cic_analysis.txt

Clock Induced Charge Analysis

This recipe from Mike Ireland:

1) Take at least 10 frames with chosen settings, which is a high gain mode and with shutter closed (definitely above a gain of 200 electrons, preferably 500 electrons) 2) Compute a median dark and subtract it. 3) Compute a robust standard deviation, e.g. stdev = 1.48*sqrt(median(data^2)) 4) Compute the number of elements above a threshold e.g. threshold = 3.5*sigma length(where(data > threshold))/length(data)

5) Finally, there is a correction you could do which is to divide the final number by: exp(-threshold / em_gain_in_dn)

[but to do step 5, you need the EM gain. As long as gain is high, this correction is small]

Further clarification from Mike:

yes by median dark I mean a bias frame. But computed by taking the median of lots of bias frames.

Threshold should be a single number, equal to 3.5 times the readout noise.

Data is a bias-subtracted data cube. Use one slice or all slices for better statistics, but definitely not a sum or average.

My analysis procedure:

Each data cube contains 500 images in rapid succession. I only use the first 30 images of each data cube for analysis. I sum the 30 images and divide by 30 to get the bias frame. I subtract the bias frame from each of the 30 images. I then cast the entire 512x512x30 data cube into a 1-D vector, use that vector to generate a robust standard deviation scalar quantity (1.48*sqrt(median(vector^2)) -- finding the median is the most computationally intensive step), troll the vector for any values that exceed the robust standard deviation by a factor of 3.5, divide that number by the total length of the 1-D vector, and divide by the gain correction factor (exp(-1.0*thresh/gain)).

Two observations:

1) The vertical shift delay makes a difference in the measured CIC. I don't understand why this would be the case. If one expects this, wouldn't we need to know the value used by Andor in their literature?

2) The higher the gain, the larger the clock induced charge. To me, this would imply that I'm not applying the correction factor correctly -- being called a correction factor implies that its application would compensate for the differences in the gains.

The summary file:

File Temperature Gain Vertical Shift CIC Fraction

... 0000.fits -50.0000 300 3.30000e-06 0.00426411 ... 0001.fits -50.0000 300 3.30000e-06 0.00390681 ... 0002.fits -80.0000 300 1.70000e-06 0.00133295 ... 0003.fits -80.0000 300 9.00000e-07 0.000842734 ... 0005.fits -80.0000 300 5.00000e-07 0.000698496 ... 0006.fits -80.0000 300 3.00000e-07 0.000715206 ... 0007.fits -80.0000 500 3.30000e-06 0.00416787 ... 0008.fits -80.0000 500 1.70000e-06 0.00221594 ... 0009.fits -80.0000 500 9.00000e-07 0.00149811 ... 0010.fits -80.0000 500 5.00000e-07 0.00114543 ... 0011.fits -80.0000 500 3.00000e-07 0.00101927 ... 0012.fits -80.0000 800 3.30000e-06 0.00643356 ... 0013.fits -80.0000 800 1.70000e-06 0.00335487 ... 0014.fits -80.0000 800 9.00000e-07 0.00227567 ... 0015.fits -80.0000 800 5.00000e-07 0.00174591 ... 0016.fits -80.0000 800 3.00000e-07 0.00155108 ... 0017.fits -80.0000 100 3.30000e-06 0.000628299 ... 0018.fits -80.0000 100 1.70000e-06 0.000518980

Sample FITS header (in case I overlooked something):

