XILS Lab's XILS 3 is a recreation of the EMS VCS3, but there's more to it than that — much, much more.
The VCS3 was the world's first integrated, portable synthesizer, predating the Minimoog and the ARP 2600 by a year. It was an arcane, unstable, unmanageable and unreliable piece of junk that you were unlikely to be able to play in tune over two octaves for more than five minutes, and once decent synths appeared it was barely worth the next‑to‑nothing that I paid for mine in the early 1990s. Alternatively, it was, and remains, the most interesting, innovative and creative audio sound generator and signal mangler ever released, worth every penny of the £6600 that someone recently paid for one on eBay.
Given the DSP power available nowadays, you might think that it would be possible to emulate the VCS3 in software. You could imitate its unstable oscillators, model its unpredictable filter, recreate its loopy envelope generator and all its other facilities, iron out its idiosyncrasies, add a few enhancements, and then stick a pretty GUI on the front that forces players to approach it in the same way as the original. But if you did so, would you destroy the essential character of the original? No doubt XILS 3 will provide the answer.
So, what is XILS 3? On the surface, it's a soft synth designed to look, feel and sound like a VCS3. However, as we delve deeper, we'll find that it's much more than that. What it is not, however, is a stand‑alone application; it can only be opened as a VST, AU or RTAS plug‑in. But no matter what else it may or may not be, the question on everybody's lips is 'How good an emulation of the VCS3 is it?' Up until now, there has never been a soft synth that has captured the quintessentially English character of EMS's synthesizers, so let's start with a step‑by‑step comparison of their common elements, all of which are contained in the right‑hand window of the soft synth.
Since the VCS3 is a modular synth, it generates no sounds until you connect its modules in an appropriate fashion. Consequently, its most recognisable feature is its pin matrix, which allows you to direct the outputs of just about everything to the inputs of just about everything else. The XILS 3 imitates this, although not precisely, with the two output channel rows missing (a bad thing), but with separate rows for the two waveforms generated by oscillators 1 and 2 (which is a good thing).
In the earliest version, the matrix offered just one sort of virtual pin, but the latest update boasts no fewer than eight, with positive and inverted connection values of 1.0 (the standard), 0.75, 0.5 and 0.25. This is excellent stuff.
Comparing the oscillators of a VCS3 and XILS 3 immediately reveals differences. For example, the sawtooth waves generated by the soft synth are brighter than those generated by the vintage synth, and its sine wave is much purer. Likewise, the pulse waves generated by XILS 3 are much more precise than those that you'll obtain from the original, although the extreme ranges of pulse width are not available from their respective knobs, so you need to delve into the additional modules (of which more later) to obtain the nasty, thin sounds that oscillators 2 and 3 on the VCS3 can generate. Shaping the triangle waves from their ramp to sawtooth extremes is perhaps more true to the original, although you'll hear clear stepping as you rotate the knobs on XILS 3.
Unfortunately, there's a significant error in XILS 3's oscillator 1, whose Shape control is linked to the sawtooth wave, not the sine wave. This is a shocking mistake!
In addition to the accuracy of the waveforms, there's also the question of aliasing to consider. XILS Lab claims that there is none. That's not true; there's a small amount at the highest frequencies, but XILS 3's performance is better in this respect than many other soft synths, so I'll say no more about it.
Finally, as far as conventional sound sources are concerned, the noise generator on the VCS3 was an aggressive beastie with a coarse texture. In XILS 3, the noise is altogether smoother, which you may or may not prefer.
The VCS3's ring modulator was one of the main facilities that helped it gain its reputation for science-fiction wackiness. No matter what signals you fed into it, it always managed to convince you that the aliens were coming, and that they were not friendly. The implementation in XILS 3 is largely faithful to the original, but it's perhaps a little too polite ("We come in peace, shoot to kill!”) with a little less character.
XILS 3's filter is immediately different from the VCS3's because it offers 12, 18 and 24 dB/octave options, the last two of which will self‑oscillate if required. This is not as extravagant as it may seem. All VCS3s and Synthi As sold from 1969 until 1974 had an 18dB/octave filter that rolled off more gently close to its cutoff frequency, so the inclusion of both the 12dB/octave and 18dB/octave modes is justified. In 1974, EMS changed the filter so that its maximum slope was 24dB/octave, so this option is also appropriate.
Trying to describe the character of the VCS3's filter is tricky. With zero resonance, it still has a slightly nasal character, and it distorts the signal significantly no matter how carefully you set the levels. With the resonance set to anything other than zero it has a unique character, especially on the point of self‑oscillation, at which it exhibits a noisy, uncertain personality. If you invoke self‑oscillation without an input, the signal thus generated changes in timbre and drops in pitch as you increase the resonance so that, if you turn the knob at the right speed, it sounds like a plaintive cat!
Inevitably, the character of the XILS 3 filter is different, most noticeably because it doesn't distort the input to anything like the same degree. What's more, with no signal applied, its self‑oscillation has a breathy quality and never really gets going. If you kick it, you obtain a stronger signal, but never does it sound the same as a VCS3. Don't get me wrong, all three of its responses are fine but, in side‑to‑side comparison, they won't fool you into thinking that you are listening to the original synth.
