silas/imdd_silas
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

MPI FFE with A1 and A2 algorithms

Browse Source
This commit is contained in:
Silas Oettinghaus
2026-07-10 09:14:34 +02:00
parent aa210e5352
commit a6cf742121
11 changed files with 863 additions and 229 deletions
Download Patch File Download Diff File Expand all files Collapse all files

661
Classes/04_DSP/Equalizer/FFE.m
View File

@@ -12,7 +12,7 @@ classdef FFE < handle
% eq_ = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ... % eq_ = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ...
% "mu_dd",1e-1,"mu_tr",0.4,"order",50, ... % "mu_dd",1e-1,"mu_tr",0.4,"order",50, ...
% "sps",2,"decide",0,"optmize_mus",1,"dd_mode",1, ... % "sps",2,"decide",0,"optmize_mus",1,"dd_mode",1, ...
% "adaption_technique","nlms","mu_dc",1.021e-05); % "adaption_technique","nlms","dc_tracking_mu",1.021e-05);
properties properties
@@ -32,8 +32,24 @@ classdef FFE < handle
mu_dd %weight update in dd mode mu_dd %weight update in dd mode
epochs_dd epochs_dd
dd_len_fraction dd_len_fraction
mu_dc
adaptive_dc_enabled % A1 moving-average input suppression
dc_avg_bufferlength_a1
dc_smoothing_a1
% A2 level-dependent residual suppression
dc_level_avg_bufferlength_a2
dc_smoothing_a2
dc_level_weights_a2
% Adaptive DC-tracking loop
dc_tracking_mu
dc_tracking_adaptive_enabled
dc_tracking_power_exponent
dc_tracking_buffer_len
% Delayed FFE tap update
ffe_update_buffer_len
e_dc e_dc
P % covariance matrix of rls P % covariance matrix of rls
@@ -51,23 +67,29 @@ classdef FFE < handle
mu_optimization_len mu_optimization_len
plot_mu_optimization = 0; plot_mu_optimization = 0;
mu_optimization_fignum = 3010; mu_optimization_fignum = 3010;
optimize_dc_params = 0; optimize_dc_tracking_params = 0;
dc_optimization dc_tracking_optimization
dc_optimization_iter = 0; dc_tracking_optimization_iter = 0;
dc_optimization_len dc_tracking_optimization_len
dc_optimization_max_evals dc_tracking_optimization_max_evals
dc_optimization_delay_weight dc_tracking_optimization_delay_weight
dc_optimization_smoothing_len dc_tracking_optimization_smoothing_len
optimize_a2_level_weights = 0;
a2_level_weight_optimization
a2_level_weight_optimization_iter = 0;
a2_level_weight_optimization_len
a2_level_weight_optimization_max_evals
a2_level_weight_max
a2_level_weight_initial_stats
end end
properties (Access = private) properties (Access = private)
dc_alpha = 0.98 dc_tracking_alpha = 0.98
dc_gamma = 1e-6 dc_tracking_gamma = 1e-6
dc_mu_min = 1e-6 dc_tracking_mu_min = 1e-6
dc_mu_max = 3e-1 dc_tracking_mu_max = 3e-1
dc_mu_eff_min = 0 dc_tracking_mu_eff_min = 0
dc_mu_eff_max = inf dc_tracking_mu_eff_max = inf
dc_power_exponent = 1
end end
methods methods
@@ -86,8 +108,20 @@ classdef FFE < handle
options.mu_dd = 1e-5; options.mu_dd = 1e-5;
options.epochs_dd = 5; options.epochs_dd = 5;
options.dd_len_fraction = 1; options.dd_len_fraction = 1;
options.mu_dc = 0;
options.adaptive_dc_enabled = false; options.dc_avg_bufferlength_a1 = 0;
options.dc_smoothing_a1 = 0;
options.dc_level_avg_bufferlength_a2 = 0;
options.dc_smoothing_a2 = 0;
options.dc_level_weights_a2 = 0;
options.dc_tracking_mu = 0;
options.dc_tracking_adaptive_enabled = false;
options.dc_tracking_power_exponent = 2;
options.dc_tracking_buffer_len = 1;
options.ffe_update_buffer_len = 1;
options.decide = false; options.decide = false;
@@ -96,11 +130,15 @@ classdef FFE < handle
options.mu_optimization_len = 2^15; options.mu_optimization_len = 2^15;
options.plot_mu_optimization = 0; options.plot_mu_optimization = 0;
options.mu_optimization_fignum = 3010; options.mu_optimization_fignum = 3010;
options.optimize_dc_params = 0; options.optimize_dc_tracking_params = 0;
options.dc_optimization_len = 2^15; options.dc_tracking_optimization_len = 2^15;
options.dc_optimization_max_evals = 20; options.dc_tracking_optimization_max_evals = 20;
options.dc_optimization_delay_weight = 1e-3; options.dc_tracking_optimization_delay_weight = 1e-3;
options.dc_optimization_smoothing_len = 501; options.dc_tracking_optimization_smoothing_len = 501;
options.optimize_a2_level_weights = 0;
options.a2_level_weight_optimization_len = 2^15;
options.a2_level_weight_optimization_max_evals = 30;
options.a2_level_weight_max = 1;
end end
@@ -109,9 +147,34 @@ classdef FFE < handle
obj.(fn{n}) = options.(fn{n}); obj.(fn{n}) = options.(fn{n});
end end
assert(obj.dc_tracking_buffer_len >= 0);
assert(obj.ffe_update_buffer_len >= 0);
assert(obj.dc_avg_bufferlength_a1 >= 0);
assert(obj.dc_level_avg_bufferlength_a2 >= 0);
obj.e = zeros(obj.order,1); obj.e = zeros(obj.order,1);
obj.e_dc = 0; obj.e_dc = 0;
obj.error = 0; obj.error = 0;
obj.a2_level_weight_initial_stats = struct();
obj.dc_avg_bufferlength_a1 = floor(obj.dc_avg_bufferlength_a1);
obj.dc_smoothing_a1 = min(max(obj.dc_smoothing_a1,0),1);
obj.dc_level_avg_bufferlength_a2 = floor(obj.dc_level_avg_bufferlength_a2);
obj.dc_smoothing_a2 = min(max(obj.dc_smoothing_a2,0),1);
obj.dc_tracking_buffer_len = floor(obj.dc_tracking_buffer_len);
obj.ffe_update_buffer_len = floor(obj.ffe_update_buffer_len);
obj.a2_level_weight_max = max(obj.a2_level_weight_max,0);
if obj.dc_avg_bufferlength_a1 > 1 && (obj.dc_tracking_mu ~= 0 || obj.optimize_dc_tracking_params)
warning("FFE:DCAvgWithDcTracking", ...
"A1 moving-average DC suppression and dc_tracking_mu/adaptive DC tracking are alternative DC suppression paths and will be combined.");
end
a2_requested = obj.dc_level_avg_bufferlength_a2 > 1 && ...
(any(obj.dc_level_weights_a2(:) ~= 0) || obj.optimize_a2_level_weights);
if a2_requested && ...
(obj.dc_tracking_mu ~= 0 || obj.optimize_dc_tracking_params || obj.dc_avg_bufferlength_a1 > 1)
warning("FFE:DCLevelAvgWithOtherDcSuppression", ...
"A2 level-dependent MPI suppression is enabled together with another DC/MPI suppression path; both corrections will be combined.");
end
end end
@@ -135,8 +198,15 @@ classdef FFE < handle
obj.P = (1/delta) * eye(obj.order); obj.P = (1/delta) * eye(obj.order);
end end
if obj.optimize_dc_params if obj.optimize_dc_tracking_params
obj.optimizeDcParams(X.signal,D.signal); obj.optimizeDcTrackingParams(X.signal,D.signal);
obj.e = zeros(obj.order,1);
obj.e_dc = 0;
obj.P = (1/delta) * eye(obj.order);
end
if obj.optimize_a2_level_weights
obj.optimizeA2LevelWeights(X.signal,D.signal);
obj.e = zeros(obj.order,1); obj.e = zeros(obj.order,1);
obj.e_dc = 0; obj.e_dc = 0;
obj.P = (1/delta) * eye(obj.order); obj.P = (1/delta) * eye(obj.order);
@@ -212,13 +282,56 @@ classdef FFE < handle
P_err = 0; P_err = 0;
err_prev = 0; err_prev = 0;
dc_alpha = obj.dc_alpha; dc_tracking_alpha = obj.dc_tracking_alpha;
dc_gamma = obj.dc_gamma; dc_tracking_gamma = obj.dc_tracking_gamma;
dc_mu_min = obj.dc_mu_min; dc_tracking_mu_min = obj.dc_tracking_mu_min;
dc_mu_max = obj.dc_mu_max; dc_tracking_mu_max = obj.dc_tracking_mu_max;
dc_mu_eff_min = obj.dc_mu_eff_min; dc_tracking_mu_eff_min = obj.dc_tracking_mu_eff_min;
dc_mu_eff_max = obj.dc_mu_eff_max; dc_tracking_mu_eff_max = obj.dc_tracking_mu_eff_max;
dc_power_exponent = obj.dc_power_exponent; dc_tracking_power_exponent = obj.dc_tracking_power_exponent;
dc_buffer_enabled = obj.dc_tracking_mu ~= 0 && obj.dc_tracking_buffer_len > 1;
if dc_buffer_enabled
e_dc_buffer = NaN(obj.dc_tracking_buffer_len,1);
end
ffe_buffer_enabled = obj.ffe_update_buffer_len > 1 && obj.adaption_technique ~= adaption_method.rls;
if ffe_buffer_enabled
ffe_update_buffer = NaN(obj.order,obj.ffe_update_buffer_len);
end
dc_avg_enabled = obj.dc_avg_bufferlength_a1 > 1;
if dc_avg_enabled
dc_avg_buffer_a1 = NaN(obj.dc_avg_bufferlength_a1,1);
dc_avg_est_a1 = 0;
dc_avg_update_blocklength_a1 = obj.dc_avg_bufferlength_a1;
dc_avg_update_blocklength_a1 = max(1,floor(dc_avg_update_blocklength_a1));
dc_avg_input_offset_a1 = floor(obj.order/2);
% Hardware-like A1 is causal; dc_smoothing_a1 is reserved for offline variants.
end
