oscura 0.5.1__py3-none-any.whl → 0.7.0__py3-none-any.whl

oscura/analyzers/signal/timing_analysis.py

@@ -0,0 +1,1086 @@
+"""Comprehensive timing analysis and clock recovery for signal synchronization.
+
+This module provides advanced timing analysis capabilities including:
+- Clock recovery using multiple methods (ZCD, histogram, autocorrelation, PLL, FFT)
+- Baud rate detection for serial protocols
+- Timing jitter and drift analysis
+- Phase-locked loop (PLL) simulation
+- Signal-to-noise ratio (SNR) calculation
+- Eye diagram generation
+
+Example:
+    >>> import numpy as np
+    >>> analyzer = TimingAnalyzer(method="autocorrelation")
+    >>> signal = np.sin(2 * np.pi * 1e6 * np.linspace(0, 1e-3, 100000))
+    >>> result = analyzer.recover_clock(signal, sample_rate=100e6)
+    >>> print(f"Clock rate: {result.detected_clock_rate / 1e6:.3f} MHz")
+    Clock rate: 1.000 MHz
+    >>> print(f"Confidence: {result.confidence:.2f}")
+    Confidence: 0.95
+
+References:
+    - Digital Communications by John G. Proakis
+    - Clock Recovery in High-Speed Optical Fiber Systems (IEEE)
+    - Phase-Locked Loop Design Handbook by Dan H. Wolaver
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import TYPE_CHECKING, Any, ClassVar
+
+import numpy as np
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+
+
+@dataclass
+class TimingAnalysisResult:
+    """Timing analysis and clock recovery result.
+
+    Attributes:
+        detected_clock_rate: Recovered clock frequency in Hz.
+        confidence: Confidence score from 0.0 to 1.0.
+        jitter_rms: RMS jitter in seconds.
+        drift_rate: Clock drift in parts per million (ppm).
+        snr_db: Signal-to-noise ratio in decibels.
+        method: Method used ("zcd", "histogram", "autocorrelation", "pll", "fft").
+        statistics: Additional method-specific statistics.
+
+    Example:
+        >>> result = TimingAnalysisResult(
+        ...     detected_clock_rate=10e6,
+        ...     confidence=0.95,
+        ...     jitter_rms=1e-12,
+        ...     drift_rate=2.5,
+        ...     snr_db=45.0,
+        ...     method="autocorrelation"
+        ... )
+        >>> print(f"Clock: {result.detected_clock_rate / 1e6:.1f} MHz")
+        Clock: 10.0 MHz
+    """
+
+    detected_clock_rate: float
+    confidence: float
+    jitter_rms: float
+    drift_rate: float
+    snr_db: float
+    method: str
+    statistics: dict[str, float] = field(default_factory=dict)
+
+
+class TimingAnalyzer:
+    """Timing analysis and clock recovery engine.
+
+    Provides comprehensive timing analysis including multiple clock recovery
+    methods, jitter/drift measurement, and signal quality assessment.
+
+    Attributes:
+        method: Clock recovery method to use.
+        METHODS: Available clock recovery methods.
+
+    Example:
+        >>> analyzer = TimingAnalyzer(method="autocorrelation")
+        >>> signal = np.random.randn(10000)
+        >>> result = analyzer.recover_clock(signal, sample_rate=1e6)
+        >>> print(f"Method: {result.method}")
+        Method: autocorrelation
+    """
+
+    METHODS: ClassVar[list[str]] = ["zcd", "histogram", "autocorrelation", "pll", "fft"]
+
+    def __init__(self, method: str = "autocorrelation") -> None:
+        """Initialize timing analyzer.
+
+        Args:
+            method: Clock recovery method ("zcd", "histogram", "autocorrelation",
+                "pll", or "fft"). Default is "autocorrelation" for best general-
+                purpose performance.
+
+        Raises:
+            ValueError: If method is not in METHODS.
+
+        Example:
+            >>> analyzer = TimingAnalyzer(method="fft")
+            >>> analyzer.method
+            'fft'
+        """
+        if method not in self.METHODS:
+            raise ValueError(f"Method must be one of {self.METHODS}, got '{method}'")
+        self.method = method
+
+    def recover_clock(
+        self,
+        signal: NDArray[np.floating[Any]],
+        sample_rate: float,
+        initial_estimate: float | None = None,
+    ) -> TimingAnalysisResult:
+        """Recover clock frequency from signal.
+
+        Uses the configured method to recover the clock rate from a digital
+        or analog signal. Automatically detects periodicity and estimates
+        the dominant frequency.
+
+        Args:
+            signal: Input signal array.
+            sample_rate: Sampling rate in Hz.
