Calibration System Architecture¶

Overview¶

El Calibration System es un framework modular para calibración y ajuste fino de componentes DSP. Implementa un pipeline estándar de 5 etapas y soporta optimización multi-objetivo.

Core Concepts¶

1. Calibration Target¶

Un CalibrationTarget encapsula: - El componente a calibrar (ICalibratableComponent) - Una o más especificaciones (CalibrationSpecification) - Método para calcular error entre respuesta medida y target

CalibrationTarget target("FilterID", componentPtr);
target.addSpecification(spec1);
target.addSpecification(spec2);  // Multi-objective

2. Calibration Specification¶

Define el comportamiento target:

CalibrationSpecification spec(TargetType::FREQUENCY_RESPONSE);
spec.idealResponse = { /* target curve */ };
spec.tolerances = { /* per-point tolerances */ };
spec.importance = 1.0f;  // Weight for multi-objective
spec.errorThreshold = 0.1f;  // Convergence criterion

3. Calibration Process¶

Pipeline estándar de 5 etapas:

1. Preparation → Setup test signals, initialize optimizer
2. Measurement → Inject stimulus, capture response
3. Optimization → Iteratively adjust parameters
4. Validation → Verify stability, consistency
5. Storage → Save and apply results

Implementado como clase base CalibrationProcess con métodos virtuales.

4. Measured Response¶

Datos capturados del componente:

struct MeasuredResponse {
    std::vector<float> frequencies;
    std::vector<float> magnitudes;  // dB
    std::vector<float> phases;      // radians
    std::map<std::string, float> metrics;
};

5. Calibration Result¶

Output final:

struct CalibrationResult {
    bool success;
    std::map<std::string, float> parameters;  // Optimized values
    CalibrationError error;
    int iterationsUsed;
    MeasuredResponse finalResponse;
};

Architecture Diagrams¶

Component Hierarchy¶

ICalibratableComponent
    ↓
CalibrationTarget (wraps component + specs)
    ↓
CalibrationProcess (abstract base)
    ↓
Concrete Calibrators:
    - FilterCalibrator
    - DynamicsCalibrator
    - NonlinearCalibrator
    - etc.

Data Flow¶

Component Parameters
    ↓
Test Signal Generator → Component → Measurement
    ↓
Error Calculation (vs Target)
    ↓
Optimizer → New Parameters
    ↓
Repeat until convergence

Multi-Objective Optimization¶

Population of Solutions
    ↓
Evaluate all objectives
    ↓
Calculate Pareto Ranks
    ↓
Calculate Crowding Distances
    ↓
Select next generation
    ↓
Crossover & Mutation
    ↓
Repeat until convergence
    ↓
Select best from Pareto front

Subsystems¶

00_calibration_framework¶

Core framework con estructuras base: - CalibrationTarget: Define qué calibrar - CalibrationProcess: Pipeline estándar - MultiObjectiveOptimizer: NSGA-II implementation - CalibrationSignalGenerator: Test signal generation

01_frequency_calibration¶

Calibración de respuesta en frecuencia: - FilterCalibrator: Filtros y EQs - Log sweep o discrete tone measurement - Gradient descent optimization

02_dynamics_calibration¶

Calibración de procesadores dinámicos: - Curvas de compresión - Tiempos de attack/release - Gate/expander thresholds

03_nonlinear_calibration¶

Calibración de no-linealidades: - Contenido armónico - Transfer functions - Saturación y distorsión

04_modulation_calibration¶

Calibración de moduladores: - Rangos de LFO - Curvas de envelope - Depths musicales

05_level_calibration¶

Calibración de niveles: - Gain staging - Unity gain - Headroom management

06_latency_calibration¶

Medición y compensación de latencia: - Sample-accurate measurement - Phase-locked loops - Delay compensation

07_reference_matching¶

Matching de referencias: - Hardware emulation - Null test calibration - Component matching

08_psychoacoustic_calibration¶

Calibración perceptual: - Loudness calibration - Equal loudness curves - Masking-aware optimization

