September 27, 2023

1. Dithering for ACS/WFC3 Primes and Parallels

J. Anderson

The three workhorse imagers currently on board HST (ACS/WFC, WFC3/UVIS, and WFC3/IR) are all undersampled.  When cameras are undersampled, it is often useful to dither observations to place each source at a variety of locations relative to the pixel boundaries so that Drizzle and similar tools can create higher resolution images of the astronomical scene.  

The standard 4-point "box" dither patterns available in APT were designed with only one instrument in mind — the instrument that is “prime”.  ISR 2023-05 explores how effective these patterns are for observations taken in parallel.  There are four combinations to explore:  (1) WFC3/UVIS prime with ACS/WFC parallel, (2) WFC3/IR prime with ACS/WFC parallel, (3) ACS/WFC prime with WFC3/UVIS parallel, and (4) ACS/WFC prime with WFC3/IR parallel.  In Figure 1,  we show the pixel-phase coverage achieved by the standard "box" dithers for first two combinations:  

Two panels, each holding four sub-plots. The left four show the total dither and sub-pixel dither for the prime instrument (WFC3, with UVIS on top and IR on the bottom), and the right shows the same panels for the ACS/WFC in parallel. The subplots consist of red, green, blue, and black clusters of points.
Figure 1: The panels on the left show the total dither and sub-pixel dither for the prime instrument (WFC3, with UVIS on top and IR on the bottom).  The panels on the right show the same for the ACS/WFC in parallel.

The left shows the total dither and sub-pixel dither for the prime instrument (WFC3, with UVIS on top and IR on the bottom), and the right shows the same panels for the ACS/WFC in parallel.  The sub-pixel dither in the prime instruments covers the pixel face evenly in both cases.  But the parallel coverage leaves some large gaps in the face of the pixel.  A drizzle reconstruction of the scene in the parallel instrument would have significant sampling-related artifacts.  Depending on where a point source landed in the detector pixel grid and where that maps to in the reconstructed frame, the source could end up appearing to be sharper or broader than other point sources in the drizzle-reconstructed images.  This makes it hard to distinguish point sources from resolved objects and to determine accurate size and shape estimates for marginally resolved objects.

In an effort to remedy this, we made use of the flexibility we have to add whole-pixel offsets to the optimal sub-pixel offsets in hopes of finding a combination of whole-pixel offsets that produces a good sub-pixel dither in the parallel instrument.  This required several steps.  First, we developed an empirical mapping from the commanded POS-TARG vector to the achieved raw-pixel x and y offsets in the prime and parallel instruments.  This allowed us to take a given trial dither in the prime instrument and evaluate the corresponding dither in the parallel instrument.  Next, since we cannot expect to construct a "perfect" dither in the parallel instrument, we had to define a metric that could tell us which dither is the best (in a sub-pixel sense) among the many options.  

Using these tools, for each of the four cases described above, we took the "ideal" sub-pixel offsets in the prime instrument and cycled through the allowable whole-pixel offsets (consistent with the <0.3 arc-second restriction above) and determined the set of prime dithers that produced the best sub-pixel dither in the parallel instrument.  Figure 2 below shows the results for the two WFC3-vs-ACS cases shown in Figure 1:

Two panels, each holding four sub-plots. The left four show the total dither and sub-pixel dither for the prime instrument (WFC3, with UVIS on top and IR on the bottom), and the right shows the same panels for the ACS/WFC in parallel. The subplots consist of red, green, blue, and black points.
Figure 2: Same as Figure 1, except with the newly optimized prime-instrument dithers that produce better sub-pixel dithers in the parallel camera.

The dithers in the parallel instrument are not perfect, but they are quite good, and they are considerably better than before.  We were also able to find good dithers for the other two cases where ACS/WFC is prime and WFC3/UVIS and WFC3/IR were in parallel (see ISR 2023-05 ). The ISR provides these dithers in terms of POS-TARGs that users can input into APT as special requirements. 

Not all prime-parallel programs will benefit from these dithers; in particular, programs that don't require a high-resolution reconstruction in both cameras should use the standard prime-instrument box dithers.  As with most observing decisions, there are trade-offs involved in using these dithers.  In order to prevent pixel-phase decoherence, we had to keep the dither within 0.3" (6 ACS pixels).  This made it hard to mitigate blobs and persistence in WFC3/IR and also made us more vulnerable to bad columns in ACS.  For this reason, if the program is focused a small number of small targets in each field, then it may be important to use the standard box dithers, since they mitigate larger artifacts for all sources.  These tighter patterns should provide excellent results for most targets, but some targets may be beset by artifacts in more than one out of the four images.  

The current ISR provides only 4-point patterns.  However, the same procedure could be used to construct 2-point and 3-point patterns that provide the best possible sampling for prime and parallel.  If there is demand (contact the HST General Help Desk), we will construct the additional patterns, prioritizing the cases that are needed most.  Many programs take several orbits of the same field, which involve more than four exposures.  For these programs, it is possible to package these sets of four dithers into larger patterns, for example, by offsetting each set of four POS-TARGs by 0.5" in either x or y.  This way, the good sub-pixel sampling is guaranteed, and no source falls on the same pixel in more than one exposure.  It can take a large number of "random" dithers to get even pixel-phase coverage, so this Nx4 strategy guarantees good coverage for any multiple of four exposures.  

