ISO Spectral access and QUANTITY discussion

[posuna at iso.vilspa.esa.es]

Hi David,


we do not pretend this to be the solution for any quantity, but in our
view it is already an advance to be able to model the access in the case
of spectra where, as you say, by knowledge of the problem at hand we
know that the only parameters involved in the transformation have to be
C and LAMBDA.

The "free" parameters involved in any dimensional analysis are the
fundamental constants. It is easy to make a dimensional equation have
infinite solutions just by adding the wrong parameter in the equation.
Thus, if you add, e.g., the electron mass to our SWS/LWS equation, you
will have infinite solutions that solve the system, as there will be
more unknowns than equations. 

That's why dimensional analysis always involves some knowledge of the
problem, to decide which funfamental constants translate your problem in
"dimensions X" to your problem in "dimensions Y".

Even though it is not the "panacea", it seems to work OK for the
spectra, which we believe is already an asset, and it could also work
for other things, provided the correct "protocol" is defined (in our
case, the"protocol" is our modified version of the SIAP for spectral
access) and there is some prior knowledge of the problem (in our case,
the "knowledge" of the problem is summarised in a sentence: "We are
working with wavelength versus flux spectra").

As for the case where an axis is transformed in any non-linear form, the
dimensional equations will not work. Transformations in axes vary from
simple "log" to (e.g.) axes moving in a helix at non-constant speed. We
believe that however we ever build the IVOA, we will never be able to
tell a machine the infinite possibilities on these things. I would vote
then for simple things first, complex later (maybe never, as the end
user will always be a human being, and -in the end-, he/she will have to
have, certainly, some knowledge base to make any sense of the data we
provide him/her with).

Anyway we are happy to see that you seem to find useful this approach to
the spectrum case, as it was our goal in this case.

We could discuss these issues in more detail if you want offline.

Thanks for your mail. 


Best regards,

Pedro





On Mon, 17 Nov 2003 10:14:58 +0000 (GMT)
David Berry <dsb at ast.man.ac.uk> wrote:

> Pedro,
> 
> 
> >  --------------------------------------------
> >  Use of dimensional equation for the metadata
> >  --------------------------------------------
> >
> >  Our data fluxes, in the case of LWS, are given in "w/cm^2/um",
> >  whereas the data for SWS are given in Jy.
> >
> >  We thus need a way to convert from one to the other. Let's try
> >  to convert the LWS spectrum units (w/cm^2/s) to the SWS spectrum
> >  units(in Jy) and then display both in the same browse tool.
> >
> >  According to our server, the SWS flux (in Jy) dimensional equation
> >  is:
> >
> >  [SWS] = MT-2
> >
> >  and the LWS (w/cm^2/um) dimensional equation is:
> >
> >  [LWS] = MLT-3
> >
> >
> >  So, to represent one with respect to the other, we will have to
> >  divide both dimensional equations:
> >
> >  [SWS]     MT-2
> >  ----- = ------- = LT
> >  [LWS]   ML-1T-3
> >
> >
> >  as the only quantities we can use in the transformation are those
> > coming from the conversion between "wavelength" and "frequency", to
> > go from one to the other we will have to use a combination of
> > "LAMBDA" and"c" (wavelength and light velocity).
> 
> I can see this works nicely for the case of this spectral axis
> transformation. How would it work in the case of an arbitrary
> Quantity? In the above, because of your knowledge of the problem, you
> knew that the dimensional inequality between SWS and LSW needed to be
> balanced by some combination of "LAMBDA" and "c". I am wondering how
> this would work in the general case. If you know the MLT dimensions of
> 2 quantities, what would the general algorithm be for determining what
> other values needed to be used to equate the dimensions?
> 
> David
> 
> ----------------------------------------------------------------------
> Dr David S. Berry    (dsb at ast.man.ac.uk)
> 
> STARLINK project		 |	Centre for Astrophysics
> (http://www.starlink.ac.uk/)	 |	University of Central Lancashire
> Rutherford Appleton Laboratory	 |	PRESTON
> DIDCOT				 |	United Kingdom
> United Kingdom			 |	PR1 2HE
> OX11 0QX
> 


-- 

Pedro Osuna Alcalaya


SOFTWARE Development Group
XMM-Newton Science Archive				
e-mail: Pedro.Osuna at esa.int
Tel + 34 91 8131314  				

European Space Agency
VILLAFRANCA Satellites Tracking Station
P.O. Box 50727 
E-28080 Villafranca del Castillo
MADRID - SPAIN