The noise estimates given in Table IX.B.1 are based on baseline readings, and refer to a single spectral resolution element. Estimates for individual spectra deviate from the mean value by typically 20%.
|Detector Number||Response to Internal Reference (mV)||Noise level (mV)||NEFD (Jy)||Wavelengths (µm)|
|1||11.9||0.10||1.4||8 - 13|
|2||4.4||0.04||1.6||8 - 13|
|3||7.4||0.06||1.3||8 - 13|
|4||22.8||0.12||3.0||11 - 22|
|5||46.3||0.20||2.5||11 - 22|
The zero point of the wavelength scale could be determined by the in-scan position of the source as determined by the survey instrument during the same scan. The uncertainty in this position caused a jitter in the zero point of up to 1 sample (corresponding to 7" or one-half of a spectral resolution element). This effect degraded the correction for wavelength-dependent gain and the averaging of individual spectra. A better estimate of the zero point was obtained using the four well-defined band edges of the spectrometer. A signal-to-noise ratio of 10 measured at one of the edges gave an accuracy of 0.3 sample (2").
Figure IX.B.1 Cross-scan responsivity for each of the five spectrometer
detectors. Detectors 1, 2 and 3 cover the 8-12 µm wavelength range,
and detectors 4 and 5 cover the 11-25 µm wavelenght range. Open
circles indicate the R UMi results; the curves were used for the
correction (Section IX.C.2.e).|
The cross-scan responsivity of the detectors was determined by finely spaced raster scans across the star R UMi (as shown in Fig. IX.B.1.) where the curves are normalized over the central half of the detectors. The sharp dip in response of detector 4 was also found in the laboratory tests. Although the curves are accurate within a few percent, the derived value of the cross-scan correction is much less accurate. Because the cross-scan position determined by the survey instrument during a scan had a typical uncertainty of 1'. Except for sources passing over the very center of a detector this introduced an uncertainty in the cross-scan correction factor of 10-20%.
In order to determine the factor for conversion of sample
values into flux densities, the observed spectra of a Tau were
compared with a black body spectrum of 10,000 K. The resultant
responsivity curves (Fig. IX.B.2) show the
ratios of flux density
to sample value as a function of wavelength. The curves have
been normalized in the overlap region from 11 to 12 µm and are
and to 2% at the shortest and to 4% at the longest wavelengths.
No significant differences were found between detectors 1, 2
and 3, or between detectors 4 and 5. All detectors had somewhat
better responsivity in the center of their wavelength range than
at the edges. Except near the very edges of the band, the responsivity
correction did not exceed 30%. The feature seen between 9 and
12 µm in the short-wavelength band is a characteristic of the
instrument that is found in all raw spectra and was already known
from laboratory tests.
Figure IX.B.2 Wavelength dependent responsivity for the 8-13
and the 11-25 µm detectors, respectively, plotted on the standard
samples (Section IX.C.2.c). The responsivities are normalized
in the 11-12 µm region.|
Occasionally, the data streams of the detectors were scrambled by the multiplexer. In most cases this was evident in the spectra as single-sample spikes. The larger ones were easily recognized and removed (see Section IX.C). In some cases a series of samples was scrambled and the spectrum could not be repaired without running the risk of erasing real features.
Since the slitless spectrometer had a field-of-view of 5' by 6' (short wavelengths) or 7.5' by 6' (long wavelengths), regions that were confused in the short wavelength survey bands were also confused for the spectrometer. A spectrum could be confused by a nearby point source, or by extended structure associated with the source itself or in the background. In most cases such confusion was recognizable as a difference in the baseline level between the two sides of each spectrum-half. However, as discussed below, if the test of baseline asymmetry was made too restrictive, otherwise good spectra were lost.
No evidence was found for photon induced responsivity enhancement (Section IV.A.8 and VI.B.4c), although it would have been hard to discover. In the first place, most spectrometer sources were stars or star-like sources with flux densities gradually decreasing towards longer wavelengths. Since the spectra were scanned starting at the long wavelengths, and thus at the lowest flux density, the deformation of the spectra was minimized. Secondly, close to the Galactic plane, where the effects were greatest there were very few clean, isolated point sources. Confusion probably masked any effects of photon induced enhancement present.
Changes in the detector responsivity on a time scale of minutes due to exposure to a bright source or to a prolonged exposure to a medium bright source like the Galactic plane were found to occur. The latter effect was compensated for to within 5% by using interpolated internal reference flash responses to correct the flux densities.
Memory effects did not cause differences in response of the system between spectral lines and continuous spectra. The planetary nebula NGC 6543 was scanned at the nominal survey rate and at ½ , ¼ , and 1/8 of that speed. The line intensities did not change noticeably as a function of the different scan speeds.
It is expected that any non-linearity effects were small, although no comprehensive linearity tests were made. This statement is based on the agreement between two different determinations of the wavelength dependent responsivity curves. One set of curves was determined using -Tau, a star with a spectrum that rises steeply towards shorter wavelengths. The second set of curves was determined using asteroids -- cool objects which are brightest at the longer wavelengths. The differences between the two sets were small, and the comparison was limited by the accuracy with which the asteroid curves could be determined.
After processing a large sample of spectra with the correction techniques described in Section IX.C, the integrated fluxes in the spectra (after convolution with the survey pass-band) were compared with the fluxes measured by the survey array at 12 µm. Comparison with the 25 µm survey fluxes was not attempted because the spectrometer hardly overlaps the survey 25 µm pass-band. Systematic factors of 0.75 and 1.00, respectively, were applied to the data to make the 8-13 and 11-25 µm spectrometer fluxes were with the survey fluxes. On the average the spectrometer flux densities in the catalog are consistent with the survey flux densities to within 10-15%.