Difference between PSD from aLIGOZeroDetHighPower and real data

Dear colleagues,

The attached figure is a comparison of PSDs. The blue one was generated from aLIGOZeroDetHighPower in pycbc.psd. The orange one is the averaged energy spectrum density (np.abs(rfft(data))**2) from multiple raw data fetched by TimeSeries.fetch_open_data() in gwpy.


The peak of aLIGOZeroDetHighPower PSD happens near 0, while the peak of real data PSD ranges around 6 to 12. Why doesn’t aLIGOZeroDetHighPower PSD reflect real situation?

@yidelee Thank you for the question!

In general, model PSDs are only models, and they never capture all the features of the real noise curves.

The peak in the PSD of the real data is due to high-pass filtering. In reality, LIGO noise becomes very loud at low frequencies, both due to seismic motion and controls noise. However, in order to keep the control loops within a manageable range, high-pass filters are applied to remove signals below 8 Hz. This creates the peak at 8 Hz.

This feature of the real detector is not captured in the model, as you correctly noticed. In practice, most analyses only use frequencies above 20 Hz, because very low frequencies are dominated by noise.

@jonah Thank you very much for your reply. The power of real PSD is larger than the model PSD. Is it also a calibration or a filtering effect?