Rebin

The rebin command bins data into fixed intervals and averages all points within each bin. This is a destructive operation that reduces noise but permanently replaces original data.

(Note that the original data file remains unchanged; only the in-memory dataset is modified.)

Basic Usage 

rebin <range> <--interval <step> | --number <n>> [options]

where:

  • <range> defines the binning limits
  • Either --interval or --number specifies bin size (exactly one required)

Command Options

Option Description
<range> Range for rebinning (required). Must have finite bounds. See range specification for details.
--interval <step> Fixed bin width (mutually exclusive with --number). Must include units.
--number <n> Number of bins to create (mutually exclusive with --interval).
--axis <axis>
-x <axis>		 Axis column for binning (auto-inferred from range units if omitted).
--domain <domain> Domain in which to rebin (default: reciprocal). Options: reciprocal, fourier.
--fix-points Interpolate bin averages back to exact bin centers (preserves target axis values).

Binning Method

The command:

  1. Creates bin boundaries at half-intervals: [range[0] - step/2, ..., range[1] + step/2]
  2. Assigns each data point to a bin
  3. Computes the mean of all points in each bin (all columns)
  4. Optionally interpolates to exact bin centers if --fix-points is set

Fix Points Option

  • Default (no --fix-points): Bin centers are the mean axis values within each bin (may not align exactly with target grid)
  • With --fix-points: After averaging, cubic spline interpolation repositions points to exact bin centers

Use --fix-points when you need bin centers at precise positions (e.g., for subsequent operations requiring exact alignment between files).

Examples 

# Rebin energy axis to 1 eV intervals
rebin 8000eV 9000eV --interval 1eV

# Create exactly 100 bins in k-space
rebin 2k 14k --number 100

# Rebin Fourier domain with fixed bin centers
rebin 0A 5A --interval 0.1A --domain fourier --fix-points

# Explicit axis specification
rebin 0 1000 --interval 2 --axis E --domain reciprocal

Behavior

Rebinning:

  • Reduces noise by averaging multiple points within each bin
  • Downsamples high-resolution data for faster processing
  • Removes outliers if most points in a bin are consistent
  • Smooths irregular step sizes into uniform intervals

Destructive operation: Original data is replaced with binned averages. The original data file remains unchanged; only the in-memory dataset is modified. Data points outside the extended range [range[0] - step/2, range[1] + step/2] are discarded.

Use Cases

  • Noise reduction: Average out high-frequency noise while preserving overall shape
  • Downsampling: Reduce point count from oversampled scans
  • Uniformity: Convert variable-spacing scans to fixed intervals
  • Pre-processing for averaging: Align multiple scans to identical grids before merging

Tips and Best Practices

  1. Choose appropriate bin size:

    • Too small → minimal noise reduction
    • Too large → loss of spectral features
    • Typical: 0.5–2 eV for energy, 0.05–0.2 k⁻¹ for k-space

  2. Check data spacing: Bin width should be larger than original step size to achieve averaging

  3. Use for noise, not upsampling: Rebinning cannot increase resolution; use interpolate for upsampling (interpolation does not add information, only resamples existing data)

  4. Edge effects: Data near range boundaries may be lost if bins extend outside the data range

Comparison with Interpolate

Feature rebin interpolate
Method Binning + averaging Cubic spline fitting
Noise reduction Yes (averaging reduces noise) No
Outlier handling Robust (majority vote per bin) Sensitive (fits through all points)
Use case Noise reduction + downsampling Resampling to new grid

See also: