fix: ImpulseTracker style instrument filters

it2taud.py

@@ -300,9 +300,13 @@ def _it214_decompress_block(payload: bytes, num_samples: int,
             delta = _sign_extend(v, width)
             is_data = True
         elif width < init_width:
-            # Mode B (mid): `range_count` escape codes centred on signed midpoint.
+            # Mode B (mid): `range_count` escape codes occupy values (border, border+range_count].
+            # The encoder simply does NOT emit data values that would collide with this slot —
+            # it widens first. So values *above* the escape range are sign-extended verbatim,
+            # not collapsed. Reference: schismtracker fmt/compression.c:103-127 and
+            # MilkyTracker XModule.cpp:629-640.
             # border = (mask >> (init_width-width)) - border_sub, where border_sub
-            # = range_count / 2. Reference: libxmp it_compress.c, OpenMPT ITTools.cpp.
+            # = range_count / 2.
             # 8-bit:  width=7 → border=63-4=59, width=8 → border=127-4=123
             # 16-bit: width=7..16 with border_sub=8.
             border = (mask >> (init_width - width)) - border_sub
@@ -310,8 +314,6 @@ def _it214_decompress_block(payload: bytes, num_samples: int,
                 new_w = v - border
                 width = new_w if new_w < width else new_w + 1   # skip-self
                 continue
-            if v > border + range_count:
-                v -= range_count   # collapse escape range out
             delta = _sign_extend(v, width)
             is_data = True
         else:
@@ -467,7 +469,8 @@ class ITInstrument:
     __slots__ = ('name', 'dfp', 'gv', 'canonical_sample', 'canonical_volume',
                  'vol_envelope', 'pan_envelope', 'vol_env_sustain', 'pan_env_sustain',
                  'pf_envelope', 'pf_env_sustain', 'pf_is_filter',
-                 'ifc', 'ifr', 'fadeout', 'pps', 'ppc', 'rv', 'rp', 'nna')
+                 'ifc', 'ifr', 'fadeout', 'pps', 'ppc', 'rv', 'rp', 'nna',
+                 'dct', 'dca')
     # vol_envelope / pan_envelope / pf_envelope: list of 25 (value, minifloat_idx) tuples, or None
     # *_env_sustain: int (16-bit, 0b 0ut sssss pcb eeeee), 0 = no envelope
     # pf_is_filter: bool — pf envelope mode (False = pitch, True = filter)
@@ -489,6 +492,11 @@ def parse_instruments(data: bytes, h: ITHeader) -> list:
         inst.name = data[ptr+0x20:ptr+0x3A].rstrip(b'\x00').decode('latin-1', errors='replace')
         # NNA at IMPI+0x11 (new format). 0=cut, 1=continue, 2=note off, 3=note fade.
         inst.nna  = data[ptr + 0x11] & 0x03
+        # DCT (Duplicate Check Type) and DCA (Duplicate Check Action), per Schism iti.c:80-94.
+        # DCT: 0=off, 1=note, 2=sample, 3=instrument.
+        # DCA: 0=note cut, 1=note off, 2=note fade.
+        inst.dct  = data[ptr + 0x12] & 0x03
+        inst.dca  = data[ptr + 0x13] & 0x03
         inst.fadeout = struct.unpack_from('<H', data, ptr + 0x14)[0]   # 0..1024
         # PPS is signed -32..+32; PPC is the centre note (IT note number 0..119, C-5=60).
         inst.pps  = struct.unpack_from('b', data, ptr + 0x16)[0]
@@ -516,9 +524,12 @@ def parse_instruments(data: bytes, h: ITHeader) -> list:
         inst.canonical_sample = c5_smp   # 1-based sample index, 0 = none
         inst.canonical_volume  = min(inst.gv, 64)
 
