Infrared prism objective available at TNG

The infrared multi-mode instrument of TNG, NICS, can now perform low resolution spectroscopic observations using the Amici prism in slit-less mode. The frame format is similar to classical prism-objective images, an example is given in the figure below.

Characteristics and performances

Data acquisition

The instrument setup corresponds to the low-resolution mode with the Amici, slit-less and SW attenuator. The array should be set to detector integration time DIT=30sec and NDIT=3. To correct for the spectral overlapping of nearby objects it is convenient taking 2 series of prism-objective frames at position angles PA=0 and PA=90 degrees. An H image should also be taken to better recognize and position the objects in the field.

Data reduction

The 2D frames can be sky-subtracted and aligned as normal images using e.g. the snap reduction package. The extracted spectra can be wavelength calibrated using the Amici look-up table described in the NICS documentation and shifting the pixel vector by, to a first approximation, (Xcen-530); Xcen being the X-position of the object in the non-dispersed image. The calibration can be then refined by matching the 1.75 micron cutoff and the deep atmospheric absorption band at 1.37 microns. A first-order flux calibration can be obtained using the above zero-points which, given the good instrumental stability, are not expected to vary by more than +/-0.1 mag even on long time scales. More accurate flux calibration can be obtained by separately measuring standard stars. The atmospheric absorption features can be corrected using spectra of other stars in the same field/image.

NICS - Detector

The detector is a Rockwell 1024x1024 HgCdTe Hawaii array which, alike other such devices used in various astronomical instruments, has some peculiar characteristics which should be kept in mind when collecting and reducing the data.

Persistency and memory effects (changed on Feb06)

The array multiplexer has a stubborn attitude to remember whatever strong signal was recorded in the previous frame(s). For example, the first dark exposure taken after a saturated image will still show a ghost relic of the previous frame at a level of up to 0.4% of its original intensity regardless of the time elapsed betweeen the two frames. This ghost image will also persist in the subsequent frames with intensity slowly fading down to below the noise level (see figure on the left).
Persistency was stronger and very annoying with the old acquisition electronics, also because the system remained idle when the integration was terminated. The new FASTI-NICS control electronics, on the contrary, continuosly take short dummy frames during the idle time; thus cleaning the array.
The "clear-array" procedure, which was formely mandatory when switching from imaging to spectroscopy, should be now used only in very estreme cases, e.g. when starting a very long spectroscopic integration after having centered the target using images of a field with one or more completely saturated objects.

In the image below evolution of the level of persistency (ghost) signal after a strongly saturated frame. Please note that the level of persistency has decreased be about a factor of 3 after the new acquisition electronics (FASTI-NICS) has been put into operation in Feb 2006.

Bias

Contrary to standard CCD devices, the "bias" of this detector is not uniform over the array, but has a distinct horizontal pattern with pronounced maxima at rows Y=1 and Y=513. Moreover its level is not constant along a row but increases with X (see for example the figure here).
The amplitude of the peaks and their slopes are not constant (sic!) but strongly vary with the level of the signal. In dark exposures, the pattern is characterized by prominent maxima with quite gentle slopes (like in this figure), while at higher levels of illumination the peaks become progressively narrower.
This effect could produce annoying features when subtracting images/spectra taken under variable sky conditions. Moreover, the variation of the bias makes it basically impossible to obtain accurate differential flats. Luckily, however, these are not necessary in case of NICS (see the section describing the procedure to get suitable flats).

A typical 60s dark frame with cuts along the two axes.

Efficiency versus wavelength

NICS - Electronics

The array is controlled and read-out by the "FASTI-NICS" electronics system developed by the Infrared Group of the INAF-Arcetri Observatory of Florence. Described below are only the few aspects which could be of interest for a normal observing run, more detailed information can be found in the dedicated web site.

Noise performances and integration times (changed on February 2006)

The read-out noise of the system is about 24 electrons or, equivalently, 3 ADU counts on a single frame; The conversion factor is close to 8 electrons/ADU. To make sure that imaging observations are background limited and to limit the annoying effects related to the bias variations it is convenient using detector on-chip integration times (DIT) long enough to obtain a sky level of at least 1500 counts. However, be careful not to exceed 15,000 ADUs to avoid non-linearity, note also that the detector saturates at about 20,000 ADU counts.
For observations in conditions of low background, e.g. spectroscopy or narrow band imaging, the detector integration time can be increased up to about 15 minutes. Longer integrations are not recommended.
Short integration times should be used only to observe very bright sources. The minimum on-chip integration time is 3 seconds. If this is not short enough one can use the grey filters to attenuate the signal by up to 10 magnitudes.
The table here summarizes the values of detector integration times which should guarantee good performances for observations of faint objects. Entries marked with a "*" indicate that the maximum value of DIT is not sufficient to guarantee background-limited performances, i.e. that the data taken with this observing mode will be most likely read-out noise limited.

Please note that imaging with the small field camera (SF) should require integration times a factor of 4 longer than those with the large field camera (LF).
In spectroscopy, the data taken with low and medium resolution grisms are read-out noise limited at all the wavelengths where the sky emission is low. The only mode which achieves background limited performances at most wavelengths is low-R spectroscopy with the Amici prism. However, this requires a careful tuning of the value of DIT to obtain a reasonably high signal in the blue without saturating the red part of the spectrum. The value given in the table should guarantee good performances in the blue with saturation starting at about 2.4 microns, click here to see a representative sky spectrum taken with the Amici prism. Other examples of sky spectra taken with various grisms can be found in the section dedicated to wavelength calibration and sky spectra.

Overheads (changed on February 2006)

The table here summarizes the main overheads which noticeably, are now much lower than before the refurbishing of the control electronics of Feb 2006. The total time spent for an integration with a given on-chip integration (DIT), an averaging NDIT imaging before writing them on disk, repeated in n positions (Npos), is given by:

     Ttot=((DIT+2.5) * NDIT + 7 sec + 15 sec) * Npos

The time necessary for moving the telescope (15 sec) is virtually independent of the offset amplitude.

Cross talking (changed on September 2006)

Due to a spurious coupling between the electronic channels which simultaneously read the 4 sections of the array, whatever signal detected in a given quadrant produces negative ghost images in the other 3 quadrants.
Apart for the large improvement which occured after the intrument refurbishing of Feb-Apr 2003, the cross-talking is very stable and can be partially corrected by applying a suitable algorithm in the pre-reduction phase.
To this purpose you can use one of the following Fortran programs which eliminate most of the cross-talking effects produced by non-saturated images.
WARNING: the algorithm does not completely correct the cross talking of saturated images but leaves positive ghosts.

The program takes the original fits file of NICS and produce another fits file preserving all the original headers and keywords. Instructions for compiling and using the program are inserted as first comment lines in the files.

You may download here:

If you have different systems try using both, the wrong one will produce quasi-random results.