Space Telescope Science Institute
Table of Contents Previous Next Index Print

NICMOS Data Handbook > Chapter 5: Data Analysis > 5.5 Coronagraphic Reductions

5.5 Coronagraphic Reductions
NICMOS coronagraphic observations consist of an acquisition image and science observations. The acquisition could be an onboard acquisition performed by the NICMOS flight software (FSW) during the same visit as the science observations. Or, a real time acquisition may have been performed by sending the acquisition image(s) to the ground for analysis by the Observation Support System (OSS) and Post-Observation Data Processing System (PODPS) Unified System (OPUS) personnel during the first visit of a two visit observation set. The accompanying science observations would have been obtained during the second visit. The ACCUM, BRIGHTOBJ, and MULTIACCUM observing modes have all been used for acquisitions and for the science observations. Each type of data requires different calibration steps and dark reference files.
Coronagraphic imaging requires an acquisition sequence at the beginning of the visit to center the target in the coronagraphic hole. The size of the coronagraphic hole is smaller than typical HST blind-pointing errors. The procedure for a coronagraphic acquisition is to first image the target in Camera 2 using blind-pointing, then use either an onboard (Mode-2 Acquisition), Reuse Target Offset (RTO), or a fully interactive acquisition method. A telescope slew is calculated and commanded to move the position of the hole over the image of the target.
The target is positioned on the NIC2-CORON aperture, which has the aperture fiducial point at the image position of the coronagraphic hole on the detector. The science exposures are then specified using any of the NICMOS observing modes and any of the NIC2 filters.
During the Second Servicing Mission (SM2) Orbital Verification (SMOV) commissioning of the coronagraphic observing mode, it was determined that decentering a point source from the center of the coronagraphic hole by a small amount (x=-0.75, y=-0.25 pixels) reduced the scattered light background intensity. This offset was implemented in the NICMOS ACQ flight software for use in Cycle 7 and 7N science observations. During recommissioning following SM3B, the low scatter point was determined to be at pixel location x=-0.75, y=-0.05 from the hole center. This offset was implemented for Cycle 12 and later science observations.
The various modes of NICMOS coronagraphic acquisition and the processing required to analyze the resulting data are described in detail in a series of Instrument Science Reports (ISRs) by Schultz et al., which are available from the STScI Web pages. Users analyzing coronagraphic data, or planning future coronagraphic observations, should carefully read NICMOS ISR-97-031, ISR-98-012, ISR-98-019, and ISR-99-006. A discussion of the coronagraphic performance in Two-Gyro mode is presented in NICMOS ISR 2005-001. Some important highlights for understanding and reducing coronagraphic data are summarized in this Handbook, but the reader should consult the ISRs for more detailed discussions.
Onboard Acquisition
In Mode-2 acquisition (see NICMOS ISR-98-012, Schultz et al.), the onboard NICMOS flight software (FSW) will locate the position of the coronagraphic hole and target. It will use this information to calculate an offset to slew the telescope to position the image of the hole over the position of the target. The location of the coronagraphic hole is determined from pointed flat-field observations. Two short ACCUM F160W filter exposures (7.514 seconds each) with calibration Lamp 1 on (flat field) and two identical exposures with the lamp off (background) are obtained before the acquisition images. The telescope is NOT slewed to move the target out of the FOV during the lamp-on and background observations. The image of the target will be superimposed on these images.
The flight software combines the two background and two flat-field images, subtracts the background image from the flat-field image, and determines the position of the hole. These temporary difference images are not saved. The location of the hole is temporarily stored onboard, but it is not included in the science telemetry. The target position and the slew are stored within the embedded engineering attached to the data set. This information is extracted from the engineering telemetry and written to the science header of the observation following the ACQ, ipppss00t_raw.fits. The two flat-field images are written to image ipppssoot_rwf.fits and the two background images are written to image ipppssoot_rwb.fits.
An error in the Flight Software code prior to June 28, 1998 resulted in errors in the determination of the image position of the coronagraphic hole. Usually, this error was at the sub-pixel level and occasionally was off by a pixel. The decentered imaging affected the coronagraphic performance. It was corrected after this date.
The acquisition aperture is a square area on the detector, a 128 x 128 pixel aperture (center at 157, 128) of size 9.6 x 9.6 arcseconds. Two ACCUM images of equal exposure are obtained, and are saved together in a single fits file ipppssoot_raw.fits. A summary of the files produced in an ACQ data set is presented in Table 5.6.