2. The WFC3/UVIS G280 Grism Sky

A. Pagul, R. Ryan, B. Kuhn, D. Som

We have constructed the first publicly available sky calibration frames for the WFC3/UVIS G280 grism, for both the calibrated, flat-fielded individual FLT exposures, as well as their corresponding CTE-corrected FLC frames, utilizing publicly available data from MAST. The G280 sky frames will be published and available through the Grism Resources page in late 2023.

In this work (ISR 2023-06), we analyze sources that contribute to stray light and characterize the structures they imprint on the UVIS CCD to efficiently mitigate the effects of these backgrounds. The background includes nominal CCD features such as the flare and cross-hatch pattern which are also present in both UVIS reference flat fields and overlapping broadband flat field images. Because the magnitude of these effects varies with observing conditions (e.g. zodiacal light, galactic light, earth shine, etc.), we also offer thresholds for background levels achieved depending on observing conditions and precision requirements for observers’ science cases. Some examples are shown in Figure 3.

Four subplots arranged in a two by two setup. Each plot is made up of dots ranging in color from purple (negative sun altitude) to blue, then green, then yellow with increasing sun altitude.
Figure 3: (Top left) Median background in G280 measured as a function of Sun Altitude. (Top right) Median background in G280 as a function of ecliptic latitude. (Bottom left) Median background in G280 as a function of Sun Angle. (Bottom right) Median background in G280 as a function of Moon Angle. All data-points are color-coded by Sun Altitude.

We find that the sky calibration performs well with a single spectral component, with no statistically significant additional spectral components from Earth’s atmosphere. Implementing our sky model with HSTaXe significantly reduced background noise across low and high-background frames, with a background pixel distribution centered at 0 e-/s (Figure 4). Our model also reduced the spatial variations across both chips, resulting in a constant background value in the extracted spectra (Figure 5). 

One plot containing two histograms. The left histogram is orange and represents UVIS 2, the right is blue and represents UVIS 1. Both histograms are roughly the same height.
Figure 4: Pixel value distribution histograms of the original and corrected science frames for UVIS1 and UVIS2 from observation icwz50hbq, GO proposal 14127; PI Fumagalli. Note how the zodi-corrected (orange) histogram is centered around 0, providing a sanity check that the global correction is correctly applied. 
A figure containing three rows and two columns, holding six plots.  The top two plots are light green with yellow and show a WFC3/UVIS image. The middle two are a darker green and show the same image but “zodi-subtracted”, and the third row is a scatter plot with blue and orange markers. The blue markers (original column medians) have an increasing slope and the orange (zodi-corrected column medians)  have a mostly flat slope.
Figure 5: Examples of sky reduction for UVIS1 (left) and UVIS2 (right) for the same frames used in Figure 4. The top row corresponds to the original frame; the middle row corresponds to the zodi-subtracted frame; the bottom row corresponds to the column medians across the x-axis normalized to 0.

Future work will focus on characterizing how the sky varies across time and predicting background levels given observing conditions and benchmarks necessary for specific science cases. This first G280 sky helps enhance and improve the utility of the UVIS grism’s capabilities, enabling more precise results across a variety of science cases. 

3. Updates to WFC3/UVIS Shutter Timing Jitter

K. Huynh, P.R McCullough

A comprehensive study of shutter timing jitter has brought to light a few new details for observers modeling jitter noise. The repeatability of the UVIS exposure time is a characteristic of the mechanical shutter (WFC3 Instrument Handbook (IHB) Section 2.3). The WFC3/UVIS channel integration times are controlled by the shutter operations, but idiosyncrasies of the mechanical shutter result in slight non-repeatability in the exposure time from one exposure to another, which we refer to as “shutter timing jitter”. The current system requirements include a repeatability to 0.010 seconds (WFC3 IHB Section 6.7.1).

In ISR 2023-04 we describe our new methodology developed to calculate the shutter timing jitter for short exposures. We observe a spectrophotometric standard star with the G280 grism in order to disperse its light across hundreds of pixels. This allows for much higher photometric precision in very short exposures as well as minimizing several sources of noise so we can more precisely measure the shutter timing jitter compared to previous results using direct images of a star (ISR 2009-25) and internal flat fields (ISR 2015-12).

