Shaping the Wave
Since Yamaha popularised FM synthesis, a lot of attention has been paid to digital technology. Lorenz Rychner explains how a lot of this technology still uses analogue programming techniques.
We're living in the age of the digital synth - but lurking beneath the casing of many popular digital synths is a disguised, familiar method of sound generation: analogue synthesis.
THE SALESMAN ASSURED you it was digital. The ad said it was FM, or LA, or PD, or VS or something. The owner's manual lists features like TVA, TVF, MG, DCW. If you grew up in the days of VCOs, VCFs, VCAs, ADSRs and LFOs you're looking in vain for old friends. If it's your first keyboard, you're sticking to factory sounds. Who has the time to learn all these different programming techniques - and you can always buy more sounds on cartridges, so why bother?
Then again, it would be nice to make your own sounds. You've seen books on analogue programming but isn't analogue supposed to be dead, replaced by digital? And your synth is digital, that's why you bought it in the first place. What's going on?
I'll tell you: analogue (or subtractive) synthesis is alive and well. Don't send flowers, it's doing just fine. Unless your only keyboard is a Yamaha DX instrument, analogue programming is an integral part of your synth no matter how digital it's supposed to be.
Actually, there is one thing wrong with today's analogue: its name. "Analogue" simply means that the sound signal is at all times represented by a voltage, but that's not what happens on today's synthesisers and samplers. They're all computers, and they treat the sound signal as a set of numbers that are ultimately converted into an analogue signal that an amplifier can boost and send to the speakers. So technically speaking, analogue synthesisers are a thing of the past. We're actually dealing with analogue/digital hybrids - synthesisers that work partly with analogue signals and partly with a digitised version.
What about samplers? Aren't they totally digital? Does "analogue programming" apply there, too? You bet. Once the samples have been recorded and digitally processed, they need to be treated with good old analogue processes before they become musically useful. These processes may be hiding behind unfamiliar names, but they're analogue processes all the same. Even an instrument like E-mu's new Emulator III - they call it a Digital Sound Production System - has analogue written on its front panel.
"ANALOGUE PROGRAMMING" IS still the technique you need to control your digital instrument. And it hasn't changed all that much since the days when all synthesisers were truly analogue. As I said, the main problem is the name. To make their instruments appear more advanced some manufacturers have come up with new names for features that were around when Hammer was a tool and Miami had no vice. Time Variant Filter (TVF) and Modulation Generator (MG) may look strange, but they aren't that hard to figure out once you realise that they're comparable to a VCF and an LFO.
There are fundamental differences between analogue and digital sound generation. Analogue sounds traditionally have a warmth that is often missing from digital sounds. But there are many advantages to using digital technology: stable tuning, program recall, large selections of oscillator waveforms, complex envelope generators, the entire sampling process, onboard signal processors, MIDI... These facilities wouldn't be available if it wasn't for the bits and bytes in your instrument's computer chips. Not to mention the effect digital technology has had on instrument prices.
Does this mean that you have to learn computer programming? No, just about sound.
ALL SOUNDS CONSIST of three basic elements: timbre, pitch and amplitude. To begin the programming process you first need to select a basic sound source from the oscillator, which produces the sound. Your first choice is the waveform, which will determine the sound's timbre, and your second choice is the waveform's tuning, which determines its pitch.
The waveform puts your sound into certain broad categories, according to its characteristics. Traditionally you could select from three or four waveforms: sawtooth (or ramp), pulse, triangle and sine. A pulse wave is a square wave that is modulated to alter the pulse width, hence pulse width modulation or PWM. None of these waveforms bear more than a passing resemblance to any "real" sounds or instruments - until you modify them. Great sounds are still being made using these basic waveforms, whether they are supplied by VCOs (Voltage Controlled Oscillators) or by DCOs (Digitally Controlled Oscillators).
How has digital technology helped in this area? Digital memory has become cheap, and cheap storage of large amounts of data has made cheap synths powerful or expensive synths very powerful - it depends on your point of view.
This mix-n-match approach to waveforms can produce rich timbres on synths with more than one oscillator. Sequential went a step further with Vector Synthesis, where you can combine waveforms with a joystick, literally stirring them up until the mixture is just right. Roland took this to the extreme with their D50, where a Patch can be a layer of two Tones, and each Tone is a blend of two Partials. "Partials" here refers to 100 samples permanently resident in the D50. These samples can be treated as if they were simple waveforms - or you can choose to use sawtooth or pulse waves. Suddenly we're back to basics.
