Spectroscopic keyword issues 20130819

EFOSC2 spectra
SJS checked through EFOSC2 and SOFI spectra for basic checks of validity of important header keywords. SOFI spectra all look fine (apart from the integer shoft problem which also affects all EFOSC2 data), EFOSC2 data have some issues 

Bugs for Dave/Morgan to check/correct in purple. Bugs for Stefano to correct in next version of pipeline in red

Problem 1
Working on the *si.fits files in /data/psdb2data1/pessto_eso_phaseIII/validated_pessto_files/efsoc/spectra

There are 860 science spectra (*si.fits). ALL of the problem files below are from files from April 2012 
  • Gr13: The pixel size is 0.55nm and in most cases it is set correctly (SPEC_BIN) But in 77 files this is set to 1.52nm . In these files the value of SPEC_BW is correct, and as SPEC_BIN=SPEC_BW/N_pix ; presumably the value for N_pix is wrong. 
  • Gr11 : three files have a similar problem, SPEC_BIN = 1.27 when it should be 0.41
  • Gr16 : four files have a similar problem, SPEC_BIN = 1.88 when it should be 0.43

List of files in : SPEC_BIN_problemfiles.txt   

Maybe Dave already corrects this in the conversion to FITS tables ?  Can be corrected by Dave during conversion. Dave's Response: I don't correct these yet - will add this in.
Stefano to check in pipeline and correct for next release. 

Problem 2
A large number of files have a garbage value set for 

SPEC_RES= 432.0955936041426           / Spectral resolving power 

For example, it should be 357 or 225 for Gr13 (for 1 or 1.5" slits), but it is recorded as values between 3-30, which is obviously wrong. Similar problem for Gr11 and Gr16 exists. 

There are 70 problem files - in SPEC_RES_problemfiles.txt

Stefano - how is this value calculated ? Is it measured on the arc frames ? Can pipeline be revised to reduce these failures for SSDR2 ?

SSDR1 solution - Dave could take a SPEC_RES value from a file on the same night and insert. Or, if it is a reduction issue, Morgan an re-reduce. 

Problem 3
The pipeline shifts the spectra in wavelength with respect to the sky lines. But the value for SHIFT in the header is always an integer. Why ? It should not be ? Is it recorded wrongly, or is it calculated wrongly ?  Morgan and Stefano to check. This is worrying - if wrong, we may need to do all reductions again.  
Also, the value for SHIFT is missing in two files  


The SOFI spectra also have integer shifts. 

Stephen Smartt,
20 Aug 2013, 03:27
Stephen Smartt,
20 Aug 2013, 03:27