ISSA Explanatory Supplement
IV. ANALYSIS RESULTS
E. Noise Performance and Sensitivity
Chapter Contents
Table of Contents | Index | Previous Section | Next Section
- Cross-Scan vs. In-Scan Noise
- Noise Equivalent Surface Brightness in
ISSA
- Residual Zodiacal Emission
- Quality Estimates From
Scan-to-Scan Statistics
Six major sources contribute to the noise in the ISSA data.
Detector noise plus photon noise constitute random noise;
natural variations in the celestial background contribute
confusion noise; drifts in the calibration of the data produce
stripe noise; residual zodiacal emission contains gradients and
steps; and nonconfirming sources and radiation spikes introduce
spurious point sources. The effects of nonconfirming sources
and methods for eliminating them are discussed in
§I.D.4.
Confusion noise in the IRAS data is discussed in Gautier
et al., (1992). The remaining noise sources were measured and
analyzed as described below to give the user of ISSA an idea of
the sensitivity limits of the ISSA data and of the kinds of
errors to expect in the data.
The remaining calibration or stripe noise falls into two
spatial domains. Variations over several degrees in the scan
direction are discussed in
§IV.D.2 in terms of large- and
medium-scale baseline distortions. Calibration imperfections
produce scan-to-scan and detector-to-detector variations
in the images. The RMS stripe noise is measured by examining the
variations perpendicular to the scan direction.
The performance of the ISSA destripers in reducing this noise is
detailed in the section below on cross-scan vs. in-scan noise.
Random variations due to electronic noise and photon noise set
the noise floor and determine the ultimate sensitivity of ISSA as
described in the discussion of noise-equivalent surface
brightness density (NESB) and dimmest detectable sources.
Finally the magnitude and character of the residual zodiacal
emission is discussed.
One of the performance goals of the ISSA destriping
procedure was to reduce the cross-scan noise to the same level as
the in-scan noise. This goal was substantially achieved.
Table IV.E.1 shows typical values for a coadded
image of the
RMS variation along a ~1° cut taken
in the cross-scan and in-scan
directions. These cuts were confined to flat, low signal regions
within each image.
Values for the
individual HCONs are about 1.6 times higher. There remains a difference in
the spatial power spectrum of the noise in the two directions.
The in-scan noise spectrum is characteristic of the noise
spectrum of an individual IRAS detector. This spectrum is
characterized by a power law with a spectral index near -0.75 and
contains little power at frequencies near the resolution limit of
ISSA. In contrast, the cross-scan spectrum contains substantial
power at frequencies up to the free spectral range of the ISSA
data at (3')-1, because the cross-scan noise is caused by
variations between adjacent detectors whose noises are uncorrelated.
This difference in spectral distribution leaves
stripe-like features in the residual noise, because the period of
the noise variation is much longer in the in-scan direction than
in the cross-scan direction. The RMS variation over a
few degrees is nearly the same in the two directions, however.
Noise equivalent surface brightness (NESB), actually brightness density
here, is conveniently expressed in units of Jy sr-1sr-0.5.
Then, for instance, the expected minimum detectable surface brightness for
an object of size Omega sr can be calculated as
NESB× sqrt(Omega). NESBs for the ISSA images can be estimated
from Table IV.E.1 assuming that the appropriate
solid angle is that of the 90%
encircled energy contour of the ISSA point spread functions (about 2.4×
10^{-6}sr). This calculation yields 51, 68, 56 and 97
Jy sr-1sr-0.5 for the 12, 25, 60 and 100 µm bands,
respectively. In areas similar to those where the data for
Table IV.E.1 was
taken, the dimmest discernible features with size about 0.5° have
surface brightness above the background of about 0.02 MJy sr-1
at 12, 25 and
60 µm and about 0.07 MJy sr-1 at 100 µm. Assuming that "dimmest
discernible' means about a 3 sigma detection these dimmest surface
brightnesses are consistent with the estimates above except at 100 µm
where the higher general cirrus brightness makes selection of features as
dim as 3 sigma more difficult.
The estimates of NESB based on Table IV.E.1 are in
agreement with estimates
based on the average IRAS detector NEFDs shown in
Figure IV.A.1 of the Main Supplement.
Residual zodiacal emission causes gradients and sharp
discontinuities in the ISSA images. Discontinuities can occur
when adjacent regions of sky were observed at very different
zodiacal brightnesses. These discontinuities are small, as seen
in Table IV.E.2, but are easily identified because their boundaries
are very sharp and align in the scan direction. Residual zodiacal
gradients are more subtle and can be harder to detect. The
residual gradients in the high-ecliptic-latitude ISSA data are
most apparent near the ecliptic poles in the 12 and 25 µm
bands. The magnitude of the residual emission is largest compared
to other celestial emission at the shorter wavelengths. The
spatial scale of variation is small near the poles due to the
combination of scanning geometry and modeling errors in the
variation of polar brightness with the motion of the Earth in its
orbit. Measurements of some prominent residual zodiacal gradients
are given in Table IV.E.3.
Noise and variability statistics were kept for each pixel
during the ISSA map generation process. These noise maps were
originally intended for use with a confirmation algorithm that
failed because the variation of data within a pixel was
extremely non-Gaussian. This was presumably caused by zero-point variation
from scan to scan due to systematic errors in the zodiacal
emission model. As a result, the pixel statistics
did not reveal much about the actual noise levels in the ISSA data.
Chapter Contents
Table of Contents | Index | Previous Section | Next Section