dc_level_enabled = obj.dc_level_avg_bufferlength_a2 > 1 && any(obj.dc_level_weights_a2(:) ~= 0);
if dc_level_enabled
if isempty(obj.constellation)
builtin("error","FFE:MissingConstellation", ...
"A2 level-dependent MPI suppression requires obj.constellation to be set.");
end
n_levels = numel(obj.constellation);
if isscalar(obj.dc_level_weights_a2)
dc_level_weight_by_level = repmat(obj.dc_level_weights_a2,n_levels,1);
elseif numel(obj.dc_level_weights_a2) == n_levels
dc_level_weight_by_level = obj.dc_level_weights_a2(:);
else
builtin("error","FFE:InvalidDCLevelWeights", ...
"dc_level_weights_a2 must be scalar or have one entry per constellation level.");
end
dc_level_err_buffer = NaN(n_levels,obj.dc_level_avg_bufferlength_a2);
dc_level_mpi_est_by_level = zeros(n_levels,1);
dc_level_valid_count_by_level = zeros(n_levels,1);
dc_level_update_blocklength_a2 = obj.dc_level_avg_bufferlength_a2;
dc_level_update_blocklength_a2 = max(1,floor(dc_level_update_blocklength_a2));
dc_level_window_future_fraction = obj.dc_smoothing_a2; %#ok<NASGU> % reserved for delayed/offline A2 variants
end
debug_enabled = obj.save_debug; debug_enabled = obj.save_debug;
if debug_enabled if debug_enabled
@@ -228,8 +341,14 @@ classdef FFE < handle
obj.debug_struct.main_cursor = NaN(1,n_symbols_debug); obj.debug_struct.main_cursor = NaN(1,n_symbols_debug);
obj.debug_struct.mu_nlms = NaN(1,n_symbols_debug); obj.debug_struct.mu_nlms = NaN(1,n_symbols_debug);
obj.debug_struct.update_gradient = NaN(1,n_symbols_debug); obj.debug_struct.update_gradient = NaN(1,n_symbols_debug);
obj.debug_struct.mu_dc_eff = NaN(1,n_symbols_debug); obj.debug_struct.dc_tracking_mu_eff = NaN(1,n_symbols_debug);
obj.debug_struct.e_dc_eff = NaN(1,n_symbols_debug); obj.debug_struct.dc_tracking_est = NaN(1,n_symbols_debug);
obj.debug_struct.dc_avg_offset = NaN(1,n_symbols_debug);
obj.debug_struct.dc_level_mpi_est = NaN(1,n_symbols_debug);
obj.debug_struct.dc_level_weight = NaN(1,n_symbols_debug);
obj.debug_struct.dc_level_valid_count = NaN(1,n_symbols_debug);
obj.debug_struct.dc_level_symbol_idx = NaN(1,n_symbols_debug);
obj.debug_struct.dc_level_y_raw = NaN(1,n_symbols_debug);
if training if training
obj.debug_struct.error_tr = NaN(1,n_symbols_debug); obj.debug_struct.error_tr = NaN(1,n_symbols_debug);
@@ -246,14 +365,38 @@ classdef FFE < handle
for sample = 1 : obj.sps : N for sample = 1 : obj.sps : N
symbol = symbol+1; symbol = symbol+1;
mu_dc_eff = 0; dc_tracking_mu_eff = 0;
dc_avg_offset = 0;
dc_level_mpi_est = 0;
dc_level_weight = 0;
dc_level_valid_count = 0;
dc_level_symbol_idx = NaN;
U = x(obj.order+sample-1:-1:sample); U = x(obj.order+sample-1:-1:sample);
if dc_avg_enabled
dc_avg_offset = dc_avg_est_a1;
U = U - dc_avg_offset;
end
y(symbol,1) = obj.e_dc + (obj.e.*mask).' * U; % Calculating output of LMS __ * | y(symbol,1) = obj.e_dc + (obj.e.*mask).' * U; % Calculating output of LMS __ * |
y_raw = y(symbol);
if dc_avg_enabled
dc_avg_input_idx = dc_avg_input_offset_a1 + sample;
dc_avg_buffer_a1 = circshift(dc_avg_buffer_a1,1);
dc_avg_buffer_a1(1) = x(dc_avg_input_idx);
if mod(symbol,dc_avg_update_blocklength_a1) == 0
dc_avg_est_a1 = mean(dc_avg_buffer_a1,"omitnan");
end
end
if training if training
d_hat(symbol,1) = d(symbol); d_hat(symbol,1) = d(symbol);
if isempty(obj.constellation)
symbol_idx = NaN;
else
[~,symbol_idx] = min(abs(d_hat(symbol) - obj.constellation));
end
else else
if ~always_ideal_decision if ~always_ideal_decision
[~,symbol_idx] = min(abs(y(symbol) - obj.constellation)); % decision for closest constellation point [~,symbol_idx] = min(abs(y(symbol) - obj.constellation)); % decision for closest constellation point
@@ -263,6 +406,35 @@ classdef FFE < handle
end end
end end
dc_level_symbol_idx = symbol_idx;
if dc_level_enabled
mpi_err = y_raw - d_hat(symbol);
dc_level_mpi_est = dc_level_mpi_est_by_level(symbol_idx);
dc_level_valid_count = dc_level_valid_count_by_level(symbol_idx);
dc_level_weight = dc_level_weight_by_level(symbol_idx) * ...
min(dc_level_valid_count / obj.dc_level_avg_bufferlength_a2,1);
y(symbol,1) = y_raw - dc_level_weight * dc_level_mpi_est;
if training
d_hat(symbol,1) = d(symbol);
else
if ~always_ideal_decision
[~,symbol_idx] = min(abs(y(symbol) - obj.constellation));
d_hat(symbol,1) = obj.constellation(symbol_idx);
else
d_hat(symbol,1) = d(symbol);
end
end
dc_level_err_buffer(dc_level_symbol_idx,:) = circshift(dc_level_err_buffer(dc_level_symbol_idx,:),1,2);
dc_level_err_buffer(dc_level_symbol_idx,1) = mpi_err;
if mod(symbol,dc_level_update_blocklength_a2) == 0
dc_level_valid_count_by_level = sum(isfinite(dc_level_err_buffer),2);
dc_level_mpi_est_by_level = mean(dc_level_err_buffer,2,"omitnan");
dc_level_mpi_est_by_level(dc_level_valid_count_by_level == 0) = 0;
end
end
% err(symbol) = y(symbol) - d_hat(symbol); % Instantaneous error % err(symbol) = y(symbol) - d_hat(symbol); % Instantaneous error
err(symbol) = d_hat(symbol) - y(symbol); % Instantaneous error err(symbol) = d_hat(symbol) - y(symbol); % Instantaneous error
@@ -278,7 +450,6 @@ classdef FFE < handle
weight = mu / normU; weight = mu / normU;
grad = err(symbol) * U; grad = err(symbol) * U;
update = grad * weight; update = grad * weight;
obj.e = obj.e + update;
case adaption_method.lms case adaption_method.lms
@@ -287,7 +458,6 @@ classdef FFE < handle
weight = mu; weight = mu;
grad = err(symbol) * U; grad = err(symbol) * U;
update = grad * weight; update = grad * weight;
obj.e = obj.e + update;
@@ -300,27 +470,47 @@ classdef FFE < handle
% Gewichtsupdate: % Gewichtsupdate:
update = k * err(symbol); update = k * err(symbol);
obj.e = obj.e + update;
% P-Matrix‐Update: % P-Matrix‐Update:
obj.P = (1/lambda) * (obj.P - k * (U.' * obj.P));
end end
if obj.mu_dc ~= 0 if ffe_buffer_enabled
if obj.adaptive_dc_enabled ffe_update_buffer = circshift(ffe_update_buffer,1,2);
delta_mu = dc_gamma * err(symbol) * err_prev * (U.'*U); ffe_update_buffer(:,1) = update;
obj.mu_dc = min(max(obj.mu_dc + delta_mu,dc_mu_min),dc_mu_max); if mod(symbol,obj.ffe_update_buffer_len) == 0
obj.e = obj.e + mean(ffe_update_buffer,2,"omitnan");
end
else
obj.e = obj.e + update;
end
if obj.adaption_technique == adaption_method.rls
obj.P = (1/lambda) * (obj.P - k * (U.' * obj.P));
end
if obj.dc_tracking_mu ~= 0
if obj.dc_tracking_adaptive_enabled
delta_mu = dc_tracking_gamma * err(symbol) * err_prev * (U.'*U);
obj.dc_tracking_mu = min(max(obj.dc_tracking_mu + delta_mu,dc_tracking_mu_min),dc_tracking_mu_max);
err_prev = err(symbol); err_prev = err(symbol);
P_err = dc_alpha*P_err + (1-dc_alpha)*err(symbol)^2; P_err = dc_tracking_alpha*P_err + (1-dc_tracking_alpha)*err(symbol)^2;
mu_dc_eff = obj.mu_dc / ((P_err + eps)^dc_power_exponent); dc_tracking_mu_eff = obj.dc_tracking_mu / ((P_err + eps)^dc_tracking_power_exponent);
else else
mu_dc_eff = obj.mu_dc; dc_tracking_mu_eff = obj.dc_tracking_mu;
end end
mu_dc_eff = min(max(mu_dc_eff,dc_mu_eff_min),dc_mu_eff_max); dc_tracking_mu_eff = min(max(dc_tracking_mu_eff,dc_tracking_mu_eff_min),dc_tracking_mu_eff_max);
obj.e_dc = obj.e_dc + mu_dc_eff * err(symbol); if dc_buffer_enabled
e_dc_buffer(1) = obj.e_dc + dc_tracking_mu_eff * err(symbol);
e_dc_buffer = circshift(e_dc_buffer,1);
if mod(symbol,obj.dc_tracking_buffer_len) == 0
obj.e_dc = mean(e_dc_buffer,"omitnan");
end
else
obj.e_dc = obj.e_dc + dc_tracking_mu_eff * err(symbol);
end
end end
end end
@@ -336,8 +526,14 @@ classdef FFE < handle
obj.debug_struct.main_cursor(1,symbol) = abs(obj.e(maincursor_pos)); obj.debug_struct.main_cursor(1,symbol) = abs(obj.e(maincursor_pos));
obj.debug_struct.mu_nlms(1,symbol) = weight; obj.debug_struct.mu_nlms(1,symbol) = weight;
obj.debug_struct.update_gradient(1,symbol) = grad.'*grad; obj.debug_struct.update_gradient(1,symbol) = grad.'*grad;
obj.debug_struct.mu_dc_eff(1,symbol) = mu_dc_eff; obj.debug_struct.dc_tracking_mu_eff(1,symbol) = dc_tracking_mu_eff;
