


QC PLOTS 

CURRENT 
HISTORY 
conversion factor 


fixedpattern noise 


QC1 database (advanced users):
browse

plot

Click on CURRENT to see the current trending (Health Check).
Click on HISTORY to see the historical evolution of the trending. 
Parameters are measured for the 5 standard filters BVI Bessell and UR_SPECIAL FORS2 screen flats, and for the standard CCD modes (100Kps/high_gain/2x2 and 200Kps/low_gain/2x2).
QC1 parameters
 conversion factor (CONAD/GAIN) (QC1 database table fors2_scrflat, column conad)
If the exposure time of the first two raw screen flat fields in the input set of frames is the same (within 4%), the difference frame is computed.
Then, the whole readout port is divided into a chessboard of
16x16 boxes. For each one of these boxes, the median signal level from
the first raw frame is divided by the variance in the difference frame
scaled by 2. The median of the 16x16 values obtained is the conversion factor in e/ADU. For historical reasons (different control software) the CONAD
and GAIN parameters are often confused. What is called CONAD here is really the GAIN parameter (units e/adu). The CONAD (units ADU/e) can be calculated as the reciprocal mean of this value.
History
 20101104  20101201 The flat field lamp saturates the screen flats. Therefore there are no valid data for that time range.
 The old pipeline in use until 20080331 derived the CONAD somewhat differently: the whole readout port was divided into a chessboard of 16x16 boxes. For each one of these boxes, the median signal level from the first raw frame was divided by the variance in the difference frame scaled by 2. The median value of the 16x16 values obtained was the accepted value for the conversion factor.
 For the old FORS2 CCD (until March 2002) the conversion factor was derived from a raw file as 100x100 pixels sigma, corrected for fixedpattern contribution. The sigma was compared to the squareroot of the signal averaged across 4 ports. The conversion factor was trended only for the R_SPECIAL filter.
QC1 parameters
 relative fixedpattern noise (QC1 database table fors2_scrflat, column FPN_rel)
The fixedpattern noise is computed in the following way: The difference of two 100x100 subframes of the same master flat field image is determined. These two subframes are are shifted with respect to each other by 10 pixels in X and Ydirection. The population standard deviation of the difference frame, divided by the square root of 2, minus the photon noise (geometrically subtracted) yields the fixedpattern noise, which is then divided by the median of the respective region. The fixedpattern noise is dominated by the brick wall like pattern shown by the MITCCDs, that is particularly strong in U and is due to the CCDs thinning procedure. It is trended only for the B_BESSELL filter.
History
 20101104  20101201 The flat field lamp saturates the screen flats. Therefore there are no valid data for that time range.
 For the old FORS2 CCD (until March 2002) trending of the fixedpattern noise can be found here.
