read gain from sample level WSMP in DLS

fix volenv recalculation being called twice on note on

src/spessasynth_lib/soundfont/dls/dls_sample.js

@@ -10,6 +10,7 @@ export class DLSSample extends BasicSample
      * @param loopStart {number} sample data points
      * @param loopEnd {number} sample data points
      * @param data {Float32Array}
+     * @param sampleDbAttenuation {number} in db
      */
     constructor(
         name,
@@ -18,7 +19,8 @@ export class DLSSample extends BasicSample
         pitchCorrection,
         loopStart,
         loopEnd,
-        data
+        data,
+        sampleDbAttenuation
     )
     {
         super(
@@ -32,13 +34,20 @@ export class DLSSample extends BasicSample
             (loopEnd - 1) * 2 // -1 sample because soundfont end is last sample and dls end is next sample
         );
         this.sampleData = data;
+        this.sampleDbAttenuation = sampleDbAttenuation;
     }
 
     getAudioData()
     {
         return this.sampleData;
     }
 
+    /**
+     * in decibels of attenuation, WITHOUT EMU CORRECTION
+     * @type {number}
+     */
+    sampleDbAttenuation;
+
     /**
      * @type {Float32Array}
      */

src/spessasynth_lib/soundfont/dls/read_region.js

@@ -65,11 +65,10 @@ export function readRegion(chunk)
     );
 
     // gain correction:  Each unit of gain represents 1/655360 dB
+    // it is set after linking the sample
     const gainCorrection = readLittleEndian(waveSampleChunk.chunkData, 4);
     // convert to signed and turn into attenuation (invert)
     const dbCorrection = (gainCorrection | 0) / -655360;
-    // convert to centibels
-    const attenuation = (dbCorrection * 10) / 0.4; // make sure to apply EMU correction
 
     // skip options
     readLittleEndian(waveSampleChunk.chunkData, 4);
@@ -118,11 +117,20 @@ export function readRegion(chunk)
     readLittleEndian(waveLinkChunk.chunkData, 4);
     // sampleID
     const sampleID = readLittleEndian(waveLinkChunk.chunkData, 4);
+    /**
+     * @type {DLSSample}
+     */
     const sample = this.samples[sampleID];
     if(sample === undefined)
     {
         throw new Error("Invalid sample ID!");
     }
+
+    // this correction overrides the sample gain correction
+    const actualDbCorrection = dbCorrection || sample.sampleDbAttenuation;
+    // convert to centibels
+    const attenuation = (actualDbCorrection * 10) / 0.4; // make sure to apply EMU correction
+
     zone.setWavesample(
         attenuation, loopingMode,
         loop,

src/spessasynth_lib/soundfont/dls/read_samples.js

@@ -106,6 +106,7 @@ export function readDLSSamples(waveListChunk)
         let samplePitch = 0;
         let sampleLoopStart = 0;
         let sampleLoopEnd = sampleData.length - 1;
+        let sampleDbAttenuation = 0;
 
         // read wsmp
         const wsmpChunk = waveChunks.find(c => c.header === "wsmp")
@@ -120,8 +121,10 @@ export function readDLSSamples(waveListChunk)
                 wsmpChunk.chunkData[wsmpChunk.chunkData.currentIndex++],
                 wsmpChunk.chunkData[wsmpChunk.chunkData.currentIndex++]
             );
-            // gain is handled in regions as initialAttenuation
-            readLittleEndian(wsmpChunk.chunkData, 4);
+            // gain is applied it manually here (literally multiplying the samples)
+            const gainCorrection = readLittleEndian(wsmpChunk.chunkData, 4);
+            // convert to signed and turn into decibels
+            sampleDbAttenuation = (gainCorrection | 0) / -655360;
             // no idea about ful options
             readLittleEndian(wsmpChunk.chunkData, 4);
             const loopsAmount = readLittleEndian(wsmpChunk.chunkData, 4);
@@ -162,7 +165,8 @@ export function readDLSSamples(waveListChunk)
             samplePitch,
             sampleLoopStart,
             sampleLength,
-            sampleData
+            sampleData,
+            sampleDbAttenuation
         ));
 
         sampleID++;

src/spessasynth_lib/synthetizer/worklet_system/main_processor.js

@@ -82,6 +82,7 @@ class SpessaSynthProcessor extends AudioWorkletProcessor
 
         /**
          * Interpolation type used
+         * TODO: cubic interpolation?
          * @type {interpolationTypes}
          */
         this.interpolationType = interpolationTypes.linear;