I'm not sure whether to praise or criticise the Envelope Shaper in XILS 3. I'm inclined to praise it because it offers an additional ADSR mode not found on the VCS3, and this hugely extends the range of sounds available. What's more, it's far snappier than the VCS3's Shaper, which is a good thing. On the other hand, there is an important aspect of the original device (which you might call 'AHRO' for Attack, Hold, Release, Off‑time) that isn't quite right. Unless the Off knob was at its maximum (called 'Manual') the envelope would loop endlessly, generating a trapezoid contour determined by the values of each of its stages. This looping was an important element in many VCS3 sounds and, with the Release time (called 'Decay') controlled via the matrix, it was another of its most identifiable features. XILS 3 recreates the AHRO contour, although with hugely different calibrations of the knobs. What it gets wrong, however, is that looping requires a gate or a trigger to get it going, so you can't perform your Hawkwind impersonations without kicking it into action in some way. Oops!
The spring reverb in the VCS3 is poorly insulated from the outside world, and even tapping the synth's control panel creates a 'boing'. What's more, it's prone to acoustic feedback that causes it to howl and scream like... well, like a VCS3. The three unlabelled reverb modes on XILS 3 are far more civilised, and although the uppermost has much of the 'boing' of a genuine spring reverb, you can never get it to howl. In many ways this is a good thing, but I have no doubt that there are aficionados who will complain. Oh yes, and another thing will annoy purists: the positions of the Mix and Level controls are swapped on XILS 3. It's no big deal, but strange nonetheless.
Each of the Channel 1 and Channel 2 output amplifiers on the VCS3 includes a static filter that acts as a low‑pass (values in the range 0 to 5) or a high‑pass (value in the range 5 to 10) filter. These are basically tone controls, used to tweak the final signal once everything else has been programmed. These are recreated on XILS 3, but with a greater range of effect, which is fine.
The final controls on a VCS3 lie alongside its pin matrix. They comprise the joystick, plus level and pan controls for both Channel 1 and Channel 2. You would think that there's not much to get wrong here, but the Range controls for the joystick's horizontal and vertical movements are swapped, which is confusing if you're accustomed to the real thing. Secondly, the operation of the joystick is reversed (right and up being greater, rather than left and down). Happily, this is easy to fix: just use inverting pins when connecting it and order is restored.
If you're hoping to recreate your favourite VCS3 sounds on XILS 3, forget it. You can't. The sound generators are too clean, too stable and too close to the ideal, the modifiers are too clean, too stable and too close to the ideal, and the controllers are, umm, too clean, too stable and too close to the ideal. The thing is, the VCS3 really is a ghastly piece of engineering and it achieves what it does precisely because its signal paths are so dirty, it's so unstable and it's so unpredictable. By removing this essential un‑ness, you change the nature of the beast.
To illustrate this, I patched a set of basic VCS3 sounds that involve no more that passing the output of its noise generator through a looping trapezoid, then passing the output from this through the reverb and thence to the outside world. This setup requires just three pins, but can create all manner of effects such as footsteps, waves, rain, and some fabulous aircraft sound effects. But, even ignoring the need to gate the envelope, the same patches programmed on the XILS 3 sounded quite unlike the originals and, no matter how much I experimented with knob values, this remained the case. That doesn't mean that XILS 3's patches were unpleasant. Far from it, in fact. But they were different.
I also tried recreating some of the patches contained in EMS's original Synthi Book Of Sounds. In not one case did I obtain what I expected. Take the VCS3's 'Ethereal Voice' as an example. Because XILS 3 lacks the Output Ch 1 and Output Ch 2 sources in its main matrix, you can't patch it in the same way and, even after working out how to re‑jig the patch to obtain the same result, the suggested knob values delivered a blast of sci‑fi weirdness that was totally unlike the gorgeous vocal timbre that the VCS3 delivers. I spent hours with this patch and created some excellent new sounds (which was fun) but I couldn't make the two synths sound identical. Close, perhaps, but not identical. But perhaps this misses the point. XILS 3 was already proving to be incredibly flexible, and it sounded great.
Moving way beyond the capabilities of the VCS3, XILS 3's left‑hand window contains numerous additional modules, in the shape of a MIDI‑sync'ed S&H generator, a MIDI‑sync'ed LFO, a voltage processor, two further dual‑form AHRO/ADSR envelope generators, three further matrices, a detailed keyboard controller, input signal processing including a pitch tracker and envelope follower, and dedicated chorus and delay effects. It's a huge palette of additional facilities, so forgive me if I just concentrate on just a handful that may be of particular interest.
Let's start with unison and polyphony. Yes, XILS 3 can create huge (monophonic) unison sounds and is also polyphonic so, if you've ever wondered what an EMS polysynth might have sounded like had it ever existed (the PolySynthi doesn't count) you can now find out. Beware, however, that some modules are strictly monophonic. In particular, you can't create a musical timbre based on ring modulation and then play it polyphonically because a horrible clangourous noise will immediately ensue.
