SINFONI has one detector, (see the User Manual for details concerning read-out modes). Our QC1 checks for the detector using dark frames can be divided in three categories:

• PRODUCT QC PLOT. For each dark pipeline product and its associated raw frames we create a plot with the most significant features.
  
  
• DIFFERENCE FRAMES. Either the difference between the current and the most recent product frame (local difference) or the difference between the current and a reference product frame. This simple and efficient method is to check short-term and long-term variations.
  
  

• RELATIONS checks. This means that we associate QC1 parameters with each other and check if they follow a certain relation and we check if this relation is constant in time.
  
  

• CONTROL CHARTS. Is a continuing statistical test of the last (most recent) QC1 parameter
  
  
• TRENDING This means that selected QC1 parameters are monitored as a function time. The purpose is to find long term variations which cannot detected on a single product frame.

• The array is operated in Fowler sampling read mode.

We consider the noise in dark frames to be composed by statistical noise (random noise) and fixed pattern noise. Close to the read-out ports there might be contributions of photon noise as well. If a raw dark frame has DIT=5sec and NDIT=10, this means that 10 times 5-sec exposures have been read out from the array and averaged on the registers before being stored in a frame. The statistical noise as derived from a raw frame is reduced to NDIT=1 via sqrt(NDIT) to be comparable to other dark settings. The fixed pattern noise does not vary with NDIT by definition.

There follows the dark product qc plot. The Figure caption is equal for all four plots.

• Example SI_DARK product plot.
• Most recent SI_DARK product plot. [PS]
  

Figure caption

 

Upper Left: The mean (collapsed) column (red) and the column at X=2048/4 of the product frame.	Upper Middle Two rows at Y=2048/4 (red) and Y=2048*(3/4) (blue) of the product frame	Upper Right A parabolic function (blue) is fit to the central part (-3 to 3 counts) of the log of the raw2-raw1 difference frame (red). Histogram of the raw3-raw2 difference frame (green).
Lower Left Mean column (collapse), column x=500 of the product and of the first of the three raw input files. A smaller y-range is shown to show the difference between fixed-pattern (=pixel-to-pixel) noise and the statistical noise.	Lower Middle The Y=500 row of the product minus the Y=500 row of a reference product. Y=600 row of the product minus the Y=600 row of a reference product. X=500 column of the master minus the X=500 column of a reference product. The reference product is renewed usually once a period or after an intervention.	Lower Right Histogram of the difference frame (second raw minus first raw) (red) and Gaussian fit between -3 and 3 counts (blue). Histogram of the raw3-raw2 difference frame (green)

Inspection of dark product difference frames is a valuable way to check on product variability.

• Current Product minus Reference Product shows long-term variations. The reference product is usually updated after an intervention or less frequently (diff).
• Current Product minus Last (Most Recent) Product shows the short term variations

The performance of the instrument and detector is measured in terms of QC1 parameters. These parameters are defined per frame stack / product type. The most important QC1 parameters derived from dark frames are given below. Each QC1 parameter is checked against the mean and the dispersion of the recent QC1 parameters (e.g. those taken from the last 3 weeks). Speaking in terms of process control charts, we check if the QC1 parameters behave like a process under statistical control, but there are no external constraints that the QC1 parameter must lie in a certain interval. Hence no 'acceptance control charts' are required, but simple 'Shewhartd control charts' are sufficient.

The quality control checks are performed (beside the visual inspection) on QC1 parameters. Some of them are related with each other. We check the stability of this QC1 parameter relation and simply plot QC1_A versus QC1_B in a diagram. The relation is hence defined empirically, although it is possible to define a theoretical relation as well. Such a diagram shows the relation based on values collected of a longer period of time. The current (the newly extracted) QC1 parameters are overplotted over this relation using another graphical coding, to highlight them.

Trending describes the variation of a QC1 parameter with time. Trending plots don't consist of control limits. Their purpose is more to document long term changes of QC1 parameters, which cannot be detected in process control charts. Trending plots can be subject of 'time variation analysis'. The same QC1 parameters as the ones shown under the 'control chart' section are trended. More parameters are available via the interfaces of our QC1 parameter database.

• Random noise as delivered by the sinfo_rec_mdark recipe of the pipeline. For the SINFONI array the recipe gives one QC.RON parameter for the whole chip. These values are trended for the most frequently used DITs. The calculation of the QC parameters follows in the same way as for ISAAC/NACO. Two consecutive raw dark frames are subtracted from each other and the median standard deviation of a limited number of samples is taken and normalized to DET.NDIT=1. The most recent QC1 parameters are available :
  • DARK_RON
    [PS]
    
    

• Median Dark level. Is taken from each product frame and is trended. The parameter is derived from the product frame via the eclipse command stcube. The most recent QC1 parameters available :
  • DARK_COUNTS
    [PS]
    
    
• Fixed pattern noise (New 2005-06-01). From a master dark; take all pixels, build the histogram; fit a Gaussian to the distribution; the stddev=sigma is the Fixed Pattern Noise. The most recent QC1 parameters available for the most frequently used DITs:
  • DARK_FPN
    [PS]
    
    
• Fixed pattern noise per counts (New 2005-06-17). The fixed pattern noise, that should scale linearly with the counts on the array. Hence FPN/counts should be a constant in case of stable detector linearity properties. The most recent QC1 parameters available for the most frequently used DITs:
  • DARK_FPN/COUNTS
    [PS]
    
    

• Hot Pixel. The number of pixels with a high dark current. The threshold is given in the pipeline recipe resource file 'sinfo_rec_mdark.rc'.
  • DARK_HP
    [PS]
    
    
• OddEvenColumn Stripes in the area of slitlet #25. Take the median of all odd columns in the area of slitlet #25 (=Med(odd)), take the median of all even columns in the area of slitlet #25 (=Med(even)) and build
  OES25 = 2 * ( Med(odd) - Med(even) ) / ( Med(odd) + Med(even) ).
  This parameter OES25 is trended for each individual Dark raw frame. The area of slitlet #25 contains 64 columns starting from column 1922.
  • DARK_OES25
    [PS]