Arturia CS80V

The main view in CS80V. Here the instrument is running as a plug-in under Sonar on the review PC.

Yamaha's CS80 was one of the finest analogue polysynths of all time. How close have Arturia come to modelling its classic sound in software? We put the software up against the hardware to find out...

In almost every way, the CS80 was a seminal synthesizer, marking the moment at which polyphonic synthesis matured from unreliable and hit-or-miss (such as the Oberheim Four-Voice) or huge and unaffordable (such as the Yamaha GX1) to moderately affordable, moderately transportable, and moderately reliable. What's more, the CS80 was an instrument in ways that no previous synthesizer could claim to be. It particular, it sported a beautiful, wooden, velocity-sensitive keyboard that offered polyphonic aftertouch and had the feel of a quality piano. And, in an era when most polyphonic synths were just hyperactive string machines, each of the eight notes generated by the CS80 was formed from two complete synth voices, each boasting a dedicated oscillator, dual filters and dual contour generators. Add to this ring modulation, a chorus/vibrato generator, a huge ribbon controller, plus 22 preset memories and four user-programmable memories (a big deal in 1978), and it's no wonder that it made such an impact, even with its vast price tag (perhaps £50,000 at today's prices).

What's more surprising is that the CS80 is still revered amongst keyboard players more than 25 years later. To understand this, you have to realise that the CS80 sounded, and still sounds, wonderful. It has a timbre that is quite different from the rash of polysynths that followed, and no Prophet, Oberheim, Jupiter or Memorymoog was ever able to recreate its distinctive character. In fact, with the exception of its cut-down siblings, the CS60 and CS50, nothing sounds like a CS80. Until the arrival of Arturia's software emulation, CS80V, that is...

Updates & Formats

CS80V has been around a while, and SOS began a review of it early last year, but the software was updated twice while the review was underway, and the additions, improvements and bug fixes at each revision were quite significant (see www.arturia.com/en/cs80/csupdates.php for a full list), which necessitated changes to the review. The version used for the final draft of this review was v1.2, which has been out since October last year [as we went to press, we heard that Arturia were starting work on another round of updates — Ed].

CS80V can be used within most popular MIDI + Audio plug-in hosts. As well as in a stand-alone version, it can run as a VST, DXi, MAS, RTAS, HTDM or Audio Units plug-in. The PC version works with W95, 98, ME, 2000 and XP, while Mac owners can use Mac OS 9.x and 10.2 or higher. If you're using only the minimum recommended PC/Mac spec of 128MB of RAM and a 500MHz CPU, CS80V is likely to stretch your computer to its limits. As ever, the more RAM you have and the faster your computer, the better.

Although the main screen of CS80V (see above) superficially resembles the top panel of a real CS80, closer scrutiny reveals that Arturia have made many changes to the CS80 panel layout. What's more, as with their Minimoog V, reviewed in last month's SOS, Arturia have also extended the feature-set. Some of these additional features may be accessed by 'flipping up' the virtual panel on the left (the one marked with a signal-flow diagram in the 'closed' view), and the back of the virtual panel also 'opens' to reveal the modulation matrix, of which more later in this review.

Just as with Minimoog V review, it seems sensible to compare the various parts of CS80V 's signal path with the real thing (in this case, a mint, tuned and stable CS80), and assess how close an emulation it is, before considering it as a software instrument in its own right. Let's start, as in most SOS synth reviews, with...

The Oscillators

In traditional dual-oscillator synthesizers, it's usual to find that the outputs from both oscillators pass through a common set of filters and contour generators. Each of CS80V 's oscillators, however, enjoys its own dedicated filters and contour generators, arranged in two rows named 'Voice I' and 'Voice II'. A Mix fader to the right of the Preset selectors balances the relative volume of the voices, much as the traditional mixer would balance the contributions from the oscillators, while another further to the left allows you to detune Voice II with respect to Voice I.

Unlike the original CS80 (which offers just sawtooth, pulse and noise waveforms), each oscillator on CS80V offers four basic waveforms — triangle, sawtooth, pulse and noise. Two waveform-selector switches are provided for each voice — the first selects square on/off, the second selects triangle, or saw, or off. Thus, in principle, the pulse+saw and pulse+triangle waves can be combined for further tonal variation — or each can be used on its own. As on the original synth, there are six 'footage' options provided for each voice, available on two sliders found below the oscillator sections. The noise generator in each voice has its own level fader, adjustable from +16.13dB down to off.

The non-standard 'Initial Level/Attack Level/Attack/Decay/Release' envelope used by the filter envelope in Yamaha's original CS80 (and in CS80V).

