The duration of an observation is a critical parameter for radio astronomy observations and adequate observation length is crucial for a successful completion of an observing project. The characteristics of an observation depend on the instrument used and the objective of the observing project.
The duration of an observation is often understood as integration time. This is true when during this period of time the observing conditions do not change. For e.g. an Earth rotation synthesis instrument like the Westerbork telescope in the Netherlands, these concepts are not the same because during the observation the projection of the instrument rotates with respect to the sky. In this case, the integration time and receiver sampling time are considered to be the same.
For the time duration of an observation, data are collected either integrated over the whole obervation time or over much shorter (basic) integration/sampling times (of a duration dependent on the receiver characteristics), e.g. 10 seconds, which are later combined for the whole observation duration. The duration of an observation is also related to the kind of instrument used: for a single dish telescope for some project 2000 seconds may be sufficient while for an Earth rotation aperture synthesis instrument, such a duration must be at least 12 hours.
The duration of an observation is determined by the nature of the celestial radio source as well. Usually such a source is variable. Such a variation may have time scales ranging from milliseconds (pulsars) to years (some stars and quasars), i.e. a very wide range of values occurs. For practical reasons, it can be said that continuum observations made with telescopes operating singly are reasonably well represented by the integration time of 2000 seconds (as used in e.g. ITU-R Recommendation RA.769). On the other hand, 2000 seconds is less representative for spectral line observations.