obj.debug_struct.e_dc_eff(1,symbol) = obj.e_dc; obj.debug_struct.dc_tracking_est(1,symbol) = obj.e_dc;
obj.debug_struct.dc_avg_offset(1,symbol) = dc_avg_offset;
obj.debug_struct.dc_level_mpi_est(1,symbol) = dc_level_mpi_est;
obj.debug_struct.dc_level_weight(1,symbol) = dc_level_weight;
obj.debug_struct.dc_level_valid_count(1,symbol) = dc_level_valid_count;
obj.debug_struct.dc_level_symbol_idx(1,symbol) = dc_level_symbol_idx;
obj.debug_struct.dc_level_y_raw(1,symbol) = y_raw;
if training if training
obj.debug_struct.error_tr(1,symbol) = error_power; obj.debug_struct.error_tr(1,symbol) = error_power;
@@ -364,16 +560,16 @@ classdef FFE < handle
case adaption_method.rls case adaption_method.rls
mu_range = [0.98, 0.99999]; mu_range = [0.98, 0.99999];
end end
mu_dc_range = [1e-5, 1e-1]; dc_tracking_mu_range = [1e-5, 1e-1];
mu_tr_var = optimizableVariable("mu_tr",mu_range,"Transform","log"); mu_tr_var = optimizableVariable("mu_tr",mu_range,"Transform","log");
vars = mu_tr_var; vars = mu_tr_var;
if obj.dd_mode if obj.dd_mode
vars = [vars, optimizableVariable("mu_dd",mu_range,"Transform","log")]; vars = [vars, optimizableVariable("mu_dd",mu_range,"Transform","log")];
end end
optimize_mu_dc = obj.mu_dc ~= 0; optimize_dc_tracking_mu = obj.dc_tracking_mu ~= 0;
if optimize_mu_dc if optimize_dc_tracking_mu
vars = [vars, optimizableVariable("mu_dc",mu_dc_range,"Transform","log")]; vars = [vars, optimizableVariable("dc_tracking_mu",dc_tracking_mu_range,"Transform","log")];
end end
obj.mu_optimization_iter = 0; obj.mu_optimization_iter = 0;
fprintf("FFE mu opt uses %d samples / %d symbols\n",N_opt,numel(d_opt)); fprintf("FFE mu opt uses %d samples / %d symbols\n",N_opt,numel(d_opt));
@@ -387,18 +583,18 @@ classdef FFE < handle
if obj.dd_mode if obj.dd_mode
obj.mu_dd = obj.mu_optimization.XAtMinObjective.mu_dd; obj.mu_dd = obj.mu_optimization.XAtMinObjective.mu_dd;
end end
if optimize_mu_dc if optimize_dc_tracking_mu
obj.mu_dc = obj.mu_optimization.XAtMinObjective.mu_dc; obj.dc_tracking_mu = obj.mu_optimization.XAtMinObjective.dc_tracking_mu;
end end
if obj.dd_mode && optimize_mu_dc if obj.dd_mode && optimize_dc_tracking_mu
fprintf("\nFFE mu opt done: mu_tr=%9.3e, mu_dd=%9.3e, mu_dc=%9.3e, BER=%9.3e\n", ... fprintf("\nFFE mu opt done: mu_tr=%9.3e, mu_dd=%9.3e, dc_tracking_mu=%9.3e, BER=%9.3e\n", ...
obj.mu_tr,obj.mu_dd,obj.mu_dc,obj.mu_optimization.MinObjective); obj.mu_tr,obj.mu_dd,obj.dc_tracking_mu,obj.mu_optimization.MinObjective);
elseif obj.dd_mode elseif obj.dd_mode
fprintf("\nFFE mu opt done: mu_tr=%9.3e, mu_dd=%9.3e, BER=%9.3e\n", ... fprintf("\nFFE mu opt done: mu_tr=%9.3e, mu_dd=%9.3e, BER=%9.3e\n", ...
obj.mu_tr,obj.mu_dd,obj.mu_optimization.MinObjective); obj.mu_tr,obj.mu_dd,obj.mu_optimization.MinObjective);
elseif optimize_mu_dc elseif optimize_dc_tracking_mu
fprintf("\nFFE mu opt done: mu_tr=%9.3e, mu_dc=%9.3e, BER=%9.3e\n", ... fprintf("\nFFE mu opt done: mu_tr=%9.3e, dc_tracking_mu=%9.3e, BER=%9.3e\n", ...
obj.mu_tr,obj.mu_dc,obj.mu_optimization.MinObjective); obj.mu_tr,obj.dc_tracking_mu,obj.mu_optimization.MinObjective);
else else
fprintf("\nFFE mu opt done: mu_tr=%9.3e, BER=%9.3e\n", ... fprintf("\nFFE mu opt done: mu_tr=%9.3e, BER=%9.3e\n", ...
obj.mu_tr,obj.mu_optimization.MinObjective); obj.mu_tr,obj.mu_optimization.MinObjective);
@@ -409,43 +605,99 @@ classdef FFE < handle
end end
end end
function optimizeDcParams(obj,x,d) function optimizeDcTrackingParams(obj,x,d)
[x_opt,d_opt,N_opt] = obj.optimizationSignals(x,d,obj.dc_optimization_len); [x_opt,d_opt,N_opt] = obj.optimizationSignals(x,d,obj.dc_tracking_optimization_len);
vars = optimizableVariable("mu_dc",[1e-5,1e-1],"Transform","log"); vars = optimizableVariable("dc_tracking_mu",[1e-5,1e-1],"Transform","log");
if obj.adaptive_dc_enabled if obj.dc_tracking_adaptive_enabled
vars = [vars, ... vars = [vars, ...
optimizableVariable("dc_alpha",[0.85,0.995]), ... optimizableVariable("dc_tracking_alpha",[0.85,0.995]), ...
optimizableVariable("dc_gamma",[1e-7,3e-5],"Transform","log"), ... optimizableVariable("dc_tracking_gamma",[1e-7,3e-5],"Transform","log"), ...
optimizableVariable("dc_power_exponent",[0,1]), ... optimizableVariable("dc_tracking_power_exponent",[0,2]), ...
optimizableVariable("dc_mu_eff_max",[1e-3,3e-1],"Transform","log")]; optimizableVariable("dc_tracking_mu_eff_max",[1e-3,3e-1],"Transform","log")];
end end
obj.dc_optimization_iter = 0; obj.dc_tracking_optimization_iter = 0;
fprintf("FFE DC opt uses fixed mu_tr=%9.3e, mu_dd=%9.3e on %d samples / %d symbols\n", ... fprintf("FFE DC opt uses fixed mu_tr=%9.3e, mu_dd=%9.3e on %d samples / %d symbols\n", ...
obj.mu_tr,obj.mu_dd,N_opt,numel(d_opt)); obj.mu_tr,obj.mu_dd,N_opt,numel(d_opt));
obj.dc_optimization = bayesopt(@(p)obj.dcObjective(p,x_opt,d_opt),vars, ... obj.dc_tracking_optimization = bayesopt(@(p)obj.dcTrackingObjective(p,x_opt,d_opt),vars, ...
"MaxObjectiveEvaluations",obj.dc_optimization_max_evals, ... "MaxObjectiveEvaluations",obj.dc_tracking_optimization_max_evals, ...
"AcquisitionFunctionName","expected-improvement-plus", ... "AcquisitionFunctionName","expected-improvement-plus", ...
"IsObjectiveDeterministic",false, ... "IsObjectiveDeterministic",false, ...
"Verbose",0, ... "Verbose",0, ...
"PlotFcn",[]); "PlotFcn",[]);
best = obj.dc_optimization.XAtMinObjective; best = obj.dc_tracking_optimization.XAtMinObjective;
obj.mu_dc = best.mu_dc; obj.dc_tracking_mu = best.dc_tracking_mu;
if obj.adaptive_dc_enabled if obj.dc_tracking_adaptive_enabled
obj.dc_alpha = best.dc_alpha; obj.dc_tracking_alpha = best.dc_tracking_alpha;
obj.dc_gamma = best.dc_gamma; obj.dc_tracking_gamma = best.dc_tracking_gamma;
obj.dc_power_exponent = best.dc_power_exponent; obj.dc_tracking_power_exponent = best.dc_tracking_power_exponent;
obj.dc_mu_eff_max = best.dc_mu_eff_max; obj.dc_tracking_mu_eff_max = best.dc_tracking_mu_eff_max;
fprintf("\nFFE DC opt done: mu_dc=%9.3e, alpha=%6.3f, gamma=%9.3e, p=%5.2f, mu_eff_max=%9.3e, objective=%9.3e\n", ... fprintf("\nFFE DC opt done: dc_tracking_mu=%9.3e, alpha=%6.3f, gamma=%9.3e, p=%5.2f, mu_eff_max=%9.3e, objective=%9.3e\n", ...
obj.mu_dc,obj.dc_alpha,obj.dc_gamma,obj.dc_power_exponent,obj.dc_mu_eff_max,obj.dc_optimization.MinObjective); obj.dc_tracking_mu,obj.dc_tracking_alpha,obj.dc_tracking_gamma,obj.dc_tracking_power_exponent,obj.dc_tracking_mu_eff_max,obj.dc_tracking_optimization.MinObjective);
else else
fprintf("\nFFE DC opt done: mu_dc=%9.3e, objective=%9.3e\n", ... fprintf("\nFFE DC opt done: dc_tracking_mu=%9.3e, objective=%9.3e\n", ...
obj.mu_dc,obj.dc_optimization.MinObjective); obj.dc_tracking_mu,obj.dc_tracking_optimization.MinObjective);
end end
end end
function optimizeA2LevelWeights(obj,x,d)
if obj.dc_level_avg_bufferlength_a2 <= 1
warning("FFE:A2OptimizationDisabled", ...
"A2 level-weight optimization requires dc_level_avg_bufferlength_a2 > 1.");
return
end
if obj.a2_level_weight_max <= 0
warning("FFE:A2OptimizationDisabled", ...
"A2 level-weight optimization requires a2_level_weight_max > 0.");
return
end
if isempty(obj.constellation)
obj.constellation = unique(d);
end
[x_opt,d_opt,N_opt] = obj.optimizationSignals(x,d,obj.a2_level_weight_optimization_len);
n_levels = numel(obj.constellation);
var_names = compose("dc_level_weight_a2_%d",1:n_levels);
vars_cell = cell(1,n_levels);
for level_idx = 1:n_levels
vars_cell{level_idx} = optimizableVariable(var_names(level_idx),[0,obj.a2_level_weight_max]);
end
vars = [vars_cell{:}];
[initial_weights,stats] = obj.a2LevelWeightInitialGuess(x_opt,d_opt);
current_weights = obj.expandA2LevelWeights(n_levels);
initial_matrix = initial_weights(:).';
if any(current_weights ~= 0)
initial_matrix = [initial_matrix; current_weights(:).'];
end
initial_matrix = min(max(initial_matrix,0),obj.a2_level_weight_max);
initial_matrix = unique(initial_matrix,"rows","stable");
initial_x = array2table(initial_matrix,"VariableNames",cellstr(var_names));
obj.a2_level_weight_initial_stats = stats;
