NIRSpec MPT - Parameter Space

The JWST MSA Planning Tool has been tested using a range of parameters and a variety of catalog densities in order to give a realistic example of performance trends.

On this page

Introduction

See Also: NIRSpec MPT - Computational Performance 

Words in bold italics are buttons 
or parameters in GUI tools. Bold 
style represents GUI menus/
panels & data software packages.

The Micro-Shutter Assembly (MSA) that is used for MOS observations has ~250,000 shutters that can be opened individually, or in columns to create ‘slitlets’. For NIRSpec multi-object spectroscopy, observation plans are generated by the MSA Planning Tool (MPT) in APT. Many MPT parameters can significantly affect the number of sources in a plan. The impact of several of them are explored in this article. The runtime of plans generated in MPT is affected primarily by the number of sources in the Catalog and the number of test pointings in the search grid (which in turn depends on the Height, Width and Step Size).

To prevent spectral overlaps, each source that is accepted as a target masks out a certain area of the MSA where other targets are prevented from appearing. However, the masked area for the medium-resolution (M) and high-resolution (H) gratings are not significantly different from one another. The resulting source numbers were nearly the same, so the results for the M and H gratings were combined in the figures below. Prism spectra are much shorter in the dispersion direction, and the multiplexing on the MSA is subsequently about 3 to 4 times greater than for the other gratings.

The numbers on this page were obtained using APT version 2020.3.1. The numbers in Figures 3, 4, and 5 were later updated using APT 2020.4.1.



About the Catalogs

The Catalogs used in these tests were derived from a catalog of galaxies in the Hubble UDF (the UVUDF catalog of Rafelski, 2015). The original catalog has approximately 830 sources/arcmin2 and is distributed evenly over an area slightly smaller than the 3.4” × 3.6” MSA footprint. It is considered to be moderately dense. Subsets of the full catalog were randomly chosen to create additional catalogs with densities of 800, 400, 200, 100, 50, 25, 12.5, and 3.125 sources/arcmin2. The Aperture PA used in MPT was 45o in all cases, to approximately match the footprint of the catalog.



Nods

All results shown are for MPT plans using 3-Shutter Slitlets with Nods. Nodding repositions the telescope slightly between exposures to place the targets into different shutters within their respective slitlets. Adding nods is highly recommended for point source science to improve sampling and provide background measurements. There is little cost to nodding because the same MSA configuration is re-used at the different nod positions. Adding nods decreases the multiplexing slightly because it requires additional space above and below each planned slitlet to be clear of sources. This is needed to prevent contamination in the background spectra at all the nod positions.



Source Centering Constraint 

The Source Centering Constraint sets the bounds of the area inside a mircoshutter where sources can be observed. There are five choices: Unconstrained, Entire Open Shutter, Midpoint, Constrained, and Tightly Constrained. Constraining the target centering results in higher throughput, but, except for dense Catalogs and Candidate Sets, the number of planned, or observed, sources will be lower. 

As  shown in Figures 1 and 2, the effect of the Source Centering Constraint on the number of sources varies. The effect is greatest at lower Catalog density (below ~200 sources/arcmin2 for M or H gratings and below ~600 sources/arcmin2 for the prism). At higher Catalog density, there is an excess of potential sources, so the number of sources is instead limited by the available area of the MSA. As the number of target sets (defined below) increases, and the number of available Catalog sources decreases, the choice of Source Centering Constraint has a greater effect.



Number of target sets

A target set is a group of sources that are observed in one full set of exposures (over Nods and Dithers). The number of target sets depends on the number of Dithers that are specified and the Number of configurations that are selected. MPT will display the Number of configurations needed to complete each target set. When a multiple, N, of that value is selected, N target sets will be planned. For each new target set in the plan, the number of Catalog sources that are unobserved and available for planning decreases. As a result of there being fewer available sources, the number of observed sources in each new target set will correspondingly decrease. This effect can be seen more readily when many target sets are planned.

The number of sources in a target set depends on a number of input planning parameters discussed in this article. However, with other parameters held fixed, the number of planned sources is higher for denser catalogs, as seen in Figures 1 ad 2. At densities above ~200 sources/arcmin2 for M or H gratings, and above ~600 sources/arcmin2 for the prism the number of planned sources using a default 3-Shutter Slitlet approaches the limit that is set by the slit length and the finite area of the MSA, with its failed shutters. At lower density, the number of available, unobserved sources in a Catalog limits the number that are planned within a target set.

Figure 1. M or H gratings source numbers vs. Catalog density, and Source Centering Constraint for 1-3 target sets

A plot showing how Catalog density, Source Centering Constraint, and number of target sets affect the number of sources in plans that use M or H gratings with a 3-Shutter Slitlet and a 5” Search Step Size. Beyond Catalog densities of ~200 sources/arcmin2, target sets have typically ~60 to 65 sources.
Figure 2. Prism source numbers vs. Catalog density, and Source Centering Constraint for 1-3 target sets

A plot showing how Catalog density, Source Centering Constraint, and number of target sets affect the number of sources in plans that use the prism with a 3-Shutter Slitlet and a 5” Search Step Size. Beyond Catalog densities of ~600 sources/arcmin2, target sets have typically ~180 to 200 sources.


Search step size

The Search Step Size is the spacing of the grid that MPT uses to find the optimal set of pointings for a plan. Searching with a finer grid (smaller step size) increases the total number of pointings that are checked, and often results in plans with more sources being found. 

As shown in Figures 3 and 4, decreasing the Search Step Size generally increases the number of observable sources. For lower density Catalogs, the relative increase is more significant. The increase is not exactly linear. The results depend on the source distribution at the MSA plane; the MSA shutter bars can hide sources. The numbers also depend on the location of the test pointing relative to the Catalog center; test pointings near the edge of the Catalog contain fewer sources in the MSA. Note that the overall source numbers in the prism plots are larger simply because the spectra are shorter and more sources can be accommodated in the MSA without spectral overlap on the detector.

The real estate on the MSA is limited by its size and the number of operable shutters. It is apparent in Figures 3 and 4 that the maximum number of sources that will fit on the MSA is ~70 for M or H gratings and ~220 for the prism in the 3-shutter slitlet observing case with very dense catalogs. 

Figure 3. M or H gratings source numbers vs. Search Step Size and Catalog density

A plot showing how Catalog density and Search Step Size affect the number of sources in plans that use M or H gratings, and Midpoint source centering. Plans using step sizes smaller than the ones shown caused low memory errors in APT and could not be generated. See the tip below.
Figure 4. Prism source numbers vs. Search Step Size and Catalog density

A plot showing how Catalog density and Search Step Size affect the number of sources in plans that use the prism, and Midpoint source centering. Plans using step sizes smaller than the ones shown caused low memory errors in APT and could not be generated.  See the tip below.


Runtime

While choosing a smaller Search Step Size can result in better multiplexing, the runtimes are longer. The time needed to generate a plan can be estimated from Figure 5. These runtimes are approximate and depend on the computer processor speed and available CPU. All our tests were run using a search grid area equal to the Catalog area (about the size of the MSA). With everything else held constant, Figure 5 demonstrates that the runtime and Search Step Size are related by a power law, except at the largest steps sizesThis is to be expected since the runtime should roughly scale as the number of test pointings in the grid. When the stepsize is halved, the number of test pointings (and runtime) is quadrupled. At the largest step sizes, the runtime approaches a minimum that depends on other factors in the code.

To reduce long runtimes and avoid memory errors caused by large Catalogs or small Search Step Sizes, try using a large step size to do a coarse search over the entire catalog area, then a smaller step size over a limited area around the results of the previous plan to find the best pointing. See NIRSpec MPT - Computational Performance for more details.


Figure 5. Runtime vs Search Grid Size

A plot showing the runtime for the plans shown in Figure 3.


Dither size

Dithering allows spectra to be observed in multiple areas of the detector. It is often recommended to mitigate the effects of detector artifacts, and improve sampling. It involves moving the telescope pointing and re-configuring the MSA to observe as many of the same sources as possible. In the MPT Planner, there is a checkbox to allow Partially Completed Sources. For these tests, that option was not used, so that all sources must be drawn from the overlap region on the MSA at the two dither points. We tried tests with a single dither in X (Dispersion; Figures 6 and 7), and in Y (Cross-dispersion; Figures 8 and 9).

From Figures 6 through 9, it can be seen that even for a single dither (two dither points) the number of planned sources is generally lowered. (Though not shown here, this effect is exacerbated when the number of dithers is increased because of the increased likelihood that an individual target will fall behind an MSA bar or off the edge of the shutter array at one of the pointings.) Larger dithers result in fewer planned sources, as seen in the figures. However, the relative loss depends on the Catalog density; higher density Catalogs lose a lower percentage of sources than lower density Catalogs. Figures 6 and 7 show dips and bumps in the cases with the lowest densities, as the sample sizes are small in these tests.

Optical distortions at the MSA can result in source losses affecting dithers of all sizes, especially if the Source Centering Constraint is tight. At larger separations, additionally, fewer sources are visible at both pointings, which limits the number of observable sources.

For spatial (Y) dithers, a sharp drop in the number of sources can occur at short dither separations due to the likelihood of encountering an unusable row due to a failed open shutter, or a short in the MSA, within a distance similar to the dither separation. Though there are also shorted columns in the MSA, these contribute far less to source losses because the slit for each source occupies only a single column. The story is a little different in Y, because each slit is 3 shutters in length, not 1.

Figure 6. M or H grating source numbers vs. Dither size in X (dispersion) and Catalog density

A plot showing how dispersion direction dither size affects the number of sources in the first target set of plans that use M or H gratings, Midpoint source centering, and a 5” search step size, normalized to the results for no dither. The average of the results is also shown.
Figure 7. Prism source numbers vs. Dither size in X (dispersion) and Catalog density

A plot showing how dispersion direction dither size affects the number of sources in the first target set of plans that use the prism, Midpoint source centering, and a 5” search step size, normalized to the results for no dither. The average of the results is also shown.
Figure 8. M or H grating source numbers vs. Dither size in Y (cross-dispersion) and Catalog density

A plot showing how spatial direction dither size affects the number of sources in the first target set of plans that use M or H gratings, Midpoint source centering, and a 5” search step size, normalized to the results for no dither. The average of the results is also shown.
Figure 9. Prism source numbers vs. Dither size in Y (cross-dispersion) and Catalog density

A plot showing how spatial direction dither size affects the number of sources in the first target set of plans that use the prism, Midpoint source centering, and a 5” search step size, normalized to the results for no dither. The average of the results is also shown.


References

Rafelski, M., et al. 2015, AJ, 150, 1 (arXiv)
UVUDF: Ultraviolet Through Near-infrared Catalog and Photometric Redshifts of Galaxies in the Hubble Ultra Deep Field 




Latest updates
  •  
    Updated all figures from tests using more recent versions of APT. Also, textual changes were made throughout to provide more detailed explanation of different effects. A new link was added to the Computational Performance article.
Originally published