mirror sites: PL (internal link) HQ     [?]

Quality Control and
Data Processing


CAL | HC | refs | QC
QUALITY CONTROL
    HOME
GIRAFFE QC
Trending & QC1
   bias
Pipeline
GIRAFFE ...
QC links:
GIRAFFE: Bias
Read noise (raw and master) | BIAS level | Structure | Bad pixels

 
HC PLOTS
Bias level, RON, structure, bad pixels
QC1 database (advanced users): browse | plot

2008-05-24: GIRAFFE CCD "Carreras" operational (often called "new") (replacing "Bruce"). Both CCDs are identical in format and size.

2011-01-11: fast read mode added

2016-04-01: the 'slow' mode is offered in VM, calibration data are processed and QC parameters are available, but they are not trended because of their very rare acquisition.

A set of 5 BIAS frames is measured daily as part of the routine calibration and health check plan. The BIAS level, the read noise and the structure are monitored.

'Normal' is the standard read mode of the CCD. The QC1 database has also some data for the 'Fast' and the 'Slow' mode, but (with the exception of the BIAS plot) there are no HC plots for these rarely used modes.

Since October 2010, the fast readout mode is offered in addition to the normal readout mode. Its trending plots are set up the same way as the ones for normal read mode, find them under 'bias_Fast'.

Other CCD parameters are monitored using the darks and the image flats.

BIAS frame, old CCD. There is an upward gradient of the order of several ADU which is monitored by the STRUCTY parameter. The bright vertical feature is a bad column.
BIAS frame, new CCD. The new CCD has cleaner cosmetics, there is no bad column. The gradients in X and Y are much smaller.


Read noise (raw and master)
Read noise (raw and master) | BIAS level | Structure | Bad pixels

The read noise (RON) is measured as rms in a single raw frame and in the master bias frame, using in both cases a small central window 100x100 pixels wide. The RON of the master bias frame is about a factor sqrt(N) lower than the raw file RON, where N typically is 5. RON is given in electrons, calculated from the ADU value using the nominal transformation factor (CONAD = 1/gain) 2.25 e-/ADU. (The measured CONAD values are close to that value, check out here.)

QC1_parameters

FITS key QC1 database: table, name definition class* HC_plot** more docu
QC.OUT1.RON.RAW giraffe_bias..sigma_raw RON value of first raw frame in ADU KPI
QC.OUT1.RON.MASTER giraffe_bias..sigma_master RON value of master bias in ADUCAL
*Class: KPI - instrument performance; HC - instrument health; CAL - calibration quality; ENG - engineering parameter
**There might be more than one, always check the link 'HC'.

Algorithm Read noise (raw and master)

The read noise (RON) is measured as rms in a single raw frame and in the master bias frame, using in both cases a small central window 100x100 pixels wide. Outliers below 100 and above 300 ADU are clipped.

Trending

Box 2 of the trending plot shows the RON of single raw BIAS frames, box 3 the RON of master bias frames.

RON plots. Box 2 shows the RON of single raw BIAS frames, box 3 the RON of master bias frames, typically constructed from 5 input raw frames.

Scoring&thresholds Read noise (raw and master)

The sigma_raw is tightly scored, with dynamic thresholds following the median average of the current period. The thresholds could be made even tighter but due to the relatively small window for the noise measurement the risk of false red scores due to outlier cosmics would then increase. The window size is chosen as to minimize the larger-scale structure of the bias.

The sigma_raw is a KPI parameter, with its reference value being 4.2 el- (from the commissioning report of the new CCD), slightly higher than the current value.

History

The sigma_raw value is very stable over time.


BIAS level
Read noise (raw and master) | BIAS level | Structure | Bad pixels

Box 1 shows the median BIAS level measured across the whole CCD, given in ADU, for the 'Normal' read mode. It is calculated by the QC procedure (for historical reasons).

QC1_parameters

FITS key QC1 database: table, name definition class* HC_plot** more docu
[BIAS_level calculated by QC procedure] giraffe_bias..median_master median value of master bias in ADUCAL
*Class: KPI - instrument performance; HC - instrument health; CAL - calibration quality; ENG - engineering parameter
**There might be more than one, always check the link 'HC'.