+            initial_estimate: Initial frequency estimate in Hz (optional,
+                required for PLL method).
+
+        Returns:
+            TimingAnalysisResult with recovered clock rate and statistics.
+
+        Example:
+            >>> signal = np.tile([0, 0, 1, 1], 1000)
+            >>> analyzer = TimingAnalyzer(method="autocorrelation")
+            >>> result = analyzer.recover_clock(signal, sample_rate=1e6)
+            >>> result.detected_clock_rate > 0
+            True
+
+        References:
+            IEEE 1241-2010: Standard for Terminology and Test Methods
+        """
+        if self.method == "zcd":
+            return self._zero_crossing_detection(signal, sample_rate)
+        elif self.method == "histogram":
+            return self._histogram_method(signal, sample_rate)
+        elif self.method == "autocorrelation":
+            return self._autocorrelation_method(signal, sample_rate)
+        elif self.method == "pll":
+            if initial_estimate is None:
+                raise ValueError("PLL method requires initial_estimate parameter")
+            return self._pll_simulation(signal, sample_rate, initial_estimate)
+        elif self.method == "fft":
+            return self._fft_method(signal, sample_rate)
+        else:
+            raise ValueError(f"Unknown method: {self.method}")
+
+    def detect_baud_rate(
+        self,
+        signal: NDArray[np.floating[Any]],
+        sample_rate: float,
+        min_baud: float = 300,
+        max_baud: float = 115200,
+    ) -> TimingAnalysisResult:
+        """Detect serial baud rate from signal.
+
+        Analyzes signal to determine the most likely baud rate for
+        serial communication protocols (UART, RS-232, etc.).
+
+        Args:
+            signal: Input serial signal array.
+            sample_rate: Sampling rate in Hz.
+            min_baud: Minimum baud rate to consider (default 300).
+            max_baud: Maximum baud rate to consider (default 115200).
+
+        Returns:
+            TimingAnalysisResult with detected baud rate.
+
+        Example:
+            >>> # 9600 baud signal (104.17 µs bit period)
+            >>> bit_period = 1 / 9600
+            >>> signal = np.tile([0]*10 + [1]*10, 500)
+            >>> analyzer = TimingAnalyzer(method="autocorrelation")
+            >>> result = analyzer.detect_baud_rate(signal, sample_rate=1e6)
+            >>> abs(result.detected_clock_rate - 9600) < 1000
+            True
+
+        References:
+            EIA-232: Serial data communication standard
+        """
+        # Use autocorrelation to find bit period
+        temp_analyzer = TimingAnalyzer(method="autocorrelation")
+        result = temp_analyzer.recover_clock(signal, sample_rate)
+
+        # Filter to common baud rates
+        detected_rate = result.detected_clock_rate
+
+        # Clamp to valid range
+        if detected_rate < min_baud:
+            detected_rate = min_baud
+            result.confidence *= 0.5
+        elif detected_rate > max_baud:
+            detected_rate = max_baud
+            result.confidence *= 0.5
+
+        # Find nearest standard baud rate
+        standard_bauds = [
+            300,
+            1200,
+            2400,
+            4800,
+            9600,
+            14400,
+            19200,
+            38400,
+            57600,
+            115200,
+            230400,
+            460800,
+            921600,
+        ]
+        valid_bauds = [b for b in standard_bauds if min_baud <= b <= max_baud]
+
+        if valid_bauds:
+            nearest_baud = min(valid_bauds, key=lambda b: abs(b - detected_rate))
+            # Increase confidence if very close to standard baud
+            if abs(nearest_baud - detected_rate) / nearest_baud < 0.05:
+                result.confidence = min(1.0, result.confidence * 1.2)
+                detected_rate = float(nearest_baud)
+
+        result.detected_clock_rate = detected_rate
+        result.statistics["standard_baud_match"] = (
+            abs(detected_rate - nearest_baud) / nearest_baud < 0.05 if valid_bauds else False
+        )
+
+        return result
+
+    def analyze_jitter(
+        self,
+        transitions: NDArray[np.floating[Any]],
+        nominal_period: float,
+    ) -> dict[str, Any]:
+        """Analyze timing jitter from edge transitions.
+
+        Computes jitter statistics including RMS, peak-to-peak, and
+        histogram distribution.
+
+        Args:
+            transitions: Array of transition timestamps in seconds.
+            nominal_period: Expected nominal period in seconds.