"Analogue programming is still the technique you need to control your digital instrument. And it hasn't changed since the days when all synthesisers were truly analogue."
The sawtooth is the brightest of the basic waveforms. It lends itself to the synthesis of brass and string sounds, as well as many other sounds that require a colourful, gutsy waveform. What makes it sound like that? Two things give the sawtooth its harmonic spectrum and hence, its sound. Firstly it contains all of nature's harmonics (overtones) and secondly, the loudness of the harmonics compared to that of the fundamental follows a fractional formula: the amplitude of any harmonic is the reciprocal of the number of that harmonic. For example, the fifth harmonic is a fifth as loud as the fundamental.
The pulse wave is less easily defined. Its sound colour depends on the width of the pulse. The pulse wave has no up or down slope, it's either positive (on) or negative (off). The pulse width (duty cycle) describes the ratio of the positive to negative parts of the cycle. If that ratio is equal (50/50 or 50%), the pulse is called a square wave and it contains only odd numbered harmonics. Their loudness follows the same reciprocal formula as the sawtooth. When square, this wave sounds similar to a clarinet in its lower register, slightly woody and hollow, but well rounded. The more the pulse width changes away from 50%, the more the sound becomes nasal, piercing and bright, similar to double reed instruments and twangy strings (oboes and harpsichords).
The triangle wave is a more bland version of the square wave. It also contains the odd numbered harmonics, but they are much weaker. Their loudness is inversely proportional to the fundamental: the third harmonic is a ninth as loud, the fifth harmonic is 1/25 as loud... Not much colour there.
The sine wave is the simplest of all waveforms. A pure sine wave has no overtones. Its timbre is musically of little use. But because it is nature's building block of sound, and because computers have no trouble crunching massive amounts of numbers, there are affordable synthesisers that allow you to construct your own waveforms with sine waves. You combine them using an additive process - hence it's called additive synthesis - to produce the harmonic series of the waveform you require. The computer then produces the desired timbre from this combination of sine waves. Kawai's K3, Korg's DSS1 and DSM1 all offer this facility while Kawai's K5 goes one step further, with four separate envelopes controlling the appearance and disappearance of selectable groups of waves. Not long ago this could only be done on equipment in the Fairlight league. Now it's here for everyone, and you can load up the oscillators with your own creations.
SINCE THE "V" (for voltage) has become a "D" (for digital) in VCOs/DCOs, we don't have to worry about tuning problems due to fluctuations in mains voltage, heat, and some less explicable causes. The instability of analogue oscillators is a thing of the past.
All pitch control has to happen at the oscillators. The keyboard talks to the oscillators for pitch selection. So does the LFO when its job is that of creating pitch modulation, and so do other programmable features that can control the pitch, like Auto Bend, or Pitch Envelope Generators. As you play, you may have further pitch control available in real time, from Pitch Bias via keyboard velocity or aftertouch, and from the pitch-bend wheel or joystick. Portamento (or glide, as Moog had it) connects the played pitches in a smooth, gliding manner, and its speed can be adjusted. Beware of parameters that say something like "Porta Time" when it's really a rate of speed rather than a fixed amount of time. If the distance between two adjacent notes is covered in the same amount of time as the distance over several octaves, then it is truly a Time parameter (as on the OSCar). But if the time varies with the pitch distance, then you're dealing with a Rate.
The oscillators themselves can be set to one of several octave ranges that are sometimes listed as 16, 8, 4, with an occasional 32 or 2. These numbers are taken from pipe organ terminology where the physical length (in feet) of the organ pipe determines pitch range. Further tuning is usually available in halfsteps, on at least one of the oscillators on synths with multiple oscillators. By using this capability you can set up constant intervals between oscillators. A facility to slightly detune the oscillators from each other can be used to create oscillator beating, to fatten a sound or turn the sound of a solo instrument into that of an instrument section.
Cross modulation, where one oscillator's waveform imposes its frequency on the waveform of another oscillator, has always been a source of new and often amazing timbres. When combined with Auto Bend, this feature creates sweeping changes in timbre that are hard to describe but easy to love. Digital memory allows very precise programming of such effects, because the tuning of the contributing oscillators is memorised more accurately than on analogue equipment.
