fftfilt Class Reference

#include <fftfilt.h>

Collaboration diagram for fftfilt:

Public Types
typedef std::complex< float >	cmplx

Public Member Functions
	fftfilt (float f1, float f2, int len)
	fftfilt (float f2, int len)
	~fftfilt ()
void	create_filter (float f1, float f2)
void	create_dsb_filter (float f2)
void	create_asym_filter (float fopp, float fin)
	two different filters for in band and opposite band More...
void	create_rrc_filter (float fb, float a)
	root raised cosine. fb is half the band pass More...
int	noFilt (const cmplx &in, cmplx **out)
int	runFilt (const cmplx &in, cmplx **out)
int	runSSB (const cmplx &in, cmplx **out, bool usb, bool getDC=true)
int	runDSB (const cmplx &in, cmplx **out, bool getDC=true)
int	runAsym (const cmplx &in, cmplx **out, bool usb)
	Asymmetrical fitering can be used for vestigial sideband. More...

Protected Member Functions
float	fsinc (float fc, int i, int len)
float	_blackman (int i, int len)
cmplx	frrc (float fb, float a, int i, int len)
void	init_filter ()
void	init_dsb_filter ()

Protected Attributes
int	flen
int	flen2
g_fft< float > *	fft
cmplx *	filter
cmplx *	filterOpp
cmplx *	data
cmplx *	ovlbuf
cmplx *	output
int	inptr
int	pass
int	window

Private Types
enum	{ NONE, BLACKMAN, HAMMING, HANNING }

Detailed Description

Definition at line 17 of file fftfilt.h.

Member Typedef Documentation

cmplx

typedef std::complex<float> fftfilt::cmplx

Definition at line 21 of file fftfilt.h.

Member Enumeration Documentation

anonymous enum

anonymous enum

private

Enumerator
NONE
BLACKMAN
HAMMING
HANNING

Definition at line 18 of file fftfilt.h.

{NONE, BLACKMAN, HAMMING, HANNING};

Constructor & Destructor Documentation

fftfilt() [1/2]

fftfilt::fftfilt	(	float	f1,
		float	f2,
		int	len
	)

Definition at line 78 of file fftfilt.cpp.

References create_filter(), flen, init_filter(), pass, and window.

{
    flen    = len;
    pass    = 0;
    window  = 0;
    init_filter();
    create_filter(f1, f2);
}

Here is the call graph for this function:

fftfilt() [2/2]

fftfilt::fftfilt	(	float	f2,
		int	len
	)

Definition at line 87 of file fftfilt.cpp.

References create_dsb_filter(), flen, init_filter(), pass, and window.

{
    flen    = len;
    pass    = 0;
    window  = 0;
    init_filter();
    create_dsb_filter(f2);
}

Here is the call graph for this function:

~fftfilt()

fftfilt::~fftfilt

(

)

Definition at line 96 of file fftfilt.cpp.

References data, fft, filter, filterOpp, output, and ovlbuf.

{
    if (fft) delete fft;

    if (filter) delete [] filter;
    if (filterOpp) delete [] filterOpp;
    if (data) delete [] data;
    if (output) delete [] output;
    if (ovlbuf) delete [] ovlbuf;
}

Member Function Documentation

_blackman()

float fftfilt::_blackman ( int  i, int  len  )

inlineprotected

Definition at line 59 of file fftfilt.h.

References cos(), and M_PI.

Referenced by create_asym_filter(), create_dsb_filter(), and create_filter().

    {
        return (0.42 -
                 0.50 * cos(2.0 * M_PI * i / len) +
                 0.08 * cos(4.0 * M_PI * i / len));
    }

Here is the call graph for this function:

Here is the caller graph for this function:

create_asym_filter()

void fftfilt::create_asym_filter	(	float	fopp,
		float	fin
	)

two different filters for in band and opposite band

Definition at line 174 of file fftfilt.cpp.

References _blackman(), abs(), g_fft< FFT_TYPE >::ComplexFFT(), fft, filter, filterOpp, flen, flen2, fsinc(), and i.

Referenced by ATVDemod::applySettings(), and ATVMod::applySettings().

