next up previous contents
Next: Reduction of Long Slit Up: Introduction Previous: A typical reduction.

Notes on specific commands

- - SET/IRSPEC flat=....
The InSb array used in IRSPEC is not linear and the relative response of its pixels (i.e. the flat-field) varies slightly with the level of the signal (i.e. the count-level) of the image. The flat used should therefore have a count-level close to the astronomical frames, better if within a factor <2. There is also a small dependence on wavelength which makes it advisable to have at least one flat per photometric band (J,H,K,L). Up to K ( $\lambda<2.5$ $\mu$m) the flats can be derived from short exposures of the halogen lamp; at longer wavelengths one can directly use the sky frames.

- - SET/IRSPEC dark=....
The dark frame is used in
     CALIBRATE/IRS ... mode=d
and must be taken with the same on-chip integration time (DIT) as the astronomical frame(s). However, in SKYSUB/IRS you can avoid caring about the dark if - and only if - you use this command to make a straight subtraction (see last example of TUTORIAL/SKYSUB), in which case the dark frame is effectively not used. In case you do not have darks at a given DIT you may try to derive them by scaling darks taken with different (longer) DITs.

When you have more than one pair of object and sky frames it is not trivial to decide whether it is better to apply SKYSUB to each pair and then average them or first average the objects and the skies together and then apply SKYSUB (N.B. you must average, not sum the frames otherwise you will get into trouble with the flux calibration). The best approach is probably to first use the second method (SKYSUB on the averaged frames) and try the second if you have obvious problems with the cancellation of the sky lines.

The wavelength calibration of IRSPEC data is much simplified by the fact that the pixel-wavelength dispersion is linear to very high accuracy, and by the very precise mechanical positioning of the grating. The on-line (mechanical) wavelength calibration that you retrieve using
      CALIB/IRS image
is already quite accurate (usually within one pixel) and may be enough for many purposes. It can still be improved using by manually determining with       CALIB/IRS ima_ref mode=d
the shift (error) on an image containing one or more reference lines (e.g. the OH airglow lines in a sky frame); with this you can easily achieve accuracies of 1/3 of a pixel. When using this mode remember to apply - using the above command - the same calibration to the object and to the standard otherwise you will not be able to run FLUX/IRSPEC. Small shifts between the object and standard frames can be corrected within FLUX/IRSPEC using the    shift=...    option.

The analytical formula used to derive the tilt angle at a given central wavelength is quite accurate and should produce rectified spectra within a fraction of a pixel. The errors introduced by assuming that the slit images are straight and parallel to each other are also small.

I decided to force the users to create `flux' files themselves containing their best guesses for the flux of the `standard' stars because the use of the photometric points is not necessarily accurate, even though there is little more one could do. I hope that a list of spectrophotometric - or even just spectroscopic - standard stars for the infrared will be soon available. In the meantime, I apologize to the users for the inconvenience which, however, should make them aware of the problems one encounters to find a star with a precisely known flux distribution and with a featureless spectrum.

These procedures are meant to give you a complete list of examples for the use of these relatively complicated commands. To work they need a number of small files (called irstut....) in the `DEMO' area (.../midas/demo/data), contact your local wizard to have them installed.

=01 =11 =1994

next up previous contents
Next: Reduction of Long Slit Up: Introduction Previous: A typical reduction.
Petra Nass