We have studied the atoll source 4U 1608-52 using a large data set obtained with the Rossi X-ray Timing Explorer. We find that the timing properties of 4U 1608-52 are almost exactly identical to those of the atoll sources 4U 0614+09 and 4U 1728-34 despite the fact that contrary to these sources 4U 1608-52 is a transient covering two orders of magnitude in luminosity. The frequencies of the variability components of these three sources follow a universal scheme when plotted versus the frequency of the upper kilohertz QPO, suggesting a very similar accretion flow configuration. If we plot the Z sources on this scheme only the lower kilohertz QPO and HBO follow identical relations. Using the mutual relations between the frequencies of the variability components we tested several models; the transition layer model, the sonic point beat frequency model, and the relativistic precession model. None of these models described the data satisfactory. Recently, it has been suggested that the atoll sources (among them 4U 1608-52) trace out similar three-branch patterns as the Z sources in the color-color diagram. We have studied the relation between the power spectral properties and the position of 4U 1608-52 in the color-color diagram and conclude that the timing behavior is not consistent with the idea that 4U 1608-52 traces out a three-branched Z shape in the color-color diagram along which the timing properties vary gradually, as Z sources do.

We present the results of a multi-Lorentzian fit to the power spectra of two kilohertz QPO sources; 4U 0614+09 and 4U 1728-34. This work was triggered by recent results of a similar fit to the black-hole candidates (BHCs) GX 339-4 and Cyg X-1 by Nowak in 2000. We find that one to six Lorentzians are needed to fit the power spectra of our two sources. The use of exactly the same fit function reveals that the timing behaviour of 4U 0614+09 and 4U 1728-34 is almost identical at luminosities which are about a factor 5 different. As the characteristic frequency of the Lorentzians we use the frequency, ν max , at which each component contributes most of its variance per log frequency as proposed by Belloni, Psaltis & van der Klis in 2001. When using ν max instead of the centroid frequency of the Lorentzian, the recently discovered hectohertz Lorentzian is practically constant in frequency. We use our results to test the suggestions by, respectively, Psaltis Belloni and van der Klis in 1999 and Nowak in 2000 that the two Lorentzians describing the high-frequency end of the broad-band noise in BHCs in the low state can be identified with the kilohertz QPOs in the neutron star low mass X-ray binaries. The prediction for the neutron star sources is that if the two kilohertz QPOs are present, then these two highfrequency Lorentzians should be absent from the broad-band noise. We find, that when the two kilohertz QPOs are clearly present, the low-frequency part of the power spectrum is too complicated to draw immediate conclusions from the nature of the components detected in any one power spectrum. However, the relations we observe between the characteristic frequencies of the kilohertz QPOs and the band-limited noise, when compared to the corresponding relations in BHCs, hint towards the identification of the second-highest frequency Lorentzian in the BHCs with the lower kilohertz QPO. They do not confirm the identification of the highest-frequency Lorentzian with the upper kilohertz QPO.

We have studied the aperiodic X-ray timing and color behavior of the accreting millisecond pulsars SAX J1808.4À3658, XTE J1751À305, XTE J0929À314, and XTE J1814À338 using large data sets obtained with the Rossi X-Ray Timing Explorer. We find that the accreting millisecond pulsars have timing properties very similar to atoll sources and low-luminosity bursters. On the basis of the correlation of timing and color behavior, SAX J1808.4À3658 can be classified as an atoll source, and XTE J0929À314 and XTE J1814À338 are consistent with atoll sources, but the color behavior of XTE J1751À305 is different. Unlike the color in other atoll sources, the hard color in the extreme island state of XTE J1751À305 is strongly correlated with soft color and intensity and is not anticorrelated with any of the characteristic frequencies. We found previously that the frequencies of the variability components of atoll sources follow a universal scheme of correlations. The frequency correlations of the accreting millisecond pulsars are similar, but in the cases of SAX J1808.4À3658 and XTE J0929À314 they are shifted compared with those of the other atoll sources in a way that is most easily described as a shift in upper and lower kilohertz quasi-periodic oscillation (QPO) frequency by a factor close to 1.5. However, we note that the shift in the lower kilohertz QPO frequency is based on only one observation for SAX J1808.4À3658. XTE J1751À305 and XTE J1814À338, as well as the low-luminosity bursters, show no or possibly only small shifts.

We study the correlations between timing and X-ray spectral properties in the low mass X-ray binary 4U 0614+09 using a large (265-ks) data set obtained with the Rossi X-ray Timing Explorer. We find strong quasi-periodic oscillations (QPOs) of the X-ray flux, like the kilohertz QPOs in many other X-ray binaries with accreting neutron stars, with frequencies ranging from 1329 Hz down to 418 Hz and, perhaps, as low as 153 Hz. We report the highest frequency QPO yet from any low mass X-ray binary at 1329 ± 4 Hz, which has implications for neutron star structure. This QPO has a 3.5 σ single-trial significance, for an estimated 40 trials the significance is 2.4 σ. Besides the kilohertz QPOs, the Fourier power spectra show four additional components: high frequency noise (HFN), described by a broken power-law with a break frequency between 0.7 and 45 Hz, very low frequency noise (VLFN), which is fitted as a power-law below 1 Hz, and two broad Lorentzians with centroid frequencies varying from 6 to 38 Hz and 97 to 158 Hz, respectively. We find strong correlations between the frequencies of the kilohertz QPOs, the frequency of the 6−38 Hz broad Lorentzian, the break frequency of the HFN, the strength of both the HFN and the VLFN and the position of the source in the hard X-ray color vs. intensity diagram. The 1 frequency of the 97−158 Hz Lorentzian does not correlate with these parameters. We also find that the relation between power density and break frequency of the HFN is similar to that established for black hole candidates in the low state. We suggest that the changing mass accretion rate is responsible for the correlated changes in all these parameters.