The pulse widths of the square and triangle waves are variable, and applying PWM (pulse-width modulation) to the pulse wave has the expected effect, while PWM of the triangle shifts its rather strange 'shark's-tooth' wave from a quasi-ramp (ie. rising sawtooth) wave to a quasi-sawtooth and back again. Oh yes, and while we mention it, just as on Minimoog V, the sawtooth wave, which should offer the usual falling profile, is actually a ramp (ie. a rising sawtooth). The waveform is modulated by a dedicated LFO with a choice of sine, sawtooth, ramp, square, noise and random waveforms, the speed of which ranges from 0.01Hz to 50Hz. The original CS80 has no such waveform selector — it offers just sine-wave modulation.

Further additions are found in CS80V's ability to sync the waveform-modulation LFOs (either to internal clock or MIDI), and in the three LFO trigger options: Mono, which applies the PWM at the same phase to all notes played; Trig, which resets the PWM phase for each new note played; and Free, which allows the PWM modulators to run freely for each note played, without any phase reset. When MIDI sync is engaged, the LFO speed control becomes a tempo multiplier, locking the speed to a multiple of the song tempo. Of these, the least interesting option is the one available on the original CS80. On this, there is just one waveform LFO per eight-voice section so, if you apply PWM, all the pulse waves in the section sweep in synchronisation.

CS80V also has oscillator sync, which the CS80 did not, but true to the original CS80 there is no 'fine-tune' control for VCO1, so this only works as anticipated if you modulate the pitch of VCO1 using the modulation matrix (more on this later). On the Mac version, a fault sometimes caused the sync'ed sound to disappear and be replaced by a dull click. But when carefully set up, the sync is suitably 'tearing', and produces an aggressive range of sounds that were unobtainable on the original synth.

There is also a new button labelled 'Link'. At first sight this appears to re-route VCO2 through the same filters and contour generators as VCO1, mimicking the more common dual-oscillator architecture. However, closer inspection shows that it applies the filter and envelope settings of Voice I to the filters and envelopes of Voice II, still allowing you to adjust other parameters such as waveforms and pitches independently. This might seem strange, but it's ideal for when you're programming sounds in which you want to use the two sections as partials within a composite.

That's the specification. How about the sound? To be honest, the raw timbres of CS80V 's oscillators are not quite authentic when compared to those of the test CS80. To make this comparison, we set the cutoff of the low-pass filter at its highest setting on the original CS, set the cutoff of the high-pass at its lowest setting, and turned the resonance to zero, thus minimising the effect of the filters. With the filters in CS80V switched off altogether, we could directly compare the oscillators. Of course CS80V 's triangle wave has no equivalent on the CS80, but the sawtooth waves sounded similar when playing single notes. However, the same cannot be said for the pulse waves. These sounded similar at high frequencies when set to produce square waves, but CS80V sounds far squarer and more hollow as you play down the keyboard. By the time you reach the bottom notes, the two sound quite dissimilar (possibly because the square waveforms on the real CS80 aren't particularly square!). Narrower pulses fare no differently, and the differences continue to surface when you apply modulation to the pulse width.

All these tests were made by playing single notes, and the differences between the instruments become more obvious when you play chords, because CS80's tuning is less perfect than CS80V 's. Consequently, CS80V sounds a little thinner and less 'organic' than the original, although this is something that can be overcome by using CS80V's multi-mode section, of which more later.

The oscillators in versions 1.0 and 1.1 of CS80V also exhibited some bugs and flaws, but happily, these were cured in v1.2. However, strange residual sounds still appear when you switch the oscillator waveforms on and off. Plenty has been made of this on CS80V 's user forum, and even though we are now on the third major revision of CS80V and the noises are quieter than before, the problem persists.

The last thing to check was the noise generator. In versions 1.0 and 1.1, the noise on CS80V was too 'blue' (ie. too weighted toward high-frequency content) except at the upper limit of the keyboard range. This was because CS80V 's filters did not track the keyboard CV in the earlier versions, unlike on the original CS80. In version 1.2, the filters can now be made to track the keyboard, thanks to a new 'key-follow' source in the modulation matrix.

Filters & Amps

As on the CS80, each CS80V voice has two filters, a resonant high-pass and a resonant low-pass, and you can select their profiles to be either 12 or 24dB-per-octave, although not independently. The filters' cutoff frequencies are modified by an unusual IL/AL/ADR contour generator rather than the ADSRs found on most other synths of the era (aside from CS-series polysynths, this unconventional envelope was also used on the flagship Yamaha GX1). Controlled by five sliders, this emulates the conventional ADSR plus its associated 'Amount' control, and generates an equivalent four-stage contour.