obj.a2_level_weight_optimization_iter = 0;
max_evals = max(obj.a2_level_weight_optimization_max_evals,height(initial_x));
fprintf("FFE A2 opt uses fixed mu_tr=%9.3e, mu_dd=%9.3e on %d samples / %d symbols\n", ...
obj.mu_tr,obj.mu_dd,N_opt,numel(d_opt));
fprintf("FFE A2 opt init: var_slope=%9.3e, weights=%s\n", ...
stats.variance_slope,mat2str(initial_weights(:).',3));
obj.a2_level_weight_optimization = bayesopt(@(p)obj.a2LevelWeightObjective(p,x_opt,d_opt),vars, ...
"MaxObjectiveEvaluations",max_evals, ...
"InitialX",initial_x, ...
"AcquisitionFunctionName","expected-improvement-plus", ...
"IsObjectiveDeterministic",false, ...
"Verbose",0, ...
"PlotFcn",[]);
best = obj.a2_level_weight_optimization.XAtMinObjective;
obj.dc_level_weights_a2 = obj.a2LevelWeightsFromParams(best);
fprintf("\nFFE A2 opt done: weights=%s, BER=%9.3e\n", ...
mat2str(obj.dc_level_weights_a2(:).',3),obj.a2_level_weight_optimization.MinObjective);
end
function [x_opt,d_opt,N_opt] = optimizationSignals(obj,x,d,opt_len) function [x_opt,d_opt,N_opt] = optimizationSignals(obj,x,d,opt_len)
if nargin < 4 if nargin < 4
opt_len = obj.mu_optimization_len; opt_len = obj.mu_optimization_len;
@@ -469,20 +721,22 @@ classdef FFE < handle
function objective = muObjective(obj,params,x,d) function objective = muObjective(obj,params,x,d)
old_debug = obj.save_debug; old_debug = obj.save_debug;
old_mu_dc = obj.mu_dc; old_dc_tracking_mu = obj.dc_tracking_mu;
obj.save_debug = 0; obj.save_debug = 0;
has_mu_dc = any(strcmp(params.Properties.VariableNames,"mu_dc")); has_dc_tracking_mu = any(strcmp(params.Properties.VariableNames,"dc_tracking_mu"));
if has_mu_dc if has_dc_tracking_mu
obj.mu_dc = params.mu_dc; obj.dc_tracking_mu = params.dc_tracking_mu;
end end
obj.e = zeros(obj.order,1); obj.e = zeros(obj.order,1);
obj.e_dc = 0; obj.e_dc = 0;
obj.P = (1/0.05) * eye(obj.order); obj.P = (1/0.05) * eye(obj.order);
obj.debug_struct = struct(); obj.debug_struct = struct();
N_tr = min(obj.len_tr,numel(x)); N_tr = min(obj.len_tr,numel(x));
[signal,~] = obj.equalize(x,d,params.mu_tr,obj.epochs_tr,N_tr,1,0); obj.equalize(x,d,params.mu_tr,obj.epochs_tr,N_tr,1,0);
if obj.dd_mode if obj.dd_mode
[signal,~] = obj.equalize(x,d,params.mu_dd,obj.epochs_dd,numel(x),0,0); [signal,~] = obj.equalize(x,d,params.mu_dd,obj.epochs_dd,numel(x),0,0);
else
[signal,~] = obj.equalize(x,d,0,1,numel(x),0,0);
end end
[ber,errors] = obj.berObjective(signal,d); [ber,errors] = obj.berObjective(signal,d);
@@ -491,28 +745,58 @@ classdef FFE < handle
objective = inf; objective = inf;
end end
obj.mu_optimization_iter = obj.mu_optimization_iter + 1; obj.mu_optimization_iter = obj.mu_optimization_iter + 1;
if obj.dd_mode && has_mu_dc if obj.dd_mode && has_dc_tracking_mu
fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, mu_dd=%9.3e, mu_dc=%9.3e, BER=%9.3e, errors=%d", ... fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, mu_dd=%9.3e, dc_tracking_mu=%9.3e, BER=%9.3e, errors=%d", ...
obj.mu_optimization_iter,params.mu_tr,params.mu_dd,params.mu_dc,ber,errors); obj.mu_optimization_iter,params.mu_tr,params.mu_dd,params.dc_tracking_mu,ber,errors);
elseif obj.dd_mode elseif obj.dd_mode
fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, mu_dd=%9.3e, BER=%9.3e, errors=%d", ... fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, mu_dd=%9.3e, BER=%9.3e, errors=%d", ...
obj.mu_optimization_iter,params.mu_tr,params.mu_dd,ber,errors); obj.mu_optimization_iter,params.mu_tr,params.mu_dd,ber,errors);
elseif has_mu_dc elseif has_dc_tracking_mu
fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, mu_dc=%9.3e, BER=%9.3e, errors=%d", ... fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, dc_tracking_mu=%9.3e, BER=%9.3e, errors=%d", ...
obj.mu_optimization_iter,params.mu_tr,params.mu_dc,ber,errors); obj.mu_optimization_iter,params.mu_tr,params.dc_tracking_mu,ber,errors);
else else
fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, BER=%9.3e, errors=%d", ... fprintf("\rFFE mu opt %02d: mu_tr=%9.3e, BER=%9.3e, errors=%d", ...
obj.mu_optimization_iter,params.mu_tr,ber,errors); obj.mu_optimization_iter,params.mu_tr,ber,errors);
end end
obj.save_debug = old_debug; obj.save_debug = old_debug;
obj.mu_dc = old_mu_dc; obj.dc_tracking_mu = old_dc_tracking_mu;
end end
function objective = dcObjective(obj,params,x,d) function objective = a2LevelWeightObjective(obj,params,x,d)
state = obj.captureObjectiveState(); state = obj.captureObjectiveState();
cleanup = onCleanup(@()obj.restoreObjectiveState(state)); cleanup = onCleanup(@()obj.restoreObjectiveState(state));
obj.applyDcObjectiveParams(params); obj.dc_level_weights_a2 = obj.a2LevelWeightsFromParams(params);
obj.save_debug = 0;
obj.e = zeros(obj.order,1);
obj.e_dc = 0;
obj.P = (1/0.05) * eye(obj.order);
obj.debug_struct = struct();
N_tr = min(obj.len_tr,numel(x));
obj.equalize(x,d,obj.mu_tr,obj.epochs_tr,N_tr,1,0);
if obj.dd_mode
[signal,~] = obj.equalize(x,d,obj.mu_dd,obj.epochs_dd,numel(x),0,0);
else
[signal,~] = obj.equalize(x,d,0,1,numel(x),0,0);
end
[ber,errors] = obj.berObjective(signal,d);
objective = ber;
if ~isfinite(objective)
objective = inf;
end
obj.a2_level_weight_optimization_iter = obj.a2_level_weight_optimization_iter + 1;
fprintf("\rFFE A2 opt %02d: weights=%s, BER=%9.3e, errors=%d", ...
obj.a2_level_weight_optimization_iter,mat2str(obj.dc_level_weights_a2(:).',3),ber,errors);
end
function objective = dcTrackingObjective(obj,params,x,d)
state = obj.captureObjectiveState();
cleanup = onCleanup(@()obj.restoreObjectiveState(state));
obj.applyDcTrackingObjectiveParams(params);
obj.save_debug = 1; obj.save_debug = 1;
obj.e = zeros(obj.order,1); obj.e = zeros(obj.order,1);
obj.e_dc = 0; obj.e_dc = 0;
@@ -520,60 +804,165 @@ classdef FFE < handle
obj.debug_struct = struct(); obj.debug_struct = struct();
N_tr = min(obj.len_tr,numel(x)); N_tr = min(obj.len_tr,numel(x));
[signal,~] = obj.equalize(x,d,obj.mu_tr,obj.epochs_tr,N_tr,1,0); obj.equalize(x,d,obj.mu_tr,obj.epochs_tr,N_tr,1,0);
if obj.dd_mode if obj.dd_mode
[signal,~] = obj.equalize(x,d,obj.mu_dd,obj.epochs_dd,numel(x),0,0); [signal,~] = obj.equalize(x,d,obj.mu_dd,obj.epochs_dd,numel(x),0,0);
else
[signal,~] = obj.equalize(x,d,0,1,numel(x),0,0);
end end
[ber,errors] = obj.berObjective(signal,d); [ber,errors] = obj.berObjective(signal,d);
[delay_symbols,delay_corr] = obj.dcDelayObjective(signal,d); [delay_symbols,delay_corr] = obj.dcTrackingDelayObjective(signal,d);
delay_penalty = obj.dc_optimization_delay_weight * abs(delay_symbols) / max(numel(d),1); delay_penalty = obj.dc_tracking_optimization_delay_weight * abs(delay_symbols) / max(numel(d),1);
objective = ber + delay_penalty; objective = ber + delay_penalty;
if ~isfinite(objective) if ~isfinite(objective)
objective = inf; objective = inf;
end end
obj.dc_optimization_iter = obj.dc_optimization_iter + 1; obj.dc_tracking_optimization_iter = obj.dc_tracking_optimization_iter + 1;
if obj.adaptive_dc_enabled if obj.dc_tracking_adaptive_enabled
fprintf("\rFFE DC opt %02d: mu_dc=%9.3e, alpha=%6.3f, gamma=%9.3e, p=%5.2f, mu_eff_max=%9.3e, BER=%9.3e, delay=%7.0f, corr=%6.3f, obj=%9.3e, errors=%d", ... fprintf("\rFFE DC opt %02d: dc_tracking_mu=%9.3e, alpha=%6.3f, gamma=%9.3e, p=%5.2f, mu_eff_max=%9.3e, BER=%9.3e, delay=%7.0f, corr=%6.3f, obj=%9.3e, errors=%d", ...
obj.dc_optimization_iter,params.mu_dc,params.dc_alpha,params.dc_gamma,params.dc_power_exponent,params.dc_mu_eff_max, ... obj.dc_tracking_optimization_iter,params.dc_tracking_mu,params.dc_tracking_alpha,params.dc_tracking_gamma,params.dc_tracking_power_exponent,params.dc_tracking_mu_eff_max, ...
ber,delay_symbols,delay_corr,objective,errors); ber,delay_symbols,delay_corr,objective,errors);
else else
fprintf("\rFFE DC opt %02d: mu_dc=%9.3e, BER=%9.3e, delay=%7.0f, corr=%6.3f, obj=%9.3e, errors=%d", ... fprintf("\rFFE DC opt %02d: dc_tracking_mu=%9.3e, BER=%9.3e, delay=%7.0f, corr=%6.3f, obj=%9.3e, errors=%d", ...
obj.dc_optimization_iter,params.mu_dc,ber,delay_symbols,delay_corr,objective,errors); obj.dc_tracking_optimization_iter,params.dc_tracking_mu,ber,delay_symbols,delay_corr,objective,errors);
end end
end end
function applyDcObjectiveParams(obj,params) function applyDcTrackingObjectiveParams(obj,params)
var_names = string(params.Properties.VariableNames); var_names = string(params.Properties.VariableNames);
if any(var_names == "mu_dc") if any(var_names == "dc_tracking_mu")
obj.mu_dc = params.mu_dc; obj.dc_tracking_mu = params.dc_tracking_mu;
end end
if any(var_names == "dc_alpha") if any(var_names == "dc_tracking_alpha")
obj.dc_alpha = params.dc_alpha; obj.dc_tracking_alpha = params.dc_tracking_alpha;
end end
if any(var_names == "dc_gamma") if any(var_names == "dc_tracking_gamma")
obj.dc_gamma = params.dc_gamma; obj.dc_tracking_gamma = params.dc_tracking_gamma;
end end
if any(var_names == "dc_power_exponent") if any(var_names == "dc_tracking_power_exponent")
obj.dc_power_exponent = params.dc_power_exponent; obj.dc_tracking_power_exponent = params.dc_tracking_power_exponent;
end end
if any(var_names == "dc_mu_eff_max") if any(var_names == "dc_tracking_mu_eff_max")
obj.dc_mu_eff_max = params.dc_mu_eff_max; obj.dc_tracking_mu_eff_max = params.dc_tracking_mu_eff_max;
end end
end end
function weights = a2LevelWeightsFromParams(~,params)
var_names = string(params.Properties.VariableNames);
weight_names = var_names(startsWith(var_names,"dc_level_weight_a2_"));
weight_idx = extractAfter(weight_names,"dc_level_weight_a2_");
weight_idx = str2double(weight_idx);
[weight_idx,sort_idx] = sort(weight_idx);
weight_names = weight_names(sort_idx);
weights = zeros(numel(weight_names),1);
for n = 1:numel(weight_names)
weights(weight_idx(n),1) = params.(char(weight_names(n)));
end
end
function weights = expandA2LevelWeights(obj,n_levels)
if isscalar(obj.dc_level_weights_a2)
weights = repmat(obj.dc_level_weights_a2,n_levels,1);
elseif numel(obj.dc_level_weights_a2) == n_levels
weights = obj.dc_level_weights_a2(:);
else
builtin("error","FFE:InvalidDCLevelWeights", ...
"dc_level_weights_a2 must be scalar or have one entry per constellation level.");
end
weights = min(max(weights,0),obj.a2_level_weight_max);
end
function [weights,stats] = a2LevelWeightInitialGuess(obj,x,d)
levels = obj.constellation(:);
if isempty(levels)
levels = unique(d(:));
end
n_levels = numel(levels);
n_symbols = min(numel(d),floor(numel(x) / obj.sps));
d = d(1:n_symbols);
rx_symbols = x(1:obj.sps:obj.sps*n_symbols);
rx_symbols = rx_symbols(:);
level_mean = NaN(n_levels,1);
level_variance = NaN(n_levels,1);
for level_idx = 1:n_levels
level_samples = rx_symbols(d == levels(level_idx));
level_mean(level_idx) = mean(level_samples,"omitnan");
level_variance(level_idx) = var(level_samples,0,"omitnan");
end
valid = isfinite(level_mean) & isfinite(level_variance);
variance_fit = level_variance;
slope_axis = "mean";
if nnz(valid) >= 2
if numel(unique(level_mean(valid))) >= 2
p = polyfit(level_mean(valid),level_variance(valid),1);
variance_fit = polyval(p,level_mean);
else
slope_axis = "level";
level_axis = (1:n_levels).';
p = polyfit(level_axis(valid),level_variance(valid),1);
variance_fit = polyval(p,(1:n_levels).');
end
else
p = [0, mean(level_variance(valid),"omitnan")];
end
if any(valid)
baseline = min(variance_fit(valid),[],"omitnan");
else
baseline = 0;
end
variance_score = max(variance_fit - baseline,0);
finite_score = isfinite(variance_score);
if ~any(finite_score) || max(variance_score(finite_score),[],"omitnan") <= 0
if any(valid)
baseline = min(level_variance(valid),[],"omitnan");
else
baseline = 0;
end
variance_score = max(level_variance - baseline,0);
finite_score = isfinite(variance_score);
end
weights = zeros(n_levels,1);
usable_score = valid & finite_score;
if any(usable_score)
max_score = max(variance_score(usable_score),[],"omitnan");
else
max_score = 0;
end
if isfinite(max_score) && max_score > 0
initial_weight_ceiling = min(0.7,obj.a2_level_weight_max);
weights(usable_score) = initial_weight_ceiling * variance_score(usable_score) ./ max_score;
end
weights = min(max(weights,0),obj.a2_level_weight_max);
stats = struct( ...
"levels",levels, ...
"mean",level_mean, ...
"variance",level_variance, ...
"variance_fit",variance_fit(:), ...
"variance_slope",p(1), ...
"slope_axis",slope_axis, ...
"initial_weights",weights);
end
function state = captureObjectiveState(obj) function state = captureObjectiveState(obj)
state.e = obj.e; state.e = obj.e;
state.e_dc = obj.e_dc; state.e_dc = obj.e_dc;
state.P = obj.P; state.P = obj.P;
state.save_debug = obj.save_debug; state.save_debug = obj.save_debug;
state.debug_struct = obj.debug_struct; state.debug_struct = obj.debug_struct;
state.mu_dc = obj.mu_dc; state.dc_tracking_mu = obj.dc_tracking_mu;
state.dc_alpha = obj.dc_alpha; state.dc_tracking_alpha = obj.dc_tracking_alpha;
state.dc_gamma = obj.dc_gamma; state.dc_tracking_gamma = obj.dc_tracking_gamma;
state.dc_power_exponent = obj.dc_power_exponent; state.dc_tracking_power_exponent = obj.dc_tracking_power_exponent;
state.dc_mu_eff_max = obj.dc_mu_eff_max; state.dc_tracking_mu_eff_max = obj.dc_tracking_mu_eff_max;
state.dc_level_weights_a2 = obj.dc_level_weights_a2;
state.a2_level_weight_initial_stats = obj.a2_level_weight_initial_stats;
end end
function restoreObjectiveState(obj,state) function restoreObjectiveState(obj,state)
@@ -582,11 +971,13 @@ classdef FFE < handle
obj.P = state.P; obj.P = state.P;
obj.save_debug = state.save_debug; obj.save_debug = state.save_debug;
obj.debug_struct = state.debug_struct; obj.debug_struct = state.debug_struct;
obj.mu_dc = state.mu_dc; obj.dc_tracking_mu = state.dc_tracking_mu;
obj.dc_alpha = state.dc_alpha; obj.dc_tracking_alpha = state.dc_tracking_alpha;
obj.dc_gamma = state.dc_gamma; obj.dc_tracking_gamma = state.dc_tracking_gamma;
obj.dc_power_exponent = state.dc_power_exponent; obj.dc_tracking_power_exponent = state.dc_tracking_power_exponent;
obj.dc_mu_eff_max = state.dc_mu_eff_max; obj.dc_tracking_mu_eff_max = state.dc_tracking_mu_eff_max;
obj.dc_level_weights_a2 = state.dc_level_weights_a2;
obj.a2_level_weight_initial_stats = state.a2_level_weight_initial_stats;
end end
function [ber,errors] = berObjective(~,signal,d) function [ber,errors] = berObjective(~,signal,d)
@@ -603,23 +994,23 @@ classdef FFE < handle
"returnErrorLocation",1); "returnErrorLocation",1);
end end
function [delay_symbols,delay_corr] = dcDelayObjective(obj,signal,d) function [delay_symbols,delay_corr] = dcTrackingDelayObjective(obj,signal,d)
delay_symbols = 0; delay_symbols = 0;
delay_corr = 0; delay_corr = 0;
if ~isfield(obj.debug_struct,"e_dc_eff") || isempty(obj.debug_struct.e_dc_eff) if ~isfield(obj.debug_struct,"dc_tracking_est") || isempty(obj.debug_struct.dc_tracking_est)
return return
end end
smooth_len = obj.dc_optimization_smoothing_len; smooth_len = obj.dc_tracking_optimization_smoothing_len;
e_dc_eff_s = movmean(obj.debug_struct.e_dc_eff(:),smooth_len,"omitnan"); dc_tracking_est_s = movmean(obj.debug_struct.dc_tracking_est(:),smooth_len,"omitnan");
avg_lvl_dc = obj.averageLevelTrace(signal,d,smooth_len); avg_lvl_dc = obj.averageLevelTrace(signal,d,smooth_len);
xcorr_len = min(numel(e_dc_eff_s),numel(avg_lvl_dc)); xcorr_len = min(numel(dc_tracking_est_s),numel(avg_lvl_dc));
if xcorr_len < 2 if xcorr_len < 2
return return
end end
inv_dc_xcorr = -e_dc_eff_s(1:xcorr_len); inv_dc_xcorr = -dc_tracking_est_s(1:xcorr_len);
avg_lvl_xcorr = avg_lvl_dc(1:xcorr_len); avg_lvl_xcorr = avg_lvl_dc(1:xcorr_len);
inv_dc_xcorr = fillmissing(inv_dc_xcorr,"linear","EndValues","nearest"); inv_dc_xcorr = fillmissing(inv_dc_xcorr,"linear","EndValues","nearest");
avg_lvl_xcorr = fillmissing(avg_lvl_xcorr,"linear","EndValues","nearest"); avg_lvl_xcorr = fillmissing(avg_lvl_xcorr,"linear","EndValues","nearest");

