next up previous contents index
Next: 7.4 Error Propagation in Up: 7. Standard Product Generation Previous: 7.2 From ERD to

Subsections


7.3 From SPD to AAR

Auto-Analysis7.2 (AA) is the processing stage that starts from SPD, corrects for all instrumental effects lasting longer than one reset interval and generates sets of spectra, fluxes (in Jy) against wavelength (in $ \mu $m). It applies further corrections to the slope data to transform it into fully calibrated astronomical data. These spectra are generated irrespective of the observation specifications. Figure 7.6 gives an overview of AA processing. Most of the SWS AA processing steps are independent of AOT number as most of the special options (e.g. reference scans) are used in several different AOTs and thus are implemented as part of the main processing flow.

These steps are described in more detail in the following pages.

Figure 7.6: Auto-Analysis flow-diagram
\begin{figure} % latex2html id marker 6194\begin{center} \setlength {\unitleng... ...put(8.0, 1.0){\makebox(1.0,0.4){$\vdots$}} \end{picture}\end{center}\end{figure}


7.3.1 Reading data from the SPD

The data and header information are read from the input SPD. Contrary to the ERD, where we needed to read in several files, in this stage all information is contained in the SPD. The AOT number, needed for the further processing of the data, is determined from the EOHAAOTN keyword in the SPD FITS header. The spacecraft's heliocentric velocity, and, for SWS01s, the scan speed, are also obtained.

7.3.2 Transient correction

As the first few samples of dark current measurements may suffer from memory effects, or unknown shutter status, the first three datapoints of dark currents for bands 1, 3, 4, 5 and 6 are rejected at this processing step. This is all that is currently done for transient correction in bands other than band 2. We have seen no transient problems in bands 1 or 3. In bands 4, 5 and 6 they are certainly present but no other correction is available. Band 2 transients are treated as changing dark currents, see Section 7.3.3.2.

7.3.3 Subtraction of dark currents

For each observation dark currents are subtracted in all bands for all detectors separately. In band 2 a different dark current model is used than in the other bands because of the transient effects present there and because a model is available which can actually handle these transients.


7.3.3.1 Dark currents in general

For every block of dark current data, the median of valid dark current data is calculated. For every non-dark current scan two dark blocks are selected, one immediately before and one immediately after the scan, which have the same gain and reset lengths as the scan. All AOTs are devised such that this is the case. A linearly interpolated slope connecting the medians of both darks is subtracted from the scan.

In few cases only one dark current block can be associated with a scan. If so, the median of this block is subtracted from the scan. In even fewer cases there is no valid dark current associated with the scan. Then the mean value of all dark currents in the observation with the same gain and reset is used. If there is no valid dark current at all with the same gain and reset length the block is flagged as having no data.

For the photometric checks only the dark current preceding the measurement is subtracted. The calculated dark current values are also subtracted from the dark current frames. The effect on the data of one of the band 1 detectors is shown in Figure 7.7.

Figure 7.7: The dark current correction. In the upper panel detector 10 of an SWS02 is displayed. The samples in the dark current blocks are indicated in black. The lines show the linear interpolation. Blue points indicate high gain data and red points indicate low gain data. Note that there is a clear jump between different gains. In the lower panel the dark current is subtracted.
\rotatebox {90}{\resizebox{!}{12cm}{\includegraphics{dd10.ps}}}

7.3.3.2 Dark currents in band 2

The Fouks-Schubert model, as described in more detail in Section 9.2.1, is used to calculate the varying dark current for the band 2 detectors.

$\displaystyle S_{k}=\beta S^{\infty}_{k}+\alpha _{k} + Z(t_k)$ (7.10)

$\displaystyle \alpha _{k} = \frac{ ( 1 - \beta ) \alpha_{k-1} S^{\infty}_{k} } ... ...( 1 - \beta ) S^{\infty}_{k} ) \exp( -\Lambda \beta S^{\infty}_{k} \Delta t ) }$ (7.11)

Here $ S_{k}$ is the signal through the detector at time $ k$, $ S^{\infty}_{k}$ is the relaxed signal at time $ t = \infty$. It is proportional to the incoming flux. $ \beta$ is a constant, representing the instantaneous increase at a flux step, and $ \alpha_{k}$ comprises the memory effects. $ \alpha_{k}$ is dependant on its previous value, $ \alpha_{k-1}$, on the present relaxed signal, $ S^{\infty}_{k}$, and on the constants $ \beta$ and $ \Lambda$. $ \Delta t$ is the duration of the time step from $ k-1$ to $ k$. $ Z(t_k)$ is the zero level function which represents the level to which the signal would decay when no flux is entering the system. The signal is related to the relaxed signal (or flux) via a constant gain $ \beta$ and a changing dark current $ \alpha $. This changing dark current is subtracted from the data.

The model described in Equations 7.10 and 7.11 is used with a fixed value for $ \beta$ = 0.82 and $ \Lambda$ = 63 V$ ^{-1}$; they are material parameters verified to be constant over the mission. For every observation the zero-level function and the starting value for $ \alpha $ had to be estimated anew. They depend on the flux history as seen by the detectors.

The effect on the data of one of the band 2 detectors is shown in Figure 7.8.

Figure 7.8: The dark current correction in band 2. The layout in this figure is the same as in Figure 7.7 with the exception that there is medium gain data too, shown in green. Note the changing dark currents and how they make the individual scans in the lower panel much more symmetric; they are all up-down scans. Note also that in cases of low flux there is not much difference with the `standard' dark current subtraction.
\rotatebox {90}{\resizebox{!}{12cm}{\includegraphics{dd14.ps}}}


7.3.4 Responsivity calibration

A Relative Spectral Response Function (RSRF) has been determined for every band (Section 5.4). It provides the relative response of the instrument as a function of wavelength. They are stored in calibration files Cal-G 25_xx, where `xx' is the band designation.

