Reason's Malström Synthesizer

During this class, we looked at Reason's Malström synthesizer, which Propellerheads refers to as a "Graintable" Synth. The name comes from a combination of two synthesis methods: Granular Synthesis and Wavetable Synthesis.

In Subtractive Synthesis, a raw waveform is used as the basic sonic building block; in Sampler-Based Instruments, a recording of an actual sound is the basic sonic element. Graintable Synthesis takes a different approach: Sounds are built using a short segment of audio called a wavelet, which falls somewhere between the two in length.

Think of a wavelet as a few cycles of a sampled sound. Typically, a wavelet lasts for only a couple dozen milliseconds. At the beginning of the wavelet, there is a short fadeup and the end of the wavelet has a similarly short fadeout. This allows the wavelet to crossfade when it is looped. This A-Gate-R envelope is sometimes called a Windowing Envelope as it acts as the "window" through which the wavelet is "seen."

When a wavelet is looped, it's as though the timbre of the sample at that moment was "frozen." Some very interesting and impressionistic "clouds" of sound can result. For recorded examples of Granular Synthesis, check out the work of Greek composer Iannis Xenakis.

Wavelets can also be used as a way of time-stretching a sound. A typical sound has attack, decay and sustaining portions. A sequence of wavelets could be created for each of the various moments in each portion of the sampled wave. Playing each wavelet twice before moving to the next wavelet in the sequence would have the effect of doubling the duration of the sample.

True Granular Synthesis is difficult to work with because it requires a tremendous amount of computing power to create and perform the wavelets. As a result, it has limited popularity among electronic composers. Still there are hardware and software packages that can let you dabble in it, notably the Kyma System from Symbolic Sound, the Barry Vercoe's Csound programming language and Cycling 74's MAX/MSP environment (which we have at Mercy).

(Incidentally, it's easy to hear the timbre of a single wavelet in Reason. Choose a graintable from the Oscillator A browser. Set the Motion control to -64 (all the way left). This stops the Malström from cycling through the table, so that it plays one wavelet repeatedly. Now you can use the Index control to scroll through the wavelets until you find one you like.)

Wavetable Synthesis takes a different approach to synthesis. Basically, a number of sampled waves are arranged in a ROM or RAM table. The Wavetable synth can access any portion of the table at any time. By instructing the Wavetable synth to scroll through various parts of the table, you can create wildly dynamic timbres. In some synths, the tables can even be scrolled in real time via a joystick contoller! One early synth using wavetables, the Roland D-50, used the sampled attacks and decays from acoustic instruments and substituted sustain waveforms from a subtractive synthesizer. A more sophisticated wavetable synth is the Waldorf PPG.

In a sense, every sample-based "synthesizer" can be thought of as a kind of wavetable synth—even the Roland Juno-D! But to really take advantage of the power of the wavetable, there must be some process that lets you scroll or cycle through the table, and a way to control that process.

In the Malström, the two techniques are combined. The wizards at Propellerheads have collected 80 sampled waveforms and for each sample, they've created a wavetable consisting of a sequence of wavelets—the Graintable.

If you play the wavelets in the normal order, and with the normal speed and tuning, you will hear what seems to be the original sampled waveform. In reality, what you hear is the reconstituted waveform. If the wavlets are played at the normal speed, but in reverse order, the sound will be played backward. If the wavelets are played in the normal sequence, but moving through the table at a slower speed (using the Motion control), the sound will take longer to attack and decay—but with no change to pitch. You can slow the Motion way down to create unearthly textures. You can also get the Motion to stop completely and use the Index control to manually scroll through the table.

An interesting aspect of the Malström is that the pitch tuning of the graintable is separate from the tuning of the notes you play. By using the Shift control, you can tune the table up or down. This creates a formant shift effect, similar to the timbral effects that occur when a recording is sped up or slowed down, but again without changing the pitch of the sample or its playback speed. Raising the Shift value makes the sound get brighter and "chipmunky"; lowering shift gives a filtered and logy effect.

Things gets totally off the hook when you take into account that the Malström has two triggerable or cyclic Modulators that can be routed to the Motion, Shift and Index controls of the two Oscillators. You can also route Pitch, ModWheel and external controllers to these controls as well.


