Depletion Voltage of 1.3K Detector Through Charge Collection
Mia T. Onodera
Summer Research Final Report
August 19, 1998
ABSTRACT
The setup and initial data for characterizing
the radiation hardness of the pre-radiated 1.3 kohm*cm Brookhaven (low
resistivity) detector is discussed. The depletion voltage determined
by using the CV curve (electronics only) and charge collection (through
the detector) methods are compared.
INTRODUCTION
Although I initially began working with Tim
Dubbs on the Cafe-M comparator chip, the majority of my research this summer
was working with Tim Nissen and his radiation hardness tests of low resistivity
detectors. There are 20 detectors with resistivity from 100 ohm*cm
to 6 kohm*cm that will be measured before and after irradiation.
Each group consists of four similar resistivity detectors that are characterized
before irradiation by determining the depletion voltage using CV curves
and charge collection. One detector will be kept as a control while
the other three will be irradiated at different fluences. After radiation,
the detectors’ depletion voltage will be measured again, this time as a
function of proton fluence.
My work, and subsequently this paper, will
discuss the setup and the data collected for one of the 1.3 kohm*cm detectors.
The CV curve for this resistivity detector was completed prior to my summer
program (figure 1) and the CV curves for the other resistivity detectors
are being completed at this time by other researchers.
SETUP
To test the charge collection
of the detectors, a strontium source was used. Strontium is a beta
emitter with a maximum energy of 2.2 MeV. The detector board placed
above the source includes the detector, capacitor chip, several LBIC (low
power bipolar integrated circuit) chips, and the CDP (clock driven pipeline
chip) which in turn is connected to the bias card, DSP (digital signal
processor), and computer. The scintillator is situated within millimeters
above the detector board, directly over the detector. Attached to
the scintillator are the front-end electronics that change the analog signal
from the scintillator to the appropriate TTL signal and the pulse generator
that eventually reaches the bias card and computer (figure 2). The
effective doping concentration of the detectors is found to be directly
proportional to the depletion voltage and inversely proportional to the
square root of the depth by Vdep = (Neff / 2E)*d2 .
Relative to the detectors we are using, the effective doping concentration
is inversely proportional to the resistivity, meaning we need more “impurities”
or p-type material in our low resistivity detectors. In addition,
the n-type bulk of our detectors will invert to p-type over time after
irradiation, so it is necessary to have a layer of n-type material on the
back plane so that a depletion area remains after type inversion.
As particles are detected, the charge produced
on the detector strips is amplified and converted to voltage in the LBIC
and then the voltage is compared to a set threshold. In the binary
code, if the voltage is above the threshold, a 1 is sent out. If
it is not, a 0 is recorded. This data is held in the CDP until the
DSP is ready to read the data. There are 256 clock cycles in the
CDP, each 27 ns long. Thus, the first particle’s data is not ready
to be collected until 256*27ns or 6.912 microseconds after the data is
stored in the CDP. The scintillator picks up the same particle as
it was emitted from the detector but the scintillator can send its information
in about 300 ns. Therefore, the scintillator is delayed so that its
trigger corresponds to the data of the same particle. This trigger
tells the CDP to send its data to the front-end electronics. The
binary code stored in the CDP translates to “hits” once it reaches the
computer. Using this method of so many “hits” per channel, efficiency
and occupancy is determined.
OBJECTIVE
Our objective in this study is to obtain results
that imply that lower resistivity (below 6 kohm*cm) detectors will have
longer life spans in accelerators such as the LHC (Large Hadron Collider)
at CERN. The assumption we are making about our test method is that
the depletion voltage we determine through the CV curves forms a linear
relationship with the depletion voltage found using charge collection for
each of the detectors. From prior studies it has been determined
that “pulse height information can be recovered by varying the threshold
and measuring the count rate, which is the integral of the pulse height
spectrum.” Therefore, the most probable pulse height at a given bias
will be determined by plotting efficiency as a function of threshold or
charge (figures 3-4), and charge (fC) as a function of bias for each median
point or 50% of maximum efficiency should yield the depletion voltage.
As shown in figure 5, when the charge remains constant, the depletion voltage
has been reached.
INITIAL DATA
The first factor that needed to be addressed
was the timing. Tim initially had a non-digital gate generator with
several discriminators attached by cable,
but we simplified this setup with the implementation of a pulse generator
that is capable of multiple functions. We determined that the test
pattern was strongest using a clock cycle of 27 ns rather than 25 ns and
then used the pulse generator to delay the scintillator trigger.
Figure 6 shows that early data was not very consistent even though the
error appeared to fall within one standard deviation. Another point
to note is that the maximum efficiency lied around 60%. Using a bias
of 230V (above predicted depletion voltage), efficiency was determined
as a function of threshold. With our setup changes, the efficiency
improved, but these improvements revealed higher than expected efficiency
results (noise) at low thresholds which were inconsistent with the other
threshold values (figure 7).
IMPROVEMENTS
To deal with the inconsistent data points,
we incorporated two methods. The first method was to examine
channel maps (figure 8) to determine if certain channels were “hot”.
If the plots showed that a channel was not working properly, we masked
that channel. The second method was an obvious one in which we took
more events for each run in subsequent testing. To improve efficiency,
we compared the number of triggers shown on the scintillator counter and
the DSP and found that the DSP was not reading all of the triggers.
