The deglitching performed by Derive-SPD is believed to remove the majority of the glitches in the data. However, some glitches may still remain undetected. In particular, any glitches which occur during the period of time discard at the beginning of each ramp are not currently detected.
In addition it has been found that some (large) glitches have a long lasting effect on the detector responsivity. They can cause one scan for one detector to be significantly higher for some period of time. At this time no processing solution is known for this problem.
Deleted
Deleted
For very strong sources the integration ramps become non-linear. This is seen in detectors LW1-4 (LW3 being the worst affected), where the adjacent detector sub-spectra have different slopes/shapes and `sag' in the middle. Around the middle of the filter profile more flux reaches the detector than at the edges. Because of the non-linear response the photocurrent at the middle of the profile is proportionately less than at the edges of the filter response where the flux on the detector is less.
This causes the sub-spectra shapes to sag - to be higher at the edges and lower near the centre (ie the sub-spectra are `saggy').
In figure 7.1 an example of saggy sub-spectra. It is quite clear that the spectra do not represent the correct fluxes as two detectors, with an overlapping wavelength range, do not give the same shape of spectrum!
This non-linearity can be corrected for using Saturn. A model of Saturn (Davis G.R. et al. 1996, A&A 315, L393) converted to photocurrents, using the low-flux linear response, is compared with Saturn's photocurrents from the LWS giving a relationship between the two.
The figure 7.2 shows the model's
photocurrents versus the LWS data for Saturn, over-plotted with the
second order polynomial forced through zero:
Saturnmodel = a * Saturndata + b * (Saturndata)2 | (7.1) |
Sourcecorrected = a * Sourcedata + b * (Sourcedata)2 | (7.2) |
This method produces sub-spectra which agree with each other and the correction gives a relative calibration (see section 10.X for instructions to find out if you need the correction and how to apply it to your data).
Fig 7.3 shows the full corrected spectrum of the source from fig 7.1. It has the correction applied to the data and the detector sub-spectra have been scaled together. It can be seen that the sub-spectra are no longer saggy and the overlapping parts agree on the same spectral shape.
The lvdt versus wavelength relationship has proven to vary with time, not in a continuous manner. The cause of this grating behaviour is not yet understood and is under study. The variations are largest for the extreme values of lvdt (around 1000 or lower and around 3000 and higher; i.e. close to the overlap regions of the detectors), where in the worse case they have reached 50 but there is no variation for the mid-values of lvdt, around 2000 (i.e. close to the center of the detector range).
In particular there has been a sudden change in the grating behaviour between rev 342 and rev 349, followed by a relatively stable state up to the end of the mission. In consequence, the present calibration is time dependent and considers two time periods: the first relationship is used up to revolution 345, the second one from revolution 346 upto revolution 875 (the end of the mission).
Prior to the jump, there has been a progessive noise increase in the lvdt readings, introducing increased wavelength uncertainties for some observations. LWS users should be warned that errors of almost up to 1/3 of a resolution element (i.e. 0.1 m for the short wavelength detectors and 0.2 m for the long wavelength detectors) have been observed in a few observations in the revolutions preceding revolution 346.
Observers are warned of spurious features which may appear in LWS spectra. In contrast to the broad features caused by the near-infrared leak (see section 7.11), these features reside in the calibration files and are transferred to the observed spectrum during the calibration process.
The primary calibration source for the LWS is Uranus. For each detector, the measured photocurrent is plotted as a function of wavelength. A model of Uranus is then used to predict the flux-density falling on the entrance pupil of the telescope, also as a function of wavelength. From these two plots, the overall responsivity of the instrument, in terms of current per unit flux-density, is recorded, detector by detector. For historical reasons, these responsivities are known as the detector "filter profiles".
The calibration of any other source is then achieved by dividing the observed photocurrents for that source by the filter profiles. Clearly, any spurious features in the profiles will be transferred to the calibrated spectrum.
The signal-to-noise ratio obtained in the calibration observations of Uranus - and therefore in deriving the filter profiles - is comparatively modest. It limits the signal-to-noise ratio on the calibrated spectrum of any other source, no matter how strong. The situation will be improved when we re-derive the profiles using Mars. However, observers should be very circumspect in picking out unknown "features" in their spectra: these should be checked against the published filter profiles (in the LCGR file on the CD ROMs).
We have identified lines and features in the spectrum of NGC 7027 arising from Uranus. The 112 micron HD line emission line in 7027 has exactly the same equivalent width as the same line in absorption in Uranus. Similarly, the 56 m HD line has an equivalent width of 3.4 millimicrons in NGC 7027 compared with 2.8 millimicrons in eta Car, which is 100 times brighter. We suggest that only HD lines with equivalent widths very much greater than 3 millimicrons should be accepted as real.
