1
0
mirror of https://github.com/f4exb/sdrangel.git synced 2026-02-14 05:33:39 -05:00
sdrangel/sdrbase/dsp/firfilterrrc.cpp
2026-02-12 23:23:27 +01:00

193 lines
6.7 KiB
C++

///////////////////////////////////////////////////////////////////////////////////
// Copyright (C) 2012 maintech GmbH, Otto-Hahn-Str. 15, 97204 Hoechberg, Germany //
// written by Christian Daniel //
// Copyright (C) 2015-2019, 2023 Edouard Griffiths, F4EXB <f4exb06@gmail.com> //
// Copyright (C) 2026 SDRangel contributors //
// //
// This program is free software; you can redistribute it and/or modify //
// it under the terms of the GNU General Public License as published by //
// the Free Software Foundation as version 3 of the License, or //
// (at your option) any later version. //
// //
// This program is distributed in the hope that it will be useful, //
// but WITHOUT ANY WARRANTY; without even the implied warranty of //
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the //
// GNU General Public License V3 for more details. //
// //
// You should have received a copy of the GNU General Public License //
// along with this program. If not, see <http://www.gnu.org/licenses/>. //
///////////////////////////////////////////////////////////////////////////////////
#include "dsp/firfilterrrc.h"
#include <algorithm>
#include <cmath>
FIRFilterRRC::FIRFilterRRC() :
m_ptr(0),
m_symbolRate(0.0f),
m_rolloff(0.0f),
m_samplesPerSymbol(0)
{
}
void FIRFilterRRC::create(float symbolRate, float rolloff, int symbolSpan, Normalization normalization)
{
m_symbolRate = symbolRate;
m_rolloff = rolloff;
// Clamp rolloff to valid range [0, 1]
if (m_rolloff < 0.0f) {
m_rolloff = 0.0f;
} else if (m_rolloff > 1.0f) {
m_rolloff = 1.0f;
}
// Calculate samples per symbol
m_samplesPerSymbol = static_cast<int>(std::round(1.0f / symbolRate));
// Calculate number of taps (always odd for symmetry)
int numTaps = symbolSpan * m_samplesPerSymbol + 1;
if ((numTaps & 1) == 0) {
numTaps++; // Ensure odd number of taps
}
// Allocate tap storage (only half + center due to symmetry)
int halfTaps = numTaps / 2 + 1;
m_taps.resize(halfTaps);
// Calculate filter taps
int center = numTaps / 2;
for (int i = 0; i < halfTaps; i++)
{
// Time offset from center in symbol periods
float t = static_cast<float>(i - center) / static_cast<float>(m_samplesPerSymbol);
m_taps[i] = computeRRCTap(t, m_rolloff);
}
// Apply normalization
if (normalization == Normalization::Energy)
{
// Normalize energy: sqrt(sum of squares) = 1
float sum = 0.0f;
for (int i = 0; i < halfTaps - 1; i++) {
sum += m_taps[i] * m_taps[i] * 2.0f; // Two symmetric taps
}
sum += m_taps[halfTaps - 1] * m_taps[halfTaps - 1]; // Center tap
float scale = std::sqrt(sum);
if (scale > 0.0f) {
for (int i = 0; i < halfTaps; i++) {
m_taps[i] /= scale;
}
}
}
else if (normalization == Normalization::Amplitude)
{
// Normalize amplitude: max output for bipolar sequence ≈ 1
float maxGain = 0.0f;
for (int offset = 0; offset < m_samplesPerSymbol; offset++)
{
float gain = 0.0f;
for (int i = offset; i < halfTaps - 1; i += m_samplesPerSymbol) {
gain += std::abs(m_taps[i]) * 2.0f; // Both sides
}
// Add center tap if aligned
if ((center - offset) % m_samplesPerSymbol == 0) {
gain += std::abs(m_taps[halfTaps - 1]);
}
if (gain > maxGain) {
maxGain = gain;
}
}
if (maxGain > 0.0f) {
for (int i = 0; i < halfTaps; i++) {
m_taps[i] /= maxGain;
}
}
}
else if (normalization == Normalization::Gain)
{
// Normalize for unity gain: sum of taps = 1
float sum = 0.0f;
for (int i = 0; i < halfTaps - 1; i++) {
sum += m_taps[i] * 2.0f; // Two symmetric taps
}
sum += m_taps[halfTaps - 1]; // Center tap
if (sum > 0.0f) {
for (int i = 0; i < halfTaps; i++) {
m_taps[i] /= sum;
}
}
}
// Initialize sample buffer
m_samples.resize(numTaps);
std::fill(m_samples.begin(), m_samples.end(), Complex(0.0f, 0.0f));
m_ptr = 0;
}
float FIRFilterRRC::computeRRCTap(float t, float rolloff) const
{
constexpr float Ts = 1.0f; // Symbol period (normalized)
const float beta = rolloff;
// Handle special case: t = 0
if (t == 0.0f) {
// h(0) = (1/T) * (1 + beta*(4/π - 1))
return (1.0f / Ts) * (1.0f + beta * (4.0f / M_PI - 1.0f));
}
// Check for zeros of denominator: 1 - (4*β*t/T)² = 0
// This occurs at t = ±T/(4*β)
const float denomCheck = 4.0f * beta * t / Ts;
if (std::abs(std::abs(denomCheck) - 1.0f) < 1e-6f && beta > 0.0f)
{
// h(±T/4β) = (β/(T*√2)) * [(1+2/π)*sin(π/4β) + (1-2/π)*cos(π/4β)]
const float arg = M_PI / (4.0f * beta);
return (beta / (Ts * std::sqrt(2.0f))) *
((1.0f + 2.0f / M_PI) * std::sin(arg) +
(1.0f - 2.0f / M_PI) * std::cos(arg));
}
// General case
const float numerator = std::sin(M_PI * t / Ts * (1.0f - beta)) +
4.0f * beta * t / Ts * std::cos(M_PI * t / Ts * (1.0f + beta));
const float denominator = M_PI * t / Ts * (1.0f - denomCheck * denomCheck);
return numerator / (denominator * Ts);
}
FIRFilterRRC::Complex FIRFilterRRC::filter(const Complex& input)
{
const int numTaps = static_cast<int>(m_samples.size());
const int halfTaps = static_cast<int>(m_taps.size()) - 1;
// Store input sample in circular buffer
m_samples[m_ptr] = input;
// Perform convolution using symmetry
Complex acc(0.0f, 0.0f);
int a = m_ptr;
int b = (a == numTaps - 1) ? 0 : a + 1;
// Process symmetric pairs
for (int i = 0; i < halfTaps; i++)
{
acc += (m_samples[a] + m_samples[b]) * m_taps[i];
a = (a == 0) ? numTaps - 1 : a - 1;
b = (b == numTaps - 1) ? 0 : b + 1;
}
// Add center tap
acc += m_samples[a] * m_taps[halfTaps];
// Advance circular buffer pointer
m_ptr = (m_ptr == static_cast<size_t>(numTaps) - 1) ? 0 : m_ptr + 1;
return acc;
}