SWS Calibration Status

(Last update:  29 March 1999)

Introduction

This note presents a top-level summary of the current status of the Short Wavelenght Spectrometer's (SWS) in-flight calibration and characterisation, as well as improvements that are hoped to be achieved during the post mission phase.  The SWS instrument operated at medium and high spectral resolution in the wavelength range from 2.38 to 45 µm.

The SWS calibration is progressing very well.  The goals originally set before the mission have, in general, already been achieved for well-behaved data sets.  In the context of SWS well-behaved implies point sources in the several hundred to several thousand Jansky flux density range, that were observed with correct pointing and suffered no instrumental or operational problems.  An area where improvements over the current status are expected to be possible is in the characterisation of detector memory effects.

At this time, an 'error- recipe' for the combination of all individual error contributions to a useful overall error estimate for SWS data products does not exist.  The main reason being that the relevance of the various error contributions depends strongly on the properties of the observed source (e.g. extended or point source) as well as on the type of scientific analysis that is aimed for (e.g. line fluxes or spectral energy distribution). It is hoped to provide such a recipe at some point in the post operations phase (see below).

In this note no error budget in the true sense will be given, only an indication of the currently known error contributors.  Scientists are encouraged to evaluate which are the most important error terms for their particular scientific application of SWS data.
 
 

Calibration Accuracy Status for Version 7 Pipeline Products

In this section the calibration accuracies are described, as currently valid for the standard off-line processing (OLP, V7.0) products for well-behaved data sets.  For SWS well behaved firstly implies that it is a point source, i.e. smaller than a few seconds of arc.  The source should have a moderate intensity in the several hundred to several thousand Jansky flux density range.  The observation must have been carried out with correct pointing with the source at the centre of the SWS entrance aperture. Finally the data should not be suffering from any instrumental problems (e.g. strong glitches or detector sensitivity jumps) or operational problems (e.g. a dark current missing due to missing telemetry).

Wavelength Calibration:  The pre-launch goal was to achieve an accuracy of about 10% of a resolution element for the grating sections. This goal has been reached.  For the SW section (2.3 to 12 µm), the achieved accuracy is 1/16 of a resolution element and, for the LW section (12 to 44.5 µm), this is 1/8.  Both Fabry-Pérots's are accurate to about 25% of a wavelength resolution element.

Wavelength stability is very good. During the first year of the mission, the shift was only about 3/8 of a resolution element. This shift is, of course, corrected in the respective calibration tables.

Flux Calibration: The pre-launch goal for the absolute flux calibration was to achieve an accuracy of 30% or better.  This goal has also been achieved.  The status of the accuracy of the SWS flux calibration is given in Table 1.  The accuracies given here are as determined in June 1997 for version 6 of the off-line processing (OLP).  A table including minor corrections as applicable for OLP V7.0 is currently in preparation.  The F-P calibration is tied to that of the LW grating section.  The quoted 'worst' uncertainties are primarily valid at the band edges of the AOT bands, representing a maximum product of the absolute and relative spectral responsivity uncertainties.  With the ongoing improvements in the determination of the Relative Spectral Response Function and the corrections for the hysteresis effects, these values will certainly go down, particularly at the long wavelengths.

The flux reproducibility was checked during the mission by regular observations of the calibration star HR6705. The various measurements agreed to within 6% for band 1 and 12 to 15% for band 2 to 4.
 
 

