TECHNICAL
REPORT
Title: NIRSpec Dithering Strategy Part 2:
The Integral Field Unit (IFU)
Authors: J. Tumlinson
1.0
Phone: 410338-4553
Doc #:
JWST-STScI-001749, SM-12
Date:
April 30. 2009
Rev:
-
Release Date: 25 August 2009
Abstract
This study concerns need for dithering of NIRSpec data, the strategies for obtaining
dithered data, the effect of dithering on data calibration, and the effect of dithering on
mechanism lifetimes. For simplicity the study has been broken into three parts for each of
NIRSpec’s major observing modes (Fixed Slits, Part 1; IFU, Part 2; MSA, Part 3).
2.0
Introduction
NIRSpec data needs dithering for a number of important reasons. First, the PSF of the
JWST OTE is undersampled at the FPA, where each 18 µm pixel spans 0.1˝ on the sky
compared with a PSF FWHM >~ 0.1 – 0.2˝ over 1-5 µm. Dithering by small spatial
offsets (< 0.1˝) can improve the spatial sampling of the PSF. For many observations
dithering with a larger offset will be needed to cover the gap in wavelength caused by the
~ 3 mm (17.8”) gap between SCA segments. Dithering can increase the sky coverage of
observations efficiently and cover small gaps between detector segments. Finally,
dithering can reduce noise associated with pixel-to-pixel sensitivity variations by
obtaining measurements of the same point on the sky at different places on the detector,
averaging out the variations.
For NIRSpec, finding the optimal dithering strategy is complicated by the complexity of
operating the instrument, particularly the MSA. Also, the usage of the mechanisms for
positioning the filter wheel (FWA) and grating wheel (GWA) can increase depending on
the number of dither positions that will be observed per visit (see Section 8.2 of the
NIRSpec Ops Concept Document). It may not be possible to optimize data quality from
dithering while also minimizing the number of instrument component moves, so we will
need to find a balance between these competing desires.
3.0
Assumed Accuracy of Small Angle Motions
From the mission IRD, the JWST slew accuracy for small-angle motions are (per axis): 5
mas up to 0.5”, linearly increasing between 5 mas and 20 mas from 0.5 – 2” (1% of the
motion), 20 mas from 2 – 20”, and 90 mas at 20-30”. Informal guidance from Ed Nelan
Operated by the Association of Universities for Research in Astronomy, Inc., for the National
Aeronautics and Space Administration under Contract NAS5-03127
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
JWST-STScI-001749
SM-12
and the line-of-sight working group indicates that medium and large angle offsets will
improve by a factor of two on these requirements, so that slew accuracy between 0.5 and
2” will rise from 5 to 10 mas and will be 10 mas from 2 – 20”. Thus positioning error
associated with small angle motions down the slit should be small, of order 10 mas or
less, and present no impediment to executing small dithers.
Figure 1: Small angle motion accuracies from mission IRD requirements.
Figure 2: Layout of the IFU virtual slits on the NIRSpec detector. The IFU entrance
aperture is the circled box at right. Note that slitlets that adjoin each other in the focal
plane (slitlets 1 and 2, for example) fall on widely separated regions of the detector. The
spectra from the IFU slitlets fall in horizontal lines across the detector, at the vertical
location of the corresponding slitlet.
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
-2-
JWST-STScI-001749
SM-12
4.0
Dithers for data obtained through the IFU
NIRSpec’s Integral Field Unit (IFU) will obtain simultaneous spatial and spectral
coverage over a 3" x 3" field of view. This field is sliced into 30 slitlets that are 0.1”
wide, arranged with their long dimension in the x coordinate. Each slitlet is mapped to a
dedicated detector area. These slitlets are laid out on the detector so that slitlets that
adjoin each other in the entrance aperture lie far apart on the detector (see Figure 2). Four
types of “dithers” are under consideration for the IFU: (1) a primary pattern that can tile
of a large extended object, and three secondary dithers: (2) “slitlet stepping” to place light
from the same part of the sky on very different parts of the detector, and subpixel dithers
in the (3) spatial and/or (4) spectral dimensions. The first of these patterns is not the type
of “subpixel sampling” that traditionally counts as a dither, but since most such patterns
should be executable by a single set of guide stars they can count as “dithers” by that
more generic definition. JWST / STScI project discussions of exact terminology to
describe these motions and patterns to the flight software and to the user are still under
discussion at this writing: here, these patterns are all described as “dithers”, either
primary or secondary, and more specific naming is deferred.
