Nick Magnus continues his quest to emulate analogue sounds on digital instruments. This month, he shows you how to recreate analogue strings.
The most important area to consider when programming string tones is that of movement within the waveform. Unlike brass or piano tones, whose character depends greatly upon the filter and amplifiers' envelope shapes, the waveform content of a string sound is crucial to one's perception of the end result. This movement is traditionally provided on analogue instruments in one of several ways:
1. The first method applies to both single and dual oscillator synths, and involves the use of a pulse width modulated (PWM) waveform. This can be seen as a square wave whose shape is modulated slowly by a low frequency oscillator (LFO). This modulation sounds not unlike chorus, but sometimes has the effect of making the tuning of lower notes appear rather nebulous. Where two oscillators are available, the second also uses PWM, and is slightly detuned with respect to the other.
2. A second method is to use two oscillators, both producing a sawtooth wave. Again, these would be slightly detuned, but the desired movement would be provided in a different way. Assuming one of the oscillators could have pitch modulation (vibrato) applied to it independently using an LFO, the LFO is set to modulate somewhere around the 'natural' speed of a violinist (something like eight cycles per second). The oscillator mix is set at equal levels and the depth of the LFO adjusted to produce a smooth ensemble effect.
3. A third method which could apply to single‑oscillator synths again uses a PWM or a sawtooth wave which is gently modulated as in example 2, but the ensemble quality is augmented by an internal chorus unit. This is typical of the Juno 60 or Korg Poly 6.
Now to apply one of these methods to our digital synth. It will be assumed that, alongside the sampled 'real' sounds, there is at least a basic complement of plain waveforms on board the synth, including sawtooth, square and sine waves. Due to the long cycle time of a PWM waveform (thus making it very hungry in terms of sample memory), we will also assume there is no PWM wave available. So the first port of call is the classic Jupiter 8/Prophet/Oberheim string sound using sawtooth waves. Incidentally, if you are in a position to have the genuine analogue instrument in front of you, it will serve as an excellent reference point to compare filter settings, LFO speeds and the like. If not, don't switch your gear off! Memory and/or imagination will serve you perfectly well.
It will help to set up your Digital Sample Synth (referred to as DSS from now on) with an initialised tone; that is, only one voice or tone playing, filter wide open or bypassed, and a gate (or organ) envelope. Also de‑assign all modulation functions like mod wheel, aftertouch, and velocity sensitivity. Naturally, turn off all the internal effects and reverb. Make sure the other muted tones are identically set up, then save this 'vanilla' patch, as it will be a very often‑used starting point.
The first sound creation task will be to select a waveform. Some DSSs have more than one of each type. The Roland JV synths, for example, have three each of saw and square waves, which on the surface sound identical. I was initially baffled as to why this should be. If you couldn't hear the difference, what was the point? The answer came when I played two tones, both using the Saw 1 waveform. Upon detuning them, they produced slight phase cancellation, resulting in a thinning of the sound rather than the smooth fatness of their analogue counterparts. This was remedied, however, by replacing one Saw wave with either of the other two. Clearly, the three Saw waves were samples of different oscillators, just as an analogue synth has physically separate oscillators. The phase cancellation was a result of two identical samples playing at the same time. Armed with this knowledge, the creation of fat, detuned sounds becomes a reality.
It is common for both oscillators of an analogue synth to share the same filter and envelopes. (Yamaha's CS80 is one exception where the oscillators are virtually independent synthesizers). For this reason it is extremely useful if your DSS has the ability to 'globally' edit tones, that is, any parameter changes affect all the assigned tones simultaneously, otherwise there is a fair amount of tedious copying to do! Failing that, in most cases, there is usually a Tone Copy function, enabling you to copy the settings of one tone onto another. Remember that this also duplicates the tuning and waveforms of the tones, so you will need to readjust these.
Presuming that we now have two tones at equal level, different Saw waves on each if possible, suitably detuned, we can set about adding some modulation to supply the necessary movement. Select only one of the waves for editing, and go to the LFO page. Choose a sine or triangle wave for the LFO if a choice is presented, then add some pitch modulation depth until its effect is sufficiently audible, and adjust the rate of modulation to approximate the average speed at which a violinist might play. Now set the depth of modulation until you obtain what seems to you to be a smooth ensemble effect. If you have the actual analogue sound for comparison, try muting the non‑modulating tone of both synths and check how the modulations compare. A further development of this is to modulate both oscillators identically, except that the LFO cycle of one tone is the inverse of the other (the Roland Super Jupiter is capable of this). This is usually achieved on a DSS by making one tone's LFO depth into a negative value. This can give a particularly smooth result, but watch the depth setting!
