DC removal

[sbonnefo]

Hi Marc,

I can not remember if the mean value is automatically supressed from signal before the computation of H/V ratio and/or FK processing?

If not, is it possible to implement it?
If yes, forget my post !

Thanks a lot
Sylvette

[admin]

For H/V and for all array processing the mean is removed on a time window basis. So no need for further implementations.

For spectra (H/V and spectrum tools), the time windows are processed this way:

Code: Select all

  psig->suppressDC( true, 0 );
  if ( specParam->doTaper() ) {
    double winLen = tw.lengthSeconds();
    psig->taper( 0, winLen, specParam->taperWidthRatio() * winLen );
  }
  QtbDoubleSignal * spec = new QtbDoubleSignal( 0 );
  psig->fft( QtbDoubleSignal::Spectrum );
  spec->copyAmplitudeFrom( psig );
  delete psig;
  spec->smooth( specParam->smoothing, specParam->sampling.min(), specParam->sampling.max() );
  psig = spec;

For FK analysis, the signals for each time window are processed this way:

Code: Select all

  _signals[0].processed->suppressDC(true,0);
  // Set a 10% taper in time domain
  double winLen=_signals[0].processed->duration();
  _signals[0].processed->taper(0,winLen,0.1*winLen);
  // Switch to frequency domain
  _signals[0].processed->fft(QtbDoubleSignal::Spectrum);

[sbonnefo]

So, this mean that it is not mandatory to remove the mean before H/V and array processing?

[admin]

For the processing signals without any particular pre-selection of time windows, it is of course useless.

If you want to reject time windows with STA/LTA criteria, it is the same, the average value of the global signal has no influence.

So in any case, doing such processing is useless. If your data suffer from strong baseline drift, correct first.