The Semester 00B Call for Proposals refers often to the preliminary nature of some of the numerical descriptors of the SCUBA system since the October 1999 upgrades . This document attempts to explain not only the nature of these numerical uncertainties, but also the new understandings we have gained about SCUBA recently. In parallel, we explain what we feel should now be standard photometry reduction techniques for data taken with the new wideband filters. It is clear from the improved understanding of skydips that older photometry would benefit from repeated reduction with the new formulae. We apologize in advance to those of you for whom this may mean considerable work. Any similar new insights into, or suggested reduction techniques for, jiggle-map and scan-map data may appear here later.
The Cookbook describes the necessary sequence of commands (reduce_switch, flatfield, extinction, resmky, scuphot, scucat). The items that concern us here are the extinction values to be used in extinction, and the options available within scuphot.
The opacity (extinction) of the sky is determined from measurements of
the sky brightness at several (typically 10 equally spaced) airmasses -
a
skydip. These
data are fit by a model of the sky, one parameter
of which is the zenith opacity. (This assumes, of course, that the
sky is well-behaved - i.e. has structures compatible with those
assumed in the model). The measured brightness of the sky is calibrated
by measures of two instrumental 'loads' - one 'hot' (essentially
'ambient'), and one 'cold' (a partial view of a source of liquid
nitrogen). The values for these measured load temperatures are
then adjusted to match the bandwidth and central wavelengths of the
various filters.
It has been our (and your ?) experience that the 450 micron skydips
were seldom well fit by the model and, indeed, we generally favoured
calculating tau450 from tau850 using
empirical linear relationships , and discarding the 450 micron
skydip data themselves. Following the
upgrades
this situation initially persisted, but, following
careful measurement of the loads, it became clear that
previous assessments of these temperatures have been in
error, both for the older narrowband filters and the new wideband filters.
A substantial
reanalysis of skydips was undertaken and
the following load temperatures and telescope efficiencies are recommended :
The ostensibly optional parameter b should be allowed to float
within the algorithm by accepting the default value of -1.
SKYDIP data taken after 13 March 2000 will have the correct
values for Tcold in the file headers.
Similar upgrading to support filter-dependent values for Thot
is a trickier task and is pending.
Comparison of SKYDIP results using these correct load temperatures
with the
CSO measurement of
opacity at 225GHz , taucso, have
yielded new relations :
We are indebted to Jeff Wagg, a co-op student from UBC, Victoria,
Canada, for these analyses, and to Ed Chappin, also of UBC, and
Mark Amure (Univ. Cardiff) for similar work during their extended
stays at the JAC. Jeff's work includes fitting polynomials to the
CSO data
for each night. This characterizes the major
temporal variations of opacity and marginalizes the occasional spurion
within the
CSO records .
He found that the scatter in tau850w -vs- taucso
during any night is improved considerably using this technique over,
say, linear interpolation between
CSO records .
The new relationships
will yield maximum benefit in post-observing analysis.
A full description of how these relationships were derived is reported
here. It includes the relations for
the narrowband filters, describing how they differ from previous estimates and
the likely consequences for data reduction.
The
SCUPHOT documentation says it (scuphot) " . . takes the
extinction-corrected
data and averages the nine points together". Well, this is one
way of doing it, known as AVERAGES, while SAMPLES and PARABOLA are the
other two options. The nine points in question are the 9 positions of the
jiggle map, forming a square pattern of grid size 2 arcseconds. Photometry
is taken in this way rather than in a STARE mode because the LONG and
SHORT arrays are not perfectly coaxial; being offset by 1.00 arcseconds in
the Nasmyth-X- and 0.25 arcseconds in the Nasmyth-Y- directions. The
little jiggle pattern is an attempt to ensure that the peak signal from
the source is captured by the central bolometers of both arrays at some
stage in the jiggle, following an assumed POINTING at 850 microns.
While AVERAGES first generates the average value of the 9
measurements of each integration and then produces the mean and standard
deviation of these N integrations, the SAMPLES option operates on all
9*N individual measurements and provides better accounting of errors.
While in practice the mean results of AVERAGES and SAMPLES are identical,
the associated error is erroneously smaller in the AVERAGES
case, and increasingly so for the smaller number of integrations used
typically for bright calibrators.
The statistical veracity of the SAMPLES method makes it
preferable to AVERAGES.
The PARABOLA option fits a 2-D parabola to the 9 measurements in each of
N integrations and provides a peak value for each integration.
The subsequent analysis then does AVERAGES on the N peak values.
PARABOLA really only works with a strong signal, such as from calibrators.
For 850 (450) microns, where the beams are gaussians
with HPBWs of 14 (8) arcseconds, the peak signal derived from PARABOLA
should be 1.075 (1.25) times that derived from SAMPLES. If pointing is
always done at 850 microns, the 1.03 arcseconds pointing difference between
the arrays increases the 450 micron factor to 1.30. These theoretical
factors are substantially confirmed in practice although we know that the
beams are not true gaussians, are likely to change as the dish
shape changes and so must be confirmed on a nightly basis if these factors
are to be used. Additionally, the planets and some of our calibrators
are not point sources and will deviate from these ideal circumstances.
As a matter of principle, rigourous analysis of photometry of both bright
calibrators and fainter targets should use the same SCUPHOT option.
However, the PARABOLA method is to be recommended for the brighter
calibrators since it offers the best chance of catching the peak flux
in the presence of pointing errors etc., while SAMPLES is preferred for
fainter targets. In this case the factors above must be used to
convert FCFs from PARABOLA to FCFs for SAMPLES.
Data taken carefully with the wideband filters
(450w:850w)
since the
upgrades were analysed for Flux Conversion factors (FCFs) and
Noise Equivalent Flux Densities (NEFDs).
The former are derived
from calibrator data and we distinguish between the SAMPLES and PARABOLA
methods as described above, and the latter are derived from long
integrations (n_integrations of 50 or more) on fainter targets analysed
using SAMPLES. The FCF results are as follows :
The empirical ratios of FCFs (in the sense PARABOLA : SAMPLES) are
1.077 and 1.29, which superbly confirm the
theoretical expectations .
The NEFDs - see
plot - were calculated initially in volts using the SAMPLES method.
The appropriate FCFs with which to convert these volts to Janskys are then
the PARABOLA FCFs, derived from the calibrators observed that same night,
corrected by the factors above.
After that date SKYDIP file header information, particularly the
filter-dependent, measured values of Thot, will be correct.
(28 June 2000 : SURF v.1.6 and above should handle
Tcold and Thot correctly.)
(Jy/volt)
(Jy/volt)
Watch for temperature dependence.
No apparent variation with
tau850
or
airmass
or
UT
or
seeing (when seeing <1")
No apparent variation with
tau450
or
airmass
.
Highly temperature dependent
* - Strong monotonic decrease with
UT
Iain M. Coulson, Elese N. Archibald