Table 5.6: NICMOS ACQ Data Image Extensions
During Cycle 7 and 7N, the exptime keyword for NICMOS ACQs was incorrectly populated. A correction was installed in OPUS 9.4, on October 19, 1998.
Reuse Target Offset (RTO) and Interactive Acquisitions
A variation of the Reuse Target Offset (RTO) capability is sometimes used to acquire and position bright targets into the coronagraphic hole. This is known as Mode-1 acquisition (see Schultz et al., NICMOS ISR-98-019). Any target that would saturate the detector in the shortest possible Mode-2 ACQ ACCUM exposure time, 0.228 seconds, was considered to be a bright target. The following discussion describes the necessary steps for a Mode-1 Reuse Target Offset (RTO) acquisition. The RTO acquisitions are performed during the first visit and the science observations are obtained in a following second visit.
Images of the target and coronagraphic hole are obtained a few orbits in advance of the coronagraphic observations, and sent to the ground for analysis (RT ANALYSIS). The same roll of the spacecraft is used for both the acquisition and science visits. Usually, the same dominant and subdominant guide stars are used for both the acquisition and science observations. However, in a few instances, this is not the case; after the slew is performed, the target is found outside of the hole, either on the edge of the hole or far from the hole. The coronagraphic observer as well as future Archive users are advised to check the ipppssoot_spt.fits files for the guide stars used for the ACQ and science observations (keywords dgestar, sgestar).
The background and flat-field observations are usually offset as much as 18-25 arcseconds from the target position to avoid the diffraction spike from the image of an overexposed target crossing the coronagraphic hole and introducing errors in the measured position of the coronagraphic hole. OPUS staff assist the PI in identifying the target, centroiding, and determining offsets.
Information that an RTO has executed is currently stored in an operational database and not available to GO’s or archival users.
The location of the target and the slew are saved in the embedded engineering data attached to the science observations. These values are recovered from the embedded engineering data and written to keywords (NXCENT, NYCENT, NOFFSETX, NOFFSETY) in the RAW file. The first observation following the NICMOS ACQ will contain the updated values. The target location and slew values of the most recent target acquisition will be reflected in following observations until the next acquisition executes, or until the values are re-initialized to zero. These values are not initialized during normal operation of NICMOS.
As presented, the keyword values are scaled engineering units, and not in pixels or arcseconds. They must be converted from engineering units to detector and image coordinates.1
Target Position in Detector Coordinates
The target position coordinates (NXCENT, NYCENT, NOFFSETX, NOFFSETY) can be converted into detector coordinates by dividing by 256. However, the slew values are written in 2’s compliment. A value less than 32757 is positive, while a value larger than 32757 is negative. Slew values need to be divided by 128 during the conversion into detector coordinates. For example, the following target and slew values were obtained from a RAW file and converted to detector coordinates.
> hedit n4q832nrq_raw.fits[1] NXCENT,NYCENT,NOFFSETX,\
The position of the coronagraphic hole can then be inferred by subtracting the offset slew and the coronagraphic hole offset from the position of the target.
A fix was included in OPUS 9.2, which was installed July 16, 1998. The slew and target position keywords NXCENTP, NYCENTP, NOFFSTXP, and NOFFSTYP were added to the SPT file. These keyword values are in detector coordinates.
The image coordinate system for NICMOS Cameras 1, 2, 3 have been defined such that the origin will be in the lower left hand corner when displayed, while detector coordinates are defined by the readout directions for each camera. Any NICMOS camera image, when displayed using the IRAF display command, will be displayed relative to the HST field of view as depicted in the NICMOS Instrument Handbook. The conversion from detector coordinates to image coordinates is performed during OPUS pipeline processing. For Camera 2, the OPUS +x direction is detector -y direction, and correspondingly, the OPUS +y direction is detector -x direction. Care must be exercised when converting from one coordinate system to the other.
Measuring the Hole Position
The coronagraphic hole location can be determined from off-line processed ACQ images and should be very close to the inferred location of the hole determined by subtracting the FSW slew from the FSW target position. However, the hole pattern is not symmetric about the low scatter point (pixel with least counts) of the OPUS pipeline processed coronagraphic hole image due to the impression of the flat field on top of the image. The onboard ACQ background and flat-field images need to be reduced in a similar manner as performed by the FSW to achieve a meaningful comparison. The coronagraphic acquisition n4q832npq will be used as an example to demonstrate off-line processing.
The background image (n4q832npq_rwb.fits) must be subtracted from the image of the hole (n4q832npq_rwf.fits), and the resulting image flat fielded using a preflight flat or SMOV flat with the hole at a different position. In this example, the pre-flight flat field h1s1337dn_flt.fits was used to process the hole image. The task imcalc is used to combine the pair of background and hole images, taking the minimum value at each pixel in order to eliminate cosmic rays.
One can then invert the flat-fielded hole image to make the hole positive, and then use the IRAF task center to determine the centroid of the reversed hole image.
The position of the coronagraphic hole measured from the pipeline processed image is (73.384, 213.082), while the measured position of the hole from the off-line processed image is (73.273, 213.163). The difference is ~0.1 pixel in the x- and y-positions. This measured offset in hole position is most probably due to the different calibrations performed on the ACQ images. The pipeline processed ACCUM images are not dark or background corrected. The location of the coronagraphic hole in the pipeline on-orbit flat was patched and could contribute to error in determining the position of the hole.
For comparison with the FSW determined position of the hole, the IRAF positions of the hole need to be converted into detector coordinates by subtracting them from 256.5. For example:
In this example, for back-to-back orbit ACQ images n4q832npq and n4q833nzq (obtained prior to June 28, 1998), the difference between the two positions derived for the hole (IRAF and FSW) in detector coordinates is ~0.9 pixels in y and ~0.3 pixels in x. This is much larger than can be explained by summing the errors of the target and hole positions in quadrature, and is due to a bug in the early version of the FSW, which introduced errors in the derived hole positions.
Measuring the Target Position
The IRAF task center can be used to determine the centroid of the target in the NIC2 field of view. For comparison with the FSW positions of the target, the centroid values need to be converted into detector coordinates by subtracting them from 256.5. For example:
In this example, the target position we have measured agrees quite well with that determined by the FSW. The slight differences are most probably due to the different algorithms used to determine the centroids.
The user may wish to recalibrate coronagraphic data using the best reference files currently available (see Section 3.5). Unfortunately, at this time there are no standard dark reference files available from the calibration database for ACCUM and BRIGHTOBJ mode observations (see Section 4.2.3 for a discussion). Some coronagraphic observers obtained their own on-orbit ACCUM mode darks, which are available from the HST Archive and can be used for calibrating the science data. If you have questions about calibrating coronagraphic ACCUM images, please contact the STScI help desk (
Any given on-orbit flat-field observation with NIC2 will include the coronagraphic hole. Because the hole moved with time, it is unlikely that the hole will be in the same location for a given observation as it was in a flat-field taken at a different time. This can have particularly catastrophic consequences for coronagraphic data. It is therefore important to use a flat-field where the hole is in a substantially different location. For Cycle 7 and 7N, the pre-flight flat-fields from thermal vacuum testing show this feature and can be used (however, since the preflight flat-fields were obtained at a different temperature compared to on-orbit, they do not provide a perfect match). A flat-field where the hole has been "patched" can also be used. Patched flat-fields should be suitable for most purposes, although the patches are not perfect, and it is possible that small mismatches may still affect coronagraphic science. Both the pre NCS (patched and unpatched) and post NCS (only un-patched) flat-fields are available at the NICMOS Web page:
Contemporary “Hole” Flats
Calibration using contemporary flat fields would remove the very strong hole-edge gradient resulting from calibration with a non-contemporaneous hole image or with a patched flat. The F160W filter acquisition lamp on/off paired images obtained as part of the Mode-2 target acquisition process can be used to create a flat-field reference file. The hole position in these images is at the same location as the hole position in the associated coronagraphic observations. Reference files created from these images would be appropriate for regions close to the coronagraphic hole. The S/N for these F160W filter reference files would be S/N ~100.
Note: the recommended standard calibration reference files have high S/N (typically ~ 1200) and should be used for regions far from the coronagraphic hole. The Mode-2 acquisition F160W filter lamp and background images will usually contain an overexposed image of the target.
For completeness, the following processing steps to create a contemporary flat field from the ACQ images are listed to assist the user.
ni> hedit n4xj13_f160w_flt.fits[0] PEDIGREE \
>>> "Contemporary 31/08/98" add+
Recalibrating the ACQ ACCUM Images
Some observers have expressed the desire to obtain photometry from the ACQ images of their targets. These images are not dark corrected and will need to be recalibrated. As discussed in Section 4.2.3, however, at the present time there are no standard dark reference files for ACCUM mode images. Appropriate DARK exposures matching the ACQ image exposure times may be available in the HST Archive. If you are a NICMOS GO, please speak with your contact scientist (CS) about dark subtraction for ACCUM images, or contact
We note here that due to a software bug, the exposure time stated in the header keyword exptime is in error for ACQ data taken prior to October 19, 1998. The commanded exposure time listed in the SHP and UDL data file (_spt.fits) should be used for photometry. For example:
The samptime keywords have to be changed in both SCIENCE groups of the raw, uncalibrated imageset (_raw.fits) before calnica processing.
The OPUS pipeline will combine multiple coronagraphic exposures using calnicb. This can cause difficulties, however, for coronagraphic data analysis. If the individual images that form a mosaic require registration, some smoothing will be introduced by the bi-linear interpolation that is used by calnicb to shift the images. Even if the images are well registered to begin with, slight differences in the diffraction pattern due to movement of the coronagraphic hole may cause the calnicb cross-correlation routine to compute non-zero offsets between the various images. This will lead to misregistration and again, some smoothing. One way around this problem is to force calnicb not to shift any of the images before averaging them together. This can be accomplished by adding columns of XOFFSET and YOFFSET values to the input association table (_asn.fits) used by calnicb and setting their data values to zero. However, in general, PSF subtraction is best performed with individual images and not with the mosaic image.
Pedestal Removal
The residual bias or “pedestal” effect was described in Section 4.2.2, and should be removed if possible before analyzing coronagraphic data. The pedsky task described in Section 4.2.4 may not be suitable for coronagraphic images where a bright object dominates the field of view. In this case, the pedsub task may be more appropriate. This method has been used successfully with coronagraphic images from the SMOV calibration program 7052 (1997 July 23). The pedsub task was run with the parameters filter=mask (i.e., applies unsharp-masking filter to remove low spatial frequency information) and doquadeq=no (i.e., do not force quadrant boundaries to be continuous). The images of the stellar PSF in the coronagraphic images of the 7052 data fill the quadrant with the coronagraphic hole and spill over into adjacent quadrants. This constrains the determination of the pedestal contribution in that quadrant. A mask file was created to flag bright pixels (threshold limit= 4.0 cts/sec); the flagged pixels were not used for estimating the pedestal contribution. This mask was added to the mask of known bad pixels included in the DQ image extension, and the pedsub parameters dqon and dqpar were used to tell the task which mask values correspond to pixels that should be ignored when fitting the pedestal.
Removing Other Image Artifacts
The identification and removal of bad pixels, “grot,” and other NICMOS data anomalies are discussed in Chapter 4. For coronagraphic data, the presence of a bright source commonly induces the vertical streaks known as the “Mr. Staypuft” effect (see Section 4.8.4). Because of the bright coronagraphic target, it is not possible to fit simple medians to columns along the entire y-axis length of the image in order to measure and subtract the Mr. Staypuft streaks. It may be possible to remove (or at least reduce) the streaks using a correction image derived from the bottom rows of the image only, far from the coronagraphic target. The first 19 rows are usually the only area of the image that is not intersected by the bright diagonal diffraction spike emanating from the hole. You may use blkavg to average the first 19 rows, and then blkrep to “stretch” this row average to form a 2-dimensional image to be subtracted from your data. Care must be taken to ensure that bad pixels (hot or cold pixels, or “grot”) do not bias the row average: if necessary, interpolate over bad pixels using fixpix.
Figure 5.9: Coronagraphic image, F110W filter.
Star centered in hole with image recalibrated (left), pedestal removed (middle), and with “Mr. Staypuft” striping partially corrected (right). Images are displayed with the same stretch
In Figure 5.9 above, the correction for the Mr. Staypuft bands is not perfect, and shows the limitations of doing this. The amplitude may modulate from one quadrant to another for electronic reasons, and flat fielding also introduces variations along the column so that a constant correction derived from the bottom rows may not be appropriate. Nevertheless, the process results in first-order cosmetic improvement, and may be worthwhile.
Persistent afterimages from previous exposures of the target star (see Section 4.8.1) can also be a problem for coronagraphic images, and users should be aware of this effect.
PSF Subtraction
The light distribution within the PSF and the pattern of light scattered about the hole can change significantly on both short and long time scales, from orbit to orbit and over the lifetime of NICMOS (see Section 5.4 for a general discussion of focus and PSF issues). These changes may be attributed to several factors, including OTA focus variations, thermally induced motions in NICMOS fore-optics, possible changes in the position of the cold mask, and coronagraphic hole motion. All of these changes are related to either the thermal input to the telescope, changes in Sun angle (attitude) and spacecraft roll, or to the dewar thermal short. The PSF also varies as a function of position across the field of view, which may affect the quality of PSF subtractions (see Figure 5.10). In addition, the HST focus position is known to oscillate with a period of one HST orbit. Changes in the focus position are attributed to thermal contraction/expansion of the optical telescope assembly (OTA) resulting from the telescope warm up and cool down during an orbital period.2 These short term focus position variations are usually referred to as “OTA breathing”, “HST breathing”, “focus breathing”, or simply “breathing”.
During Cycle 7 and 7N, the NICMOS dewar anomaly caused the coronagraphic hole to migrate to different locations on the detector. The position of the hole on the detector had been observed to move as much as ~0.25 pixel in three orbits. The movement of the hole was found not to be uni-directional, but rather, the hole “jitters” back and forth along an X-Y diagonal by as much as 0.5 pixel.3
The direct subtraction of two unregistered coronagraphic images that were not obtained back-to-back in the same orbit with a change in roll can yield large residuals. Coronagraphic images need to be subpixel shifted and/or convolved with a Gaussian function to match the observed PSF before subtraction.
The following discussion applies both subpixel shifting and Gaussian convolution reduction techniques to two F110W filter images from the NICMOS calibration program 7052. The data used in the examples below are n45j22lam and n45j23mym. The data were obtained in consecutive orbits with a spacecraft roll of 36 between orbits.
Subpixel Shifting
The images to be subtracted must be registered to sub-pixel accuracy in order to achieve good results. The IRAF task xregister provides one way to determine the offsets between the images, and to shift one into alignment with the other. For the example considered here, a 995 pixel region ([27:121,162:256]) around the hole was used as the area for cross-correlation, and a shift of dx = -0.286, dy = 0.136 pixels was measured between the images. The example shown in Figure 5.10 and Figure 5.11 displays the subtraction of the two images, using both unregistered (left) and registered (right) data. The subtraction using the registered images exhibits a more symmetrical residual light pattern about the hole. Occasionally, cross correlation does not yield the best measure of the image offsets. You may find that “trial-and-error” shifting by various fractional pixel amounts, minimizing the subtraction residuals “by eye,” can yield the best results.
Figure 5.10: PSF Subtraction.
F110W filter images obtained in back-to-back orbits with a roll of the spacecraft between orbits. Direct subtraction of images (left) and a subtraction with the second image shifted to match the first image (right).
Gaussian Convolution
The interpolation introduced by sub-pixel shifting has the effect of smoothing the image slightly. Subtracting the blurred image from the unblurred image has the effect of edge enhancement. One way to suppress high frequency variations in the images is to convolve one or both by a Gaussian filter. This may help reduce residuals in the subtracted image. The IRAF task gauss may be used to smooth the images. The resulting images for two different convolutions, with σ = 0.4 and σ = 0.8, are presented in Figure 5.11.
Figure 5.11: PSF Subtraction.
F110W filter images obtained in back-to-back orbits with a roll of the spacecraft between orbits. Images convolved with a Gaussian function, σ = 0.4 (left) and σ = 0.8 (right) and subtracted.
In this example, the σ = 0.4 convolution results in a slight improvement over the subtractions shown in Figure 5.10. Possibly, this choice is undersmoothing the data. The σ = 0.8 convolution more closely matches the low frequency components. Determining a suitable degree of smoothing will require experimentation by the user.


NICMOS Instrument Science Report, NICMOS ISR-98-015

NICMOS ISR 98-012.

Table of Contents Previous Next Index Print