27
Functions/EQ_blocks/ffe.m
View File

@@ -74,17 +74,22 @@ end
% Create FFE results structure % Create FFE results structure
ffe_results = struct(); ffe_results = struct();
try config_clear_props = { ...
eq_.e = []; "e", ...
eq_.e2 = []; "e2", ...
eq_.e3 = []; "e3", ...
eq_.b = []; "b", ...
eq_.b2 = []; "b2", ...
eq_.b3 = []; "b3", ...
end "mu_optimization", ...
try "dc_optimization", ...
eq_.mu_optimization = []; "dc_tracking_optimization", ...
eq_.dc_optimization = []; "a2_level_weight_optimization"};
for config_clear_idx = 1:numel(config_clear_props)
prop_name = config_clear_props{config_clear_idx};
if isprop(eq_,prop_name)
eq_.(prop_name) = [];
end
end end
ffe_results.config = Equalizerstruct(); ffe_results.config = Equalizerstruct();

2
Functions/EQ_recipes/dsp_recipe_minimal.m
View File

@@ -21,7 +21,7 @@ function output = dsp_recipe_minimal(Scpe_sig_raw, Symbols, Tx_bits, options)
"debug_plots", options.debug_plots); "debug_plots", options.debug_plots);
eq_ffe = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",4096*2,"mu_dd",1e-1,"mu_tr",0.4,"order",25,... eq_ffe = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",4096*2,"mu_dd",1e-1,"mu_tr",0.4,"order",25,...
"sps",2,"decide",0,"optmize_mus",0,"dd_mode",options.userParameters.dd_mode,"adaption_technique","nlms","mu_dc",1.021e-05); "sps",2,"decide",0,"optmize_mus",0,"dd_mode",options.userParameters.dd_mode,"adaption_technique","nlms","dc_tracking_mu",1.021e-05);
ffe_results = ffe(eq_ffe, options.M, Scpe_sig, Symbols, Tx_bits, ... ffe_results = ffe(eq_ffe, options.M, Scpe_sig, Symbols, Tx_bits, ...
"precode_mode", options.duob_mode, ... "precode_mode", options.duob_mode, ...

6
Functions/EQ_recipes/dsp_scope_signal.m
View File

@@ -70,8 +70,8 @@ function output = dsp_scope_signal(Scpe_sig_raw, Symbols, Tx_bits, options)
pf_ = Postfilter("ncoeff", pf_ncoeffs, "useBurg", 1); %#ok<NASGU> pf_ = Postfilter("ncoeff", pf_ncoeffs, "useBurg", 1); %#ok<NASGU>
mlse_ = MLSE("duobinary_output", 0, 'M', M, 'trellis_states', PAMmapper(M,0).levels); %#ok<NASGU> mlse_ = MLSE("duobinary_output", 0, 'M', M, 'trellis_states', PAMmapper(M,0).levels); %#ok<NASGU>
eq_post = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",4096*2,"mu_dd",1e-4,"mu_tr",0,"order",2001,"sps",1,"decide",0,"adaption_technique","lms","mu_dc",mu_dc); %#ok<NASGU> eq_post = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",4096*2,"mu_dd",1e-4,"mu_tr",0,"order",2001,"sps",1,"decide",0,"adaption_technique","lms","dc_tracking_mu",mu_dc); %#ok<NASGU>
eq_post = FFE("epochs_tr",5,"epochs_dd",2,"len_tr",2^13,"mu_dd",mu_dd,"mu_tr",mu_tr,"order",25,"sps",2,"decide",0, "adaption_technique",adaption_method(adaption),"dd_mode",use_dd_mode,"mu_dc",mu_dc); %#ok<NASGU> eq_post = FFE("epochs_tr",5,"epochs_dd",2,"len_tr",2^13,"mu_dd",mu_dd,"mu_tr",mu_tr,"order",25,"sps",2,"decide",0, "adaption_technique",adaption_method(adaption),"dd_mode",use_dd_mode,"dc_tracking_mu",mu_dc); %#ok<NASGU>
mlse_db_enc = MLSE("DIR", [1,1], "duobinary_output", 0, "M", M, "trellis_states", PAMmapper(M,0).levels); %#ok<NASGU> mlse_db_enc = MLSE("DIR", [1,1], "duobinary_output", 0, "M", M, "trellis_states", PAMmapper(M,0).levels); %#ok<NASGU>
eq_db_enc = EQ("Ne", ffe_order_dbtgt, "Nb", dfe_order_dbtgt, "training_length", len_tr, ... eq_db_enc = EQ("Ne", ffe_order_dbtgt, "Nb", dfe_order_dbtgt, "training_length", len_tr, ...
@@ -94,7 +94,7 @@ function output = dsp_scope_signal(Scpe_sig_raw, Symbols, Tx_bits, options)
if use_ffe if use_ffe
eq_ffe = EQ("Ne",ffe_order_ffe,"Nb",[0,0,0],"training_length",len_tr,"training_loops",5,"dd_loops",5,"K",2,"DCmu",mu_dc,"DDmu",[mu_ffe mu_dfe],"DFEmu",0.005,"FFEmu",0,"plotfinal",0,"ideal_dfe",0); eq_ffe = EQ("Ne",ffe_order_ffe,"Nb",[0,0,0],"training_length",len_tr,"training_loops",5,"dd_loops",5,"K",2,"DCmu",mu_dc,"DDmu",[mu_ffe mu_dfe],"DFEmu",0.005,"FFEmu",0,"plotfinal",0,"ideal_dfe",0);
eq_ffe = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",4096*2,"mu_dd",1e-1,"mu_tr",0.4,"order",ffe_order_ffe(1),... eq_ffe = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",4096*2,"mu_dd",1e-1,"mu_tr",0.4,"order",ffe_order_ffe(1),...
"sps",2,"decide",0,"optmize_mus",0,"dd_mode",1,"adaption_technique","nlms","mu_dc",1.021e-05); "sps",2,"decide",0,"optmize_mus",0,"dd_mode",1,"adaption_technique","nlms","dc_tracking_mu",1.021e-05);
ffe_results = ffe(eq_ffe, M, Scpe_sig, Symbols, Tx_bits, ... ffe_results = ffe(eq_ffe, M, Scpe_sig, Symbols, Tx_bits, ...
"precode_mode", duob_mode, ... "precode_mode", duob_mode, ...

