Calculations have been carried out to simulate the effect that data ov
erflows have on X-ray timing observations when the data values are bin
ned in equal time intervals, as is anticipated for some bright X-ray s
ources in the USA experiment (Wood et al. 1993). Simulated data sets h
ave been generated with Poisson noise and two 5% amplitude cosine wave
s which were then subject to standard Fast Fourier Transform (FFT) ana
lysis. This was done for a series of average count rates and 1-bit, 2-
bit, and 4-bit ''data'' values, assuming that any overflows would rese
t the counter to zero. We find that overflows produce a significantly
worse loss of sensitivity for 1-bit data recording than for P-bit data
recording. In both cases a factor of 10 power response loss occurs wh
en the dimensionless parameter x = S-0 Delta t/(2(n) - 1) (where S-0 i
s the mean source count rate, Delta t is the time sampling interval, a
nd n is the number of data bits available) is 0.65, but for n = 2 the
loss of response is more gradual up until x approximate to 0.5. With P
oisson noise the effect of overflows can be seen from the mean value o
f the power spectrum if the normalization of Leahy et al. (1983) is us
ed.