SoapySDRThread: more comments about the relation between MTU and sample rate

This commit is contained in:
vsonnier 2017-05-21 09:20:41 +02:00
parent c1fef72103
commit fc2716518d

View File

@ -9,6 +9,7 @@
#include <algorithm>
#include <SoapySDR/Logger.h>
#define TARGET_DISPLAY_FPS 60
SDRThread::SDRThread() : IOThread(), buffers("SDRThreadBuffers") {
device = NULL;
@ -99,7 +100,7 @@ bool SDRThread::init() {
int streamMTU = device->getStreamMTU(stream);
mtuElems.store(streamMTU);
std::cout << "Stream MTU: " << mtuElems.load() << std::endl << std::flush;
std::cout << "Device Stream MTU: " << mtuElems.load() << std::endl << std::flush;
deviceInfo.load()->setStreamArgs(currentStreamArgs);
deviceConfig.load()->setStreamOpts(currentStreamArgs);
@ -131,7 +132,8 @@ bool SDRThread::init() {
device->setGainMode(SOAPY_SDR_RX,0,agc_mode.load());
numChannels.store(getOptimalChannelCount(sampleRate.load()));
numElems.store(getOptimalElementCount(sampleRate.load(), 30));
numElems.store(getOptimalElementCount(sampleRate.load(), TARGET_DISPLAY_FPS));
//fallback if mtuElems was wrong.
if (!mtuElems.load()) {
mtuElems.store(numElems.load());
}
@ -188,6 +190,7 @@ void SDRThread::assureBufferMinSize(SDRThreadIQData * dataOut, size_t minSize) {
//Called in an infinite loop, read SaopySDR device to build
// a 'this.numElems' sized batch of samples (SDRThreadIQData) and push it into iqDataOutQueue.
//this batch of samples is built to represent 1 frame / TARGET_DISPLAY_FPS.
void SDRThread::readStream(SDRThreadIQDataQueue* iqDataOutQueue) {
int flags;
long long timeNs;
@ -196,6 +199,13 @@ void SDRThread::readStream(SDRThreadIQDataQueue* iqDataOutQueue) {
int nElems = numElems.load();
int mtElems = mtuElems.load();
// Warning: if MTU > numElems, i.e if device MTU is too big w.r.t the sample rate, the TARGET_DISPLAY_FPS cannot
//be reached and the CubicSDR displays "slows down".
//To get back a TARGET_DISPLAY_FPS, the user need to adapt
//the SoapySDR Device to use smaller buffer sizes, because
// readStream() is suited to device MTU and cannot be really adapted dynamically.
//TODO: Add in doc the need to reduce SoapySDR device buffer length (if available) to restore higher fps.
//0. Retreive a new batch
SDRThreadIQData *dataOut = buffers.getBuffer();
@ -241,10 +251,6 @@ void SDRThread::readStream(SDRThreadIQDataQueue* iqDataOutQueue) {
if ((n_read + n_stream_read) > nElems) {
//n_requested is the exact number to reach nElems.
//note: setting n_requested = n_stream_read;
//lead to push the whole n_stream_read, so no overflow management occurs,
//so dataOut->data can exess nElems by at most mtElems.
//TODO: doesn't change CubicSDR behaviour, so why not ?
int n_requested = nElems-n_read;
//Copy at most n_requested CF32 into .data liquid_float_complex,
@ -399,6 +405,7 @@ void SDRThread::updateSettings() {
}
if (rate_changed.load()) {
device->setSampleRate(SOAPY_SDR_RX,0,sampleRate.load());
// TODO: explore bandwidth setting option to see if this is necessary for others
if (device->getDriverKey() == "bladeRF") {
@ -406,9 +413,10 @@ void SDRThread::updateSettings() {
}
sampleRate.store(device->getSampleRate(SOAPY_SDR_RX,0));
numChannels.store(getOptimalChannelCount(sampleRate.load()));
numElems.store(getOptimalElementCount(sampleRate.load(), 60));
numElems.store(getOptimalElementCount(sampleRate.load(), TARGET_DISPLAY_FPS));
int streamMTU = device->getStreamMTU(stream);
mtuElems.store(streamMTU);
//fallback if mtuElems was wrong
if (!mtuElems.load()) {
mtuElems.store(numElems.load());
}
@ -416,7 +424,9 @@ void SDRThread::updateSettings() {
overflowBuffer.data.resize(mtuElems.load());
free(buffs[0]);
buffs[0] = malloc(mtuElems.load() * 4 * sizeof(float));
//clear overflow buffer
numOverflow = 0;
rate_changed.store(false);
doUpdate = true;
}