performs the bias correction in two steps (see Section 3.4.1
: a “superbias” reference image is subtracted from the science image to remove the fixed bias structure and during-readout dark current1
. The superbias, constructed from individual bias frames obtained three times a week, samples the fixed bias structure at a high signal-to-noise ratio and is free of cosmic ray artifacts. This calibration step is applied to the image before its conversion to units of electrons, so the superbias is in units of DN.
: this step subtracts the bias level from the image, after the image has been converted from DNs to electrons. For pre-SM4 data, it fits the bias level from the physical overscan, and subtracts it from the science data. For post-SM4 data, additional steps are required to also remove the effects of bias shift, followed by the effects of cross-talk, amplifier offsets, and bias stripes.
Each quadrant of the WFC focal plane (A & B for WFC1, C & D for WFC2) has
two overscan regions: a 24 pixel-wide leading physical pre-scan at columns 1 - 24 and a 20 row-wide virtual overscan at rows 2049 - 2068. The physical pre-scan is produced by 24 extra pixels in the CCDs’ serial registers between the readout amplifiers and the imaging region of the CCD. The virtual overscans are obtained by over-clocking the last rows in the imaging regions of each CCD 20 times.
After each vertical row shift, the bias level requires some time to reach its nominal
level. The bias levels in the first 18 columns of the physical pre-scans associated with each WFC amplifier decay quasi-exponentially to their nominal levels. The bias levels in the imaging areas of the CCDs can be safely measured using the six columns of the physical pre-scans adjacent to the imaging areas, i.e., columns 19
24 for amplifiers A and D, and columns 4121
4126 for amplifiers B and C. For more information, see the 2001 JHU-ACS internal report by Sirianni, Martel, & Hartig, available at:
The virtual overscan is not used to estimate the bias level. This region exhibits
large scale structure that is quadrant- and gain-dependent. Moreover, this region can be contaminated by deferred charge due to degradation in the parallel charge transfer efficiency.
The WFC bias frames show small differences between the bias levels of the
physical pre-scans and the imaging region of the CCD (Sirianni, et al., 2002, HST Calibration Workshop, STScI, page 82
). These bias offsets vary from amplifier to amplifier and they can be as large as 3.5
DN. If these offsets were constant, a full frame bias subtraction (doBias
) would remove any differences between the pre-scans and the imaging region. Unfortunately, the offsets show random variations of about 0.3
DN that may be caused by interference between the WFC integrated electronics module and the telescope and/or other science instruments. The accuracy of the bias level subtraction in a single quadrant is limited by this random effect. Consequently, sky background levels often appear discontinuous across the boundaries of adjacent image quadrants after calacs
processing (Figure 4.1
). Automated photometry of point or extended sources that span the quadrant boundaries should therefore be considered suspect. In such cases, measurement and subtraction of the sky background levels in each quadrant separately is recommended.
Since Servicing Mission 4 (SM4), the WFC bias frames exhibit two-dimensional
spatial gradients of 5 DN
DN within each image quadrant (Figure 4.2
). These gradients are stable within the time spanned by each superbias reference image, and so they are completely removed (along with other fixed pattern noise) in the doBias
step of calacs
. These gradients are characteristics of the dual-slope integrator (DSI) implemented in the replacement CCD electronics to reduce the noise incurred during pixel sampling. The gradients are caused by slow drifts of the bias reference voltages during and after the readout of each row of pixels.
These gradients were not produced by the pre-SM4 CCD electronics, which used
the clamp-and-sample technique of pixel sampling. The replacement electronics also offer the clamp-and-sample option and, consequently, gradient-free biases, but this option increases the read noise of the images by about 0.5
. Consequently, the DSI is the default mode for WFC operations.
The linear stretch is identical, and is equivalent to σ
of the single bias. The local pixel-to-pixel noise (including both read noise and striping noise) of the superbias, which comprises 16 biases (including the one shown on the left) is a factor of ~3.6 times lower than the noise level of the single bias.