-        # Initial filter cutoff/resonance (high bit = enabled, low 7 bits = value)
+        # Initial filter cutoff/resonance (high bit = enabled, low 7 bits = value).
-        ifc_raw = data[ptr + 0x39]
+        # Per Schism iti.c struct it_instrument: name[26] occupies 0x20..0x39,
-        ifr_raw = data[ptr + 0x3A]
+        # ifc is at 0x3A, ifr at 0x3B. Off-by-one would silently disable filters
+        # on every IT instrument because name's last byte is always NUL.
+        ifc_raw = data[ptr + 0x3A]
+        ifr_raw = data[ptr + 0x3B]
         # Taud uses full 0..255 range (double IT's resolution): IT 0..127 → Taud 0..254,
         # IT "off" (high bit clear) → Taud 255.
         inst.ifc = (ifc_raw & 0x7F) * 2 if (ifc_raw & 0x80) else 255
@@ -1114,7 +1125,7 @@ def build_sample_inst_bin_it(samples_or_proxy: list,
         vol_env, vol_sus, pan_env, pan_sus, pf_env, pf_sus, pf_is_filter,
         inst_gv, fadeout, vib_speed, vib_depth, vib_sweep, vib_rate, vib_wave,
         default_pan, pps, ppc_taud, pan_swing, vol_swing, ifc, ifr,
-        sample_detune, nna.
+        sample_detune, nna, dct, dca.
     All optional; missing keys default to neutral values.
 
     Returns (bin_bytes[SAMPLEINST_SIZE], offsets_dict).
@@ -1282,7 +1293,13 @@ def build_sample_inst_bin_it(samples_or_proxy: list,
         inst_bin[base + 187] = idata.get('vib_depth', 0) & 0xFF
         # Byte 188: vibrato rate (0..255 full range, IT samplewise Vir).
         inst_bin[base + 188] = idata.get('vib_rate', 0) & 0xFF
-        # Bytes 189-191: reserved (already zeroed).
+        # Byte 189: duplicate-check / action (IT-only — bits 0-1 = DCT, bits 2-3 = DCA).
+        #   DCT: 0=off, 1=note, 2=sample, 3=instrument.
+        #   DCA: 0=note cut, 1=note off, 2=note fade.
+        dct = idata.get('dct', 0) & 0x03
+        dca = idata.get('dca', 0) & 0x03
+        inst_bin[base + 189] = (dca << 2) | dct
+        # Bytes 190-191: reserved (already zeroed).
 
         vprint(f"  instrument[{taud_idx}] '{s.name}' ptr:{ptr} c5spd:{s.c5_speed}")
 
@@ -1622,14 +1639,14 @@ def assemble_taud(h: ITHeader, samples: list, instruments: list,
                 default_pan = 0x80
 
             # Auto-vibrato lives on the canonical sample (not the IT instrument).
-            # IT samplewise auto-vibrato: Vis (speed 0..64), Vid (depth 0..32),
+            # IT samplewise auto-vibrato: Vis (speed 0..64), Vid (depth 0..64),
             # Vir (rate 0..255 — IT-style ramp-in), Vit (waveform 0..3).
             # Taud byte 175 (Vibrato Speed) follows FT2 0..255 scale: rescale Vis.
-            # Taud byte 187 (Vibrato Depth) is full 0..255: rescale Vid 0..32 → 0..255.
+            # Taud byte 187 (Vibrato Depth) is full 0..255: rescale Vid 0..64 → 0..255.
             # Taud byte 188 (Vibrato Rate) is IT Vir verbatim.
             # Taud byte 176 (Vibrato Sweep) is FT2-only — leave 0 for IT.
             vib_speed_taud = min(255, round(src_smp.av_speed * 255 / 64))
-            vib_depth_taud = min(255, round(src_smp.av_depth * 255 / 32))
+            vib_depth_taud = min(255, round(src_smp.av_depth * 255 / 64))
             # IT NNA (0=cut, 1=continue, 2=note off, 3=note fade) →
             # Taud NNA  (00=note off, 01=cut, 10=continue, 11=fade).
             it_to_taud_nna = (0b01, 0b10, 0b00, 0b11)
@@ -1658,6 +1675,8 @@ def assemble_taud(h: ITHeader, samples: list, instruments: list,
                 'ifr':        inst.ifr,
                 'sample_detune': 0,                    # IT samples have no finetune
                 'nna':        nna_taud,
+                'dct':        inst.dct,
+                'dca':        inst.dca,
             }
         sampleinst_raw, _ = build_sample_inst_bin_it(proxy, instr_data_by_slot)
     else:

