Swedish designer Wowa Cwejman has built a reputation for exclusive analogue synths. Now he's going modular, starting with the VM1 Voice module...
It has been more than two years since the Cwejman Sound S1 Mk2 appeared and, during that time, the company has restructured itself as simply 'Cwejman' and garnered a reputation for building exclusive, high-quality monophonic synthesizers. For me, it was quite amusing to watch this happen because, as one of the fortunate few to own one of these synths, I watched with amazement as arguments raged over whether it was a hoax or not. While I twiddled its knobs, people on various synth forums presented cogent arguments that it didn't exist. Fortunately, it didn't take long for Wowa Cwejman to consolidate an enviable reputation and when, in the autumn of 2005, he uploaded PDFs describing eight forthcoming modules for a proposed modular synthesizer, few doubted that they would appear. And that's just as well because, nine months later, shipments have begun.
All the modules share the design ethic of the S1 Mk2, as well as its knobs, switches and connectors. Wowa Cwejman claims that the modules are built to the same quality as the S1 Mk2, and I'm happy to accept this, but will point out that their performance could be compromised if they were used with an inappropriate or poorly adjusted power supply. This might not be an issue if Cwejman manufactured his own rack and PSU, but he has so far chosen to adopt the standard Eurorack format and Doepfer power connectors. This means that the modules have no Swedish home of their own, but are instead compatible with Doepfer and Analogue Systems frames. (Note, however, that if you mount them alongside Analogue Systems' modules in an RS Integrator rack, you'll find that there is a slight gap between any AS modules and any Cwejman modules mounted beside them). I understand that Cwejman is now designing his own cases, and that these will be much shallower than other manufacturers', in keeping with the shallow dimensions of the modules themselves.
Cwejman has eschewed conventional wisdom, which suggests that synthesizers sound best when designed with large, heavy, discrete components, and has instead employed miniaturised, surface-mount technology on boards protected from prying fingers and stray electromagnetic noise by metal cages. Happily, the cages permit access to trimmers that permit calibration as and when necessary. More evidence of modern thinking is the lack of wires. All the modules are of single-board construction, with every component mounted directly to them, so there's no need for any wiring other than the power connectors. This bodes well for low noise and long-term reliability.
Looking at the modules, it's obvious that you can build a powerful modular synth using them. However, one of them stands slightly apart from the others. This is the VM1 Voice Module, which is (almost) a complete synthesizer in a single module.
The idea of the Voice Module is not a new one: in 1979, the Roland 110 module for the System 100M combined a VCO, a low-pass VCF, and a VCA. This had a pre-patched audio path, but was semi-modular in that you could override the internal connections by inserting plugs into the appropriate sockets. Sure, there were no on-board contour generators, no LFO, and no noise generator — but there was still a great deal that you could do with it. In particular, you could connect four modules in parallel and control them from the Roland 184 keyboard to build a primitive four-voice polyphonic synthesizer.
Interesting though the Roland was, it wasn't a patch (sorry!) on the much earlier Oberheim SEM (Synthesizer Expander Module). Released in 1974, the SEM combined two VCOs, two contour generators, a multi-mode filter, an LFO and a VCA in a mains-powered box. Sold as both a synthesizer expander and a laboratory signal generator, the SEM's finest hour came when four of them were placed alongside one another in the Oberheim 4-Voice, and a little later when four more found a home in the synthesizer's lid, to create the 8-Voice. With primitive patch-storage capabilities, the 4-Voice was the first true, commercially available, polyphonic synthesizer, pre-dating even the Yamaha GX1 by a gnat's wotsit. And a well calibrated 4-Voice is still a joy to hear, more than three decades later.
In many ways, the Cwejman VM1 looks like the product of sex between a Roland 110 and an Oberheim SEM. But although the three products can be used in the same ways, they sound different and their architectures are very different. The SEM scores by having two oscillators and an LFO but the VM1 wins over its predecessors in every other department: its oscillator is more flexible, its filter is more flexible, the dual contour generators are more flexible and, of course, it's semi-modular (which unmodified SEMs are not).
The upper four fifths of the VM1 are densely populated with 19 knobs and three small switches but, like the Cwejman S1 Mk2 on which it's modelled, it avoids appearing cluttered through careful layout and legending. The VM1 also boasts 11 CV inputs and, as on the S1 Mk2, you'll find them along the lower part of the panel. This is a neat arrangement, although it means that some sockets are positioned away from the knobs that control them. However, unlike the S1 Mk2, it has only two types of I/O legend: white text on a dark background for an input, and dark text on a white surround for an output.
The oscillator offers the same seven waveforms as the S1 Mk2. These are sine, triangle, sawtooth, saw + triangle, pulse + triangle, saw + pulse, and variable-width pulse, with a nominal duty-cycle range of five percent to 95 percent. As I observed when I reviewed the S1 Mk2, the pulse wave responds to pulse-width modulation in all three of the waves to which it contributes, and you can modulate it to zero percent at one extreme and 100 percent at the other in the composite waves, to reveal the unadulterated triangle and sawtooth waves. In contrast, the fine-tuning range has been reduced to ±4 semitones, and the S1 Mk2's range control has been replaced by a Coarse tuning knob that ranges from 16Hz to a little over 15kHz with no CVs applied, and from unmeasurably low (on my analyser) to supersonic if you ask it to.
Given the quality of the S1 Mk2's waveforms, it's no surprise to find that the VM1's are good, although they are not mathematically accurate. For example, there are slight bumps in the somewhat asymmetrical 'sine' wave, and the sawtooth wave exhibits a pronounced shark's tooth profile (and is, in fact, a ramp wave). Nevertheless, it would be wrong to judge the VM1 harshly; I have seen many synthesizers' waveforms that conform far less closely to their ideals.
The oscillator section boasts four control inputs. The first two — CV1 and CV2 — conform to the one-volt-per-octave standard and I found that they tracked accurately over a large number of octaves. Alongside these, there's a PWM input, and the associated knob attenuates the CV to the desired level. Finally, there's a sync input that slaves the VM1 oscillator to an outside source. Note, however, that inserting a signal here means that the VM1's pitch will not change as you play up and down the keyboard (or use any form of controller) unless you alter the pitch of the oscillator providing the sync signal.
A fifth input, located in the filter section, allows you to insert an external audio signal into the VM1. You can then balance the amount of oscillator-generated audio and the amount of external signal presented to the filter's audio input with the Mix knob in the oscillator section. The geometry of the controls is a little confusing, but simple when you've worked out what's happening. Unfortunately, once you insert a patch-lead here, there is bleed from the oscillator even when you turn the knob fully clockwise to the 'external audio only' position. Until corrected, this makes the VM1 unsuitable for filtering and warping external music signals from CDs and so on.
The four-pole filter itself has a remarkably high feature count. Not just a multi-mode filter with low-pass, band-pass and high-pass modes, it allows you to morph from one mode to the next, and provides a CV input that allows you to determine its action between low-pass (at one extreme) and high-pass (at the other) using an appropriate control voltage.
Setting the filter to LP, I tested its response and was impressed. With the Q-Peak (resonance) set to maximum so that the filter was self-oscillating, the spectrum analyser showed that, even without CVs applied, the filter cutoff frequency ranged from 17Hz to 19.5kHz. By applying CVs, I was able to reduce the lowest frequency to just 1Hz, while the highest disappeared into supersonic territory. Interestingly, self-oscillation does not occur unless you give the filter a kick, either by passing a signal through it or by changing the cutoff frequency. I don't think that I have encountered this before on an analogue filter, and it suggests that there is some form of noise reduction within it.
I found that self-oscillation produces a wave quite happily when used in LP and BP modes, but that there is a 'dead spot' between the two at which there is little signal generated. When moving from BP to HP there's a similar dead spot, but the waveform becomes more complex as you approach and pass through it, and the resulting sound is brighter. A quick peek at the analyser showed that, far from being a sine wave, the oscillation has a harmonic series reaching up to 10kHz and beyond. This complexity is maintained as you sweep the mode all the way to HP, so it's clear that the audio filtering will be more complex than the simple, idealised view of a LP/BP/HP filter leads you to expect.
Everything so far has been straightforward, but the Osc/AR and Mix switches, the Ext CM (Cut-off Modulation) input, and the CM1 knob warrant detailed explanation...
Firstly, the Osc/AR switch determines whether the internal CM1 control voltage is derived from the AR contour generator or from the oscillator's output, the latter of which makes all manner of audio frequency modulation (FM) effects possible. Secondly, the Ext CM input allows you to insert an external signal into CM1. With me so far? OK... The CM1 knob then determines the source and depth of the modulation, either from the internal sources (anti-clockwise from 12 o'clock) or from the external source (clockwise from 12 o'clock). However, if you flip the second switch to Mix, the CM1 knob becomes a mixer, mixing the internal and external signals from 100 percent internal to 100 percent external, with a 50/50 mix at the 12 o'clock position. Phew!
Complex and powerful though this appears to be, there's another pre-wired CV input to the filter! This is the ADSR contour generator, which can be applied in normal and inverse polarity, with the polarity and depth determined by the CM2 knob.
Despite its patching capabilities, the VM1 is not truly modular; it's a semi-modular voice module, able to respond to external CVs. This means that all the internal functions are already connected in sensible ways, so that you can obtain conventional (and many less conventional) sounds without using any patch cables other than those that provide the pitch CV and Gate signals that let you play it.
As implied above, the VM1 hosts two contour generators — an AR and an ADSR — the standard configuration of the ARP Odyssey and ARP 2600. I've mentioned that the ADSR is pre-patched to the filter CM2 input, but you can select which of the two contours controls the VCA. This nod in the direction of ARP is all the more apparent when you spot the Trig input alongside the Gate, which allows you to retrigger the ADSR even while the Gate is held open.
I tested the speed of the ADSR using noise inserted via the Audio input and the Osc/Audio Mix knob, and by blipping the contour using an external LFO. I was impressed! The Attack was just 0.5ms, which is extremely rapid. The Release was 2ms, which is also way faster than the norm. With a total AR cycle of just 2.5ms, one would have to say that the VM1's contours are up there with the very best of them. However, I also found that the Sustain level of the ADSR was somewhat lower than the peak of the Attack phase of the contour. It turns out that there is a trimmer for this, but I still found it impossible to get the level right. Despite moving the trimmer by microscopic amounts, the level was always a tad high or a tad low. Of course, you wouldn't notice this unless you were using the ADSR to control the self-oscillating filter, but it's not a problem that I've encountered before, so it's worth noting.
The final input is marked Ampl Level and this allows you to control the amplifier gain using an external CV. There are many ways to use such a control, but one jumps out at me: the VM1 is velocity sensitive! That's an unexpected bonus.
Finally, we come to the two outputs. The first — marked Filter — taps the signal before the VCA, and is useful for filtering external signals when no keyboard is on hand to offer CV and Gate signals. The second — marked Voice — is what one would normally recognise as the output of a conventional synthesizer.
It's important to remember that the VM1 is not intended to be a complete synthesizer. It lacks an LFO and a noise generator, it has no multiples (there's no way to direct a single CV to multiple inputs, for example) and there's no slew generator for portamento. Nonetheless, it's remarkably flexible, and only the lack of an LFO stops it competing with (and beating) many single-oscillator analogue monosynths, both in terms of features and sound quality. You only have to consider the flexibility of the three CM1 inputs to realise that you can do things with the VM1 that would normally be the preserve of a true modular synth, not an integrated one.
Quality is also a byword for Cwejman products. I was very complimentary about the S1 Mk2, and I feel the same way about the VM1 — especially its 3.5mm sockets, which feel much more solid and robust than those of its competitors.
Listening to the VM1, I found that it shares a sonic character with the S1 Mk2, one that I have previously described as 'precise'. It is neither overpowering like some American synthesizers, nor thin and unimposing like some Oriental ones. Instead, it produces strong, clean sounds that it colours and shapes as you ask. For example, I patched a bass sound using the oscillator's square-wave output as the sound source and also used this to modulate the filter. I then set the cutoff frequency quite low and the resonance quite high so that the ADSR could create a filter sweep, and connected an external LFO to add a slow PWM. The result was excellent; deep and warm, but without any muddiness. The same was true for lead-synth sounds and imitative patches such as trumpets and flutes. But what I would really love to try is creating some polyphonic sounds using multiple VM1s. Oh well... one can dream.
Firstly, I'm not a fan of the coarse- and fine-tuning arrangement. While it's suitable for a signal generator, it's less appropriate for playing melodies in tune with other instruments. I would thus welcome a return to the S1 Mk2's octave- and fine-tuning arrangement. I've also mentioned the audio bleed in the filter's input mixer, and I must criticise the lack of a manual. Two other points are (perhaps) not so much criticisms as observations: the VM1 would be a much better module if it had a multiple and — as the output is very hot — an output level control. It would have to be a little larger to accommodate these, but it would not compromise the concept, and would eliminate the need for two extra modules (multiple and amplifier) for most uses.
Finally, what about the price? When viewed in context, the VM1 is not particularly expensive. An RS95e oscillator, an RS110 multi-mode filter, an RS180 VCA and a pair of RS60 contour generators from Analogue Systems would offer the same set of facilities (albeit with more patching points and greater flexibility) for a total of £445, while the VM1 will set you back £400 or so (the Sterling equivalent of its 549 Euro price at the time of going to press). That's close enough for you to base a purchasing decision on factors other than price.
In conclusion, the VM1 is a well engineered voice module that almost (but not quite) elevates itself to the status of a single-module synthesizer. It sounds excellent, and the price won't give your bank manager apoplexy. If I were buying my first modular synth today, I would look at including at least one of these, whether the rest of the system was sourced from Cwejman, Analogue Systems, Doepfer, or any combination of the three.