Topics of Historical Interest

Q. How Far Away Should I Place The Reference Position?
A. We recommend 5' or greater from the source.

Q. What are the Ways to Save on Overheads?
A. Fortunately, compared to the other instruments, the LWS overheads are fairly straight forward. Here is a list:

Acquisition Overhead - 180 seconds. If an AOT is not concatenated to the previous AOT, then the observer will be charged 180 seconds to acquire this position. If the source is concatenated to the previous AOT, then the observer should instead add 20 seconds for the acquisition of the position. To estimate the total (Target Dedicated) time of an AOT, the observer should add this additional time, i.e., 180 seconds, to the output of the time estimator. Even if the position of a new AOT is the same as the position of the previous AOT, the user must add in the 180 seconds of overhead if it is not concatenated to the previous AOT. The reason for this is that the new AOT may not actually follow the previous AOT in the satellite's schedule.

Note that the concatenation of one AOT to the next is a bit of a dicy subject. While it does save acquisition time for the observer, it represents an unwanted constraint to the satellite schedulers at ESTEC. Therefore, the policy of ESTEC, roughly speaking, is that concatenations of any type, (e.g., LWS02 to LWS02, LWS01 to LWS02, SWS02 to LWS02) require scientific justification to be accepted, not just the "it saves time" reason. All the PGA-MDB files have had all their concatenations broken and all concatenations must be reestablished. Some proposal's concatenations were disallowed in the individual technical assessments of the proposals. (For US observers, these assessments are available from IPAC.) Making new concatenations, not originaly part of one's Phase 1 proposal, requires a scientific justifications, explained at IPAC to the support scientists and writing into the "exit survey" regarding your observation.

What are valid reasons? If the science requires you that acquire all of several map positions on a source for a conclusion, then the points may be concatenated. If the science requires that you have the output of several instruments at one position to make your conclusions, e.g. an SWS and LWS line ratio, they may be concatenated. But you should be aware that the more concatenations you require, the harder it is to schedule your observation! For example, you will be less likely to win in a competition to get data from an oversubscribed location if your many positions in this region are concatenated. Likewise, there may be some orbits where just one instrument will be used (this remains a possibility). Concatenated AOTs from different instruments would be disallowed from these orbits.

Mapping Overheads - user calculates. The lws-te does compute SLEW TIME in its time estmates [new revelation, 30 March]. The user does not need add any further overheads. It has been asked if time can be saved on instrument overheads by slewing and taking care of instument set-up at the same time, either when acquiring a source or when making a raster map. The answer is "No." Essentially, nothing happens with any instrument when you're slewing to a target so one can't save time by setting-up an instrument while slewing. Within a raster map the instruments basically do not reset, and therefore require minimal overhead each time they hit a point in the raster. In fact, the instruments are continually "observing," and they simply synchronize the slewing with start of data taking.

LWS Overheads - there are overheads associated with all LWS observations. These, very roughly speaking, are about 6 minutes, in the sense that one cannot do anything with any AOT in less than this time. The observer is encouraged to use the time estimator to explore this issue. (Your IPAC computer account will be made available ahead of your visit with access to all time estimators if you request this from The efficiency improves when you do multiple lines with the LWS02 and LWS04 AOTs, and when you do faint sources in general. The efficiency improves dramatically when you use raster maps. Maps, however carry the constraint that all positions are observed with the same instrument procedure, and so the user can not optimize the observation as per the anticipated source flux.

Strategies for Reducing Overheads

As mentioned above, there is significant overhead in making a single observation, so the strategy should be to get everything one can out of it. A broad LWS01 range scan can be somewhat non-optimal in that it per force uses the same integration time at each grating position (wavelength). Since the detector sensitivities increase greatly from the short to the long wavelengths, one can reach quite different sensitivities at the two ends of the scan (e.g., unnecessarily high sensitivities may be reached at the long wavelengths in a full wavelength scan). If one is fishing for faint lines anywhere in an LWS01 range, one might break the range into segments to achieve more uniform sensitivities across the range (trading off against the additional overheads of doing two AOTs). One nice attribute of the time estimator, is that it will tell you the achieved (as well as the requested) signal to noise.

If one has requested an LWS01 range scan, one might reconsider identifying the lines that are of interest inside the range and doing them in the LWS02 line AOT, spending, within a single AOT, the optimal time on each line. This is allowed if no new science (wavelengths and very different sensitivities) are reached, with permission from ESTEC (helpdesk).

If one has an observing strategy where one repeats an observation at an off position, one might consider doing the on and off measurement as a two position map. This is a little risky however because LWS beamshape is not yet well known.

If one has a bright source (>10 Jy) that one is mapping, then using a raster map within a single AOT can save a great amount of time over individual AOT pointings, in all AOTs, since the observations are overhead dominated and using the same instrument parameters for all positions, (those that pertain to the faintest position) will not incur very much additional times, while losing the overhead for each position will save much time.

