WIP symbol sync
Wojciech Kaczmarski
parent
408fad3744
commit
5c4d4975f7
|
|
@ -14,7 +14,7 @@ parameters:
|
|||
- id: max_dev
|
||||
label: Maximum deviation
|
||||
dtype: float
|
||||
default: 0.005
|
||||
default: 0.05
|
||||
|
||||
templates:
|
||||
imports: from gnuradio import m17
|
||||
|
|
|
|||
|
|
@ -51,6 +51,8 @@ namespace gr
|
|||
;
|
||||
}
|
||||
|
||||
static constexpr int INTERPS_PER_SYMBOL = 16;
|
||||
|
||||
void symbol_sync_impl::set_in_sps(int in_sps)
|
||||
{
|
||||
_in_sps = in_sps;
|
||||
|
|
@ -72,15 +74,23 @@ namespace gr
|
|||
d_omega = d_omega_mid;
|
||||
d_omega_lim = _max_dev;
|
||||
|
||||
// very standard Gardner loop gains
|
||||
d_gain_omega = 0.25f * _loop_bw * _loop_bw;
|
||||
d_gain_mu = _loop_bw;
|
||||
// ---- PLL coefficients (zeta = 1.0, TED gain = 1.0) ----
|
||||
const float zeta = 1.0f;
|
||||
// const float Kp = 1.0f; // Gardner TED gain (normalized)
|
||||
const float Bn = _loop_bw; // normalized loop bandwidth
|
||||
|
||||
const float denom = 1.0f + 2.0f * zeta * Bn + Bn * Bn;
|
||||
d_gain_mu = (4.0f * zeta * Bn) / denom;
|
||||
d_gain_omega = (4.0f * Bn * Bn) / denom;
|
||||
|
||||
d_mu = 0.0f;
|
||||
d_tau = 0.0f;
|
||||
|
||||
d_prev_sample = 0.0f;
|
||||
d_mid_sample = 0.0f;
|
||||
d_have_mid = false;
|
||||
d_prev_mid = 0.0f;
|
||||
d_prev_symbol = 0.0f;
|
||||
|
||||
d_interp_tick = 0;
|
||||
d_phase = 0.0f;
|
||||
}
|
||||
|
||||
inline float clampf(float x, float lo, float hi)
|
||||
|
|
@ -92,6 +102,22 @@ namespace gr
|
|||
return x;
|
||||
}
|
||||
|
||||
static inline float interp_cubic(const float *x, float mu)
|
||||
{
|
||||
// x[-1], x[0], x[1], x[2] must be valid
|
||||
const float xm1 = x[-1];
|
||||
const float x0 = x[0];
|
||||
const float x1 = x[1];
|
||||
const float x2 = x[2];
|
||||
|
||||
// const float a0 = x1 - x0;
|
||||
const float a1 = 0.5f * (x1 - xm1);
|
||||
const float a2 = xm1 - 2.5f * x0 + 2.0f * x1 - 0.5f * x2;
|
||||
const float a3 = 0.5f * (x2 - xm1) + 1.5f * (x0 - x1);
|
||||
|
||||
return ((a3 * mu + a2) * mu + a1) * mu + x0;
|
||||
}
|
||||
|
||||
/*
|
||||
* Our virtual destructor.