Whether it's called FM through cross-modulation as on the Korg DS8, or ring modulation as on the D50, or simply oscillator sync or hard sync, try it on your synth. It may get you the metallic sounds that the DX7 has become famous for.
ONCE YOU'VE SELECTED and tuned your oscillator's waveforms, the fun starts in earnest. Whatever the timbre is that you've come up with, you'll want to modify it. At least one, possibly several filters let you do just that. The main filter goes under many names, but it always acts as a low-pass filter (LPF). You'll hardly ever see this name, but it says more about what the filter really does to the waveform than any of the names you see on today's instruments: VCF (Voltage Controlled Filter) on the majority of synths; DCW (Digitally Controlled Waveform) on Casio CZ synths; TVF (Time Variable Filter) on Roland's D50 and MT32.
"Analogue sounds have a warmth often missing from digital sounds - but there are advantages to digital technology: stable tuning, program recall, onboard signal processors..."
The LPF works in a subtractive manner. It can't add overtones to the waveforms from the oscillators, but it can remove them. It does that "from the top down". When the filter is fully open, all overtones pass unaltered. As soon as the filter starts to close down, the highest overtones are eliminated. The more the filter closes, the fewer high overtones pass, until the filter closes completely, by which time even the lowest overtones and eventually the fundamental are blocked. No more waveform: no more sound.
The parameter that lets you open or close the filter may be called Cutoff, Frequency, Fc or something similar.
Frequencies higher than the cutoff frequency will be eliminated, while lower ones will pass through the filter. But the term cutoff is misleading, because the eliminated frequencies are faded out or rolled off rather than cut.
Let's assume that you've picked a nice, fat waveform and you play and hold the middle C key. If your oscillator waveform is tuned to the 8' octave range and the keyboard is in its normal transposition range, you hear the pitch of middle C. If you find the sound too bright, you adjust the filter's frequency cutoff by ear.
Let's assume that the cutoff falls at 1000Hz, which happens to be just a hair below high C two octaves above middle C. Does this mean that you don't hear the frequency of the fourth harmonic, in this case 1046Hz, any longer? If the cutoff were a true cutoff, then the answer would be yes, but no filter acts this sharply so you still hear it a bit. In fact, you still hear frequencies as far as one octave away from the cutoff point.
The fade of overtones above the cutoff point is gradual, and the curve can be adjusted on many synths. If you see a switch for 2-pole/4-pole selection, alternate between the two and see if you can hear the difference. A pole is a measurement of attenuation by 6dB per octave. Without getting involved in mathematical formulae, think of it this way: the 2-pole setting causes frequencies that are one octave higher than the cutoff point to be softer by 12dB than the frequencies at the cutoff point. With a 4-pole setting, this changes to 24dB. In practice, you'll always judge the effect by ear, but it's nice to know what to expect. String ensemble and other lush sounds tend to sound better with a 2-pole setting, whereas sounds that need extra bite work best with the 4-pole setting.
While most VCFs eliminate all sound, overtones plus fundamental, when their cutoff point is low, there are several filters that only remove the overtones, leaving a sine wave at the lowest cutoff value. Korg's DS8 and the Casio CZ series of synths come to mind.
Next to the cutoff parameter you have a control called Resonance, Emphasis, Regeneration, Feedback or simply Q. This is a specialised loudness regulator that affects only a small portion of the overtones of your waveform. It concentrates on the frequencies at the cutoff point. In the example above, a high amount of Resonance would make the frequencies around 1000Hz a lot louder than they normally are in this waveform when compared to the fundamental's loudness. This gives a biting edge to your sound. If you leave the Resonance at a high value and you adjust the cutoff point up and down, you'll hear individual harmonics stand out as individual pitches.
On most synths, a maximum value of Resonance over-emphasises the frequencies at the cutoff point, and the filter self-oscillates on those frequencies. You hear a sine wave that may not follow the pitches from the keyboard the way you expect. If you adjust the values for the parameter that is called Keyboard Track, or Key Follow, you'll get microtonal scales. The filter is now acting as the sound source, and the cutoff point determines the tuning. Consequently, Keyboard Track is the only link between the filter and the keyboard.
