Mrk 279 was observed with two of the instruments on board the ISO satellite: nine narrow-band filter images were obtained with ISOCAM (Cesarsky et al. 1996), while spectra were recorded with the PHT-S spectrometer, a subsystem of the ISOPHOT instrument (Lemke et al. 1996). The PHT-S spectrometer covers the 2.5-12 $\mu$ m spectral range at a mean resolution of $\sim$ 3150kms^-1, with a gap between 4.9 and 5.9 $\mu$ m. Its entrance aperture projects to $24\hbox{$^{\prime\prime}$ }\times 24\hbox{$^{\prime\prime}$ }$ on the sky. All PHT-S observations were carried-out in an identical fashion. The integrating amplifiers were reset every 32s and on-source measurements were interleaved with background measurements by ``chopping'' on the sky at a frequency of 1/256 Hz. The chopper throw was set to 300''. For each of the observations, total on-source integration time was 2048 s and total observing time (including background measurements and instrument overheads) was 4236 s.

The CAM observations were performed in staring mode at a magnification of $3\hbox{$^{\prime\prime}$ }$ per pixel and with a gain of 2. Nine different filters were used. The unit integration time was 2.1s per readout and there were between 72 and 197 readouts per image, depending on the filter used. The particular sequence of filters was chosen to go from high to low illumination so as to minimize the detector stabilization time. Good stabilization was also guaranteed by the relatively large number of readouts per exposure.

The PHT-S observations were made at 16 different epochs, from 1996, February 5 to 1997, February 13 (Table 1). The ISOCAM observations of Mrk 279 were all carried out on 1996 February 5, contemporaneously with the first of the PHT-S spectra.

Standard procedures from the CAM Interactive Analysis (CIA) software package were used for the reduction of the ISOCAM data (Ott et al. 1997). The full width at half maximum (FWHM) of Mrk 279 varies with the filter wavelength from $3\hbox{$.\!\!^{\prime\prime}$ }3$ to $5\hbox{$.\!\!^{\prime\prime}$ }0$ (Table 2), but is always consistent with that of a point source. Monochromatic intensities were obtained by integration of the source flux within a circle of radius 6 pixels (18'') and subtraction of a normalised background measured in a concentric circular ring. The intensity of Mrk 279 in the different filters is listed in Table 2. The accuracy of these measurements is $\pm 10$ %.

The ISOPHOT-S data were reduced with the PHOT Interactive Analysis (PIA; Gabriel 1998) software package. However, because ISOPHOT-S was operating close to its sensitivity limit, special reduction and calibration procedures had to be applied. After a change of illumination, the responsivity of the Si:Ga photoconductors immediately jumps to an intermediate level. This initial jump is followed by a characteristic slow transient to the final level. At the faint flux limit, this time constant is extremely long, and in practice only the initial step is observed in chopped-mode. The spectral response function for this particular mode and flux-level was derived directly from observations of a faint standard star HD 132142 whose flux ranges from 0.15 to 2.54 Jy. The calibration star observation was performed with the same chopper frequency and readout-timing as the AGN observations. The S/N of the ISOPHOT-S spectra was considerably enhanced by two additional measures: i) the 32-s integration ramps were divided into sub-ramps of 2 s and no de-glitching (removal of cosmic ray hits) was performed at ramp-level ii) after slope-fitting and de-glitching at slope-level, the maximum of the distribution of the slopes was determined by fitting a gaussian to the histogram. The resulting ISOPHOT-S fluxes are accurate to within $\pm 10$ %.

The 1996 February 5 PHT-S spectrum is shown in Fig. 1, together with the monochromatic intensities measured with ISOCAM on the same day.

$\begin{figure} \includegraphics[width=8.8cm,clip]{ms9537f1.eps} \end{figure}$

Figure 1: The ISOPHOT-S spectrum from 1996 February 5 together with the ISOCAM photometric fluxes from the same date. The x-axis shows the rest frequency at the bottom and the rest wavelength in microns at the top, both on a logarithmic scale

2 IR observations and data reduction