ISP Macro Processing Pipeline

The Macro module in ISP allows users to apply a series of signal processing steps to seismic data using a flexible, GUI-driven interface. Macros are applied sequentially and can include filtering, denoising, normalization, and transformation operations. This macro system enables reproducible, streamlined signal processing for all ISP toolboxes. Adjust each block interactively and view the effect in real time.

This page describes the macro structure, available processing tools, and typical workflows with illustrative examples.

Macro Structure

The macro tree defines each processing step in logical order. Each node represents a category or method with customizable parameters.

Detrending Methods

Remove mean / trend from the signal

  • linear — Remove a best-fit line
  • demean — Subtract mean value
  • polynomial — Fit and subtract a polynomial (e.g., order 3)
  • spline — Remove a smooth spline trend

Taper

Applies a window function to reduce edge effects

  • Parameter: max_percentage (e.g., 0.05 = 5%)
  • Supported Windows:
  • cosine, barthann, bartlett, blackman, blackmanharris
  • bohman, boxcar, chebwin, flattop, gaussian
  • general_gaussian, hamming, hann, kaiser, nuttall
  • parzen, slepian, triang

Normalize

Scales data globally

  • 0: Normalize by maximum
  • Any other value: Divides trace by specified value

Filtering

Band filter operations

  • Parameters: freqmin, freqmax, zerophase, poles
  • Methods:
  • bandpass, lowpass, highpass, bandstop
  • cheby1, cheby2, elliptic, bessel

freqmin must be less than freqmax.

Wiener Filter

1D Wiener de-noising

  • Parameters: time_window (sec), noise_power
  • Tip: Use 1s window; if noise_power=0, it's estimated from variance.

Resample

Adjust signal's sampling rate

  • Parameters: new_sampling_rate, pre_filter (recommended True)

Fill Gaps

Reconstruct missing data

  • Methods: latest or interpolate

Differentiation

  • Methods: gradient

Integration

  • Methods: cumtrapz, spline

Time Shifting

Shift traces by custom time offsets

  • Parameter: List of shifts (in seconds)

Remove Instrument Response

  • Parameters: pre-filter corners, water_level (60–90), output units

Add White Noise

Simulate noise at defined power

  • Parameter: Noise power in dB relative to signal max

Spectral Whitening

Flattens spectrum using a defined bandwidth

  • Parameters: spectral_bandwidth, taper (True/False)

Time Normalization

Normalize signal in time

  • Parameters: time_window, method
  • Methods:
  • time-normalization, 1-bit, clipping (with iterations)

Wavelet Denoise

Wavelet-based signal denoising

  • Parameters: wavelet_family (e.g., sym8), threshold (e.g., 0.05)

Smoothing

Reduces high-frequency noise

  • Parameters: time_window, fwhm
  • Methods: mean, gaussian, tkeo

Spike Removal

Remove outliers or spikes

  • Parameters: time_window, threshold (in std units)

Filter Design Example

In this example, we use a regional earthquake trace expected to contain energy in the 0.5–8 Hz band. We:

  1. Remove linear trend
  2. Apply a cosine taper (5%)
  3. Use a bandpass filter with:
  4. freqmin=0.5
  5. freqmax=8
  6. poles=4
  7. zerophase=True

Screenshot

The top panel shows the raw signal, and the bottom shows the filtered version:

Screenshot

Signal Denoising Example

We now show how to denoise a noisy ocean-bottom seismometer trace using:

  • Wiener filter
  • Wavelet denoising

Step 1: Wiener Filter

Screenshot

Step 2: Wavelet Transform

Screenshot

Final Output

Screenshot

Removing Spikes Example

In this example we use the Hampel algorithm to remove the waveform spikes. In the image is shown the effect in the time domain and in the spectrogram after applying the "removing spikes" macro.

remove_spikes