Known problems in the SINFONI pipeline public version 1.8.2
	HOME INDEX SEARCH HELP NEWS

Known problems in the SINFONI pipeline public release 1.8.2

Last update: Fri Oct 19 16:13:46 UTC 2007.

•  sinfo_rec_detlin.  LINEARITY,LAMP data taken in the period 24/11/2006-28/02/2007 have been taken with increased lamp intensities. This means that in particular frames taken with high values of DET.DIT (for example DET.DIT=45) have several pixels in the non linear detector's range. This affects particularly the J band. As a consequence if the user uses recipe parameter defaults may obtain non linear pixel maps (ESO.PRO.CATG=BP_MAP_NL) collecting significantly more pixels than usual, and this would propagate in the MASTER_BP_MAP. Those pixels consequently would not be taken in consideration in the wavelength calibration, despite the level of the signal in the arc lamp frames is in the linear regime, this possibly leading to a lower quality WAVE_MAP and wavelength calibration solution. To solve this problem the user may:
  • Not use raw frames at higher DET.DIT to build the non linear bad pixel map.
  • Use all the raw frames and increase the parameter bp_lin-thresh_sigma_fct from the default 10 to 15 or more.
  
   
•  sinfo_rec_distortion.  Occasionally this recipe may fail indicating that too less slitlets have been found. The user may try to increase the parameter smooth_rad or the parameter arcs_min_arclen_factor.
   
•   sinfo_rec_wavecal This recipe sometimes fails to determine the slitlets positions. A possible reason is an improper data reduction parameter setting. For example the parameter wcal-pixel_tol has a default value of 5.0 chosen to insure good accuracy. Greater values equal may insure higher robustness (particularly reducing H+K-band data) at the price of a minor accuracy.  Another parameter to eventually change to improve robustness is wcal-pixel_dist. It has a default value of 15. One may try to increase it a bit. Other parameters that can be changed are the ones having aliases wcal-min_diff, wcal-na_coeffs, wcal-nb_coeffs, wcal-pixel_tol, wcal-y_box, which in the current release, for accuracy have default values as specified in the first column of the following table but in the previous releases they had default values as specified in the other table raws, values which might offer greater robustness at a price of a minor accuracy.



band	wcal_min_diff	wcal_na_coeffs	wcal_nb_coeffs	wcal_pix_tol	wcal_y_box	comment
any	1.0	4	2	5.0	5.0	default setting
H	10.0	3	2	7.0	2.0	old setting
H+K	1.0	4	2	5.5	5.0	old setting
K	1.0	4	2	12.0	5.0	old setting
J	1.0	4	2	3.0	5.0	old setting

Another possible explanation of the failure is the usage of an improper bad pixel map. We suggest to use the REF_BP_MAP as input of the sinfo_rec_mflat recipe which determines the MASTER_BP_MAP which is input of sinfo_rec_wavecal.
 
 

• sinfo_rec_jitter.  Sky subtraction quality. Despite we provide an improved routine to correct for possible sky line residuals, we suggest the user to use it only when (at least) a sky frame is available. Otherwhise (if the user uses such correction in absence of a sky frame, case in which the next in time object frame is used as ``dummy'' sky), there is the risk that object lines cancel out after subtraction. The user should also choose proper values of the critical parameters skycor-sky_bkg_filter_width, skycor-line_hw, skycor-rot_cor, skycor-fit_obj_noise. A check of the results provided by the product table(s) with PRO.CATG QC_SPECTRA is always recommended to compare the extracted spectra before and after the correction.

 
• sinfo_rec_jitter. This recipe assumes as default for the parameter objnod-scale_sky the value TRUE. This is appropriate in case the user has no sky frame and has chosen for the parameter  objnod-ks_clip the value TRUE in case the subtraction of the fake sky has not completely removed the sky backround. If the user has sky frames, using such defaults may lead to an over estimation of the sky and consequently to portion of spectra with a negative value. To prevent this problem, we suggest the user to set the parameter objnod-scale_sky to FALSE.

 
•  sinfo_rec_jitter.  WCS coordinates in the coadded cube may be imprecise. We invite the user to check the accuracy of the reference catalog used, it may be that the VLT-SINFONI combo is more accurate; eventually correct object coordinate to take into account of its possible proper motion.

 
•  sinfo_rec_jitter. Seldomly, this recipe may originate coadded cubes whose spectra along the z axis may have spikes. This occurrence may be an indication that the kappa sigma clipping algorithm has removed some object point. In those cases we suggest to increase the value of kappa.

 
•  sinfo_rec_jitter. Some user may have observations taken with the telescope position angle not equal to 0. This occurrence can be revealed verifying that  the value of the FITS keyword ADA POSANG is not 0. In those cases the pipeline may coadd cube components incorrectly.

 
• sinfo_rec_jitter. This recipe may fail to perform a 2D Gaussian fit to the observed target. In this case the STD_STAR_SPECTRA and STD_STAR_SPECTRUM pipeline products will not be generated. The user may try to reduce the value of std_star-fwhm_factor  to get also those products.

 
• sinfo_rec_jitter This recipe has performance limits. The fastest execution may be obtained setting product-density to 0.
• sinfo_rec_jitter This recipe is much demanting in terms of RAM. The amount of RAM required is approximatively given by the formula:

RAM=[150+4 *(2 * sX * sY * sZ + 64 * 64 * sZ * N_frames)] MB,

where sX, sY, sZ indicate the size of the coadded cube, and N_frames is the number of coadded frames. sX and sY are a function of the minimum and the maximum x and y offsets of the coadded cube components with respect to the first one:

sX=2*floor(offx_max-offx_min+0.5)+64,  sY=2*floor(offy_max-offy_min+0.5)+64.

The floor(x) function computes the largest integral value not greater than x.We suggest the user to have at least 1 GB of RAM and a swap area at least equal to the amount of RAM. In case sX=sY=130, sZ=2300, N_frames=30, the user needs at least 1.6 GB of RAM, the same amount of swap, dedicated to run the pipeline.
 

We also suggest the user before starting the data reduction to check the values of  the FITS keywords SEQ CUMOFFSETX and SEQ CUMOFFSETY. Sometimes those may be very different. Their values may be such that the size of the output coadded cube is large enough to not allow a full data reduction, due to the limited amount of available RAM. In those cases or if the single cube components corresponds to observations of portions of sky which do not overlap with each other, we suggest the user to set objnod-jit_ind=FALSE.
 

• EsoRex may report at the end of a recipe execution the following wrong information:

  [ INFO  ] esorex:                 (no ESO PRO CATG keyword found...)
  

  The user should ignore it, the SINFONI pipeline products have the ESO.PRO.CATG keyword properly set.
•   Some FITS tables from the file sinfo-calib-1.8.2.tar.gz of the kit release are corrupted. The user is invited to use master calibration data from this file.
   

 

Send comments to <usd-help@eso.org> Last modified: Fri Oct 19 16:13:46 UTC 2007 MET