The ACS CCD Electronics Box Replacement includes a SIDECAR
Application-Specific Integrated Circuit (ASIC) that exhibits a low frequency noise (1 mHz to 1 Hz) on the bias and reference voltages it generates for the WFC CCDs. This noise contribution does not matter for bias voltages going to the CCD since it is canceled out by correlated double sampling (CDS). However, there is one reference voltage from the ASIC that is used to offset the signal applied after the CDS stage. Here, the noise does not cancel out, and manifests as a slow moving variation of the baseline. In practice, “striping
” is observed in all post-SM4 WFC images that is virtually uniform across both amplifier readouts (the entire 4096 columns) of each WFC CCD (Figure 4.3
Because of the uniformity of the striping across WFC rows, it is straightforward to
characterize and remove this low-level 1/f
noise from WFC bias frames. The amplitude distribution is well fit by a Gaussian of σG
= 0.74 e−
with an enhanced negative tail, giving an overall σ = 0.9 e−
. (See Figure 4.4
). This is under 25% of the WFC read noise. Averaging N bias images reduces the 1/f
noise by nearly a factor of N1/2
, so the total noise in the post-SM4 WFC superbias reference images approaches pre-2007 levels. The striping is not well estimated by the limited WFC overscan regions alone, complicating stripe removal in non-bias frames. A WFC superdark reference image (average of ~24 darks) is effectively stripe-free, but science targets rarely comprise so many exposures.
HRC images were read out using only one amplifier. Each image has three
overscan regions: a physical pre-scan and overscan of width 19 pixels at columns 1
19 and 1044
1062, and virtual overscan of width 20 pixels at rows 1025
1044. The first 10 columns of the physical pre-scan exhibit bias-settling behavior similar to that described for the WFC quadrants. The bias level is therefore measured in the last six columns of the pre-scan. There is not a significant difference between the bias levels in the pre-scan and imaging regions of the CCD.
WFC images obtained before SM4 showed intermittent bias variations of a few
tenths of a DN during readout. Bias frames occasionally exhibited horizontal bias jumps in one or more quadrants that lasted for several hundreds of rows (Figure 4.5
). The probable cause of these jumps was electronic interference from other scientific instruments and/or spacecraft activities. There is no automatic detection of these bias jumps within the calibration pipeline. Bias jumps at the sub-DN level are not important for most science applications, but users should be aware of their possible existence in their calacs
The DSI mode (Dual-slope Integrator, implemented in the replacement CCD
electronics during SM4) of WFC operation induces a signal-dependent bias shift whose cause is closely related to that of the bias gradient described in Section 4.2.1
. The DC level of the DSI mode is sensitive to changes in the CCD output voltage in such a way that the pixel bias level is shifted positively by 0.02%
0.30% (depending on the amplifier) of the signal from the previously integrated pixel. This phenomenon is well characterized for ACS/WFC full frame images and can be analytically removed using a parametric algorithm described in Golimowski et al.(ACS-ISR 2012-02
). The calacs
pipeline now performs this correction, only for ACS/WFC post-SM4 full frame images, during the doBlev
Before SM4, the bias reference frames for science images obtained with WFC and
HRC sub-array readout modes were simply extracted from the appropriate locations of the full frame bias reference images, provided the sub-arrays did not cross quadrant boundaries. Tests showed that sub-array science images that were calibrated with the relevant extracted regions from a full frame reference image were just as good as those calibrated using reference images with the same sub-array readout patterns. Users were advised not to use sub-arrays that spanned amplifier quadrants because doing so required the procurement of single amplifier sub-array bias images at the expense of the users’ observing time.
Unfortunately, this convenient use of full frame reference images cannot be applied
to post-SM4 sub-array science images because the two-dimensional bias gradients imposed by the DSI (see Section 4.2.1
) are dependent on the timing patterns used to read out the CCD. The bias gradients seen in the standard 512 x 512 and 1024 x 1024 sub-array images are significantly different from the gradients seen in the full-quadrant and full frame readout modes.
is designed to perform bias subtraction of both pre-SM4 and post-SM4 WFC sub-array images; no special directions for the calibration pipeline are needed. The pipeline initially performs a search for contemporary superbias images with the appropriate sub-array dimensions and, if unsuccessful, reverts to the pre-SM4 procedure of extracting the corresponding region from a contemporary full frame superbias.