+
+        Returns:
+            Dictionary with jitter statistics:
+                - rms: RMS jitter in seconds
+                - peak_to_peak: Peak-to-peak jitter in seconds
+                - mean_period: Mean measured period in seconds
+                - std_period: Standard deviation of period in seconds
+                - histogram_bins: Histogram bin edges
+                - histogram_counts: Histogram counts
+
+        Example:
+            >>> transitions = np.array([0.0, 1e-6, 2e-6, 3.01e-6, 4e-6])
+            >>> analyzer = TimingAnalyzer()
+            >>> stats = analyzer.analyze_jitter(transitions, nominal_period=1e-6)
+            >>> stats['rms'] >= 0
+            True
+            >>> 'peak_to_peak' in stats
+            True
+
+        References:
+            IEEE 2414-2020: Standard for Jitter and Phase Noise
+        """
+        if len(transitions) < 2:
+            return {
+                "rms": np.nan,
+                "peak_to_peak": np.nan,
+                "mean_period": np.nan,
+                "std_period": np.nan,
+                "histogram_bins": np.array([]),
+                "histogram_counts": np.array([]),
+            }
+
+        # Calculate periods
+        periods = np.diff(transitions)
+
+        if len(periods) == 0:
+            return {
+                "rms": np.nan,
+                "peak_to_peak": np.nan,
+                "mean_period": np.nan,
+                "std_period": np.nan,
+                "histogram_bins": np.array([]),
+                "histogram_counts": np.array([]),
+            }
+
+        # Jitter is deviation from nominal period - handle NaN values
+        with np.errstate(invalid="ignore"):
+            deviations = periods - nominal_period
+
+        rms_jitter = float(np.std(deviations))
+        pp_jitter = float(np.max(periods) - np.min(periods))
+        mean_period = float(np.mean(periods))
+        std_period = float(np.std(periods))
+
+        # Generate histogram
+        if len(deviations) >= 10:
+            counts, bins = np.histogram(deviations, bins=50)
+        else:
+            counts = np.array([])
+            bins = np.array([])
+
+        return {
+            "rms": rms_jitter,
+            "peak_to_peak": pp_jitter,
+            "mean_period": mean_period,
+            "std_period": std_period,
+            "histogram_bins": bins,
+            "histogram_counts": counts,
+        }
+
+    def analyze_drift(
+        self,
+        transitions: NDArray[np.floating[Any]],
+        window_size: int = 1000,
+    ) -> float:
+        """Analyze clock drift over time.
+
+        Measures the rate of change in clock frequency over the observation
+        window, expressed in parts per million (ppm).
+
+        Args:
+            transitions: Array of transition timestamps in seconds.
+            window_size: Number of transitions to use for drift calculation.
+
+        Returns:
+            Clock drift in ppm (parts per million).
+
+        Example:
+            >>> # Perfect clock (no drift)
+            >>> transitions = np.arange(0, 1000) * 1e-6
+            >>> analyzer = TimingAnalyzer()
+            >>> drift = analyzer.analyze_drift(transitions, window_size=100)
+            >>> abs(drift) < 10  # Very low drift
+            True
+
+        References:
+            IEEE 1588: Precision Time Protocol (PTP)
+        """
+        if len(transitions) < window_size:
+            window_size = len(transitions)
+
+        if window_size < 10:
+            return np.nan
+
+        # Split into windows and calculate average period in each
+        n_windows = max(2, window_size // 100)
+        window_length = len(transitions) // n_windows
+
+        window_frequencies: list[float] = []
+
+        for i in range(n_windows):
+            start_idx = i * window_length
+            end_idx = min((i + 1) * window_length, len(transitions))
+
+            if end_idx - start_idx < 2:
+                continue
+
+            window_transitions = transitions[start_idx:end_idx]
+            periods = np.diff(window_transitions)
+
+            if len(periods) > 0:
+                mean_period = np.mean(periods)
+                if mean_period > 0:
+                    window_frequencies.append(1.0 / mean_period)
+
+        if len(window_frequencies) < 2:
+            return 0.0
+
+        # Linear fit to frequency vs time
+        time_points = np.linspace(0, len(transitions), len(window_frequencies))
+        coeffs = np.polyfit(time_points, window_frequencies, 1)
+        slope = coeffs[0]  # Hz per sample
+        mean_freq = np.mean(window_frequencies)
+
+        if mean_freq == 0:
+            return 0.0
+
+        # Convert to ppm
+        drift_ppm = (slope * len(transitions)) / mean_freq * 1e6
+
+        return float(drift_ppm)
+
+    def calculate_snr(
+        self,
+        signal: NDArray[np.floating[Any]],
+        signal_freq: float,
+        sample_rate: float,
+    ) -> float:
+        """Calculate signal-to-noise ratio in dB.
+
+        Estimates SNR by separating signal power (at fundamental frequency)
+        from noise power (all other frequency components).
