VOPlot - Tool for plotting VOTables released.
dtody at nrao.edu
Tue Feb 18 11:36:22 PST 2003
While at ESO a couple of weeks ago the topic of spectra came up in
connection with SIA - the ESO folks were interested in this as well.
Images and spectra are two sides of the same coin. SIA could be extended
to prototype some sort of simplified uniform access to spectra, or we
could do something new. Both images and spectra should be dealt with as
components of VO data access.
The main concern I have with this is the complexity of the problem.
If we try to compose spectra within a VOtable, even for a prototype, we
will need to develop a general multiwavelength data model for spectra,
which is a daunting task. The problem is not just 1D spectra: we have
images, 1D and 2D spectra, and spectral data cubes to deal with, so it
would be hard to even prototype this in XML without addressing the data
model problem. We need to do so anyway of course, but it is hard to see
us doing much without first developing some infrastructure first, such as
standard way to represent structured data and data models in XML/Votable.
One approach would be to simply go ahead and do so, developing a prototype
representation for images and spectra in VOtable.
An alternative short term solution might be to devise some prototype
FITS representation for spectra, using VOtable to describe the metadata
as in SIA currently. We already have formats for images and image cubes
in FITS. A format for spectra could be devised as well which could be
used by services to return data to a VO client. This could allow us to
do something useful now, similar to what was done for SIA, while the VO
data model infrastructure is developed.
Both approaches would work. We would need to explore this further to
decide which way to go.
Note there is already a related effort underway to develop a WCS standard
for spectra (Greisen, Valdes, et al). This already exists in draft form,
and addresses a major portion of the problem.
This is a data access problem which we should up within the DAL subgroup.
We can discuss it in Cambridge if not sooner. It should be integrated
into an overall plan for the next phase of DAL development.
On Tue, 18 Feb 2003, Niall Gaffney wrote:
> This brings me to a soapbox of mine. Being a fomer spectrscopist I know
> that there is no standard way to store even 1d spectra in a FITS file.
> Tables of varying formats and columns, 1d images with WCS. I also know
> that many users simply take the format and dump it to ascii columns to
> work with.
> All that said, will there be a standard VOTable way to write a 1d
> spectrum or spectral energy distribution? At first, people want to take
> the FITS way they know how to work with and implement or embed in a
> VOTable, but that would leave us in the same state we are today with
> FITS. For example, SpecView, a FITS spectral plotting and analysis tool
> from Ivo Busko at STScI, has a different FITS reader for each of the
> different instruments on HST that measure spectra and a factory to get
> the correct one based on a rather knowedgefull scanning of the FITS
> headers and table column names A cascade of java try and catches are
> the key here, which is not really the way to go for a general tool.
> Unifying this in the NVO (with a standard set of UCDs to lead you to
> the correct columns and standardized descritpion of the units in each
> column) would be fantastic.
> I don't think we can hope to use VOTable for 2d or 3d data, as that
> really is to big for ASCII data and should be binary (but there should
> be a VOTable way to describe how to read and interpret the units of the
> binary data). But to do this for 1d and SED data would help
> significantly and enable things like Simbad and NED to do SED exports
> for cross-IDed objects in their catalogs rather than making the client
> do the join by getting selected photometry from each catalog.
> The underlying short term motivation for this is to try to get a VOTable
> reader into SpecView
> (http://www.stsci.edu/resources/software_hardware/specview) to work with
> VO data. I'd really like to be able to look at a VOTable and look for
> the UCDs for Wavelength, Flux, and Errors (as well as the units for each
> in standard formats) to then be able to load these spectra into SpecView
> for analysis, overplotting, combining, and/or fitting. Additional
> metadata like resolution for the points would be great, but that goes
> beyond what we do today with FITS in most cases.
> If anyone has any solutions to this or suggestions on where to get
> started on this, I would like to try to get a first cut of a Spectral
> VOTable reader into SpecView sometime soon to allow us to play with
> spectra and SED data in addition to catalogs and images in the VO.
> Michael Harris wrote:
> > Hello
> > This looks pretty interesting. I was wondering if it plots data that
> > is not presented in a table, possibly CDATA or from a stream link.
> > Also, does anyone have any test VOTables that represent actual spectra
> > or event lists so we can see how it works on real astronomical data?
> > Thanx
> > m
More information about the votable