In ``On-The-Fly'' (OTF) mode, the telescope continuously scans along a row or column with a given velocity instead of tracking the individual raster points. Spectra are written every couple of arc-seconds which corresponds to typical integration times of a few seconds or less (in the future (new AOS control) down to 100 milliseconds). These portions of data are called ``OTF dumps''.
OTF is more efficient because the telescope system is not initialized for each point along the scanning axis. For the above map it would save 40 subscans or 400 seconds and the efficiency would go up to about 91% ! In addition, one can also reduce the time spent on the ``Off'' position since OTF is essentially a ``several On's per Off'' mode. The ``Off'' time needs to be only where is the time per OTF dump, and n is the number of Nyquist-spaced positions along the scanning axis. Since large regions are mapped rapidly, weather effects are minimized. Thus the final result is more reliable.
Therefore, it is recommended to use OTF whenever mapping a region larger than a few beams in size. There should not be a limit concerning the line temperatures, although weaker lines require several maps that have to be co-added to reach a sufficiently high signal-to-noise ratio. OTF mapping has the same limits concerning baseline quality as the traditional single-point or raster ``On-Off'' observations. In the future, we plan to introduce a Wobbler On-The-Fly mode (WMAP) as well as a Frequency-Switching On-The-Fly mode (FMAP; already working at the Effelsberg 100m telescope).