PhaseWarp

PhaseWarp

Overview

PhaseWarp can be thought of as a special kind of filter for normalized phasor signals. When used with a phasor and a table-lookup oscillator this produces what is known as phase distortion synthesis, famously used on Casio synthesizers like the VZ-1.

Algorithm

The algorithm for PhaseWarp works by adjusting the midpoint position where the phasor reaches 0.5. The adjustment amount is usually a normalized bipolar signal in the range -1 to 1. When the position value is positive, it warps the midpoint towards the upper bounds, causing a greater slope in the upper half, while decreasing the slope in the lower half. A negative position does opposite, creating a higher slope in the lower half, and a lower slope in the upper half.

(The whole thing makes a lot more sense when you graph it. Someday, I'll do that here. For now, you'll just have to take my word for it.)

Tangled Files

phasewarp.c and phasewarp.h.

<<phasewarp.c>>=
#include "phasewarp.h"
<<funcs>>

<<phasewarp.h>>=
#ifndef SK_PHASEWARP_H
#define SK_PHASEWARP_H

#ifndef SKFLT
#define SKFLT float
#endif

<<funcdefs>>
#endif

Compute

This algorithm is so simple, it can be done in a stateless function called sk_phasewarp_tick. Its arguments are the input phasor signal and the warp factor.

<<funcdefs>>=
SKFLT sk_phasewarp_tick(SKFLT in, SKFLT warp);

<<funcs>>=
SKFLT sk_phasewarp_tick(SKFLT in, SKFLT warp)
{
    SKFLT out;
    SKFLT wmp;

    out = 0;

    <<calculate_warped_midpoint>>
    <<determine_side_and_warp>>
    return out;
}

Calculate warped midpoint. Call it wmp.

<<calculate_warped_midpoint>>=
wmp = (warp + 1.0) * 0.5;

Determine side and compute. If the phasor position is less than the warped midpoint, than compute with lefthand slope. Otherwise, compute righthand slope.

<<determine_side_and_warp>>=
if (in < wmp) {
    <<compute_with_lefthand_slope>>
} else {
    <<compute_with_righthand_slope>>
}

Compute With Lefthand Slope. This slope is equal to 0.5/wmp. This value scales the phasor input slope * in.

<<compute_with_lefthand_slope>>=
if (wmp != 0) out = ((SKFLT)0.5 / wmp) * in;

Compute Righthand Slope. The righthand slope is computed as 0.5/(1 - wmp). The output is computed as slope * (in - wmp) + 0.5.

<<compute_with_righthand_slope>>=
if (wmp != 1.0) {
    out = ((SKFLT)0.5 / (SKFLT)(1.0 - wmp)) * (in - wmp) + 0.5;
}