The five controls are labeled 'IL' (Initial Level), 'AL' (Attack Level), 'A' (Attack Time), 'D' (Decay Time), and 'R' (Release Time). As you can see, the Sustain Level control of the standard ADSR is missing so, instead of adding the ADSR contour to the cutoff frequency defined elsewhere, Yamaha defined the sustain level to be the cutoff frequency, subtracting a voltage down to an Initial Level, and adding one for the Attack Level. The diagram above should make this clearer.

Clicking on the name below the HPF (high-pass filter) or LPF (low-pass filter) sliders activates or deactivates that element of the filter. If both are off, the sound passes straight through the filter section to be shaped by the appropriate amplifier. If both filters are on, they combine to produce a band-pass filter, assuming of course that the HPF slider is positioned below the LPF — otherwise all the frequencies are attenuated and you'll hear not a sausage.

The filters are further controlled by two tabs found in the main section of the panel. Labelled 'Brill' and 'Reso', these are global offsets for the filters' cutoff frequencies and resonances, and they apply to all four filters equally and simultaneously. You can make significant changes to a sound using these alone, but they are not part of the 'patch' as such, and on the original synth, their settings could not be saved. Fortunately, CS80V allows you to save every setting, so it's no longer necessary to mark the controls with chinagraph to reproduce sounds correctly!

Comparing the filter response to the original, we found that the maximum amount of frequency sweep on CS80V is much greater than on the original synth. This is good, because you can always get close to the CS80's response by tempering the contours appropriately. Likewise, at 20 seconds, the longest Attack and Decay times are much longer on the software synth, as is the maximum Release of 30 seconds.

The most complex 'view' of CS80V, with the virtual top panel open to reveal the software's modulation matrix, keyboard-zoning and voice-assignment controls.

The filters sound quite similar in 12dB-per-octave mode, although they're not indistinguishable. CS80V 's filters are more resonant than the originals, and the former ring if you sweep them quickly. You can also hear a slight bumpiness to CS80V 's filter responses when you sweep the controls quickly. Finally, the Release on either voice pops when its Release Time is set to zero, and you have to increase it a little for the sound to be useable. This was not the case on the CS80 itself, and the clicks are not being generated by super-fast envelopes, but strange artefacts that shouldn't be there. Of course, these can be eliminated by increasing the Attack and Release settings by a fraction, but that's not the point.

After the filter section, the signal in each voice passes to its amplifier section — exactly how much is determined by the VCF Level fader. You can use the two of these to adjust the relative levels of the two voices, but the CS80 (and, therefore, CS80V) offers another little trick: to the right of the VCF Level lies a fader that introduces an additional sine wave to the mix, post-filter. When active, it reinserts the fundamental frequency of the voice, adding 'oomph' to bass patches and high-pass filtered sounds. Unfortunately, on CS80V, there's a bug here; if you switch on the sine wave in Voice II, everything else in that voice disappears.

The signal then passes through the amp itself, which is shaped by a conventional ADSR contour generator, with the same maximum times as the filter's IL/AL/ADR contour generator. Finally, another Level slider determines the output level.

Alongside the amplifiers lie the Touch Response sections. These govern how the filter cutoff frequencies and amplifier gain of each voice react to Initial touch (velocity) and Aftertouch (key pressure). Unlike most synths, in which the velocity and aftertouch work subtractively — ie. the unaffected VCA level is also the loudest you'll get when playing or pressing hard — CS80V works the opposite way round. If you raise the Initial Level slider to introduce some gain controlled by velocity or aftertouch (or both), the unaffected VCA level is the quietest you'll get when playing softly. Playing harder can add massive amounts of gain to the signal, so you have to exercise a lot of caution using these controls: be sure to pull the VCA Level fader way down before you start to increase the Level in the Touch Response section. If you don't, and you use any degree of touch sensitivity, you'll find that it's far too easy to create harsh, aggressive digital distortion in the signal path. You need to reduce the amp levels considerably to eliminate this, and it can take just a single note played harder than you anticipated to reintroduce the distortion.

Likewise, the Brilliance (filter) response operates additively, although this lacks the dangers of the amplifier gain response. To be fair to Arturia, the original CS80 works in the same way, but the maximum amount of additional gain is much less than applied within CS80V. It's worth noting here that while the touch-response faders add positive 'voltages' to the filter cutoff frequency and final output gain, it's possible to set up a negative response using the modulation matrix.

