flanger_1191.xml

<?xml version="1.0"?>
<!DOCTYPE ladspa SYSTEM "ladspa-swh.dtd">
<?xml-stylesheet href="ladspa.css" type="text/css"?>

<ladspa>
  <global>
    <meta name="maker" value="Steve Harris &lt;steve@plugin.org.uk&gt;"/>
    <meta name="copyright" value="GPL"/>
    <meta name="properties" value="HARD_RT_CAPABLE"/>
    <code><![CDATA[
#include "ladspa-util.h"
    ]]></code>
  </global>

  <plugin label="flanger" id="1191" class="FlangerPlugin">
    <name>Flanger</name>
    <p>A digital flanger implementation. Uses a novel zero excursion, controlled bandwidth modulation function, which should make the modulation less repetitive and noticable.</p>
    <p>This effect is similar in character to a phaser (see section \ref{lfoPhaser}). The main difference is that a phaser sounds more regular and stable.</p>

    <callback event="instantiate"><![CDATA[
int min_size;

sample_rate = s_rate;

prev_law_peak = 0.0f;
next_law_peak = 1.0f;
prev_law_pos = 0;
next_law_pos = 10;

min_size = sample_rate * 0.04f;
for (delay_size = 1024; delay_size < min_size; delay_size *= 2);
delay_tbl = malloc(sizeof(LADSPA_Data) * delay_size);
delay_pos = 0;
count = 0;
old_d_base = 0;
    ]]></callback>

    <callback event="activate"><![CDATA[
memset(delay_tbl, 0, sizeof(LADSPA_Data) * delay_size);
delay_pos = 0;
count = 0;
old_d_base = 0;
    ]]></callback>

    <callback event="cleanup"><![CDATA[
free(plugin_data->delay_tbl);
    ]]></callback>

    <callback event="run"><![CDATA[
unsigned long pos;
long d_base, new_d_base;
LADSPA_Data out;
float delay_depth;
float dp; // float delay position
float dp_frac; // fractional part
long dp_idx; // integer delay index
long law_p; // period of law
float frac = 0.0f, step; // Portion the way through the block
float law; /* law amplitude */
float n_ph, p_ph;
const float fb = f_clamp(feedback, -0.999f, 0.999f);

// Set law params
law_p = (float)sample_rate / law_freq;
if (law_p < 1) {
	law_p = 1;
}

// Calculate base delay size in samples
new_d_base = (LIMIT(f_round(delay_base), 0, 25) * sample_rate) / 1000;

// Calculate delay depth in samples
delay_depth = f_clamp(detune * (float)sample_rate * 0.001f, 0.0f, delay_size - new_d_base - 1.0f);

step = 1.0f/sample_count;
for (pos = 0; pos < sample_count; pos++) {
	if (count % law_p == 0) {
		// Value for amplitude of law peak
		next_law_peak = (float)rand() / (float)RAND_MAX;
		next_law_pos = count + law_p;
	} else if (count % law_p == law_p / 2) {
		// Value for amplitude of law peak
		prev_law_peak = (float)rand() / (float)RAND_MAX;
		prev_law_pos = count + law_p;
	}

	// Calculate position in delay table
	d_base = LIN_INTERP(frac, old_d_base, new_d_base);
	n_ph = (float)(law_p - abs(next_law_pos - count))/(float)law_p;
	p_ph = n_ph + 0.5f;
	while (p_ph > 1.0f) {
		p_ph -= 1.0f;
	}
	law = f_sin_sq(3.1415926f*p_ph)*prev_law_peak +
		f_sin_sq(3.1415926f*n_ph)*next_law_peak;

	dp = (float)(delay_pos - d_base) - (delay_depth * law);
	// Get the integer part
	dp_idx = f_round(dp - 0.5f);
	// Get the fractional part
	dp_frac = dp - dp_idx;

	// Accumulate into output buffer
	out = cube_interp(dp_frac, delay_tbl[(dp_idx-1) & (delay_size-1)], delay_tbl[dp_idx & (delay_size-1)], delay_tbl[(dp_idx+1) & (delay_size-1)], delay_tbl[(dp_idx+2) & (delay_size-1)]);

	// Store new delayed value
	delay_tbl[delay_pos] = flush_to_zero(input[pos] + (fb * out));
	// Sometimes the delay can pick up NaN values, I'm not sure why
	// and this is easier than fixing it
	if (isnan(delay_tbl[delay_pos])) {
		delay_tbl[delay_pos] = 0.0f;
	}

	out = f_clamp(delay_tbl[delay_pos] * 0.707f, -1.0, 1.0);
	buffer_write(output[pos], out);

	frac += step;
	delay_pos = (delay_pos + 1) & (delay_size-1);

	count++;
}

plugin_data->count = count;
plugin_data->prev_law_peak = prev_law_peak;
plugin_data->next_law_peak = next_law_peak;
plugin_data->prev_law_pos = prev_law_pos;
plugin_data->next_law_pos = next_law_pos;
plugin_data->delay_pos = delay_pos;
plugin_data->old_d_base = new_d_base;
    ]]></callback>

    <port label="delay_base" dir="input" type="control" hint="default_low">
      <name>Delay base (ms)</name>
      <range min="0.1" max="25"/>
      <p>This is the offset from the input time that the detune delay moves around.</p>
      <p>10 is probably a good starting value.</p>
    </port>

    <port label="detune" dir="input" type="control" hint="default_low">
      <name>Max slowdown (ms)</name>
      <range min="0" max="10"/>
      <p>This is the maximum delay that will be applied to the delayed signal, relative to the dry signal.</p>
    </port>

    <port label="law_freq" dir="input" type="control" hint="default_low,logarithmic">
      <name>LFO frequency (Hz)</name>
      <range min="0.05" max="100"/>
      <p>This is the core frequency that the 'LFO' will move at. The LFO isn't actually an oscillator, but it does vary periodically.</p>
    </port>

    <port label="feedback" dir="input" type="control" hint="default_0">
      <name>Feedback</name>
      <range min="-1" max="1"/>
      <p>Feedback applied from the output to the input, increases the depth of the effect, but makes it sound less like a real flanger.</p>
    </port>

    <port label="input" dir="input" type="audio">
      <name>Input</name>
    </port>

    <port label="output" dir="output" type="audio">
      <name>Output</name>
    </port>

    <instance-data label="sample_rate" type="long"/>
    <instance-data label="count" type="long"/>
    <instance-data label="prev_law_peak" type="float"/>
    <instance-data label="next_law_peak" type="float"/>
    <instance-data label="prev_law_pos" type="int"/>
    <instance-data label="next_law_pos" type="int"/>
    <instance-data label="delay_tbl" type="LADSPA_Data *"/>
    <instance-data label="delay_pos" type="long"/>
    <instance-data label="delay_size" type="long"/>
    <instance-data label="old_d_base" type="long"/>
  </plugin>
</ladspa>