spectrum services 2007

21
VO Spectroscopy Workshop, ESAC 2007 1 EÖTVÖS UNIVERSITY BUDAPEST Department of Physics of Complex Systems Spectrum Services 2007 László Dobos (ELTE) [email protected] Tamás Budavári (JHU), István Csabai (ELTE)

Upload: tuan

Post on 06-Jan-2016

20 views

Category:

Documents


0 download

DESCRIPTION

Spectrum Services 2007. László Dobos (ELTE) [email protected] Tamás Budavári (JHU), István Csabai (ELTE). MOTIVATIONS. SDSS Data Archive Server offers very limited possibility to search among spectra SkyServer has nice search features, but does not contain the spectra - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 1

EÖTVÖS UNIVERSITY BUDAPEST

Department of Physics of Complex Systems

Spectrum Services 2007

László Dobos (ELTE)[email protected]

Tamás Budavári (JHU), István Csabai (ELTE)

Page 2: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 2

MOTIVATIONS

• SDSS Data Archive Server offers very limited possibility to search among spectra

• SkyServer has nice search features, but does not contain the spectra

• Observed spectra are in many formats:– binning– wavelength scale (air / vacuum)– extinction correction– file format

Page 3: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 3

SPECTRUM & FILTER SERVICES

• Database of spectra• Wide range of search possibilities• Preprocessing spectra on the fly• Data analysis• Access to data in many file formats• Plot spectra• Web interface• Web service to program against

• Filter service:database of optical instrument band-pass curves

Page 4: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 4

Page 5: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 5

Page 6: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 6

AVAILABLE SPECTRA

• Repository of observed and theoretical spectra– SDSS DR1-DR5 (1M spectra in DR5 only)– 2dF (250k)– BaSeL stellar library– Bruzual-Charlot stochastic burst model library

(100k)

• 'My Spectrum' is a public repository where users can upload their own measurements or save results of spectrum processing.

Page 7: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 7

SEARCH POSSIBILITIES

• Cone search - search by coordinates and radius

• ID search - if you know the database identifier (ie. from SkyServer)

• Object search - by a list of coordinates• Redshift search• Advanced search - with many basic options

• Region search - gives back spectra of objects in a complex, arbitrary area

• Similarity search - based on PCA

Page 8: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 8

3. REGION SEARCH

• The new Footprint Service allows defining complex regions on the sphere

• Spectrum Services link to the Footprint Service to get region description

• Spectra observed in the given region can be retrieved from the database

Page 9: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 9

Page 10: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 10

4. SIMILAR SPECTRUM SEARCH

• Spectra are expanded on a predefined basisPCA (see Ching-Wa's talk) or model spectra

• Coefficients are stored in a multi-dimensional database

• Database is indexed by special spatial indices

• N-Nearest neighbour search in the coefficient space

• Not necessarily physical, still interesting

Page 11: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 11

GALAXIES

Page 12: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 12

PREPROCESSING TOOLS

• General functions to convert spectra to the same format to use in analysis

• Functions:– Correction for galactic extinction– Rest frame– Normalization– Rebinning– Vacuum - Air wavelength conversion

• Download preprocessed spectra (XML, VOTable, ASCII)

• Store them in 'MySpectrum' for later use

Page 13: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 13

ANALYSIS TOOLS

• Composite spectrum:– average, median

• Convolution with filters from Filter Servicesynthetic magnitude calculation

• Continuum and line fitting

• The new version will be really fast:– Parallel computation on multiple processor– optimized LAPACK libraries

Page 14: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 14

Page 15: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 15

SUMMARY & FUTURE WORKS

• Spectrum database and public repository• Web site & web service• Integration with other VO services:

– Filter Service, Footprint Service

• Preprocessing and analyzis tools

• Improved line fitting algorithm• Make similarity search public• Latest VO standard implementation

Page 16: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 16

URLs OF INTEREST

• Spectrum Serviceshttp://voservices.net/spectrum

• Filter Servicehttp://voservices.net/filter

• Footprint Servicehttp://voservices.net/footprint

Page 17: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 17

Page 18: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 18

Distribution of SDSS objects in the space of SDSS galaxy eigenspectra

Galaxies

Stars

QSOs

Page 19: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 19

WEB INTERFACE

• For interactive use by humans• Concept: try the functions on the web page,

but use the web service for batch processing

• Register / login• Search and display results in a nice format• 'Wizards' to guide users through the pages

of many parameters of scientific functions• Download spectra in XML, VOTable, ASCII• (No implementation of the latest DM yet)

Page 20: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 20

2. SPEEDING UP SPECTRUM SERVICES

• Objective:processing hundreds of thousands of

spectra• Moved to .Net 2.0 - Better computing performance

• Streamlined processing of spectra– Minimal memory usage– No upper limit for number of spectra

• Multi-processor support– parallel execution of processing routines

• Optimized LAPACK integration:– MKL (intel), ACML (AMD)

Page 21: Spectrum Services  2007

VO Spectroscopy Workshop, ESAC 2007 21

4.1 THE ECOEFF SPACE

• DR4 galaxy eigenspectra used as basis(Yip et. al 2002) from gappy PCA algorithm

• All spectra (even stars and QSOs) are expressed on this basis

• Our method: minimal χ2 fitting (with masks etc.)

• Applied to:– SDSS DR5– Charlot's stochastic 100k spectrum set

• Objective: find physical parameters very quickly