Minutes of SV Meeting in Edinburgh 15/12/05 (Author: S.Dye)

Those Present

Nick Cross, Simon Dye, Alastair Edge, Nigel Hambly, Simon Hodgin, Mike Irwin, Richard Jameson, Andy Lawrence, Nick Lodieu, Phil Lucas, Daniel Mortlock, Richard Mc Mahon, Steve Warren

Presentations

1. Steve Warren

• SV2: better PSFs, better sky subtraction
• sky flatness: using interlaced stacked image, and bin 2x2, 4x4, 16x16. Noise characteristics initially scale with binning as expected, but then departures are seen when binning at 16x16 and beyond.
• sky flatness YJH ~ 0.2 sig, K ~0.4 sig.
• bad channels: typically 1sig above rms of bgnd. (up to 50 adu) MJI says this is not simply a pedestal offset, but often get increased noise levels in elevated channels. nasty effect in interlaced data (1 in 4 pixels affected) arrays 2 & 4 worst. Larger effect in Y & J (since lower sky)
• Bad decurtaining seen in case where m49 in Virgo lies right in centre of array. Decurtaining causes troughs of size 0.6 sigma c.f. sky.
• big blotches appear randomly due to sky subtraction. typically: Y 0.3sig (.5%), J 0.4sig (.5%), H 0.2 sig (.1%), K 0.3sig (.15%) compared to average sky level. MJI expected effect to disapppear with cleaner, later survey data.
• Interference spikes at 128 pix intervals not being clipped out (eg. only appears in 1 in 4 frames). MJI says shouldn't have got through

ACTION SJW to send spiked frame IDs to MJI

• Moon ghosts: bright butterflies (moon 20-30 degrees away, phase 60%), long arcs (moon >~50 degrees away, phase 90-100%), spider shadows (moon >~50 degrees away, phase ~100%).

2. Daniel Mortlock

• bad channel problem causes large fluctuations in no. density of detected objects. channels that are biased low relative to the average sky level show fewer detected objects.
• 10 times as many objects detected in Y cf other bands (mainly as a result of channel prob, but also the butterflies, field lens dirt etc ...)
• Would be nice to have SQL queries written into FITS headers

ACTION NCH to add SQL queries to FITS headers.

• Virgo strip mainly ok but one pointing has bad sky subtraction in Y causing anomalous detections.

3. Simon Dye

• Comparison between manually calibrated LAS SV targets and pipeline calibration shows better agreement with SV2 than SV1.
• Faulty Y band frames spotted due to bad sky subtraction frames. This causes 1000's of false detections and is the cause of the anomalously high number counts seen in Y by Anthony Smith previously.

ACTION SD to send frame IDs to MJI.

• Comparison between SExtracted aperture mags of pipeline processed image and the corresponding aperture mags of the same objects from the WSA shows good agreement. A 1sig scatter of 0.14 in the distribution of detla_mag significances is found. This compares to an analysis where SExtractions from a manually reduced frame were compared to the WSA aperture mags and a 1sig scatter of 1.62 in the distribution of delta_mag significances was found. Implication is that the larger scatter in the latter case is a result of differing dark subtraction, flatfield and sky subtraction (even though same raw science frame used) rather than differences between SExtractor and the pipeline extraction algorithm.

4. Richard Mc Mahon

• UKIDSSLASImageClassifications20051215rgm.ppt: ppt
• UKIDSSLASImageClassifications20051215rgm.pdf: pdf
• Comparison between source classifications in LAS from SDSS (spectral and photom.) with UKIDSS pipeline classifications. SDSS: 0.4% of gals photometrically classed as stars, 4% of stars classed as gals, 23% of qso classed as gals (mainly due to host gals at low z). UKIDSS: 1% of gals mis-classified (as stars and noise), 8% of stars mis-classified, 36% of z<2.3 qsos classed as gals (presumably host resolved), 2/12 z>2.3 QSOs classed as gals.
• Caveat is that SDSS not representative in sense that its biased towards gals not stars.