Ring Modulator, Sub-oscillator & Arpeggiator

The ring modulator was one of the CS80's most distinguishing features, and Arturia have done a good job of replicating it on CS80V. As on the original, there are faders for speed and modulation amount, and the speed may be varied (if you wish) by a dedicated Attack/Decay envelope. Given the vintage of the original synth, it's not surprising that there was only one modulator and one envelope for the whole instrument, and to Arturia's credit, they have left this unaltered. In contrast, the maximum modulation depth, envelope depth and modulation speed are all greater than on the CS80, so you can go far beyond what was previously possible if you wish. Unfortunately, there were bugs in the ring modulator, and at times it failed to work at all on the Mac version of CS80V. But when it did, it was good, and with care, it could be set up to sound similar to the original.

The Sub-oscillator is a single dedicated LFO that can be applied to the overall patch, and can affect the VCOs' pitch, the filters' cutoffs and the amplifiers' gains. Its controls should be used in conjunction with the Touch Response levers that allow you to control the Sub-Oscillator's speed, pitch modulation depth and filter modulation depth, all with independent amounts as desired (if you want to imitate Vangelis, this is the place to do it). You'll find these levers next to the Initial Pitch Bend control, which adds a quick, velocity-sensitive, upward pitch swoop to the start of each note.

CS80V 's arpeggiator is basic by modern standards, but nevertheless welcome. It offers up, down, up/down, random and note order options, over a range of one to four octaves, with between one and four repeats of each octave in the pattern. This can be sync'ed to a host sequencer's tempo or left to run freely, and can be latched 'on' using the Hold button. As is the current norm, the tempo knob becomes a tempo multiplier when the arpeggiator is running in Tempo Sync mode.

Added Value

Whilst the virtual 'top flap' of CS80V is closed, it behaves like a single, eight-voice synth. When opened, CS80V goes far beyond the spec of the original CS80, becoming an eight-part multitimbral monster. The modulation matrix, which is only operative when the lid is open, is also accessed here. In version 1.2, there are 10 modulation slots, each allowing positive and negative modulation. There's a total of 13 modulation sources (LFOs 1 & 2, the sub-oscillator, both filter and amp envelopes, velocity, aftertouch, the ribbon controller, mod wheel, expression pedal, and keyboard follow), and no less than 38 destinations, including the frequency and pulse-width of both oscillators, both noise sources, filter cutoff and resonance, amp volume, the speed and volume of the LFOs, and the individual stages of the amp and filter envelopes. This is commendable, and only die-hard purists will be concerned at the potential for straying so far from the capabilities of the original.

CS80V's more compact keyboard-only view gives you access to a more limited crop of patch-selection and editing controls, and disallows access to many of the main synth parameters, but saves considerably on monitor space!

Behind the mod matrix on the 'open' version of CS80V 's front panel lie the controls that allow you to assign any combination of up to eight different patches to separate zones, each with its own key range, MIDI channel and voice assignments. There are seven types of voice assignment, including 'Rotate' (in which the voices cycle round in numerical order), 'CSAssign' (random voice assignment) and Unison High, Low or Last (lowest-note, highest-note and last-note priority respectively.) Of course, you still only have eight notes of polyphony to play with, but if you set all active zones to Rotate, there is a fair chance that voices will be assigned to where they're needed, so long as you're sensible about what you expect CS80V to play.

Each voice can be adjusted for transpose, tuning, volume, pan, effect on/off, Ring Mod on/off, and portamento on/off. Given these options, you can set up a multi-patch to play an eight-note cyclic sequence in which every note plays a different sound, triggered by repeatedly playing one note on the keyboard. Alternatively, this is where you can create differences in each voice to emulate the inconsistencies always present on the real CS80's eight voice cards. Detune each voice slightly, shift the volume just a tad and suddenly CS80V sounds more authentic.

Arturia's web site claims that multiple outputs are now supported, but we could find no way to put this into operation on the PC test system, and upon inserting an instance of CS80V into Sonar v4, we were still presented with only the stereo Primary Output. Matters were much better on the Mac system, and Plogue Bidule 0.7001 immediately recognised all 10 outputs — eight for the individual parts, and two that formed a stereo master pair.

Like Arturia's other software synths, CS80V has built-in effects. However, this is one of the weakest parts of the product. For some reason, they have never worked correctly, from the pre-release beta version to version 1.2. On its own, the tremolo simply doesn't work. Secondly, there's some bizarre signal routing going on here — turn on both chorus and delay, and only the delayed signal is chorused. On the PC version of CS80V, this is all but inaudible unless you have the delay mix level full up. If you now engage the Tremolo switch, the dry signal vanishes altogether, leaving only a delayed and heavily chorused signal, which can only be heard when the delay mix knob is turned up. We can't understand why, after two revisions, these effects still don't work properly — particularly when dozens of affordable software synths offer simple effects like these.

Alongside the effects, which are accessed via the toolbar at the top of CS80V 's window, lie the recreations of the CS80's sustain, glide, and pedal controllers. Fortunately, these do work correctly.

In Use

Having compared the component parts of CS80 and CS80V, what's the sound like overall, and what's it like to play? It's not sufficient to compare presets; checking the software settings against the original Yamaha patch charts, which we were able to refer to, it's clear that the patch parameters are very different. Nevertheless, the software's presets do bear some sonic resemblence to those of the hardware, suggesting perhaps that Arturia's engineers recreated the presets by ear.

In order to carry out a better comparison, we therefore proceeded to program and use a wide range of sounds that were typical of the CS80 in the '70s and '80s. Of these, the most definitive and recognisable is the slow, swelling solo brass patch so beloved of Vangelis on soundtracks such as Chariots Of Fire and, in particular, Blade Runner. If CS80V could capture this, it would bode well.

Remarkably, given the differences we've noted, CS80V is capable of recreating this sound. We set up the CS80 correctly, playing it through a Quadraverb to imitate the software synth's delay, and then set the controls identically on CS80V. The calibrations of the controls were different, but with perseverance, we were able to make the two almost identical, swapping between them without any perceptible change in the sound. We were less successful at recreating the beautiful strings that were also a strong point on the original CS80; there seemed to be no way to get CS80V to recreate simultaneously both the cut and warmth of the original. On the other hand, setting the PWM mode to 'Free' and patching a traditional saw+PWM ensemble patch resulted in a new range of textures that you could never have coaxed from an original CS80. The same was true when using the 24dB-per-octave filter settings... new sounds poured forth, although you could no longer describe them as particularly CS80-like. Overall, we'd have to say that the sound of CS80V is a winner.

Unfortunately, when using CS80V to playing real music rather than being analysed in an academic fashion, more faults revealed themselves. For example, when you assign filter keyboard tracking in the modulation matrix on the Mac version of CS80V, each synth voice plays with the filter settings applicable to the previous note. This means that if one of the voices is allocated to a low note followed by a high note, the resulting sound is far too dull. Conversely, if a high note is followed by a markedly lower one, the sound is excessively bright. And the more we played, the more the signal path distortion/overload problems mentioned earlier began to vex us. Aside from the audio overloads, the program proved to be quite CPU-hungry, and despite both test machines being well above the suggested minimum specification, it proved easy to create the tell-tale audible signs of the software running out of processor power.

We also spotted a problem with the emulation of the CS80's huge ribbon controller, which takes the place of the pitch-bend wheel on most other instruments. In version 1.2 of CS80V, Arturia claim to have corrected the bugs in the response of the ribbon controller, but their implementation still does not behave the way the ribbon does on the CS80. When you release your finger from the CS80's ribbon, the pitch of the sound reverts immediately to the unaffected pitch. This does not happen in CS80V; the pitch glides back. This makes it impossible to play trills on CS80V 's ribbon, which was a rather neat performance feature used by many CS80 players.

Finally, we had some concerns about the stability of the stand-alone Mac version of CS80V. While using this, we occasionally needed to change the I/O settings in the configuration window. This would work the first time, but a second invocation of the window would cause the program to crash. Saving and loading patches in the stand-alone Mac version also caused numerous crashes, sometimes leaving the CS80V engine running in the background, but with no user interface or other means to silence it other than OS X's 'Force Quit' command. Of course, Arturia can't test every possible configuration on which their software will be run, but if we had problems with our thoroughly ordinary test G4, so can other users.

Conclusions

CS80V has the potential to be great, but leaves us in a bit of a quandary. The sound of the instrument can be remarkable, emulating the CS80 with an accuracy that can be quite uncanny, and this makes us keen to use it, but the problems we experienced discouraged us — especially from attempting to use it live. This dual-platform test has also had the benefit of allowing us to draw up comparisons concerning the reliability and idiosyncrasies of both the Mac and PC versions. As our findings have revealed, the stand-alone Mac version appears to have some stability problems and several 'bonus' bugs not shared by the PC version, which prospective Mac-oriented users should take into account. Having said all that, the most recent version 1.2 has corrected many faults, and is a must-have upgrade for any user of the earlier versions.

If Arturia can sort out the problems, CS80V could get an energetic 'thumbs-up'. But until then, you'll have to decide whether you can live with its problems, or whether you're going to wait until they're fixed.