Team Only Data Release 3 - GOODS-N Complete Dataset v1

This release provides the completed GOODS-N dataset of the HDUV survey, i.e. 8 out of 8 pointings. Nothing was changed relative to the previous release in terms of data reduction steps, and we anticipate that these are the data to be used for first scientific analyses by the team.
We do not plan to publicly release these data until a few months after the full survey is complete (which should be end of November).

GOODS-N 60mas (drizzled to same pixels as 3D-HST images)

  • hduv_goodsn_60mas_3dhst_f275w_v1.0_drz.fits.gz   F275W science image on 3DHST pixel frame
  • hduv_goodsn_60mas_3dhst_f275w_v1.0_rms.fits.gz   F275W rms map
  • hduv_goodsn_60mas_3dhst_f336w_v1.0_drz.fits.gz   F336W science image on 3DHST pixel frame
  • hduv_goodsn_60mas_3dhst_f336w_v1.0_rms.fits.gz   F336W rms map

GOODS-N 30mas (cut to enclose full HDUV survey footprint)

  • hduv_goodsn_30mas_f275w_v1.0_drz.fits.gz   F275W science image trimmed
  • hduv_goodsn_30mas_f275w_v1.0_rms.fits.gz   F275W rms map
  • hduv_goodsn_30mas_f336w_v1.0_drz.fits.gz   F336W science image trimmed
  • hduv_goodsn_30mas_f336w_v1.0_rms.fits.gz   F336W rms map

You can download all the files using this script or access them individually here.


Zeropoints:

As a reminder, the Galactic dust-corrected WFC3/UVIS AB zeropoints to be used in GOODS-N are:

GOODS-N F275W 24.065
GOODS-N F336W 24.615







Team Only Data Release 2 - Complete Epoch 1

This is the second distribution of HDUV imaging data. It contains background-subtracted, drizzled F275W and F336W images of the first complete epoch of the HDUV survey.

We are now about half-way through the survey, meaning that we got one epoch of F275W and F336W imaging data over all 5 pointings in GOODS-S, and we got 4 out of 8 pointings in GOODS-N completed.

The raw files were CTE-corrected, and WCS-aligned, before drizzling onto a common output grid. Compared to the first image release we now perform a better destriping and background subtraction of the data. Specifically, after a first pass reduction, we mask all objects and create a super sky dark image. This is shown in the figure below. The sky dark is subtracted from all images before the final drizzle iteration.

Again, the WCS registration was performed relative to the 3DHST HST images (Skelton et al. 2014; available here), and the GOODS-N F275W data also contains all the CANDELS imaging taken over this field. Those images were aligned using shifts determined by the F350LP exposures taken just prior to the F275W frames.

We now also correctly stack the data using an inverse variance map weighting (including the postflash electrons), and we provide the corresponding RMS maps of the final drizzled images in the release. These were renormalized to account for correlated noise using the standard correction factors from Casertano et al. (2000).

We provide images at 60mas and 30mas pixel scales. The 60mas images are drizzled to the same frames as the 3D-HST data, while the 30mas images were trimmed to avoid excessively large file sizes.


The images in this release are thus:

GOODS-S 60mas

  • hduv_goodss_epoch1_60mas_3dhst_f275w_v0.5_drz.fits.gz   F275W science image on 3DHST pixel frame
  • hduv_goodss_epoch1_60mas_3dhst_f275w_v0.5_rms.fits.gz   F275W rms map
  • hduv_goodss_epoch1_60mas_3dhst_f336w_v0.5_drz.fits.gz   F336W science image on 3DHST pixel frame
  • hduv_goodss_epoch1_60mas_3dhst_f336w_v0.5_rms.fits.gz   F336W rms map

GOODS-N 60mas

  • hduv_goodsn_epoch1_60mas_3dhst_f275w_v0.5_drz.fits.gz   F275W science image on 3DHST pixel frame
  • hduv_goodsn_epoch1_60mas_3dhst_f275w_v0.5_rms.fits.gz   F275W rms map
  • hduv_goodsn_epoch1_60mas_3dhst_f336w_v0.5_drz.fits.gz   F336W science image on 3DHST pixel frame
  • hduv_goodsn_epoch1_60mas_3dhst_f336w_v0.5_rms.fits.gz   F336W rms map

GOODS-S 30mas

  • hduv_goodss_epoch1_30mas_f275w_v0.5_drz.fits.gz   F275W science image trimmed
  • hduv_goodss_epoch1_30mas_f275w_v0.5_rms.fits.gz   F275W rms map
  • hduv_goodss_epoch1_30mas_f336w_v0.5_drz.fits.gz   F336W science image trimmed
  • hduv_goodss_epoch1_30mas_f336w_v0.5_rms.fits.gz   F336W rms map

GOODS-N 30mas

  • hduv_goodsn_epoch1_30mas_f275w_v0.5_drz.fits.gz   F275W science image trimmed
  • hduv_goodsn_epoch1_30mas_f275w_v0.5_rms.fits.gz   F275W rms map
  • hduv_goodsn_epoch1_30mas_f336w_v0.5_drz.fits.gz   F336W science image trimmed
  • hduv_goodsn_epoch1_30mas_f336w_v0.5_rms.fits.gz   F336W rms map

You can download all the files using this script or access them individually here.


