The PadSynth engine was designed by Paul Nasca and to the best of our knowledge there is no comparable type generally available. It starts with a waveform that is virtually identical to the one used by AddSynth. However, each harmonic of this waveform is then widened and shifted and blurred, by rendering it with a harmonic profile (Frequency Distribution). The resulting sound is similar to Granular Synthesis, but can be generated very efficiently from a set of perfectly looping wavetables. These wavetables need to be precomputed after a change, so none of the harmonic profile controls are real-time. Also, it is the wavetables that are called for actual sound generation after passing through the usual envelope controls (and this part of sound generation does work real-time).
There are two tabbed views in a single window. You will normally be shown the harmonic structure, the envelopes one (as it's name suggests) consisting mostly of envelope and LFO inserts.
Harmonic Structure
(left side)
The way some of these controls operate is rather obscure. The effect on the sound is often quite subtle, while making dramatic changes to the visual aspect of the harmonic profile. Generally speaking, when widening a harmonic profile, the corresponding sound starts to beat and fluctuate irregularly, like a swarm of bees hovering around common focal points.
Base Type Base shape used to widen and spread each harmonic: Gauss, Square or Double Explonential.
Width (spread) of a single peak within the profile.
Repeat Repeat the same base shape multiple times in the generated profile.
Str Modulate and spread the base shape, thereby creating several side bands with frequency shifted slightly above / below the centre — side bands create a chorus like quality.
Freq Increase the frequency of this modulation to increase the density of side bands.
Scale The profile as a whole can be stretched or squeezed by this parameter. Note: when AutoScale is on (see below), the effect of this parameter is almost completely compensated.
Profile Which part of the profile to use: Full (default), or only the upper or lower half.
AmpMultiplier Apply a secondary modulation on top of the profile built thus far; the modulating shape can be a bell function (Gauss), a Sine wave or just a plateau in the centre (Flat).
AmpMode The way this secondary modulation is worked into the profile.
Sum: a linear blend between the secondary modulator, and the original profile; the fade is controlled by par2
Multiply: the profile is filtered by (multiplied with) the modulator; the strength of filtering is controlled by par2
Div1: the profile is divided by the modulator — where the latter is strong, the profile is damped
Div2: the modulator is divided by the profile, i.e. the profile is carved out of the modulator shape
Par1 Squeeze or spread the secondary modulating shape.
Par2 Controls how the secondary modulation is faded or combined with the harmonic profile.
AutoScale Automatically stretches or squeezes the resulting profile, so that all the various profile shapes generate a similar blurring effect. What is taken as nominal bandwidth is indicated in the profile display by the vertical bars and the dark background. If AutoScale is disabled, this nominal bandwidth is fixed and thus reshaping the profile also increases or decreases the actual spread.
Bandwidth This is the most important control, and defines the effective Bandwidth of the harmonic profile in cents. By increasing this value, the sound transitions gradually from the precise waveform to a sonic cloud.
Bandwidth Scale How the Bandwidth is adjusted with the increasing frequency of each harmonic.
Normal: Bandwidth grows with frequency, and thus perceptually the spread is the same on each harmonic.
EqualHz: Bandwidth is constant, independent of frequency. Perceptually this means that the bandwidth on higher harmonics seems to diminish.
Quarter, Half, 75%, 150%, Double: all these setting increase the Bandwidth for higher harmonics to various degrees
Inv.Half: here the Bandwidth is even reduced for higher harmonics by half an octave per octave
(right side)
PadSynth builds a series of wavetables to cover the most important middle octaves. Each tone is drawn from the closest wavetable starting at a random point (phase) and sampled up or down to the required note frequency. The quality of this resampling interpolation can be increased with the PadSynth Interpolation global setting. These quality settings (and the number of octaves) can be relevant for sharp and dense sounds, and for very high or low notes.
base The note value for the lowest wavetable generated.
smp/oct The number of wavetables generated within each octave.
no.oct The number of octaves to cover with dedicated wavetables.
