sdrangel/sdrgui/dsp/spectrumvis.cpp

#include "dsp/spectrumvis.h"
#include "gui/glspectrum.h"
#include "dsp/dspcommands.h"
#include "util/messagequeue.h"

#define MAX_FFT_SIZE 4096

#ifndef LINUX
inline double log2f(double n)
{
	return log(n) / log(2.0);
}
#endif

MESSAGE_CLASS_DEFINITION(SpectrumVis::MsgConfigureSpectrumVis, Message)
MESSAGE_CLASS_DEFINITION(SpectrumVis::MsgConfigureScalingFactor, Message)

const Real SpectrumVis::m_mult = (10.0f / log2f(10.0f));

SpectrumVis::SpectrumVis(Real scalef, GLSpectrum* glSpectrum) :
	BasebandSampleSink(),
	m_fft(FFTEngine::create()),
	m_fftBuffer(MAX_FFT_SIZE),
	m_powerSpectrum(MAX_FFT_SIZE),
	m_fftBufferFill(0),
	m_needMoreSamples(false),
	m_scalef(scalef),
	m_glSpectrum(glSpectrum),
	m_averageNb(0),
	m_avgMode(AvgModeNone),
	m_linear(false),
	m_ofs(0),
    m_powFFTDiv(1.0),
	m_mutex(QMutex::Recursive)
{
	setObjectName("SpectrumVis");
	handleConfigure(1024, 0, 0, AvgModeNone, FFTWindow::BlackmanHarris, false);
}

SpectrumVis::~SpectrumVis()
{
	delete m_fft;
}

void SpectrumVis::configure(MessageQueue* msgQueue,
        int fftSize,
        int overlapPercent,
        unsigned int averagingNb,
        AvgMode averagingMode,
        FFTWindow::Function window,
        bool linear)
{
	MsgConfigureSpectrumVis* cmd = new MsgConfigureSpectrumVis(fftSize, overlapPercent, averagingNb, averagingMode, window, linear);
	msgQueue->push(cmd);
}

void SpectrumVis::setScalef(MessageQueue* msgQueue, Real scalef)
{
    MsgConfigureScalingFactor* cmd = new MsgConfigureScalingFactor(scalef);
    getInputMessageQueue()->push(cmd);
}

void SpectrumVis::feedTriggered(const SampleVector::const_iterator& triggerPoint, const SampleVector::const_iterator& end, bool positiveOnly)
{
	feed(triggerPoint, end, positiveOnly); // normal feed from trigger point
	/*
	if (triggerPoint == end)
	{
		// the following piece of code allows to terminate the FFT that ends past the end of scope captured data
		// that is the spectrum will include the captured data
		// just do nothing if you want the spectrum to be included inside the scope captured data
		// that is to drop the FFT that dangles past the end of captured data
		if (m_needMoreSamples) {
			feed(begin, end, positiveOnly);
			m_needMoreSamples = false;      // force finish
		}
	}
	else
	{
		feed(triggerPoint, end, positiveOnly); // normal feed from trigger point
	}*/
}

void SpectrumVis::feed(const SampleVector::const_iterator& cbegin, const SampleVector::const_iterator& end, bool positiveOnly)
{
	// if no visualisation is set, send the samples to /dev/null

	if (m_glSpectrum == 0) {
		return;
	}

    if (!m_mutex.tryLock(0)) { // prevent conflicts with configuration process
        return;
    }

	SampleVector::const_iterator begin(cbegin);

	while (begin < end)
	{
		std::size_t todo = end - begin;
		std::size_t samplesNeeded = m_refillSize - m_fftBufferFill;

		if (todo >= samplesNeeded)
		{
			// fill up the buffer
			std::vector<Complex>::iterator it = m_fftBuffer.begin() + m_fftBufferFill;

			for (std::size_t i = 0; i < samplesNeeded; ++i, ++begin)
			{
				*it++ = Complex(begin->real() / m_scalef, begin->imag() / m_scalef);
			}

			// apply fft window (and copy from m_fftBuffer to m_fftIn)
			m_window.apply(&m_fftBuffer[0], m_fft->in());

			// calculate FFT
			m_fft->transform();

