NIRCam Imaging Recommended Strategies
Guidance is provided for astronomers preparing JWST NIRCam Imaging observations using the Astronomers' Proposal Tool (APT).
Here we give advice on handling the gaps between detectors, selecting dither patterns and readout patterns, and reducing data volume and overheads.
Dithering is required to mitigate bad pixels and improve overall image quality. Larger primary dithers are useful to fill gaps in sky coverage. Smaller subpixel dithers are optimized to improve sampling in stacked images. Programs can use both, but programs covering very large areas, or viceversa targeting single compact sources may implement only primary or secondary dithers, depending on their science goals and the observing overheads added for each dither and exposure. Primary dithers also offer some improvement in image sampling, so these may be preferred unless an optimal image sampling is the highest priority.
In general, it is preferable to divide each exposure into as many dithers as possible, considering noise floor and overheads, to improve image quality.
Of the primary dither patterns, a few options recommended for general use are:
- FULLBOX 6TIGHT – fills all gaps between detectors and modules; requires greater overheads than the INTRAMODULE dither types
- INTRAMODULEBOX 4 – fills the short wavelength gaps, leaving the gap between modules; maximizes the deep area with full exposure time
- INTRAMODULEX 3 or 4 – fill the short wavelength gaps, leaving the gap between modules; larger but more uneven areal coverage than INTRAMODULEBOX 4.
- Depth across the stacked image will be uneven, especially for FULLBOX or FULL dithers. This should be considered when calculating your required exposure time with the JWST Exposure Time Calculator.
- Any area covered by only 1 dither cannot be corrected for bad pixels. This may leave holes in the final stacked image.
Of the subpixel dither patterns, SMALL-GRID-DITHER has lower overheads and is recommended for use when primary dithers are also being obtained. If no primary dithers are performed, then the larger STANDARD dithers are recommended to better mitigate bad pixels.
Subpixel dithering is particularly relevant when the PSF is undersampled by the detector pixels, i.e. when the pixels are too large to provide adequate Nyquist sampling of the PSF in a single exposure. In such cases, multiple exposures obtained with a subpixel dither pattern provide additional information about the PSF structure, enabling reconstruction of an improved PSF that is closer to the intrinsic one. For NIRCam, the PSF is well sampled at wavelengths longer than about 2 µm and 4 µm in the short and long wavelength channels, respectively. This is because the nominal pixel scale of 31 mas and 64 mas have been defined using the criteria of sampling with 2 pixels the lambda/D core of the PSF at these wavelengths.
For shorter wavelengths, where undersampling can become significant, a good rule of thumb is that the number N of subpixel dithers should scale approximately as the square of the wavelength ratio, relative to the well-sampled regime. For example, images obtained in the short wavelength channel with F115W are undersampled by a factor ~ 2µm / 1.15µm, ie a factor of ~1.73 (relative to the well-sampled 2 µm regime), so a minimum of (1.73)2, i.e. 3 subpixel dithers, would be needed to recover the PSF information. In practice, observers may wish to consider 4 subpixel dithers, which can provide somewhat more regular subsampling of the square detector pixels on a 2x2 half-pixel grid. For more detailed background information on principles of dithered observations with JWST, see Koekemoer & Lindsay (2005), Anderson (2011), and Anderson (2014).
Note when performing coordinated parallel observations, additional custom subpixel dither patterns are available that achieve optimal pixel phase sampling for both the prime and parallel instruments. However, NIRCam's small grid dithers are not available for parallel observing, since they are not large enough for good sampling of the PSF for other instruments.
Mind the gaps
We recommend using the APT Aladin Viewer to check the target placement and dither coverage. We remind that Primary dithers fill the gaps between NIRCam detectors but result in uneven depth across the stacked image. Consider this when calculating your required exposure time using the Exposure Time Calculator.
When observing with both ("ALL") NIRCam modules, the science target is by default centered in the ~44” gap between the modules in the field of view. Depending on the size of the target, we recommend the following strategies to ensure your target does not fall between the gaps:
a) Large scene
If the goal is to image a scene larger than 5' × 2' without gaps:
- Use FULLBOX primary dithers to fill the gaps. This will cover the full area while sacrificing depth in some areas.
b) Mid-size target
For a target that fits within one module (2' × 2'), say well within the field of view of an HST camera, we recommend centering the target within Module B by either:
- Adding a Special Requirement OFFSET of ~82” in X. Here, X is given in the “Ideal” (X, Y) coordinate system, which is fixed relative to the NIRCam detectors. (Note with this offset, the center of the target will be in the gap between the short wavelength detectors, so primary dithering should be used to fill these gaps.)
- Observing with NIRCam Module B only. (The center of the target will be near the center of Module B, in the corner of detector B4.)
In either case, we recommend primary dithering with INTRAMODULEBOX or INTRAMODULEX to fill the short wavelength chip gaps. Most of the area imaged by the two modules will be at full depth. Observing overheads will also be lower than when using FULLBOX dithers to cover the full scene.
c) Compact target
For a target that fits within one short wavelength detectors (1' × 1'), you may choose to center it within one of those detectors using a different OFFSET. For example:
- Add a Special Requirement OFFSET of ~55” in X and ~35" in Y to center the target within detector B3 while observing with both modules. (To choose a different detector, or if observing with Module B only, you may calculate the offset using Table 1 in NIRCam Apertures, or simply use the Aladin viewer to move your target manually.)
This strategy ensures the small target is observed at full depth by avoiding the gaps between the short wavelength detectors. INTRAMODULEBOX, INTRAMODULEX, or SMALL INTRASCA primary dithers and/or subpixel dithers are recommended in this case. Even point sources will benefit as dithering mitigates bad pixels and flat field uncertainties.
We recommend integration times less than 1000 seconds. For longer integrations, the majority of pixels would likely be affected by cosmic rays. Total exposure times can be made longer by using dithers and by increasing the number of integrations. See discussion in MIRI Recommended Strategies.
To achieve a desired exposure time and signal-to-noise, users must choose among the 9 available NIRCam detector readout patterns, as well as numbers of groups (of each pattern) and integrations at each dither position. We provide the following guidance:
- The readout patterns RAPID, BRIGHT2, SHALLOW4, and MEDIUM8 maximize signal-to-noise for a given integration time < 1000 seconds. See NIRCam Imaging Sensitivity for details.
- Greater numbers of groups are preferred to mitigate cosmic rays for all pixels. The minimum number should be 5 groups, except for the shortest exposures, according to the recommendations for maximum sensitivity. For example, we recommend 8 groups of SHALLOW4 rather than 4 groups of MEDIUM8, as long as the data volume is manageable.
- A single integration at each dither position will be sufficient for most programs. Given the choice between extra integrations or extra dithers, dithers are preferred as they improve data quality in multiple ways. The one drawback is that dithers increase overheads.
- DEEP8 and DEEP2 result in integrations >1000s, which are expected to be substantially affected by cosmic rays. These patterns are thus not recommended, but may be required in some cases to reduce data volume.
APT places some limits on Data Rates and Volume. For each visit, the Data Volume may not exceed 58 GB, roughly the memory capacity of the onboard Solid State Recorder. Exceeding this limit will generate an error in APT. Exceeding half that limit (29 GB) will generate a warning.
Users should also check the ratio of Data Volume / Total Charged Time for each observation. If this ratio exceeds 0.654 MB/s for a total time of ~12 hours or more, then the program will likely be difficult or prohibitive to schedule. APT does not issue a warning in this case.
To reduce the data volume, try the following (assuming a fixed total exposure time to achieve the required signal-to-noise):
- Use a longer readout pattern that generates less saved data, for example DEEP8 instead of MEDIUM8.
- Observe with fewer detectors.
- Observing with both ("ALL") modules yields data from 10 detectors.
- Observing with Module B only and full subarrays yields data from 5 detectors.
- Some NIRCam subarrays yield data from as few as 2 detectors.
- Note that choosing a smaller subarray does not necessarily decrease the data volume / total charged time significantly, since smaller subarrays are read out more quickly.
- If necessary, obtain fewer dithers, each with a longer exposure. This may sacrifice data quality somewhat. One additional benefit is that observing overheads will decrease.
Our advice to reduce observing overheads and improve efficiency can be summarized simply as: sit and stare. Minimize the numbers of filter changes, dithers, exposures, and (importantly) visits requiring new guide stars after larger pointing shifts (see Slew Times and Overheads). These operations may take several minutes (see Instrument Overheads). Of course, filter changes and dithers improve data quality and may be required for your science.
Bear in mind your dither pattern(s) will be repeated for each filter pair (short and long wavelength). Larger dithers with larger overheads should be avoided if possible, in part by using the dither patterns introduced in APT 25.4.1 (Coe 2017). Whenever possible, use a more compact dither pattern that does not split your observation into multiple visits. This will depend on the Visit Splitting Distance assigned to your observation. Targets at lower Galactic latitude will have larger Visit Splitting Distances and may be more efficient to observe, especially with the FULLBOX dither pattern that fills all gaps between detectors and the modules.
Dither Patterns for NIRCam Imaging
Anderson, J. 2011, JWST-STScI-002199
NIRCam Dithering Strategies I: A Least Squares Approach to Image Combination
Anderson, J., 2014, JWST-STScI-002473
NIRCam Dithering Strategies II: Primaries, Secondaries, and Sampling
Coe, D. 2017, JWST-STScI-005798
More Efficient NIRCam Dither Patterns
Koekemoer, A. M. & Lindsay, K. 2005, JWST-STScI-000647
An Investigation of Optimal Dither Strategies for JWST