+
+        Args:
+            signal: Input signal array.
+            signal_freq: Expected signal frequency in Hz.
+            sample_rate: Sampling rate in Hz.
+
+        Returns:
+            SNR in decibels (dB).
+
+        Example:
+            >>> signal = np.sin(2 * np.pi * 1000 * np.linspace(0, 0.1, 10000))
+            >>> analyzer = TimingAnalyzer()
+            >>> snr = analyzer.calculate_snr(signal, signal_freq=1000, sample_rate=100e3)
+            >>> snr > 40  # Clean signal should have high SNR
+            True
+
+        References:
+            IEEE 1057: Standard for Digitizing Waveform Recorders
+        """
+        if len(signal) < 64:
+            return np.nan
+
+        # Remove DC component - handle inf/nan
+        with np.errstate(invalid="ignore"):
+            signal_clean = np.where(np.isfinite(signal), signal, 0.0)
+            signal_centered = signal_clean - np.mean(signal_clean)
+
+        # Compute FFT
+        n = len(signal_centered)
+        nfft = int(2 ** np.ceil(np.log2(n)))
+        spectrum = np.fft.rfft(signal_centered, n=nfft)
+        freqs = np.fft.rfftfreq(nfft, d=1.0 / sample_rate)
+        magnitude = np.abs(spectrum)
+
+        # Find bin closest to signal frequency
+        signal_bin_idx = np.argmin(np.abs(freqs - signal_freq))
+
+        # Signal power is in the signal bin and immediate neighbors
+        signal_bins = [signal_bin_idx]
+        if signal_bin_idx > 0:
+            signal_bins.append(signal_bin_idx - 1)
+        if signal_bin_idx < len(magnitude) - 1:
+            signal_bins.append(signal_bin_idx + 1)
+
+        signal_power = float(np.sum(magnitude[signal_bins] ** 2))
+
+        # Noise power is everything else (excluding DC at bin 0)
+        noise_mask = np.ones(len(magnitude), dtype=bool)
+        noise_mask[0] = False  # Exclude DC
+        for idx in signal_bins:
+            noise_mask[idx] = False
+
+        noise_power = float(np.sum(magnitude[noise_mask] ** 2))
+
+        if noise_power == 0 or signal_power == 0:
+            return np.nan
+
+        snr = 10 * np.log10(signal_power / noise_power)
+
+        return float(snr)
+
+    def generate_eye_diagram(
+        self,
+        signal: NDArray[np.floating[Any]],
+        symbol_rate: float,
+        sample_rate: float,
+        output_path: Path,
+    ) -> None:
+        """Generate eye diagram for signal quality assessment.
+
+        Creates an eye diagram by overlaying multiple symbol periods.
+        A wide, open eye indicates good signal quality; a closed eye
+        indicates high jitter or noise.
+
+        Args:
+            signal: Input signal array.
+            symbol_rate: Symbol rate in Hz.
+            sample_rate: Sampling rate in Hz.
+            output_path: Path to save eye diagram image.
+
+        Example:
+            >>> import tempfile
+            >>> signal = np.sin(2 * np.pi * 1e6 * np.linspace(0, 1e-3, 100000))
+            >>> analyzer = TimingAnalyzer()
+            >>> with tempfile.NamedTemporaryFile(suffix='.png', delete=False) as f:
+            ...     analyzer.generate_eye_diagram(signal, 1e6, 100e6, Path(f.name))
+
+        References:
+            - Telecommunications Measurement Analysis (Tektronix)
+            - IEEE 802.3: Ethernet eye diagram templates
+        """
+        try:
+            import matplotlib
+
+            matplotlib.use("Agg")
+            import matplotlib.pyplot as plt
+        except ImportError as e:
+            raise ImportError(
+                "matplotlib required for eye diagram generation. "
+                "Install with: pip install matplotlib"
+            ) from e
+
+        # Calculate samples per symbol
+        samples_per_symbol = int(sample_rate / symbol_rate)
+
+        if samples_per_symbol < 4:
+            raise ValueError(
+                f"Insufficient samples per symbol: {samples_per_symbol}. "
+                f"Need at least 4 samples per symbol for eye diagram."