SIMPLE = T / file does conform to FITS standard BITPIX = -32 / number of bits per data pixel NAXIS = 3 / number of data axes NAXIS1 = 512 / length of data axis 1 NAXIS2 = 512 / length of data axis 2 NAXIS3 = 500 / length of data axis 3 EXTEND = T / FITS dataset may contain extensions COMMENT FITS (Flexible Image Transport System) format is defined in 'Astronomy COMMENT and Astrophysics', volume 376, page 359; bibcode: 2001A&A...376..359H HEAD = 'DU897_BV' / Head model ACQMODE = 'Kinetics' / Acquisition mode ACT = 0.0246 / Integration cycle time KCT = 0.0246 / Kinetic cycle time NUMACC = 1 / Number of integrations NUMKIN = 500 / Series length READMODE= 'Image ' / Readout mode IMGRECT = '1, 512, 512, 1' / Image format HBIN = 1 / Horizontal binning VBIN = 1 / Vertical binning SUBRECT = '1, 512, 512, 1' / Subimage format DATATYPE= 'Counts ' / Data type XTYPE = 'Pixel number' / Calibration type XUNIT = 0 / Calibration units TRIGGER = 'Internal' / Trigger mode CALIB = '0,1,0,0 ' / Calibration DLLVER = '4.21.30006.0' / Software Version EXPOSURE= 1.0E-05 / Total Exposure Time TEMP = -50. / Temperature READTIME= 5.882353E-08 / Pixel readout time OPERATN = 0 / Type of system GAIN = 300 / Gain EMREALGN= 0 / EM Real Gain VCLKAMP = 0 / Vertical Clock Amplitude VSHIFT = 3.3E-06 / Vertical Shift Speed OUTPTAMP= 'Electron Multiplying' / Output Amplifier PREAMP = 3. / Pre Amplifier Gain SERNO = 7103 / Serial Number UNSTTEMP= -999. / Unstabilized Temperature BLCLAMP = T / Baseline Clamp PRECAN = 0 / Prescans FLIPX = 0 / Horizontally Flipped FLIPY = 0 / Vertically Flipped CNTCVTMD= 0 / Count Convert Mode CNTCVT = 1 / Count Convert DTNWLGTH= 550. / Detection Wavelength SNTVTY = 5.44053 / Sensitivity SPSNFLTR= 0 / Spurious Noise Filter Mode THRSHLD = 0. / Threshold PCNTENLD= 0 / Photon Counting Enabled NSETHSLD= 0 / Number of Photon Counting Thresholds PTNTHLD1= 0. / Photon Counting Threshold 1 PTNTHLD2= 0. / Photon Counting Threshold 2 PTNTHLD3= 0. / Photon Counting Threshold 3 PTNTHLD4= 0. / Photon Counting Threshold 4 AVGFTRMD= 0 / Averaging Filter Mode AVGFCTR = 1 / Averaging factor FRMCNT = 1 / Frame Count USERTXT1= ' ' / User text USERTXT2= ' ' / User text USERTXT3= ' ' / User text USERTXT4= ' ' / User text DATE = '2012-07-06T01:22:34' / file creation date (YYYY-MM-DDThh:mm:ss) FRAME = '2012-07-06T01:22:34.000' / Start of Frame Exposure ESHTMODE= -1 / Electronic Shuttering Mode END

If anyone wishes to do further reduction, the data can be found in my home directory ( ~nils/tmp/andor/andor_cic_tests ) on the office side network.

In comparison with the literature included with the camera, we are very close. The CIC number with the camera tests was done at an EM gain of 1000x, 10 MHz horizontal, and 0.5 usec vertical shift. For our tests, we did 300x, 500x, and 800x, 17 MHz horizontal with varying vertical shifts. The relevant entries from above are:

... 0005.fits -80.0000 300 5.00000e-07 0.000698496 ... 0010.fits -80.0000 500 5.00000e-07 0.00114543 ... 0015.fits -80.0000 800 5.00000e-07 0.00174591

These match well with the value of 0.00108 events per pixel measured by Andor. My conclusion -- the camera meets spec.

2012-09-11:

CIC tests using subregions:

All tests done using an EM gain of 1000x, a temperature of -75C, a readout rate of 10MHz, a 0.5 usec vertical shift rate, and a 90x90 region of interest.


 * Lower left corner of ROI || Preamp gain || CIC ||
 * (1,1) || 3 || 0.00578088 ||
 * (1,1) || 2 || 0.00447576 ||
 * (1,1) || 1 || 0.00284685 ||
 * (211,1) || 3 || 0.00581897 ||
 * (211,1) || 2 || 0.00419993 ||
 * (211,1) || 1 || 0.00273999 ||
 * (211,211) || 3 || 0.00609663 ||
 * (211,211) || 2 || 0.00490706 ||
 * (211,211) || 1 || 0.00275234 ||

2012-09-26:

CIC tests on the six cameras delivered 2012-09-24: