1994A&A...292..175B

We have studied the fast timing behaviour of Cygnus X-3 using the entire EXOSAT ME dataset on this source, consisting of 22 observations that cover a total of 320000seconds. This large amount of data allowed us to measure the rapid (>1Hz) X-ray variability of the source to an unprecedented accuracy of 0.2% fractional rms amplitude. Above 256Hz, the power of the X-ray count rate fluctuations is significantly below the level predicted from Poisson statistics modified by the known dead time processes in the instrument. We also saw this in other X-ray sources, as well as in a background observation. We developed an improved empirical method for predicting the Poisson level in EXOSAT ME HTR3 and HTR5 data, and use it in our present analysis. A weak but formally very significant high-frequency noise (HFN) component was found in the overall average of all power spectra of Cygnus X-3, with an rms amplitude of 3.3±0.2% that appears to be independent of binary phase, and a cut-off frequency of 93±28Hz. Four other X-ray sources, including Crab, were studied to check this result. None of them had a weaker noise-component in the region 1-256 Hz. An instrumental effect could therefore not be excluded. If the effect is instrumental, the 99% confidence upper limit on the 1-256 Hz rapid variability of Cygnus X-3 is 0.6% rms. The consequences of these results for previously reported EXOSAT HTR observations are briefly discussed. We compare our results to the predictions of the stellar wind model for Cygnus X-3. Monte Carlo simulations were carried out to investigate the signal-attenuating effect of the wind. For the most likely wind parameters, the intrinsic source variability is found to be either very strong (∼60% rms), unlike seen in any other X-ray source at similar luminosity, or to be ≲12%, which would be consistent with a black hole candidate in the high state, or a low magnetic field neutron star.