Section	Band	Key	Grating	Aperture	Aperture	Detector	Detector	Wavelength	Accuracy	Accuracy
		wavelength	order	number	area	type	pixels	range	key	worst
		µm			"			µm	%	%
SW	1A	2.48	SW4	1	14x20	InSb	1 - 12	2.38 - 2.60	5	7
	1B	2.87	SW3	1	14x20	InSb	1 - 12	2.60 - 3.02	5	7
	1D	3.08	SW3	2	14x20	InSb	1 - 12	3.02 - 3.52	5	7
	1E	3.80	SW2	2	14x20	InSb	1 - 12	3.52 - 4.08	5	7
	2A	4.50	SW2	2	14x20	Si:Ga	1 - 12	4.08 - 5.30	7	12
	2B	5.90	SW1	2	14x20	Si:Ga	1 - 12	5.30 - 7.0	7	15
	2C	7.70	SW1	3	14x20	Si:Ga	1 - 12	7.0 - 12.0	11	25
LW	3A	14.0	LW2	1	14x27	Si:Ga	1 - 12	12.0 - 16.5	11	25
	3C	17.0	LW2	2	14x27	Si:Ga	1 - 12	16.5 - 19.5	11	20
	3D	24.0	LW1	2	14x27	Si:Ga	1 - 12	19.5 - 27.5	12	20
	3E	28.5	LW1	3	20x27	Si:Ga	1 - 12	27.5 - 29.0	20	30
	4	32.0	LW1	3	20x33	Ge:Be	1 - 12	29.0 - 45.2	30	35
	5A	11.8	LW3	1	10x39	Si:Sb	1 - 2	11.4 - 12.2	11	25
	5B	14.0	LW2	1	10x39	Si:Sb	1 - 2	12.2 - 16.0	11	20
	5C	17.0	LW2	2	10x39	Si:Sb	1 - 2	16.0 - 19.0	11	20
	5D	24.0	LW1	2	10x39	Si:Sb	1 - 2	19.0 - 26.0	12	30
	6	27.0	LW1	3	17x40	Ge:Be	1 - 2	26.0 - 44.5	30	35

 

Table 1 Overview of the SWS AOT bands and flux calibration accuracy.
 
 

Calibration Accuracy Status for OSIA Products

Generally, for well-behaved data sets, use of the Observers' SWS  Interactive Analysis (OSIA) tools does not lead to significant improvement in the calibration.  The main use of OSIA processing for such well-behaved data sets is to boost the confidence of the researchers in their results. The only area where OSIA can give higher quality products than OLP 7.0 is by removing internal interference fringes as seen in band  3 and, to a lesser degree, also in band 2 data. The proper removal of  these fringes depends heavily on knowledge of the properties of the source (point source or extended, general spectral shape etc.).

For data sets that are not well-behaved as described above, the quoted accuracies are not reached with standard OLP type processing.  However, often manual processing using OSIA tools will allow scientists to reach the same level of accuracy.  Currently, techniques are available to circumvent a number of the problems that are found in data sets that are not well behaved. The following paragraphs describe these cases, as well as the some of the problems for which solutions are less obvious.

Detector memory effects: Currently there is no foolproof method to correct SWS data for detector memory effects.  Bands 2 (Si:Ga) and 4 (Ge:Be) are especially affected.  The signature of memory effects is that the 'up' and the 'down' scans -that are carried out at least once in all SWS AOTs- are different in flux level.  The 'down' scan normally succeeds the 'up' scan in the AOT and appears to be already 'accustomed' to the flux level.  Thus, one procedure that has been used to improve data affected by memory effects is to take the 'down' scan as the reference for the flux level when flat-fielding the data within OSIA.

Dark current problems: In all SWS AOTs, only limited amounts of dark current data are taken.  These dark current data can be 'corrupted' by memory effects (e.g. when a dark current measurement follows an internal calibrator scan) or e.g. strong glitches.  A badly subtracted dark current usually results in systematic differences between 'up' and 'down' scans. For large errors in the subtracted dark current, the final spectra will resemble the spectral responsivity. Unfortunately small dark current problems and detector memory effects have a similar signature in the final spectra.  Correcting for these effects requires manual deletion of dark current data.

In flight it was realised that detector memory effects could have a severe impact on the dark current data. Therefore, the scanning logic was changed for all AOTs.  As a result observations carried out early in the mission are often more problematic in terms of dark current subtraction.

Fringes: Fringes due to FP type interference inside the SWS detector material are present in data for bands 2 and 3.  The amplitude and frequency of the fringes depends on the source morphology, its brightness and its position in the SWS entrance aperture.  Fringe amplitudes of about 5% of the continuum level are seen in band 2 and up to 20% in band 3.  Fourier filtering or sine fitting within OSIA can be used to remove the fringes 'manually'.

