It's strange, but in a bizarre parallel to the 1970s and early 1980s, the world has been rather resistant to the notion that high-quality synths extend much beyond the revered American models. Other than the occasional oddity such as G-Media's ImpOSCar, Arturia's CS80V and Korg's recreations of their own vintage synths, there have been few mainstream attempts to step beyond the obvious Moogs, ARPs and Prophets. Would you be interested in a 'soft' EMS VCS3, Minikorg 700S or Roland SH5? I know I would. But if there's one synth that should have been modelled long before now, it's the one I have in front of me today. It's Arturia's recreation of the mighty (and beautiful) JP8 'Jupiter 8': Jupiter 8V.
For the uninitiated, the Jupiter 8 was an eight-voice polysynth with two VCOs per voice, a 24dB/octave low-pass filter, a pair of traditional ADSR envelope generators, a sophisticated arpeggiator and a wide range of modulation options. When it was released, in 1981, the JP8 represented a step away from the 'American sound' established by the likes of Sequential Circuits and Oberheim, and its distinctive character has led to it becoming a classic and desirable Instrument. You can read more about the Jupiter 8 in the February 1998 issue of SOS.
Installing 8V is not quite as simple as Arturia's first products, because it requires the company's Syncrosoft dongle and on-line authorisation. I have heard people complain about this, but I think it's an excellent system, because you can carry the dongle around and legitimately use copies of the software on different machines. It should also mean that Arturia reap the rewards of their labours, rather than suffering from piracy. Once installed, you can invoke 8V as a plug-in or a stand-alone instrument. Arturia's programmers are good at graphics and they have really got it right this time. The Jupe was a sexy synth, and 8V's front-end has captured this superbly. Everything falls easily to hand with no nasty surprises, and even the additions (which are subtly blended into the original look and feel) are integrated sympathetically.
Comparing my (hard) Jupiter 8 and the (soft) Jupiter 8V, I was pleased to find that their oscillator sections are all but identical, with merely cosmetic differences to distinguish them. Using VCO1 as a test source on both synths, I found the sawtooth waves very close, as are the square waves. The triangle waves are similar but sound more full on the original, while the rectangular (pulse) waves are, for a given setting, a little fatter on the soft synth. Even with added PWM, the '8' and 8V can produce sounds that are all but indistinguishable for normal uses.
Much more significant are the differences that you hear when you introduce cross-modulation. Given that it's almost impossible to find two big, analogue polysynths that offer the same results when cross-mod is introduced, it's no surprise that my Jupiter 8 and 8V sound different. Remarkably, and without being pressured, Arturia have confessed to me that the cross-mod on 8V is not quite right, and have undertaken to improve it in a future revision. I am very impressed by that. However, I really dislike how Arturia have implemented their so-called 'free running oscillators'. Every time you hit a note, their phases are initialised randomly. This is irrelevant when using one oscillator, but results in different tones when you use both oscillators in unison, or when sync is on. Indeed, if sawtooth waves are initialised 180 degrees out of phase, the fundamental is eliminated and the note sounds an octave higher than it should. If two sine waves are selected, the note can disappear entirely! Furthermore, once the relative phases are calculated, they don't drift, so you obtain static sounds with different tones as you press successive notes: horrid. So, while I understand the engineering issues (truly free-running oscillators would increase DSP load), I hope Arturia find a solution soon, because it also leads to a second problem...
When you mix the Jupiter 8's oscillators, the loudness increases a little compared with a single-oscillator version of the same sound. For Jupiter 8 sawtooth waves in unison this increase is around 2.5dB, but when you mix 8V's oscillators, the increase can be as much as 6dB, or can even be a decrease if the oscillators are out of phase! If you're looking for a reason why the soft synth does not precisely emulate the original's character, look no further.
The layouts and functions of the HPF, VCF, VCA and ENV sections again appear to be all-but-identical on the original Jupiter 8 and on 8V. Since the JP8's (and, therefore, 8V's) filters do not self-oscillate in either 12dB/octave or 24dB/octave modes, I passed white noise through them and averaged the spectrum to obtain profiles. I found that the highest cut-off frequency of the 8V filter is around 15.5kHz, rising to a maximum of a little over 18kHz when positive Env or Key Follow is applied. In contrast, my Jupiter 8 has a maximum cutoff frequency of around 7kHz with no CVs applied, but disappears way into the supersonic as you increase the modulation. Surprised by this result, I tested the Jupiter 8 that Arturia had used to model the filter response and found that 8V matched their synth very closely. Maybe I need to look at the calibration of my Jupe!
Likewise, I found that the envelopes on a real Jupe and on 8V are somewhat dissimilar. My Jupiter 8 offers a maximum attack of eight seconds, but a more generous decay of 50 seconds or so, and a release in the region of 45 seconds. Arturia's Jupiter 8 is similar, with a maximum attack of six seconds, and a maximum decay and release exceeding a minute. The soft synth offers the maximum attack of six seconds, and around 35 seconds for decay and release. The latter figure was not an oversight by Arturia's engineers; they found that the controls could be implemented with less stepping in the useful range if they limited the maximum. Very sensible, in my opinion.