Note that the GOODS-S images do NOT include the UVUDF images. The lack of large scale dithers makes those data very difficult to reduce and we found that our dark correction is likely still not as good as the one developed by the UVUDF team (see also: future improvements). In particular, the small dither steps do not allow us to create a sky dark from the UVUDF data themselves. The reduced frames over the Hubble-Ultra Deep field by the UVUDF team can be accessed here.



Zeropoints:

The dust-free WFC3/UVIS AB zeropoints are:

  • F275W: 24.1305
  • F336W: 24.6682

The Schlafly & Finkbeiner (2011) E(B-V) for GOODS-S and -N are 0.0069 mag and 0.0105 mag (see here). These result in a dust correction of 0.0434 mag and 0.0349 mag in GOODS-S and 0.066 mag and 0.053 mag in GOODS-N, for F275W and F336W, respectively.


The Galactic dust-corrected WFC3/UVIS AB zeropoints to be used are thus:

GOODS-S F275W 24.087
GOODS-S F336W 24.633
GOODS-N F275W 24.065
GOODS-N F336W 24.615



Future Plans:

Over the last month we incorporated most of the TBDs from the first release, however, some further improvements can of course be made. In particular, for a v1 dataset we are planning to work on:

  1. 1) CTE corrected darks, and generally improved handling of dark subtraction
  2. 2) refining alignment of individual exposures in a second pass




Notes on Data Reduction

Sky Dark

As we had seen in the previous data release, the CTE-corrected files revealed clear structure in the background as well as clear striping along the readout columns. We now significantly improved upon our previous treatment and now create a super sky dark image from the data directly. Specifically, after a first iteration of WCS alignment, we mask all sources in the image (using the B-band segmentation maps). The first drizzle iteration also masked all the cosmic rays in the images. This therefore allows us to easily create a masked median from all the input images. The resulting sky dark is shown on the left. It nicely captures all the features and subtracting it from the science frames improved the sky by a lot. This completely eliminates the need for an additional de-striping which we had done previously (see bottom of this page).

Ideally, however, this kind of dark image would be subtracted already before flat-fielding. And in a future improvement, we will thus try to create CTE corrected darks from all the DARK frames in an anneal cycle to create our own darks to subtract in the CALWFC3 step. This is effectively what the UVUDF and Siana teams do.







SUPERSEDED: Team Only Data Release 1

This is the first distribution of HDUV imaging data. It contains background-subtracted, drizzled F275W and F336W images of the first two HDUV pointings in GOODS-N.

The raw files were CTE-corrected, WCS-aligned, background subtracted, de-striped, and finally drizzled onto a common output grid. The WCS registration was performed relative to the 3DHST HST images (Skelton et al. 2014; available here). The F275W data also contains all the CANDELS imaging taken over this field. Those images were aligned using shifts determined by the F350LP exposures taken just prior to the F275W frames.

The files in this distribution are:

  • hduv_gn23_uvis_f275w_30mas_v0.5_drz_sci.fits.gz   F275W science frame
  • hduv_gn23_uvis_f275w_30mas_v0.5_drz_wht.fits.gz   F275W weight frame (inverse variance map)
  • hduv_gn23_uvis_f336w_30mas_v0.5_drz_sci.fits.gz   F336W science frame
  • hduv_gn23_uvis_f336w_30mas_v0.5_drz_wht.fits.gz   F336W weight frame (inverse variance map)

The pixel scale of those images is 0.03"/pix, but the distribution also includes images with 0.06"/pix (replace 30mas with 60mas in filename).

Download the files using this script or access them individually here.


For comparison, the reduced frames over the Hubble-Ultra Deep field by the UVUDF team can be accessed here. These have about twice the exposure time compared to our data.




Future Plans (mostly implemented by now):

These images should be science ready. However, we are working on several improvements which include:

  1. 1) improved background subtraction in a second pass
  2. 2) improved de-striping of columns
  3. 3) refining alignment of individual exposures in a second pass
  4. 4) improved cosmic ray identification and flagging of bad pixels
  5. 5) improved inverse variance maps including post-flash background values




SUPERSEDED: Some Notes on Data Reduction

Backgrounds

The direct astrodrizzle stacks of CTE-corrected files reveals clear structure in the background as well as clear striping along the readout columns (see left panel). We performed a simple background subtraction (per amplifier) using a median filtering and also subtracted the median pixel values along the readout direction. This results in the image in the right panel (shown at same stretch). Clearly, some oversubtraction along some readout columns which contain bright galaxies is still visible. We will attempt to remove this in a next release.

This figure shows the bias along readout columns after subtracting a median filtered background in Chip 1. Gray lines show 24 individual exposures, while the black curve shows their mean. Some columns show clear biases of up to 1-2 electrons.