+            )
+
+        # Extract symbol periods (2 symbols per trace for eye diagram)
+        num_symbols = len(signal) // samples_per_symbol
+        eye_traces: list[NDArray[np.floating[Any]]] = []
+
+        for i in range(num_symbols - 1):
+            start = i * samples_per_symbol
+            end = start + 2 * samples_per_symbol  # 2 symbols for eye
+            if end <= len(signal):
+                eye_traces.append(signal[start:end])
+
+        if len(eye_traces) == 0:
+            raise ValueError("Insufficient data for eye diagram")
+
+        # Plot overlaid traces
+        plt.figure(figsize=(10, 6))
+        time_axis = np.linspace(0, 2, 2 * samples_per_symbol)
+
+        for trace in eye_traces:
+            plt.plot(time_axis, trace, alpha=0.1, color="blue", linewidth=0.5)
+
+        plt.xlabel("Time (symbol periods)")
+        plt.ylabel("Amplitude")
+        plt.title(f"Eye Diagram (Symbol Rate: {symbol_rate / 1e3:.1f} kHz)")
+        plt.grid(True, alpha=0.3)
+        plt.xlim(0, 2)
+        plt.tight_layout()
+
+        # Save to file
+        plt.savefig(output_path, dpi=300, bbox_inches="tight")
+        plt.close()
+
+    def export_statistics(
+        self,
+        result: TimingAnalysisResult,
+        output_path: Path,
+    ) -> None:
+        """Export timing statistics as JSON.
+
+        Args:
+            result: TimingAnalysisResult to export.
+            output_path: Path to save JSON file.
+
+        Example:
+            >>> import tempfile
+            >>> result = TimingAnalysisResult(
+            ...     detected_clock_rate=10e6,
+            ...     confidence=0.95,
+            ...     jitter_rms=1e-12,
+            ...     drift_rate=2.5,
+            ...     snr_db=45.0,
+            ...     method="autocorrelation"
+            ... )
+            >>> analyzer = TimingAnalyzer()
+            >>> with tempfile.NamedTemporaryFile(suffix='.json', delete=False) as f:
+            ...     analyzer.export_statistics(result, Path(f.name))
+        """
+        import json
+
+        data = {
+            "detected_clock_rate_hz": result.detected_clock_rate,
+            "detected_clock_rate_mhz": result.detected_clock_rate / 1e6,
+            "confidence": result.confidence,
+            "jitter_rms_seconds": result.jitter_rms,
+            "jitter_rms_picoseconds": result.jitter_rms * 1e12,
+            "drift_rate_ppm": result.drift_rate,
+            "snr_db": result.snr_db,
+            "method": result.method,
+            "statistics": result.statistics,
+        }
+
+        with open(output_path, "w") as f:
+            json.dump(data, f, indent=2)
+
+    def _zero_crossing_detection(
+        self,
+        signal: NDArray[np.floating[Any]],
+        sample_rate: float,
+    ) -> TimingAnalysisResult:
+        """Zero-crossing based clock recovery.
+
+        Detects zero crossings (rising edges) and calculates the most
+        common interval between crossings to determine clock period.
+
+        Args:
+            signal: Input signal array.
+            sample_rate: Sampling rate in Hz.
+
+        Returns:
+            TimingAnalysisResult with recovered clock rate.
+        """
+        # Normalize signal
+        signal_norm = signal - np.mean(signal)
+
+        # Find zero crossings (rising edges)
+        crossings: list[float] = []
+        for i in range(len(signal_norm) - 1):
+            if signal_norm[i] <= 0 and signal_norm[i + 1] > 0:
+                # Linear interpolation to find exact crossing
+                if abs(signal_norm[i + 1] - signal_norm[i]) > 1e-12:
+                    frac = abs(signal_norm[i]) / (abs(signal_norm[i]) + signal_norm[i + 1])
+                    crossing_idx = i + frac
+                    crossings.append(crossing_idx / sample_rate)
+
+        if len(crossings) < 2:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="zcd",
+            )
+
+        # Calculate intervals between crossings
+        intervals = np.diff(crossings)
+
+        if len(intervals) == 0:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="zcd",
+            )
+
+        # Find most common interval (mode) using histogram
+        hist, edges = np.histogram(intervals, bins=min(100, len(intervals)))
+        mode_idx = int(np.argmax(hist))
+        mode_interval = (edges[mode_idx] + edges[mode_idx + 1]) / 2
+
+        # Detected clock rate
+        clock_rate = 1.0 / mode_interval if mode_interval > 0 else 0.0
+
+        # Calculate jitter (RMS of deviations from mode interval)
+        jitter_rms = float(np.std(intervals - mode_interval))
+
+        # Confidence based on histogram peak sharpness
+        confidence = float(hist[mode_idx] / len(intervals))
+
+        # Calculate drift
+        drift_rate = self.analyze_drift(np.array(crossings))
+
+        # Calculate SNR
+        snr_db = self.calculate_snr(signal, clock_rate, sample_rate)
+
+        return TimingAnalysisResult(
+            detected_clock_rate=float(clock_rate),
+            confidence=confidence,
+            jitter_rms=jitter_rms,
+            drift_rate=drift_rate,
+            snr_db=snr_db,
+            method="zcd",
+            statistics={"num_crossings": len(crossings), "mode_interval": mode_interval},
+        )
+
+    def _histogram_method(
+        self,
+        signal: NDArray[np.floating[Any]],
+        sample_rate: float,
+    ) -> TimingAnalysisResult:
+        """Histogram-based clock recovery.
