fun::QAM< NumBits > Class Template Reference

Template for QAM class. More...

#include <qam.h>

Public Member Functions
	QAM (double power, int gain=0)
	QAM Constructor. More...
void	encode (const char *bits, double *sym)
	Encode recursively to match the decoding process. More...
void	decode (double sym, unsigned char *bits)
	Decode recursively. More...

Static Public Member Functions
static int	sign (int v)
	sign More...
static int	clamp (int i)
	clamp More...

Private Attributes
int	d_gain
double	d_scale_e
double	d_scale_d

Detailed Description

template<int NumBits>
class fun::QAM< NumBits >

Template for QAM class.

fast QAM (uses at most 4x imull to decode) Tested on 7600 bogomips yields 600-1200Mbps encoding and 300Mbps decoding Compile with -O3

Definition at line 24 of file qam.h.

Constructor & Destructor Documentation

template<int NumBits>

fun::QAM< NumBits >::QAM ( double  power, int  gain = 0  )

inline

QAM Constructor.

Parameters

power	desired symbol power
gain	gain on the decoded confidence (power of 2)

Definition at line 35 of file qam.h.

References fun::QAM< NumBits >::d_gain, fun::QAM< NumBits >::d_scale_d, and fun::QAM< NumBits >::d_scale_e.

        {
            d_gain = gain + CHAR_BIT - NumBits;
            const int nn = (1<<(NumBits-1));
            // sum((2k+1)^2,k=0..n-1)
            #if 0
            int sum2 = 0;
            for (int i = 0; i < nn; ++i) {
            sum2+= (2*i + 1)*(2*i + 1);
            }
            #else
            int sum2 = (4*nn*nn*nn-nn)/3;
            #endif
            double sf = sqrt(power * double(nn) / double(sum2));
            d_scale_e = sf;
            d_scale_d = (1 << d_gain) / sf;
        }

Member Function Documentation

template<int NumBits>

static int fun::QAM< NumBits >::clamp ( int  i)

inlinestatic

clamp

Parameters

i

Returns

saturate between 0 and 255

Definition at line 68 of file qam.h.

Referenced by fun::QAM< NumBits >::decode().

        {
            return i < 0 ? 0 : (i > 255 ? 255 : i);
        }

template<int NumBits>

void fun::QAM< NumBits >::decode ( double  sym, unsigned char *  bits  )

inline

Decode recursively.

We decode recursively because we want meaningful confidences. The alternative would be to simply divide and round, which only yields the smallest per-bit confidence.

output bit confidence is between 0 and 255.

Parameters

sym
bits

Definition at line 110 of file qam.h.

References amp, fun::QAM< NumBits >::clamp(), fun::QAM< NumBits >::d_gain, fun::QAM< NumBits >::d_scale_d, and fun::QAM< NumBits >::sign().

Referenced by fun::modulator::demodulate().

        {
            int pt = sym * d_scale_d;
            int flip = 1; // +1 or -1 -- for gray coding
            int amp = (1 << (NumBits-1)) << d_gain;
            // unrolled with -O3
            for (int i = 0; i < NumBits; ++i)
            {
                *bits = clamp(flip * pt + 128);
                int bit = sign(pt);
                pt -= bit * amp;
                flip = -bit;
                amp /= 2;
                ++bits;
            }
        }

template<int NumBits>

void fun::QAM< NumBits >::encode ( const char *  bits, double *  sym  )

inline

Encode recursively to match the decoding process.

This could have been implemented by a gray + multiply. gray(i) = (i> >1)^i see: http://www.dspguru.com/dsp/tricks/gray-code-conversion

Parameters

bits
sym

Definition at line 83 of file qam.h.

References fun::QAM< NumBits >::d_scale_e.

Referenced by fun::modulator::modulate().

        {
            int pt = 0; // constellation point
            int flip = 1; // +1 or -1 -- for gray coding
            // unrolled with -O3
            for (int i = 0; i < NumBits; ++i)
                {
                int bit = *bits * 2 - 1; // +1 or -1
                pt = bit * flip + pt * 2;
                flip *= -bit;
                ++bits;
                }
            *sym = pt * d_scale_e;
        }

template<int NumBits>

static int fun::QAM< NumBits >::sign ( int  v)

inlinestatic

sign

Parameters

v

Returns

+1 or -1

Definition at line 58 of file qam.h.

Referenced by fun::QAM< NumBits >::decode().

        {
            return +1 | (v >> (sizeof(int) * CHAR_BIT - 1));
        }

Member Data Documentation

template<int NumBits>

int fun::QAM< NumBits >::d_gain

private

Definition at line 26 of file qam.h.

Referenced by fun::QAM< NumBits >::decode(), and fun::QAM< NumBits >::QAM().

template<int NumBits>

double fun::QAM< NumBits >::d_scale_d

private

Definition at line 28 of file qam.h.

Referenced by fun::QAM< NumBits >::decode(), and fun::QAM< NumBits >::QAM().

template<int NumBits>

double fun::QAM< NumBits >::d_scale_e

private

Definition at line 27 of file qam.h.

Referenced by fun::QAM< NumBits >::encode(), and fun::QAM< NumBits >::QAM().

---

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

• /home/benjamin/Documents/fun_ofdm/src/qam.h