Teraminutes for the meeting of Friday 3rd June 2005

Present: Yannick, Anthony, Mireille, Jean-Christophe, Fred, Laurent, Emmanuel, Henry

 Status of the release (Henry, Yannick, Mireille)

 D1,D2,D4 are ready

 D3 i-band stack not ready. Still a very large dispersion in D3 in r-i vs i-z colour-colour plots which have to be investigated: this probably from double images in the i-band image. One solution to provide a fast release is to re-release the T0001/D3 images and use this to make the chisquared images and catalogues.

 In some images the PSF is doubled, and this is just about visible in the output of PSFex, if one knows what one is looking for. Some of these images have been flagged according by CFHT, but not all. Yannick proposes the CFHT meta-data are used to grade all CFHT-grade 5-4 images as "C" in QFITS-in. Laurent will implement this automated grading process, so that it keeps QualityFITS decoupled from a specific CFHT data management.

 W1,W2, et W3 - offsets still need to be calculated. Offsets also need to be calculated for the images which have just gri. Henry will do the offset analysis from catalogues produced by Mireille.

 D1,D2,D3,D4 are almost ready. Henry will attempt to finalise the stacks for the D3 field by using the D3-t02 release.

 Yannick will provide the masks for images for the wide on Monday.

 In summary: the D1, D2, D3 (+D3-i from T0001), D4 and all W1 done un ugriz will be prepared for a "non-release" delivery. Mireille and Henry will show Jean-Christophe where these data are. Jean-Christophe will contact Luc Simard to organise the data transfer and the organisation of the "non-release" data distribution.

 Emmanuel suggests adding flags to the database indicating further quality control flags so that these images can be easily retrieved later and be used to train neural networks to identify problems. A preliminary list has be discussed that will be updated over the next week. The flags to train the neural network should focus on tasks that demand visual inspection. The other flags should focus on the most common defect/features Mireille, Henry or Yannick found on images during the grading phase of QFITS-in or those found by Yannick and Jean-Christophe during the grading phase of QFITS-out.

Other matters: organisational changes at TERAPIX

 Matthias Schultheis has been recruited at CNAP with TERAPIX as part of his 'Tache de Service'. Mathias will work at TERAPIX two days per week, and will most probably be located in Henry's office.

 Fred will become project manager.

 Henry will take full responsibility for all photometric issues related to the survey

 Gilles Missonnier will leave TERAPIX and vacate his office; Fred will take Gilles' office.

 Yannick proposes that Chiara Marmo moves to Fred' office.

 Some work will be done in Fred's and Emmanuel's offices in order to reduce the temperature in the router cabinet. Emmanuel is not happy because there may be some noise or heat coming from the new air-conditioning system that will be on the terrace.

Plans for T0003

 Yannick emphasises the need for Quality assessment (QA) tools. QA turn out to be the most demanding steps in terms of manpower. Automated tools that provide quantitative measurements of output images and catalogues are necessary.

 Emmanuel will try to output the offsets with respect to the reference catalogues

 The problem of detecting 'elliptical' astrometric residuals in SCAMP was mentioned.

 Yannick and all Terapix discussed the planning for the new release. It is decided that T003 should be at CADC by December 15. Goal: about 15000 images to process.

 The following steps are required to process the images: Transfer images /QFITS_in/SPICA/transfer to CADC

 If we can distribute the images automatically to all the machines we can utilise the machines much more effectively. Using all the machines in the cluster we should be able to get 1000/images day.

 Jean-Christophe will make a prototype of a 'web-service' which will allow load balancing across the cluster. The prototype will be presented to Terapix by June 10.

 Another urgent problem is eliminate communication with the CDS for astrometric reference catalogues: during many weeks of QFs it is almost certain that the connection with the CDS will be broken.

 We aim to run the QFs on all images in 14 days. Laurent thinks a 'realistic' limit will be 21 days.

 Downloading images from CADC will take around 30 days.

 Yannnick makes a calculation about the total amount of time required to make T0003, it looks like it is going to be around 4.5 months.

 Yannick suggests a release date of 15 december 2005

 A list of meeting dates / deadlines was proposed by Yannick

 8 June: a tool that list the duplicated images between T0002 and T0003 QFITS-in from CADC. (Jean-Christophe)

 9 june : web service prototype.

 10 june: SExtractor with new error computations. Needed for QFITS-in. (Emmanuel)

 15 june for QFITS_in with parallelisation + SCAMP analysis. Implementation new SExtrator. (Fred)

 14 june meeting to decide what quality assessment tools are necessary for the T0003 release (Mireille, Yannick and Henry).

 ? june for the list of flags we need for QFITs-in comments and to train a neural network (Mireille, Henry, Yannick, Emmanuel).

 20 june images will be transfered [missed that bit]

 Skywatcher (Jean-Christophe) openGS library added

 Defectix (Anthony): nothing new about defectix, this past week most work was done on EFIGI. Apparently haloes around bright stars can be found. Anthony will make a presentation on Monday June 6, at 15.30 about the work he has done for EFIGI.