+
+        Uses histogram of signal values to detect logic levels,
+        then finds transitions and computes intervals.
+
+        Args:
+            signal: Input signal array.
+            sample_rate: Sampling rate in Hz.
+
+        Returns:
+            TimingAnalysisResult with recovered clock rate.
+        """
+        # Find logic levels using histogram
+        hist, bin_edges = np.histogram(signal, bins=100)
+        bin_centers = (bin_edges[:-1] + bin_edges[1:]) / 2
+
+        # Find peaks in histogram (logic levels)
+        mid_idx = len(hist) // 2
+        low_peak_idx = int(np.argmax(hist[:mid_idx]))
+        high_peak_idx = mid_idx + int(np.argmax(hist[mid_idx:]))
+
+        low_level = bin_centers[low_peak_idx]
+        high_level = bin_centers[high_peak_idx]
+
+        # Threshold is midpoint
+        threshold = (low_level + high_level) / 2
+
+        # Find transitions (rising or falling edges)
+        transitions: list[float] = []
+        for i in range(len(signal) - 1):
+            if (signal[i] < threshold <= signal[i + 1]) or (signal[i] >= threshold > signal[i + 1]):
+                transitions.append(i / sample_rate)
+
+        if len(transitions) < 2:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="histogram",
+            )
+
+        # Calculate intervals
+        intervals = np.diff(transitions)
+
+        if len(intervals) == 0:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="histogram",
+            )
+
+        # Mode interval
+        mode_interval = float(np.median(intervals))
+        clock_rate = 1.0 / mode_interval if mode_interval > 0 else 0.0
+
+        jitter_rms = float(np.std(intervals))
+        confidence = min(1.0, 1.0 / (1.0 + jitter_rms / mode_interval))
+
+        drift_rate = self.analyze_drift(np.array(transitions))
+        snr_db = self.calculate_snr(signal, clock_rate, sample_rate)
+
+        return TimingAnalysisResult(
+            detected_clock_rate=float(clock_rate),
+            confidence=confidence,
+            jitter_rms=jitter_rms,
+            drift_rate=drift_rate,
+            snr_db=snr_db,
+            method="histogram",
+            statistics={
+                "num_transitions": len(transitions),
+                "low_level": low_level,
+                "high_level": high_level,
+                "threshold": threshold,
+            },
+        )
+
+    def _autocorrelation_method(
+        self,
+        signal: NDArray[np.floating[Any]],
+        sample_rate: float,
+    ) -> TimingAnalysisResult:
+        """Autocorrelation-based clock recovery.
+
+        Autocorrelation shows periodicity in signal. Peak at lag τ
+        indicates period τ.
+
+        Args:
+            signal: Input signal array.
+            sample_rate: Sampling rate in Hz.
+
+        Returns:
+            TimingAnalysisResult with recovered clock rate.