			// extract power spectrum and reorder buckets
			const Complex* fftOut = m_fft->out();
			Complex c;
			Real v;
			std::size_t halfSize = m_fftSize / 2;

			if (m_avgMode == AvgModeNone)
			{
                if ( positiveOnly )
                {
                    for (std::size_t i = 0; i < halfSize; i++)
                    {
                        c = fftOut[i];
                        v = c.real() * c.real() + c.imag() * c.imag();
                        v = m_linear ? v/m_powFFTDiv : m_mult * log2f(v) + m_ofs;
                        m_powerSpectrum[i * 2] = v;
                        m_powerSpectrum[i * 2 + 1] = v;
                    }
                }
                else
                {
                    for (std::size_t i = 0; i < halfSize; i++)
                    {
                        c = fftOut[i + halfSize];
                        v = c.real() * c.real() + c.imag() * c.imag();
                        v = m_linear ? v/m_powFFTDiv : m_mult * log2f(v) + m_ofs;
                        m_powerSpectrum[i] = v;

                        c = fftOut[i];
                        v = c.real() * c.real() + c.imag() * c.imag();
                        v = m_linear ? v/m_powFFTDiv : m_mult * log2f(v) + m_ofs;
                        m_powerSpectrum[i + halfSize] = v;
                    }
                }

                // send new data to visualisation
                m_glSpectrum->newSpectrum(m_powerSpectrum, m_fftSize);
			}
			else if (m_avgMode == AvgModeMovingAvg)
			{
	            if ( positiveOnly )
	            {
	                for (std::size_t i = 0; i < halfSize; i++)
	                {
	                    c = fftOut[i];
	                    v = c.real() * c.real() + c.imag() * c.imag();
	                    v = m_movingAverage.storeAndGetAvg(v, i);
	                    v = m_linear ? v/m_powFFTDiv : m_mult * log2f(v) + m_ofs;
	                    m_powerSpectrum[i * 2] = v;
	                    m_powerSpectrum[i * 2 + 1] = v;
	                }
	            }
	            else
	            {
	                for (std::size_t i = 0; i < halfSize; i++)
	                {
	                    c = fftOut[i + halfSize];
	                    v = c.real() * c.real() + c.imag() * c.imag();
	                    v = m_movingAverage.storeAndGetAvg(v, i+halfSize);
	                    v = m_linear ? v/m_powFFTDiv : m_mult * log2f(v) + m_ofs;
	                    m_powerSpectrum[i] = v;

	                    c = fftOut[i];
	                    v = c.real() * c.real() + c.imag() * c.imag();
	                    v = m_movingAverage.storeAndGetAvg(v, i);
	                    v = m_linear ? v/m_powFFTDiv : m_mult * log2f(v) + m_ofs;
	                    m_powerSpectrum[i + halfSize] = v;
	                }
	            }

	            // send new data to visualisation
	            m_glSpectrum->newSpectrum(m_powerSpectrum, m_fftSize);
	            m_movingAverage.nextAverage();
			}
			else if (m_avgMode == AvgModeFixedAvg)
			{
			    double avg;

                if ( positiveOnly )
                {
                    for (std::size_t i = 0; i < halfSize; i++)
                    {
                        c = fftOut[i];
                        v = c.real() * c.real() + c.imag() * c.imag();

                        if (m_fixedAverage.storeAndGetAvg(avg, v, i))
                        {
                            avg = m_linear ? avg/m_powFFTDiv : m_mult * log2f(avg) + m_ofs;
                            m_powerSpectrum[i * 2] = avg;
                            m_powerSpectrum[i * 2 + 1] = avg;
                        }
                    }
                }
                else
                {
                    for (std::size_t i = 0; i < halfSize; i++)
                    {
                        c = fftOut[i + halfSize];
                        v = c.real() * c.real() + c.imag() * c.imag();

                        if (m_fixedAverage.storeAndGetAvg(avg, v, i+halfSize))
                        { // result available
                            avg = m_linear ? avg/m_powFFTDiv : m_mult * log2f(avg) + m_ofs;
                            m_powerSpectrum[i] = avg;
                        }

                        c = fftOut[i];
                        v = c.real() * c.real() + c.imag() * c.imag();

                        if (m_fixedAverage.storeAndGetAvg(avg, v, i))
                        { // result available
                            avg = m_linear ? avg/m_powFFTDiv : m_mult * log2f(avg) + m_ofs;
                            m_powerSpectrum[i + halfSize] = avg;
                        }
                    }
                }

                if (m_fixedAverage.nextAverage()) { // result available
                    m_glSpectrum->newSpectrum(m_powerSpectrum, m_fftSize); // send new data to visualisation
                }
			}
			else if (m_avgMode == AvgModeMax)
			{
			    double max;

                if ( positiveOnly )
                {
                    for (std::size_t i = 0; i < halfSize; i++)
                    {
                        c = fftOut[i];
                        v = c.real() * c.real() + c.imag() * c.imag();