Sample Size Size of the generated wavetable(s). A large wavetable captures more fine details of the harmonic profile and gives more time until the patterns repeat, but it costs more time to build and it takes up more memory.
128k: the default value provides 2.6sec of sound at 48kHz sampling rate and requires ½ MiB.
1M (big): the largest possible wavetable holds 21sec of sound until repetition, and requires 4 MiB of RAM per table.
Crossfade The time it takes to fade from the old wavetable to the new one (for Background and Auto-Apply modes).
Retrigger This button opens a window for Random regeneration of the wavetable.
There is a button to give access to an overall Resonance that can be applied to the engine.
Below this is a Waveform button which gives access to a highly detailed waveshape editor, which is identical to the one available in AddSynth — with the exception of the phase control on individual partials; PadSynth ignores these and always picks completely random phases, whenever building a new wavetable. Note also that you can copy and paste the entire settings of this waveform editor between AddSynth and PadSynth; you may create a clear and pronounced sound with AddSynth and then take it into PadSynth to soften, spread and blur it, while retaining its character.
(bottom)
Spectrum Mode defines the way PadSynth generates the spectrum, which is then rendered into wavetables.
Bandwidth(default): widen each harmonic of the base waveform spectrum with the harmonic profile, by an amount controlled through the bandwidth setting and bandwidth scale.
Discrete: similar to AddSynth, each harmonic is retained as a sharp line, not using bandwidth and profile, yet the overtones position can still be shifted and unharmonic.
Continuous: likewise ignoring bandwidth and profile, but this time taken to the other extreme; the outline of all harmonics is connected into a common distribution and thus rendered into a form of coloured noise.
Overtones Position Since PadSynth re-renders the partials with high resolution, it is possible to shift overtones to non-harmonic positions, to create a wide array of metallic and noisy flavours. For Harmonic there is no control, so the other parameters are inactive. Similarly Par 2 does nothing for Shift so is disabled for that variation.
Harmonic(default): Overtones are located at exact multiples of the base frequency.
Shift: all overtones are spread towards higher pitches
Power: here the spread is guided by a power function and thus increases excessively for higher harmonics; par2 defines the exponent (i.e. the acceleration)
ShiftU / ShiftL: par2 defines a threshold, harmonics above are spread or condensed
PowerU: par1 defines a turning point, par2 the strength of the effect; harmonics are condensed around the turning point by a power function
PowerL: par1 controls a linear blend between the harmonic positions and positions shifed by a power function; harmonics are here spread away from a fixed turning point at the 10-th overtone
Sine: harmonics are alternatingly shifted up or down, causing them to cluster; par2 defines the frequency and thus the density of these overtone clusters
Par 1 spreads the harmonics according to the 'Overtones Position' type.
Par 2 provides a further variation on the harmonics spread, depending on type.
ForceH moves the shifted harmonics by a variable amount back towards the nearest actual multiple of the fundamental. This allows you to reduce and fine-tune the actual amount of noisiness
Note
Apply Changes: If any of the above controls for the harmonic profile are altered it will be necessary to rebuild the wavetables to hear the effect. Also, be aware that with a very big sample size and/or octave range and samples/octave this could take many seconds to complete. Yoshimi provides several modes to handle these PadSynth wavetable builds, which can be configured in the Global Settings
Mute: build is triggered manually with the Apply button — disable part while wavetable is assembled. This is the legacy mode and should be used if the other modes cause audible clicks.
Background(default): build triggered manually — continue using old wavetables while building the new ones in the background; smooth transition with crossfade when ready.
Auto-Apply: automatically launch a wavetable build whenever a parameter is changed, then crossfade when ready.
Envelopes and LFOs
Once again we have all the standard inserts. There are also the same local controls found in AddSynth and SubSynth. None of these affect the wavetable itself, so there is no need for the apply button here.
The Export button that appears at the bottom of both of these views takes you to a filer window for exporting a complete wavetable with the total number of samples you have currently set. These don't include any changes in the sound produced by controls in the envelopes window.