Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.eso.org/~qc/concept/MEF_QC.pdf Дата изменения: Mon Aug 7 11:09:40 2006 Дата индексирования: Tue Oct 2 01:19:40 2012 Кодировка: Поисковые слова: п п п п п п п п п п п п п п п п п п п р п р п р п р п р п р п

QC process extended to multi-extension files
Reinhard Hanuschik, 2006-08-03 This document collects novel concepts developed and discussed by the QC working group for multi-extension QC: Wolfgang Hummel, Mark Neeser, Reinhard Hanuschik, Burkhard Wolff Issues: · creating QC reports · display products · certification process · QC parameters: global and detector parameters Principles: · information on demand · as much as possible rely on auto-evaluation, scoring etc.

Prologue: Keywords in MEF files
· · all keywords required for association must be present in primary array; they describe the file, not an extension; this comprises raw and product files only the following keys go into extensions (and therefore come 16x or 32x): o PRO REC keywords (if PARAMs are specific for an extension) o QC parameters o DET keywords (e.g. TEMP)

1.
issue: · · ·

Creating QC reports for MEF files
procedure has to be enabled to read all QC1 parameters from extensions calculate if necessary non-existing QC1 parameters per extension calculate in any case global QC1 parameters as average/stddev from per-detector QC1 parameters

enabled: measureQuality ­a to prepare for calling QC procedures in parallel (and if necessary also per detector): aligned to processAB · · · need a TQS module qc1Reader to be called within/before QC procedure, to read QC parameters from header into procedure need another TQS module qc1Writer to collect all QC1 parameters and ingest them useful for all instruments, not just for MEFs!


qc1Reader

QC procedure

qc1Writer

novel · · · ·

approach: python scripting language, capable of doing arithmetics on command line wrapping pymidas, pyraf etc could be useful for MEFs since simple and powerful looping etc. powerful graphics

Issue with MIDAS procedures: · to be enabled to execute in parallel, loading/storing display info (load/ima) is an issue · not simple to implement · omit if possible? alternatives?

2.
· ·

Display, certification
need to be able to display all or some extensions (raw and products) QC procedures should be configurable to support all/ext

displays: PANORAPIX, rtd issue: displays don't show all pixels (not new but getting more serious problem) use zoom functions concepts: · master QC report, detail (extension) QC reports o detail reports could either come per detector or per QC1 parameter o opened on demand only details QC details QC report details report report master QC report

·

ratio to reference o reduces information o focuses on changes, ignores what is the same as before


·

VIMOS implementation: o 4 quadrants simultaneous implemented o reference products used

3.

Scoring

Necessary step towards more automated certification Idea: · define key QC1 parameters, check new values against trending for those · evaluate compliance: o if new within defined range, ok o if outside: nok · user may focus on nok ones, while ok is certified automatically in `confidence mode' Required: · configure set of key parameters per product · configure comparison mode o fixed: check against configured value (useful if nominal/reference value exists) o dynamic: calculate average and scatter from sample, check if outlier (useful to detect sudden changes) · should be supported by trendPlotter in CERTIF mode
measureQ qc1Ingest trendPlotter scoreProducts score certifyProducts

report/ alert

·

·

score: for each key parameter get flag ok/nok 0 or 1; for instance: o 01000 for 5 key parameters with 4 times ok and once nok o have configured which result sets are considered fully ok: green light possibly ok, need check: yellow light likely nok, need check: red light conceptually related to OB grading


· · ·

color codes used e.g. on AB monitor; user then wants to investigate only the yellow or red ones; the others go certified (in JUMP mode) can be easily scaled to MEF case, to focus on `suspicious' extensions can have a reporting system

Example for scoreProducts configuration: BIAS Parameter RON_raw RON_master DATANCOM green yellow red

Check mode dynamic dynamic fixed 000 100, 010 else

Check value Median +/- 3sigma Median +/- 3sigma >= 5

4.

Trending, QC1 database

Issue: · MEF files have QC1 vectors rather than parameters · we may want to keep their coherence, rather than treat them independently store · · · `QC files' in cdb? not on-line available but available for trendPlotter QC1db: only mean values and scatter? needs det_id values plus `all'

store QC parameters in the traditional way plus `all_scatter'; seems doable

for each QC1 parameter, define avg and scatter (global QC1 params): · helps to disentangle coherent (all change by 5%) and incoherent changes (det16 changes, the others stay) · follow in trending mainly the global QC1 params · any det-trending plots? issue of complexity!