From: Andrew Fruchter (fruchter@stsci.edu)
Date: Sat Feb 16 2008 - 09:41:10 PST
Hi Kyle,
I would recommend against using the Lanczos kernel for photometry.
It does not preserve flux. The errors are probably small, but
real. I generally only recommend this choice of kernel if you are
drizzling singly dithered images. This is what the Jee et al test
images are -- singly-dithered. What Jee et al did makes sense for
singly-dithered data, but not for the well-dithered data we will have
for lensing work. For well-dithered data, I suggest you use an
output scale less than one (~0.6 for ACS) and avoid the Lanczos
kernel. I would also recommend against this choice for SN
photometry.
The skew seen in the 47 Tuc data is due to a time-dependent skew in
the ACS distortion that is not removed at present. This effect is
probably caused by motion of the chip in the distorted focal plane.
Velocity aberration produces a first order change in location in the
center of the field, and a second order apparent change of scale, not
skew.
Our software group is working on implementing Jay Anderson's solution
to the time-dependent skew, and as soon as this is done and tested
the correction will be made public. If you are using wdrizzle or
straight drizzle, rather than the ACS multidrizzle, you could
probably get rid of this with a little bit of mathematica and some
playing with the coeffs file. Let me know if this is important to
you now. I hope we can have some software that removes the time-
dependence available on a test basis fairly soon.
Cheers,
Andy
On Feb 15, 2008, at 10:12 PM, Kyle Dawson wrote:
> Hello All,
> After some discussion with the GTO team, we decided to re-run the
> image processing one more time. These new data should be
> consistent with the published data and analysis techniques (i.e.
> Jee et al, 2007). Below is a summary of the reduction:
>
>
> Feb 15, 2008 by Nao Suzuki (LBNL)
>
> 1) Summary
> We post a new stacked image, v98, which is drizzled by lanczos3
> kernel following the procedure of the GTO team. This is the same
> reduction used for analysis of the red-sequence scatter in high z
> clusters and the same reduction used by James Jee in his weak
> lensing analyses. In addition, James has generated a set of tools
> that can be used to create a pixel-by-pixel PSF across the full ACS
> FOV for images created with the new multidrizzle parameters.
>
> As a summary, now we have the followings stacked image posted on
> the web
> which can be used for different scientific purposes.
>
> Ben Koester of University of Chicago has a program to generate the
> PSFs from the v98 data. Please contact him if such a PSF model is
> required for your science objectives. The model has previously
> been applied to F775W data for weak lensing measurements (Jee et al).
>
> Ben Koester : bkoester@oddjob.uchicago.edu
>
> Advantage and disadvantage of the choice of kernels are discussed in
> the following papers.
>
> v98 : 0.05"/pix sampling with lanczos3 kernel, pixfrac=1.0
> M. J. Jee et al. : astroph/0710.5560
> v46 : 0.03"/pix sampling with gaussian kernel, pixfrac=0.8
> J. D. Rhodes et al. : 2007 ApJS, 172, 203
> v45 : 0.05"/pix sampling with default STScI multidrizzle parameters
> STScI hand book
>
> Stacked image are posted here:
>
> http://supernova.lbl.gov:8083/HSTclusterSN/(cityname)/stacked/README
> http://supernova.lbl.gov:8083/HSTclusterSN/(cityname)/stacked/v45/
> http://supernova.lbl.gov:8083/HSTclusterSN/(cityname)/stacked/v46/
> http://supernova.lbl.gov:8083/HSTclusterSN/(cityname)/stacked/v98/
>
> WCS fixed and sky subtracted images can be found in the following
> directory so that you can reproduce the same results.
>
> http://supernova.lbl.gov:8083/HSTclusterSN/(cityname)/wcsfixed_flts/
> v46/
> http://supernova.lbl.gov:8083/HSTclusterSN/(cityname)/wcsfixed_flts/
> v46/drz_script_v98.py
>
> Note: Basic process such as alightment, bias jump correction,
> sky subtracion are the same which you can find reports in
> earlier versions. (python 2.5)
>
> 2) What's new
> a) Why lanczos kernel?
> b) Zeropoint offset adjustment
> c) Issues of distortion correction
> d) Issues of photometric zeropoint
>
> a) Why lanczos kernel?
> M.J. Jee et al. reports that lanczos3 kernel produces a sharp PSF
> without introducing correlated noise, and he made PSF models
> publically
> available. Once you measure star PSFs from an image, we can deduce
> PSF
> at any given pixel pisition using M.J. Jee's PCA-PSF. We expect this
> would be useful for weaklensing study or precise photometry of
> galaxies.
>
> However, lanczos3 is known to introduce ringing and cosmic ray
> rejection
> needs extra masking than others, and it destroys the core of the
> bright
> stars. We exchanged info with STScI people and believe we have
> minimized these side-effects. It may be useful to cross-compare
> photometry on several bright stars across several multidrizzle
> kernels as a sanity check.
>
> See a figure on the same image but different kernel:
> http://supernova.lbl.gov/nsuzuki/acs/lanczos/
>
> The following is a quote of e-mail exhange with STScI help desk.
>
> >One thing I have found from drizzling with the lanczos
> >kernel as opposed to the square kernel (in the final step)
> >is that it creates a narrower PSF, but it also introduces
> >a lot of negative pixels. I have also heard that it
> >can cause a 'ringing' structure around cosmic rays.
>
> b) Zeropoint offset adjustment
> ACS zeropoint has moved because of the CCD temperature has changed
> on July 4th 2006.
>
> http://www.stsci.edu/hst/acs/analysis/zeropoints
>
> Please note the number found on the web above sometimes changes.
> The followins is a snap shot of what we found in Dec 2007.
>
>
> Before July 4th 2006 Temperature=-77C
> AB Vega
> F775W 25.678 25.291 <= This is what we use
> F850LP 24.867 24.347 <=
>
> After July 4th 2006 Temperature=-80C
> AB Vega
> F775W 25.665 25.277
> F850LP 24.842 24.323
>
> Since the majority of our 2005-06 observation was taken before
> July 4th 2006, we have rescaled the images to match 'before 7/4/06'
> zeropoint for a few clusters with data after 7/4/06. The -77C
> zeropoints should therefore be correct for all of the data.
>
> Please use Temperature=-77C zeropoint for your photometry.
>
> c) Issues of distortion correcion
> We have reported the image offsets of order of 0.2-0.3pix in the
> past, although we didn't know the reason. We have investigated
> 47Tuc which is a dense stellar field for calibration and found
> it is a function of rotation. The structure of distortion correction
> error shows a skewness and similar to velocity aberration. I expect
> slight adjustment of velocity aberration can potentially fix the
> problem. However, this adjustment has not been applied to the data
> and we do not plan to make this
> correction in the near future.
>
> http://supernova.lbl.gov/nsuzuki/acs/distortion/
>
> d) Issues of photometric zeropoint
> COSMOS projects report there exists a photometric offset between
> ACS and ground based data such as SDSS.
>
> See Capak et al. 2007 ApJS, 171, 99
> http://adsabs.harvard.edu/abs/2007ApJS..172...99C
>
> We began to investigate if this is happening to our filters,
> F775W and F850LP. If anybody has a good photometric field between
> HST ACS or NICMOS which can be tied to ground based data, please
> let us know.
>
> For questions or requests, please contact:
>
> Nao Suzuki
> nsuzuki@lbl.gov
> (510)486-5218
>
>
>
This archive was generated by hypermail 2.1.4 : Sat Feb 16 2008 - 09:45:03 PST