Elixir Detrend Data notes

Since the Elixir system started at CFHT, there have been a variety of changes to the recipes used to create the detrend data, and also changes to the camera which are reflected in the detrend data. On this page, we note any ways in which the Elixir processing or the status of the camera have significantly affected the data.

1999 July A major change to the CFH12K chips. To date (July 2002), no Elixir data has been created for the single semester prior to this major overhaul.

2000 May 25 The primary amplifier in chip 03 died. The readout was switched to the secondary amplifier, resulting in an X-direction flip of the pixels in the data files, and low-level changes in the gain & bias. All detrend data, including masks, but excluding the scattered light map for this chip are affected. Make sure to use the mask 00Ak02.mask.0.03.00.fits after this date.

fringe versions We are on the second major version of the fringe creation recipe. The first version used a single median image to correct both large-scale and small-scale structures: these are the fringe-1.0 images. In 2001 September, we implemented the mode / fringe correction system (see Fringe Creation): these are the fringe-2.0 & fringe-2.1 images. All data distributed before 2001 September 22 used fringe-1.0, while all data since then has used fringe-2.X. The minor numbers (2.0 vs 2.1) indicate a change in the software used to measure the fringe patterns and construct the master fringe frame. This difference has relatively little impact, though the 2.0 version resulted in certain I-band fringe frames having sky levels which were clipped at -1000. All of the affected images were re-distributed using fringe-2.1 in 2002 January. All data distributed after 2002 February 01 used fringe-2.1.

photflat versions We have used two variations on the flat-field photometric correction image (see discussion in Scattered Light ). The first version used an ad hoc image created by covering the bottom of the mirror covers and scaling the resulting (vignetting) pattern to minimize the standard star scatter. The improved correction was determined using stellar photometry from multiple images at offset positions and construction a correction frame to minimize the scatter. We call the first version 'scatter-A.0' and the second 'scatter-B.0'. The flat-field images corrected with these scattered light frames are noted as 'photflat-A.0' and 'photflat-B.0'. All data distributed after 2002 July 15 uses scatter-A.0.

There are two ways to check which scattered-light correction was applied to the flat used to process your data: from the flat-field images distributed with your QSO distribution packet, or from the CFHT Elixir Web pages. In either case, the first step is to identify the name of the flat-field used on your image. To do this, look at your image header. There are a variety of image processing HISTORY statements written by the FLIPS processing system. Look for the line that identifies the flat-field image:

 HISTORY imred: Flat-field: * 1.0 / 02Bk01.flat.I.00.01                         

This says the image was normalized by the flat-field image 02Bk01.flat.I.00.01. The first part of the line shows that no re-normalization was applied; this is generally the case for Elixir-processed data since the registered flats are defined to have normalization set to 1.0. The name of the flat-field image defines 5 quantities separated by dots: the camera run ID (02Bk01), the detrend type (flat), the detrend filter (I), the CCD number (00), and the version number (01). See top for more details on these definitions. Having identified the flat-field image, it is nessecary to determine how that image was processed.

All QSO-distributed data comes with the associated detrend frames. If these are conveniently available, examine the header of the appropriate flat-field frame, in this case 02Bk01.flat.I.00.01.fits. If the image has been corrected for scattered light, it will include a comment in this form:

 COMMENT  Scattering Correction B.0 applied                                     

The alternative method is to find your flat-field in the CFHT detrend database tables on the Web. Start from the table of camera run IDs. Select the appropriate run ID (in this case 02Bk01) and you will see a list of all possible detrend types available on the web. Select the 'all' entry for your flat & filter, and you will see a table of all versions of flat-field images for this run ID, type and filter, along with a comment. The comment will either be 'elixir' for the uncorrected images, or it will indicate the version of the scattered light correction (photflat-A.0, photflat-B.0).

 To correct images already flattened with either uncorrected flats or
 the old scattered light version (scatter-A.0), follow the discussion
 We have an uncorrected flat F 
 We have a scattered light image version A: SA
 We have a scattered light image version B: SB
 The corrected flats FA, FB are created with:
 FA = F - SA
 FB = F * SB
 Given a raw image Io, it is flattened by doing: Ix = Io / Fx where Fx
 may be one of F, FA, FB, and the result is called I, IA, or IB:
 I   = Io / F
 IA  = Io / (F - SA)
 IB  = Io / (F * SB)
 So, if you have I and want IB, that is straightforeware:
 IB = I / SB
 A bit more tricky is going from IA to IB:
 IB = IA * (F - SA) / (F * SB) = IA * (1 - SA/F) / SB
 This would be fine if you have the uncorrected flats, but it can be
 simplified with a minimal loss of accuracy, as follow.  The difficulty
 in applying the above is that each camera run has a different set of
 flats for each filter, so there will be alot of work identifying those
 flats, downloading them, etc.  To avoid this, we make use of the fact
 that all flats of a given filter are stable across many months at a
 level of about 2%.  Since the error introduced by using the wrong flat
 for F is of the order SA*F1/F2, the amplitude is quite small
 (~5%*2% -> ~0.1%).  Thus, we can construct a correction frame Rab
 using a single uncorrected flat from a single camera run and apply
 this to all camera runs with a reasonable accuracy.  This would be
 generated as:
 Rab = (1 - SA / F) / SB
 and would be applied as:
 IB = IA * Rab
You can download SA and SB from our website. go to:
and you will get a list of both scatter-A.0 and scatter-B.0 images for the B filter (change B to VRI in the above query to get the other filters).