06. Filter Descriptions

Because Peakaboo uses a plug-in system for fitlers, this cannot be an exhaustive list, but it does cover all built-in filters.

Background Filters

Brukner

This filter removes background over several iterations by smoothing the data and taking the minimum of the unsmoothed and smoothed data for each channel on each pass.

Linear Trim

This filter examines all pairs of points which are n channels apart (ie (1, 10), (2, 11), … where n = 10). For each pair of points, any signal which exceeds a straight line connecting the two points is truncated.

Polynomial

This filter attempts to fit a series of parabolic (or higher order single-term) curves under the data, with a curve centred at each channel, and attempting to make each curve as tall as possible while still staying completely under the spectrum. The union of these curves is calculated and subtracted from the original data.

Square Snip

This is a very fast background removal method based on the Peak Stripping algorithm. It iteratively replaces signal with the average of the points (-window, +window) channels apart if that average is less than the existing signal. By taking a double square root of the signal and then reversing it afterwards, the number if iterations required is greatly reduced. Because noise-reduction filters are separate and composable, this version of the algorithm does not do any smoothing itself.

Spectrum Subtraction

This background filter loads another dataset and subtracts the other dataset's average spectrum from each spectrum in this one.

Custom Compton

This filter requires manual configuration to work properly. It will attempt to fit the compton scattering curve between start and end channels.

Signal Smoothing Filters

Fourier Low-Pass

This filter transforms the spectral data with a Fourier Transformation into a frequency domain. Data from a high frequency range (noise) is filtered out, while lower frequencies (peaks, background) are passed through.

Weighted Averaging

This filter refines the values of each point in a scan by sampling it and the n points to either side of it, and replacing it with an exponentially weighted average of the sampled points.

Savitsky-Golay

This filter attempts to remove noise by fitting a polynomial to each point P_i and its surrounding points P_(i-n) … P_(i+n), and then taking the value of the polynomial at P_i. This filter is generally fast, while minimizing peak distortion. A moving average may be considered a special case of this filter with a polynomial of order 1.

Spring Smoothing

This filter operates on the assumption that weak signal should be smoothed more than strong signal. It treats each pair of adjacent points as if they were connected by a spring. With each iteration, a tension force draws neighbouring points closer together.

The Force Multiplier controls how strongly a pair of points are pulled together, and the Force Falloff Rate controls how aggressively stronger signal is anchored in place, unmoved by tension forces. This prevents stronger intensity points such as peak shapes from being distorted by the smoothing algorithm.

Low Statistics

This filter smooths signal per-channel by shrinking a moving-average window until either:

• The signal in the window is less than Max Signal

• Both of the following are true:

  • The signal in the window is less than Threshold × sqrt(centerpoint)

  • The slopes of the left vs. right windows is less than Max Slope

Mathematical Filters

Add

This filter adds a constant value to all points on a spectrum.

Subtract

This filter subtracts a constant value to all points on a spectrum.

Multiply

This filter multiplies all points on a spectrum by a constant value.

Derivative

This filter transforms the data such that each channel represents the difference between itself and the channel before it.

Integral

This filter transforms the data such that each channel represents the sum of itself and all channels prior to it.

Advanced Filters

Spectrum Normalizer

This filter scales each spectrum so that the intensity of the selected channel(s) matches the given noramlized intensity. Channel selection is one of max intensity, average intensity, or region of interest.

Dataset Normalizer

This filter scales each spectrum's intensity against the entire dataset based on the options selected.

Filter Partial Spectrum

This filter allows the application of another filter to a portion of a spectrum.