spessasynth_lib 3.25.4 → 3.25.6

package/synthetizer/worklet_system/worklet_methods/controller_control/master_parameters.js

@@ -6,7 +6,9 @@ import { SYNTHESIZER_GAIN } from "../../main_processor.js";
  */
 export function setMIDIVolume(volume)
 {
-    this.midiVolume = volume;
+    // GM2 specification, section 4.1: master-volume is squared though,
+    // according to my own testing, e seems like a better choice
+    this.midiVolume = Math.pow(volume, Math.E);
     this.setMasterPan(this.pan);
 }
 

package/synthetizer/worklet_system/worklet_methods/program_change.js

@@ -45,6 +45,7 @@ export function programChange(programNumber, userChange = false)
         this.presetUsesOverride = false;
     }
     this.setPreset(preset);
+    console.log(preset.presetName, this.isXGChannel, this.drumChannel, bank);
     this.synth.callEvent("programchange", {
         channel: this.channelNumber,
         program: preset.program,

package/synthetizer/worklet_system/worklet_methods/soundfont_management/reload_sound_font.js

@@ -16,8 +16,6 @@ export function reloadSoundFont(buffer, isOverride = false)
         if (isOverride)
         {
             this.overrideSoundfont = loadSoundFont(buffer);
-            // assign sample offset
-            this.overrideSoundfont.setSampleIDOffset(this.soundfontManager.totalSoundfontOffset);
         }
         else
         {

package/synthetizer/worklet_system/worklet_methods/worklet_soundfont_manager/worklet_soundfont_manager.js

@@ -23,24 +23,11 @@ export class WorkletSoundfontManager
          * @type {Function}
          */
         this.ready = readyCallback;
-        this.totalSoundfontOffset = 0;
         this.reloadManager(initialSoundFontBuffer);
     }
     
-    _assingSampleOffsets()
-    {
-        let offset = 0;
-        this.soundfontList.forEach(s =>
-        {
-            s.soundfont.setSampleIDOffset(offset);
-            offset += s.soundfont.samples.length;
-        });
-        this.totalSoundfontOffset = offset;
-    }
-    
     generatePresetList()
     {
-        this._assingSampleOffsets();
         /**
          * <"bank-program", "presetName">
          * @type {Object<string, string>}

package/synthetizer/worklet_system/worklet_utilities/lowpass_filter.js

@@ -4,9 +4,11 @@ import { generatorTypes } from "../../../soundfont/basic_soundfont/generator.js"
 /**
  * lowpass_filter.js
  * purpose: applies a low pass filter to a voice
- * note to self: a lot of tricks and fixes come from fluidsynth.
+ * note to self: a lot of tricks and come from fluidsynth.
  * They are the real smart guys.
  * Shoutout to them!
+ * Give their repo a star over at:
+ * https://github.com/FluidSynth/fluidsynth
  */
 
 export const FILTER_SMOOTHING_FACTOR = 0.1;
@@ -123,9 +125,8 @@ export class WorkletLowpassFilter
         
         if (!filter.initialized)
         {
-            // filter initialization
+            // filter initialization, set the current fc to target
             filter.initialized = true;
-            // don't smooth, override
             filter.currentInitialFc = initialFc;
         }
         else
@@ -161,6 +162,7 @@ export class WorkletLowpassFilter
         }
         
         // filter the input
+        // initial filtering code was ported from meltysynth created by sinshu.
         for (let i = 0; i < outputBuffer.length; i++)
         {
             let input = outputBuffer[i];
@@ -201,19 +203,23 @@ export class WorkletLowpassFilter
         }
         let cutoffHz = absCentsToHz(cutoffCents);
         
-        // fix cutoff on low frequencies (fluid_iir_filter.c line 392)
+        // fix cutoff on low sample rates
         cutoffHz = Math.min(cutoffHz, 0.45 * sampleRate);
         
+        // the coefficient calculation code was originally ported from meltysynth by sinshu.
+        // turn resonance to gain, -3.01 so it gives a non-resonant peak
         const qDb = qCb / 10;
-        // correct the filter gain, like fluid does
-        const resonanceGain = decibelAttenuationToGain(-1 * (qDb - 3.01)); // -1 because it's attenuation, and we don't want attenuation
+        // -1 because it's attenuation, and we don't want attenuation
+        const resonanceGain = decibelAttenuationToGain(-(qDb - 3.01));
         
-        // reduce the gain by the Q factor (fluid_iir_filter.c line 250)
+        // the sfspec asks for a reduction in gain based on the Q value.
+        // note that we calculate it again,
+        // without the 3.01-peak offset as it only applies to the coefficients, not the gain.
         const qGain = 1 / Math.sqrt(decibelAttenuationToGain(-qDb));
         
         
-        // initial filtering code was ported from meltysynth created by sinshu.
-        let w = 2 * Math.PI * cutoffHz / sampleRate; // we're in the AudioWorkletGlobalScope so we can use sampleRate
+        // note: no sin or cos tables here as the coefficients are cached
+        let w = 2 * Math.PI * cutoffHz / sampleRate;
         let cosw = Math.cos(w);
         let alpha = Math.sin(w) / (2 * resonanceGain);
         
@@ -248,7 +254,7 @@ export class WorkletLowpassFilter
     }
 }
 
-// precompute all the cutoffs for 0q
+// precompute all the cutoffs for 0q (most common)
 const dummy = new WorkletLowpassFilter();
 dummy.resonanceCb = 0;
 // sfspec section 8.1.3: initialFilterFc ranges from 1500 to 13,500 cents