terranmon.txt

@@ -2108,6 +2108,10 @@ TODO:
     [x] cue and pattern compression of the Taud format (taud_common.py, taud.mjs)
     [x] figure out how IT (0..256) and FT2 (0..FFF + cut) handles volume fadeout numbers, and come up with a compatible Taud spec, then implement
     [x] Pitchbend on Amiga frequency mode sometimes works right, sometimes works wrong. (effect underdelivers) Affects every song with Amiga picth mode, AND ON THE fresh taut.js session only
+    [x] Fix 4THSYM.it filters
+    [ ] 4THSYM.it: pitchbend is wrong, some notes keep playing (loudly!) even if new notes are emitted
+    [ ] some notes are emitted with wrong volset (tested with .mod, may affect others as well)
+    [ ] nearly_there_.mod: `C#5 SD300 / ... / C-5 SD200 / A#4 / G#4`: every C-5 SD200 (there are four occurances) gets skipped
     [ ] implement bitcrusher and overdrive  (eff sym '8' and '9')
 
 

tsvm_core/src/net/torvald/tsvm/peripheral/AudioAdapter.kt

@@ -1570,6 +1570,63 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
         if (voice.panbrelloRetrig) voice.panbrelloLfoPos = 0
     }
 
+    /**
+     * IT-style Duplicate Check (DCT/DCA). Runs *before* NNA on every fresh foreground
+     * trigger: existing voices on this channel — the foreground itself plus any of its
+     * own background ghosts — that match the new note under DCT have DCA applied.
+     * Reference: schismtracker effects.c:1664-1764 (csf_check_nna).
+     *
+     * DCT (per existing voice's instrument):
+     *   1 = note        — same noteVal AND same instrumentId
+     *   2 = sample      — same canonical sample (matched by samplePtr+sampleLength)
+     *   3 = instrument  — same instrumentId
+     * DCA: 0 = note cut, 1 = note off (release sustain), 2 = note fade.
+     *
+     * Note: the foreground voice will be replaced by triggerNote() right after this,
+     * so applying DCA to it is mostly relevant for ghosts spawned *from* it via NNA
+     * — the ghost is cloned from the (already-DCA-modified) foreground state.
+     */
+    private fun applyDuplicateCheck(ts: TrackerState, channel: Int, newInstId: Int, newNote: Int) {
+        if (newInstId == 0) return
+        val newInst = instruments[newInstId]
+
+        fun isDuplicate(v: Voice): Boolean {
+            val existInst = instruments[v.instrumentId]
+            return when (existInst.duplicateCheckType) {
+                1 -> v.noteVal == newNote && v.instrumentId == newInstId
+                2 -> v.instrumentId == newInstId &&
+                     existInst.samplePtr == newInst.samplePtr &&
+                     existInst.sampleLength == newInst.sampleLength
+                3 -> v.instrumentId == newInstId
+                else -> false
+            }
+        }
+
+        fun applyAction(v: Voice) {
+            val existInst = instruments[v.instrumentId]
+            when (existInst.duplicateCheckAction) {
+                0 -> { v.fadeoutVolume = 0.0; v.active = false }
+                1 -> v.keyOff = true
+                2 -> v.noteFading = true
+            }
+        }
+
+        val fg = ts.voices[channel]
+        if (fg.active && instruments[fg.instrumentId].duplicateCheckType != 0 && isDuplicate(fg)) {
+            applyAction(fg)
+        }
+
+        val it = ts.backgroundVoices.iterator()
+        while (it.hasNext()) {
+            val bg = it.next()
+            if (bg.sourceChannel != channel || !bg.active) continue
+            if (instruments[bg.instrumentId].duplicateCheckType == 0) continue
+            if (!isDuplicate(bg)) continue
+            applyAction(bg)
+            if (!bg.active) it.remove()
+        }
+    }
+
     /**
      * On a fresh foreground trigger, optionally migrate the existing voice into the
      * mixer-private background pool per the New Note Action setting (instrument default
@@ -1751,6 +1808,7 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
                         voice.delayedInst = row.instrment
                         voice.delayedVol = if (row.volume >= 0) row.volume else -1
                     } else {
+                        applyDuplicateCheck(ts, vi, row.instrment, row.note)
                         maybeSpawnBackgroundForNNA(ts, voice, vi)
                         triggerNote(voice, row.note, row.instrment, -1)
                     }
@@ -2027,6 +2085,7 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
             // Note delay — fire deferred trigger when the requested tick arrives.
             // NNA fires now (not at row parse) so that delayed retriggers ghost correctly.
             if (voice.noteDelayTick == ts.tickInRow) {
+                applyDuplicateCheck(ts, vi, voice.delayedInst, voice.delayedNote)
                 maybeSpawnBackgroundForNNA(ts, voice, vi)
                 triggerNote(voice, voice.delayedNote, voice.delayedInst, voice.delayedVol)
                 voice.noteDelayTick = -1
@@ -2161,12 +2220,17 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
             val finalPitch = (pitchToMixer + autoVibDelta + pitchEnvDelta).coerceIn(0, 0xFFFE)
             voice.playbackRate = computePlaybackRate(inst, finalPitch)
 
-            // Filter envelope (filter mode): scale current cutoff by env value (0..1, 0.5 = unity).
+            // Filter envelope (filter mode): scale baseCut by envValue (0..1, 0.5 = unity).
+            // Schism filters.c:80-86 computes `cutoff_used = chan->cutoff * (flt_modifier+256)/256`
+            // where flt_modifier = (env_value_0..64 - 32) * 8. Mapping TSVM's [0..1] env to Schism's
+            // [-256..+256] modifier and accounting for our pre-doubled defaultCutoff (it2taud.py
+            // stores IFC*2 in 0..254) gives `currentCutoff = baseCut * envPfValue` — at unity (0.5)
+            // the filter sits at IFC, at max (1.0) it opens to 2*IFC, at min (0.0) it closes.
             // If the instrument has no initial cutoff (255 = off), the envelope drives the filter
             // from the maximum active value (254) so the filter can become audible during the note.
             if (voice.hasPfEnv && voice.pfEnvOn && voice.envPfIsFilter) {
                 val baseCut = if (inst.defaultCutoff < 255) inst.defaultCutoff else 254
-                voice.currentCutoff = (baseCut * (voice.envPfValue * 2.0)).toInt().coerceIn(0, 254)
+                voice.currentCutoff = (baseCut * voice.envPfValue).toInt().coerceIn(0, 254)
             }
 