{
    // in band
    // initialize the filter to zero
    memset(filter, 0, flen * sizeof(cmplx));

    for (int i = 0; i < flen2; i++) {
        filter[i] = fsinc(fin, i, flen2);
        filter[i] *= _blackman(i, flen2);
    }

    fft->ComplexFFT(filter); // filter was expressed in the time domain (impulse response)

    // normalize the output filter for unity gain
    float scale = 0, mag;
    for (int i = 0; i < flen2; i++) {
        mag = abs(filter[i]);
        if (mag > scale) scale = mag;
    }
    if (scale != 0) {
        for (int i = 0; i < flen; i++)
            filter[i] /= scale;
    }

    // opposite band
    // initialize the filter to zero
    memset(filterOpp, 0, flen * sizeof(cmplx));

    for (int i = 0; i < flen2; i++) {
        filterOpp[i] = fsinc(fopp, i, flen2);
        filterOpp[i] *= _blackman(i, flen2);
    }

    fft->ComplexFFT(filterOpp); // filter was expressed in the time domain (impulse response)

    // normalize the output filter for unity gain
    scale = 0;
    for (int i = 0; i < flen2; i++) {
        mag = abs(filterOpp[i]);
        if (mag > scale) scale = mag;
    }
    if (scale != 0) {
        for (int i = 0; i < flen; i++)
            filterOpp[i] /= scale;
    }
}

Here is the call graph for this function:

Here is the caller graph for this function:

create_dsb_filter()

void fftfilt::create_dsb_filter

(

float

f2

)

Definition at line 148 of file fftfilt.cpp.

References _blackman(), abs(), g_fft< FFT_TYPE >::ComplexFFT(), fft, filter, flen, flen2, fsinc(), and i.

Referenced by AMDemod::applyAudioSampleRate(), SSBDemod::applyAudioSampleRate(), SSBMod::applyAudioSampleRate(), AMDemod::applySettings(), SSBDemod::applySettings(), SSBMod::applySettings(), fftfilt(), and ChannelAnalyzer::setFilters().

{
    // initialize the filter to zero
    memset(filter, 0, flen * sizeof(cmplx));

    for (int i = 0; i < flen2; i++) {
        filter[i] = fsinc(f2, i, flen2);
        filter[i] *= _blackman(i, flen2);
    }

    fft->ComplexFFT(filter); // filter was expressed in the time domain (impulse response)

    // normalize the output filter for unity gain
    float scale = 0, mag;
    for (int i = 0; i < flen2; i++) {
        mag = abs(filter[i]);
        if (mag > scale) scale = mag;
    }
    if (scale != 0) {
        for (int i = 0; i < flen; i++)
            filter[i] /= scale;
    }
}

Here is the call graph for this function:

Here is the caller graph for this function:

create_filter()

void fftfilt::create_filter	(	float	f1,
		float	f2
	)

Definition at line 107 of file fftfilt.cpp.

References _blackman(), abs(), g_fft< FFT_TYPE >::ComplexFFT(), fft, filter, flen, flen2, fsinc(), and i.

Referenced by SSBDemod::applyAudioSampleRate(), SSBMod::applyAudioSampleRate(), WFMDemod::applyChannelSettings(), BFMDemod::applyChannelSettings(), WFMMod::applyChannelSettings(), DATVDemod::applyChannelSettings(), ATVMod::applyChannelSettings(), FreeDVMod::applyFreeDVMode(), FreeDVDemod::applyFreeDVMode(), WFMDemod::applySettings(), UDPSource::applySettings(), BFMDemod::applySettings(), WFMMod::applySettings(), SSBDemod::applySettings(), SSBMod::applySettings(), DATVDemod::applySettings(), ATVMod::applySettings(), fftfilt(), and ChannelAnalyzer::setFilters().

{
    // initialize the filter to zero
    memset(filter, 0, flen * sizeof(cmplx));

    // create the filter shape coefficients by fft
    bool b_lowpass, b_highpass;
    b_lowpass = (f2 != 0);
    b_highpass = (f1 != 0);

    for (int i = 0; i < flen2; i++) {
        filter[i] = 0;
    // lowpass @ f2
        if (b_lowpass)
            filter[i] += fsinc(f2, i, flen2);
    // highighpass @ f1
        if (b_highpass)
            filter[i] -= fsinc(f1, i, flen2);
    }
    // highpass is delta[flen2/2] - h(t)
    if (b_highpass && f2 < f1)
        filter[flen2 / 2] += 1;

    for (int i = 0; i < flen2; i++)
        filter[i] *= _blackman(i, flen2);

    fft->ComplexFFT(filter); // filter was expressed in the time domain (impulse response)