136
Functions/EQ_recipes/mpi_recipe_dev.m
View File

@@ -21,23 +21,61 @@ function output = mpi_recipe_dev(Scpe_sig_raw, Symbols, Tx_bits, options)
"debug_plots", options.debug_plots); "debug_plots", options.debug_plots);
output = struct(); output = struct();
eq_settings = { ... eq_core_settings = { ...
"epochs_tr", 5, ...
"epochs_dd", 5, ...
"len_tr", 4096, ...
"mu_tr",0.04, ...
"mu_dd",0.012, ...
"mu_dc", 0, ...
"adaptive_dc_enabled", 0, ...
"order", 25, ...
"sps", 2, ... "sps", 2, ...
"order", 25, ...
"decide", 0, ... "decide", 0, ...
"optmize_mus", 0, ... "adaption_technique", "nlms"};
eq_training_settings = { ...
"len_tr", 4096, ...
"epochs_tr", 5, ...
"mu_tr", 0.04};
eq_dd_settings = { ...
"dd_mode", 1, ... "dd_mode", 1, ...
"adaption_technique", "nlms", ... "epochs_dd", 5, ...
"plot_mu_optimization", options.debug_plots,... "mu_dd", 0.012};
"optimize_dc_params", false, ...
"save_debug",true}; eq_a1_settings = { ...
"dc_smoothing_a1", 0, ...
"dc_avg_bufferlength_a1", 0};
eq_a2_settings = { ...
"dc_smoothing_a2", 0, ...
"dc_level_avg_bufferlength_a2", 0, ...
"dc_level_weights_a2", [0.692 0.979 0.98 0.138]};
eq_dc_tracking_settings = { ...
"dc_tracking_mu", 0.05, ...
"dc_tracking_adaptive_enabled", 0, ...
"dc_tracking_buffer_len", 1024};
eq_ffe_update_settings = { ...
"ffe_update_buffer_len", 1};
eq_optimizer_settings = { ...
"optmize_mus", 0, ...
"plot_mu_optimization", options.debug_plots, ...
"optimize_dc_tracking_params", 1, ...
"optimize_a2_level_weights", 0, ...
"a2_level_weight_optimization_len", 2^15, ...
"a2_level_weight_optimization_max_evals", 30, ...
"a2_level_weight_max", 1};
eq_debug_settings = { ...
"save_debug", true};
eq_settings = [ ...
eq_core_settings, ...
eq_training_settings, ...
eq_dd_settings, ...
eq_a1_settings, ...
eq_a2_settings, ...
eq_dc_tracking_settings, ...
eq_ffe_update_settings, ...
eq_optimizer_settings, ...
eq_debug_settings];
eq_ffe = FFE(eq_settings{:}); eq_ffe = FFE(eq_settings{:});
@@ -73,8 +111,8 @@ function output = mpi_recipe_dev(Scpe_sig_raw, Symbols, Tx_bits, options)
output.ffe_debug.smoothing_window = smooth_len; output.ffe_debug.smoothing_window = smooth_len;
error_first_s = movmean(dbg.error_first_epoch(:),smooth_len,"omitnan"); error_first_s = movmean(dbg.error_first_epoch(:),smooth_len,"omitnan");
error_last_s = movmean(dbg.error(:),smooth_len,"omitnan"); error_last_s = movmean(dbg.error(:),smooth_len,"omitnan");
mu_dc_eff_s = movmean(dbg.mu_dc_eff(:),smooth_len,"omitnan"); dc_tracking_mu_eff_s = movmean(dbg.dc_tracking_mu_eff(:),smooth_len,"omitnan");
e_dc_eff_s = movmean(dbg.e_dc_eff(:),smooth_len,"omitnan"); dc_tracking_est_s = movmean(dbg.dc_tracking_est(:),smooth_len,"omitnan");
figure(460); clf; figure(460); clf;
tiledlayout(2,1,"TileSpacing","compact","Padding","compact"); tiledlayout(2,1,"TileSpacing","compact","Padding","compact");
@@ -90,9 +128,9 @@ function output = mpi_recipe_dev(Scpe_sig_raw, Symbols, Tx_bits, options)
nexttile;hold on nexttile;hold on
% plot(eq_sig_mov,"DisplayName","err dc eff"); % plot(eq_sig_mov,"DisplayName","err dc eff");
plot(mu_dc_eff_s,"DisplayName","mu DC - either fixed or adaptive"); plot(dc_tracking_mu_eff_s,"DisplayName","mu DC - either fixed or adaptive");
e_dc_scale = max(abs(e_dc_eff_s),[],"omitnan") + eps; e_dc_scale = max(abs(dc_tracking_est_s),[],"omitnan") + eps;
plot(-1.*e_dc_eff_s./e_dc_scale,"DisplayName","Inverted DC-tracking; Value that is subtracted during EQ"); plot(-1.*dc_tracking_est_s./e_dc_scale,"DisplayName","Inverted DC-tracking; Value that is subtracted during EQ");
avg_lvl_dc=mean(avg_for_lvl,1,"omitnan"); avg_lvl_dc=mean(avg_for_lvl,1,"omitnan");
plot(avg_lvl_dc./0.01,"DisplayName","Smoothed EQ output signal; calc'd by showLevelScatter"); plot(avg_lvl_dc./0.01,"DisplayName","Smoothed EQ output signal; calc'd by showLevelScatter");
grid on; grid on;
@@ -101,7 +139,7 @@ function output = mpi_recipe_dev(Scpe_sig_raw, Symbols, Tx_bits, options)
title("Effective adaptive DC step"); title("Effective adaptive DC step");
legend legend
inv_dc_track = -e_dc_eff_s(:); inv_dc_track = -dc_tracking_est_s(:);
avg_lvl_dc = avg_lvl_dc(:); avg_lvl_dc = avg_lvl_dc(:);
xcorr_len = min(numel(inv_dc_track),numel(avg_lvl_dc)); xcorr_len = min(numel(inv_dc_track),numel(avg_lvl_dc));
inv_dc_xcorr = inv_dc_track(1:xcorr_len); inv_dc_xcorr = inv_dc_track(1:xcorr_len);
@@ -135,33 +173,33 @@ function output = mpi_recipe_dev(Scpe_sig_raw, Symbols, Tx_bits, options)
%
%
%% MPI Reduction % %% MPI Reduction
if 1 % if 1
%
% dc_buffer_len = 0; % % dc_buffer_len = 0;
% ffe_buffer_len = 0; % % ffe_buffer_len = 0;
% smoothing_buffer_length = options.userParameters.smoothing_length; % % smoothing_buffer_length = options.userParameters.smoothing_length;
% smoothing_buffer_update = 1; % % smoothing_buffer_update = 1;
%
% eq_ffe_dcr = FFE_DCremoval_adaptive_mu(eq_settings{:}, ... % % eq_ffe_dcr = FFE_DCremoval_adaptive_mu(eq_settings{:}, ...
% "dc_buffer_len",dc_buffer_len, ... % % "dc_buffer_len",dc_buffer_len, ...
% "ffe_buffer_len",ffe_buffer_len,... % % "ffe_buffer_len",ffe_buffer_len,...
% "smoothing_buffer_length",smoothing_buffer_length,... % % "smoothing_buffer_length",smoothing_buffer_length,...
% "smoothing_buffer_update",smoothing_buffer_update); % % "smoothing_buffer_update",smoothing_buffer_update);
%
eq_ = FFE_DCremoval("epochs_tr",5,"epochs_dd",3,"len_tr",4096*2,"mu_dd",... % eq_ = FFE_DCremoval("epochs_tr",5,"epochs_dd",3,"len_tr",4096*2,"mu_dd",...
0.0002,"mu_tr",0,"order",25,"sps",2,"decide",0,... % 0.0002,"mu_tr",0,"order",25,"sps",2,"decide",0,...
"mu_dc",0.005,"dc_buffer_len",1); % "mu_dc",0.005,"dc_buffer_len",1);
%
ffe_results_dcr = ffe(eq_, options.M, Scpe_sig, Symbols, Tx_bits, ... % ffe_results_dcr = ffe(eq_, options.M, Scpe_sig, Symbols, Tx_bits, ...
"precode_mode", options.duob_mode, ... % "precode_mode", options.duob_mode, ...
'showAnalysis', options.debug_plots, ... % 'showAnalysis', options.debug_plots, ...
"postFFE", [], ... % "postFFE", [], ...
"eth_style_symbol_mapping", 0); % "eth_style_symbol_mapping", 0);
%
ffe_results_dcr.metrics.print("description",'FFE DCR'); % ffe_results_dcr.metrics.print("description",'FFE DCR');
output.ffe_dcr_package = ffe_results_dcr; % output.ffe_dcr_package = ffe_results_dcr;
end % end
end end

191
Tests/04_DSP/Equalizer/FFE_test.m
View File

@@ -12,8 +12,19 @@ classdef FFE_test < IMDDTestCase
testCase.verifyEqual(ffe.dd_mode, 1); testCase.verifyEqual(ffe.dd_mode, 1);
testCase.verifyEqual(ffe.mu_dd, 1e-5); testCase.verifyEqual(ffe.mu_dd, 1e-5);
testCase.verifyEqual(ffe.epochs_dd, 5); testCase.verifyEqual(ffe.epochs_dd, 5);
testCase.verifyFalse(logical(ffe.optimize_dc_params)); testCase.verifyFalse(logical(ffe.optimize_dc_tracking_params));
testCase.verifyEqual(ffe.dc_optimization_delay_weight, 1e-3); testCase.verifyEqual(ffe.dc_tracking_optimization_delay_weight, 1e-3);
testCase.verifyEqual(ffe.dc_tracking_power_exponent, 2);
testCase.verifyEqual(ffe.dc_avg_bufferlength_a1, 0);
testCase.verifyEqual(ffe.dc_smoothing_a1, 0);
testCase.verifyEqual(ffe.dc_level_avg_bufferlength_a2, 0);
testCase.verifyEqual(ffe.dc_smoothing_a2, 0);
testCase.verifyEqual(ffe.dc_level_weights_a2, 0);
testCase.verifyEqual(ffe.dc_tracking_buffer_len, 1);
testCase.verifyEqual(ffe.ffe_update_buffer_len, 1);
testCase.verifyFalse(logical(ffe.optimize_a2_level_weights));
testCase.verifyEqual(ffe.a2_level_weight_optimization_max_evals, 30);
testCase.verifyEqual(ffe.a2_level_weight_max, 1);
testCase.verifyFalse(ffe.decide); testCase.verifyFalse(ffe.decide);
testCase.verifyEqual(ffe.e, zeros(ffe.order, 1)); testCase.verifyEqual(ffe.e, zeros(ffe.order, 1));
testCase.verifyEqual(ffe.error, 0); testCase.verifyEqual(ffe.error, 0);
@@ -71,6 +82,182 @@ classdef FFE_test < IMDDTestCase
testCase.verifyNotEqual(y.signal, zeros(size(x.signal))); testCase.verifyNotEqual(y.signal, zeros(size(x.signal)));
testCase.verifyEqual(length(y.signal), length(x.signal)); testCase.verifyEqual(length(y.signal), length(x.signal));
end end
function dcTrackingBufferDelaysDcUpdateUntilBlockBoundary(testCase)
x = zeros(4, 1);
d = ones(4, 1);
ffe = FFE( ...
"sps", 1, ...
"order", 1, ...
"dc_tracking_mu", 0.1, ...
"dc_tracking_buffer_len", 2, ...
"dd_mode", 0, ...
"adaption_technique", adaption_method.lms);
ffe.constellation = unique(d);
ffe.equalize(x, d, 0, 1, numel(x), true, false);
testCase.verifyEqual(ffe.e_dc, 0.19, "AbsTol", 1e-12);
end
function ffeBufferDelaysTapUpdateUntilBlockBoundary(testCase)
x = ones(4, 1);
d = ones(4, 1);
ffe = FFE( ...
"sps", 1, ...
"order", 1, ...
"dc_tracking_mu", 0, ...
"ffe_update_buffer_len", 2, ...
"dd_mode", 0, ...
"adaption_technique", adaption_method.lms);
ffe.constellation = unique(d);
ffe.equalize(x, d, 0.1, 1, numel(x), true, false);
testCase.verifyEqual(ffe.e, 0.19, "AbsTol", 1e-12);
end
function dcAvgA1BufferlengthOneLeavesInputUnchanged(testCase)
x = (1:4).';
d = zeros(4, 1);
ffe = FFE( ...
"sps", 1, ...
"order", 1, ...
"dc_avg_bufferlength_a1", 1, ...
"save_debug", true, ...
"dd_mode", 0, ...
"adaption_technique", adaption_method.lms);
ffe.e = 1;
[y,~] = ffe.equalize(x, d, 0, 1, numel(x), true, false);
testCase.verifyEqual(y, x, "AbsTol", 1e-12);
testCase.verifyEqual(ffe.debug_struct.dc_avg_offset, zeros(1,4), "AbsTol", 1e-12);
end
function dcAvgA1SubtractsCausalMovingAverageByDefault(testCase)
x = (1:4).';
d = zeros(4, 1);
expected_offset = [0, 0, 0, 2];
expected_y = [1; 2; 3; 2];
ffe = FFE( ...
"sps", 1, ...
"order", 1, ...
"dc_avg_bufferlength_a1", 3, ...
"save_debug", true, ...
"dd_mode", 0, ...
"adaption_technique", adaption_method.lms);
ffe.e = 1;
[y,~] = ffe.equalize(x, d, 0, 1, numel(x), true, false);
testCase.verifyEqual(y, expected_y, "AbsTol", 1e-12);
testCase.verifyEqual(ffe.debug_struct.dc_avg_offset, expected_offset, "AbsTol", 1e-12);
end
function dcAvgA1SmoothingIsReservedForOfflineVariants(testCase)
x = (1:4).';
d = zeros(4, 1);
expected_offset = [0, 0, 0, 2];
expected_y = [1; 2; 3; 2];
ffe = FFE( ...
"sps", 1, ...
"order", 1, ...
"dc_avg_bufferlength_a1", 3, ...
"dc_smoothing_a1", 0.5, ...
"save_debug", true, ...
"dd_mode", 0, ...
"adaption_technique", adaption_method.lms);
ffe.e = 1;
[y,~] = ffe.equalize(x, d, 0, 1, numel(x), true, false);
testCase.verifyEqual(y, expected_y, "AbsTol", 1e-12);
testCase.verifyEqual(ffe.debug_struct.dc_avg_offset, expected_offset, "AbsTol", 1e-12);
end
function dcAvgA1WarnsWhenCombinedWithDcTracking(testCase)
testCase.verifyWarning(@() FFE( ...
"dc_avg_bufferlength_a1", 3, ...
"dc_tracking_mu", 0.1), "FFE:DCAvgWithDcTracking");
end
function dcLevelAvgA2SubtractsScalarWeightedResidual(testCase)
x = (1:4).';
d = zeros(4, 1);
expected_y = [1; 2; 1.5; 2.5];
ffe = FFE( ...
"sps", 1, ...
"order", 1, ...
"dc_level_avg_bufferlength_a2", 2, ...
"dc_level_weights_a2", 1, ...
"save_debug", true, ...
"dd_mode", 0, ...
"adaption_technique", adaption_method.lms);
ffe.e = 1;
ffe.constellation = 0;
[y,~] = ffe.equalize(x, d, 0, 1, numel(x), true, false);
testCase.verifyEqual(y, expected_y, "AbsTol", 1e-12);
testCase.verifyEqual(ffe.debug_struct.dc_level_mpi_est, [0, 0, 1.5, 1.5], "AbsTol", 1e-12);
testCase.verifyEqual(ffe.debug_struct.dc_level_weight, [0, 0, 1, 1], "AbsTol", 1e-12);
testCase.verifyEqual(ffe.debug_struct.dc_level_valid_count, [0, 0, 2, 2]);
end
function dcLevelAvgA2UsesVectorWeightsInConstellationOrder(testCase)
x = [1; 3; 1; 4];
d = [0; 2; 0; 2];
ffe = FFE( ...
"sps", 1, ...
"order", 1, ...
"dc_level_avg_bufferlength_a2", 2, ...
"dc_level_weights_a2", [0, 1], ...
"dd_mode", 0, ...
"adaption_technique", adaption_method.lms);
ffe.e = 1;
ffe.constellation = [0; 2];
[y,~] = ffe.equalize(x, d, 0, 1, numel(x), true, false);
testCase.verifyEqual(y, [1; 3; 1; 3.5], "AbsTol", 1e-12);
end
function dcLevelAvgA2WarnsWhenCombinedWithDcTracking(testCase)
testCase.verifyWarning(@() FFE( ...
"dc_level_avg_bufferlength_a2", 2, ...
"dc_level_weights_a2", 1, ...
"dc_tracking_mu", 0.1), "FFE:DCLevelAvgWithOtherDcSuppression");
end
function a2InitialWeightGuessFollowsLevelVarianceSlope(testCase)
x = [ ...
0.98; 1.02; 1.00; 1.01; ...
1.90; 2.10; 2.00; 2.08; ...
2.50; 3.50; 3.00; 3.40];
d = [ ...
ones(4,1); ...
2*ones(4,1); ...
3*ones(4,1)];
ffe = FFE("sps", 1, "a2_level_weight_max", 1);
ffe.constellation = [1; 2; 3];
[weights,stats] = ffe.a2LevelWeightInitialGuess(x,d);
testCase.verifySize(weights, [3, 1]);
testCase.verifyGreaterThan(weights(3), weights(2));
testCase.verifyGreaterThan(weights(2), weights(1));
testCase.verifyEqual(max(weights), 0.7, "AbsTol", 1e-12);
testCase.verifyGreaterThan(stats.variance_slope, 0);
end
end end
end end

