Engineering summary
A practical walkthrough of cleaning raw infrasound records: correct baseline drift, bandpass-filter the sub-20 Hz band of interest, calibrate to Pascals, then detrend, window and analyze.
In short: Infrasound signal processing follows six steps: pre-process the raw record (digitize, time-sync, remove outliers); correct baseline drift by subtracting a constant, a polynomial fit or a moving average; bandpass-filter to the band of interest, below 20 Hz, with a Butterworth, Chebyshev or Bessel design; convert the filtered signal to Pascals using the sensor sensitivity; detrend, window and apply extra noise reduction; then measure event amplitude, frequency content, phase and duration.
Processing infrasound signals is a critical step in analyzing data from phenomena that generate low-frequency acoustic waves. These waves can originate from various sources, including natural events (volcanic eruptions, tornadoes, etc.), man-made explosions, and large machinery. The goal of post-processing infrasound data is to enhance signal quality, making it easier to detect and analyze these phenomena. Here’s a detailed explanation of the post-processing steps, including baseline correction and bandpass filtering:
1. Pre-Processing:
Before diving into specific post-processing techniques, it’s essential to ensure that the raw infrasound data is correctly pre-processed. This might include steps like digitization (if working with analog signals), ensuring correct time synchronization, and initial data cleaning to remove any obvious errors or outliers.
2. Baseline Correction:
Infrasound signals can be affected by drift and shifts in the baseline, which can obscure the true signal or make analysis more difficult. Baseline correction aims to adjust the signal so that its baseline is stable over time, which is crucial for accurate measurement and analysis.
- Identify the Baseline: Using statistical methods or by visually inspecting the signal, determine the baseline level. This could be a constant value that the signal should nominally return to in the absence of any events.
- Correction Methods: Apply a method to correct the baseline drift. This might involve subtracting the identified baseline value from the entire signal or using more sophisticated methods like polynomial fitting or moving average subtraction to adjust dynamically for baseline changes over time.
3. Bandpass Filtering:
Bandpass filtering is used to remove noise and irrelevant frequencies that do not contribute to the signal of interest. By focusing on a specific frequency band, it enhances the signal’s detectability and clarity.
- Determine Frequency Band: Based on the source and nature of the infrasound signal, identify the relevant frequency range. Infrasound signals typically fall below 20 Hz, but the exact band of interest can vary depending on the source and environment.
- Apply Filter: Use a bandpass filter to retain only the frequencies within the desired range. Common types of bandpass filters include Butterworth, Chebyshev, and Bessel filters, each with its characteristics in terms of phase shift and roll-off rate. The choice of filter depends on the analysis requirements and the characteristics of the signal.
- Filter Design: The filter can be designed digitally in software, specifying the passband (the range of frequencies to keep), the stopband (frequencies to be attenuated), and the filter order (which affects the steepness of the roll-off). Higher-order filters provide sharper cutoffs but can introduce phase distortion.
4. Convert to Pascal Values:
After filtering, the signal is often converted into physical units (e.g., Pascals) for analysis. This step involves calibrating the signal based on the sensitivity of the infrasound sensors used to record the data and any known reference levels. Calibration ensures that the signal amplitude reflects the true pressure variations caused by the infrasound source.
For detailed information on this step, visit this link:
5. Additional Processing Steps:
Depending on the application, further processing steps might be necessary, such as:
- Detrending: Removing linear trends from the data to focus on the signal fluctuations.
- Windowing: Applying a window function to manage the signal’s start and end points, useful for Fourier analysis.
- Noise Reduction: Implementing additional noise reduction techniques, such as spectral subtraction or signal enhancement algorithms, to improve signal quality.
6. Analysis:
After post-processing, the signal is ready for analysis, which could involve identifying specific events, measuring their characteristics (amplitude, frequency content, phase, duration), and interpreting their source and impact.
In summary, the post-processing of infrasound raw signals, including baseline correction and bandpass filtering, is essential for accurately interpreting the data. These steps help in enhancing the signal quality by eliminating noise and irrelevant information, thereby facilitating a more precise analysis of the infrasound phenomena captured by the sensors.
Questions?
Email us at support@quakelogic.net or call us at +1-916-899-0391.