    // normalize the output filter for unity gain
    float scale = 0, mag;
    for (int i = 0; i < flen2; i++) {
        mag = abs(filter[i]);
        if (mag > scale) scale = mag;
    }
    if (scale != 0) {
        for (int i = 0; i < flen; i++)
            filter[i] /= scale;
    }
}

Here is the call graph for this function:

Here is the caller graph for this function:

create_rrc_filter()

void fftfilt::create_rrc_filter	(	float	fb,
		float	a
	)

root raised cosine. fb is half the band pass

Definition at line 222 of file fftfilt.cpp.

References abs(), filter, flen, frrc(), and i.

Referenced by ChannelAnalyzer::applySettings(), ChannelAnalyzer::setFilters(), and FreqTracker::setInterpolator().

{
    std::fill(filter, filter+flen, 0);

    for (int i = 0; i < flen; i++) {
        filter[i] = frrc(fb, a, i, flen);
    }

    // normalize the output filter for unity gain
    float scale = 0, mag;
    for (int i = 0; i < flen; i++)
    {
        mag = abs(filter[i]);
        if (mag > scale) {
            scale = mag;
        }
    }
    if (scale != 0)
    {
        for (int i = 0; i < flen; i++) {
            filter[i] /= scale;
        }
    }
}

Here is the call graph for this function:

Here is the caller graph for this function:

frrc()

cmplx fftfilt::frrc ( float  fb, float  a, int  i, int  len  )

inlineprotected

RRC function in the frequency domain. Zero frequency is on the sides with first half in positive frequencies and second half in negative frequencies

Definition at line 68 of file fftfilt.h.

References cos(), and M_PI.

Referenced by create_rrc_filter().

    {
        float x = i/(float)len; // normalize to [0..1]
        x = 0.5-fabs(x-0.5); // apply symmetry: now both halves overlap near 0
        float tr = fb*a; // half the transition zone

        if (x < fb-tr)
        {
            return 1.0; // in band
        }
        else if (x < fb+tr) // transition
        {
            float y = ((x-(fb-tr)) / (2.0*tr))*M_PI;
            return (cos(y) + 1.0f)/2.0f;
        }
        else
        {
            return 0.0; // out of band
        }
    }

Here is the call graph for this function:

Here is the caller graph for this function:

fsinc()

float fftfilt::fsinc ( float  fc, int  i, int  len  )

inlineprotected

Definition at line 52 of file fftfilt.h.

References M_PI, and sin().

Referenced by create_asym_filter(), create_dsb_filter(), and create_filter().

    {
        int len2 = len/2;
        return (i == len2) ? 2.0 * fc:
                sin(2 * M_PI * fc * (i - len2)) / (M_PI * (i - len2));
    }

Here is the call graph for this function:

Here is the caller graph for this function:

init_dsb_filter()

void fftfilt::init_dsb_filter ( )

protected

init_filter()

void fftfilt::init_filter ( )

protected

Definition at line 51 of file fftfilt.cpp.

References data, fft, filter, filterOpp, flen, flen2, inptr, output, and ovlbuf.

Referenced by fftfilt().

{
    flen2   = flen >> 1;
    fft = new g_fft<float>(flen);

    filter      = new cmplx[flen];
    filterOpp   = new cmplx[flen];
    data        = new cmplx[flen];
    output      = new cmplx[flen2];
    ovlbuf      = new cmplx[flen2];

    memset(filter, 0, flen * sizeof(cmplx));
    memset(filterOpp, 0, flen * sizeof(cmplx));
    memset(data, 0, flen * sizeof(cmplx));
    memset(output, 0, flen2 * sizeof(cmplx));
    memset(ovlbuf, 0, flen2 * sizeof(cmplx));

    inptr = 0;
}

Here is the caller graph for this function:

noFilt()

int fftfilt::noFilt	(	const cmplx &	in,
		cmplx **	out
	)

Definition at line 248 of file fftfilt.cpp.

References data, flen2, and inptr.

