Understanding Signals

Signals are the fundamental data carriers in SimFusion. This guide covers signal types, properties, and best practices.

Signal Basics

A signal in SimFusion consists of:

• Data: The actual values (NDArray)
• Sample Rate: Samples per second (Hz)
• Metadata: Units, timestamps, channel names

Creating a Signal

from simfusion import Signal
import numpy as np

# Create a 1kHz sine wave
t = np.linspace(0, 1, 1000)
data = np.sin(2 * np.pi * 10 * t)  # 10 Hz sine

signal = Signal(
    data=data,
    sample_rate=1000,
    unit="V",
    name="Voltage Signal"
)

Signal Types

Continuous vs Discrete

Common Signal Patterns

Audio

Shape: (channels, samples)
Example: (2, 48000)  # Stereo, 1 second at 48kHz

Video

Shape: (frames, height, width, channels)
Example: (30, 1080, 1920, 3)  # 1 second at 30fps

Sensor Array

Shape: (sensors, timepoints)
Example: (64, 1000)  # 64 EEG channels, 1 second at 1kHz

Sampling Theory

Nyquist Rate

To avoid aliasing, sample at >2x the highest frequency:

Signal bandwidth: 0-20 kHz
Minimum sample rate: 40 kHz
Recommended: 44.1 kHz or 48 kHz

Resampling

Change sample rate while preserving information:

# Downsample from 48kHz to 16kHz
downsampled = signal.resample(target_rate=16000)

# Upsample with interpolation
upsampled = signal.resample(target_rate=96000, method='cubic')

Signal Operations

Arithmetic

Signals support vectorized operations:

# Element-wise operations
result = signal1 + signal2
result = signal1 * 2.5
result = Signal.concat([sig1, sig2], axis=0)

Windowing

Apply time windows for spectral analysis:

# Hanning window
windowed = signal.window('hanning', size=1024)

# Custom window
window = np.hamming(512)
windowed = signal.apply_window(window)

Metadata and Annotations

Adding Metadata

signal.set_metadata({
    "sensor_id": "TEMP_01",
    "location": "Room A",
    "calibration_date": "2024-01-15"
})

Time Stamps

For irregular sampling or event data:

timestamps = np.array([0.0, 0.1, 0.25, 0.3, 0.5])
event_signal = Signal(data=values, timestamps=timestamps)

Best Practices

1. Always specify units — Prevents calculation errors
2. Check sample rates — Mismatched rates cause issues
3. Use appropriate dtypes — float32 vs float64 tradeoffs
4. Chunk large signals — Memory-efficient processing

See Also

• Block System