Astron. Astrophys. 362, 75-96 (2000)

3. ISOPHOT observations and data reduction

Photometric data at several (up to 11) wavelengths between 3.6 and 200 µm were obtained for each object using the single-element P1 and P2 detectors plus the two array cameras, C100 (3 pixels 3 pixels) and C200 (2 pixels 2 pixels). Detector and observing parameters are listed in Table 4. Most of the observations (124 in total for 16 objects) were performed in chopper mode, and the remaining (37 for 10 objects) by mapping the region surrounding the target (scans or rasters).

Table 4. ISOPHOT detector properties and covered sky region in mapping observations

In chopper mode the radiation beam is deflected from the source (on-source position) to adjacent fields on the sky (off-source position) several times in order to measure the background emission. Triangular (T) and rectangular (R) chopping modes were used. In the triangular chopper mode the background emission is measured in two different regions, while in the rectangular chopper mode it is measured in only one position. Observing dates, filters, apertures, exposure times, chopping mode, and measured fluxes are reported in Table 5 for each ISOPHOT chopper observations.

Table 5. Details of ISOPHOT Observations performed in Chopper Mode

Table 5. (continued)

Table 5. (continued)

Table 5. (continued).
Notes:
) In units of mJy.
) Doubtful data. Discussed in Sect. 3.3 and 4.1.

In mapping mode the telescope moves in a pattern around the source, providing more sky coverage than in the chopper mode (Table 4). P1 detector maps were performed with an aperture of 52" during all observations, except one (B2 2201+31A) during which the chosen aperture was 23". Observing dates, filters, exposure times, and measured fluxes are reported in Table 6. More details on mapping mode are reported in Sect. 3.2.1.

Table 6. Details of ISOPHOT Observations performed in Raster Mode.
Notes:
) In units of mJy. a) 42 map. b) 24 map. c) Step amplitude of 90". d) Telescope nodding mode.

Table 6. (continued).
Notes:
) In units of mJy. a) 42 map. b) 24 map. c) Step amplitude of 90". d) Telescope nodding mode.

3.1. First steps of the data reduction: from ERD to AAP level

The first part of the data reduction was performed using version 8.1 of the PHT Interactive Analysis (PIA) ⁴ tool (Gabriel et al. 1997). We started the reduction with the raw data processed with version 8.7 of the Off-Line Processing (Laureijs et al. 1998). The raw data form a sequence of detector read-outs distributed in 2ⁿ (n=2-6) sets of four response curves or ramps, as function of time (Edited Raw Data: ERD in Volts).

Each set of four ramps represents a sky position. Each ramp is corrected for the non-linearity of the detector response, and for contamination of cosmic particle events (glitches). The removal of read-outs affected by glitches is carried out by applying two median filtering techniques: the single-threshold technique that uses a threshold of 4.5 standard deviations () for flagging bad read-outs and the two-threshold technique that uses a threshold of 3.0 for flagging and 1.0 for re-accepting read-outs. After applying the non-linearity correction and the deglitching to the ERD, a straight line is fitted to each ramp, in order to determine its slope or Signal per Ramp Data (SRD in Volt/s).

In most of the cases the first 25 or 50% (1 or 2 ramps of 4) of the signals per chopper plateau at the SRD level are discarded to enable the detector response to stabilize at the level corresponding to the source flux density. The remaining data are further corrected for highly discrepant points (value at more than 3 from the average signal) still contaminated by glitches, for the orbital dependent dark current, and for the signal dependence on the ramp integration time (reset time interval) to obtain an average Signal per Chopper Plateau (SCP in Volt/s).

After applying flat-fielding correction using PIA values, the SCP data are calibrated to obtain the Standard Processed Data (SPD in unit of Watts). Since the detector response varies with time, it is determined at the time of the observation by measuring the flux emitted by two thermal Fine Calibration Sources (FCS1 and FCS2) on board. The FCS measurements are reduced in the same way as the scientific measurements up to this step. Data from FCS1 are used because they are the best calibrated. The FCS1 signal is checked in order to remove data with large uncertainties (this step is equivalent to computing the weighted mean of the FCS1 data).

In the case of mapping observations, the FCS1 is observed twice, before and after the observation of the source. The photometric calibration we use is the average value of the two FCS1 measurements.

After the flux calibration the AAP (Auto Analysis Product) data are obtained. They are a sequence of 2ⁿ off- and on-source flux measurements (in Jy) each corresponding to a sky position. The reduction from the AAP level to the final results is performed using our own IDL routines, and not following the standard pipeline. This procedure was also applied in the reduction of ISOPHOT chopper data of a sample of Seyfert galaxies (Polletta & Courvoisier 1999).

3.2. From the AAP level to final results

The last steps of the data reduction before determining the source flux are the background subtraction, the deletion of remaining highly discrepant points, and the correction for effects depending on chopper plateau time, vignetting (only for chopper observations) and point spread function.

In the case of chopper observations with the C100 detector only the central pixel pointed on the source is considered to derive the flux density, since the eight border pixels contain only a small fraction of the central point source and summing these values would hence only increase the noise.

3.2.1. Background subtraction

In chopper observations the background is measured at each off-source position. Since in some cases the instruments show long term drift effects, the background signal is estimated near the time of each on-source measurement and subtracted. In the case of chopper observations the background estimates are obtained by computing the weighted mean of each pair of consecutive off-source measurements. The weights are computed from PIA statistical uncertainties. Since the sequence of chopper plateaux ends with an on-source position, we used the weighted mean of the two last on-source measurements, and the flux observed in the last off-source position to determine the last pair of on- and off-source values, for a total of 2 flux values.