{
    data[inptr++] = in;
    if (inptr < flen2)
        return 0;
    inptr = 0;

    *out = data;
    return flen2;
}

runAsym()

int fftfilt::runAsym	(	const cmplx &	in,
		cmplx **	out,
		bool	usb
	)

Asymmetrical fitering can be used for vestigial sideband.

Definition at line 360 of file fftfilt.cpp.

References g_fft< FFT_TYPE >::ComplexFFT(), data, fft, filter, filterOpp, flen, flen2, i, inptr, g_fft< FFT_TYPE >::InverseComplexFFT(), output, and ovlbuf.

Referenced by ATVDemod::demod(), and ATVMod::modulateVestigialSSB().

{
    data[inptr++] = in;
    if (inptr < flen2)
        return 0;
    inptr = 0;

    fft->ComplexFFT(data);

    data[0] *= filter[0]; // always keep DC

    if (usb)
    {
        for (int i = 1; i < flen2; i++)
        {
            data[i] *= filter[i]; // usb
            data[flen2 + i] *= filterOpp[flen2 + i]; // lsb is the opposite
        }
    }
    else
    {
        for (int i = 1; i < flen2; i++)
        {
            data[i] *= filterOpp[i]; // usb is the opposite
            data[flen2 + i] *= filter[flen2 + i]; // lsb
        }
    }

    // in-place FFT: freqdata overwritten with filtered timedata
    fft->InverseComplexFFT(data);

    // overlap and add
    for (int i = 0; i < flen2; i++) {
        output[i] = ovlbuf[i] + data[i];
        ovlbuf[i] = data[i+flen2];
    }

    memset (data, 0, flen * sizeof(cmplx));

    *out = output;
    return flen2;
}

Here is the call graph for this function:

Here is the caller graph for this function:

runDSB()

int fftfilt::runDSB	(	const cmplx &	in,
		cmplx **	out,
		bool	getDC = true
	)

Definition at line 327 of file fftfilt.cpp.

References g_fft< FFT_TYPE >::ComplexFFT(), data, fft, filter, flen, flen2, i, inptr, g_fft< FFT_TYPE >::InverseComplexFFT(), output, and ovlbuf.

Referenced by ChannelAnalyzer::processOneSample(), AMDemod::processOneSample(), SSBDemod::processOneSample(), and SSBMod::pullAF().

{
    data[inptr++] = in;
    if (inptr < flen2)
        return 0;
    inptr = 0;

    fft->ComplexFFT(data);

    for (int i = 0; i < flen2; i++) {
        data[i] *= filter[i];
        data[flen2 + i] *= filter[flen2 + i];
    }

    // get or reject DC component
    data[0] = getDC ? data[0] : 0;

    // in-place FFT: freqdata overwritten with filtered timedata
    fft->InverseComplexFFT(data);

    // overlap and add
    for (int i = 0; i < flen2; i++) {
        output[i] = ovlbuf[i] + data[i];
        ovlbuf[i] = data[i+flen2];
    }

    memset (data, 0, flen * sizeof(cmplx));

    *out = output;
    return flen2;
}

Here is the call graph for this function:

Here is the caller graph for this function:

runFilt()

int fftfilt::runFilt	(	const cmplx &	in,
		cmplx **	out
	)

Definition at line 260 of file fftfilt.cpp.

References g_fft< FFT_TYPE >::ComplexFFT(), data, fft, filter, flen, flen2, i, inptr, g_fft< FFT_TYPE >::InverseComplexFFT(), output, and ovlbuf.

Referenced by WFMDemod::feed(), BFMDemod::feed(), DATVDemod::feed(), ChannelAnalyzer::processOneSample(), FreqTracker::processOneSample(), and WFMMod::pull().

{
    data[inptr++] = in;
    if (inptr < flen2)
        return 0;
    inptr = 0;

    fft->ComplexFFT(data);
    for (int i = 0; i < flen; i++)
        data[i] *= filter[i];

    fft->InverseComplexFFT(data);

    for (int i = 0; i < flen2; i++) {
        output[i] = ovlbuf[i] + data[i];
        ovlbuf[i] = data[flen2 + i];
    }
    memset (data, 0, flen * sizeof(cmplx));

    *out = output;
    return flen2;
}

Here is the call graph for this function:

Here is the caller graph for this function:

runSSB()

int fftfilt::runSSB	(	const cmplx &	in,
		cmplx **	out,
		bool	usb,
		bool	getDC = true
	)

Definition at line 284 of file fftfilt.cpp.