5. Phil Lucas

• Large scatter (times 2-3) in GPS aperture photometry compared to PSF fitted photometry due to crowding effects. MJI says aperture mags can match PSF mags in terms of scatter if the right aperture is used (scatter seen by PL is because aperture is too large).
• Apparent variation of colour terms between arrays previously reported by PL has now gone away.
• Poor object detection in regions of bright nebulosity. MJI says can't expect pipeline to handle such regions perfectly. Users will have to run their own extractions on pixel data.
• Depth in K less than expected (~1 mag). MJI says probably mainly due to poor ~1" seeing in SV data.

6. Nick Lodieu

• up to ~1mag difference between 2MASS and UKIDSS in H-K colour of L/T dwarfs. SJW says this could be largely due to different filter system
• obtain large fraction of objects with odd colours in various col-col diagrams drawn from merged table. NCH says reason is that merging has 5" matching tolerance to allow for proper motion and therefore will make several misidentifications. user needs to impose a more stringent matching tolerance when selecting objects from the WSA. MJI says can also cut out lots of objects by applying quality control criteria (eg. ellipticity)

ACTION NCH to update WSA documentation to spell this out more clearly. Need a 'Frequently Made Cockups' page.

7. Simon Hodgkin

• new calibration document on CASU web pages
• 70%-80% of nightly zero point RMSs from 2MASS calibration are <3%
• In 600 images over 19 nights have 46 unique standard stars. 'very few' of the magnitudes of these stars fall outside of the 2% photometric accuracy criterion.
• Some stars however show a systematic offset - could be UFTI calibration error (which is only 2% itself) or colour terms
• STH proposes using 2 hour intervals for std field obs rather than 1 hour. This won't impinge on JHK obs.
• on non-photo nights, 2MASS calib good enough - no need to obs standard fields.
• should continue to observe UKIRT FS for independence from 2MASS.
• now need to measure accuracy of photometry in ZY
• will be able to quantify calib systematics in EDR.

8. Mike Irwin

• summary of different pipeline + data versions presented.
• photometric systematics (relative to 2MASS) show distinct pattern across the arrays. In April 05 data, max-min size of resids is ~0.05mags with 2" diameter apertures, ~0.02mag with 4" diameter apertures. This is due to image quality (ellipticity). In May and Jun 05 data, get much smaller difference between 2" and 4" apertures because image quality improved.

9. Nigel Hambly

• suggested quality control of SV data: frame by frame histograms of seeing, ellip & depth
• where to make QC cuts?; 1) post-ingest: flagging as opposed to deletions, 2) pre-release: remove unwanted data from release database product, 3) query time: user filters on QC attributes.

General discussion

• In EDR, could define a 'complete EDR' and 'total EDR', former for areas with full filter coverage, latter for all data, even if not covered in all filters. EDR paper to define both of these. User can select from either at archive end.
• For DXS, could select fields for producing stacked images that get released in the EDR.

ACTION MJI and AE to iterate on this.

• Generally agreed that users should query the EDR by default. To access more, users have to read the relevant documentation that gives full database products. This is easy for WFAU to implement.
• possible parameters to check for QC: 1) SKYSUB - currently embedded in a string in the FITS headers. 2) Various others ....

ACTION NCH to dig out SKYSUB from FITS header and incorporate into Multiframedetector as an attribute

ACTION SJW to send list of parameters to MJI

ACTION MJI to return their explanation

• data QC tests: seeing, trailing, depth not reached, missing arrays, bad channels, moon ghosts, high airmass, varying sky level between arrays.
• Generally agreed that only 05A should be released in 'complete EDR' and that the 'total EDR' includes 05B as well, even though an earlier version of the pipeline applies to 05A data.
• processing QC tests: astrometry accuracy, photometric accuracy,