Other features may also be present. Again, the situation should improve as soon as Mars is available.
Deleted
In the data products derived for the FPs, some data can be marked as bad, although there is in principle nothing wrong with this data in the SPD file. The reason for the discrepancy lies in the way the responsivity correction is performed for FP data. The processing software determines for each line where the FP is expected to be for the position of the grating that was used. This expected FP position depends on the order that the FP is working in. The software then checks that the FP was indeed scanned around this expected position. If this is the case, the responsivity correction is done for this grating and FP position. If the FP however was not scanned around the expected value, the software flags the data as invalid, since no valid responsivity can be found for this FP position.
With the current calibration files for some wavelength regions (especially at points where the order of the FP that is used changes) the FP position can not be found in the calibration files, although the FP was scanned over the correct range. This is due to a difference in the FP order between the processing software and the AOT planning software. Although the new calibration files, that are used since Version 7 of the Off-line processing, have been derived from the uplink files, there still are some problems with the order selection remaining. To solve these remaining cases will probably require significant changes to the processing software.
The observer can check if there were problems with the responsivity correction for his data by checking the status words in the Auto Analysis product file (see section 8.2.8).
The LWS processing corrects the data for the effective aperture of the instrument, assuming that the source that is observed is a point source in the center of the LWS aperture. For an extended source a correction has to be applied to the calculated fluxes, because for an extended source (that is bigger than the LWS aperture) the diffraction of the telescope shall not be wavelength dependent. For a point source the instrument is diffraction limited at about 110 m, so for longer wavelengths the effective aperture of the instrument will be smaller (part of the flux will be defracted out of the aperture). For a source that is in between these to cases, the procedure is more complicated and would involve a deconvolution with the instrument beam pattern. At this moment the procedure to do this is not yet known. In Table 7.1 the correction factor for the center wavelengths of the ten detectors is given for point and extended sources.
In the first data that was obtained with the LWS instrument it became clear that there were fringes in the spectra. During the Performance verification phase the fringing was seen in a number of spectra of different sources. It has been shown that the fringes are strongest in extended sources and in point sources that are observed off-axis. Also sources on a strong background will have strong fringing. The exact cause of the fringing is not yet known. The period of the fringes is constant in wavenumber in all sources at about 3.6 cm-1. The fringe period can be described with an interference formula:
(7.3) |
where d is the gap that causes the fringes, n is the order of the fringe and is the wavelength of the fringes. For two adjacent fringe peaks at wavelengths 1 and 2 we get:
(7.4) |
The gap that was derived from the data taking in PV phase that contained fringes was 1.455 mm. The observers can use this gap to identify the fringes. the amplitude of the fringes depends on the wavelength they are observed at, the extend of the source, the strength of the source and the position of the source in the beam. Currently a procedure is being written that can correct the data for fringes. This procedure may later be implemented in the ISO Spectrometer Analysis Package (ISAP), see also Chapter refch-post.
Another anomaly that was found in one of the early observations with the LWS instrument was the occurrence of straylight features in the spectra. These show up as broad emission features, and could easily be identified as dust emission features. However if the spectra of different detectors are compared the features move in wavelength from one detector to the next. The features are observed on detectors that are close to each other in the LWS instrument. The order of the detectors in the instrument is not in increasing wavelength, instead the detectors are paired with a short and a long wavelength detector in each pair. The order of the detectors in the instrument is SW1, LW1, SW2, LW2, SW3, LW3, SW4, LW4, SW5, LW5. In the following two sections the two straylight features that were identified upto now are described.
LWS spectra of stars that are bright in the near-IR sometimes contain features which resemble broad spectral features, but which do not occur at the same wavelength in detectors which cover the same range. The origin of these features is believed to be a near-infrared leak in the blocking filters located in front of the detectors. This supposition is supported by a statistical test, which showed a correlation between these features and the J, H, and K bands. The strength of these features is best correlated with the strength of the sources in the H-band (1.6 um). In a small sample of post-main-sequence stars observed in the Core Programme, all those which had H-band emission brighter than about 2.2 magnitudes ( 140 Jy) were seen to exhibit these spurious features.
Follow-up observations of one of the affected sources gave results consistent with a near-IR leak as the origin of the features. The spurious features are now known to be reproducible, in terms of wavelength, shape and FWHM from one source to another and for repeated observations of the same source. This means that a template profile can be created which observers can fit to their data and use to remove the spurious features. A more sophisticated profile has yet to be derived, but to first order the profile is Gaussian for each of the three most-affected detectors: SW2, LW1 and LW2. The short-wavelength wing of a similar feature also affects the longest-wavelengths of detector SW1's range, while the long-wavelength wing of a spurious feature affects the shortest wavelengths of detector SW3's range. The best estimates of central wavelength and FWHM of the gaussians are given in table 7.2.
