.3)
 Finally analysis of the whole scan is performed and the dependence of eciency,
noise occupancy and cluster size on the scanned variable (usually threshold) is plot-
ted.The results are:
{ Median
{ Cluster size at 1 fC threshold
{ Noise occupancy at 1 fC threshold
{ Visualization of the monitored values of bias voltage,leakage current,temper-
ature and power consumption of the readout electronics
{ A simple dierentiation of the multiplicity by threshold to get the spectrum
and Gaussian noise is performed and shown
{ Rate of hits of cluster sizes 1,2,...is drawn to compare the median of single
width clusters with 2- or more-wide clusters
The software became a part of the standard DAQ and analysis software for the SCT
module tests [27] and was used for all the measurements in Prague and CERN.
40
6.3 Analysis methods
The most important results from source tests are the eciency (or median),cluster
size,noise occupancy and checking of the connection (bonding) between strips and chips'
channels.This characteristics can be calculated separately for both detector planes or
for some group of strips (for example the ones connected to the same chip).The methods
of analysis are described below:
 Eciency - because there are no devices except the tested one able of precision po-
sition measurement of particle passing through the detector the eciency cannot be
calculated in the same way as in beam tests analysis.The simplest way is to mark
the event to be ecient if there is at least one hit on the detector plane.But this
absolute method is sensitive to fake triggers from the scintillator (or photomulti-
plier) that lead to eciency drop.This problem can be partially solved by selecting
a part of the beamarea and only events with hit inside this region are assumed to be
ecient.This is valid until higher noise occupancy of tested module (at low thresh-
olds) or high rate of fake triggers is reached.To avoid this,another method based
on comparison of both detector planes of the module is used.Let n
2
be the number
of events with a hit inside the selected part of the beam area at the second detector
plane.Because the angle of rotation of the planes is very small (40 mrad),if particle
passes through the module,the hit strips numbers on both planes should be close
(accounting the geometry of the module,the dierence in the strips'numbers is not
more than 60).Let's get the n
2
mentioned events and select from them the n
1both
ones where the hit strip numbers on both planes are close enough.Typical used
maximal dierence that corresponded to the beam prole coming from the setup
of the source and the scintillator,was 15.Than the eciency of the rst detec-
tor plane can be dened as the ratio n
1both
=n
2
.The error on this eciency comes
from binomial distribution:
Prob
p
(n;k) =

n
2
k
!
 p
k
 (1 p)
n
2
k
(54)
Selecting a condence level C
L
(95%),the high and low errors are determined
by probabilities p
high
and p
low
so that:
Prob
p
high
(n
2
;k  n
1both
) = 1 C
L
=2 (55)
respectively
Prob
p
low
(n
2
;k  n
1both
) = 1 C
L
=2 (56)
To speed up the error calculation (nding the values p
high
and p
low
) Gaussian ap-
proximation of binomial distribution is used:
Prob
p
(n;k) 
1
q
2np(1 p)
 exp


(k np 0:5)
2
2np(1 p)

