Cosmic Origins Spectrograph Instrument Handbook for Cycle 17

7.6 FUV Dispersed-Light Acquisitions

---

COS includes flight software that can find and center a source in the selected aperture by working with the dispersed spectrum. This can be done with either the FUV or NUV detector. A dispersed-light acquisition has the advantage of analyzing the same image that will then be integrated to form the science spectrum. However, there are some disadvantages to acquiring in dispersed light:

• Instead of obtaining only a single image that is then analyzed to determine a centroid (as in ACQ/IMAGE mode), in dispersed light the telescope is moved a number of times to create a spiral search pattern on the sky, and the accumulated counts are then analyzed. At each dwell point a separate exposure is needed, and then HST must be moved a small amount. Those exposures and motions are each fairly short, but they add up, resulting in a fairly slow acquisition.
• Mostly because of lower S/N, the initial dispersed-light acquisition achieves pointing precision of about 0.1 arcsec, which is not as good as ACQ/IMAGE. This can be improved significantly with ACQ/PEAKXD and ACQ/PEAKD.
• Airglow (or geocoronal) emission features fill the aperture, and at those wavelengths they can produce high count rates that make source detection difficult. This problem is most severe in the FUV and is averted by ignoring portions of the detector illuminated by airglow features. This is done by using sub-arrays on the detector, and this is carried out automatically by the flight software. Thus in practice airglow lines should not impede acquisitions.

7.6.1 FUV Dispersed-light Acquisition Summary

Airglow lines and sub-arrays

Nearly all the strong airglow lines are in the FUV (see a list of lines and strengths in Table 10.3). Of these, Lyman- is by far the most important. To avoid the airglow lines, the dispersed-light acquisition process reads discrete sub-arrays on the XDL detector. In addition, segment B, which records the shortest wavelengths, gets very little light when grating G140L is used, and therefore only segment A is used for an acquisition with G140L.

Steps in a Dispersed-Light Acquisition

There are three steps needed to center a target with a dispersed-light acquisition:

1. A search (using ACQ/SEARCH) is carried out, in a spiral pattern, making a square with 2, 3, 4, or 5 points on a side. At each scan point the telescope stops and an integration is taken. After completion of the full n × n pattern, the data are analyzed and the telescope is moved to center the object.
2. A peak-up in the cross dispersion direction is performed to improve the centering (PEAKXD).
3. A peak-up in the along-dispersion direction is done as well (PEAKD).

The last two steps are optional and should be done in the order indicated (PEAKXD then PEAKD). Also, any one step may be done more than once (such as doing a 3 × 3 spiral search followed by a 2 × 2 one to improve the centering). As a result, there is a huge number of possible ways to acquire a target and improve its centering. Here we will concentrate on some specific scenarios that achieve good results in a reasonable time.

7.6.2 Mode=ACQ/SEARCH: The Spiral Target Search

The initial target search is done with the ACQ/SEARCH command for the COS/FUV configuration. In ACQ/SEARCH mode you can command the COS aperture to be moved in a spiral pattern to cover a square grid up to 5 × 5 steps in size. With a STEP-SIZE of 1.767 arcsec (the default and recommended value) and a 3 × 3 search pattern, your target will be found if it is within about 3 arcsec of the initial pointing.

You will need to specify:

• The aperture to use, either PSA or BOA.
• The spectrum element (i.e., which grating) to be used and the wavelength setting. In general this will be the same as the grating and wavelength to be used for the science spectrum that follows. However, an observer may acquire with a different grating + wavelength combination than the one to be used for the science spectrum, and there may be advantages to doing so.
• The SCAN-SIZE, which is 2, 3, 4, or 5, corresponding to spiral patterns of 2 × 2, 3 × 3, etc.
• The exposure time per dwell point.

To calculate the total time needed for an FUV ACQ/SEARCH:

1. Add 20 sec to the exposure time to be used at each dwell point.
2. Multiply this value by the number of dwell points.
3. Add any overheads from Table 9.2 that apply to putting in place the spectral element you will use. Note that the "home" position for OSM1 leaves grating G130M in position. This means that G130M is in position by default at the start of a new visit.

Large SCAN-SIZE values should only be used in cases where the target coordinates are mediocre, which should occur only rarely. A 3 × 3 pattern should be adequate in virtually all cases. Note that the even SCAN-SIZE values (2 or 4) entail some additional overhead time because there is an additional movement of the telescope needed to displace the aperture by half of STEP-SIZE in both x and y (the coordinate system at the aperture). This is so the overall pattern remains centered on the initial pointing.

Figure 7.8: Example of a 3 × 3 Spiral Search Pattern.


 
This example was executed with the default STEP-SIZE of 1.767 arcsec. The blue circles represent the nine positions of the aperture, each 2.5 arcsec in diameter, and the numbers show the sequence of steps. The large outer circle in red has a radius of 3 arcsec. Thus an initial pointing that was good to 1 arcsec (1) would result in a successful acquisition with a 3 × 3 pattern 99.5% of the time.
 

The STEP-SIZE parameter determines the spacing, in arcsec, between dwell points in the pattern. It may be set at any value from 0.2 to 2.0 arcsec, but we strongly recommend using the default value of 1.767 arcsec. This default value has been chosen so that no part of the sky is missed, given the 2.5 arcsec diameter aperture (2.5/2 = 1.767).

