wiki:PreProcessing

The preprocessing workflow supports

  • decimation - data are resampled at a lower sampling frequency. This functions as an anti-alias filter
  • frequency domain windowing - data can be windowed in the frequency domain, such filters are zero-phase but can cause some distortion. As such, artificial dead time is imposed when applying this filter based on the impulse response (plotted when designing).
  • bandpass filtering - A zero phase bandpass filter can further be applied to the data. The filter is run in both directions in order to achieve the zero-phase behaviour (similar to Matlab's filtfilt).
  • time-domain noise suppression - noise suppression using reference coils is applied in the time domain. One benefit of this approach is that noise spikes to not invalidate an entire time series. As such, for noisy sites with many spikes, more data is generally retained. Outlier detection is applied at stack.
  • data summing - for multiple loop datasets, composite data can be generated.

They can be implemented in any order but following the order of the GUI tends to produce the best results. Reasonable defaults are provided for all parameters, but users may want to test with varying settings depending on data specifics.

At any point during the processing, you may notice on the log page a human readable description of the processing steps. This description is valid YAML and will become part of the exported data. For example:

!AkvoData
Akvo_VERSION: 1.0.5
Import:
  GMR Header: /path/to/datafile
  data channels: [1]
  instrument dead time: 0.0055
  opened: '2017-10-17T15:35:05.216686'
  pulse Type: FID
  pulse records: Pulse 1
  reference channels: [3, 5]
  stacks: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
Processing:
  Resample:
    downsample factor: ['5']
    truncate length: ['0']
  Window filter: {centre: '2245.0', type: Hamming, width: '600.0'}
  Bandpass filter: {central_nu: '2245.0', gpass: '0.01', gstop: '5.0', passband: '50.0',
    stopband: '280.0', type: Butterworth}
  TD noise cancellation: {PCA: 'Yes', lambda: '0.99', mu: '0.01', n_Taps: '200', truncate: '800.0'}

Following this processing flow you should see something like this: preproc

Last modified 3 years ago Last modified on Oct 17, 2017 10:03:55 PM

Attachments (1)

Download all attachments as: .zip