diff --git a/plugin/module.cpp b/plugin/module.cpp
index c2405b0..c174ec9 100644
--- a/plugin/module.cpp
+++ b/plugin/module.cpp
@@ -58,16 +58,16 @@ void Module::lfoRampOn() {
 
 void Module::runLFO() {
     if (lfoDelayState == 1) {
-        lfoDelayTimer += lfoDelayTable[patchRam.lfoDelay >> 4];
-        if (lfoDelayTimer & 0xc000) lfoDelayState = 2;
+        lfoDelayTimer += attackTable[patchRam.lfoDelay];
+        if (lfoDelayTimer > 0x3fff) lfoDelayState = 2;
     }
     if ((lfoDelayState == 2)) {
-        lfoDelay += attackTable[patchRam.lfoDelay];
+        lfoDelay += lfoDelayTable[patchRam.lfoDelay >> 4];
     }
 
-    if (lfoDelay & 0xc000) {
+    if (lfoDelay > 0xff) {
         lfoDelayState = 0;
-        lfoDelay = 0x3fff;
+        lfoDelay = 0xff;
     }
 
     lfoRate = lfoRateTable[patchRam.lfoRate];  // FIXME move to parameters
@@ -83,15 +83,16 @@ void Module::runLFO() {
     }
     lfo = (lfoState & 0x02) ? -lfoPhase : lfoPhase;
 
-    pw = (lfoState & 0x02) ? lfoPhase + 0x2000 : 0x2000 - lfoPhase; // PW LFO is unipolar
-    pw = (patchRam.switch2 & 0x01) ?  0x3fff : pw;  // either LFO or "all on"
-    pw =  0x3fff - ((pw * patchRam.pwmLfo) >> 7);   // scaled by PWM pot
-    // lfo = (lfo * lfoDelay) >> 14;
+    pw = (lfoState & 0x02) ? lfoPhase + 0x2000 : 0x2000 - lfoPhase;  // PW LFO is unipolar
+    pw = (patchRam.switch2 & 0x01) ? 0x3fff : pw;                    // either LFO or "all on"
+    pw = 0x3fff - ((pw * patchRam.pwmLfo) >> 7);                     // scaled by PWM pot
 }
 
 void Module::run(Voice* voices, uint32_t blockSize) {
     // run updates for module board
 
+    int16_t lfoToVco = 0, lfoToVcf = 0;
+
     // FIXME break these out to the patch setter
     a = attackTable[patchRam.env_a];  // attack time coeff looked up in table
     d = decayTable[patchRam.env_d];   // decay time coeff looked up in table
@@ -116,10 +117,11 @@ void Module::run(Voice* voices, uint32_t blockSize) {
 
     runLFO();
 
-    float pwf = pw / 32768.0f;
+    // calculate "smoothed" parameters
+    // these are single outputs with heavy RC smoothing
     for (uint32_t i = 0; i < blockSize; i++) {
         vcaRC = (master - vcaRC) * subTC + vcaRC;
-        pwmRC = (pwf - pwmRC) * pwmTC + pwmRC;
+        pwmRC = ((pw / 32768.0f) - pwmRC) * pwmTC + pwmRC;
         subRC = (sub - subRC) * vcaTC + subRC;
 
         vcaBuf[i] = vcaRC;
@@ -129,13 +131,27 @@ void Module::run(Voice* voices, uint32_t blockSize) {
         if (bufPtr < bufferSize) bufPtr++;
     }
 
-    int16_t vcf = (patchRam.vcfEnv << 7) * ((patchRam.switch2 & 0x02) ? -1 : 1);
+    lfoToVco = (lfoDepthTable[patchRam.vcoLfo] * lfoDelay) >> 8;  // lookup table is 0-255
+    lfoToVco += /* lfo from modwheel FIXME */ 0;
+    if (lfoToVco > 0xff) lfoToVco = 0xff;
+    lfoToVco = (lfo * lfoToVco) >> 11;  // 8 for normalisation plus three additional DSLR EA
 
