Charge injection has been explored as a method to reduce
or eliminate CTI effects in radiation damaged devices. The ACIS
and Astro-E multi grid are examples of techniques which
do reduce the apparent CTI for both
100 keV and
proton damage ccid17s. However, in these cases it is still desirable
to measure the CTI in order to restore the best possible
performance. Recently it has been suggested that charge injection
techniques may provide an improved method for measuring CTI.
Tests have been made to test the feasibility of CTI determination
via charge injection for use on the current ACIS and projected
When compared to earlier charge injection experiments where
relatively large charge levels were used to fill all possible
traps the charge levels required to measure CTI must be
much smaller. Furthermore the charge level idealy should
be adjustable over a range corresponding to the x-ray energy
band of interest, especially at lower energies.
If this level of control can be achieved
the method offers obvious advantages over conventional CTI
measurements determined with x-ray calibration sources. The
limited flux and emission lines of typical sources would no longer
be limiting factors. The serious problem with the charge
injection technique is that it does not cause charge to
appear in the desired location in the manner of an x-ray, but
rather moves the charge there from some source location.
This transfer and other deviations from nominal ccd readout
impact the measured CTI. What this means is that the CTI
measured by charge injection must be corrected before it
can be applied to x-ray data.
The experiments were made with the ACIS dea electronics for
reasons of convenience only. Nothing was done which could
not be accomplished with the existing Astro-E electronics.
Likewise the 100 keV proton damaged ccid17
was used strictly for convenience, and also because it happens
to behave very much like the current ACIS devices. The
injection technique uses a pre-existing pram which was created
to produce single rows of injected charge in the image array
prior to exposure. The only modifications required were those
necessary to reduce the level of injected charge to the lower
levels required for CTI measurement. A simplified version
was found to work just as well but has the virtue of using only
existing ACIS pram/sram modules.
The charge injection method requires a source of electrons, the ability
to control their number and the ability to place them at a
desired location. All this must be reproducible to the degree
required by the accuracy of the CTI measurement needed.
Surplus charge, however created, is swept into the serial
register which is then opened (all phases high) to allow
charge to be uniformly distributed. This charge is then
reverse clocked through the frame store to the desired location
(row) in the image array. After some dwell time the image array
is read out in the usual fashion producing an image with
a single bright row. The signal level of the row is reduced
by the effects of CTI as the location is moved further into
the image array. By averaging readouts for charge injection
at a number of rows a CTI value for every ccd column can be
In the course of the experiment it was soon discovered that
accumulation of charge and subsequent manipulation of the
serial register was not needed nor was it particularly useful.
Charge could be accumulated at the bottom of image array just
by running the reversed frame store clocks for some number of
cycles. The amount of charge is accurately controlled by the number of
cycles and is influenced by the parallel clock levels used.
Some type of an as yet not understood dark current seems to
be the source of the charge. The source of charge seems related
to the ACIS "jitter" dacs situation. The level of injected
charge is seen to decrease violently after a reset or power
cycle. It is also important that the transfer of charge into
image array not be accompanied by a "fat zero" - this is crucial
at low charge levels and will seriously distort the cti if not
Cosmic ray and x-ray events
are not needed or desired, although it appears that an ACIS
level of background can be tolerated if some considerations
Fig.(1) shows a histogram of charge levels injected into
a single row near the bottom of the image array where
CTI will produce minimum distortion. This injection was
made without employing the serial register, a consequence
of which is that some columns will have more charge.
This means that the CTI is determined for different charge
levels for different columns. If this is a problem the
serial register can be employed to smooth the charge. Averaging
over 20 readouts reduces the FWHM to 80 eV which was felt
to be adequate for these tests.
Fig.(2) illustrates the degree of control of charge level
determined by the number of reverse clock cyckes. The break in the slope
at 1026 clock cycles is an artifact of the technique caused
by a finite integration (idle) time in the frame store.
The rather alarming effect of the clock level is not understood,
but is clearly related to the underlying source of charge.
Since a clock cycle represents 40 microseconds the actual
time required to produce the charge row is rather small, much
smaller than would be expected from thermal dark current. It
also means no special time constraints are placed on the readout
(such as the Astro-E 8 second rule.)
Fig.(3) shows the data used to determine CTI for a particular
column at the charge level of Fig.(2). For our purposes
a 2 parameter linear fit is used although we have every
reason to believe a more sophisticated fit is appropriate.
This linear fit to 7 points each the average of 20 readouts
is the basis for all further CTI measurements. Bearing in
mind that only 7 minutes worth of data are needed to measure
CTI for every column it is easy to understand why this
technique is of interest. In fact, it is possible to dispense
with fitting altogether by measuring the CTI for every
row (and thus every pixel).
A problem with this charge injection technique is that moving
the charge to the desired location fills the very traps which
are the source of the CTI which is to be measured. For this
reason the CTI measured by charge injection will overestimate
the true CTI.
be seen by increasing the dwell time - the time delay between
moving the charge row "up" to the desired location and
subsequently moving it "down" for readout.
Fig.(4) shows how the measured CTI increases with the dwell
a time constant of near 1 second for some component of the
CTI producing traps. This is not a surprising result given
what we already know about this device. In general a dwell
time of 2.76 seconds was employed.
Fig.(5) compares the CTI as a function of energy for the charge injection
and conventional x-ray technique. That the charge injection method overestimates
the x-ray CTI is an unavoidable consequence of the finite dwell time. However,
it is reassuring that over the energy range of concern the methods appear
to disagree only by a simple proportionality factor. For levels above 1 keV
it is necessary to adjust the parallel clocks in order to produce sufficient
charge in reasonable time.
The efforts to date show that stable charge injection corresponding to low energy levels is possible using the ACIS and Astro-E electronics without any seriously restrictive clocking operations. The degree of control over the injected charge is quite good and also reproducible even thought the mechanism responsible for producing the charge is not adequately understood. A reasonable agreement between CTI measured via charge injection and with x-rays has been achieved.
Before this technique can be used on ACIS there are a number of issues which must be resolved. One important consideration is the tolerance of cosmic ray background. This can be tested with the 60Co cosmic ray simulator.
Fig.(6) compares the charge level vs. row curves (CTI plot) for the cases with and without the 60Co source. The source leaves some residual charge in the frame store which causes the overall injected level to rise slightly. However, the quality of data is not affected by very much. We conclude that the ACIS level cosmic ray background can be tolerated for this charge injection scheme. The cosmic ray background does make it necessary to minimize the reverse clocking time used to accumulate the injected charge. For ACIS flight application it seems reasonable to fix the number of reverse clocks and use the frame store parallel clock voltage swing to adjust the charge level.
Fig.(7) shows the injected charge level as a function of frame store clock voltage swing. The 100 eV - 6 Kev region is covered for reasonable clock voltages and the degree of precision which can be achieved ranges from 2 eV at 277 eV (C_k) to 160 eV at 6 keV (Mn_k).