Synthesizer progress

Implemented:
- LERP_BUFFER_2
- OnePole_LPF/HPF
- AWProcess
- AllpassProcess
This commit is contained in:
alexguirre 2022-03-16 21:25:56 +01:00
parent 8998b8c808
commit 8b188a13e5

View File

@ -432,8 +432,25 @@ namespace CodeWalker.Utils
a[i] = Lerp(a[i], min, max);
}
break;
//case Dat10Synth.Opcode.LERP_BUFFER_2:
// break;
case Dat10Synth.Opcode.LERP_BUFFER_2: // TODO: some better name for LERP_BUFFER_2
{
var t = GetScalar(param[1]);
float[] min, max;
if ((param[0].Value & 0xFF) == (param[2].Value & 0xFF))
{
min = GetBuffer(param[2]);
max = GetBuffer(param[3]);
}
else
{
min = GetBuffer(param[3]);
max = GetBuffer(param[2]);
}
for (int i = 0; i < BufferSize; i++)
min[i] = Lerp(t, min[i], max[i]);
}
break;
case Dat10Synth.Opcode.LERP_SCALAR:
{
var t = GetScalar(param[1]);
@ -567,18 +584,24 @@ namespace CodeWalker.Utils
GetScalar(param[5]),
ref StateBlocks[param[6].Value]);
break;
//case Dat10Synth.Opcode.OnePole_LPF_BUFFER_BUFFER:
// break;
//case Dat10Synth.Opcode.OnePole_LPF_BUFFER_SCALAR:
// break;
//case Dat10Synth.Opcode.OnePole_LPF_SCALAR_SCALAR:
// break;
//case Dat10Synth.Opcode.OnePole_HPF_BUFFER_BUFFER:
// break;
//case Dat10Synth.Opcode.OnePole_HPF_BUFFER_SCALAR:
// break;
//case Dat10Synth.Opcode.OnePole_HPF_SCALAR_SCALAR:
// break;
case Dat10Synth.Opcode.OnePole_LPF_BUFFER_BUFFER:
OnePoleLPF(GetBuffer(param[0]), GetBuffer(param[1]), ref StateBlocks[param[2].Value]);
break;
case Dat10Synth.Opcode.OnePole_LPF_BUFFER_SCALAR:
OnePoleLPF(GetBuffer(param[0]), GetScalar(param[1]), ref StateBlocks[param[2].Value]);
break;
case Dat10Synth.Opcode.OnePole_LPF_SCALAR_SCALAR:
SetRegister(param[0], OnePoleLPF(GetScalar(param[1]), GetScalar(param[2]), ref StateBlocks[param[3].Value]));
break;
case Dat10Synth.Opcode.OnePole_HPF_BUFFER_BUFFER:
OnePoleHPF(GetBuffer(param[0]), GetBuffer(param[1]), ref StateBlocks[param[2].Value]);
break;
case Dat10Synth.Opcode.OnePole_HPF_BUFFER_SCALAR:
OnePoleHPF(GetBuffer(param[0]), GetScalar(param[1]), ref StateBlocks[param[2].Value]);
break;
case Dat10Synth.Opcode.OnePole_HPF_SCALAR_SCALAR:
SetRegister(param[0], OnePoleHPF(GetScalar(param[1]), GetScalar(param[2]), ref StateBlocks[param[3].Value]));
break;
case Dat10Synth.Opcode.OSC_RAMP_BUFFER_BUFFER:
OscillatorRamp(GetBuffer(param[0]), ref StateBlocks[param[1].Value]);
break;
@ -816,8 +839,9 @@ namespace CodeWalker.Utils
for (int i = 0; i < BufferSize; i++)
a[i] = scalar;
break;
//case Dat10Synth.Opcode.AWProcess:
// break;
case Dat10Synth.Opcode.AWProcess:
AWFilter(GetBuffer(param[0]), GetBuffer(param[1]), GetBuffer(param[2]), ref StateBlocks[param[3].Value]);
break;
case Dat10Synth.Opcode.LERP_BUFFER_BUFFER:
{
var t = GetBuffer(param[0]);
@ -910,10 +934,12 @@ namespace CodeWalker.Utils
SetRegister(param[0], result);
}
break;
//case Dat10Synth.Opcode.AllpassProcess_BUFFER_SCALAR:
// break;
//case Dat10Synth.Opcode.AllpassProcess_BUFFER_BUFFER:
// break;
case Dat10Synth.Opcode.AllpassProcess_BUFFER_SCALAR:
AllpassFilter(GetBuffer(param[0]), GetScalar(param[1]), ref StateBlocks[param[2].Value]);
