Working with phasor re-rotation

ch32v/ch32v203-goertzel/adcgoertzel.c

@@ -81,6 +81,12 @@ volatile uint16_t adc_buffer[ADC_BUFFSIZE];
 
 int g_volume_pwm = 127; // 0 - 127 (100%) (but you can go over 100)
 
+int32_t g_goertzel_phasor_r = 32768;
+int32_t g_goertzel_phasor_i = 0;
+
+int32_t g_goertzel_advance_r = 32768;
+int32_t g_goertzel_advance_i = 0;
+
 #if 0
 int g_pwm_period = (30-1);
 int g_goertzel_buffer = (752);
@@ -390,6 +396,36 @@ void DMA1_Channel1_IRQHandler( void )
 				int32_t rr = (((int64_t)(g_goertzel_coefficient  ) * (int64_t)zp<<1)>>32) - (zp2);
 				int32_t ri = (((int64_t)(g_goertzel_coefficient_s) * (int64_t)zp<<1)>>32);
 
+				// Advanced the current goertzel advance 
+				// phasor = phasor * advance;
+				//  real = real * real - imag * imag;
+				//  imag = real * imag + real * imag;
+				// Sometimes you would bias the output here so that when truncating down you don't perpetually decay.
+				// But experimentally, it didn't make a difference.
+				int32_t temp = (g_goertzel_phasor_r * g_goertzel_advance_i + g_goertzel_phasor_i * g_goertzel_advance_r) >> 15;
+				g_goertzel_phasor_r = (g_goertzel_phasor_r * g_goertzel_advance_r - g_goertzel_phasor_i * g_goertzel_advance_i) >> 15;
+				g_goertzel_phasor_i = temp;
+
+				// Fixup phasor over time to prevent it from dacaying.
+				#define ABS(x) (((x)<0)?-(x):(x))
+				int s_phasor = g_goertzel_phasor_r * g_goertzel_phasor_r + g_goertzel_phasor_i * g_goertzel_phasor_i;
+				int intensity_phasor = (ABS(g_goertzel_phasor_r) + ABS(g_goertzel_phasor_i)) * 26100 / 32768 + 1; // Found experimentally (Also try to avoid divide-by-zero.
+				intensity_phasor = (intensity_phasor + s_phasor/intensity_phasor)/2;
+				intensity_phasor = (intensity_phasor + s_phasor/intensity_phasor)/2;
+				if( intensity_phasor < 32760 )
+				{
+					// It is decaying, this is equivelent to f = f * 1.000244141
+					g_goertzel_phasor_r += g_goertzel_phasor_r >> 12;
+					g_goertzel_phasor_i += g_goertzel_phasor_i >> 12;
+				}
+
+				// Now, rotate rr, ri by that phasor.
+				temp = (g_goertzel_phasor_r * ri + g_goertzel_phasor_i * rr) >> 15;
+				rr = (g_goertzel_phasor_r * rr - g_goertzel_phasor_i * ri) >> 15;
+				ri = temp;
+
+				// rr, ri are now in the correct frame of reference.  Continue computing.
+
 				qibaselogs[qibaselogs_head] = ((uint16_t)rr) | (((uint16_t)ri)<<16);
 				qibaselogs_head = ( qibaselogs_head + 1 ) & ((LOG_GOERTZEL_LIST)-1);
 
@@ -400,8 +436,9 @@ void DMA1_Channel1_IRQHandler( void )
 				//int intensity = 1<<( ( 32 - __builtin_clz(s) )/2);
 				#define ABS(x) (((x)<0)?-(x):(x))
 				int intensity = (ABS(rr) + ABS(ri)) * 26100 / 32768; // Found experimentally (Also try to avoid divide-by-zero.
-				if( intensity == 0 )
+				//if( intensity == 0 )
-					intensity = 1;
+				//	intensity = 1;
+				intensity++;
 				intensity = (intensity + s/intensity)/2;
 				intensity = (intensity + s/intensity)/2;
 				intensity_average = intensity_average - (intensity_average>>12) + (intensity>>6);
@@ -714,6 +751,8 @@ void HandleHidUserReportOutComplete( struct _USBState * ctx )
 		if( numconfigs > 3) g_goertzel_coefficient = configs[5];
 		if( numconfigs > 4) g_goertzel_coefficient_s = configs[6];
 		if( numconfigs > 5) g_exactcompute = configs[7];
+		if( numconfigs > 6) g_goertzel_advance_r = configs[8];
+		if( numconfigs > 7) g_goertzel_advance_i = configs[9];
 