09_adaptive_calibration¶

Auto-calibración: - Continuous calibration - ML-based parameter prediction - User preference learning

10_standards_compliance¶

Conformidad con estándares: - AES17 audio measurement - EBU R128 loudness - ITU-R BS.1770

Optimization Algorithms¶

Gradient Descent¶

Para espacios de parámetros suaves: - Learning rate adaptativo - Momentum para convergencia más rápida - Backtracking line search

Genetic Algorithm¶

Para espacios complejos: - Tournament selection - Uniform crossover - Gaussian mutation

Particle Swarm¶

Para optimización rápida: - Swarm intelligence - Global y local best - Inertia weight

Multi-Objective (NSGA-II)¶

Para objetivos competing: - Pareto optimization - Crowding distance para diversidad - Elitism

Extension Points¶

Custom Calibrators¶

Inherit from CalibrationProcess:

class MyCustomCalibrator : public CalibrationProcess {
protected:
    bool prepare(CalibrationTarget& target) override {
        // Custom preparation
    }

    MeasuredResponse measure(CalibrationTarget& target) override {
        // Custom measurement
    }

    CalibrationResult optimize(/* ... */) override {
        // Custom optimization
    }

    bool validate(/* ... */) override {
        // Custom validation
    }

    void store(/* ... */) override {
        // Custom storage
    }
};

Custom Components¶

Implement ICalibratableComponent:

class MyComponent : public ICalibratableComponent {
public:
    std::string getComponentId() const override;
    void process(const float* input, float* output, size_t numSamples) override;
    void setParameter(const std::string& name, float value) override;
    float getParameter(const std::string& name) const override;
    // ... etc
};

Custom Objectives¶

Add to CalibrationSpecification:

CalibrationSpecification spec(TargetType::CUSTOM);
spec.idealResponse = computeCustomTarget();
spec.importance = 0.8f;

Performance Considerations¶

Measurement Speed¶

Log sweeps más rápidos que discrete tones
Averaging reduce noise pero aumenta tiempo
Cache resultados de mediciones idénticas

Optimization Speed¶

Gradient descent rápido pero puede quedar stuck
GA/PSO más robusto pero más lento
Multi-objetivo significativamente más costoso

Memory Usage¶

Response data puede ser grande (muchas frecuencias)
Population-based algorithms usan mucha memoria
Stream processing para señales largas

Error Handling¶

Component Errors¶

try {
    result = calibrator.calibrate(target);
} catch (const std::exception& e) {
    // Handle calibration failure
}

Convergence Failures¶

if (!result.success) {
    // Check result.notes for reason
    // Use best found parameters anyway?
}

Validation Failures¶

if (result.success && !calibrator.validate(target, result)) {
    // Calibration succeeded but validation failed
    // May indicate instability
}

Thread Safety¶

CalibrationProcess is NOT thread-safe
Each thread should have its own calibrator instance
Components should be thread-safe for parallel evaluation
Use mutex for shared resources (databases, files)

Future Enhancements¶

Planned Features¶

Performance Improvements¶

Parallel evaluation in GA/PSO
GPU acceleration for FFT
Cache optimization
SIMD for signal processing

Usability Improvements¶

GUI for calibration visualization
Preset calibration profiles
Automatic component discovery
Calibration scheduling

References¶

Academic Papers¶

"Multi-Objective Optimization Using Evolutionary Algorithms" (Deb, 2001)
"A Fast Elitist Non-Dominated Sorting Genetic Algorithm: NSGA-II" (Deb et al., 2002)
"Audio Measurement Handbook" (Metzler, 2005)

Standards¶

AES17: Digital Audio Measurement
EBU R128: Loudness normalisation
ITU-R BS.1770: Loudness measurement
IEC 61606: Digital Audio Interfaces

Tools¶

Audio Precision APx series
QuantAsylum QA400 series
MATLAB Audio Toolbox
Python librosa, scipy.signal

Version: 1.0.0 Last Updated: 2025-10-14 Maintainer: AudioLab Team