Preprocessing

From QuB

Jump to: navigation, search
Prev: How to import data Outline Next: Single Channel Analysis


Contents

Baseline Correction

Baseline correction acts on the active channel.

Constant offset

QuB subtracts a constant baseline offset from each data point, whether or not "apply baseline" is selected. The offset defaults to 0. To change it, select a piece of data that should be 0, right-click, and choose "Set baseline."

"Set baseline" also puts the selection in a list named "Baseline selection". Data in the baseline selection is used to Erase (see Editing, below).

Piecewise linear

To correct a moving baseline by approximating it with line segments, you define several baseline nodes. QuB constructs a baseline by interpolating between nodes, and can subtract it for analysis or saving.

Make sure "bline apply" (at the top) is not checked
Define the baseline nodes
  1. select the first bit of closed in your data, then right-click->Add a baseline node (or press N). If you don't see it appear, make sure "Draw baseline nodes" is checked in Data Properties.
  2. add more nodes at inflection points, or wherever the data begins to diverge from the (pink) baseline.
Check "bline apply"
your data should be beautiful

By default, algorithms operate on data "processed as displayed." If you have changed this setting, make sure it is applying the baseline (or not) as you expect.

Algorithmic

The button Modeling:Idl/Base tracks baseline changes while idealizing. It may not be accurate for noisy data.

The button Preprocessing:Baseline has limited success adding baseline nodes.

Filtering

To apply a low-pass filter to the active channel, check the box next to "Fc [kHz]" and type in a frequency or use the slider. A faster "moving average" filter is available in Data Properties.

By default, algorithms operate on data "processed as displayed." If you have changed this setting, make sure it is applying the filter (or not) as you expect.

Cluster Detection

Recordings often have long stretches of baseline punctuated by clusters (bursts) of channel activity. The desensitized baseline stretches are poorly understood, and the clusters may come from different channels, so you want to analyze each burst separately. Here's a simple procedure to create a selection list of clusters. A more advanced procedure is covered in the heterogeneous bursts tutorial.

  • Idealize the whole file -- Half-amp without stats is fastest
  • Determine a threshold between short closed intervals and inter-burst intervals.
  • Click ChopIdl (under Preprocessing) and enter your threshold as the "burst terminator"
  • Change data source to "list" and do a nice idealization e.g. with SKM

Editing

Cut, copy, paste, and delete all work normally. You can also make a segment break (from the right-click menu), and move a data point up and down (ctrl-shift-arrows).

You might want to delete a hopelessly noisy piece of data, but it can cause problems: selection lists are kept as a list of offsets, and aren't updated when you delete. The solution is to select the piece and "Erase" instead. Erase replaces the selected data with the contents of the "Baseline selection" (see "Constant Offset", above).

Extraction and Conversion

QuB can save all or part of a data file in a number of formats, preprocessed as you like. i.e. you can correct the baseline, build a selection list of clean bursts, and extract the baseline-corrected selections to a new file for further analysis. See Preprocessing:Extract for details.


Prev: How to import data Outline Next: Single Channel Analysis
Personal tools