Having analysed the bits and pieces, I decided to start programming some real sounds. Things started very well. I created a simple sound, that I called 'Plink', on my Jupiter 8. This comprised a single sawtooth wave passed through the VCF with maximum envelope follow and a simple AD contour with instant Attack and a short Decay. I found that I could recreate this sound precisely on 8V if I reduced the amount of envelope follow on the original synth to about 85 percent of its maximum. Having done so, I sequenced the two instruments, alternating notes from one to the other, and it was impossible to tell that two different synths were contributing to the sound.
When I began to experiment with more complex patches (for example, by introducing resonant filter sweeps) things started to look a little less rosy. I returned to 'Plink', increased the resonance to maximum, and then slowed the attack so that the JP8 went 'Woooo-ooow' very smoothly. I made the same modifications on 8V, and it went 'Woooo-ding!-ooow' with a noticeable discontinuity as it transited from the attack to the decay stage. I tried an ASR 'trapezoid' shape and found the same thing; at the transit between the attack and the sustain, there was a bump. I switched from ENV1 to ENV2 and obtained the same result. Depending upon the patch, this can be a subtle problem or a serious one, although it can be masked if you use copious effects. Nonetheless, sounds such as string ensembles, brass and pads require smooth transitions between the attack and decay stages or attack and sustain stages, so I'm relieved that Arturia have already undertaken to fix the problem.
There are four effects slots provided within 8V: two that are applied on a voice-by-voice basis, and two that affect the patch as a whole.
The four stereo, patch-based effects — chorus/flanger, delay, reverb and phaser — are rather basic, and they lack the modulation capabilities of the per-voice effects. Furthermore, Arturia's programmers have admitted to me that they have been kept simple (in signal-processing terms) to minimise CPU drain. The delay is useful but, to be honest, the others don't impress me; the chorus has none of the lushness that I crave, at flange settings I couldn't obtain the classic 'jet' effects, the phaser didn't 'whoosh' as I wanted... and so on. I think serious users will bypass these and play 8V through their favourite plug-in effects.
The first (minor) functional differences appear when we look at the arpeggiator, playing modes and key assignment controls. The volume, Upper/Lower Balance knob (which controls the relative loudness of the Upper and Lower patches in both split and layered modes) and the master tune controls are as before. Understandably, however, the hardware 'Tune' button of the original synth has been discarded, to be replaced by a Detune knob (which fattens unison sounds and adds so-called 'analogue instability' to non-unison patches) plus an LFO MIDI Sync button. Other than this, all is as it was. Likewise, the arpeggiator (which was one of the most important features on the Jupiter 8) is duplicated lovingly, with the Up, Down, Up/Down and Random modes working correctly at all ranges.
Moving on, the performance panel looks and acts like an exact recreation of the original, although the effect ranges are different. For example, pitch-bend on the Jupe had a maximum range of approximately plus or minus 16 semitones; on 8V it's precisely plus or minus 24 semitones. Likewise, LFO modulation on my Jupe has a maximum depth (from the performance panel) of plus 10 semitones and minus nine semitones, whereas 8V offers precisely plus or minus 12 semitones... and so on. Personally, I'm glad that Arturia's programmers have made these changes; they are welcome corrections of analogue limitations.
The statements posted on a well-known synth web site that, "quite honestly this thing sounds awful... don't believe the claims it sounds the same or similar to a real JP8 because it seriously does not come even a little bit close" are palpable nonsense. Given that I'm sitting here with a Jupiter 8 under my left hand and my Mac running 8V in front of me, I'm very confident of my ground here: when used to create simple Jupiter 8 patches, 8V can sound indistinguishable from the original synth. For typical (ie. slightly more complex) Jupiter-esque pads and washes where real-time tweaking is not an issue, 8V comes very close, and for all music-making purposes can be considered interchangeable with the original. For more extreme patches with cross modulation, high levels of modulation, high resonance and so on, it does not sound the same as the Jupiter 8; it's in the same sonic territory, but it's not identical.
Inevitably for a version one soft synth, there are bugs. In addition to those I've already mentioned, 8V would often decline to launch the first time I tried to invoke it, although it was perfectly happy the second time. I also encountered a wonderful new bug: on one occasion it refused to play MIDI Note C3, although everything else worked correctly. How analogue is that? I also found an obvious bug in the effects: select the phaser and the default parameters create an uncomfortable discontinuity in the sound where none should be. Change the sweep waveform and then change it back again, and the problem disappears.
But, ignoring these, all of which will probably be addressed in the next revision or two, what's the verdict?