 		// Need to reset so we don't blast by.
 		g_goertzel_samples = 0;

ch32v/ch32v203-goertzel/ttest/Makefile

@@ -1,4 +1,10 @@
-all : test floattest
+all : test floattest phasor_rotation_test
+
+phasor_rotation_testt : phasor_rotation_test.c
+	gcc -o $@ $^ -lm -g
+
+phasor_rotation_test : phasor_rotation_testt
+	./phasor_rotation_testt
 
 floattest : floattestt
 	./floattestt

ch32v/ch32v203-goertzel/ttest/floattestt.c

@@ -0,0 +1,44 @@
+#include <stdio.h>
+#include <math.h>
+
+int main()
+{
+	const float omegaPerSample = 3.1415926*2.0 / 128; // pi / 200
+
+	// Only divide by 128 to show two cycles.
+
+	const int numSamples = 256; // enough to go from 0 to 2pi
+
+	float phase = 0;
+	//for( float phase = 0; phase < 3.1415926*2.0; phase += 0.01 )
+	{
+		float coeff = 2 * cos( omegaPerSample );
+		int i;
+
+		// TRICKY: When you want a sinewave, initialize with omegaPerSample.  This
+		// is crucial.  The initial state will have massive consequences.
+		float sprev = omegaPerSample;
+		float sprev2 = 0;
+		for( i = 0; i < numSamples; i++ )
+		{
+			// If you wanted to do a DFT, set SAMPLE to your incoming sample.
+			float SAMPLE = 65536 * sin( phase + i * omegaPerSample );
+
+			// Here is where the magic happens.
+			float s = SAMPLE + coeff * sprev - sprev2;
+			sprev2 = sprev;
+			sprev = s;
+
+		// For DFT, your power will be:
+		float power = sprev*sprev + sprev2*sprev2 - (coeff * sprev * sprev2);
+		//printf( "Power: %f\n", power );
+
+		float coeff_s = 2 * sin( omegaPerSample );
+		double rR = 0.5 * coeff   * sprev - sprev2;
+		double rI = 0.5 * coeff_s * sprev;
+		printf( "%d,%f,%f\n", i, rR, rI);
+
+		}
+
+	}
+}

ch32v/ch32v203-goertzel/ttest/phasor_rotation_test.c

@@ -0,0 +1,43 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <math.h>
+
+int main()
+{
+	int phasor_r = 32768;
+	int phasor_i = 0;
+	double phasor = 0;
+
+	double omega = 0.1;
+	int omega_r = cos( omega ) * 32768;
+	int omega_i = sin( omega ) * 32768;
+
+	int i;
+	for( i = 0; i < 10000; i++ )
+	{
+		
+		int32_t temp = (phasor_r * omega_i + phasor_i * omega_r ) >> 15;
+		phasor_r = (phasor_r * omega_r - phasor_i * omega_i ) >> 15;
+		phasor_i = temp;
+		phasor += omega;
+
+
+		// Approximate sqrt(x*x+y*y)
+		#define ABS(x) (((x)<0)?-(x):(x))
+		int s = phasor_r * phasor_r + phasor_i * phasor_i;
+		int intensity = (ABS(phasor_r) + ABS(phasor_i)) * 26100 / 32768 + 1; // Found experimentally (Also try to avoid divide-by-zero.
+		intensity = (intensity + s/intensity)/2;
+		intensity = (intensity + s/intensity)/2;
+
+		if( intensity < 32763 )
+		{
+			phasor_r += phasor_r >> 12;
+			phasor_i += phasor_i >> 12;
+		}
+
+		double fA = atan2( phasor_i, phasor_r );
+		printf( "%6d %6d / %6d %6d / %d / %f %f %f\n", omega_r, omega_i, phasor_r, phasor_i, intensity, fA, phasor, fA-phasor );
+		if( phasor >= 3.141592653589 ) phasor -= 3.141592653589*2;
+	}
+
+}

ch32v/ch32v203tx/ch32v203tx.c

@@ -84,10 +84,10 @@ int main()
 		TIM2->CH1CVR = 2;
 		TIM2->CCER = TIM_CC1E | TIM_CC1P;
 		funDigitalWrite( LEDPIN, 1 );
-		Delay_Us( 2000 );
+		Delay_Us( 20000 );
 		TIM2->CCER = TIM_CC1E;
 		TIM2->CH1CVR = 2;
 		funDigitalWrite( LEDPIN, 0 );
-		Delay_Us( 2000 );
+		Delay_Us( 20000 );
 	}
 }

ch32v/lib/calculator.html

@@ -24,12 +24,20 @@ function DrawSpan( rowspan, colspan, freq, target, docolor, extrastr = "" )
 
 function Goertz( n, mhz, fr, brf, exact_compute )
 {
-	let omega = fr * 3.1415926535*2.0;
+	let tau = 3.1415926535*2.0;
+	let omega = fr * tau;
 	var textarea = document.getElementById("goertzeloutput");
 
-	var g_goertzel_omega_per_sample = Math.round( ( omega*2*(1<<29)) );
+	var goertzel_omega_per_sample_real = ( omega*2*(1<<29));
+	var g_goertzel_omega_per_sample = Math.round( goertzel_omega_per_sample_real );
 	var g_goertzel_coefficient = Math.round( (2 * Math.cos( omega ) * (1<<30)) );
 	var g_goertzel_coefficient_s = Math.round( (2 * Math.sin( omega ) * (1<<30)) );
+
+	var omega_per_group = omega * brf;
+	var goertzel_phasor_advance_radians_per_sample = tau * (Math.round( omega_per_group / tau ) - omega_per_group / tau); 
+	var g_goertzel_advance_r = Math.cos( goertzel_phasor_advance_radians_per_sample ) * 32768;
+	var g_goertzel_advance_i = Math.sin( goertzel_phasor_advance_radians_per_sample ) * 32768;
+
 	var g_exactcompute = exact_compute;
 	textarea.value = 
 		"int g_pwm_period = ("+n+"-1);\n" +
@@ -37,7 +45,9 @@ function Goertz( n, mhz, fr, brf, exact_compute )
 		"int g_goertzel_buffer = ("+brf+");\n" +
 		"int32_t g_goertzel_omega_per_sample = " + g_goertzel_coefficient.toFixed(0) + "; // " + ( omega / (3.1415926535*2.0)).toFixed(6) + " of whole per step / " + mhz.toFixed(6) + "MHz\n" +
 		"int32_t g_goertzel_coefficient = " + g_goertzel_coefficient.toFixed(0) + ";\n" +
-		"int32_t g_goertzel_coefficient_s = " + g_goertzel_coefficient_s.toFixed(0) + ";\n";
+		"int32_t g_goertzel_coefficient_s = " + g_goertzel_coefficient_s.toFixed(0) + ";\n" +
+		"int32_t g_goertzel_advance_r = " + g_goertzel_advance_r.toFixed(0) + ";\n" +
+		"int32_t g_goertzel_advance_i = " + g_goertzel_advance_i.toFixed(0) + ";\n";
 
 	// Highlight its content
 	textarea.select();
@@ -45,7 +55,16 @@ function Goertz( n, mhz, fr, brf, exact_compute )
 	// Copy the highlighted text
 	document.execCommand("copy");
 
-	updateWebHidDeviceWithParameters( [ (n-1)|0, brf|0, g_goertzel_omega_per_sample|0, g_goertzel_coefficient|0, g_goertzel_coefficient_s|0, exact_compute|0 ] );
+	updateWebHidDeviceWithParameters( [ 
+		(n-1)|0,
+		brf|0,
+		g_goertzel_omega_per_sample|0,
+		g_goertzel_coefficient|0,
+		g_goertzel_coefficient_s|0,
+		exact_compute|0,
+		g_goertzel_advance_r|0,
+		g_goertzel_advance_i|0,
+		 ] );
 }
 
 function computeTable()
@@ -88,7 +107,7 @@ function computeTable()
 			"<TABLE BORDER=1>" +
 			"<TR><TD>Goertzel</TD></TR>" +
 			"<TR><TD>Goertzel (Inverse)</TD></TR>" +
-			"</TABLE><TEXTAREA ROWS=6 COLS=120 ID=goertzeloutput></TEXTAREA>" +
+			"</TABLE><TEXTAREA ROWS=8 COLS=120 ID=goertzeloutput></TEXTAREA>" +
 			"<P>Click on a ordinal offset to create the C code needed for that tuning parameter. Clicking will copy-to-clipboard.</P>" +
 			"<P>N Divisor #30 (row 3) is usually pretty good.  And, try to select things near 0.25 / 0.75, and avoid 0.0, 0.5, and 1.0.</P>" +
 			"<P>Goertzel's mode is for the ch32v203</P>";