References g_fft< FFT_TYPE >::ComplexFFT(), data, fft, filter, flen, flen2, i, inptr, g_fft< FFT_TYPE >::InverseComplexFFT(), output, and ovlbuf.

Referenced by UDPSink::feed(), UDPSource::modulateSample(), ATVMod::modulateSSB(), ChannelAnalyzer::processOneSample(), AMDemod::processOneSample(), SSBDemod::processOneSample(), FreeDVDemod::processOneSample(), FreeDVMod::pullAF(), and SSBMod::pullAF().

{
    data[inptr++] = in;
    if (inptr < flen2)
        return 0;
    inptr = 0;

    fft->ComplexFFT(data);

    // get or reject DC component
    data[0] = getDC ? data[0]*filter[0] : 0;

    // Discard frequencies for ssb
    if (usb)
    {
        for (int i = 1; i < flen2; i++) {
            data[i] *= filter[i];
            data[flen2 + i] = 0;
        }
    }
    else
    {
        for (int i = 1; i < flen2; i++) {
            data[i] = 0;
            data[flen2 + i] *= filter[flen2 + i];
        }
    }

    // in-place FFT: freqdata overwritten with filtered timedata
    fft->InverseComplexFFT(data);

    // overlap and add
    for (int i = 0; i < flen2; i++) {
        output[i] = ovlbuf[i] + data[i];
        ovlbuf[i] = data[i+flen2];
    }
    memset (data, 0, flen * sizeof(cmplx));

    *out = output;
    return flen2;
}

Here is the call graph for this function:

Here is the caller graph for this function:

Member Data Documentation

data

cmplx* fftfilt::data

protected

Definition at line 45 of file fftfilt.h.

Referenced by init_filter(), noFilt(), runAsym(), runDSB(), runFilt(), runSSB(), and ~fftfilt().

fft

g_fft<float>* fftfilt::fft

protected

Definition at line 42 of file fftfilt.h.

Referenced by create_asym_filter(), create_dsb_filter(), create_filter(), init_filter(), runAsym(), runDSB(), runFilt(), runSSB(), and ~fftfilt().

filter

cmplx* fftfilt::filter

protected

Definition at line 43 of file fftfilt.h.

Referenced by create_asym_filter(), create_dsb_filter(), create_filter(), create_rrc_filter(), init_filter(), runAsym(), runDSB(), runFilt(), runSSB(), and ~fftfilt().

filterOpp

cmplx* fftfilt::filterOpp

protected

Definition at line 44 of file fftfilt.h.

Referenced by create_asym_filter(), init_filter(), runAsym(), and ~fftfilt().

flen

int fftfilt::flen

protected

Definition at line 40 of file fftfilt.h.

Referenced by create_asym_filter(), create_dsb_filter(), create_filter(), create_rrc_filter(), fftfilt(), init_filter(), runAsym(), runDSB(), runFilt(), and runSSB().

flen2

int fftfilt::flen2

protected

Definition at line 41 of file fftfilt.h.

Referenced by create_asym_filter(), create_dsb_filter(), create_filter(), init_filter(), noFilt(), runAsym(), runDSB(), runFilt(), and runSSB().

inptr

int fftfilt::inptr

protected

Definition at line 48 of file fftfilt.h.

Referenced by init_filter(), noFilt(), runAsym(), runDSB(), runFilt(), and runSSB().

output

cmplx* fftfilt::output

protected

Definition at line 47 of file fftfilt.h.

Referenced by init_filter(), runAsym(), runDSB(), runFilt(), runSSB(), and ~fftfilt().

ovlbuf

cmplx* fftfilt::ovlbuf

protected

Definition at line 46 of file fftfilt.h.

Referenced by init_filter(), runAsym(), runDSB(), runFilt(), runSSB(), and ~fftfilt().

pass

int fftfilt::pass

protected

Definition at line 49 of file fftfilt.h.

Referenced by fftfilt().

window

int fftfilt::window

protected

Definition at line 50 of file fftfilt.h.

Referenced by fftfilt().

---

The documentation for this class was generated from the following files:

• sdrbase/dsp/fftfilt.h
• sdrbase/dsp/fftfilt.cpp