Pointing offsets: Because of non-uniformity of the detector material and small (internal) misalignments, the SWS detector 'beam' patterns are not simple flat-topped functions.  As a result, if a point source is observed with a pointing offset from the centre of the SWS aperture, the signals in the different AOT bands are reduced by band-dependent factors. Pointing offsets can be recognised in the data by jumps in the continuum level at band edges in AOTs 1 and 6.  In AOT 2 data, the effect is very difficult to see. Corrections for this effect can only be applied when the detailed beam profiles have been determined and the actual satellite pointing is known with sufficient accuracy.

A pointing offset in the dispersion direction furthermore leads to errors in the wavelength scale. This is visible as errors in the wavelength of emission or absorption features and a poorly-applied correction for the relative spectral responsivity of the system.  If the pointing offset is known, this can be corrected by incorporating it in the wavelength calculation. Note that this error can lead to misinterpretation in the kinematics of the observed object.

Pointing offsets have been significantly reduced during the mission;therefore observations in the first year are more likely to be affected by this problem.

Glitches: All SWS detector data suffer from cosmic ray hits.  A number of the 'glitches' in the data following from these hits are removed automatically, but especially in the long wavelength bands this by no means complete.  In these bands, a significant increase in the noise level is seen, as well as the presence of tails due to the detector memory effect after a cosmic ray hit.  Currently these can only be removed manually within OSIA.

Extended sources: All of the SWS calibrations are based on point source emission.  As a result, the fluxes derived are not appropriate for extended sources (partially) filling the SWS aperture.  This is clearly seen for AOT 1 and AOT 6 scans crossing band edges.  At the band edges, jumps in the signal level are seen, which are due to the change in the effective size of the detectors.  To obtain reliable flux densities for extended sources, the SWS detector beam profiles should be convolved with the source structure at the appropriate wavelength.  As a first order correction, the ratio of the aperture sizes can be taken.

Very weak sources: For very weak sources careful inspection of the data is required to be able to filter out e.g. data affected by glitches and to be able to determine the true dark current level, fringes are normally insignificant.  Proper calibration and analysis of such data is only possible with much manual labour.

Very bright sources:  Very bright sources suffer mainly from memory effects and fringes.  However, the data may also be affected by detector saturation which starts to become important at the several 10000 Jy level.  Again much manual labour is needed with such observations.
 
 

Projected Calibration Accuracy in Final Archive

One of the main aims of the post operations phase is to characterise better the detector memory effects and their consequences for SWS observations.  It is hoped that these efforts will lead to algorithms that can be incorporated into the off-line processing (OLP).  ln this case, a direct improvement of the spectral shape of band 2 and band 4 data can be expected for observations of data with strong signal in those bands.  A better understanding of the memory effects will likely also lead to higher accuracies in the band 2 and band 4 flux calibration scale.  It is unfortunately not possible to quantify this improvement at the present time.

A second important aim for the off-line processing (OLP) for the final archive is to make the errors as given in the SWS data products (SPD and AAR) a useful quantity.  The errors given in the SWS data products -as calculated currently in OLP 7.0- are entirely dominated by the systematic errors introduced by the uncertainty in the absolute flux scale (see Table 1).  As a result, these error values cannot be used e.g. as weights when averaging data.  It is the aim to design an error propagation mechanism for OLP that overcomes this limitation.

In the area of extended sources,  calibration improvements are expected as well.  When the detailed beam profiles of the SWS detection system have been determined, correction factors for the spatial morphology of the observed sources can be derived and applied to the data. Along the same vein, pointing offset correction factors can be applied to SWS data.  It is however not obvious how to automatically apply such corrections because this requires the introduction of information about source structure into the OLP environment.

Some improvement is expected in the wavelength calibration in the sense that the temporal variations will be characterised somewhat better.  The final accuracy for the wavelength scale will be only slightly higher than quoted at this time.