Finding the Source

Once the integrations are all done, the flight software determines what point in the array to return to, and there are three options. The default, and recommended, option is CENTER=FLUX-WT. This algorithm uses a flux-weighted centroiding procedure to determine the center of the light and has been shown in simulations to be effective in locating a source. The algorithm contains a check that removes dwell points from the calculation if the number of counts at that point is below a certain percentage of the maximum counts seen in the brightest dwell point. That threshold ("LOCAL-THRESHOLD") is set at 10% and is not selectable by the observer. The optimum threshold value will be checked during characterization of COS after launch.

A variation on FLUX-WT is to use CENTER=FLUX-WT-FLR. In this case a floor is subtracted from all the array's data points before the centroid is computed, and that floor is taken as the minimum number of counts seen in any one dwell point. FLUX-WT-FLR has the advantage of getting rid of background counts, but leaves one point in the array with zero. This can cause computational problems, and, as a result, FLUX-WT-FLR may not be used with SCAN-SIZE=2.

The last option for centering is to use CENTER=BRIGHTEST which simply centers the dwell point with the most counts. This is straightforward but not as accurate as the centroiding methods, although it may be appropriate for some situations that involve structured objects in which one wishes to center, say, a bright knot.

Exposure Times

Figure 7.9 allows you to estimate the exposure time needed for an FUV acquisition in dispersed light. The COS acquisition ETC should be used to get actual values, of course. Note that these exposure times apply to each separate dwell point of a pattern, which is the quantity entered into APT in Phase II.

Figure 7.9: Exposure Times Needed for FUV Dispersed-light Acquisitions.


 
The calculations have been made for a flat source spectrum and are based on achieving
S/N = 40.
 

Quality of Centering After ACQ/SEARCH

The ability of COS to center objects accurately will be calibrated once it is installed in HST, and so this discussion relies on computational simulations. Those simulations, using realistic estimates of source brightness, coordinate accuracy, and noise levels, predict that the ACQ/SEARCH stage, by itself, together with CENTER=FLUX-WT should lead to a source being centered to within 0.2 arcsec in the along-dispersion direction and 0.1 arcsec in the cross-dispersion direction. However, statistical effects play a role, and the worst-case errors were 1.3 arcsec. If CENTER=BRIGHTEST is used instead, simulations show that the centering can often be off by 0.4 arcsec or more. FLUX-WT-FLR also produced good results, but not as good as FLUX-WT.

7.6.3 PEAKXD: Peaking up in the Cross-dispersion Direction

As noted, in most cases an ACQ/SEARCH by itself will center a source well in the cross-dispersion direction, generally well enough for most purposes. However, an additional command, ACQ/PEAKXD, exists to enable that centering to be improved.

ACQ/PEAKXD works very much like ACQ/SEARCH mode except that no movement of the telescope occurs. As with an ACQ/SEARCH, with PEAKXD you specify the aperture to use (PSA or BOA, the same as for your science exposure, in general); the grating and central wavelength, and the exposure time. You can optionally choose to just use one of the segments, A or B, but use of the default is recommended. The default uses both segments except that only segment A is used with G140L set at 1105 Å. The specific steps executed in ACQ/PEAKXD are:

• A spectrum is recorded in TIME-TAG mode for a time and using a sub-array tailored to each grating setting.
• The mean location of the spectrum in the cross-dispersion direction is computed.
• This mean is compared to a similar calculation for a short exposure of the wavelength calibration lamp, and a shift is then computed to apply to the target spectrum to center it in the default location.
• The telescope is then slewed by this offset to center the target.

Simulations show that use of PEAKXD should end up centering a source to within 0.03 to 0.04 arcsec in almost all cases.

The total duration of an FUV ACQ/PEAKXD is 80 sec plus the exposure time plus any overhead time for an OSM movement from Table 9.2.

7.6.4 PEAKD: Peaking up in the Along-dispersion Direction

A COS point-source spectrum as imaged onto the FUV detector has some aberrations, but is still basically a single line along the direction of dispersion. This makes the determination of the spectrum's center in the cross-dispersion direction straightforward, but centering the source in the aperture in the along-dispersion direction using the dispersed spectrum is not as easy. At the same time, as noted above, the centering in the along-dispersion direction is more important for the quality of the spectrum because it helps assure the wavelength zero point.

The ACQ/PEAKD command works very much like ACQ/SEARCH except that instead of a spiral, a linear motion of HST is made to integrate the spectrum. As with ACQ/SEARCH, the centroid is then computed. The number of steps may be chosen as 3, 5, 7, or 9, with 3 being the default. The STEP-SIZE can be 0.01 to 2.0 arcsec, and there is no default value, although 1.2 arcsec is recommended.

As with ACQ/SEARCH, there are three options for the centering algorithm, CENTER=FLUX-WT, =FLUX-WT-FLR, and =BRIGHTEST, and they work in the same way as described above. We recommend that you specify CENTER=DEFAULT, which uses FLUX-WT if NUM-POS=3, but uses FLUX-WT-FLR if NUMPOS=5, 7 or 9.

The duration of an FUV ACQ/PEAKD is the number of dwell points times the sum of the exposure at each dwell point plus 20 sec. OSM1 movements overheads (Table 9.3) must be added as well.