break;
case Dat10Synth.Opcode.AllpassProcess_BUFFER_BUFFER:
AllpassFilter(GetBuffer(param[0]), GetBuffer(param[1]), ref StateBlocks[param[2].Value]);
break;
case Dat10Synth.Opcode.FINISH:
frameFinished = true;
break;
@ -1007,32 +1033,19 @@ namespace CodeWalker.Utils
return (max - min) * t + min;
}
private float HardKnee(float y, float threshold)
private float HardKnee(float sample, float threshold)
{
/*
ENVELOPE_GEN__R_LINEAR_T_ONE_SHOT 0, 0.025, 0, 1, 0, 0, 1 => B0 [0]
COPY_BUFFER B0 => B1
HARD_KNEE_BUFFER B1, 0.25 => B1
COPY_BUFFER B0 => B2
HARD_KNEE_BUFFER B2, 0.85 => B2
COPY_BUFFER B0 => B4
HARD_KNEE_BUFFER B4, 0.75 => B4
FILL_BUFFER 0 => B3
FINISH =>
*/
// TODO(alexguirre): better names for HardKnee and maybe some comments
float result;
if (y < 0.0f)
if (sample < 0.0f)
result = 0.0f;
else
result = (y / threshold) * 0.5f;
result = (sample / threshold) * 0.5f;
if ((y - threshold) >= 0.0f)
result = (((y - threshold) / (1.0f - threshold)) + 1.0f) * 0.5f;
if (sample >= threshold)
result = (((sample - threshold) / (1.0f - threshold)) + 1.0f) * 0.5f;
if (y >= 1.0f)
if (sample >= 1.0f)
result = 1.0f;
return result;
@ -1048,10 +1061,11 @@ FINISH =>
return (float)Math.Pow(Math.Max(0, a), b);
}
private float NoteToFrequency(float note)
// https://newt.phys.unsw.edu.au/jw/notes.html
private float NoteToFrequency(float midiNote)
{
// TODO(alexguirre): figure out the meaning of this formula and constants in NoteToFrequency
return (float)Math.Pow(2.0f, (note + 36.376301f) / 12.0f);
// A4 (note #69) = 440Hz
return 440.0f * (float)Math.Pow(2.0f, (midiNote - 69) / 12.0f);
}
private void Decimate(float[] buffer, float scalar1, float deltaPerSample, ref StateBlock stateBlock)
@ -1220,7 +1234,7 @@ FINISH =>
float w = (float)(2.0 * Math.PI * centerFrequency / SampleRate);
float cosW = (float)Math.Cos(w);
float v14 = 1.0f / (float)Math.Tan((freq2 * 0.000020833333f) * Math.PI);
float v14 = 1.0f / (float)Math.Tan(Math.PI * freq2 / SampleRate);
b0 = 1.0f;
b1 = 0.0f;
@ -1243,7 +1257,7 @@ FINISH =>
float w = (float)(2.0f * Math.PI * centerFrequency / SampleRate);
float cosW = (float)Math.Cos(w);
float v15 = (float)Math.Tan((freq2 * 0.000020833333f) * Math.PI);
float v15 = (float)Math.Tan(Math.PI * freq2 / SampleRate);
b0 = 1.0f;
b1 = -2.0f * cosW;
@ -1268,7 +1282,7 @@ FINISH =>
float q = centerFrequency / freq2;
float alpha = (float)Math.Sin(w) / (2.0f * q);
float cosW = (float)Math.Cos(w);
float A = Math.Max(0.000001f, gain * 1.4141999f); // TODO(alexguirre): this should be amp = 10.0 ** (gain_db/40.0), where does 1.41 come from? maybe gain not in dB?
float A = Math.Max(0.000001f, gain * (float)Math.Sqrt(2)); // TODO(alexguirre): this should be amp = 10.0 ** (gain_db/40.0), where does sqrt(2) come from? maybe gain not in dB?
b0 = 1.0f + alpha * A;
b1 = -2.0f * cosW;
@ -1338,53 +1352,96 @@ FINISH =>
private void BiquadFilter2Pole(float[] buffer, float b0, float b1, float b2, float a1, float a2, ref StateBlock stateBlock)
{
// State block:
// X -> ...