Small maps of the regions immediately surrounding ten of the targets were constructed in one or both of the following ways (Fig. 1): multiple linear scans across the source and rastering the detector about the source. A scan with the P1 and C100 detectors consisted of three steps of the telescope, with this sequence repeated three times. Only the middle row of C100 pixels (8, 5, and 2, as depicted in Fig. 1) viewed the source. The C200 scans contained four steps, repeated twice, with the source centered between two pixels. Note that the source was observed in only the middle two steps of the C200 scan. The raster patterns were 33, 33, and 42 or 24, for the P1, C100, and C200 detectors, respectively. Each pixel viewed the source once in the raster maps. The step size between exposures for both scans and rasters was approximately equal to the pixel size, a little more to take into account the gap between pixels, for the C100 and C200 cameras, and equal to the aperture size for the P1 detectors, which resulted in a different total sky coverage for each detector (see Table 4). A background estimate for each on-source measurement was obtained from a weighted average of the flux measured in the raster or scan positions immediately preceding and following the source position by the same pixel. Using the same pixel to determine the background reduced the impact of uncertainties in the flat field. The weighted average background was subtracted from each on-source measurement providing a sequence of source flux values.

Fig. 1. Schematic representation of the two observing modes used for mapping observations: telescope nodding (scan) on the left side and classical raster on the right side. The position of the source in the different cases is represented by a star. The bold square represents the initial position of the detector. The numbering corresponds to the pixels numbers. The detectors move first from the left to the right, then in the opposite direction. In the case of classical raster, at each change of the horizontal direction, the detector shifts down one step. All the steps are separated by the pixel size. The reported pixel sizes do not correspond to the real proportions. The numbers in parenthesis refer to the number of horizontal steps the number of repetitions for nodding and vertical steps of telescope motion in the raster maps.

Residual effects of detector instabilities, in both chopper observations and maps, produced occasional discrepant points, which were culled by one-pass sigma clipping. The threshold number of standard deviations to reject a flux value depended on the number of points in the sequence (Chauvenet's criterion in Taylor (1982)), ranging from 1.15 to 2.66.

In the case of chopper observations with the C200 camera, the source flux was computed by adding together the fluxes measured by each of the four pixels. The source flux was divided between pairs of adjoining pixels in the C200 scans; these were averaged by weighting with their uncertainties. Table 6 lists the weighted mean of each flux sequence after clipping, or 3 upper limits for non-detection, where is the quadratic sum of the statistical and systematic uncertainties (see Sect. 3.3), also reported in the tables.

3.2.2. Vignetting correction

In the case of chopper observations with the C100 and C200 detectors, the data are further corrected for the signal loss outside the beam of the telescope (vignetting). The PIA default vignetting correction factors were computed considering the dependency only on the distance of the chopper positions and on the filter, but recent investigations have shown that they depend also on the time per chopper plateau (M. Haas, private communication).

The PIA default vignetting factors were applied directly to the data, resulting in little change to the flux values obtained with the C100 camera. However, the fluxes from the C200 detector are 80%, in median, larger after the vignetting correction. Given the uncertainty over the accuracy of the vignetting factors, we averaged the fluxes derived with and without the correction and added the respective uncertainties in quadrature.

3.2.3. Correction for effects depending on the chopper plateau time and point spread function

In the case of chopper observations a correction for the signal loss due to effects depending on the time per chopper plateau is applied to the computed flux value. Since short integration times with the C100 and C200 detectors do not reach the full signal, the observed flux can be reduced by large factors, typically up to 68% for the C100 detector, and up to 12% for the C200 detector for the shortest observations. The correction factors we use are the PIA default values.

All the computed fluxes are finally corrected for the point spread function (psf) of each detector using the default PIA values derived empirically in most of the cases. The available psf correction values correspond to a source position centered in a pixel or located in a corner of the pixel. In case of scans with the C200 detector we needed the psf correction corresponding to the target located in the middle of a side of the pixel. We derived it by assuming a bi-dimensional Gaussian function for the psf and constraining its parameters using the other two known values for each wavelength (30.6% at 150 µm, and 29.1% at 170 µm).

3.3. Calculation of systematic uncertainties

During the data reduction only statistical errors were taken into account, and not those in the absolute flux density calibration. The accuracy of the absolute photometric calibration depends mainly on systematic errors (detector transient effects, calibration response, dark current, point spread function), and it is currently known to be better than 30% (Klaas et al. 1998). The associated statistical and systematic uncertainties are reported in the last two columns of Table 5 and Table 6. We associate to the measured flux an uncertainty that is the quadratic sum of their statistical uncertainties and 30% (for the C200 measurements obtained in chopper mode we use 40% that corresponds to half of the uncertainty due to the vignetting correction) of the measured value. Among our ISOPHOT observations the datasets of B2 1721+34 were identified as failed after inspection of their quality by the ISO team because they were heavily affected by cosmic rays and thus scientifically useless. We report these data in Table 5, and indicate that they are doubtful in a footnote.

© European Southern Observatory (ESO) 2000

Online publication: October 30, 19100
helpdesk.link@springer.de