+        """
+        # Handle edge cases
+        if len(signal) == 0 or sample_rate <= 0:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="autocorrelation",
+            )
+
+        # Calculate autocorrelation - handle inf/nan
+        with np.errstate(invalid="ignore"):
+            signal_clean = np.where(np.isfinite(signal), signal, 0.0)
+            signal_norm = signal_clean - np.mean(signal_clean)
+
+        # Use FFT-based autocorrelation for efficiency
+        n = len(signal_norm)
+        fft_signal = np.fft.fft(signal_norm, n=2 * n)
+        autocorr = np.fft.ifft(fft_signal * np.conj(fft_signal)).real
+        autocorr = autocorr[:n]  # Keep positive lags
+
+        # Normalize
+        if autocorr[0] != 0:
+            autocorr = autocorr / autocorr[0]
+
+        # Skip small lags to avoid noise (minimum 1 MHz period = 1 µs)
+        min_lag_samples = max(1, int(sample_rate / 1000000))
+
+        # Find first significant peak (after lag 0)
+        peaks: list[tuple[int, float]] = []
+        for i in range(min_lag_samples, len(autocorr) - 1):
+            if (
+                autocorr[i] > autocorr[i - 1]
+                and autocorr[i] > autocorr[i + 1]
+                and autocorr[i] > 0.3
+            ):  # Significant peak threshold
+                peaks.append((i, autocorr[i]))
+
+        if not peaks:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="autocorrelation",
+            )
+
+        # Use first peak
+        peak_lag, peak_value = peaks[0]
+        period = peak_lag / sample_rate
+        clock_rate = 1.0 / period if period > 0 else 0.0
+
+        # Find secondary peaks for jitter estimation
+        secondary_peaks = [p for p in peaks[1:4] if p[1] > 0.2]
+        if secondary_peaks:
+            # Calculate jitter from peak spread
+            all_peak_lags = [peak_lag] + [p[0] for p in secondary_peaks]
+            # Expected harmonic positions
+            expected_lags = [peak_lag * (i + 1) for i in range(len(all_peak_lags))]
+            jitter_samples = np.std(
+                [abs(a - e) for a, e in zip(all_peak_lags, expected_lags, strict=True)]
+            )
+            jitter_rms = float(jitter_samples / sample_rate)
+        else:
+            jitter_rms = 0.0
+
+        snr_db = self.calculate_snr(signal, clock_rate, sample_rate)
+
+        return TimingAnalysisResult(
+            detected_clock_rate=float(clock_rate),
+            confidence=float(peak_value),
+            jitter_rms=jitter_rms,
+            drift_rate=0.0,  # Autocorrelation doesn't directly measure drift
+            snr_db=snr_db,
+            method="autocorrelation",
+            statistics={
+                "peak_lag": peak_lag,
+                "peak_value": peak_value,
+                "num_peaks": len(peaks),
+            },
+        )
+
+    def _pll_simulation(
+        self,
+        signal: NDArray[np.floating[Any]],
+        sample_rate: float,
+        initial_freq: float,
+    ) -> TimingAnalysisResult:
+        """Phase-locked loop simulation for clock recovery.
+
+        Simulates a PLL to track signal frequency. The PLL adjusts
+        its frequency to minimize phase error with the input signal.
+
+        Args:
+            signal: Input signal array.
+            sample_rate: Sampling rate in Hz.
+            initial_freq: Initial PLL frequency estimate in Hz.
+
+        Returns:
+            TimingAnalysisResult with recovered clock rate.
+
+        References:
+            - Phase-Locked Loop Design Handbook by Dan H. Wolaver
+            - Digital Communications by Proakis & Salehi
+        """
+        damping_factor = 0.707
+        natural_freq = initial_freq * 0.1
+        kp = 2 * damping_factor * natural_freq
+        ki = natural_freq**2
+
+        phase_errors, frequencies = _run_pll_loop(signal, sample_rate, initial_freq, kp, ki)
+        recovered_freq, confidence = _analyze_pll_convergence(frequencies, initial_freq)
+        jitter_rms = float(np.std(phase_errors)) / (2 * np.pi * float(recovered_freq))
+        drift_ppm = _compute_pll_drift(frequencies, recovered_freq)
+        snr_db = self.calculate_snr(signal, recovered_freq, sample_rate)
+
+        return TimingAnalysisResult(
+            detected_clock_rate=recovered_freq,
+            confidence=float(confidence),
+            jitter_rms=jitter_rms,
+            drift_rate=float(drift_ppm),
+            snr_db=snr_db,
+            method="pll",
+            statistics={
+                "initial_freq": initial_freq,
+                "final_freq": recovered_freq,
+                "damping_factor": damping_factor,
+                "natural_freq": natural_freq,
+            },
+        )
+
+    def _fft_method(
+        self,
+        signal: NDArray[np.floating[Any]],
+        sample_rate: float,
+    ) -> TimingAnalysisResult:
+        """FFT-based clock recovery.
+
+        Uses FFT to find the dominant frequency component in the signal.
+
+        Args:
+            signal: Input signal array.
+            sample_rate: Sampling rate in Hz.
+
+        Returns:
+            TimingAnalysisResult with recovered clock rate.
