mirror of
https://github.com/f4exb/sdrangel.git
synced 2024-11-05 00:11:16 -05:00
405 lines
13 KiB
C++
405 lines
13 KiB
C++
///////////////////////////////////////////////////////////////////////////////////
|
|
// Copyright (C) 2018-2020 Edouard Griffiths, F4EXB <f4exb06@gmail.com> //
|
|
// //
|
|
// 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/samplesinkfifo.h"
|
|
|
|
#include "bladerf2inputthread.h"
|
|
|
|
BladeRF2InputThread::BladeRF2InputThread(struct bladerf* dev, unsigned int nbRxChannels, QObject* parent) :
|
|
QThread(parent),
|
|
m_running(false),
|
|
m_dev(dev),
|
|
m_nbChannels(nbRxChannels),
|
|
m_iqOrder(true)
|
|
{
|
|
qDebug("BladeRF2InputThread::BladeRF2InputThread");
|
|
m_channels = new Channel[nbRxChannels];
|
|
|
|
for (unsigned int i = 0; i < nbRxChannels; i++) {
|
|
m_channels[i].m_convertBuffer.resize(DeviceBladeRF2::blockSize, Sample{0,0});
|
|
}
|
|
|
|
m_buf = new qint16[2*DeviceBladeRF2::blockSize*nbRxChannels];
|
|
}
|
|
|
|
BladeRF2InputThread::~BladeRF2InputThread()
|
|
{
|
|
qDebug("BladeRF2InputThread::~BladeRF2InputThread");
|
|
|
|
if (m_running) {
|
|
stopWork();
|
|
}
|
|
|
|
delete[] m_buf;
|
|
delete[] m_channels;
|
|
}
|
|
|
|
void BladeRF2InputThread::startWork()
|
|
{
|
|
m_startWaitMutex.lock();
|
|
start();
|
|
|
|
while(!m_running) {
|
|
m_startWaiter.wait(&m_startWaitMutex, 100);
|
|
}
|
|
|
|
m_startWaitMutex.unlock();
|
|
}
|
|
|
|
void BladeRF2InputThread::stopWork()
|
|
{
|
|
m_running = false;
|
|
wait();
|
|
}
|
|
|
|
void BladeRF2InputThread::run()
|
|
{
|
|
int res;
|
|
|
|
m_running = true;
|
|
m_startWaiter.wakeAll();
|
|
|
|
unsigned int nbFifos = getNbFifos();
|
|
|
|
if ((m_nbChannels > 0) && (nbFifos > 0))
|
|
{
|
|
int status;
|
|
|
|
if (m_nbChannels > 1) {
|
|
status = bladerf_sync_config(m_dev, BLADERF_RX_X2, BLADERF_FORMAT_SC16_Q11, 64, 8192, 32, 10000);
|
|
} else {
|
|
status = bladerf_sync_config(m_dev, BLADERF_RX_X1, BLADERF_FORMAT_SC16_Q11, 64, 8192, 32, 10000);
|
|
}
|
|
|
|
if (status < 0)
|
|
{
|
|
qCritical("BladeRF2InputThread::run: cannot configure streams: %s", bladerf_strerror(status));
|
|
}
|
|
else
|
|
{
|
|
qDebug("BladeRF2InputThread::run: start running loop");
|
|
while (m_running)
|
|
{
|
|
if (m_nbChannels > 1) {
|
|
res = bladerf_sync_rx(m_dev, m_buf, DeviceBladeRF2::blockSize*m_nbChannels, NULL, 10000);
|
|
} else {
|
|
res = bladerf_sync_rx(m_dev, m_buf, DeviceBladeRF2::blockSize, NULL, 10000);
|
|
}
|
|
|
|
if (res < 0)
|
|
{
|
|
qCritical("BladeRF2InputThread::run sync Rx error: %s", bladerf_strerror(res));
|
|
break;
|
|
}
|
|
|
|
if (m_nbChannels > 1)
|
|
{
|
|
callbackMI(m_buf, DeviceBladeRF2::blockSize);
|
|
}
|
|
else
|
|
{
|
|
if (m_iqOrder) {
|
|
callbackSIIQ(m_buf, 2*DeviceBladeRF2::blockSize);
|
|
} else {
|
|
callbackSIQI(m_buf, 2*DeviceBladeRF2::blockSize);
|
|
}
|
|
}
|
|
}
|
|
qDebug("BladeRF2InputThread::run: stop running loop");
|
|
}
|
|
}
|
|
else
|
|
{
|
|
qWarning("BladeRF2InputThread::run: no channels or FIFO allocated. Aborting");
|
|
}
|
|
|
|
|
|
m_running = false;
|
|
}
|
|
|
|
unsigned int BladeRF2InputThread::getNbFifos()
|
|
{
|
|
unsigned int fifoCount = 0;
|
|
|
|
for (unsigned int i = 0; i < m_nbChannels; i++)
|
|
{
|
|
if (m_channels[i].m_sampleFifo) {
|
|
fifoCount++;
|
|
}
|
|
}
|
|
|
|
return fifoCount;
|
|
}
|
|
|
|
void BladeRF2InputThread::setLog2Decimation(unsigned int channel, unsigned int log2_decim)
|
|
{
|
|
if (channel < m_nbChannels) {
|
|
m_channels[channel].m_log2Decim = log2_decim;
|
|
}
|
|
}
|
|
|
|
unsigned int BladeRF2InputThread::getLog2Decimation(unsigned int channel) const
|
|
{
|
|
if (channel < m_nbChannels) {
|
|
return m_channels[channel].m_log2Decim;
|
|
} else {
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
void BladeRF2InputThread::setFcPos(unsigned int channel, int fcPos)
|
|
{
|
|
if (channel < m_nbChannels) {
|
|
m_channels[channel].m_fcPos = fcPos;
|
|
}
|
|
}
|
|
|
|
int BladeRF2InputThread::getFcPos(unsigned int channel) const
|
|
{
|
|
if (channel < m_nbChannels) {
|
|
return m_channels[channel].m_fcPos;
|
|
} else {
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
void BladeRF2InputThread::setFifo(unsigned int channel, SampleSinkFifo *sampleFifo)
|
|
{
|
|
if (channel < m_nbChannels) {
|
|
m_channels[channel].m_sampleFifo = sampleFifo;
|
|
}
|
|
}
|
|
|
|
SampleSinkFifo *BladeRF2InputThread::getFifo(unsigned int channel)
|
|
{
|
|
if (channel < m_nbChannels) {
|
|
return m_channels[channel].m_sampleFifo;
|
|
} else {
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
void BladeRF2InputThread::callbackMI(const qint16* buf, qint32 samplesPerChannel)
|
|
{
|
|
// TODO: write a set of decimators that can take interleaved samples in input directly
|
|
int status = bladerf_deinterleave_stream_buffer(BLADERF_RX_X2, BLADERF_FORMAT_SC16_Q11 , samplesPerChannel*m_nbChannels, (void *) buf);
|
|
|
|
if (status < 0)
|
|
{
|
|
qCritical("BladeRF2InputThread::callbackMI: cannot de-interleave buffer: %s", bladerf_strerror(status));
|
|
return;
|
|
}
|
|
|
|
for (unsigned int channel = 0; channel < m_nbChannels; channel++)
|
|
{
|
|
if (m_channels[channel].m_sampleFifo)
|
|
{
|
|
if (m_iqOrder) {
|
|
callbackSIIQ(&buf[2*samplesPerChannel*channel], 2*samplesPerChannel, channel);
|
|
} else {
|
|
callbackSIQI(&buf[2*samplesPerChannel*channel], 2*samplesPerChannel, channel);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
void BladeRF2InputThread::callbackSIIQ(const qint16* buf, qint32 len, unsigned int channel)
|
|
{
|
|
SampleVector::iterator it = m_channels[channel].m_convertBuffer.begin();
|
|
|
|
if (m_channels[channel].m_log2Decim == 0)
|
|
{
|
|
m_channels[channel].m_decimatorsIQ.decimate1(&it, buf, len);
|
|
}
|
|
else
|
|
{
|
|
if (m_channels[channel].m_fcPos == 0) // Infra
|
|
{
|
|
switch (m_channels[channel].m_log2Decim)
|
|
{
|
|
case 1:
|
|
m_channels[channel].m_decimatorsIQ.decimate2_inf(&it, buf, len);
|
|
break;
|
|
case 2:
|
|
m_channels[channel].m_decimatorsIQ.decimate4_inf(&it, buf, len);
|
|
break;
|
|
case 3:
|
|
m_channels[channel].m_decimatorsIQ.decimate8_inf(&it, buf, len);
|
|
break;
|
|
case 4:
|
|
m_channels[channel].m_decimatorsIQ.decimate16_inf(&it, buf, len);
|
|
break;
|
|
case 5:
|
|
m_channels[channel].m_decimatorsIQ.decimate32_inf(&it, buf, len);
|
|
break;
|
|
case 6:
|
|
m_channels[channel].m_decimatorsIQ.decimate64_inf(&it, buf, len);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
else if (m_channels[channel].m_fcPos == 1) // Supra
|
|
{
|
|
switch (m_channels[channel].m_log2Decim)
|
|
{
|
|
case 1:
|
|
m_channels[channel].m_decimatorsIQ.decimate2_sup(&it, buf, len);
|
|
break;
|
|
case 2:
|
|
m_channels[channel].m_decimatorsIQ.decimate4_sup(&it, buf, len);
|
|
break;
|
|
case 3:
|
|
m_channels[channel].m_decimatorsIQ.decimate8_sup(&it, buf, len);
|
|
break;
|
|
case 4:
|
|
m_channels[channel].m_decimatorsIQ.decimate16_sup(&it, buf, len);
|
|
break;
|
|
case 5:
|
|
m_channels[channel].m_decimatorsIQ.decimate32_sup(&it, buf, len);