                        if (m_max.storeAndGetMax(max, v, i))
                        {
                            max = m_linear ? max/m_powFFTDiv : m_mult * log2f(max) + m_ofs;
                            m_powerSpectrum[i * 2] = max;
                            m_powerSpectrum[i * 2 + 1] = max;
                        }
                    }
                }
                else
                {
                    for (std::size_t i = 0; i < halfSize; i++)
                    {
                        c = fftOut[i + halfSize];
                        v = c.real() * c.real() + c.imag() * c.imag();

                        if (m_max.storeAndGetMax(max, v, i+halfSize))
                        { // result available
                            max = m_linear ? max/m_powFFTDiv : m_mult * log2f(max) + m_ofs;
                            m_powerSpectrum[i] = max;
                        }

                        c = fftOut[i];
                        v = c.real() * c.real() + c.imag() * c.imag();

                        if (m_max.storeAndGetMax(max, v, i))
                        { // result available
                            max = m_linear ? max/m_powFFTDiv : m_mult * log2f(max) + m_ofs;
                            m_powerSpectrum[i + halfSize] = max;
                        }
                    }
                }

                if (m_max.nextMax()) { // result available
                    m_glSpectrum->newSpectrum(m_powerSpectrum, m_fftSize); // send new data to visualisation
                }
			}

			// advance buffer respecting the fft overlap factor
			std::copy(m_fftBuffer.begin() + m_refillSize, m_fftBuffer.end(), m_fftBuffer.begin());

			// start over
			m_fftBufferFill = m_overlapSize;
			m_needMoreSamples = false;
		}
		else
		{
			// not enough samples for FFT - just fill in new data and return
			for(std::vector<Complex>::iterator it = m_fftBuffer.begin() + m_fftBufferFill; begin < end; ++begin)
			{
				*it++ = Complex(begin->real() / m_scalef, begin->imag() / m_scalef);
			}

			m_fftBufferFill += todo;
			m_needMoreSamples = true;
		}
	}

	 m_mutex.unlock();
}

void SpectrumVis::start()
{
}

void SpectrumVis::stop()
{
}

bool SpectrumVis::handleMessage(const Message& message)
{
	if (MsgConfigureSpectrumVis::match(message))
	{
		MsgConfigureSpectrumVis& conf = (MsgConfigureSpectrumVis&) message;
		handleConfigure(conf.getFFTSize(),
		        conf.getOverlapPercent(),
		        conf.getAverageNb(),
		        conf.getAvgMode(),
		        conf.getWindow(),
		        conf.getLinear());
		return true;
	}
    else if (MsgConfigureScalingFactor::match(message))
    {
        MsgConfigureScalingFactor& conf = (MsgConfigureScalingFactor&) message;
        handleScalef(conf.getScalef());
        return true;
    }
	else
	{
		return false;
	}
}

void SpectrumVis::handleConfigure(int fftSize,
        int overlapPercent,
        unsigned int averageNb,
        AvgMode averagingMode,
        FFTWindow::Function window,
        bool linear)
{
//    qDebug("SpectrumVis::handleConfigure, fftSize: %d overlapPercent: %d averageNb: %u averagingMode: %d window: %d linear: %s",
//            fftSize, overlapPercent, averageNb, (int) averagingMode, (int) window, linear ? "true" : "false");
	QMutexLocker mutexLocker(&m_mutex);

	if (fftSize > MAX_FFT_SIZE)
	{
		fftSize = MAX_FFT_SIZE;
	}
	else if (fftSize < 64)
	{
		fftSize = 64;
	}

	if (overlapPercent > 100)
	{
		m_overlapPercent = 100;
	}
	else if (overlapPercent < 0)
	{
		m_overlapPercent = 0;
	}
	else
	{
        m_overlapPercent = overlapPercent;
	}

	m_fftSize = fftSize;
	m_fft->configure(m_fftSize, false);
	m_window.create(window, m_fftSize);
	m_overlapSize = (m_fftSize * m_overlapPercent) / 100;
	m_refillSize = m_fftSize - m_overlapSize;
	m_fftBufferFill = m_overlapSize;
	m_movingAverage.resize(fftSize, averageNb > 1000 ? 1000 : averageNb); // Capping to avoid out of memory condition
	m_fixedAverage.resize(fftSize, averageNb);
	m_max.resize(fftSize, averageNb);
	m_averageNb = averageNb;
	m_avgMode = averagingMode;
	m_linear = linear;
	m_ofs = 20.0f * log10f(1.0f / m_fftSize);
	m_powFFTDiv = m_fftSize*m_fftSize;
}

void SpectrumVis::handleScalef(Real scalef)
{
    QMutexLocker mutexLocker(&m_mutex);
    m_scalef = scalef;
}