Terameeting 5th of January 2006

Present: Yannick, Henry, Anis, Laurent, Emmanuel, Jean-Christophe (for the start of the meeting), Fred, Chiara, Mireille, Florence Durret, Mathias (by telecon),

First terameeting of 2006. Yannick provided two fine galettes and a large quantity of champagne. Vive la France!

Skywatcher (Jean-Christophe) Doesn't use any more the database, everything is loaded into memory, and is much faster than before. Uses an XML file which can be generated by a separate program.
 Need to write documentation for Skywatcher.

 Still work do be done on Web-services. Yannick proposes that JC provides a "Web service" interface for scamp. The web service for QF has been finished, but the documentation still needs to be written.

 Jean-Christophe's contract ends by January 15; his new contract with Observatoire de Paris starts at January 16, and lasts until September 14. Jean-Christophe will continue to spend two days per week at TERAPIX and the rest of his time at the Observatoire de Paris.

Processing T03 (Laurent)

 All the stacks in the list for T0003 have been processed.

 There are some images with high skyprobe values (10) and these have been removed from the pipeline; however this high values come from a skyprobe malfunction, rather than lots of extinction

 Yannick proposes that we add these observations to the pipeline, but we won't use these for the intercalibration of the fields (as there are now more and more observations on the same fields).

 Fred suggests that it would be preferable if the 'bad' skyprobe measurements were indicated with, for example, -1. This is indeed the case since the skyprobe value is set to 10 only when the device does not work.

 There could be at most 200 images affected, but probably there is less than this.

 Laurent will start the processing the very-wide data. We will provide a 'release candidate' to the CADC comprising the deep and wide before this (see below)

CADC/TERAPIX interface

 Yannick tells us that he has suggested to CADC that the CDS at Strasbourg makes a copy of the CFHTLS data release which will make it much easier for users in France to download data products

 There has been some problem with the corruption of images in transfer between TERAPIX and CADC; Laurent will insert checksum keywords in the headers for T0004. Fred is already including them in T0003. JJ Kavalaars asked that we use the same software as CADC for the checksum. Fred will check this

 FITS stacks, weight maps, chi2 detection images, images chi2, combined catalogues will be sent to CADC

 However, the per-filter chisquared catalogues will be distributed by TERAPIX

 The metadata, dust images, and individual catalogues will be also be managed by TERAPIX and distributed to the community

 Yannick proposes to make a 'release candidate' comprising the deep and the wide data, which will be made available to the community

 Mireille raised the point the "Deep" and "Wide" catalogues should have the same format and contain the same fields

Quality assessment (Mireille, Yannick)

 Mireille has made 35 stacks (chisquared images and catalogues + tiff C_35 or C_33 colored images) with at least three filters. Henry has also processed all the deep stacks.

 Mireille and Yannick found only 5 CCDs with doubled images in one CCD (3 in u-band, 2 in r-band, concern CCD#8, 28 or 35). It is suggested we use "Extreme Color Objects" as indication of doubly imaged area. Seems interesting idea since each time one extreme object is detected, if this is due to an offset between filter, another extreme with opposite color should be close to it. Should be much faster than auto-correlation.

 Yannick must produce 5 special masks to maks the 5 areas with double objects.

 Laurent will devise an automatic procedure for generating the data summary table for T0003.

 Henry asks if the field-to-field offsets, computed using the stellar locii, are the same between T0002 and T0003 are the same. Apparently the are *not* the same. Mathias will investigate if the offsets between the sloan and the CFHTLS is *less* after the application of these offsets.

 Fred will write a program to compute the E(b-v) from the dust maps produced by Mireille and Yannick to add to the wide survey fields catalogues; this information is already provided by Henry for the deep fields.

 The deep fields have all been processed, there is just a small amount of work to do the D4 field (star-galaxy separation and merging the catalogues).

T0004

 Emmanuel proposes for T0004 to run Scamp between the qfits and Scamp by QRUNID in order to compute the .head. Fixing this problem will be a priority for T0004

SCAMP (Emmanuel)
 Still around 20 pages of documentation to write.

 Emmanuel spend the 10th of January in Grenoble to work on SExtractor in order to implement psf fitting photometry

 Scamp needs to work to accelerate the pattern matching, maybe a speedup of x1.5 is feasible

 Emmanuel has found that the speed of Scamp doesn't scale linearly with the number of processors; right now there is only a small gain going from 4 to 8 processors

WIRCAM (Chiara)

 Bad pixel detection implemented in the WIRCAM reduction pipeline. Unfortunately, we only have engineering grade data to test the procedure. Chiara is about to receive Cuby's data, so she will be able to make good progress on the WIRCAM processing. Cuby is expetced to visit us next Tuesday

 Some work on Weightwatcher to see if the effective area can be computed

 At the same time Emmanuel will make a release of EYE where many bugs have been fixed, with also the retinas for several detectors

 Emmanuel will download Megacam Darks in order to run EYE on Megacam images (so far, we still use past retinas created for EMMI data)

 It seems only the only objection to using an average to combine images now is the detection of satellite trails.

Visualisation tool (Anis)

 Some tests made with displaying large images. The 'double buffering' function allows displaying larger images, but is still limited by amount of physical ram on the machines

 Standard OpenGL library doesn't use floating point extensions

 Tests are underway with the floating point extensions; unfortunately the floating point libraries which anis has found until now are only available for windows.

 There doesn't seem to be any library in Linux which manages floating point textures which is 'card independent', we are forced to use card-dependent library

Hardware

 Yannick suggests that we buy a WIFI card similar to the one which was in the forum and was moved to the library.

 Fred sent mix5 back to the supplier.

 yannix still doesn't work properly; one solution could be to try a different RAID-5 card; will order the adaptec card.

Next TERA-meeting: