 semitones, which, musically, make much more sense.

In Use

Out of the box, the 3900 looks and feels great; solid and chunky despite its diminutive size. With 16 potentiometers, 19 faders, 12 buttons and its 39 sockets it presents a lot of controls in a small space, but I never found it to be cluttered, which is a compliment to the design team. The use of arrows on the control panel to illustrate the internal connections between modules is useful, and the blue LEDs on the fader heads adds to the feel-good factor, especially since many of them vary in brightness to show what's going on at any given moment. Consequently, I can't understand Pittsburgh's apparent desire to make the 3900 seem more esoteric than it is. Had the company offered a specification and eliminated the obfuscation and numerous errors in the manual it would have been a far more pleasant experience to come to grips with it. How do the two contour generators interact? I still don't know. What the heck is a "highly tuned ADSR envelope generator"? I have no idea, particularly since I'm not aware of having ever encountered a poorly tuned one. How do you "blow out" a waveform, and why should you want to? You could argue that it doesn't matter whether you understand why the filter is doing what it's doing, or what the function generator is, or what the relationship between the Pluck mode and the ADSR is; if you like what's coming out of the 3900, why should you care? If you're happy with that, great... but I don't accept it.

But then there's the sound, which can be gentle and clean, thick, distorted and aggressive, and all stages between. With the filter in its stable mode, I obtained some lovely lead sounds from the Harmonic Sine and Fold waveforms, so it's a great shame that the 3900 lacks portamento because it would be a more than passable lead synth if that were available. Likewise, it was easy to programme beefy basses, especially when I removed the sine wave from the input to the filter and used the patch bay to mix it directly into the VCA. (That's a Yamaha CS-series trick that's always worth remembering.) I was also able to obtain interesting AM, FM and formant-type sounds, and I was very impressed by the arpeggios that I could obtain using sounds shaped by the Low‑pass Gate and Pluck functions. If anything eluded me, it was orchestral imitations; the 3900 refused to recreate my standard repertoire of brass, woodwind and strings patches. But no matter... I doubt that those types of sound would be why people might be attracted to it. As Pittsburgh state in many places, it's designed for experimentation and you'll find it much simpler to wrest bleeps, squeals, bloops and thwaps from it, especially with the filter in Unstable mode.

Ah yes, Unstable mode. I thought that I was going to report how exasperating I found this until I discovered the immense percussion sounds that it's capable of generating. It's no exaggeration to say that the 3900 may have created the meatiest kick drums I've ever programmed on an analogue synth! If there's a problem with the sound, it's the presence of the thumps (which are not the clicks that can be generated by rapid contours) that occur whenever the VCA attack and release times are at or close to zero. Depending upon the sound you're trying to create and how you intend to play it, this may or may not deter you but, whether it's a problem with all units or just the review unit, I would much prefer that Pittsburgh eliminate it.

Inevitably, given its price and size, the 3900 has limitations, the most obvious of which are its single VCO and its limited patching capabilities. To overcome the former, I tried to create meaningful sounds using the Function Generator and LFO as bass oscillators, but the results always fell short of my expectations. Regarding the latter, its dearth of accessible VCAs, mixers and multiples greatly limits its capabilities. To understand this, try to create a standard monosynth patch with pitch tracking and voltage-controlled modulation of the oscillator pitch and filter cutoff frequency while the filter is being controlled by the ADSR contour. No, don't bother... you can't. Were I ever to use a 3900, it would almost certainly be in conjunction with other modular equipment to overcome issues such as these.

Finally, it's worth noting that the 3900 uses an external power supply but, for once, that's forgiveable because of the shape and diminutive size of the unit. What's less forgiveable is the lack of any cable stress relief that might stop the barrel connector being removed accidentally.


When I saw the pre-release information circulated before the Microvolt 3900 was announced, I was impressed. It seemed to pack a lot of functionality into a small footprint, both in terms of the modules delineated on its front panel, and the amount of connectivity offered by its patchbay. I was also intrigued because it seemed to promise some unusual capabilities. My preconceptions were correct.

If you're looking for a conventional synthesizer on which to play conventional solos and bass lines with a minimum of fuss, the 3900 is probably not for you. Nor is it an ideal entry-level instrument because Pittsburgh seem intentionally to have made it arcane in places. But if you're looking for a system on which to experiment with sound to a greater degree than is possible using an integrated synth, or for something that will spice up a quotidian Eurorack setup, it should certainly be of interest. Despite some shortcomings, there's a lot to like about the 3900, but only you can decide whether the format and the mix of features is right for you.

MIDI & Clock

Pittsburgh Microvolt 3900 MIDI panel.Pittsburgh Microvolt 3900 MIDI panel.Found at the top of the patch bay, the MIDI panel offers a single input and three outputs. MIDI is delivered to the synth via a 3.5mm socket, so Pittsburgh supply a 5-pin DIN to 3.5mm TRS converter cable for this. You use a combination of buttons in the MIDI/Arpeggiator section to set the channel to which the 3900 responds and, having done so, you can play the synth without any need to patch CVs or gates to get things working.

The first of the two outputs in the MIDI panel carry conventional 1V/oct pitch CV and +5V gate signals, and you can direct these anywhere you want within the synth. The third is called Mod. Although the manual states boldly that the 3900 responds to MIDI control change messages, it actually responds to just a single assignable MIDI CC at any given time.

The 3900 responds to three clock sources. These are the internal clock, MIDI Clock with six ratios ranging from 1:1 to 1:32, and external (analogue) clock, again with six ratios ranging from 1:1 to 1:32.


  • It's an attractive box of tricks.
  • It can sound excellent.
  • The Low-pass Gate is particularly euphonic.
  • It packs a lot of features into a small space without being cluttered.


  • It's not clear what some functions are doing.
  • There are insufficient mixers, multiples and programmable VCAs.
  • The manual is riddled with errors.
  • The sequencer function is pretty much pointless.


The Microvolt 3900 was designed as a sonic toolkit for experimentation and sound generation rather than as a conventional melodic synthesizer although, within its limitations, it can do the latter rather well too. It offers a lot of features, it's sensibly priced, it's nicely built and it can sound excellent. But it's not for everybody. If you're tempted, check it out carefully to ensure that it's right for you.