[Passband] a useful self-contained model?

[Anita Richards]

> > 2.- Names: There are various names for each bandpass; the obvious one is
> > the name of the filter
> > as known within the observatory (e.g. F555W for HST/ACS), but sometimes
> > an alternate name (e.g. "V") is used to indicate that the filter is
> > similar (but almost never identical) to a certain standard filter. I do
> > not like to see a filter named V when it is not really V.
>
> Name sounds good.  I hope that UCDs might cover this also; I think each Passband
> will need to hold a list of UCDs that it might represent.  For standard filters
> (such as V) the transmission curve can be soft-coded and included in a library.
>    Perhaps we should have a way of saying 'nearly-V', but hopefully UCDs will
> cover this.

I thank that is much too fine-grained for UCDs.  One of my continual moans
about current tools is that they only find data if it is described by a
UCD e.g. FLUX_5GHz and you can miss things far too easily.  I think that
for data which are not fully described we need a generic default
V(optical), K(IR), K(Radio) (beware - letters are recycled!) but even if
we don't run to a full bandpass look-up table every bandpass must be
described by a start and stop wavelength(or other approved unit) or a
centre+extent (supplied by default if necessary for generic V(optical)
etc. along with the zeropoint etc).  T

>
> > 3.- Min/Max: in the RM (if I remember) these two numbers are not
> > mathematically defined
> > but are chosen by the data provider in a way to ensure that no search by
> > fravergies
> > will miss the given filter;
>
> OK - this sounds important when Passband is part of the metadata.

But they may be general for a data collection, e.g. a catalogue may
contain data taken between 1.3 and 1.7 GHz but any individual data set may
have a passband of only 0.016 GHz.



> >
> > 4.- Wavelengths: different observatories use different characteristic
> > wavelengths.
> > The one I remember now are: central wavelength, pivot wavelength,
> > and someone uses the Schneider, Gunn, and Hoessel (1983, ApJ 264, 337)
> > definition
> > which has the property that such "mean" wavelength equals the speed of
> > light over
> > the corresponding "mean" frequency, and is half way between a "mean"
> > wavelength, and a "mean" frequency.
>
> OK - We might have a think about whether these are included as part of the core
> Passband model (ie all Passband code writers have to implement them) or whether
> they are put in separate 'helper' classes used by particular disciplines, with
> methods that analyse the passrate shape to calculate them.

The Registry would use a single (undefined) characteristic wavelength
(approx the centre) and then as Martin suggests a look-up reference to
more complicated whings can be supplied if necessary.  But where such
corrections are usually applied before the VO can do anything with the
data maybe we just need to pass the information that a correction based on
the Schneider (or whatever) system has been applied....

> Are there blue leaks too? I feel that the passrate shape (and min/max) should
> include complete information about *actual* throughput, not *intended*
> throughput.   Is there a case for having both?
>
*actual* throughput... I agree with Martin, again this is the point that
we have to see what VO tools can do, and represent the outcome of
observatory processing where possible, not reproduce it (apart from
providing a history verbatim).

cheers
a