src/spessasynth_lib/synthetizer/worklet_system/worklet_methods/note_on.js

@@ -59,6 +59,7 @@ export function noteOn(channel, midiNote, velocity, enableDebugging = false, sen
         const exclusive = voice.generators[generatorTypes.exclusiveClass];
         if(exclusive !== 0)
         {
+            // kill all voices with the same exclusive class
             channelVoices.forEach(v => {
                 if(v.generators[generatorTypes.exclusiveClass] === exclusive)
                 {
@@ -72,7 +73,9 @@ export function noteOn(channel, midiNote, velocity, enableDebugging = false, sen
         }
         // compute all modulators
         computeModulators(voice, channelObject.midiControllers);
-        WorkletVolumeEnvelope.intialize(voice);
+        // set the current attenuation to target,
+        // as it's interpolated (we don't want 0 attenuation for even a split second)
+        voice.volumeEnvelope.attenuation = voice.volumeEnvelope.attenuationTarget;
         // set initial pan to avoid split second changing from middle to the correct value
         voice.currentPan = ((Math.max(-500, Math.min(500, voice.modulatedGenerators[generatorTypes.pan] )) + 500) / 1000) // 0 to 1
     });

src/spessasynth_lib/synthetizer/worklet_system/worklet_utilities/volume_envelope.js

@@ -147,22 +147,11 @@ export class WorkletVolumeEnvelope
         WorkletVolumeEnvelope.recalculate(voice);
     }
 
-    /**
-     * Initializes a volume envelope
-     * @param voice {WorkletVoice}
-     */
-    static intialize(voice)
-    {
-        WorkletVolumeEnvelope.recalculate(voice, true);
-        voice.volumeEnvelope.attenuation = voice.volumeEnvelope.attenuationTarget;
-    }
-
     /**
      * Recalculates the envelope
      * @param voice {WorkletVoice} the voice this envelope belongs to
-     * @param setupInterpolated {boolean} if we should initialize the interpolated values (attenuation and sustain)
      */
-    static recalculate(voice, setupInterpolated = false)
+    static recalculate(voice)
     {
         const env = voice.volumeEnvelope;
         const timecentsToSamples = tc =>
@@ -172,10 +161,6 @@ export class WorkletVolumeEnvelope
         // calculate absolute times (they can change so we have to recalculate every time
         env.attenuationTarget = Math.max(0, Math.min(voice.modulatedGenerators[generatorTypes.initialAttenuation], 1440)) / 10; // divide by ten to get decibels
         env.sustainDbRelative = Math.min(DB_SILENCE, voice.modulatedGenerators[generatorTypes.sustainVolEnv] / 10);
-        if(setupInterpolated)
-        {
-            env.attenuation = env.attenuationTarget;
-        }
         const sustainDb = Math.min(DB_SILENCE, env.sustainDbRelative);
 
         // calculate durations

src/spessasynth_lib/synthetizer/worklet_system/worklet_utilities/worklet_modulator.js

@@ -116,6 +116,8 @@ export function computeModulators(voice, controllerTable, sourceUsesCC = -1, sou
     const { modulators, generators, modulatedGenerators } = voice;
 
     // Modulation envelope is cheap to recalculate
+    // why here and not at the bottom?
+    // I dunno, seems to work fine
     WorkletModulationEnvelope.recalculate(voice);
 
     if (sourceUsesCC === -1)

src/website/js/dls_to_sf2.js

@@ -27,8 +27,8 @@ document.getElementById("dls_upload").oninput = e => {
             return;
         }
         document.getElementById("sf_info").classList.remove("hidden");
-        document.getElementById("dls_name").innerText = sfont.soundFontInfo["INAM"];
-        document.getElementById("dls_description").innerText = sfont.soundFontInfo["ICMT"].replace("\nConverted from DLS to SF2 with SpessaSynth", "");
+        document.getElementById("dls_name").innerText = sfont.soundFontInfo["INAM"] || "Unnamed"            ;
+        document.getElementById("dls_description").innerText = (sfont.soundFontInfo["ICMT"] || "No description").replace("\nConverted from DLS to SF2 with SpessaSynth", "");
         document.getElementById("dls_presets").innerText = sfont.presets.length;
         document.getElementById("dls_samples").innerText = sfont.samples.length;
         message.innerText = "Loaded!";