More filtering added.

AX25Demodulator.cpp

@@ -30,9 +30,9 @@ const float32_t SAMPLES_PER_SYMBOL = SAMPLE_RATE / SYMBOL_RATE;
 const float32_t PLL_LIMIT          = SAMPLES_PER_SYMBOL / 2.0F;
 
 // XXX This is for the wrong sample rate
-const uint32_t LPF_FILTER_LEN = 96;
+const uint32_t LPF_FILTER_LEN = 96U;
 
-q15_t lpfCoeffs[] = {
+q15_t LPF_FILTER_COEFFS[] = {
     0,     1,     3,     5,     8,    11,    14,    17,    19,    20,    18,    14,
     7,    -2,   -16,   -33,   -53,   -76,  -101,  -126,  -151,  -174,  -194,  -208,
  -215,  -212,  -199,  -173,  -133,   -79,   -10,    74,   173,   287,   413,   549,
@@ -48,7 +48,9 @@ q15_t lpfCoeffs[] = {
 //      loop_coeffs = firwin(9, [64.0/(1200/2)], width = None,
 //          pass_zero = True, scale = True, window='hann')
 //
-float32_t pllLoopCoeffs[] = {3.196252e-02F, 1.204223e-01F, 2.176819e-01F, 2.598666e-01F, 2.176819e-01F, 1.204223e-01F, 3.196252e-02F};
+const uint32_t PLL_FILTER_LEN = 7U;
+
+float32_t PLL_FILTER_COEFFS[] = {3.196252e-02F, 1.204223e-01F, 2.176819e-01F, 2.598666e-01F, 2.176819e-01F, 1.204223e-01F, 3.196252e-02F};
 
 CAX25Demodulator::CAX25Demodulator(float32_t* coeffs, uint16_t length) :
 m_audioFilter(),
@@ -72,11 +74,11 @@ m_pllCount(0.0F)
 
   m_lpfFilter.numTaps = LPF_FILTER_LEN;
   m_lpfFilter.pState  = m_lpfState;
-  m_lpfFilter.pCoeffs = lpfCoeffs;
+  m_lpfFilter.pCoeffs = LPF_FILTER_COEFFS;
 
-  m_pllFilter.numTaps = 7U;
+  m_pllFilter.numTaps = PLL_FILTER_LEN;
   m_pllFilter.pState  = m_pllState;
-  m_pllFilter.pCoeffs = pllLoopCoeffs;
+  m_pllFilter.pCoeffs = PLL_FILTER_COEFFS;
 }
 
 bool CAX25Demodulator::process(const q15_t* samples, uint8_t length, CAX25Frame& frame)

AX25RX.cpp

@@ -20,19 +20,100 @@
 #include "Globals.h"
 #include "AX25RX.h"
 