Trending

The median bias level is plotted in box 1 of the trending plot.

Scoring&thresholds BIAS level

Normal read mode:
- The median_master is tightly scored, with dynamic thresholds following the median average of the current period. The only requirement on the bias level is stability on the calibration timescale.
- The sigma_raw (read noise) is scored statistically, it is very stable. The same is true for sigma_master which however could produce artificial outliers if less than 5 (or more than 5) input frames get combined.
- The structure parameters are loosely scored, there is no requirement other than stability on the calibration timescale.
- For the number of "bad" pixels there is no requirements other than the detection of sudden outliers.

Fast read mode:
The same applies to the Fast read mode bias plots, with the exception of median_master which is scored in a less restrictive way then the normal data because this mode is very rarely used and sufficiently monitored with the 'normal' mode.

History

The mean bias has slowly decreased over the years.

Algorithm BIAS level

The median of the whole master_bias is calculated, values below 100 and above 300 ADU are clipped.


Structure
Read noise (raw and master) | BIAS level | Structure | Bad pixels

The GIRAFFE BIAS frames contain some structure which is monitored here.

QC1_parameters

FITS key QC1 database: table, name definition class* HC_plot** more docu
QC.OUT1.STRUCT.X giraffe_bias..struct_row structure in X direction in ADUHC
QC.OUT1.STRUCT.Y giraffe_bias..struct_col structure in Y direction in ADUHC
*Class: KPI - instrument performance; HC - instrument health; CAL - calibration quality; ENG - engineering parameter
**There might be more than one, always check the link 'HC'.

Trending

Box 4 and 5 of the trending plot show the structure of the BIAS, derived from collapses of columns (structure in X direction, left) and rows (Y direction, right).

Structure plots. The structure in X direction is derived in the master bias, collapsed along the rows (box 4). The structure in Y is obtained along the columns (box 5). The collapses are shown below.

Collapsed master bias. While the master bias (both for the old and the new CCD) has virtually no structure in X direction (along the rows), it is quite structured in Y direction, along the columns.









Old CCD, Y direction: Roughly the first 1000 rows increase from a level about 4 ADU below average to the average level. The step is due to the bad columns around column 420 starting at about row 1460.

New CCD (normal read mode)











Y direction: the increase is much smoother and has an amplitude of less than 2 ADU. There is no bad column.

New CCD, fast read mode











Y direction: structure looks very similar to the normal read mode.

History

Both structure parameters are very stable with time.

Scoring&thresholds Structure

Both structure parameters are scored, with fixed thresholds set such as to be insensitive to normal fluctuations. A red score indicates a real issue with the detector. There are no reference values available.

Algorithm Structure

The master_bias is collapsed in X or Y direction, the rms of the result frame is measured.


Bad pixels
Read noise (raw and master) | BIAS level | Structure | Bad pixels

The number of bad pixels in GIRAFFE BIAS frames is monitored here. The pipeline delivers proper values only since 2008, for the new CCD.

QC1_parameters

FITS key QC1 database: table, name definition class* HC_plot** more docu
PRO.BPM.NPIX in bpm product giraffe_bias..bpm_npix number of bad pixels HC
*Class: KPI - instrument performance; HC - instrument health; CAL - calibration quality; ENG - engineering parameter
**There might be more than one, always check the link 'HC'.

Trending

Box 6 of the trending plot show the number of bad pixels which is usually between 0 and 200 in the new CCD.

History

The pipeline delivered proper values only since 2008, for the new CCD. The number of bad pixels for the old CCD is much higher than currently, due to the bad columns. It would need to be corrected for those columns in order to gain significance.

Scoring&thresholds Bad pixels

N_badpix is scored, with a loose upper threshold. If flagging red, there is a serious detector issue. There are no reference values available.

Algorithm Bad pixels

Pipeline recipe.


Send comments to <qc_flames@eso.org> powered by QC [webCMS v1.0.2]