Revealing Traps in Charged Couple Devices using Pocket Pumping Technique Jeremy Hyde, Thomas Smith, Ivan Kotov Jeremy Hyde, Physics Department, Carnegie Mellon University, Pittsburgh, PA, 15289 Ivan Kotov, Instrumentation Division, Brookhaven National Laboratory, Upton, NY, 11973 1. ABSTRACT In modern charged couple devices (CCD) there exists small electron trap sites that degrade the overall charge transfer efficiency of a CCD. In order to characterize these CCDs and their we used pocket pumping techniques to produced regions and catalogs for the astronomical analysis software ds9 in order to locate traps in the pixel data from the CCD image files (.fits files). The region and catalog files were then parsed and the traps were analyzed in order to study where the traps were located and how these traps’ amplitudes were affected by various amounts of pocket pumping. In order to improve this method, we took several images of the same CCD for the same exposure time and averaged the images together in order to obtain a narrower peak for the trap amplitude. Finally, the CCD was cooled to various temperatures ranging from -140° C to -60° C and the pocket pumping was run at various timing sequences in order to discern the energy levels of the traps and what might be causing them. By analyzing the CCD we hope to optimize the characterization process for CCDs being used in sensors of the LSST. As a result of this project, I have gained greater proficiency in the programming languages, C++, python, and java. I had also worked with CCDs before but this project gave me a chance to great how they function and what problems CCD developers must overcome.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Revealing Traps in Charged Couple Devices using Pocket Pumping Technique
Jeremy Hyde, Thomas Smith, Ivan Kotov
Jeremy Hyde, Physics Department, Carnegie Mellon University, Pittsburgh, PA, 15289 Ivan Kotov, Instrumentation Division, Brookhaven National Laboratory, Upton, NY, 11973
1. ABSTRACT
In modern charged couple devices (CCD) there exists small electron trap sites that degrade
the overall charge transfer efficiency of a CCD. In order to characterize these CCDs and
their we used pocket pumping techniques to produced regions and catalogs for the
astronomical analysis software ds9 in order to locate traps in the pixel data from the CCD
image files (.fits files). The region and catalog files were then parsed and the traps were
analyzed in order to study where the traps were located and how these traps’ amplitudes
were affected by various amounts of pocket pumping. In order to improve this method, we
took several images of the same CCD for the same exposure time and averaged the images
together in order to obtain a narrower peak for the trap amplitude. Finally, the CCD was
cooled to various temperatures ranging from -140° C to -60° C and the pocket pumping was
run at various timing sequences in order to discern the energy levels of the traps and what
might be causing them. By analyzing the CCD we hope to optimize the characterization
process for CCDs being used in sensors of the LSST. As a result of this project, I have
gained greater proficiency in the programming languages, C++, python, and java. I had also
worked with CCDs before but this project gave me a chance to great how they function and
what problems CCD developers must overcome.
2. INTRODUCTION
Pocket pumping is a method by which electron traps can be found. CCDs are exposed to a
flat field and the charge is transferred up and down many times, this causes electrons to be
drawn from one pixel and deposited in a neighboring one in the same column, this method is
known as pocket pumping. This withdrawal and deposit causes sets of light and dark pixels
to appear next to one another due to excess electrons be in one and lack of electrons in the
other. By comparing the amplitude of these pixels to the background amplitude it is possible
to locate the traps.
To study the amplitude of the traps in the CCD, the CCD was enclosed in a black box
to previous extraneous light from reaching it. It was then exposed to a flat field for an
exposure time ranging from one-second to ten seconds.
The CCD was then pocket pumped between 1,000 and 20,000 times with a pump depth
of one line and read out into a FITS file using code provided by Ivan Kotov on the
ccdtest@lsst2.
In order to locate traps in the “.fits” file we searched for dipole pairs of pixels
that were light and dark. We searched through possible pairs and found the ones that were 3-
sigmas from the normal background value and labeled them as traps.
In order to try to improve the data, multiple images were taken in sequences and
vaerged together using the PP_trap_final.cpp in which each image from a provided directory
was base line subtracted and new pixel values averaged together and then treated like one
image.
By varying the timing of the pumps and temperature we hope to be able to determine an
approximate value for the energy level of each trap in the CCD. Once its energy level has
been determined we will be able to determine the cause for each trap whether it be a foreign
particle, valence in the atomic matrix or an extraneous atom. In the timing experiment, we
varied the speed of the pocket pumping, this along with the emission time can predict which
pixel a trapped electron with me release into to. For a given speed there is a time tph in which
the trapped charge sits under a pixel. The electron is only pumped to the next pixel if it is
emitted between t=tph and t= 2tph. The probability of an electron being pumped is given in
equation 1 where tph can be found from the speed of the pumping.
(1)
Therefore the amplitude of the dipole is give by the equation below where N is the
number of pumps, Pc is the probability of the electron being captured and assuming the donor
pixel has not been depleted.
(2)
By tracking the I value for various tph or pumping speeds one can determine the
emission time τe. Coupling this with variations in temperature we should be able to determine
the energy of a trap using the equation below where C is a constant based on various masses,
capture cross section and the entropy factor associated with the electron emission.
(3)
3. DATA PROCESSING
At room temperature a sensor with bad segments was used to gather data with a three
second exposure time and the following number of pumps: 1000, 2000, 3000, 4000, 5000,
6000, 7000, 8000, 9000, 10000, 11000, 12000, 15000, and 20000. In order to perform
the temperature experiment we used a different, much better, CCD and cooled it down to -
140° C and took data at 20° C increments until -60° C. At each temperature the CCD was
exposed to it was pumped 1000, 2000, 4000, 8000, and 12000 times with a two second
exposure time. In addition to that it was also pumped at similar values with 0.5, 4, 16 and 32
second exposure times as well as with a 0.1 second exposure time for which it was pumped
200, 400, 800, 1600 and 5000 times. The timing experiment was implemented by varying the
speed of the pocket pumping between 5 µs and 160 µs and pumping the CCD at -120° C,
8000 times and a one second exposure time.
Once this data was gathered it was run through PP_trap_and_ regions.cpp, after first
running run_first.cpp. (Note: A similar program was written to use catalog, .xml files, as
well). This created a region file, which included data such as the amplitude, above or below
background, each pixel, x-coordinate, y-coordinate, and tile number for each trap.
Fig. 1 Pocket pumping regions shown in ds9 with red rectangles around them
This file can be loaded along with the FITS file into ds9 to show the location of the traps on
the CCD. This was repeated for several files with the same exposure time but with differing
number of pumps.
In PP_trap_analysis.cpp I wrote code that would parse all region files inputted into it
by looking for key words and extracting data such as the location of the trap, amplitude of the
pixels affected by the trap compared to background and which tile it is in. This information
was stored and values such as number of traps per segments and average amplitude were
calculated. The data was then graphed using root’s TGraph and TMultiGraph.
The next step was to try to improve the signal to noise levels by averaging together
multiple exposures of the same image. Multiple exposures were taken one after another and
added to a directory. I then manipulated the PP_trap.cpp code provided by Ivan Kotov to be
able to take an inputted directory instead of just a single file. The files in the directory were
added to a file list. For each file we performed a base line subtraction, which removed the
overscan value from the pixels. After this base line subtraction (BaLiS) was performed on
each image the pixel values were averaged together into one image and analyzed
accordingly. The data taken is saved in the base directory /data2/e2v/112-04 on the lsst2 in
the format seen in Table 1.
Table 1 The data is sorted in the directories above