|
||||
*/
|
||||
|
|
@ -116,53 +142,58 @@ namespace gr
|
|||
float *out = (float *)output_items[0];
|
||||
|
||||
int ni = ninput_items[0];
|
||||
int ii = 0;
|
||||
int ii = 1; // input index
|
||||
int oo = 0;
|
||||
|
||||
while (ii < ni && oo < noutput_items)
|
||||
while (ii + 2 < ni && oo < noutput_items)
|
||||
{
|
||||
// fractional interpolation (linear is enough here)
|
||||
int i0 = ii;
|
||||
int i1 = ii + 1;
|
||||
if (i1 >= ni)
|
||||
break;
|
||||
// --------------------------------------------------
|
||||
// Uniform interpolation clock (fixed rate)
|
||||
// --------------------------------------------------
|
||||
float interp_mu = d_phase;
|
||||
float interp_out = interp_cubic(&in[ii], interp_mu);
|
||||
|
||||
float s = in[i0] + d_mu * (in[i1] - in[i0]);
|
||||
|
||||
// midpoint sample
|
||||
if (!d_have_mid && d_mu >= 0.5f)
|
||||
// --------------------------------------------------
|
||||
// Gardner TED at half-symbol spacing
|
||||
// --------------------------------------------------
|
||||
if (d_interp_tick == INTERPS_PER_SYMBOL / 2)
|
||||
{
|
||||
d_mid_sample = s;
|
||||
d_have_mid = true;
|
||||
d_prev_mid = interp_out;
|
||||
}
|
||||
|
||||
d_mu += 1.0f / d_omega;
|
||||
|
||||
if (d_mu >= 1.0f)
|
||||
// --------------------------------------------------
|
||||
// Symbol clock
|
||||
// --------------------------------------------------
|
||||
if (d_interp_tick == 0)
|
||||
{
|
||||
d_mu -= 1.0f;
|
||||
float curr = interp_out;
|
||||
|
||||
// symbol boundary
|
||||
float error = 0.0f;
|
||||
if (d_have_mid)
|
||||
{
|
||||
error = (d_mid_sample - d_prev_sample) * s;
|
||||
d_have_mid = false;
|
||||
}
|
||||
float error = (d_prev_mid - interp_out) * curr;
|
||||
|
||||
// loop filter
|
||||
d_omega += d_gain_omega * error;
|
||||
d_omega = clampf(d_omega,
|
||||
d_omega_mid - d_omega_lim,
|
||||
d_omega_mid + d_omega_lim);
|
||||
|
||||
d_mu += d_gain_mu * error;
|
||||
d_phase += d_gain_mu * error;
|
||||
|
||||
out[oo++] = s;
|
||||
d_prev_sample = s;
|
||||
out[oo++] = curr;
|
||||
}
|
||||
|
||||
ii++;
|
||||
// --------------------------------------------------
|
||||
// Advance interpolator phase
|
||||
// --------------------------------------------------
|
||||
d_phase += d_omega / INTERPS_PER_SYMBOL;
|
||||
|
||||
while (d_phase >= 1.0f)
|
||||
{
|
||||
d_phase -= 1.0f;
|
||||
ii++;
|
||||
}
|
||||
|
||||
d_interp_tick++;
|
||||
if (d_interp_tick == INTERPS_PER_SYMBOL)
|
||||
d_interp_tick = 0;
|
||||
}
|
||||
|
||||
consume_each(ii);
|
||||
|
|
|
|||
|
|
@ -21,22 +21,27 @@ namespace gr
|
|||
private:
|
||||
int _in_sps = 10;
|
||||
float _loop_bw = 0.001f;
|
||||
float _max_dev = 0.005f;
|
||||
float _max_dev = 0.05f;
|
||||
|
||||
// timing
|
||||
float d_mu; // fractional timing [0, 1)
|
||||
float d_omega; // samples per symbol (≈ in_sps)
|
||||
float d_omega_mid; // nominal in_sps
|
||||
float d_omega_lim; // max deviation
|
||||
float d_mu; // fractional phase [0,1)
|
||||
float d_tau; // interp phase [0,1)
|
||||
float d_omega; // samples per symbol (adaptive)
|
||||
float d_omega_mid; // nominal SPS
|
||||
float d_omega_lim; // max deviation
|
||||
|
||||
// loop filter
|
||||
float d_gain_omega; // proportional
|
||||
float d_gain_mu; // integral
|
||||
float d_mu_prev;
|
||||
|
||||
// TED history
|
||||
float d_prev_sample;
|
||||
float d_mid_sample;
|
||||
bool d_have_mid;
|
||||
float d_gain_mu; // loop gain
|
||||
float d_gain_omega;
|
||||
|
||||
float d_prev_symbol; // previous symbol sample
|
||||
float d_mid_sample; // mid-sample
|
||||
float d_prev_mid; // x(k - 1/2)
|
||||
|
||||
bool d_have_mid;
|
||||
|
||||
int d_interp_tick; // [0 .. INTERPS_PER_SYMBOL-1]
|
||||
float d_phase; // interpolator phase in input samples
|
||||
|
||||
public:
|
||||
symbol_sync_impl(int in_sps, float loop_bw, float max_dev);
|
||||
|
|
|
|||
Loading…
Reference in New Issue