I compiled the following list of problems with the fits files generated by the WAPP spectrometer from various people. Most of these are format problems. There are also errors in the use of the columns, only a few of which are shown here.

Comments to Mike Nolan: nolan   at   naic.edu.

Suggestion or Problem Synopsis Comment Reference Priority A: trivial fix B: Fix if easy C: Can wait D: Discuss Resolution
Calabretta
frequencies are backwards CDELT1 is always positive. Fix. B, but important
Add a version number A
No TDIMn cards SDFITS requires TDIMn, and it's a good idea. Needed for all vector-valued cells TDIMn probably has to be a table entry, unless all lines are the same.   SDFITS B
Make the primary axes (CRVAL2, CRVAL3) be instantaneous RA and Dec J2000 Most users will want instantaneous RA and Dec, not the requested coordinates, in the primary axis. But, that means RA and dec will jitter with pointing error. It could be the instantaneous requested position. Make sure that all CRblah have all of the entries: CTYPEnA, CUNITnA, CRPIXnA CRVALnA, CDELTnA. There are a lot of these, but some (most?) can be "virtual columns".  C
CRVAL2A,C are in hours FITS "requires" (5.4.2.5) that one follow the "recommended" (5.3) degrees Fix. 
RA_HOURS or something would be OK. Also CDELT2A, CROFF2A
 NOST FITS units B
Polarizations are split into different rows, rather than being a 2-d array of freq x pol (x ra x dec x time) SDFITS allows 2-d array, and it's elegant Pols have different Tsys, which can be a vector. We should decide soon. Discuss
Time in CRVAL5 is in hours Axis units should be SI or degrees Fix cunit5 as well   SDFITS
 NOST FITS units
B
Beam runs from 0 to 7, but only 7 exist. Beam 7 doesn't really exist We're likely to put something in beam "7". Add a comment. B
CDELT2-CDELT5 misspelled as CRDELT Typos Fix. A
Cblah scattered through header Cblah are hard to find when they're not together Doesn't affect performance, but a good idea when convenient. C
CTYPE1 is FREQ-F2V FREQ-F2V is meaningless Should be FREQ. Not critical, but easy to fix. Paper 3 A
CTYPE2 and 3 should be RA, not RA---TAN For the point position that's true. for the map?  Paper 2 A
CTYPE5 has faulty comment A
SPECSYS should be SPECSYSV A
SPECSYSV should be BARYCENT? A
CRVAL1V has faulty units Listed as m/s_or_Z, which is not a legal si unit. Fix. NOST FITS units B
TDISP17-TDISP20, TUNIT19, and TUNIT20 are screwed up. They have garbage in them. Looks like a problem in an initialization .h file. B
RESTFRQ missing Is RESTFRQ meaningful? Allow user to specify? Paper 3 C
TSYS always 0 TSYS is a required SDFITS keyword, so it's included even though it's wrong. Should be fixed, but should it be a vector? It has to be read from a file. Gain should be read in too. SDFITS C
Nolan
RADESYSA and EQUINOXA I think we need these. B
CRPIX1V set to 4000f Probably fixing TDISP will fix C
CUNIT1V m/sd should be m/s typo A
CTYPE1V VOPT-F2V not legal. must be VOPT-F2W? paper 3 A
REQ_TOL arcseconds should be " or use degrees D
TCAL a table of 64 alternating very large then plausible numbers should be 2x32 array B
GAIN Gain of control room amps (if2_something?) A somewhat low-level thing to call just "gain". C
Alfalfa
CUR_TOL "Not interpolated" What is this? Is it the error? If so it needs a better name. C
Phil
ID sdfits reserves PROJID Go with standard . SDFITS A
DGRPS
NMATRIX Superseded by TDIMn? SDFITS B
STARTON, OBSMODE, STIME, ... should be table columns True if they can change within a file. C
CRVAL4 Where do we specify time domain for (ACFs?) Specify by changing axis type. C
EXPOSURE Is it dump or scan B
ENC_AXIMUTH, ENC_ELEVATION on sky Why is "ENC" on sky? C
USER_OFFAZ Great circle? say in comment B
TOT_POWER val read does not match mean Is that a problem? C
SYNFRQ 4 synths There are now 8, yes? C
NIFS What does this refer to? B
CRVAL4 The numbering goes nuts for stokes data When stokes data are taken, the CRVAL4 numbers are crazy, and the beams get counted wrong C
Mikael
CRVAL4 CRVAL4 shows pol 1 even when pol2 is selected CRVAL4 is computed from IFVAL, which really tells how many pols, not which one. In addition, the spectrometer may really be selecting pol 1. IFVAL is an engineering keyword, and can be whatever we want, but CRVAL4 needs to be fixed B
I haven't gone through the notes yet.

Valid HTML 4.01!

Last modified: Fri Dec 10 12:10:05 AST 2004