4.1
Coverage of an extended object using dithers defined in APT
The IFU is well suited to map out a large (> 3”) extended object or field, such as a
Galactic star-forming region or a nearby galaxy. Provided the total motion across all the
positions (that is, the largest distance that separates any two positions) is much smaller
than the FOV of FGS (2.3’ at 2048x2048 pixels and 0.69”/pixel), these motions can
count as dithers (i.e. FGS will need to reacquire lock on the guide stars at each position,
but the same guide stars should be available). For this type of “dither”, to keep planning
Figure 3: Screen capture from the APT MIRI mosaic tool. This method should work very
well for enabling NIRSpec users to cover extended objects with the IFU.
and implementation simple, we envision the user defining a primary pattern of positions
on the sky using a simple tool in APT. For illustration, and by analogy, we show in
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
-3-
JWST-STScI-001749
SM-12
Figure 3 a screen capture of the MIRI APT tool for defining a mosaic for MIRI imaging1.
The pattern of fields on the sky is specified by a central position and by 6 additional
parameters that describe the size of the grid, the degree of overlap in x and y, the skew in
x and y, and the orientation of the whole set. For NIRSpec IFU as for MIRI these will be
the only selectable parameters of the large-scale primary pattern. Individual subpositions
in the grid can be omitted if the user desires, but the positions cannot be chosen
individually or arbitrarily. We may define optimal degrees of overlap later on.
We anticipate that the full IFU primary pattern will be executed with a single grating /
filter setting (choice of band), followed by a grating change and then another walk
through the pattern. The alternative, to obtain all three bands at one position before
moving to the next, requires approximately MxN more movements of the grating and
filter wheels, where M and N are number of rows and columns in the large-scale pattern.
Some users may desire to use the IFU to map very large regions, greater than 1 arcmin or
even greater than the FGS field, which would entail multiple sets of guide stars. For these
users, the IFU mosaic tool would ideally be able to package the many planned positions
into the correct and minimal number of visits with optimal use of guide stars; one visit for
patterns that subtend ~< 1 arcmin and use a single guide star, and multiple visits with
multiple guide stars for larger fields. Ideally this optimal packaging into multiple visits
would be transparent to the user. However, the complexity of this approach may not be
attainable in APT. If not, then simple rules could be implemented; e.g. a single visit is
assigned if N “tiles” occupy much less than the FGS field and can go in a single visit, N
visits are assigned if not. The actual implementation will of course depend on the details
of APT. At least, the user should be able to manually define many single-visit <1’
patterns that collectively map out a large region.
4.2
Slitlet stepping
The layout of the virtual slits on the detector plane places adjoining slitlets on very
different y positions on the detector (see Figures 2 and 4). This design provides the
opportunity for obtaining data at some wavelength from some position on the sky at very
different locations on the detector to mitigate the effects of hot pixels and other detector
features. Figure 4 shows a schematic for how this process works using an extended
galaxy with a bright HII region (the star symbol). Because of the alternating y positions
of adjoining slitlets, moving 1 slitlet over (moving the HII region from 0 to 1) moves a
fixed point on the sky to different y-positions on the detector. Note also that in this case,
only the inner two slices of the galaxy are covered at all positions – this will be
generically true of any extended target that is close to the same size as the IFU field.
Moves as small as one full slitlet in x (as shown here) can accomplish this purpose.
However, since the FWHM of the JWST PSF occupies 100, 130, and 163 mas at 3, 4, and
5 µm, respectively, light from the same position on the sky overlaps significantly with
multiple slitlets. Thus to obtain independent measurements at different positions it is
1
This is an analogy only – in reality this MIRI tool is used to define a mosaic pattern, i.e. each position in
the grid is a different visit. For NIRSpec’s IFU this tool will used to define a dither pattern since the IFU
FOV is small enough for the required motions to count as dithers (i.e. executed within a single visit),
provided the total size of the pattern is smaller than the FGS field and can use the same guide stars.
