Observations of the three galaxies reported in this paper
were carried out in the photometric night of Feb. 25, 1994,
with the Calar Alto 
telescope equipped with
the MAGIC NICMOS3, 2562 pixels infrared array
(Herbst et al. 1993).
The optical set-up of the detector was chosen to give
the largest possible field of view (
),
with a pixel size of 1.61 arcsec.
The expected deviation from linearity of the detector
is 
0.5% at the count rate of the present observations.
Table 1 (click here) lists the relevant parameters of the galaxies
observed in this work. The table is arranged as follows:
Column 1: NGC denominations;
Columns 2, 3: adopted (1950) celestial coordinates;
Columns 4, 5: major and minor diameters as given in the RC2
(de Vaucouleurs et al. 1976);
Column 6: morphological type.
Obtaining a satisfactory background subtraction is the main
difficulty of IR astronomy. For the observations reported here
the sky brightness was typically 13.8 and 13.1
mag arcsec-2 at H and 
respectively, varying by
several percent over the duration each source was observed.
A careful monitoring of the sky is thus necessary,
requiring the use of mosaicing techniques in which almost
an equal amount of time is devoted to the target of interest
and to the surrounding sky.
Mosaic maps were obtained by programming the telescope pointing
along a pattern selected ad-hoc for each galaxy according to
its shape and size (see Fig. 1 (click here)).
NGC 2366 (whose angular size is comparable with the MAGIC field
of view) was observed alternating 8 fields centered on the target
(displaced one-another by 10 arcsec) with 4 observations
of the sky taken 7 arcmin to the N and S of the galaxy.
Each pointing was observed for 42 s in H (40 s in
) by adding 14 (20) elementary integrations
of 3 (2) s.
The whole mosaic pattern was repeated twice giving a total
of 672 (640) s integration on source and 336 (320) on the sky.
Figure 1: Mosaic patterns for NGC 2366 a), NGC 2403 b), NGC 4236
c)
The galaxy NGC 2403, which has an angular extent
of 
, was mosaiced in three overlapping
on-source fields taken at equal declination and displaced by
in right ascension.
Each on-source field was followed by sky measurement taken
at equal right ascension, displaced by 10 arcmin from the galaxy.
The whole mosaic pattern was repeated twice
in the band, but the W subframes were not combined
with the central and E ones because affected by residual
large-scale inhomogeneities (see Fig. 2 (click here)). The total integration
time was 126 s in H (240 s in ) on source
(and an equal amount on the sky).
Similarly, the galaxy NGC 4236 (angular extent of
) was mosaiced in three overlapping
on-source fields taken along the galaxy major axis, dithered
by 2.5 arcmin.
Each on-source field was followed by sky measurement taken
at equal declination, on each side of the galaxy.
We integrated on the two peripheral pointings twice as long as
on the central region.
Since this pattern was repeated, the total integration time
was 420 s in H (400 s in ) on source
(and 4/5 of that on the sky).
The observations were calibrated and transformed
into the H photometric system and to
using the three standard stars HD 40335, HD 84800 and HD 106965 in the
catalogue of Elias et al. (1982), observed hourly along the
night.
The calibration stars were observed with a pointing sequence
which consisted of five positions, starting with the star
near the center of the array, followed by positioning the star
in each of the four array quadrants.
The telescope was defocused to avoid saturation.
Only HD 40335 has a reference mag
(Wainscoat & Cowie 1992). For HD 84800 and HD 106965,
having almost null spectral slopes (H-K=0.00 and 0.02 mag
respectively), we assumed 
.
The typical uncertainty of the photometric zero points is 0.03 mag. We also assessed the accuracy of our flat field correction by determining the differences of the star brightness at the five positions of the observing pattern. No significant differences were devised within 0.022 mag.
The reduction of two-dimensional IR frames follows
a procedure based on the IRAF data reduction package developed
by NOAO and on the SAOIMAGE and PROS packages developed
at the Center for Astrophysics.
To remove the detector response each image is processed as follows.
Two sets of flat-field exposures were obtained on the telescope
dome with (lamp-on) and without (lamp-off) illumination with
a quartz lamp.
The response of the detector alone is contained in the normalized
frame 
.
For each galaxy all the sky exposures (SKYi) are combined using
a median filter to obtain the image 
if the deviation among their mean values
is smaller than 5% of their average.
Otherwise, three contiguous sky measures are used.
This allows removal of bad pixels and of unwanted star images.
The mean counts 
and
are determined,
rispectively for the 
target observations
and the median sky.
Individual "normalized'' 
frames
are produced such that 
. This allows removal of the time
variations of the sky level (which are usually of the order of 5%
during a full mosaic measurement), but introduces an (additive)
off-set to the background which is subsequently removed.
Each target frame (Ti) is then processed to obtain a
flat-field, sky subtracted, corrected frame: 
The individual 
target frames are then registered
using field stars and combined together with a median filter.
This allows removal of the bad pixels in the final combined image.
Star-subtracted frames are produced by manual "editing''
of the contribution from pointlike sources which are clearly
not associated with the target galaxies, being most probably stars
of our Galaxy.
Extended features, possibly associated with HII regions
belonging to the target galaxies are not subtracted.
Figure 2 (click here) gives a grey-scale/contour representation
of the three galaxies under study, smoothed with a 0.8 pixel gaussian.
For each galaxy we give the H and mosaics.
The faintest levels shown are 21.5 (H) and 21.0 ()
mag arcsec-2 for NGC 2366 (Fig. 2 (click here)a) and NGC 4236 (Fig. 2 (click here)c) and
20.0 (H and ) mag arcsec-2 for NGC 2403 (Fig. 2 (click here)b), with a
step of 0.5 mag arcsec-2. These limits are consistent with the rms of each image
(see further here).
The orientation of the frames is such that North is up
and East to the left.
Figure 2: Grey-scale rapresentations and contour levels of h and K'
band images of NGC 2366 a), NGC 2403 b), NGC 4236 c)
Special care was taken in determining the image properties.
This analysis is summarized in Table 2 (click here).
For each galaxy (and band-pass) we give the total on target
integration time, followed by the estimate of the seeing,
by the sky brightness and by the background noise
on a scale of 
pixel, i.e. not much smaller
than the scale of the measured galaxies.
This figure, varying from 0.03% to 0.16% of the sky,
represents the dominant source of error in low S/N regions.
The large-scale noise carachteristics
and those determined on a much smaller angular scale
(
pixel) are in agreement.
