LWS Data Frequently Asked Questions
Q. I tried running ISAP on an LWS Fabry-Perot
observation and I get detector buttons for only detectors 6,7,8,9.
A. The problem has nothing to do with ISAP, IDL
or the computer; it has to do with the fact that the observation is an
L03/L04 AOT (FP). If a specific wavelength range or line was required with
LWS which falls on one of the ten detectors, the instrument will also give
you for free the wavelength ranges that are simoultaneously falling on
the other nine detectors. With the FP it often happens that these extra
bits of spectra can not be calibrated. In this case the Pipeline leaves
those bits (and hence those detectors) out of the product file (the LSAN
Q. I would like to know some things about the
origin of the uncertainty in the integrated line flux as given by ISAP.
A. The integrated line flux uncertainty is propagated
from the uncertainties in 2 free parameters of the fit: height and width.
Those two uncertainties come from the diagonal element of the regression
matrix multiplied by the rms from the baseline fit (if available from 10
or more baseline points) or the residual of the gaussian fit itself. Bevington
has a discription of this calculation.
Q. How do you get pointing information out (ie.
is there a program written, or do I need to do it the hard way) ?
A. the INSPECT_RASTER routine is suited for this
purpose. Check out the documentation for this routine in the LIA www page
Q. I'd like to know the aperture for the various
LWS observations, do I just go back to the manuals or can I get it from
the header ?
A. The LWS aperture is in principle fixed. In
the paper by Swinyard et al. in the A&A special issue an aperture of
80" is quoted. Infact, it is found that the aperture (the beam-size actually)
varies with wavelength. Check out this www page for the most recent estimates
of the LWS aperture for the 10 detectors: http://isowww.estec.esa.nl/instr/LWS/note/beam_mem.html
Q. How do I adjust L04 (FP) data to account for
A. It is not possible to correct L04 data for
fringes; you could do something if you also have a full grating scan for
that source (L01 AOT). In that case you defringe the L01 observation and
use the fitted fringe to correct the L04 spectrum.
Q. Where are the LWS resolutions listed? I've
looked in the handouts from the LWS IPAC workshop, but can't seem to find
A. what it should be listed are the spectral element
sizes. The original estimates are 0.29um for the first 5 detectors and
0.6um for the last 5 detectors.
Q. I would like to know whether ISAP is available
for an NT machine.
A. some people are successfully using ISAP on
a PC but only under Linux. We have no reports of people using it under
Q. How believable is the continuum level in our
L04 data for weak/strong lines? Although it is not generally recommended
to use FP spectra to determine continuum fluxes and shapes, are there cases
where the continuum can be determined from FP spectra with high s/n?
A. The transparency of the FP is so low that the
signal from the source (unless very bright) is comparable to the dark current;
hence a small uncertainty in the determination of the dark current is enormously
amplified in the calibrated spectrum. On a source like Orion BN (100,000
Jy at 100um) the estimated continuum was within 30% of the values observed
from KAO. On a fainter source, the continuum determination has essentially
an unpredictable uncertainty.
Q. What cases justify the need for grating adjustments
in L04 data? The documentation in fp_proc suggest that the only criteria
for L04 data is to flatten out the continuum in the spectra. So should
a grating adjustment be made for all non-flat spectra or just for certain
A. For all non-flat spectra.
Q. Are there some general guidelines about the
maximum safe grating position adjustment one can make in FP_PROC ?
A. No guidelines; within the shown range, we do
not know what the exact position of the grating was during the observation.
Q. In some of our lines observed with the LWS
FP (AOT L04), we are noticing some sag (a bowl shaped depression) in the
middle of our spectra. What can we attribute this to ? Adjusting the grating
position with FP_PROC doesn't seem to help. Section 7.4 of the LWS IDUM
manual mentions that non-linear detector behavior can cause 'sags' in the
shape of the spectra. Can this also be causing this depression in the FP
data? And how can we correct for this?
A. Non-linearity in LWS is something that arises
with very bright objects in grating mode. With the FP the flux incident
on the detectors is generally so low that saturation (hence non-linearity)
never arises. The reason for the sag may be due to an incorrect dark current
estimate. Reprocessing with FP_PROC is advised. However, this is not very
important for the determination of a line flux since this is not influenced
by an incorrect dark current subtraction.
Q. In defringing, the cookbook now says to only
defringe LW detectors. For a few of our sources, there are what certainly
look like fringes in some of the SW detectors (usually SW2 and 3, once
SW5). In previous versions of ISAP, trying to defringe these detectors
failed, but in ISAP 1.6a, they cleaned up nicely for at least some sources.