(57)
The eciency of the second plane is obtained analogically with the rst plane as
reference.
There are 2 consequences of this method that have to be taken into account:higher
statistical errors compared to the absolute method due to lower statistics espe-
cially at higher thresholds and some articial eects at low thresholds connected
to high noise occupancy.This eects are shown in gure 27.At very low thresholds
(< 0.7 fC) the noise occupancy is so high that no matter if there was a fake trigger
41
Threshold [fC]
0 1 2 3 4 5 6
Efficiency [%]
0
20
40
60
80
100
Figure 27:The eciency dependence on the threshold.
there will nearly always be hits inside the selected part of the beam area on both
sides of the module with close strips numbers.With rising threshold the probability
to nd the close noisy hits decreases while the probability to nd noisy hit at least
at one side of the module inside the selected part of the beam area is still high.
The result is eciency drop in the surrounding of 0.8 fC.At higher thresholds also
the latter mentioned probability decreases and so the fake triggers from scintillator
result in events with empty occupancy in the selected parts of beamarea on both de-
tector planes and these events are,as described above,excluded from the eciency
calculation.
Threshold [fC]
1 2 3 4 5 6
Cluster size
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
Figure 28:The cluster size dependence on threshold.Dotted curve shows the values not
corrected to noise occupancy.
 Cluster size - to calculate the average size of the clusters it is especially at low
thresholds needed to distinguish noisy hits from the real ones caused by electron
42
passing through the detector.A simple calculation that averages over all clusters
inside a selected part of the beam area leads to underestimation of the cluster sizes
at low thresholds as the noisy clusters have usually the width of 1 (the probability
to nd 2 neighbouring noisy hits is proportional to the noise occupancy squared).
So to avoid this underestimation the largest ecient cluster inside the beam region
is considered to be caused by passing electron.Because this method would lead
to overestimation of the cluster size,corrections to noise occupancy are applied.
Let n
i
be the number of clusters of width i than this number is eectively decreased
by probability that there were 1 to i-1 noisy hits in the cluster.Consequently
n
i
must be also eectively increased by the probability that there were k noisy hits
in clusters of width i+k.Example of the graph of cluster size versus threshold is
shown in gure 28.
 Noise occupancy - the noise occupancy is dened as the number of channels,where
read noisy signal was greater than the threshold,divided by the total number
of channels.To avoid counting in the hits coming from the electron passing through
the detector,the channel clusters that are assumed to be caused by the electron are
not counted in the occupancy.The total number of channels is lowered appropri-
ately as well.Example of the graph of noise occupancy versus threshold is shown
in gure 29.The errors follow a binomial distribution approximated by the Gaussian
one of the summed noise occupancy over all events.
Threshold [fC]
0 1 2 3 4 5 6
Noise occupancy
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Figure 29:The threshold scan of noise occupancy.
 Bonding checking - the connection between strips and chips'channels can be checked
using the cluster sizes.If the connection is wrong than for example instead of clusters
of width 2 ("001100") one would observe hit patterns of 2 hits surrounding 2 channels
("010010") with no hit.Example and principles are shown in gure 30.As it
is important to have as many clusters of width greater than 1 as possible,it is
feasible by proper geometry,when electrons pass through the detector at an angle
according to the z-axis (see gure 5) in the x-direction (across the strips).To check
the bonding,histogramof readout digit-map patterns of types (see gure 30):"101",
"1001","10001",...is drawn.
43
110 0000000
1 10 0000000
Signal Signal
Detector
Bond-pad
Strip
Fan-in
Readout patterns at channels
Readout electronics
Readout electronics
Figure 30:The principle of checking for correct bonding.Correct bonding scheme and
appropriate pattern is shown in the left part,while the center gure shows readout pattern
when the connections between strips and readout channels is shifted by one pad at the top
row.The right gure shows photography of the connection between detector and fan-in
pads.
For the signal determination alternative method can be used.It is based on the deriva-
tive of multiplicity by threshold.The result is the spectrum of deposited energy and
the Gaussian spectrum of the noise.The result of this method is the most probable de-
posited energy instead of the median one calculated by the analysis above and measured
in beamtests.The most probable and median deposited energy are dierent due to asym-
metry of the spectrumof energy loss (for example see the Landau distribution in gure 4).
Because the derivative is approximated by dierence of 2 close values of the neighbouring
scan-points,high statistics and ne step of the threshold in the scan is needed.The re-
sults of this method developed by CERN SCT group are consistent with the ones coming
from the analysis described above.Due to not well dened relation between the median
and most probable energy loss and accounting the fact that the results of beam tests are
the medians,this method is not very suitable for the beam and source tests comparison.
6.4 Measurement results
The measurements were done in Prague and in CERN.In Prague 2 unirradiated for-
ward modules were tested,but the tests were focused on the development of the DAQ and
analysis software,and so only single result of 1 unirradiated module is presented.Because
of preparation of Endcap Module Final Design Review the emphasis was on tests of ir-
radiated modules.Due to formal diculties in transporting such modules from CERN,
the measurements could not be done in Prague.In CERN foremost irradiated forward
modules were tested.Several of them were tested in CERN SPS beam as well and so
comparison of source and beam tests was possible.The overall results are summed in ta-
ble 4,irradiated modules are marked by asterisk.For all the measurements my software
described in section 6.2 was used.The tests took place in August and in November and
44
December 2002.In both periods members of the Prague SCT group attended the mea-
surements in CERN.I have taken part in the tests of K5-503

module and the rst 7 test
scans of K5-504

module described in the table.All the used beam tests results come
from preliminary analysis of May and August beam tests that can be found in the SCT
beam tests WWWpage [19].
Because of multiple scattering of electrons inside the detectors,all the presented re-
sults,that were also used for comparison with the beam tests data,are the medians
from the detector plane closer to the radioactive source,where the incidence directions
of electrons'momenta were well dened.So the compared data were not so strongly in u-
enced by the multiple scattering as were the results fromthe further detector plane,where
the impact directions were hardly dened due to multiple scattering in the plane nearer
to the source.The detector plane further fromthe source has eectively the same function
as the aluminum plate or scintillator threshold:a minimum energy cut on the electrons
that passed through the whole module and additional material if present.The dicul-
ties in results from the further plane can be well understood from the angular dependence
of medians and cluster sizes,when the multiple scattering leads to longer paths of the elec-
trons in the detector and so higher medians,but the multiple scattering suppresses the me-
dians due to charge sharing.This fact is re ected in the data by systematically higher
measured cluster sizes in the detector plane further from the source,while the medians
from these detector planes were in part of the measurements higher and in the rest lower
than the ones measured in the nearer to source detector planes.The relation probably
depended on the used geometry,because the same relation of the 2 medians was ob-
served for every group of consequent measurements.The dierences between the results
from both detector planes can be also seen on the position dependence of the cluster
size (see gure 31),where a U-shape structure corresponding to the angular distribution
of incident electrons is observable in the nearer plane and quite homogenous dependence
on the further one.
Important aim of the signal measurement is its dependence on the bias voltage
1
.
Graphs of this measurements with irradiated K5-504

