[fitswcs] comments on WCS Paper III

[William Pence]

To subscribers of the fitswcs mail list:

Attached below are some comments on the Spectral WCS paper III (which is
available from Greisen's web site at http://www.aoc.nrao.edu/~egreisen).
My overall impression is that this paper is now close to being ready for 
submission for publication and for approval by the FITS committees.   Most
of my comments are relatively minor, however, I have 2 questions that are
more addressed to the other members of this mail list rather than to the 
authors of the paper:

1.  How many of you have carefully read the latest version of Paper III?  I
don't claim to be an expert in this field, so before voting to approve the
paper I would prefer to know that a substantial number of you (there are 67
subscribers to this mail list) have actually read the paper and agree (or
at least don't disagree) with everything that is in it.

2.  Which observatories, missions, or data analysis systems are prepared to
buy into this new way of expressing spectral coordinate systems and plan to
start generating FITS data products that use these conventions?  The paper
gives examples of how this spectral WCS can be applied to data from KPNO,
Space Telescope, and the VLA.  Can representatives from these observatories
confirm that they will use these new WCS conventions in future data
products?  If not, why not?

There will be time for discussion of the current status of Paper III at the
FITS 'Birds-of-a-Feather' session at the Strasbourg ADASS meeting 2.5 weeks
from now, so it would be most helpful if people could read and comment on
the paper here before then.

Here are the remainder of my comments on the paper (version dated September
8, 2003)
--------------------------------------------------------------------

1 As a convenience to the reader it would be helpful to include the
definitions of V, Z and z when these important terms are first introduced
in the 2nd paragraph of Section 2.  Otherwise, it takes a while to find 
the definitions in Table 4.

2 Figure 1:  it might be of interest to provide an enlarged plot covering
the range V/c < 0.1 to show how Z, and V differ from v at relatively small
velocities.

3 In section 3.2, the need to always have a factor of ln(10) in the CDELTka
keyword values seems like an unfortunate complication.  The explanation of
why natural logarithms are preferred to base 10 logs seems reasonable, but
is there no way to allow base 10 logs to be used?

4 It would be useful to explain the distinction between the CNAMEia keyword
introduced in section 3.2 and the WCSNAMEa keyword defined in Paper I.  Some
users might get them confused.

5 The first paragraph of section 3.3 says that the algorithm code is given
as 'X2S', but later on in the 6th paragraph it states that the 3rd character
defines 'P' and not 'S'.  I found this to be confusing.  Also, in the last
sentence of that 6th paragraph it would be helpful to state exactly which
equation in Table 4 you are referring to.

6 In the last paragraph of Section 4, it would be more proper to say that
Eq. 45 would be obtained from Eqs. (19), (7), and (66), not (19), (11), and
(66).

7 There's a typo in the second paragraph of section 5.1: "... assumes that
the all the dispersion ...".

8 Section 5.2 says that "small adjustments to some of the parameters" were
made to get the best fits that are shown in the figures.  Can you offer any
practical tips for the reader on which parameters you held fixed and which
you allowed to vary to get the best fit?

9 Is it necessary to state in section 5.2 that leading white space as been
removed in the keyword examples?  That white space is optional for
free-format keyword values, so the keyword records are valid as shown.  Some
readers might get the mistaken impression that the leading white space is
required.

10 In the first sentence of Section 6.1, why does it refer to
"sub-matrices" and not just "matrices".

11 In the first paragraph of Section 6.1 it would probably be better to not
use 'TIME' as an example type of coordinate because a WCS for TIME has not
yet been developed, and it raises a number of implementation questions.

12 If the table lookup coordinate vector has more than one axis, then it
presumably should be stated that the dimensions must be given by the TDIMn
keyword.  (This raises an interesting issue because the TDIMn keyword is
only defined in an unofficial appendix in the FITS Standard.  There is
currently a pending motion before the FITS committees to make the TDIMn
keyword official, however).

13 In the discussion of the case of N non-separable axes in Section 6, I'm a
little confused as to how to linearly interpolate the N+1 coordinate vector
if N > 1.  Is more explanation needed to state how to do this, or is it
really trivial?

14 In Section 6, the sentence beginning with "The dimension of this array
must be given in the header ..." has a syntax problem.

15 the last sentence of Section 6.1.1 should read "... keywords iPVn_1 and
iPVn_2 ...".

16 In Section 6.2.1 in the STIS example, I'd suggest adding a sentence for
clarity to the effect that "The 1PSi_0 keywords are not present, signifying
that the coordinate lookup vector column is in the same table as the
spectra."

17 In Table 12, in the definition of VSOURCE and ZSOURCE, should 'SOUR' be
changed to  'SOURCE'?

18 I had difficulty following the next to last paragraph in Section 7.
Maybe giving an example would help.

19 In section 8.1, what should a FITS reader assume if, for example, the
OBSGEO-X and OBSGXn keywords are both present?  Should OBSGNn apply to
column n, and OBSGEO-X apply to the other columns that don't have an
individual OBSGNn keyword?

20 In Table 14, the BLOCKED keyword should be removed because it is now
deprecated.

21 There is no mention of instrumental spectral line broadening due to
scattering or other effects.  The observed flux at a particular wavelength
is a convolution of the incident flux over a broader wavelength range. 
Should keywords be reserved to describe the apparent width and/or line
profile that would be observed for an idealized monochromatic source?

22 Should keywords be reserved to describe the expected accuracy of the
spectral fit parameters (either for each individual parameter, or for
the combined final calculated value).  You might need 2 parameters: 1 
for the random statistical error, and 1 for the estimated systematic error.

23 Would it be possible to add a section at the end of the paper that
summarizes the known types of spectroscopic data that cannot be described
using the conventions in Paper III?  There is some mention of these
scattered throughout the paper.  Presumably this would include: various
types of interferometric data (Fabry-Perot, Michelson), high energy spectral
data obtained with pulse-height analyzers or calorimeters, data from
spectrographs that have more than 1 disperser, objective prism data, and
what else?

Bill
-- 
____________________________________________________________________
Dr. William Pence                          William.D.Pence at nasa.gov
NASA/GSFC Code 662         HEASARC         +1-301-286-4599 (voice)     
Greenbelt MD 20771                         +1-301-286-1684 (fax)