             // Refresh biquad filter coefficients once per tick (only recomputes when changed).
@@ -2250,7 +2314,7 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
             // Filter-mode pf envelope: same scaling rule as foreground.
             if (bg.hasPfEnv && bg.pfEnvOn && bg.envPfIsFilter) {
                 val baseCut = if (inst.defaultCutoff < 255) inst.defaultCutoff else 254
-                bg.currentCutoff = (baseCut * (bg.envPfValue * 2.0)).toInt().coerceIn(0, 254)
+                bg.currentCutoff = (baseCut * bg.envPfValue).toInt().coerceIn(0, 254)
             }
             refreshVoiceFilter(bg)
             // Reap fully-faded ghosts so the pool stays drained.
@@ -2969,7 +3033,10 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
      *                   waveform: 0=sine, 1=ramp-down, 2=square, 3=random, 4=ramp-up (FT2)
      *   187    u8 vibrato depth (0..255 full range)
      *   188    u8 vibrato rate  (0..255 full range — IT samplewise Vir)
-     *   189..191 byte[3] reserved
+     *   189    u8 duplicate-check / action (IT-only — 0b 0000 aadd)
+     *                   dd  = DCT (Duplicate Check Type) 0=off, 1=note, 2=sample, 3=instrument
+     *                   aa  = DCA (Duplicate Check Action) 0=note cut, 1=note off, 2=note fade
+     *   190..191 byte[2] reserved
      */
     data class TaudInst(
         var index: Int,
@@ -3002,7 +3069,8 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
         var sampleDetune: Int,              // bytes 184-185 (signed 4096-TET stored as u16)
         var instrumentFlag: Int,            // byte 186 (NNA + vibrato waveform)
         var vibratoDepth: Int,              // byte 187 (0..255 full range)
-        var vibratoRate: Int                // byte 188 (IT samplewise Vir)
+        var vibratoRate: Int,               // byte 188 (IT samplewise Vir)
+        var dupCheckFlag: Int               // byte 189 (DCT bits 0-1, DCA bits 2-3)
     ) {
         constructor(index: Int) : this(
             index, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xFF,
@@ -3010,7 +3078,7 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
             Array(25) { TaudInstEnvPoint(0x80, ThreeFiveMiniUfloat(0)) },
             Array(25) { TaudInstEnvPoint(0x80, ThreeFiveMiniUfloat(0)) },
             0, 0, 0, 0, 0, 0x80, 0x5000, 0, 0, 0xFF, 0,
-            0, 0, 0, 0
+            0, 0, 0, 0, 0
         )
 
         /** Sample-flag byte 14 bit 2 — when set, the sample loop is a sustain loop:
@@ -3024,9 +3092,13 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
         val vibratoWaveform: Int get() = (instrumentFlag ushr 2) and 0x07
         /** Sample detune as a signed 4096-TET delta. */
         val sampleDetuneSigned: Int get() = sampleDetune.toShort().toInt()
+        /** Duplicate Check Type — 0=off, 1=note, 2=sample, 3=instrument (IT semantics). */
+        val duplicateCheckType: Int get() = dupCheckFlag and 0x03
+        /** Duplicate Check Action — 0=note cut, 1=note off, 2=note fade. */
+        val duplicateCheckAction: Int get() = (dupCheckFlag ushr 2) and 0x03
 
-        // Reserved padding at offsets 189..191 (3 bytes per instrument).
+        // Reserved padding at offsets 190..191 (2 bytes per instrument).
-        private val reserved = ByteArray(3)
+        private val reserved = ByteArray(2)
 
         // Funk repeat (S$Fx00) bit-mask — non-destructive XOR overlay across the loop region.
         // Lazily allocated; a 1-bit flips the byte, a 0-bit leaves it intact.
@@ -3108,7 +3180,8 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
             186 -> instrumentFlag.toByte()
             187 -> vibratoDepth.toByte()
             188 -> vibratoRate.toByte()
-            in 189..191 -> reserved[offset - 189]
+            189 -> dupCheckFlag.toByte()
+            in 190..191 -> reserved[offset - 190]
             else -> throw InternalError("Bad offset $offset")
         }
 
@@ -3163,7 +3236,8 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
             186 -> { instrumentFlag = byte and 0xFF }
             187 -> { vibratoDepth = byte and 0xFF }
             188 -> { vibratoRate = byte and 0xFF }
-            in 189..191 -> { reserved[offset - 189] = byte.toByte() }
+            189 -> { dupCheckFlag = byte and 0x0F }   // DCT (bits 0-1) + DCA (bits 2-3)
+            in 190..191 -> { reserved[offset - 190] = byte.toByte() }
             else -> throw InternalError("Bad offset $offset")
         }
     }