,and unirradiated K5-303 module
are shown in gure 32.It can be seen that even at the bias of 500 V,which is the limit
that ATLAS SCT power supplies can provide,further rise of the bias would probably
lead to higher measured signal on the irradiated module,while on the unirradiated one
the almost plateau of this dependence is reached at 150 V.These facts are in agreement
with the beam tests 2001 results of the module prototypes.The measurement at 100 V
on the K5-503

module resulted in signicantly higher cluster size at 1 fC threshold
compared to the congurations with higher bias voltage.This fact conrms charge sharing
increase due to diusion with decreasing electric eld in the detector (see gure 22).
The results of K5-503

,K5-304

irradiated modules showed low sensitivity of the me-
dians measured in the source tests setups on minor geometrical changes like the dis-
tance of the source from the module,positioning of the source so that the electrons
passed through the mechanical basement,or placement of 1 mm thick aluminum plate
between the module and the scintillator.The later change had the same eect as in-
creasing the threshold on the pulse from scintillator:eective set of minimum energy
cut on the electrons that passed through the whole module.While the median hasn't
signicantly changed,the cluster size especially on the further detector plane decreased
with the minimum energy cut application.The cluster sizes on that plane of K5-503

module are in between 1.58 and 1.61 in the basic conguration and 1.35 when aluminum
plate was used.This shows that the large cluster sizes are as expected caused by low
1
The bias voltage presented in the tables was corrected to voltage drop caused by leakage current
on serial resistor of 11 k

45
Module
Bias
Chips
Median
Noise
Cluster
Comment
[V]
[fC]
occupancy
size
Prague August:
K4-203
100
S0,S1,S2
3.30.3
(2.30.1)10
3
1.270.04
CERN August:
K5-303
100
S2,S3
2.560.07
(3.80.5)10
4
1.530.05
K5-303
150
S2,S3
3.510.06
(5.70.3)10
4
1.380.02
K5-303
200
S2,S3
3.560.13
(3.70.4)10
4
1.390.04
K5-303
249
S2,S3
3.600.08
(4.80.5)10
4
1.390.04
K5-303
300
S2,S3
3.540.14
(3.40.4)10
4
1.380.04
K5-303
350
S2,S3
3.600.07
(3.40.5)10
4
1.390.04
K5-310