Key Takeaways
- Infrasound signals fall below 20 Hz; the exact passband depends on the source (volcano, tornado, explosion, machinery).
- Baseline correction options: constant subtraction, polynomial fitting or moving-average subtraction.
- Butterworth, Chebyshev and Bessel bandpass filters differ in phase shift and roll-off; higher order sharpens cutoff but adds phase distortion.
- Convert to Pascals after filtering using sensor sensitivity and known reference levels.
- Optional steps: detrending, windowing for Fourier analysis, and spectral subtraction for noise reduction.
Frequently Asked Questions
Infrasound lies below 20 Hz, the lower limit of human hearing, so processing typically keeps a passband somewhere under that limit. The exact band depends on the source: volcanic eruptions, tornadoes, explosions and large machinery each concentrate energy in different parts of the infrasound range, and local environmental noise also shapes the choice. The passband, stopband and filter order are then specified when the bandpass filter is designed.
First identify the baseline, the level the signal should return to with no events, either statistically or by visual inspection. Then remove the drift: subtract a constant baseline value from the whole record for simple offsets, or use polynomial fitting or moving-average subtraction when the baseline wanders over time. A stable baseline is needed before filtering and conversion to Pascals so that amplitudes are measured correctly.
Each filter trades off differently. Butterworth gives a flat passband with moderate roll-off; Chebyshev gives a steeper roll-off at the cost of passband ripple; Bessel preserves phase and waveform shape best but rolls off most gently. Higher filter orders sharpen the cutoff but introduce more phase distortion. Choose by what the analysis needs: waveform timing favors Bessel, sharp band separation favors Chebyshev, general use favors Butterworth.
Raw infrasound records are in sensor output units, such as volts or digitizer counts. Converting to Pascals calibrates the signal using the sensor's sensitivity and any known reference levels so that amplitude represents the true pressure variation produced by the source. Physical units make results comparable across sensors and sites and let analysts quantify event strength. QuakeLogic's separate guide on converting infrasound sensor data to Pascal details the calculation.
Six steps: pre-processing (digitization if analog, time synchronization, removal of obvious errors and outliers); baseline correction to remove drift; bandpass filtering to retain the frequency band of interest; conversion to Pascals using sensor sensitivity; optional detrending, windowing for Fourier analysis and noise reduction such as spectral subtraction; and finally analysis, where events are identified and their amplitude, frequency content, phase and duration are measured and interpreted.
Related Articles
Need help with a project? QuakeLogic designs and supplies seismic monitoring, earthquake early warning, structural health monitoring, infrasound, and shake table systems. Browse QuakeLogic products or contact our engineering team with your site, structure, or test requirements.
Discover more from QuakeLogic
Subscribe to get the latest posts sent to your email.
Topic cluster
Related engineering knowledge areas
- Structural Health MonitoringMonitoring for bridges, buildings, dams, tunnels, industrial facilities, and resilient infrastructure.
- Earthquake Early WarningOn-site detection, alerting workflows, seismic switches, and critical infrastructure warning systems.
- Infrasound MonitoringLow-frequency acoustic sensing for environmental noise, blast, UAV, volcano, and defense applications.
- Shake TablesUniaxial, biaxial, vertical, geotechnical, and multi-axis shake table testing systems.
Definitions and references
Terms, standards, and source cues
- SHM: related to Structural Health Monitoring in this QuakeLogic knowledge cluster.
- damage detection: related to Structural Health Monitoring in this QuakeLogic knowledge cluster.
- earthquake early warning: related to Earthquake Early Warning in this QuakeLogic knowledge cluster.
- seismic switch: related to Earthquake Early Warning in this QuakeLogic knowledge cluster.
- infrasound sensors: related to Infrasound Monitoring in this QuakeLogic knowledge cluster.
- low-frequency noise: related to Infrasound Monitoring in this QuakeLogic knowledge cluster.
- shake tables: related to Shake Tables in this QuakeLogic knowledge cluster.
- AC156: related to Shake Tables in this QuakeLogic knowledge cluster.
Need project support?
Talk with QuakeLogic about monitoring, testing, or warning systems.
Get engineering guidance for seismic monitoring, structural health monitoring, infrasound, vibration, earthquake early warning, and shake table applications.