// Y -> ...
float stateX = stateBlock.X;
float stateY = stateBlock.Y;
float pole1 = stateBlock.X;
float pole2 = stateBlock.Y;
for (int i = 0; i < BufferSize; i++)
{
float x = (buffer[i] * b0) + stateX;
stateX = (buffer[i] * b1) - (x * a1) + stateY;
stateY = (buffer[i] * b2) - (x * a2);
float s = buffer[i];
float x = (s * b0) + pole1;
pole1 = (s * b1) - (x * a1) + pole2;
pole2 = (s * b2) - (x * a2);
buffer[i] = x;
}
stateBlock.X = stateX;
stateBlock.Y = stateY;
stateBlock.X = pole1;
stateBlock.Y = pole2;
}
private void BiquadFilter4Pole(float[] buffer, float b0, float b1, float b2, float a1, float a2, ref StateBlock stateBlock)
{
// State block:
// X -> ...
// Y -> ...
// Z -> ...
// W -> ...
float stateX = stateBlock.X;
float stateY = stateBlock.Y;
float stateZ = stateBlock.Z;
float stateW = stateBlock.W;
float pole1 = stateBlock.X;
float pole2 = stateBlock.Y;
float pole3 = stateBlock.Z;
float pole4 = stateBlock.W;
for (int i = 0; i < BufferSize; i++)
{
float x = (buffer[i] * b0) + stateX;
float y = (x * b0) + stateZ;
stateX = (buffer[i] * b1) - (x * a1) + stateY;
stateY = (buffer[i] * b2) - (x * a2);
stateZ = (x * b1) - (y * a1) + stateW;
stateW = (x * b2) - (y * a2);
float s = buffer[i];
float x = (s * b0) + pole1;
pole1 = (s * b1) - (x * a1) + pole2;
pole2 = (s * b2) - (x * a2);
float y = (x * b0) + pole3;
pole3 = (x * b1) - (y * a1) + pole4;
pole4 = (x * b2) - (y * a2);
buffer[i] = y;
}
stateBlock.X = stateX;
stateBlock.Y = stateY;
stateBlock.Z = stateZ;
stateBlock.W = stateW;
stateBlock.X = pole1;
stateBlock.Y = pole2;
stateBlock.Z = pole3;
stateBlock.W = pole4;
}
// https://www.earlevel.com/main/2012/12/15/a-one-pole-filter/
// TODO: verify OnePoleLPF/HPF results
private void OnePoleLPF(float[] buffer, float[] frequencies, ref StateBlock stateBlock)
{
for (int i = 0; i < BufferSize; i++)
{
buffer[i] = OnePoleLPF(buffer[i], frequencies[i], ref stateBlock);
}
}
private void OnePoleLPF(float[] buffer, float frequency, ref StateBlock stateBlock)
{
for (int i = 0; i < BufferSize; i++)
{
buffer[i] = OnePoleLPF(buffer[i], frequency, ref stateBlock);
}
}
private float OnePoleLPF(float sample, float frequency, ref StateBlock stateBlock)
{
float previousSample = stateBlock.X;
float b1 = (float)Math.Exp(-2.0f * Math.PI * frequency * 256.0f / SampleRate);
float a0 = 1.0f - b1;
float s = a0 * sample + (b1 * previousSample);
stateBlock.X = s;
return s;
}
private void OnePoleHPF(float[] buffer, float[] frequencies, ref StateBlock stateBlock)
{
for (int i = 0; i < BufferSize; i++)
{
buffer[i] = OnePoleHPF(buffer[i], frequencies[i], ref stateBlock);
}
}
private void OnePoleHPF(float[] buffer, float frequency, ref StateBlock stateBlock)
{
for (int i = 0; i < BufferSize; i++)
{
buffer[i] = OnePoleHPF(buffer[i], frequency, ref stateBlock);
}
}
private float OnePoleHPF(float sample, float frequency, ref StateBlock stateBlock)
{
return sample - OnePoleLPF(sample, frequency, ref stateBlock);
}
private void EnvelopeFollower(float[] buffer, float a2, float a3, ref StateBlock stateBlock)
