portapack-mayhem/firmware/baseband/proc_audiotx.cpp

124 lines
3.5 KiB
C++
Raw Normal View History

/*
* Copyright (C) 2015 Jared Boone, ShareBrained Technology, Inc.
* Copyright (C) 2016 Furrtek
*
* This file is part of PortaPack.
*
* 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; either version 2, 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 for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; see the file COPYING. If not, write to
* the Free Software Foundation, Inc., 51 Franklin Street,
* Boston, MA 02110-1301, USA.
*/
#include "proc_audiotx.hpp"
#include "portapack_shared_memory.hpp"
#include "sine_table_int8.hpp"
//#include "audio_output.hpp"
#include "event_m4.hpp"
#include <cstdint>
void AudioTXProcessor::execute(const buffer_c8_t& buffer){
if (!configured) return;
if( stream ) {
const size_t bytes_to_read = (buffer.count / 32); // /32 (oversampling) should be == 64
bytes_read += stream->read(audio_buffer.p, bytes_to_read);
}
// Fill and "stretch"
for (size_t i = 0; i < buffer.count; i++) {
if (!(i & 31))
audio_sample = audio_buffer.p[i >> 5] - 0x80;
sample = tone_gen.process(audio_sample);
2016-09-23 21:08:54 +00:00
// FM
2017-11-10 00:25:04 +00:00
delta = sample * fm_delta;
phase += delta;
sphase = phase + (64 << 24);
re = (sine_table_i8[(sphase & 0xFF000000U) >> 24]);
im = (sine_table_i8[(phase & 0xFF000000U) >> 24]);
buffer.p[i] = { (int8_t)re, (int8_t)im };
}
spectrum_samples += buffer.count;
if( spectrum_samples >= spectrum_interval_samples ) {
spectrum_samples -= spectrum_interval_samples;
txprogress_message.progress = bytes_read; // Inform UI about progress
txprogress_message.done = false;
shared_memory.application_queue.push(txprogress_message);
}
//AudioOutput::fill_audio_buffer(preview_audio_buffer, true);
}
void AudioTXProcessor::on_message(const Message* const message) {
switch(message->id) {
case Message::ID::AudioTXConfig:
audio_config(*reinterpret_cast<const AudioTXConfigMessage*>(message));
break;
case Message::ID::ReplayConfig:
configured = false;
bytes_read = 0;
replay_config(*reinterpret_cast<const ReplayConfigMessage*>(message));
break;
case Message::ID::SamplerateConfig:
samplerate_config(*reinterpret_cast<const SamplerateConfigMessage*>(message));
break;
case Message::ID::FIFOData:
configured = true;
break;
default:
break;
}
}
void AudioTXProcessor::audio_config(const AudioTXConfigMessage& message) {
fm_delta = message.deviation_hz * (0xFFFFFFULL / baseband_fs);
tone_gen.configure(message.tone_key_delta, message.tone_key_mix_weight);
}
void AudioTXProcessor::samplerate_config(const SamplerateConfigMessage& message) {
baseband_fs = message.sample_rate;
baseband_thread.set_sampling_rate(baseband_fs);
spectrum_interval_samples = baseband_fs / 20;
}
void AudioTXProcessor::replay_config(const ReplayConfigMessage& message) {
if( message.config ) {
stream = std::make_unique<StreamOutput>(message.config);
// Tell application that the buffers and FIFO pointers are ready, prefill
shared_memory.application_queue.push(sig_message);
} else {
stream.reset();
}
}
int main() {
EventDispatcher event_dispatcher { std::make_unique<AudioTXProcessor>() };
event_dispatcher.run();
return 0;
}