+// XXX There are for the wrong sample rate
+
+// 1200Hz = -9dB, 2200Hz = 0dB; 3252Hz cutoff, 3.7 gain; cosine.
+float32_t dB9[] = {
+        -0.00232554104135,
+        -0.142858725752,
+        -0.449053780255,
+        -0.770264863826,
+        2.77651146344,
+        -0.770264863826,
+        -0.449053780255,
+        -0.142858725752,
+        -0.00232554104135};
+
+// 1200Hz = -6dB, 2200Hz = 0dB; 2640Hz cutoff, 2.59 gain; cosine.
+float32_t dB6[] = {
+        -0.0209448226653,
+        -0.130107651829,
+        -0.299004731072,
+        -0.45336946386,
+        2.0629448761,
+        -0.45336946386,
+        -0.299004731072,
+        -0.130107651829,
+        -0.0209448226653};
+
+// 1200Hz = -3dB, 2200Hz = 0dB; 1700Hz cutoff, 1.68 gain; cosine.
+float32_t dB3[] = {
+        -0.0231416146776,
+        -0.0833375337803,
+        -0.147937602401,
+        -0.197411259519,
+        1.46066084756,
+        -0.197411259519,
+        -0.147937602401,
+        -0.0833375337803,
+        -0.0231416146776};
+
+/*
+ * Generated with Scipy Filter, 152 coefficients, 1100-2300Hz bandpass,
+ * Hann window, starting and ending 0 value coefficients removed.
+ *
+ * np.array(
+ *  firwin2(152,
+ *      [
+ *          0.0,
+ *          1000.0/(sample_rate/2),
+ *          1100.0/(sample_rate/2),
+ *          2350.0/(sample_rate/2),
+ *          2500.0/(sample_rate/2),
+ *          1.0
+ *      ],
+ *      [0,0,1,1,0,0],
+ *      antisymmetric = False,
+ *      window='hann') * 32768,
+ *  dtype=int)[10:-10]
+ */
+const uint32_t FILTER_LEN = 132U;
+
+// XXX This is for the wrong sample rate
+q15_t FILTER_COEFFS[] = {
+    4,     0,    -5,   -10,   -13,   -12,    -9,    -4,    -2,    -4,   -12,   -26,
+  -41,   -52,   -51,   -35,    -3,    39,    83,   117,   131,   118,    83,    36,
+   -6,   -32,   -30,    -3,    36,    67,    66,    19,   -74,  -199,  -323,  -408,
+ -421,  -344,  -187,    17,   218,   364,   417,   369,   247,   106,    14,    26,
+  166,   407,   676,   865,   866,   605,    68,  -675, -1484, -2171, -2547, -2471,
+-1895,  -882,   394,  1692,  2747,  3337,  3337,  2747,  1692,   394,  -882, -1895,
+-2471, -2547, -2171, -1484,  -675,    68,   605,   866,   865,   676,   407,   166,
+   26,    14,   106,   247,   369,   417,   364,   218,    17,  -187,  -344,  -421,
+ -408,  -323,  -199,   -74,    19,    66,    67,    36,    -3,   -30,   -32,    -6,
+   36,    83,   118,   131,   117,    83,    39,    -3,   -35,   -51,   -52,   -41,
+  -26,   -12,    -4,    -2,    -4,    -9,   -12,   -13,   -10,    -5,     0,     4};
+
 CAX25RX::CAX25RX() :
-m_demod1(),
-m_demod2(),
-m_demod3(),
+m_filter(),
+m_state(),
+m_demod1(dB3, 9U),
+m_demod2(dB6, 9U),
+m_demod3(dB9, 9U),
 m_lastFCS(0U)
 {
+  m_filter.numTaps = FILTER_LEN;
+  m_filter.pState  = m_state;
+  m_filter.pCoeffs = FILTER_COEFFS;
 }
 
-void CAX25RX::samples(const q15_t* samples, uint8_t length)
+void CAX25RX::samples(q15_t* samples, uint8_t length)
 {
+  q15_t output[RX_BLOCK_SIZE];
+  ::arm_fir_fast_q15(&m_filter, samples, output, RX_BLOCK_SIZE);
+
   CAX25Frame frame;
 
-  bool ret = m_demod1.process(samples, length, frame);
+  bool ret = m_demod1.process(output, length, frame);
   if (ret) {
     if (m_lastFCS != frame.m_fcs) {
       DEBUG1("Decoder 1 reported");
@@ -41,7 +122,7 @@ void CAX25RX::samples(const q15_t* samples, uint8_t length)
     }
   }
 
-  ret = m_demod2.process(samples, length, frame);
+  ret = m_demod2.process(output, length, frame);
   if (ret) {
     if (m_lastFCS != frame.m_fcs) {
       DEBUG1("Decoder 2 reported");
@@ -50,7 +131,7 @@ void CAX25RX::samples(const q15_t* samples, uint8_t length)
     }
   }
 
-  ret = m_demod3.process(samples, length, frame);
+  ret = m_demod3.process(output, length, frame);
   if (ret) {
     if (m_lastFCS != frame.m_fcs) {
       DEBUG1("Decoder 3 reported");

AX25RX.h

@@ -27,13 +27,15 @@ class CAX25RX {
 public:
   CAX25RX();
 
-  void samples(const q15_t* samples, uint8_t length);
+  void samples(q15_t* samples, uint8_t length);
 
 private:
-  CAX25Demodulator m_demod1;
-  CAX25Demodulator m_demod2;
-  CAX25Demodulator m_demod3;
-  uint16_t         m_lastFCS;
+  arm_fir_instance_q15 m_filter;
+  q15_t                m_state[160U];    // NoTaps + BlockSize - 1, 132 + 20 - 1 plus some spare
+  CAX25Demodulator     m_demod1;
+  CAX25Demodulator     m_demod2;
+  CAX25Demodulator     m_demod3;
+  uint16_t             m_lastFCS;
 };
 
 #endif