-    int16_t pitchBase = 0x1818;
-    pitchBase += (lfo * lfoDepthTable[patchRam.vcoLfo]) >> 11;
+    lfoToVcf = (patchRam.vcfLfo * lfoDelay) >> 7;  // value is 0-127
+    lfoToVcf = (lfo * lfoToVcf) >> 9;              // 8 for normalisation plus one additional DSLR EA
 
+    int16_t pitchBase = 0x1818, vcfBase = 0;
+    pitchBase += lfoToVco;
+    pitchBase += /* pitch bend FIXME */ 0;
+
+    // int16_t vcf = (patchRam.vcfEnv << 7) * ((patchRam.switch2 & 0x02) ? -1 : 1);
+    vcfBase = (patchRam.vcfFreq << 7) + /* vcf bend FIXME */ 0;
+    vcfBase += lfoToVcf;
+    if (vcfBase > 0x3fff) vcfBase = 0x3fff;
+    if (vcfBase < 0x0000) vcfBase = 0x0000;
+
+    // per-voice calculations
     for (uint32_t i = 0; i < NUM_VOICES; i++) {
-        // maybe move all this into voice.cpp FIXME
+        // run one step of the envelope
         Voice* v = &voices[i];
         switch (v->envPhase) {
             case 0:                           // release phase FIXME use an enum I guess
@@ -154,21 +170,20 @@ void Module::run(Voice* voices, uint32_t blockSize) {
         }
 
         // pitch
-        // FIXME clean this all up a bit
-        int16_t pitch = pitchBase + (v->note << 8);
-        int16_t semi = pitch >> 8;
+        uint16_t pitch = pitchBase + (v->note << 8);
+        uint8_t semi = pitch >> 8;
         float frac = (pitch & 0xff) / 256.0;
 
         float p1 = pitchTable[semi], p2 = pitchTable[semi + 1];
-        int16_t px = ((p2 - p1) * frac + p1);
+        int16_t px = ((p2 - p1) * frac + p1);  // interpolated pitch from table
 
+        // octave divider
         px *= (patchRam.switch1 & 0x07);
 
-        v->omega = px / (sampleRate * 8.0f);  // fixme use proper scaler
+        v->omega = px / (sampleRate * 8.0f);  // FIXME recalculate table using proper scaler
 
         // per voice we need to calculate the key follow amount and envelope amount
-        v->vcfCut = (patchRam.vcfFreq << 7) + ((vcf * v->env) >> 14);
-        v->vcfCut += (lfo * patchRam.vcfLfo) >> 9;
+        v->vcfCut = vcfBase + (((v->env * patchRam.vcfEnv)>>7) * ((patchRam.switch1 & 0x02) ? -1 : 1));
 
         v->vcfCut += (int)((v->note - 36) * (patchRam.vcfKey << 1) * 0.375);
 
diff --git a/plugin/module.hpp b/plugin/module.hpp
index 3f75a8c..2b6e1f8 100644
--- a/plugin/module.hpp
+++ b/plugin/module.hpp
@@ -156,11 +156,11 @@ class Voice {
     uint8_t pulseStage = 1;       // pulse wave phase
     float subosc = 1;             // sub oscillator flipflop output
 
-    uint8_t envPhase = 0;
-    int16_t env = 0;  // output amplitude
-    uint16_t vcfCut;
-    int16_t vcaEnv;
-    float vcaRC = 0, vcfRC = 0;
+    uint8_t envPhase = 0;   // current running state of envelope
+    int16_t env = 0; // calculated envelope amount
+    int16_t vcfCut; // calculated cutoff to filter
+    int16_t vcaEnv; // calculated level to VCA (env/gate)
+    float vcaRC = 0, vcfRC = 0; // RC circuit state values
 
     uint8_t note = 0;