Tim found that if he increased the width of the trigger then the DSP could
properly generate the signal for every trigger. From this, we raised
the maximum efficiency about 10% but it is hypothesized that something
in the geometry of our setup was limiting our maximum efficiency to about
70%. Lastly, a noise scan was run for subsequent tests to combat
the unrealistically high efficiency at low thresholds where the efficiency
of the signal (Es) was determined by Es = (Erun – Enoise) / (1 – Enoise).
FIRST RESULTS
All of the plots of efficiency as a function
of threshold are included with the corresponding graph of threshold as
a function of bias (appendix A) to show that the 50% of maximum (or threshold
at 35% efficiency) was arbitrary in some cases. This data did not
include the channel masks or increasing the number of events. In
addition, because we are measuring charge collection, we want our final
plot to be charge as a function of bias, not threshold. Several calibration
conversion curves have been tested in the past. Figure 9 shows the
effects of two curves. The function described in Teela Pulliam’s
thesis, V= A + B[1+(B/CQ)2] -1/2 , where A, B, and C are parameters, V
is voltage in mV, and Q is charge in fC, is a better fit because it does
not “turn over.” Using the other function previously, we were forced
to mask several data points that went beyond the limits of the function.
SECOND RESULTS
Prior to repeating our data run after the
improvements discussed above, the VME crate that holds the bias card broke
down. We moved the setup into Tim Dubbs area, using his rack for
electronics. Our efficiency improved to 88%. Tim Nissen’s hypothesis
is that it has something to do with our bias card because he tried both
as he was setting up our system in the new location. It was also
pointed out that as we mask channels, we will obviously lose some efficiency
there as well. Our second run was completed in 24 hours. Tim
spent one evening testing the timing, noise, and current changes with the
new setup. The next morning (the day of my first real earthquake,
August 12, 1998!) I began a new test run using many more events, but only
choosing six threshold values for each bias voltage. We finished
the 18 bias voltages ranging from 50–350V that evening. My hypothesis
is that the data collected is better in the morning and evening when no
one else is in the lab. I think body movement and heat and the use
of other instruments creates a lower efficiency reading (whether this is
supported by physics or not). I showed preliminary data from this
test run in my final presentation that morning. Although all of the
remaining bias voltages weren’t as smooth, the overall results looked promising
(appendix B).
ANALYSIS
Using a maximum efficiency of 88%, the threshold
(mV) of the curve graphing efficiency as a function of threshold at 44%
was recorded. These values were converted to charge (fC) using
Teela’s calibration conversion factor and plotted as median pulse height
(50% of maximum efficiency) as a function of bias (V) (figure 10).
The data points were broken down into two parts. Where the points
were horizontal, a line was drawn. All of the other points, including
the origin were fitted to a square root function. We did this because
the charge is directly proportional to the square root of the voltage,
shown by C = E*(A/d) ~ 1/V.5. The point in which the curve and the
line meet indicates the depletion voltage using the method of charge collection.
Error in this method involves our uncertainty of the threshold value at
the 50% point by the manual method in which we chose the value. Error
bars are not drawn in the final graph because it didn’t seem necessary
since they would not be much larger than the data points themselves.
CONCLUSION
The predicted square root relationship of
charge vs. the square root of the voltage was supported with a 0.99898
correlation and 2.8405 chi squared value. Even without further analysis
of these values, they show a strong relationship between charge and voltage.
The depletion voltage determined by charge collection is 212V compared
to a depletion voltage of 200V determined by CV curve. Therefore,
it can be concluded that based on this 1.3 kohm*cm detector and our current
setup, CV and charge collection testing produce similar results.
FUTURE STUDIES
Our conclusion that CV and charge collection
testing produce similar results is not unique. Two of the three resources
that I have cited in this report had already completed testing of this
nature. The hope is to determine that the lower resistivity detectors
we have will show the same relationship between these two methods and that
it takes a longer time before the n-type bulk inverts requiring higher
and higher depletion voltages is also supported. The information
regarding the latter will be shown by plotting depletion voltage as a function
of proton fluence after radiation.
ACKNOWLEDGEMENTS
This summer was a great experience.
I openly admit that I did not have much interest in particle physics at
the beginning, but I have grown to appreciate all of the enthusiasm and
knowledge shown to me by everyone in the lab. It was also amazing
to learn that my father was involved in similar studies throughout his
career. It surprised me to discover that he made a gamma ray detector
when I was in grade school, and many years later I’m just discovering what
they are. Thank you to Tim Nissen for all of his time, wisdom, and
support and Hartmut Sadrozinski for his guidance. It was wonderful
working with and receiving extra explanations from Rachel Cannara and Jason
Hancock. Lastly, whenever I had an immediate question, Wilko Kroeger
did not hesitate to stop what he was doing to provide an answer.
References
1. Macroscopic Characterisation Methods and Results (thesis of unknown
author).
2. H. Sadrozinski, Monitoring the performance of silicon detectors
with binary readout in the ATLAS beam test,1996.
3. T. Pulliam, Noise Studies in Silicon Microstrip Detectors,
UCSC thesis 1995.
Appendices
Appendix A: initial threshold scans
Appendix B: final threshold scans
Appendix C: final graph representing depletion voltage through
charge collection
return to teacher web site home page