> ## Documentation Index > Fetch the complete documentation index at: https://mintlify.com/opentrack/opentrack/llms.txt > Use this file to discover all available pages before exploring further. # Filter Overview > Understand OpenTrack's filtering algorithms and choose the right filter for your use case Filters in OpenTrack smooth the raw tracking data from your camera or sensor to reduce jitter while maintaining responsiveness. Each filter uses a different approach to balance smoothness and lag. ## Available Filters OpenTrack provides several filtering algorithms, each optimized for different use cases: Advanced adaptive filter with multi-head composition and motion prediction Exponentially weighted moving average with dynamic noise detection Quaternion-based filter using Slerp interpolation for rotations Spline-based acceleration filter for velocity-dependent smoothing Combined filter using Accela smoothing with Hamilton's quaternion rotations No filtering - raw tracking data passes through unmodified **Accela + Hamilton** combines Accela's velocity-dependent smoothing with Hamilton's quaternion-based rotation handling for smoother gimbal-free rotations. **None** can be useful for debugging or when using external smoothing. ## Filter Comparison | Filter | Best For | Complexity | CPU Usage | Adaptive | | --------------------- | ---------------------------------------------- | ---------- | ----------- | -------- | | **Alpha Spectrum** | Advanced users wanting maximum control | High | Medium-High | Yes | | **EWMA** | General purpose, automatic noise adaptation | Medium | Low | Yes | | **Hamilton** | Smooth rotation with quaternion interpolation | Medium | Low | Partial | | **Accela** | Fast action games, velocity-dependent response | Low | Low | Yes | | **Accela + Hamilton** | Combined smoothing with quaternion rotations | Medium | Low | Yes | | **None** | No filtering, raw data | Minimal | None | No | ## Choosing a Filter ### For Simulators and Precision Work Use **EWMA** or **Alpha Spectrum** (Stable profile): * Maximum smoothness when stationary * Minimal jitter for precise aiming * Responsive enough for slow pans ```text EWMA Settings theme={null} Min Smoothing: 0.02 Max Smoothing: 0.7 Curve: 0.8 ``` ```text Alpha Spectrum Settings theme={null} Rotation Min/Max: 0.03 / 0.55 Rotation Curve: 1.8 Adaptive Mode: Off ``` ### For Fast Action Games Use **Accela** or **Alpha Spectrum** (Responsive profile): * Minimal lag during rapid movements * Velocity-based acceleration * Quick response to fast turns ```text Accela Settings theme={null} Rotation Sensitivity: 1.5 Translation Sensitivity: 1.0 Rotation Deadzone: 0.03° ``` ```text Alpha Spectrum Settings theme={null} Rotation Min/Max: 0.06 / 0.85 Rotation Curve: 0.9 Adaptive Mode: On ``` ### For Smooth Cinematic Movement Use **Hamilton**: * True quaternion interpolation (Slerp) * No gimbal lock artifacts * Smooth, natural rotation paths ```text Hamilton Settings theme={null} Max Rotation: 0.01 Rotation Power: 0.01 Rotation Deadzone: 0.01° ``` ## Key Concepts All filters balance two competing goals: * **More smoothing** = less jitter but more lag * **Less smoothing** = faster response but more visible shake Adaptive filters try to automatically adjust this balance based on detected motion. Deadzones suppress tiny movements below a threshold to eliminate noise: * **Rotation deadzone**: ignores small rotational movements (degrees) * **Translation deadzone**: ignores small positional movements (millimeters) Too high = inputs feel sticky. Too low = visible jitter. The alpha parameter controls the filter weight: * `output = output + alpha * (input - output)` * Alpha of 1.0 = no smoothing (instant response) * Alpha of 0.0 = infinite smoothing (no movement) Adaptive filters dynamically adjust alpha based on motion energy. Advanced filters (EWMA, Alpha Spectrum) measure noise statistics: * Track variance of recent deltas * Distinguish intentional movement from sensor noise * Adapt smoothing strength automatically This provides heavier smoothing when still, lighter smoothing when moving. ## Filter Pipeline Filters process tracking data in the pipeline: ```mermaid theme={null} graph LR A[Tracker] --> B[Raw Input] B --> C[Filter] C --> D[Mapping] D --> E[Output] ``` 1. **Tracker** provides raw 6DOF pose data 2. **Filter** smooths the data (this module) 3. **Mapping** applies curves and response adjustments 4. **Output** sends to game/simulator Filters operate on the raw tracking data before mapping curves are applied. This means filter settings are independent of your response curve configuration. ## Common Parameters Most filters share these parameters: Minimum smoothing factor for rotation (0.005-0.4). Lower = more smoothing at rest. Maximum smoothing factor for rotation (0.02-1.0). Higher = faster response during motion. Shape of the transition from min to max (0.2-8.0). Higher = stays smooth longer. Suppresses rotation below this threshold in degrees (0.0-0.3°). Minimum smoothing factor for X/Y/Z position (0.005-0.4). Maximum smoothing factor for X/Y/Z position (0.02-1.0). Shape of the transition from min to max for translation (0.2-8.0). Suppresses translation below this threshold in millimeters (0.0-2.0 mm). ## Troubleshooting Before adjusting filter settings, ensure your tracker is working properly and securely mounted. No filter can fix tracking hardware issues. **View jitters when stationary:** * Increase minimum smoothing or deadzone * Check for Brownian/noise contribution in status (Alpha Spectrum) **Feels laggy during turns:** * Increase maximum smoothing * Decrease curve value * Enable Adaptive Mode (Alpha Spectrum) **Small movements ignored:** * Decrease deadzone values * Check that tracker sensitivity is adequate **Feels too twitchy/nervous:** * Decrease maximum smoothing * Increase curve value * Increase minimum smoothing ## Next Steps Learn about the most advanced filter with predictive heads Understand the automatic noise-adaptive filter Explore quaternion-based smooth rotations Configure velocity-dependent acceleration filtering