@ -1469,13 +1526,12 @@ FINISH =>
// State block:
// X -> latch SET
// TODO(alexguirre): understand logic of TriggerLatch
float stateSet = stateBlock.X;
bool set = stateBlock.X != 0.0f;
float latch = 0.0f;
if (triggered && stateSet == 0.0f)
if (triggered && !set)
latch = 1.0f;
if (triggered != (stateSet != 0.0f))
if (triggered != set)
{
stateBlock.X = triggered ? 1.0f : 0.0f;
}
@ -1811,21 +1867,6 @@ FINISH =>
// TODO(alexguirre): TimedTrigger may not be equivalent, game code has like 10 states, here I'm using the same states as the envelope gen
// TODO(alexguirre): verify how TimedTrigger works in-game
/*
ENVELOPE_GEN__R_LINEAR_T_ONE_SHOT 0.1, 0.2, 0.3, 0.5, 0.4, 0.35, 1 => B6, R0 [1]
TIMED_TRIGGER__T_ONE_SHOT 1, 0.1, 0.2, 0.3, 0.4, 0.35 => R1, R2, R3, R4, R5 [2]
;FILL_BUFFER R0 => B0 ; envelope finished
FILL_BUFFER R1 => B1 ; timed trigger finished
;FILL_BUFFER R2 => B2 ; attack
;FILL_BUFFER R3 => B3 ; decay
;FILL_BUFFER R4 => B4 ; hold
;FILL_BUFFER R5 => B5 ; release
FILL_BUFFER 0 => B7
FINISH =>
*/
private TimedTriggerResult TimedTrigger(EnvelopeTriggerMode triggerMode, ref StateBlock stateBlock, float trigger, float predelay, float attack, float decay, float hold, float release)
{
// State block:
@ -2070,6 +2111,65 @@ FINISH =>
}
}
private void AWFilter(float[] buffer, float[] buffer2, float[] buffer3, ref StateBlock stateBlock)
{
var s = stateBlock.X;
for (int i = 0; i < BufferSize; i++)
{
var v10 = buffer3[i] * 64.0f;
var v12 = (float)rnd.NextDouble() * v10 + buffer2[i] * 127.0f + 1.0f;
var v13 = 1.0f / v12;
var v14 = (1.0f - (v13 * v12)) * v13 + v13;
s = s - ((s - buffer[i]) * v14);
buffer[i] = s;
}
stateBlock.X = s;
}
// https://thewolfsound.com/allpass-filter/#first-order-iir-allpass
private void AllpassFilter(float[] buffer, float breakFrequency, ref StateBlock stateBlock)
{
float previousUnfilteredSample = stateBlock.X;
float previousSample = stateBlock.Y;
float a1 = (float)Math.Tan(Math.PI * breakFrequency / SampleRate);
a1 = (a1 - 1.0f) / (a1 + 1.0f);
for (int i = 0; i < BufferSize; i++)
{
float sample = a1 * buffer[i] + previousUnfilteredSample - a1 * previousSample;
sample = Math.Max(Math.Min(sample, 1.0f), -1.0f);
previousUnfilteredSample = buffer[i];
previousSample = sample;
buffer[i] = sample;
}
stateBlock.X = previousUnfilteredSample;
stateBlock.Y = previousSample;
}
private void AllpassFilter(float[] buffer, float[] breakFrequencies, ref StateBlock stateBlock)
{
float previousUnfilteredSample = stateBlock.X;
float previousSample = stateBlock.Y;
for (int i = 0; i < BufferSize; i++)
{
float a1 = (float)Math.Tan(Math.PI * breakFrequencies[i] / SampleRate);
a1 = (a1 - 1.0f) / (a1 + 1.0f);
float sample = a1 * buffer[i] + previousUnfilteredSample - a1 * previousSample;
sample = Math.Max(Math.Min(sample, 1.0f), -1.0f);
previousUnfilteredSample = buffer[i];
previousSample = sample;
buffer[i] = sample;
}
stateBlock.X = previousUnfilteredSample;
stateBlock.Y = previousSample;
}