Keyboard Filter Tracking is used to adjust the brightness changes across the whole keyboard. It acts by adjusting the cutoff point relative to the keyboard range. As a rule of thumb, give it a medium value when you start on a sound. Play your sound-in-progress at both ends of the keyboard and adjust the value by ear.
"Digital memory has become cheap, and cheap storage of large amounts of data has made cheap synths powerful or expensive synths very powerful -depending on your point of view."
MOST OF YOUR sounds will need to undergo changes in brightness during every note you play. But you can't make music with one hand on a data entry slider or increment/decrement button. You need something that will make those changes predictably and automatically each time you strike a key. For this you have the Filter Envelope and the LFO.
Envelope Generators have come a long way from the classic ADSR but no matter what your Filter Envelope looks like, you need to program the amount of change that's going to occur to the cutoff point during every note you play. For this you have a parameter called Envelope Amount, ENV, EG Depth Level or something similar. The higher the value, the further the cutoff travels from the value that you gave it with the Cutoff parameter. But there comes a point of overkill. If your cutoff is set to a very high value, and you assign a very high Envelope Amount, then you're asking the Envelope to open the filter wider than is possible. Depending on the design of your synth, either the filter opens fully, for which it would only need a small Envelope Amount and it ignores the rest of the Envelope's exaggerated message, or the filter opens fully and uses the extra Envelope Amount as a timing instruction, by staying fully open for a longer time than the rates in the envelope suggest. This can be confusing when you've set a fast decay rate to reach a low sustain level quickly and the filter cutoff stays up there, seemingly forever.
On most synths you have a choice about the direction in which the cutoff point travels when you play a note. Normally the cutoff point rises during the attack then falls during the decay. This is termed positive polarity. The alternative is negative polarity, which is achieved by inverting the filter envelope.
The LFO (Low Frequency Oscillator) simply shifts the cutoff point up and down in a regular pattern, from wherever the cutoff happens to be. If the Envelope moves the cutoff point, then the LFO follows the new point.
You may have more control over the cutoff point available as you play from other modulation sources. Keyboard velocity, aftertouch, pedals, mod wheels and joysticks let you modify the cutoff point, the envelope amount, the envelope timings, and the amount of LFO influence over the cutoff point. This is an area where fast-acting computers can help your musical expression. But when it comes to the sound itself, the digital technology hasn't had the same impact on filters as it had on oscillators.
Many synths have another filter called either a High-Pass Filter (or HPF) or a Low-Cut Filter. On most synths it acts as a static tone control for the bass, attenuating some of the lower frequencies as you raise its cutoff point. It's normally used to thin synth sounds that otherwise tend to clog the lower frequency band of a mix - because the filter is letting all the lower frequency components of the synth sounds through.
WHATEVER'S LEFT OF your waveform after it has passed the filter section must now be programmed for maximum volume, and for changes in volume during the duration of a note. The maximum volume for any sound is set with the VCA Level. This is important - don't just set it at maximum for every sound. If you have soft sounds or heavily filtered waveforms in some sounds, and loud and aggressive sounds in others, you'll get wildly differing volume levels when you change patches. Recording engineers and input stages of mixing desks respond very badly to unexpected changes in level.
For predictable changes in loudness use the VCA Envelope Generator. It takes care of the speed at which the loudness rises when you first strike a key, and, depending on the envelope's design, can allow several programmed changes until the loudness returns to zero after you release the key.
The LFO, when assigned to the VCA, can raise and lower the volume in its repetitious manner. It follows the changes in loudness level that the VCA EG causes. More control over the loudness is available as you play - Keyboard velocity, aftertouch, pedals, mod wheels and joysticks let you change the VCA Level, the envelope timings and the amount of LFO influence on the loudness.
A FEW TRENDSETTING instruments incorporate digital sound processors that are normally found only as outboard gear. Yamaha (TX81Z), Korg (DSS1, DS8) and Roland (D50, MT32) instruments let you program echo, delay, reverb and equalisation effects, and they are sure to be followed by others. Programmable output assignment and stereo panning is another luxury that is bound to become standard. Let's not forget MIDI, with its possibilities for remote programming and patch alterations via system exclusive. To be able to virtually re-program a sound while the synth is producing music under the command of a sequencer is more than useful.
So whether your keyboard uses FM, LA, PD, VS or even sampling, a good working knowledge of analogue synth programming helps you to get your money's worth. Go for it.
Feature by Lorenz Rychner
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!