Whilst pondering the seemingly uncomplicated idea of recreating analogue style bass sounds on a digital, sample-based synth, it gradually became clear that this area of sound is less straightforward than might be supposed. As I set about cloning some of my favourite timbres, it became apparent that a whole different set of considerations comes into play than in the manufacturing of brass, strings or other such generic tones. Because analogue bass sounds, on the whole, tend to be synthetic in nature (that is to say, not necessarily imitative of a real-life instrument), their very characteristics are defined by (and, indeed, often dependent on) the idiosyncrasies of whatever analogue synth we have as our model. To put this another way, the particular features and design quirks of a synth can be fundamental to the way it sounds when we're dealing with totally synthetic timbres.
Take some of the following into consideration: does our "model" synth have VCOs (voltage controlled oscillators) or DCOs (digital controlled oscillators)? If it has the former, might we reasonably expect any subtle detuning between the oscillators to be slightly inconsistent across the width of the keyboard? DCOs, of course, have their tuning constantly checked by the synth's processor chip so tuning is predictable. What about the filter; would it offer low, band or high pass? Or even 12 or 24dB per octave? Both of these will dictate the sort of colouration we can apply to our waveforms. There are many other factors, such as whether or not a sub-oscillator is present, or if Unison (stacked) mode is detunable, or how many oscillators are available per note. We will tackle some of these in the context of our digital synth in due course.
Perhaps the single most important factor separating the sound of one analogue synth from another is envelope generators, and the markedly different ways in which they behave across the vast range of instruments. Brand name alone carries no guarantee of performance; one manufacturer's entire history of analogue synths could have totally different ADSR characteristics across the range. Some classic examples can be cited: the Roland SH101 has what many would argue is one of the fastest ADSR attack times. This accounts for its value when punchy bass sounds with loads of 'spit' are what get you drooling, and, together with its O-So-Lo sub-octave, this feature has helped it to become the darling of the dance floor. On the other hand, the ADSR speeds of Roland's JX8/10 synths are comatose by comparison. Few people would opt to do bass duties on one of these if transient response was of crucial importance. Similarly, the Minimoog has always been noted for the fine tailoring possible with its extremely responsive ADSRs. Compare this to the Moog Source or Memorymoog, both of which had parameter access editing. This division of parameters into 100 steps already begins to compromise the ability to fine tune such sensitive settings as an attack or decay time.
Even more factors (word of the month) come into play, such as the linearity of the envelope curve. The ARP Odyssey had a notably pronounced exponential curve, which contributed to its aggressive nature, compared to the rather more linear envelope characteristic of the Minimoog, which was considered by some to be more 'musical'.
The reason for going over this background becomes clear when you realise that, in all likelihood, our digital synth has in its software just one ADSR algorithm with its own qualities and limitations. With this ADSR we are hoping to emulate those of a Juno 60, Prophet, SH101, Jupiter, Prodigy, Pro One, whatever it may be. In reality, it cannot be all these things. The limitations imposed by parameter access editing may mean that the setting we want is exactly between two values and is therefore inaccessible. While a decay time value of 1 seems just a gnat's too slow, a value of 0 does nothing at all. This means that a certain level of compromise must be accepted whilst attempting to convey an impression of the target sound. Frustrating? Maybe not; what the following aims to do is to point to one or two crafty ways round the problems to get us through the turnstile and into the right ballpark. Time now to look at some examples of typical sounds and to offer some suggestions along the right road.
The aforementioned SH101 presents us with the problem of ADSR speed. While this instrument is tonally simple to emulate, how can we provide the requisite kick at the front? Looking at the basic sound, we find that detuning is unnecessary, as the 101 is a single oscillator synth; therefore, all tones used on our digital synth should be exactly tuned to each other, and any LFO assignments should be copied and the key sync turned on so that all tones behave as one. Saw and pulse waveform levels can be balanced freely as on the original synth, but a modulating pulse wave is (currently) almost certain to be unavailable. This is not too much of a tragedy with bass sounds, as PWM tends to blur the tuning of low frequencies. The sub octave is pretty straightforward; on the 101 it is a square wave with a choice of three octave ranges. So the basic sound could consist of up to three tones: the saw wave, the pulse/square wave and the sub octave. The SH101 has only one ADSR (plus a gate setting) which acts on the filter and amplifier simultaneously, so our filter and amp ADSR values across all tones should initially be set identically. This rule can, of course, be broken if it gets you nearer to the sound you're after.
You should be able to get fairly close following these guidelines, but the front end will still not 'spit' like the original SH101. The "cheat" is this: using the fourth tone (if present), select a short wave such as an organ key click, hi-hat, sidestick or similar. Now tune it upwards until it is very short but still audible (being a sample, tuning will naturally change its playback speed). Then raise the filter resonance and adjust the cutoff frequency until the right flavour is obtained. Careful balancing with the other tones should provide the required 'spit' at the front of the sound. In addition, try turning off the keyboard pitch follow for this tone, or greatly reduce its range, so that it changes very little or not at all across the keyboard. Lastly, setting the key mode to monophonic legato (so the envelope retriggers only when all keys are released) should complete the picture.
The classic Moog Taurus pedal sound is not difficult to achieve, but the saw waves alone provided on a digital synth may not have the floor-shaking bottom end associated with this doyen of progressive rock. The following suggestion should add more power to your elbow: set the tuning of the two saw waves to approximately +7 and -7 cents respectively, then set up a third tone as a copy of one of the first two. Select a sine waveform for this tone and set its fine tuning to 0 and its mix level accordingly. As it's a sine wave, you will find the filter cutoff frequency is by and large irrelevant, as there are no high harmonics to be filtered out. But don't set it too low, as the tone will disappear altogether!
One phenomenon experienced with synths using VCOs, as opposed to DCOs, is the tendency for detuned bass sounds to 'drift' in and out of the mix; one moment the bass is bulldozing the track into oblivion, the next it's all but sunk without trace -- especially the stacked oscillator types of sound. This is due to the oscillators slowly wandering in and out of phase and subsequently cancelling each other out. A Juno 106 (which has DCOs) in unison mode is more likely to remain at a constant level than an OB8a or Jupiter (which have VCOs). Digital synths seem largely immune to this problem, due to their total tuning stability and the inherent repeatability of events whenever a note is sounded. As a result, the sound can seem more sharply focused than the analogue equivalent. Have you ever noticed how a good sample of an analogue bass sound often sits in the track with a noticeable clarity lacking in the original? It's much the same philosophy at work, and can certainly be seen as an advantage of pursuing this method of bass synthesis, despite the compromises.
A set of 64 classic retro sounds for the Roland JV series synths, including some sounds mentioned in these articles, is available free of charge from Roland dealers, or contact Roland UK, 0252 816181.
Unison or stacked sounds can vary from synth to synth. The Prophet 5 stacks oscillators quite literally, with no mass detuning, in contrast to the Super Jupiter, which offers a 'fanning out' detune parameter. If we simply stack tones on a digital synth with no detuning, it merely sounds like one very loud tone, so some detuning is needed to give a really fat quality. As mentioned in previous articles, use different waves, where possible, to reduce the possibility of phasing problems. To reproduce the Jupiter's unison detune effect, wide detuning between four tones may be necessary but true enormity might require some kind of chorusing to be added. Strictly speaking, this is a step away from authenticity; it also means that, unless only the mono output is used, our sound will be in stereo. This could well be fine if cutting a record to CD, but should be taken into consideration if pressing vinyl discs, as cutting lathes usually go bananas if there are phase differences between the left and right channels at low frequencies!