Why don't you recommend defringing SW detectors? Since we get an improvement
on the quality of some SW data, are you sure we shouldn't defringe?
A. We do not recommend defringing SW data because
they are more difficult to recognize in the those detectors. This is because
the frequency of the fringes (in wavelength space) increases with decreasing
wavelength and a series of bumps in the spectrum can easily be interpreted
as a fringe and corrected accordingly. Only attempt defringing if the the
period of the fringe is of the order of 3.4 (in wavenumber space).
Q. I noticed that you no longer recommend shifting
the data to align the detectors (rather emphatically, too!). Is this likely
to be the final recommendation? For most of our spectra, the offsets are
less than 15%, but a few have offsets greater than 30%, usually between
SW1 and 2 or LW 4 and 5. (I will try LIA, but it didn't make much difference
for the source I tried it on already.)
A. Moving the detectors spectra to align them
to one another is in principle an arbitrary operation. Unless you know
the cause of the misalignment (e.g. incorrect DC subtraction), you do not
know the type of shift to apply (offset or gain ?). Besides, what's a good
detector to be used as a reference one ? We cannot give a GENERAL recommendation
in this respect. Besides, we realized that a dominant cause for the misalignment
of the detectors is due to the fact that the transmission profiles we use
to calibrate the data have been derived on a point-like source. If your
source is extended, your spectrum will be incorrect to an extent that depends
on the extension and the position of emitting areas in the beam (which,
by the way is not yet fully characterized).
Q. The OI and CII line profiles with LWS show
very broad wings which look like an underlying broad component. As far
as I can see there is no physical reason for such a component, but I can't
pin down any processing artefact to be made responsible for it.
A. The problem has to do with the sampling of
the spectrum. The LWS detectors are known to be slow in response to an
increase in flux, in the sense that they rise slowly ; on the other hand
they are very good when tracing a decreasing flux: i.e. you can trust the
shape of the decreasing part of the line. This is a known problem for many
of LWS spectra which are done with fast scan and oversampled of a factor
4 but I see it really becomes critical when no oversampling is present.
If you plot the forward and backward scans separately, you will certainly
see that in the two groups of scans only one line wing at a time is apparent.
The easiest fix is to zap the wings and fit a line to what's left of the
line and baseline; fixing the linewidth to the instrumental spectral element
width might help a meaningful convergence of the fit.
Q. I asked for 4 lines observed with LWS AOT 04,
but when I open the LSAN file in ISAP I can only find three lines. Where
is the missing line ?
A. This is due to the LWS pipeline having problems
with the FP data processing. The LWS grating acts as an order sorter for
the FP; the degree of uncertainty in the positioning of the grating is
sometimes preventing the OLP from being able to determine the correct order
of the observed wavelength range. Since the number of combinations of wavelength
ranges and order is virtually limitless, only those combination which guarantee
the maximum sensitivity have been calibrated. As a result, if the wrong
order is picked-up for that grating position, the data cannot be calibrated
and are lost in the LSAN file. The fix to this, is to reprocess the data
using the LIA routine FP_PROC. This routine applies a different algorithm
with respect to the automatic pipeline, for the determination of the FP
order and the information is never lost.
Q. How it comes that when I average my LWS or
SWS data, some points have a 0 standard deviation attached ?
A. average will yield 0 standard deviations whenever
it has only one point in a delta_lambda bin. This often happens in LWS
data when you have a small number of scans. It can also happen with a high
number of scans, when you zapped many points at a particular wavelength.
In SWS it is more common at the edge of each band, where the band coverage
of the detectors was less than perfect.
Q. I can clearly see broad wings in the [CII]158um
line profile. Is there any instrumental effect that might cause this ?
A. It is possible that this result from detector
memory effects. The LWS detectors, especially the long wavelength ones,
correctly track decreasing fluxes, but tend to be slower to react to increasing
fluxes. Since a line is scanned in both directions by the grating, the
leading edge of the line will tend to exhibit a non-gaussian wing. If then
you average both scan directions together, you will have broad wings on
both sides. This effect is of course stronger on strong lines (on a relative
faint continuum), and it is illustrated in the figures below.
In the left panel the two colors represent the two grating scan direction
where is evident that the leading edge of the line
is broader (blue is scanned from right to left, green from left to
right). A straight average would yield the red profile in the right panel;
if instead, we zap the leading edge in both scan direction and then average,
the white line is obtained.