Fig. 1: XTE 2--10 keV light curve of NGC 3516. The source was sampled once
every 12.8 hr from March through July 1997 (top panel) and
quasi-continuously (except for SAA passage and interruptions for other
time-critical observations) for 4 days in the middle (bottom panel, binned
in 700 sec intervals).  The inner error bars are the statistical errors,
while the outer ones also include a 0.3 ct/s systematic uncertainty
added in quadrature due to the background.  The peak-to-peak variation was
a factor of $\sim$4, and real variability is visible down to the shortest
time scales sampled (1.5 ksec).

Fig. 2: PDS of NGC 3516, derived from the data in Fig. 1 by combining the
PDS measured from the quasi-continuous 4 day monitoring (dots on the
right) with that from the overlapping monitoring every 12.8 hr (triangles
on the left).  A fit to the short-term data shows a slope of $ a = -1.80 $
($P(f) \propto f^a $, where $P(f)$ is the fluctuation power at temporal
freqency $f$) while the long-term data has $ a = -1.05 $.  This break of $
\Delta s \approx 0.75 $ is a real flattening to longer time scales, so the
combined PDS was fitted with the function $ P(f) \propto { 1 \over {{ 1 +
f/f_t }^a} $ to derive a turnover frequency of $ f_t \approx 3 \times
10^{-7} $ Hz ($\sim$1 month; arrow in the figure) and $ a = -1.70 $. 
Although the turnover is real, its exact value is not well defined because
it lies so close to the longest time scale sampled.