Is it necessary to do an "off" for every "on" in a raster map? Answer: "No."
Is it necessary to do an "off" for the same amount of time as the "on?" In general, no, but read on.
Observers may add an off position to their AOT list if one is not there and need not get permission from the ESA help desk, but should send the email@example.com email a line saying they are doing this.
The observer pays for the time. Note that the off may be concatenated to the other observations saving the 180 sec acquisition.
Proposed strategies to save time in the off include doing a maximum separation 2-point raster, with one point being the on and one being the off. (Thus avoiding instrumental overheads of approximately 3 minutes. This is not recommended since the maximum raster separation (180") is too close to the beam size (100" + difraction effects) to assure that the off will actually completely avoid the source.
Proposed strategies to save time in the off include observing fewer lines or a smaller wavelength range there. The idea behind this is if the observer does one line in the off, the observer still gets the other 9 detector outputs, and thus a pretty good idea of the contimuum spectrum of the off position.
It usually doen't make sense to go any deeper in the reference in terms of sensitivity than the deepest sensitivity in the source.
Redundant to the above, a map may require no more than one reference position.
In the case where there is cirrus, or any other suspected background emission, off positions are definitely suggested... especially if 1) the source continuum is expected to be weak and is of interest, 2) there might be features or lines in the cirrus or ISM near wavelengths of interest, e.g. a 60 um ice absorption feature in a molecular cloud region or C+ from a galactic cloud.
Roger notes that the cirrus thermal component might be expected to present a spectral shape which falls monotonically with lambda across the full wavelength span of the LWS, while astronomical sources, being warmer, would peak at mid-range. For weak continuum sources, unscrambling the two spectra would be impossible without good cirrus measurements off source.
[Aside - running IRSKY ("telnet irsky.ipac caltech.edu 1040") is a good way to look at the cirrus environment of the source.]
Several ways to pin down the cirrus or background are possible, such as over-mapping the field so that one "runs off the source" and thus measures the peripheral regions. In the situation of variable cirrus near the source, a strategy of four positions around the source would be one approach, although it is a high price to pay. Alternatively, one could use IRSKY to identify a cirrus region off source that seems of equal magnitude to the cirrus in the source direction, and measure an off position there.
When I asked him about sources with strong lines and continua, Roger suggested that if the continuum of the source were known to dominate the possible background, by say ten times or more, one might not need to measure the background. He cautioned that if the source lines were weak, unknown features in a weak background could be a problem. But otherwise, one might be content simply to draw a baseline through the continuum points to isolate the line flux and ignore the continuum. This method would even work in the case of strong lines against weak continuum, if the continuum were not of interest. He cautioned that the continuum could _later become of interest_, and so one might do an off continuum just in case.
I asked Roger, in the absence of cirrus considerations, if an off source measurement were really necessary for any part of the LWS calibration proceedure. He said, in essence, not formally. That the only two other ways an off source measurement might help (that he could think of) was with any stray-light problem, or detector variability problems. He said the idea that an off could help with stray light was somewhat unlikely however, because the variability of stray light would have to be on time scale less than the once-a-week calibration of the instrument but on a greater scale than the source+off time. He said _anything_ is possible, that there may be unexpected sources of emission from outside the telescope, from within or from the instrument itself. Therefore an off position is not a bad idea, especially for sensitive measurements.
Detector responsivitity changes are supposed to be taken care of by the frequent illuminator flashes - and these are designed specifically to be sufficiently frequent to take care of the variablility. The only way an observer could check this themselves would be to do their own off position on a known calibrator source. This would be an extrodinarily cautious approach, only only justified by the need for much better than 15% line flux accuracy for the science. (Most of the science involves line ratios, where absolute errors may cancel.)
As for the question of the manual's draconian prescription that _each_ map on position deserves an off and emulated in the formula time ~ 2 x M x N, Roger said that this was perhaps overcaution in the manual writing. One or more off positions of sufficient sensitivity to guard against background (cirrus) confusion would probably be sufficient for a map.
If questions - contact firstname.lastname@example.org