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
-4-
JWST-STScI-001749
SM-12
desirable to move over by at least 300 mas or 3 slitlets. For users interested in the longest
wavelengths who seek to minimize the overlap of the PSF between slitlets, we should
also define a 5 slitlet offset pattern. For N slitlet offsets, only 30 – N slitlets will be
covered at both dither positions, so the effective exposure time will be reduced at the
edges of the field. Some users may not be able to tolerate this loss of coverage but still
want improved spectral sampling. For them, we recommend a one-slitlet offset, resulting
in three patterns with 1, 3, and 5 slitlet offsets in x. Note that for any choice of 1, 3, or 5
slitlets, two positions will be observed; one at the fiducial position (selected by the user to
optimize the whole field) and the other 1, 3, or 5 positions away. Though it will be rarely
used in practice, there should be an option to not execute slitlet stepping at all. Which
option to adopt will be the user’s choice based on their desired coverage of the field, S/N
requirements, and observing time request.
Figure 4: Schematic illustration of IFU “slitlet stepping”, using a galaxy with a bright HII region as a
source. Stepping by 1 slitlet in either direction places the HII region on very different y positions on
the detector. The vertical runs of slitlets represent the actual layout in the detector plane – see Figure
2. This illustrates the principle: in practice, for a 1 slitlet-step dither choice two of these positions
would be observed (i.e. the first and second, or the first and third, but not the second and third,
which are separated by 2 slitlets.)
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
-5-
JWST-STScI-001749
SM-12
4.3
Subpixel dithers in the spatial direction
The SCAs will sample the spatial dimension of the IFU at 100 mas resolution, so the user
may opt to improve this spatial sampling with secondary dithers. For the IFU these
spatial dithers should closely parallel those defined for the fixed slits - the basic pattern is
to observe with an offset of 0.15” as a secondary pattern (in addition to any slitlet
stepping). The value of 0.15” (≈1.5 pixels) is chosen to minimize overlap between the
PSF in the two positions, which is especially important for bad-pixel avoidance when the
slitlet stepping option is not chosen. See below for notes on combining these subpixel
dithers with the IFU tiling pattern.
4.4
Subpixel dithers in the spectral dimension
Since the dispersed light from the IFU cannot be moved around on the detector, there is
no need for the sort of spectral or wavelength-gap covering dithers planned for the fixed
slits and MSA. However, the user may want to increase the sampling of the spectral
resolution element and average out subpixel variations in sensitivity. Here again the
simplest approach is to observe at 0.05” offsets (≈0.5 pixel, ≈0.5 slitlet widths) as a
secondary pattern (in addition to any slitlet stepping), with or without the sub-pixel
spatial dithers. See below for notes on combining these subpixel dithers with the tiling
pattern.
4.5
Combining slitlet stepping with subpixel shifts
In contrast to the fixed slits (see Part I of this report), the IFU will often be used to cover
an extended object on the sky. So in most cases a number of dithering moves will be done
in a large-scale tiling pattern and overheads will be incurred already before any subpixel
dithers are added. Also, if slitlet stepping is done to mitigate detector effects the center
regions of the field will be observed multiple times. In light of these two effects it is most
efficient to combine the slitlet stepping and subpixel dithers into single moves if the user
chooses to include both types of dithers.
In the combined scheme, the slitlet steps illustrated in Figure 4 would occur over 1 slitlet
(or 3 or 5) plus one half slitlet (the slitlets are ≈1 pixel or ≈100 mas wide). Thus the steps
would be 1.5, 3.5, or 5.5 slitlets over in x if the user chooses to include the subpixel shifts
and 1, 3, or 5 slitlets in x if they do not (see Figure 5).
It also makes sense to combine the subpixel spatial dithers with the slitlet motions in the
same fashion, so that 1.5, 3.5, or 5.5 slitlet motions in x would be combined with ≈0.15”
(≈1.5 pixel) offsets in the y (spatial) dimension (see Figure 5). These combinations are far
more efficient than using 2 to 4 subpixel pointings at each slitlet step (as for the fixed
slits), since the slitlet step motions are being done anyway.
