Deglitching and Multiple Excitation Removal
In X-ray Absorption Spectroscopy (XAS), glitches—isolated high or low intensity spikes—can arise from detector noise, electronic interference, or artifacts like monochromator diffraction peaks. These artifacts distort the signal and must be removed prior to Fourier transforms or fitting routines.
Additionally, multiple excitations within the scanned energy range (common in some elements) can introduce discontinuities in the data. The deglitch command in EstraPy is designed to correct these issues.
Deglitch Command
The deglitch command handles the removal of spurious datapoints, visible in the acquired signals as very high (or low) intensity regions.
deglitch <range> [--options] <finder> [--options] <remover> [--options]
The command is modular, allowing users to specify:
- a range to apply the glitch removal,
- a glitch finder method to detect outliers,
- a glitch remover method to correct or interpolate affected regions.
Multiple finder and remover options can be combined to create a custom deglitching workflow. The deglitch command presents a variety of options, allowing to combine the functionality of glitch-finding and glitch-removal methods.
Main options
| Argument | Explanation |
|---|---|
<range> | The range considered to find the glitch. See Number and unit specification for the range syntax explanation. |
--column / -c <column> | Selects which column to process for glitch detection. |
Finder options
| Finder | Description | Required arguments |
|---|---|---|
force | Flags the entire selected range as glitched, without analysis. | - |
variance | Flags outliers based on standard deviation over subranges. | --width, --pvalue |
smooth | Applies LOWESS smoothing and flags the deviations. | --fraction, --width, --pvalue |
polynomial | Fits a polynomial, then flags deviations from the fit. | --degree, --width, --pvalue |
even | Compares even vs. odd indexed data via regression, then flags deviations. | --pvalue |
All methods flag a value as a glitch if its deviation exceeds the specified statistical threshold (--pvalue), relative to the calculated standard deviation.
| Argument | Explanation |
|---|---|
--width / -w <value> | Number of subranges used to compute the median standard deviation. Default is 10. |
--fraction / -f <value> | The fraction of points to use for LOWESS smoothing (smooth). Default is 0.3. |
--degree / -d <value> | The degree of the polynomial to use for the polynomial finder. Default is 1 (linear). |
--pvalue / -p <value> | Statistical cutoff (e.g. 0.01 = 99% confidence) for flagging a glitch. Default is 0.001 (99.9% confidence). |
Removal method options
After glitches are identified, a remover method is applied to correct or eliminate the affected data points.
| Remover | Description | Required arguments |
|---|---|---|
remove | Removes the identified glitch from the data, without replacing it. | - |
base | Replaces glitches using a baseline estimated during the finder phase. Only works when --column is set to a, and the finder supports baseline estimation. | - |
smooth | Fits the non-glitched data using LOWESS regression, then interpolates over the glitch regions. | --fraction |
| Argument | Explanation |
|---|---|
--fraction / -f <value> | The fraction of points to use for LOWESS smoothing (smooth). Default is 0.05. |
--noise / -n | Adds Gaussian noise to the interpolated segment to avoid visibly perfect smoothing. |
Multiple Excitation correction
Within some datasets, the data might contain secondary excitations, which change the shape of the data resulting in artifacts in the Fourier transform and poor fitting. EstraPy offers a command to manually remove these contributions, which can be modelled as either a simple function or as a shifted version of the data.
multiedge [--axis] <function> <parameters>
| Argument | Explanation |
|---|---|
--energy / -E | Calculates the function on the energy axis. This is the default behavior. |
--relenergy / -e | Calculates the function on the relative energy axis. |
--kvector / -k | Calculates the function on the k-wavevector axis. |
EstraPy offers the following functions:
| Function | Description |
|---|---|
atan <b> <a> <c> | Models the excitation as an arctangent. |
erf <b> <a> <c> | Models the excitation as an error-function. |
tanh <b> <a> <c> | Models the excitation as an hyperbolic tangent. |
exp <b> <a> <c> | Models the excitation as an exponential. |
All the step functions have been modified, so that a describes the total height of the curve, and c describes the distance from b where the curve evaluates to 10% (or 90%) of a. We can thus identify the following notable points:
| \(x\) | \(\text{atan}(x)\) | \(\text{erf}(x)\) | \(\text{exp}(x)\) | \(\text{exp}(x)\) c<0 | \(\text{tanh}(x)\) |
|---|---|---|---|---|---|
| \(-\infty\) | \(0\) | \(0\) | \(0\) | \(0\) | \(0\) |
| \(b-|c|\) | \(0.1a\) | \(0.1a\) | \(0\) | \(0.1a\) | \(0.1a\) |
| \(b\) | \(0.5a\) | \(0.5a\) | \(0\) | \(a\) | \(0.5a\) |
| \(b+|c|\) | \(0.9a\) | \(0.9a\) | \(0.9a\) | \(a\) | \(0.9a\) |
| \(\infty\) | \(a\) | \(a\) | \(a\) | \(a\) | \(a\) |
The exact mathematical description of each function is described below.
| Function | Mathematical expression |
|---|---|
atan | \(a\cdot\left(\dfrac{1}{\pi}\cdot\tan^{-1}{\left(\tan\left(0.4\pi\right)\cdot\dfrac{x-b}{c}\right)} + 0.5\right)\) |
erf | \(\dfrac{a}{2}\cdot\left(1+\text{erf}{\left(\text{erf}^{-1}{\left(0.8\right)\cdot\dfrac{x-b}{c}}\right)}\right)\) |
exp | \(a\cdot\left(1 - \exp\left(\log\left(0.1\right)\cdot\frac{x-b}{c}\right)\right)\) |
tanh | \(\dfrac{a}{2}\cdot\left(1+\tanh{\left(\tanh^{-1}(0.8)\cdot\dfrac{x-b}{c}\right)}\right)\) |