Using these parameters, it has proved possible to effectively remove the spurious features and to detect narrow emission lines that were undetectable before removal of the spurious features.
Based on the small number of affected sources in the Core Programme, we can offer the following estimates:
During the careful analysis of an LWS spectrum, one of the LWS consortium members found strange emission features on the longest wavelength detectors (LW3, LW4 and LW5), that were very similar to previously found features on shorter wavelength detectors due to a NIR leak (see note on the web dated 31 January 1997). A more thorough investigation by the LWS Instrument Team led to the discovery of about 46 cases to date. The features are only detectable in L01 spectra, since the wavelength ranges covered in L02 and L04 spectra are too small to be able to distinguish these features. For L03 data the current data processing is not in a state where a clear statement on the presence or absence of these features can be made.
The straylight features manifest themselves as broad (few micron) emission peaks at 155 micron on detector LW3, 159 micron on detector LW4, and 163 micron on detector LW5. Furthermore a rise in flux can be seen on the LW5 detector longward of 180 micron. The features are visible only on rather faint sources (typically less than a few 10-18 W/cm2 um at 160 micron), since the peak strength of the features is a few times 10-19 W/cm2 um. An example of a spectrum with the features is shown in Figure 7.4. The biggest problem of these features is that the one on the LW4 detector blends with the 157 micron [C II] line. A possible additional component of straylight has been seen in a calibration source where an increase in the continuum flux level is observed in an observation with features with respect to other observations without the features. This however needs further checks on other calibration sources.
An investigation was made to possible correlations between the occurance of the features and other spacecraft and instrument parameters such as temperatures, solar aspect angle etc., in order to try to find the cause of the straylight. Only one correlation could be found upto now. All observations of sources with low fluxes in LW3, LW4, and LW5 that have a dark current much higher than nominal (more than about a factor 2 times the nominal dark current), show the features. However, some observations that show the features show a perfectly nominal dark current. So clearly the relation is not perfect. No other correlation could be found, and thus the cause of these straylight features remains unclear.
http://www.iso.vilspa.esa.es/instr/LWS
).
The way the LWS is operated means that the spectra for different detectors will have an overlap region. Although these overlap regions of sub-spectra should after the processing be at the same flux level, it is known that in some cases there is a difference between the subspectra. These differences can be due to imperfect correction for the spectral response changes during the AOT, or to problems with the absolute flux calibration. It is expected that the in-orbit calibration solves many of these problems, but it is also inevitable that some differences will remain. It has become clear that especially the LW1 detector does not line up correctly with the other detectors when a faint source (like a galaxy) is observed. In that case the absolute flux level of the LW1 detector should not be trusted (the relative response calibration, i.e. within the detector, should be OK). Currently this effect is under investigation. An indication for a way to fit the spectra back together is given in section 10.6.
There is a known problem with the responsivity drift correction. In addition, in a few cases the drift correction does not work very well. The drift correction is only performed for L01 and L03 observations, so L02 and L04 observations are not affected by these problems.
The known problem with the implementation of the drift correction can causes up to of the dark current/straylight to be added back into the data. This will therefore only be a problem when the source flux is of the same order as the dark current/straylight.
If you are in any doubt about the how well the drift correction has been performed on your data you are advised to compare the results in the LSAN file with the results in the LSNR file. The LSNR data does not have either the absolute responsivity correction, or the responsivity drift correction applied. If required, the absolute responsivity correction can be applied to the data using the procedure described in the `Post processing' chapter.
Due to the uncertainty in the grating setting the removal of the grating spectral element from FP data (see section 6.4.6) can introduce artificial slopes in the spectrum. This is especially true for grating positions far from the rest position (about equal to the center wavelength of the detector band). This means that continuum flux and spectral shape of an FP spectrum need not necessarily reflect the true source flux and spectral shape. In general it is not recommended to use the FP spectra to determine continuum shapes and fluxes.
It is known that, especially for faint sources, the darkcurrent subtraction as done in the automatic pipeline can be wrong. For many faint sources, with flux levels close to the darkcurrent, the flux will be negative after darkcurrent subtraction. In this case the observer is advised to use the data in the LSNR file, or use the LWS Interactive Analysis (see section 9.1) to redo the processing with a fixed darkcurrent that was determined using a special calibration observation. This might give better results. The darkcurrent levels that were measured using different methods are given in table 7.3. The dark currents given in this table were measured with a special long dark current observation in revolution 650, were determined from the illuminator operations done at apogee and were determined from serendipity mode data.