For some people, it will be hard to use Jupiter 8V because it seems to be very sensitive to inefficient audio drivers and/or MIDI drivers, and if your host computer lies near (or below) the minimum recommended specification, it will run out of CPU if you make extensive use of 8V's facilities. There's also a problem with the graphics: there were innumerable times during this review when I had to wait for the 8V to display key presses from up to half a minute earlier before I could ask it to do anything else, such as modify a knob or press a switch.
At first, I was inclined to blame my three-year-old Mac for this, but before finalising this article I replaced my Korg MS20iC USB controller and Tascam US428 audio interface with an Evolution USB keyboard, and used the Mac's internal audio drivers, and matters improved considerably. Playing 8V in ways that previously slowed it to a crawl resulted in CPU usage of 80 percent, or thereabouts, but no problems.
I then spoke to SOS contributor Nick Magnus, who had downloaded the PC version for use on his 2.8GHz P4 with 2GB of RAM. He found that the stand-alone version worked with limited polyphony, but within Sonar v6.2 the CPU almost immediately reached maximum, causing the PC to grind to a halt and Sonar to crash. Another colleague, whose host is Nuendo v3.2 running on a dual 3.2GHz Xeon PC with 4GB RAM, told me his CPU usage was high but he was able to use the 8V if he didn't load the system too heavily with other VST Instruments or effects. Clearly, 8V is sensitive to the host environment. To be fair, the sound quality of soft synths is related to the amount of power you throw at them, so it's not unreasonable for the good ones to be quite hungry, but 8V seems particularly voracious, so I recommend that you download the trial version and test it before purchasing.
The Jupiter 8 was a synth that rewarded subtlety and understanding. Jupiter 8V also suffers (or benefits) from this. The factory patches are fairly restrained, and you are not going to be crying out 'oh, wow!' within moments of laying hands on it. But does the world need another big, blobby, American-sounding soft synth? Given the glut of Prophets, and that all the Minimoog, Odyssey and ARP2600 emulations are fully polyphonic, something with the more transparent and polished character of the Jupiter 8 is to be welcomed. Look at it this way: 8V sounds much like Jupiter 8, but does a zillion things that the original could not. Given that you can assign the knobs and sliders on your MIDI controllers to its screen representations, it comes even closer to the ideal. So, ignoring other issues for a moment, do the sonic differences between the Jupiter 8 and 8V matter? This is a difficult question to answer. On the one hand, the accuracy of the GUI almost demands that the imitation is perfect. On the other, the 8V is a really nice soft synth with a great sound, so you may feel that differences are only of consequence to the most retentive of analogue anoraks. I'll leave you to decide where you fall in the spectrum.
To finish, I have to say that I continue to be impressed by Arturia's willingness to listen to constructive criticism and undertake extra work to make their products first-class. I raised half a dozen or more points with them while completing this review and on every occasion they checked my findings, and then undertook to fix things. Consequently, I think that Jupiter 8V will follow the same path as the original Minimoog V. Version one is good, but version 1.5 should be excellent.
The Software Advantage: 8V's Extras
Inevitably, the architecture of 8V extends way beyond that of the original Jupiter 8 synthesizer. Obvious extensions are in the number of voices (32 of them), MIDI synchronisation, and a much more extensive patch management section. Arturia have also enhanced the sound generator itself, and you can access three new sets of controls by clicking on the Extensions icon at the far right of the toolbar. These are the Galaxy modulation generator, a step sequencer, and the dual effects sections (the last of which are explored in the separate box on the effects).
Inevitably, the majority of effects you obtain from this are... well, just effects. But with a bit of planning you can use Galaxy to program sounds that would be very hard to obtain by other means. If all this sounds a bit daunting, it isn't; eschewing unnecessary graphical affectation, Arturia's developers have come up with a display that makes it very clear what is happening. If you meet the guy who dreamed this up, buy him a decent bottle of red (or a pint, if he prefers).
The next major addition is the 32-step sequencer, which shares the grey-scale design of Galaxy. Numerous modern-day controls are provided, such as a pencil tool, an eraser, a random sequencer generator and so on, but after just a few minutes play I discovered that it is nonetheless very traditional in philosophy and implementation. Simply draw the amount of deviation at each step (which can be quantised or not, as you wish), and direct the resulting 'virtual CV' to up to three destinations simultaneously.
You can link steps, slew them, and place accents on them, determine the direction of the sequence, determine the quantisation, sync it to MIDI clock, apply swing, and all the other things you would expect from a faux-analogue sequencer. Unfortunately, you can't invert the polarity of the output for any given destination — say, to close the filter while increasing the pitch of a note, or vice-versa — but I have discussed this with Arturia, and they are looking into it as a possible upgrade. (No promises, though.) Neither can you create multiple lines: the sequencer is not polyphonic to the extent that you can determine the timing and pitch of individual notes, only to the extent that all notes played are affected simultaneously.
Interestingly, you can run the arpeggiator and the sequencer simultaneously, with the former resetting the latter each time it completes a cycle. This can create complex rhythms and modulations, and I suspect that further experimentation would yield interesting results.