Q. How to compute the number of data points in LWS02 and LWS04 lines ?
A. The scan width deterimines the number of resolution elements to observe on either side of the central resolution element in LWS02 or LWS04 AOT mode. When Fast Photometry is desired from LWS02, the width can be set to 0. Zero should be selected if only one continuum setting per detector is desired in the continuum.

If the Photometry mode is not the selected mode, then the minimum number of spectral resolution elements on either side of the central element is 2 of 3, depending on whether LWS02 or LWS04 is being requested. This number usually forms a secure baseline, unless a significant deviation from a horizontal baseline is anticipated. In such cases, the scan width may be requested to be larger.

Note that it is not determined where the satellite will be, nor where the Earth will be when the measurement is conducted. The Earth motion relative to the Solar motion is about 30 km/s. This is why the minimum number of Scan Widths for LWS03 is three. In this way the Vearth and Viso with respect to Vhelio will not offset the line more than a resolution element.

This example will let you figure out how your samples will be arranged vs. wavelength and how many will be done for LWS02 and LWS04. Suppose the scan width is 3. There will be 3 resolution elements on either side of the central one:

        X   X   X   X   X   X   X   X

wavelength -->

If there are 4 samples per resolution element,  these are filled in like this:


wavelength -->

Finally, the LWS02 and LWS04 algorithm leaves the endpoints off:


wavelength -->

As concerns the sampling rate, if the Nyquist rate were desired (assuming the structure in the line spectrum were commensurate with the width of a resolution element), then the Sampling Interval would naturally be set to 2 to recover all information. Yet, in practice for LWS01 and LWS02, there is almost certainly structure in the line profiles that will never be resolved by the LWS. The suggestion is made to record four samples per resolution element because then there will be four (non-independent) samples of the line intensity from which to form a measuremet of the strength and perhaps the width. Were there only two samples per resolution element, and they differed, it would then be difficult to reach a consensus from the data.  Two samples per resolution element are suggested if and only if time savings are critical or if the line or continuum is not expected to vary appreciably across a resolution element.

Q. How the observing strategy changed during the mission ?
A. Here is the text of a memo sent out by the ISO Science Operation Center some time during the mission.

During the course of routine observing with LWS, some observing strategies have been identified which facilitate optimal reduction of the data.  These suggestions reflect the current LWS consortium practice and encompass all LWS observing modes; you may wish to consider incorporating them in an update to your observing program.  Updates will affect any observations in your program which have not yet been scheduled. Instructions on making updates can be found at the end of this note.

A brief summary of the recommendations is in the table below. A detailed discussion follows.

Recommendation            Applicable to                              Procedure/Comments
1) Make enough              All LWS Observations            Fast scan = "Y" and
   independent scans       (<10^4 Jys)                                Tint >= 2.4s ==> 6 scans.
   to de-glitch the data                                                          For >~10^4 Jy see point 3).

2) Measure fringing        All Obs. except                         Do a sufficiently wide
   so data can be                point sources                            and deep LWS01 range scan
   de-fringed                      on axis (w/in                             within or in addition to
                                              ~5")                                             the primary observation.

3) Avoid excessive         ESSENTIAL for Sources     Tint <- 0.7s.  Forces
   non-linearity                  >~10^4 Jy                                  1/4s integration ramps.

4) Use sufficient scan     F-P Obs                                     NSPEL=5. 2 more elements
   width to cover                                                                      will be added automatically
   wings of FP spectral                                                          NOTE that this means all
   elements (+/- 5 res.                                                           LWS04 observing times will
   elements)                                                                              increase.

5) Re-consider the         Mapping                                     Sample assuming a
   spacing of well-                                                                 SMALLER beam with
   sampled maps.                                                                    beam-width ~80"


In detail:

1)  Use at least 6 scans.

Previously, fast scanning was recommended to minimise the low-frequency end of 1/f noise. It is now even more important to avoid correlated contamination os successive spectral samples after a particle hit. The recommendation to use 6 (rather than the previous 4) scans is to ensure that sufficient uncontaminated samples are available at each spectral point to allow median filtering.  For 1/2s integration ramps, used for all but very bright sources (see point 3), this implies Tint per sample must be at least 2.4s.

2) Characterise fringing with a grating range scan.