For each band the wavelength dependent responsivities are normalised with respect to the key wavelength. This is done per detector. The signals that belong to a certain band are then divided by the normalised responsivities for that band at the corresponding wavelengths. An example of this is shown in Figure 7.9.

Additionally, for SWS01s the responsivities are smoothed by carrying out a flux conserving interpolation to the SWS01 resolution. These factors depend on both the AOT speed and on the band as well as on wavelength, and are stored in the calibration file Cal-G 19.

Figure 7.9: The RSRF correction on an SWS01 observation in detector 32. In the upper panel the data are displayed after dark current subtraction, the up scan is in green and the down scan on red. Also is displayed the (scaled) RSRF for band 3A in black. In the lower panel the data have been divided by the RSRF, removing most of the instrument specific wavelength dependencies like e.g. fringes. Some fringing remains however, see Section 9.7.
\rotatebox {90}{\resizebox{!}{12cm}{\includegraphics{rsrf.ps}}}


7.3.5 Flux calibration

The flux calibration comprises three processing steps, which can be done in any order. Flat-fielding brings all detectors to one level. Photometric check removes long term drifts and flux conversion scales the $ \mu $V/s into Jy.

7.3.5.1 Flat-fielding

In the SWS grating scans, a spectral segment is observed with 12 separate detectors at any one time. In general, these detectors must be individually calibrated such that the resulting spectra from each detector within an AOT band gives the same signal. It was initially recognized that the signal-to-flux calibration per detector formally takes care of the flat-fielding. However, the responsivity of an individual detector relative to the average of the 12 detectors from the same band was found to be quite stable during the mission as determined from $ \approx$1000 observations, see Section 5.6. Therefore, an independent flat-fielding measure was introduced into the pipeline to bring the response of each detector in agreement with the average of all the detectors within an AOT band

The gains of a single detector compared with the average of all detectors in the band are stored in the calibration file Cal-G 43. This flat-fielding step applies a scaling based on Cal-G 43 and results in all detectors being placed on the same scale as an average detector. An example of flat-fielding is shown in Figure 7.10.

Although additional flat-fielding is usually performed in post processing, the application of a separate flat-fielding step here aids in removing some of the ambiguity of applying either a scaling or an offset correction. As treated in the pipeline processing, the flat-fielding is always a scaling correction. In post processing it might still be necessary to account for offsets introduced through the dark current subtraction.

Figure 7.10: The effect of flat-fielding is shown on a portion of data from band 3. Each detector is shown in a different colour. Before the correction (top panel) the signal levels are widely separated, whereas after (bottom) they more closely approximate each other.
\rotatebox {90}{\resizebox{!}{12cm}{\includegraphics{flat-field.ps}}}

7.3.5.2 Photometric check

For almost every observation a photometric check is performed using one of the internal stimulators to correct for any long term drifts of detector sensitivity, assuming that all signal changes can be traced to changes in the detectors and that the illuminator itself is constant. The wavelength calibrator was used for this purpose because its flux level was more appropriate. The wavelength calibrator source is scanned over some 20 LVDT steps.

Stimulating the band 1 detectors, however, required a high setting of the calibrator which resulted in significant memory effects in bands 2 and 4. This was noted early in the mission. Photometric checks of band 1 were seen to be quite stable so the photometric check with the high setting was discontinued in revolution 64. Band 1 was periodically monitored throughout the mission with strong illuminator flashes to guard against any possible detector changes. No change was noticed.

The detector signals during the photometric check are compared to the average photometric check over the mission, held in the calibration file Cal-G 41. The signal response is corrected based on an average photometric signal contained in Cal-G 41, see Section 5.5 for more information.

The 6 numbers that are found for the photometric gains are listed in the header of the SPD (to be copied later to the AAR header) under the keywords PHOT_BD*, where `*' is a number indicating the band.

7.3.5.3 Converting signal to flux

Once the signal, for each band, has been calibrated to a standard time by the photometric check, and to a standard detector using the flat-gain, it can be converted from $ \mu $V/s to Jy/s. These conversion factors are held in the calibration file Cal-G 42.

The effect of this processing step is shown in Figure 7.11

Figure 7.11: The effect of flux calibration is shown for a line in band 1E. The signal in $ \mu $V/s is read from the left axis while the flux in Jy is read from the right hand side axis.
\rotatebox {90}{\resizebox{!}{12cm}{\includegraphics*[2cm,0cm][12cm,25cm] {fluxcon.ps}}}

7.3.6 Velocity correction

The velocity component ($ V$) in the line of sight towards the target, due to the combined motions of the spacecraft and of the Earth is corrected for in this module. This then ensures that the AAR wavelength is heliocentric.

7.3.7 Output table files

All good data (wavelength, fluxes etc.) are sorted to wavelength and written out to a FITS file. Data for which no wavelength was assigned in Derive-SPD are skipped. The resulting file is a FITS AAR file. For a definition of this file see Section A.4.1, and an example AAR file from an SWS01 observation is shown in Figure 3.5.

next up previous contents index
Next: 7.4 Error Propagation in Up: 7. Standard Product Generation Previous: 7.2 From ERD to
ISO Handbook Volume V (SWS), Version 2.0.1, SAI/2000-008/Dc