If everything proves satisfactory, you will probably be ready to move on to the filter settings, as your simulation will most likely be considerably brighter than the sound you are aiming for. Globally editing both tones, activate the filter (if bypassed) and choose a low pass filter (LPF). Gradually reduce the cutoff frequency until you hear the tonal quality you're after, or until it compares favourably with your analogue model.
Finally, the amplifier envelope is employed to give some shape to the sound. There are of course many possibilities here, depending on whether you want a fast, nippy response or a rather more mellifluous, pad‑like quality. Pad sounds are likely to need a rounded front with a soft release time, while faster attack and release times will apply for more articulate string lines. Allowing the envelope to decay slightly after the initial attack adds emphasis, as in an Agitato playing style.
For Arco or bowed effects, an envelope applied to the filter may be desirable, making the initial attack of the note brighter than the sustain portion, but caution over the amount of filter modulation is necessary if your analogue symphony orchestra is not to become too brass‑like. You should by now be somewhere in the right ballpark. Whether or not velocity sensitivity is employed is entirely up to you, but the methods detailed above should certainly produce a passable analogue string sound when placed thoughtfully within a mix.
Let's turn our attention to another analogue string classic, the Solina String Ensemble. This has long been hailed as one of the seminal string machines of the '70s and even beyond. Much of what we've learned from the previous example can be applied to recreating this sound; however, it is a little more complex in structure. Beginning with an analysis of the workings of the actual machine, we find that the No Entry sign across the chorus effect on your synth can be removed — the Solina used chorus within its design. What concerns us first is what happens before the chorus.
A crude description of the Solina's signal path would be thus: the basic waveform (a sawtooth) is split and fed into three separate pitch modulator circuits (like a chorus but minus the dry signal). Each of these modulators applies vibrato at a different speed, and the three signals are then recombined with the dry signal. Already you should see a similarity with our first string sound. The main difference is that all the 'tones' are modulating but at unrelated rates. This in itself sounds fairly full, but to round the whole thing off, the signal is fed through our friend the chorus, which blurs the pitch sum and differences even further.
To recreate this, it's necessary to use at least three tones on the digital synth. You could use four, but the polyphony on your synth may suffer, particularly when used multitimbrally. If you don't have three different sawtooth samples at your disposal, don't worry — due to the rather more intense modulation of this sound, and the fact that the Solina has a thinnish nature in the first place, using the same wave doesn't seem to matter. As far as detuning is concerned, leave one tone alone, and tune the other two plus three and minus three cents respectively, or whatever sounds best on your machine. Now to set the modulation parameters. This requires the modulation of two of the three tones, leaving the third one straight. Experimentation revealed this to 'tie' the sound together rather nicely. Turn to the LFO page and set the pitch modulation rate of each tone individually. If there is an LFO sync switch, turn it off, as we don't want the LFOs' cycles to reset each time a note is played. The modulation rates should average fairly fast: as a rough guide, the JV880 rates I used were 104‑106, around 7‑9 cycles per second.
To complete the waveform picture, route all three tones through the chorus. The best setting for this is pretty much a matter of common sense and personal taste.
Having achieved the waveform combination, the filter is next in line for examination. Since the Solina is characteristically very bright and airy, it would not be entirely inappropriate to bypass the filter altogether. If this is simply too bright, then globally rolling just a little off the top using the LPF should bring it into line. Roll off too much, though, and we're not talking Solina anymore!
To complete the simulation, the languid amplitude envelope that characterises the Solina so well can be applied, comprising a slow attack and a long release time. The real instrument actually has a monophonic envelope generator, which means that it restarts its cycle only when all the notes are released and a new note or chord is played — an attribute I always refer to as the 'Solina Suck'. Any notes whose release time is still sounding are abruptly cut off as the envelope cycle begins anew. Some synths can be made to reproduce this effect, but otherwise we have to live without it. Some may prefer things that way!
Next month I'll be looking at some classics of the analogue brass family. Until then it seems like the right moment to rummage through my '70s record collection and chill out to some Greenslade.
A set of 64 classic retro sounds for the Roland JV series synths, including those sounds mentioned in these articles, is available free of charge from Roland dealers, or contact Roland UK on 0252 816181.