+        """
+        if len(signal) < 64:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="fft",
+            )
+
+        # Remove DC and compute FFT
+        signal_centered = signal - np.mean(signal)
+        n = len(signal_centered)
+        nfft = int(2 ** np.ceil(np.log2(n)))
+        spectrum = np.fft.rfft(signal_centered, n=nfft)
+        freqs = np.fft.rfftfreq(nfft, d=1.0 / sample_rate)
+        magnitude = np.abs(spectrum)
+
+        # Exclude DC component
+        if len(magnitude) > 1:
+            magnitude = magnitude[1:]
+            freqs = freqs[1:]
+
+        if len(magnitude) == 0:
+            return TimingAnalysisResult(
+                detected_clock_rate=0.0,
+                confidence=0.0,
+                jitter_rms=0.0,
+                drift_rate=0.0,
+                snr_db=0.0,
+                method="fft",
+            )
+
+        # Find peak
+        peak_idx = int(np.argmax(magnitude))
+        peak_freq = freqs[peak_idx]
+        peak_mag = magnitude[peak_idx]
+
+        # Parabolic interpolation for more accurate frequency
+        if 0 < peak_idx < len(magnitude) - 1:
+            alpha = magnitude[peak_idx - 1]
+            beta = magnitude[peak_idx]
+            gamma = magnitude[peak_idx + 1]
+
+            if beta > alpha and beta > gamma and abs(alpha - 2 * beta + gamma) > 1e-12:
+                freq_resolution = sample_rate / nfft
+                delta = 0.5 * (alpha - gamma) / (alpha - 2 * beta + gamma)
+                peak_freq = peak_freq + delta * freq_resolution
+
+        # Calculate confidence (ratio of peak to RMS of spectrum)
+        rms_mag = np.sqrt(np.mean(magnitude**2))
+        if rms_mag > 0:
+            confidence = min(1.0, (peak_mag / rms_mag - 1) / 10)
+        else:
+            confidence = 0.0
+
+        snr_db = self.calculate_snr(signal, peak_freq, sample_rate)
+
+        return TimingAnalysisResult(
+            detected_clock_rate=float(peak_freq),
+            confidence=float(confidence),
+            jitter_rms=0.0,  # FFT doesn't directly measure jitter
+            drift_rate=0.0,  # FFT doesn't measure drift
+            snr_db=snr_db,
+            method="fft",
+            statistics={
+                "peak_magnitude": float(peak_mag),
+                "rms_magnitude": float(rms_mag),
+            },
+        )
+
+
+def _run_pll_loop(
+    signal: NDArray[np.floating[Any]],
+    sample_rate: float,
+    initial_freq: float,
+    kp: float,
+    ki: float,
+) -> tuple[list[float], list[float]]:
+    """Run PLL loop to track signal frequency.
+
+    Args:
+        signal: Input signal array.
+        sample_rate: Sampling rate in Hz.
+        initial_freq: Initial frequency estimate.
+        kp: Proportional gain.
+        ki: Integral gain.
+
+    Returns:
+        Tuple of (phase_errors, frequencies).
+    """
+    phase = 0.0
+    frequency = initial_freq
+    phase_error_integral = 0.0
+    phase_errors: list[float] = []
+    frequencies: list[float] = []
+    dt = 1.0 / sample_rate
+
+    for sample in signal:
+        vco_output = np.sin(2 * np.pi * phase)
+        phase_error = sample * vco_output
+        phase_error_integral += phase_error * dt
+        frequency = initial_freq + kp * phase_error + ki * phase_error_integral
+        phase += frequency * dt
+        phase = phase % 1.0
+
+        phase_errors.append(phase_error)
+        frequencies.append(frequency)
+
+    return phase_errors, frequencies
+
+
+def _analyze_pll_convergence(frequencies: list[float], initial_freq: float) -> tuple[float, float]:
+    """Analyze PLL convergence from frequency history.
+
+    Args:
+        frequencies: List of frequency values over time.
+        initial_freq: Initial frequency estimate.
+
+    Returns:
+        Tuple of (recovered_freq, confidence).
+    """
+    stable_start = int(0.9 * len(frequencies))
+    final_frequencies = frequencies[stable_start:]
+
+    if len(final_frequencies) == 0:
+        return initial_freq, 0.0
+
+    recovered_freq = float(np.mean(final_frequencies))
+    freq_std = float(np.std(final_frequencies))
+    confidence = float(max(0.0, min(1.0, 1.0 - freq_std / recovered_freq)))
+
+    return recovered_freq, confidence
+
+
+def _compute_pll_drift(frequencies: list[float], recovered_freq: float) -> float:
+    """Compute frequency drift from PLL tracking.
+
+    Args:
+        frequencies: List of frequency values.
+        recovered_freq: Final recovered frequency.
+
+    Returns:
+        Drift in ppm.
+    """
+    stable_start = int(0.9 * len(frequencies))
+    final_frequencies = frequencies[stable_start:]
+
+    if len(final_frequencies) <= 10:
+        return 0.0
+
+    time_points = np.arange(len(final_frequencies))
+    coeffs = np.polyfit(time_points, final_frequencies, 1)
+    drift_hz_per_sample = coeffs[0]
+    drift_ppm = (drift_hz_per_sample * len(final_frequencies)) / recovered_freq * 1e6
+
+    return float(drift_ppm)
+
+
+__all__ = [
+    "TimingAnalysisResult",
+    "TimingAnalyzer",
+]