WFC Cookbook

Home · Search · About ING · Astronomy · Public Information · Engineering

QUALITY CHECKS OF WFC FOR SA

This page serves as a list of quality checks for Support Astronomers. SA, please use these instructions for measurement of biases, gain and readout noise of the CCD, for a construction of a bad pixel mask and also for linearity tests. In any D nights and, if possible, also at the beginning of each observing run with a visiting astronomer, please use the throughput instructions to check the quality of the CCD.

Bias level

Put the Halpha filter in (to prevent light leaking through the shutter) and take a bias.
To measure the bias level use one of the following instructions:

Type the following command on data reduction computer:

cl> wfc_display r{run number of the bias}
or
cl> imstat r{run number of the bias}.fit[1],[2],[3],[4]
where mean is the bias level.

It can happen that overscan region is not well defined, then it is best to define your own region for doing statistics:

On data reduction computer type:

cl> epar imexam

Change the size of the field used for computing statistics (in order to cover the whole bias) by changing parameters ncstat and nlstat:
ncstat=150
nlstat=150
Do imexam in IRAF:

cl> imexam

Pressing m will do statistics for the field 150x150 pixels. Do it more times in different places on bias and then do an average.

Gain, Readout noise

Take two biases and two domeflats in any filter, set the exposure time of flats in order to have 25000-30000 counts/pixel. For a homogenous illumination of the dome use side dome lights, not a dome lamp.
Now display one flat in all four CCDs (and check the other flat as well):

cl> wfc_display r{run number of the flat}

Choose the good region for computing gain and readout noise for each of the four CCDs. The good region means the region without defects on individual CCDs (for example without "screw" and "fingerprint" on the CCD4, without bad columns...).
In Iraf go to noao.obsutil package:

cl> noao
noao> obsutil

compute gain and readout noise:

ms> findgain r{flat1}.fit[1] r{flat2}.fit[1] r{bias1}.fit[1] r{bias2}.fit[1] section="[600:2500,150:1900]"
ms> findgain r{flat1}.fit[2] r{flat2}.fit[2] r{bias1}.fit[2] r{bias2}.fit[2] section="[600:2500,150:1900]"
ms> findgain r{flat1}.fit[3] r{flat2}.fit[3] r{bias1}.fit[3] r{bias2}.fit[3] section="[600:2500,150:1900]"
ms> findgain r{flat1}.fit[4] r{flat2}.fit[4] r{bias1}.fit[4] r{bias2}.fit[4] section="[600:2500,150:1900]"
where you will define your own section. The coordinates for the region section are [X1:X2,Y1:Y2].

You will get gain in electrons per ADU and the readout noise in electrons. In order to get readout noise in ADU devide it by gain in el/ADU and put the numbers on "CCD quality control" web pages.

You can see more detailes in the presentation CCDs.

Bad pixel mask

It is useful to create a bad pixel mask, because one can then interpolate over bad pixels, which are nonlinear.

Take a series of domeflats exposed at a high enough level to have several thousand counts per pixel (30000).
Take a series of domeflats exposed at a much lower count-level (50-100 counts per pixel).
In Iraf go to noao.imred.ccdred package.
Create a list of longflats and shortflats:

ccdred> files flat1,flat2,flat3 > longflats
ccdred> files flat4,flat5,flat6 > shortflats
Put appropriate names in place of flat1 to flat6.

Use "flatcombine" to combine longflats and shortflats separately:

ccdred> flatcombine @longflats out=FlatL reject=crreject scale=mode rdnoise="readnoise" gain="gain"
ccdred> flatcombine @shortflats out=FlatS reject=crreject scale=mode rdnoise="readnoise" gain="gain"
Check if "readnoise" is a keyword in headers.

Divide them to reveal the nonlinear pixels:

ccdred> imarith FlatL / FlatS Flatdiv

In Iraf in ccdred package go to nmisc routine:

ccdred> nmisc

Set the soubroutine ccdmask default:

nmisc> unlearn ccdmask

Create a bad pixel mask:

nmisc> ccdmask Flatdiv mask=badmap

You can display all the images (FlatL, FlatS, Flatdiv, badmap) to be sure that everything is well done.

Linearity

To test the linearity it is necessary to take a series of domeflats in any filter. You can do it manually or you can use the script linear_auto_s. You must set a level of illumination in the dome so that all CCDs will be saturated (65000 counts/pixel) after 32s exposure.
If you do it manually then start with 2s exposure and continue increasing by one to 32s exposure.

Throughput

Any suggestions?

Last Updated: 15 January 2007
Ovidiu Vaduvescu, ovidiuv@ing.iac.es