Some examples of severe fringing are available on the WWW:
LWS spectra of off-axis point sources or of extended sources show a modulation of the spectrum referred to as fringing. In wave-number space, these fringes are cosinusoidal with a fixed phase and a constant period of 3.6 cm^-1, independent of the detector channel. It is therefore supposed that they arise from within the instrument and are caused by the interference of two beams arriving at the detectors along paths which have a fixed difference in length. In wavelength space, the interval between successive peaks varies from about 1 micron at the shortest wavelength to about 10 microns at the longest; at the shortest wavelengths this is comparable to the grating resolution element, complicating detection of lines. If the interpretation of the origin of the fringes is correct, their effects are present in both grating and Fabry-Perot spectra of extended or off-axis sources. Note that different parts of a source may have different degrees of fringing. For example, the spectrum arising from a point source will not itself be fringed, even if it lies within a field of extended emission. The resultant overall spectrum, however, will exhibit fringes.

If you are CERTAIN that your source is point-like, well isolated and its IR position is known to within 5", you may not need to take action. In most cases, however, you will want to de-fringe your data. Information about fringe removal can be found on the LWS Home page ( and the subject will be discussed in the Addendum to the LWS Observer's manual, which will be available around the end of July from the ISO WWW Site ( as part of the material for the Supplementary call for proposals.

In brief, the current procedure makes use of the constancy of the period of the fringe to identify and isolate it. It requires at least three complete fringes on each detector. Given present understanding, IT IS IMPOSSIBLE TO DE-FRINGE spectra without this information. The case of a point-like line source with strong extended continuum emission is particularly difficult, since the amplitude of the line will be unfringed, but the line will be superimposed on a fringed continuum. In this case there is no way of determining the true strength of the lines unless a spectrum of the surrounding continuum is also taken.

In detail, it is recommended:

A) For LWS01 observations, make sure that your range is sufficient  to cover at least three fringes per detector used. (Refer to the figure on the WWW or in the Addendum, when it is available). The observation must also have sufficient S/N on the fringe (SNc) to get the desired S/N on the de-fringed line (SNl). If Nss is the number of spectral samples you have in your LWS01 spectrum  (1,2,4 or 8 times the number of resolution elements;  65, for a full grating scan. See section 6.3 in the LWS Observer's  Manual) then, at a minimum: SNc=SNl/sqrt(Nss)

B) For LWS02,03,04 and narrow-range LWS01 consider an additional, preferably concatenated, LWS01 scan meeting the above criteria.  The observer should calculate the S/N required as detailed in A). You may wish to bear in mind that a full-range minimum scan (43 to 170 microns, fast scan, 1 sample per resolution element, 3 scans) takes about 8 minutes, and that the time does not decrease much if you decrease the spectral range.

C) Note that if you are doing an off-source measurement or map you will need to characterise the fringes in ALL POINTINGs (e.g., for a 2x2 raster you would need to repeat the raster with the minimum LWS01 scan.)  For the case of a point source embedded in extended emission, it is necessary to to do an off-source scan to remove the fringed spectrum of the extended emission.

3. For bright sources, use 1/4s ramps.

For sources  ~10^4 Jy at 120 microns, 1/4s ramps are necessary to prevent the detector integration ramps from becoming too non-linear to process successfully. Unfortunately, you must use Tint < 0.7s to get 1/4s ramps, so you can only get 5 scans per AOT; to get more you must concatenate 2 or more successive AOTs or observe the same line repeatedly in one AOT.  (In most cases 5 scans should be sufficient to deglitch the data.) For sources much brighter than 10^4 Jy, you should consider Fabry-Perot observations since even 1/4s ramps may become excessively nonlinear.

4. Use a wide scan for Fabry-Perot observations.

The Airy profile of a Fabry-Perot has broad wings and sufficient width around the line should be used to get down to the continuum. The maximum number of spectral elements to each side of the line as specified in PGA is 5, but the data suggest that 7 elements on either side is preferred, especially for resolved lines.  Up-link calibration tables are therefore being modified to ADD 2 SPECTRAL ELEMENTS TO NSPEL.  This will substantially increase observing time; if you wish, you may compensate for this by reducing NSPEL by 2, but use of NSPEL=5 is strongly  recommended.

5. Reduce step size for well-sampled maps

The LWS beam width is not yet well known, however the data suggest  that it varies non-monotonically with wavelength between 58" and 82". It is probable that 80" is a good working figure but it could be less  than this at some wavelengths.  Observers who are critically dependent on full sampling are advised to reconsider their step sizes.

To make an update to your program:

The procedure is the same as it was for releasing the proposals post-launch. You should contact the ISO helpdesk at (or IPAC at if you have been doing your PGA work there) and request a visit to PDEC (or IPAC), a computer account to make remote changes or give explicit instructions for changes to made on your behalf.  When the changes are completed you will be sent a PH report for confirmation -- this is necessary to verify that the correct copy of your program will be uplinked to ISO!

Most of these changes will unfortunately increase the time needed for an observation. We remind you that the Supplemental Call for Observations will be issued soon, and that you may propose to complete or follow-up on your existing observing programmes.