Moving on, there are scores of new connection possibilities in the three additional pin matrices, and I'm very happy to see that pulse-width modulation is included in these. This makes it possible to create some of the lushest pads I have ever heard, but still with that edgy EMS timbre if desired. Once I had mastered the extra modules and patching facilities, I found that I could create all manner of 'analogue' polyphonic sounds for which one would normally turn to a Prophet, Oberheim or Jupiter. Brasses, flutes, strings and pads to die for — all are lurking inside XILS 3. This is high praise indeed.
Let's also remember the ways in which a VCS3 could accept and use an external signal, either as a mangler or something to be mangled. EMS synths were pre‑eminent in this area for many years, but the input processing modules and dozens of destinations in XILS 3 make the VCS3 look as if it's barely ready to enter primary school.
Finally, special mention has to be made of XILS 3's multi‑layer, three‑part polyphonic sequencer, which was apparently inspired by the unwieldy, radical, digital(!) EMS 256 Synthi Sequencer. Like its forebear, it's complex, arcane, and it's probably unlike anything you've tried before. You won't master it in an afternoon and, if it were to be reviewed separately, it would justify an article at least as long as this one. To check it out, load the factory patch 'XILS Three Layer Seq' and start experimenting. It's instant 1971.
So, is there anything to criticise? Unfortunately, yes. In homage to its inspiration, XILS 3 exhibits a number of bugs and idiosyncrasies. There are obvious ones such as the LFO delay not functioning correctly, and irritating ones such as clicking (even with slow Attack and Release times) when playing notes through the reverb module in Circular Poly (but not Reset Poly) mode. To be honest, I discovered several bugs and errors as I experimented, but I won't list them all here. Most are relatively benign, and when I communicated then to XILS Lab, they immediately undertook to look into them. Provided that they are fixed in the not too far distant future, I won't feel the need to belabour them.
I started out by comparing XILS 3 to the VCS3 because the chaps at XILS Lab make it clear that this is what they want potential customers to do. But by the end of the review I realised that it's not a VCS3 soft synth at all... it's much more than that. Sure, it's based on EMS's synths, and the layout and patching mechanism force you to think about sounds in a 'Synthi' sort of way. But XILS 3 is also stable and playable, and the range of facilities and the range of sounds that you can obtain from it go way beyond anything that we thought was possible in 1969. If you approach it as you would a VCS3 you'll merely scratch the surface of what it can do. Dig deeper, and you'll discover a fabulously powerful new synthesizer.
Over the years, there have been numerous false dawns regarding VCS3 soft synths, and several have been announced on the Web only to disappear shortly thereafter. At the moment, Internet searches lead you to a couple of synths that have matrix patching, but these have different architectures from the VCS3 and make no claims to model the quirks of the EMS hardware. Consequently, if you're looking for a like‑for‑like substitute for XILS 3, I think that you're going to be disappointed. But there is an alternative. As in 1969, you could ignore the EMS (British) modular synth, and travel across the pond to buy a Moog (American) modular synth. Nowadays, the journey is somewhat shorter, and if you hop across the Channel to Grenoble, you could consider Arturia's Moog Modular V, which — by a curious twist — was also developed in part by Xavier Oudin.
- Maximum polyphony: 18 voices.
- Oscillators: three per voice.
- Filter: one per voice, two‑, three‑ or four‑pole.
- Output filters: two per voice (one per output channel).
- Envelope generator: one per voice; trapezoid and ADSR options.
- Ring modulator: one per patch.
- Effects: spring reverb (three options), chorus and delay.
- Sequencer: up to 128 steps, three channels, monophonic or tri‑phonic.
- Pin matrices: four per patch (one emulating the VCS3, three more in the Matrix window).
- Additional modules: S&H generator, LFO, voltage processor, two AHRO/ADSR envelope generators, keyboard controller.
- Input signal processor: pitch tracker, envelope follower, transient detector, gate generator.
- Unison mode: mono, two, four or six layers (monophonic only).
- MIDI control: All parameters controllable.
- Formats: VST, AU, RTAS (Pro Tools 7 or later).
- Mac: OS 10.3.9 or later.
- PC: Windows XP, Vista and 7 (32‑bit).
- System requirements: 1GB RAM, 2GHz CPU.
- Protection: Syncrosoft or iLok dongles.
XILS 3 is the brainchild of Xavier Oudin, a signal‑processing engineer of no little experience. After seven years working on speech processing and noise reduction at Digigram, he spent six years at Arturia working on products such as Minimoog V, CS80V, ARP 2600V and Prophet V. This explains why the housekeeping side of XILS 3 looks so familiar, and any user of Arturia's soft synths will immediately be comfortable with things such as saving, loading, and patch organisation.
In 2005, Oudin left Arturia to co‑found a company named Eiosis, where he developed effects plug‑ins. Then, in 2008, he established XILS Lab so that he could develop XILS 3 and its sister soft synth, Poly KB, which we will review at a later date.
If you just want to dip your toe into the waters of XILS 3, there is an LE version that retains the emulation of the VCS3 in the right‑hand window but omits the additional modules and pin matrices found in the left‑hand one. At just 30 Euros, this is excellent value, and I recommend that you try it even if you have no expectation of upgrading to the full version.