63
projects/Diss/MPI_revisit/investigate_mpi_algorithms.m
View File

@@ -4,7 +4,7 @@ dsp_options.mode = "run_id";
dsp_options.recipe = @mpi_recipe_dev; dsp_options.recipe = @mpi_recipe_dev;
dsp_options.append_to_db = false; dsp_options.append_to_db = false;
dsp_options.start_occurence = 2; dsp_options.start_occurence = 2;
dsp_options.max_occurences = 10; dsp_options.max_occurences = 1;
dsp_options.debug_plots = false; dsp_options.debug_plots = false;
dsp_options.database_type = "mysql"; dsp_options.database_type = "mysql";
@@ -28,7 +28,7 @@ fp.where('Runs','run_id','GREATER_EQUAL',3153);
fp.where('Runs', 'symbolrate', 'EQUALS', 112e9); fp.where('Runs', 'symbolrate', 'EQUALS', 112e9);
fp.where('Runs', 'fiber_length', 'EQUALS', 0); fp.where('Runs', 'fiber_length', 'EQUALS', 0);
fp.where('Runs', 'interference_path_length', 'EQUALS', 1000); fp.where('Runs', 'interference_path_length', 'EQUALS', 1000);
fp.where('Runs', 'sir', 'LESS_EQUAL', 30); fp.where('Runs', 'sir', 'LESS_EQUAL', 20);
fp.where('Runs', 'db_mode', 'EQUALS', '"no_db"'); fp.where('Runs', 'db_mode', 'EQUALS', '"no_db"');
% fp.where('Runs', 'is_mpi', 'EQUALS', 1); % fp.where('Runs', 'is_mpi', 'EQUALS', 1);
fp.where('Runs', 'pam_level', 'EQUALS', 4); fp.where('Runs', 'pam_level', 'EQUALS', 4);
@@ -41,6 +41,7 @@ fp.where('Runs', 'pam_level', 'EQUALS', 4);
[~, sortIdx] = sort(dataTable.sir, 'descend'); [~, sortIdx] = sort(dataTable.sir, 'descend');
dataTable = dataTable(sortIdx, :); dataTable = dataTable(sortIdx, :);
dataTable = dataTable(1, :);
run_ids = dataTable.run_id; run_ids = dataTable.run_id;
if isempty(run_ids) if isempty(run_ids)
@@ -74,31 +75,43 @@ fprintf("-> [ %d run_id(s) × %d userParam combination(s) = %d parallel job(s) ]
%% Analyze %% Analyze
storageNames = fieldnames(wh.sto); storageNames = fieldnames(wh.sto);
x_vars = run_ids;%dsp_options.userParameters.smoothing_length; result = wh.sto.(storageNames{1});
x_vars = reshape(x_vars,1,[]); result = result(:).';
figure(2026);hold on % Each stored entry is a package cell containing one or more occurrences.
for i = 1:numel(run_ids) ber_all = cell(1,numel(result));
disp(run_ids(i)) for result_idx = 1:numel(result)
result = wh.getStoValue(storageNames{1}, x_vars); packageCell = result{result_idx};
% Each cell in result is a cell array (packageCell) containing multiple packages. if isempty(packageCell)
% Extract BERs from all packages and, for example, average them per x_var. ber_all{result_idx} = NaN;
ber_all = cellfun(@(packageCell) cellfun(@(pkg) pkg.metrics.BER, packageCell), result, 'UniformOutput', false); continue
% Convert to numeric matrix (rows = packages, cols = x_vars) by padding if needed
maxPackages = max(cellfun(@numel, ber_all));
ber_mat = nan(maxPackages, numel(ber_all));
for k = 1:numel(ber_all)
ber_mat(1:numel(ber_all{k}), k) = ber_all{k};
end end
% Choose aggregation: mean across packages (ignore NaNs)
ber = mean(ber_mat, 1, 'omitnan');
ber_values = nan(1,numel(packageCell));
x = dataTable.sir; for package_idx = 1:numel(packageCell)
y = ber; if isstruct(packageCell{package_idx}) && isfield(packageCell{package_idx},"metrics")
ber_values(package_idx) = packageCell{package_idx}.metrics.BER;
plot(x,ber); end
scatter(x,ber_mat) end
beautifyBERplot("logscale",true,"setcolors",false,"setmarkers",true); ber_all{result_idx} = ber_values;
end end
maxPackages = max(cellfun(@numel,ber_all));
ber_mat = nan(maxPackages,numel(ber_all));
for k = 1:numel(ber_all)
ber_mat(1:numel(ber_all{k}),k) = ber_all{k};
end
ber = mean(ber_mat,1,"omitnan");
x = dataTable.sir(:).';
if numel(x) ~= numel(ber)
x = 1:numel(ber);
end
figure(2026); clf; hold on
plot(x,ber);
for k = 1:numel(x)
scatter(repmat(x(k),maxPackages,1),ber_mat(:,k))
end
beautifyBERplot("logscale",true,"setcolors",false,"setmarkers",true);

BIN
projects/Diss/MPI_revisit/workerError.mat Normal file
View File

Binary file not shown.

2
projects/Diss/investigate_channel_memory.m
View File

@@ -133,7 +133,7 @@ for i = 1:numel(fsym_values)
eq_ = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ... eq_ = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ...
"mu_dd",1e-3,"mu_tr",0.4,"order",50, ... "mu_dd",1e-3,"mu_tr",0.4,"order",50, ...
"sps",2,"decide",0,"optmize_mus",0,"dd_mode",1, ... "sps",2,"decide",0,"optmize_mus",0,"dd_mode",1, ...
"adaption_technique","nlms","mu_dc",1e-3); "adaption_technique","nlms","dc_tracking_mu",1e-3);
pf_ = Postfilter("ncoeff",pf_ncoeffs,"useBurg",1); pf_ = Postfilter("ncoeff",pf_ncoeffs,"useBurg",1);
mlse_ = MLSE("duobinary_output",0,'M',M,'trellis_states',PAMmapper(M,0).levels); mlse_ = MLSE("duobinary_output",0,'M',M,'trellis_states',PAMmapper(M,0).levels);

2
projects/Diss/investigate_noise_enhancement.m
View File

@@ -134,7 +134,7 @@ for i = 1:numel(fsym_values)
eq_ = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ... eq_ = FFE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ...
"mu_dd",1e-1,"mu_tr",0.4,"order",50, ... "mu_dd",1e-1,"mu_tr",0.4,"order",50, ...
"sps",2,"decide",0,"optmize_mus",1,"dd_mode",1, ... "sps",2,"decide",0,"optmize_mus",1,"dd_mode",1, ...
"adaption_technique","nlms","mu_dc",1.021e-05); "adaption_technique","nlms","dc_tracking_mu",1.021e-05);
case "VNLE" case "VNLE"
eq_ = VNLE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ... eq_ = VNLE("epochs_tr",5,"epochs_dd",5,"len_tr",len_tr, ...
"mu_dd",[0.0004 0.0005 0.0006],"mu_tr",0, ... "mu_dd",[0.0004 0.0005 0.0006],"mu_tr",0, ...

BIN
workerError.mat
View File

Binary file not shown.