Unlike other radio sources which are taken to be statistically constant in their
strength as a function of time, pulsar signals are intrinsically
periodic signals.
The pulses have periods ranging from a few seconds for the slowest pulsars to
about a millisecond for the fastest pulsars known. Further, the pulses have a
very small duty cycle, with typical pulse widths of the order of 
of
the period. Thus typical pulse widths range from a few tens of milliseconds down
to a fraction of a millisecond. Study of such pulsar signals clearly requires
the final data to have time resolutions ranging from 
milliseconds to
microseconds.
Pulsar observations thus require very fast sampling times for the data. This leads
to a substantial increase in the speed (and therefore complexity) of the back-end
designed for pulsar observations and also in the speed of the data acquisition
system and off-line computing capabilities.
Also, the value of the sampling interval needs to be known quite accurately
in order to preserve the pulse phase coherence over a long stretch of pulsar
data spanning many periods.
The other property of the time variation of pulsar signals is that the rotation rate of pulsars is very accurate. This means that if the time of arrival of the Nth and (N+1)th pulses is known, the arrival time for the (N+M)th pulse can be predicted very accurately. Further, slow variations of the pulsar period (for example due to rotational slow down of the pulsar) can be studied if the absolute time of arrival of the pulses can be measured sufficiently accurately. This requires the availability of a very precise clock at the observatory, such as that provided by a GPS receiver (see section 17.7 for more details).