It is possible to rebin data whose wavelength/channel number relation is known in such a way that the relation is linear. Such an operation is known as `scrunching', and is performed by the `scrunch' command. Scrunched data is generally easier to handle than is un-scrunched. For example, if spectra with slightly differing wavelength scales must be added together they should be scrunched to a common wavelength scale first.
Similarly, there are cases where one needs data rebinned on a logarithmic wavelength scale-cross-correlations to determine redshift need this sort of scale, for example-and this can also be performed using `scrunch'.
`scrunch' has parameters that specify the start and end wavelengths for the resulting scrunched spectrum, and the number of elements (bins) it should have. Note that the wavelengths are those of the `centres' of the end bins; it is easy to miscalculate the number of bins by 1 when aiming to get a linear dispersion of exactly so many Angstroms per bin. The target wavelength range may also be specified in terms of the start wavelength and the increment, by means of the rather crude convention that if the end wavelength specified is less than the starting wavelength, then that `end wavelength' value will be taken as the wavelength increment. (This behaviour may be controlled explicitly by two keywords `final' and `increment', which are used to specify the interpretation to be placed on the `end wavelength' value.)
`scrunch' has one keyword that can confuse users. If the spectrum being scrunched is in flux units, then the total flux in the spectrum should be conserved by the scrunching process. Note that the distinction is between flux units and flux density units, so `raw counts' are a flux unit. If a spectrum is in flux density units, then scrunching should not change the absolute value of the data. If the scrunching is simply such that it doubles the number of bins in the spectrum, then an element that has a value of, say, 10 mJy should become two elements each with a value of 10 mJy, since Janskys are a flux density unit. Conversely, if the spectrum were still in raw counts, then an element with 10 counts should become two elements each with 5 counts. The default for the `flux' keyword in `scrunch'-whose prompt is the perhaps confusing `conserve flux (as opposed to mean counts)?' is `yes', and this is correct for raw counts. To confuse matters further, there is a concurrent parameter `mean' with the opposite meaning to `flux'.
FIGARO A general data reduction system