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
-6-
JWST-STScI-001749
SM-12
Figure 5: Examples of secondary dithers for the IFU, and their combinations. Each pair of apertures
represents the two positions observed for a given combination of slitlet step and sub-pixel shifts in x
and y. for how slitlet steps can be combined with subpixels secondary shifts for optimal sampling.
The subpixel shifts are 0.15” (≈1.5 pixel) in the spatial direction and/or 0.5 slitlet (0.05” or ≈ 0.5
pixels) in the spectral dimension. In practice users will choose the fiducial See text for discussion.
The secondary dithers proposed here are intended to deliberately shift light on the
detector pixels by a specific angular offset. However, the optical distortion of the field
across the detector, the “tilt” of the IFU apertures with respect to the y-coordinate of the
detector, and curvature of the spectra mean that there is not an exact and uniform
correspondence between position on the sky and pixel space. The effect of these nonuniformities is that a shift, say in the dispersion direction as for the slitlet stepping, will
also move the spectrum in the y-direction in some places and so change its spatial
sampling in pixel space even though this was not strictly desired. This is like an
unintentional dither. These complications cannot be avoided; their magnitude and
dependence on location will not be known until the flight instrument is tested.
Nevertheless, it is important to implement the capability to achieve deliberate shifts of
known size on the sky to ensure the general purpose of dithering, help calibrate these
distortion effects, and enable adaptation if necessary. It is for these reasons that the shifts
are usually described here in arcsec or slitlets rather than pixels.
5.0
Possible implementation in APT
From the user perspective, these dithers can be very simple. We can envision that APT
will present to the user first, the option of whether or not to create a large-scale (primary)
pattern as described in Section 4.1. This pattern would then be defined with a NIRSpectuned tool by analogy with the MIRI mosaic tool. Once this is done there will be three
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
-7-
JWST-STScI-001749
SM-12
additional choices to make: (1) whether to perform slitlet stepping at each position in the
tiling pattern, (2) whether to add on the subpixel x shifts, and (3) whether to add on the
subpixel y shifts. We can characterize these four choices with a 4-element string vector
[tiling, step, suby, subx] = [yes/no, no/1/3/5, yes/no, yes/no]. The number of possible
combinations can be derived from the range of possibilities for each choice. There will
either be a primary tile pattern or not (2 possibilities), a choice of 1, 3, or 5 or no slitlet
step (4 possibilities), and yes or no to both the subpixel offset choices, for a total of 32
possible combinations. However, since we advocate combining the subpixel shifts with
the slitlet motions, not as many actual positions will be observed in the final pattern
defined by these choices.
For instance, let’s assume the user has defined a 3 x 3 tiling pattern. At each of these the
choice is made to observe with slitlet steps with a 3-slitlet separation. There will therefore
be 3 x 3 x 2 positions observed for each grating. If the user has chosen to use subpixel
steps in y only [yes,no] the two slitlet step positions will be separated by 0.3” (=3 slitlets)
in x and 0.15” (≈1.5 pixels) in y. If the user has chosen to use subpixel steps in x only
[no,yes] the two slitlet step positions will be separated by 3.5 pixels (3 slitlets+0.5 pixel)
in x and will have the same y position. If the user has chosen both x and y subpixel offsets
the two slitlet step positions will be separated by 3.5 pixels (3 slitlets+0.5 pixel) in x and
1.5 pixels in y.
To minimize mechanism usage we recommend that exposures be obtained at each tiling
position in the pattern, including slitlet and subpixel shifts, before moving the grating
and/or filter wheel to another band. Thus, the typical observing sequence will be TA,
MxN tile positions broken into 2 slitlet steps if desired, then a grating/filter change,
followed by a backtrack through the pattern. In fact, when implemented this way these
dithers do not increase the number of mechanism motions triggered by this observation
because all the dithers occur between GWA/FWA motions that would occur anyway if
the user has chosen to observe in more than one band (e.g. G140M/F070LP at 1.3 µm,
G235M/F170LP at 2.2 µm, G395M/F290LP at 3.7 µm).
Check with the JWST SOCCER Database at: http://soccer.stsci.edu/DmsProdAgile/PLMServlet
To verify that this is the current version.
-8-