We find the actual exposure time varies by 2.43 +/- 0.32 milliseconds for exposures of nominal lengths of 1, 2, and 4 seconds, nearly 4 times better than the system requirement (repeatability to 0.010 sec). The repeatability is approximately ten times better than the requirement for the shortest possible exposures (0.5 sec), where the shutter sweeps without stopping (WFC3 IHB Section 6.7.1) during its open state for consecutive 0.5-s exposures. Contrariwise, the timing jitter is about 5.5 milliseconds for 0.7-s exposures, better than the requirement but not as good as for other short exposures. The 0.7-s exposure is the shortest exposure with a temporary stop in the open state, and as a result is the most affected by the vibration of the shutter system (WFC3 IHB Section 6.10.4). Observers should consider substituting a 0.5-s exposure whenever a 0.7-s exposure is used, although this recommendation is only based on a dozen exposures taken in series. We also find that in short exposures, the shutter jitter dominates Poisson noise as shown in Figure 6. Observers doing precision photometry with short exposures may want to include the 2.4 millisecond shutter jitter error in their noise model. For more details, see  ISR 2023-04  (Huynh & McCullough). 

One plot containing five lines in orange, green, red, purple, and blue. The blue line first increases and then decreases, while the other four lines all steadily decrease.
Figure 6: Photometric noise sources across the G280 spectra for exposures of various lengths. Two theoretical components, shutter jitter and Poisson noise, are plotted as straight lines with power law indices of -1 and -0.5 respectively. Their root-sum-of-squares combination is also shown. Jitter times of the positive orders are plotted as blue points with their associated 1 sigma uncertainties. In short exposures, the shutter jitter dominates Poisson noise, and no additional corrections for the Poisson noise need to be made when the jitter timing noise is accounted for due to the Poisson noise being negligible.

4. HSTaXe - ACS and WFC3 Cookbook Tutorials  

B. Kuhn, D. Som, A. Pidgeon, N. Hathi, R. Ryan, R. Avila, N. Pirzkal

In the recently published ISR 2023-07, we discuss a collection of six Jupyter Notebook tutorials that were released on the HSTaXe GitHub repository in Spring 2023. These ‘cookbooks’ present examples of how to preprocess data from ACS and WFC3 slitless spectroscopic modes and use the core HSTaXe routines to extract 1D spectra. The specific preprocessing procedures, described in the ISR and cookbooks, are meant to highlight three steps of the data analysis process users should consider in order to obtain optimal spectral extraction with HSTaXe. The three steps include a custom multi-component background subtraction for WFC3/IR grism data, embedding subarray data into a full-chip image, and checking that the active World Coordinate System (WCS) of dispersed images matches the corresponding direct images.

In Figure 7 we show an illustrative workflow of the recommended steps users should take to achieve optimally extracted spectra with HSTaXe. The boxes highlighted with a green background represent the three new preprocessing steps that are included in the cookbooks. While the custom background subtraction is applicable to only WFC3/IR grism data, embedding subarray exposures and verifying the active WCS is necessary for any instrument's data being used with HSTaXe.

Flowchart illustrating the recommended preprocessing steps one should take to achieve optimally extracted spectra with HSTaXe. Highlighted in green are the "custom background subtraction", "embed subarray", and "check direct image WCS" steps.
Figure 7: Flowchart illustrating the recommended preprocessing steps one should take to achieve optimally extracted spectra with HSTaXe.

The six cookbooks are split between the ACS and WFC3 instruments. Inside the ACS cookbook folder on the HSTaXe GitHub repository users will find three different cookbooks that cover the following topics:

  1. Spectral extraction with full-frame data from ACS’s Solar Blind Channel,
  2. Spectral extraction with full-frame data from ACS’s Wide Field Channel (WFC), and
  3. Spectral extraction with subarray data from ACS/WFC

The cookbooks in the WFC3 folder are separated in a very similar way. Within the folder, there are three cookbooks that encompass the following topics:

  1. Spectral extraction with data from WFC3’s IR channel (with an included advanced extraction example),
  2. Spectral extraction with full-frame data from WFC3’s UVIS channel, and
  3. Spectral extraction with subarray data from WFC3/UVIS

We will continue to maintain the cookbooks to ensure their compatibility with any future changes to the HSTaXe software, and any issues with the installation or running of the cookbooks should be reported to the HST General Help Desk. Finally, users seeking additional information and resources pertaining to slitless spectroscopic modes should visit the ACS prism/grism performance page and/or the WFC3 Grism Resources page.

5. New Documentation 

ISR 2023-04: UVIS Shutter Timing Jitter - K. Huynh & P.R. McCullough

ISR 2023-05: Dithering for ACS and WFC3 Primes and Parallels - J. Anderson & N. Grogin

ISR 2023-06: The WFC/UVIS G280 Grism Sky - A. Pagul, R. Ryan, B. Kuhn, D. Som

ISR 2023-07: HSTaXe - ACS & WFC3 Cookbook Tutorials - B. Kuhn, D. Som, A. Pidgeon, N. Hathi, R. Ryan, R. Avila, N. Pirzkal


The complete WFC3 ISR archive is available here. Additional information about WFC3 calibration, performance, data analysis, software tools, and more can be found online.

Need help?

To subscribe to the WFC3_NEWS list, please send a blank email to WFC3_NEWS-subscribe-request@MAILLIST.STSCI.EDU

To unsubscribe from the WFC3_NEWS list, please send a blank email to WFC3_NEWS-signoff-request@MAILLIST.STSCI.EDU


HST Help Desk 

Please contact the HST Help Desk with any questions.