Hard X-ray lags of Cygnus X-1 on short timescales

Calculating the time lags over different timescales using the cross-correlation technique may lead to a biased estimate of small timescales. Given a timescale for lightcurve binning, we propose to subtract the local average instead of the global average during the cross-correlation, in order to filter variations on timescales larger than the bin size. The new method allows us to make an unbiased estimate of the time lags using RXTE/PCA data on timescales as small as ∼5 ms, where the Fourier technique becomes invalid. We calculate the time lag spectra of Cygnus X-1 at different spectral states with the new method, and find that the source appears to have similar X-ray lags on small timescales independent of its spectral states.