|
|
break;
|
|
case 6:
|
|
m_channels[channel].m_decimatorsIQ.decimate64_sup(&it, buf, len);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
else if (m_channels[channel].m_fcPos == 2) // Center
|
|
{
|
|
switch (m_channels[channel].m_log2Decim)
|
|
{
|
|
case 1:
|
|
m_channels[channel].m_decimatorsIQ.decimate2_cen(&it, buf, len);
|
|
break;
|
|
case 2:
|
|
m_channels[channel].m_decimatorsIQ.decimate4_cen(&it, buf, len);
|
|
break;
|
|
case 3:
|
|
m_channels[channel].m_decimatorsIQ.decimate8_cen(&it, buf, len);
|
|
break;
|
|
case 4:
|
|
m_channels[channel].m_decimatorsIQ.decimate16_cen(&it, buf, len);
|
|
break;
|
|
case 5:
|
|
m_channels[channel].m_decimatorsIQ.decimate32_cen(&it, buf, len);
|
|
break;
|
|
case 6:
|
|
m_channels[channel].m_decimatorsIQ.decimate64_cen(&it, buf, len);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
m_channels[channel].m_sampleFifo->write(m_channels[channel].m_convertBuffer.begin(), it);
|
|
}
|
|
|
|
void BladeRF2InputThread::callbackSIQI(const qint16* buf, qint32 len, unsigned int channel)
|
|
{
|
|
SampleVector::iterator it = m_channels[channel].m_convertBuffer.begin();
|
|
|
|
if (m_channels[channel].m_log2Decim == 0)
|
|
{
|
|
m_channels[channel].m_decimatorsQI.decimate1(&it, buf, len);
|
|
}
|
|
else
|
|
{
|
|
if (m_channels[channel].m_fcPos == 0) // Infra
|
|
{
|
|
switch (m_channels[channel].m_log2Decim)
|
|
{
|
|
case 1:
|
|
m_channels[channel].m_decimatorsQI.decimate2_inf(&it, buf, len);
|
|
break;
|
|
case 2:
|
|
m_channels[channel].m_decimatorsQI.decimate4_inf(&it, buf, len);
|
|
break;
|
|
case 3:
|
|
m_channels[channel].m_decimatorsQI.decimate8_inf(&it, buf, len);
|
|
break;
|
|
case 4:
|
|
m_channels[channel].m_decimatorsQI.decimate16_inf(&it, buf, len);
|
|
break;
|
|
case 5:
|
|
m_channels[channel].m_decimatorsQI.decimate32_inf(&it, buf, len);
|
|
break;
|
|
case 6:
|
|
m_channels[channel].m_decimatorsQI.decimate64_inf(&it, buf, len);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
else if (m_channels[channel].m_fcPos == 1) // Supra
|
|
{
|
|
switch (m_channels[channel].m_log2Decim)
|
|
{
|
|
case 1:
|
|
m_channels[channel].m_decimatorsQI.decimate2_sup(&it, buf, len);
|
|
break;
|
|
case 2:
|
|
m_channels[channel].m_decimatorsQI.decimate4_sup(&it, buf, len);
|
|
break;
|
|
case 3:
|
|
m_channels[channel].m_decimatorsQI.decimate8_sup(&it, buf, len);
|
|
break;
|
|
case 4:
|
|
m_channels[channel].m_decimatorsQI.decimate16_sup(&it, buf, len);
|
|
break;
|
|
case 5:
|
|
m_channels[channel].m_decimatorsQI.decimate32_sup(&it, buf, len);
|
|
break;
|
|
case 6:
|
|
m_channels[channel].m_decimatorsQI.decimate64_sup(&it, buf, len);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
else if (m_channels[channel].m_fcPos == 2) // Center
|
|
{
|
|
switch (m_channels[channel].m_log2Decim)
|
|
{
|
|
case 1:
|
|
m_channels[channel].m_decimatorsQI.decimate2_cen(&it, buf, len);
|
|
break;
|
|
case 2:
|
|
m_channels[channel].m_decimatorsQI.decimate4_cen(&it, buf, len);
|
|
break;
|
|
case 3:
|
|
m_channels[channel].m_decimatorsQI.decimate8_cen(&it, buf, len);
|
|
break;
|
|
case 4:
|
|
m_channels[channel].m_decimatorsQI.decimate16_cen(&it, buf, len);
|
|
break;
|
|
case 5:
|
|
m_channels[channel].m_decimatorsQI.decimate32_cen(&it, buf, len);
|
|
break;
|
|
case 6:
|
|
m_channels[channel].m_decimatorsQI.decimate64_cen(&it, buf, len);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
m_channels[channel].m_sampleFifo->write(m_channels[channel].m_convertBuffer.begin(), it);
|
|
}
|