To recap:

The Motion control determine how fast the Malström cycles through the graintable. Turning it left will slow the envelope of the graintable (without affecting its pitch. This is possible by simply repeating the wavelets; for example, if each wavelet were to play twice instead of once, the graintable would take twice as long to play. When the control is set all the way to the left, the motion throught he graintable stops entirely, and the Malström repeats one wavelet endlessly. Turning the Motion control to the right cycles the graintable faster than normal. At extreme settings, the cycling occurs faster than 20 Hz, which results in some interesting audio sidebands.

The Index control determines where in the graintable playback will start. Set left, the playback begins with the first wavelet in the table. Moving it to the right, causes the table to start from a later point—perhaps to eliminate the Attack portion of a sound's envelope, for instance. When the Motion control is set fully left, the Index determines which wavelet will be repeated endlessly.

The Shift control is a bit harder to explain. Every sampled note has an inherent original pitch. On a sampler, the sample rate is varied so that when a key is played that is above the original sampled note, the sample plays faster and is rendered higher in pitch. Play a lower note on the keyboard, the sample is played back slower and lower in pitch. Simple, no?

You've probably noticed with conventional samplers that a sample played too far from its original pitch sounds very "squirrely" when pitched high and "logy" when pitched down. Most samplers address this by using multiple samples pitched relatively close together so none are played too far out of range.

In the Malström, the graintable has an original pitch, but the actual note pitch is not determined by samples in the table, but by the rate at which the wavelets repeat. Shift lets you change the inherent pitch of the table independently of the note pitch. Moving Shift to the right, pitches the table up for a more "squirrely" sound; moving it left makes it heavy and logy. What's actually happening is that the Malström resamples the wave and more or less of the wave is able to occur during each wavelet. Shifting right lets more of the sampled wave occur in a wavelet. This adds extra harmonic content for a richer, almost "phase-sync" sort of sound. Shifting left lessens the harmonic content per wavelet and acts almost like a low-pass filter.



The Malström is unusual in that the Filter sections are placed after the Amplitude Envelope Generators. Both Oscillator "A" and "B" each have an ADSR followed by a Volume stage. These, in turn, feed either Filters "A" or "B" through a variety of routing possibilities, including true-stereo output.

Malstrom Routings

Oscillator A can be fed to three possible destinations:

  • To the Shaper & Filter A [via (1)]
  • To Filter B [via (2)]
  • To the Left channel output directly (that is, with no filtering) [(1) and (2) not selected]

  • Note that Oscillator A can be fed to both Shaper/Filter A and Filter B simultaneously.

Oscillator B can be fed to two possible destinations:

  • To Filter B [via (3)]
  • To the Right channel output directly (without filtering)

To further complicate things, Filter B can itself be fed to two possible destinations:

  • To the Shaper & Filter A [via (7)]
  • To the Right channel output directly.

The two Oscillators, the two Filters and the Shaper can be individually turned on or off, so there are scads of possible combinations, including:

  • OscA feeds Left out; OscB feeds Right out. [1, 2 & 3 not selected]
  • OscA feeds the Shaper to Left out; OscB feeds Right out. [1, 4]
  • OscA feeds FilterA to Left out; OscB feeds FilterB to Right out. [1, 5; 3, 6]
  • OscA feeds FilterA to Left out; OscA also feeds FilterB to Right out. [1, 5; 2, 6]
  • OscA and OscB both feed FilterB to Right out. [2, 3, 6]
  • OscA and OscB both feed FilterB, which feeds Shaper and Filter A to Left out. [2, 3, 6, 7, 4, 5]
  • OscA feeds Shaper and FilterA to Left out; OscB feeds FilterB, and is also sent to the Shaper and FilterA. [1, 4, 5; 3, 6, 7]
  • OscA feeds Left out directly; OscB feeds FilterB, Shaper and FilterA. [3, 6, 7, 4, 5]

With all of these combinations (and more!), there may be times when the left and right output channels may not put out a balanced stereo signal. The Spread control lets you manage this. When Spread is set fully left, the Left & Right channels are summed together in mono. When Spread is set to the right, the channels are fully panned to their respective sides. Intermediate settings will vary the width of the stereo signal. A meter and stereo master Volume round out this section.

Refer to the Reason Operation Manual (ROM) pdf for specific information on the Filter and Shaper settings.