488
S2,S3
2.310.07
(3.20.4)10
4
1.220.04
K5-308

312
S2,S3
2.280.08
(1.80.5)10
2
1.270.04
K5-308

304
S2,S3
2.350.07
(1.10.5)10
2
1.270.04
K5-308

409
S2,S3
2.590.07
(1.40.5)10
2
1.300.04
K5-308

489
S2,S3
2.590.07
(9.60.4)10
3
1.280.04
K5-308

600
S2,S3
2.700.08
(1.60.5)10
2
1.370.04
K5-308

471
S2,S3
2.320.10
(2.00.4)10
3
1.250.04
K5-308

478
S10,S11
2.550.14
(2.60.4)10
3
1.200.04
K5-312

341
S4,E5
2.620.10
(9.40.3)10
4
1.370.04
K5-312

415
S4,E5
2.790.18
(9.60.3)10
4
1.370.04
K5-312

490
S4,E5
2.850.09
-
-
K5-312

489
M0,S1
2.600.27
(8.30.3)10
4
1.390.04
K5-312

550
S4,E5
2.830.07
(8.30.4)10
3
1.370.04
K5-312

489
S4,E5
2.840.07
(7.40.3)10
4
1.380.04
CERN November,December:
K5-503

473
S3,S4
2.620.06
(4.30.1)10
3
1.250.04
beam through
mechanical basement
K5-503

471
E5
2.810.06
(3.70.1)10
3
1.410.04
K5-503

470
S4,E5
2.770.06
(3.20.1)10
3
1.370.04
source shifted 3 cm
further from module
K5-503

471
E5
2.730.05
(2.70.1)10
3
1.330.04
chip analog voltage
increased to 3:8 V
K5-503

473
E5
2.950.11
(3.00.1)10
3
1.440.04
edge detect type
readout mode
K5-503

473
E5
2.850.08
(2.60.1)10
3
1.270.05
1 mm Al plate between
module and scintillator
K5-504

481
S1
2.900.05
(7.10.1)10
3
1.290.04
K5-504

432
S1
2.840.05
(7.80.1)10
3
1.310.04
K5-504

383
S1
2.750.05
(8.00.1)10
3
1.290.04
K5-504

334
S1
2.520.07
(8.50.1)10
3
1.320.04
K5-504

334
S1
2.590.06
(8.00.1)10
3
1.280.04
K5-504

284
S1
2.300.08
(5.40.1)10
3
1.260.04
K5-504

480
S1
2.850.04
(5.30.1)10
3
1.240.04
higher threshold
on the scintillator
K5-504

480
S1
2.840.11
(2.90.1)10
3
1.210.04
edge detect type
compression mode
K5-504

480
S1
2.850.06
(2.90.1)10
3
1.260.04
edge detect type
readout mode
K5-504

478
M0,S1,S2
2.980.08
(5.00.1)10
3
1.250.04
source shifted 3 cm
further from module
KB-100

468
S3
2.710.04
(3.80.1)10
3
1.240.04
KB-100

420
S3
2.540.02
(3.90.1)10
3
1.230.04
KB-100

372
S3
2.470.04
(3.30.1)10
3
1.210.04
KB-100

468
S3
2.730.03
(3.20.1)10
3
1.280.04
lower threshold
on the scintillator
Table 4:The source tests results.The hit chips match the description in section 4.1.
The noise occupancy and cluster size are taken at 1 fC threshold.
46
Channel number
100 150 200 250 300 350 400 450 500
Cluster size
1
1.5
2
2.5
3
3.5
4
Channel number
250 300 350 400 450 500 550 600 650
Cluster size
1
1.5
2
2.5
3
3.5
4
Channel number
100 150 200 250 300 350 400 450 500
Number of hits
20
40
60
80
100
120
140
160
Channel number
250 300 350 400 450 500 550 600 650
Number of hits
20
40
60
80
100
120
140
160
Figure 31:The position dependence of the cluster size measured with unirradiated module
K5-303.The left gures correspond to the detector plane nearer to the source,the right
to the further plane.The top graphs show the cluster size dependence on the channel
number (impact position).The bottom histograms show the number of hits over all
the channels.The blue histograms account clusters of non-single widths only.
Bias [V]
0 100 200 300 400 500 600
Median [fC]
2.2
2.4
2.6
2.8
3
3.2
3.4
3.6
3.8
Bias [V]
0 100 200 300 400 500 600
Median [fC]
2.2
2.4
2.6
2.8
3
3.2
3.4
3.6
3.8
Figure 32:The bias dependence of the median measured with unirradiated K5-303 module
(open circles) and irradiated K5-504

module (full circles).
47
Module
Test type
Tested
Bias [V]
Median [fC]
Noise occupancy
chips
at 1 fC threshold
K5-310

Source
S2,S3
488
2.310.07
3.210
4
August Beam
S3
490
3.1
6.010
4
K5-308

Source
S3,S4
480
2.590.12
9.610
3
Source
S2,S3
471
2.320.10
2.010
3
May Beam
S2,S3
478
2.7
2.710
3
August Beam
S3
478
2.4
2.310
3
K5-308

Source
S10,S11
478
2.550.14
2.610
3
May Beam
S10,S11
478
3.1
5.810
3
August Beam
S10
478
3.0
4.010
3
K5-312

Source
M0,S1
489
2.60.3
8.310
4
Source
S4,E5
490
2.850.11
-
Source
S4,E5
489
2.840.07
7.410
4
August Beam
S2
490
3.2
4.010
4
K5-303
Source
S2,S3
150
3.510.06
5.710
4
May Beam
S2,S3
150
3.8
5.010
5
August Beam
S3
150
3.9
1.410
4
Table 5:The comparison of source and beam tests medians.The hit chips match the de-
scription in section 4.1 and are chosen from the detector plane nearer to the source.
The beam tests results were measured in 2002 and all the source tests in August 2002.
energetic electrons,because,taking into account the shape of Bethe-Bloch curve,their
mean energy loss is higher and so are the angles of multiple scattering as can be concluded
from formula (31).
The comparison to beam tests is shown in gure 33 and written in table 5.To check
the stability and errors of beam tests results,there were used more beam tests data
for every module if available.Except measurements of K5-310

and beam tests results
of K5-308

fromAugust 2001,the ratios of medians fromsource and beamtests are similar
over the modules.Excluding the 2 mentioned results,the ratio of medians from the beam
tests to the source is about 1.1090.070.The presented value was obtained as an av-
erage weighted by inverse of the errors squared and the error is the standard deviation
of the ratios from the average value.Due to multiple scattering,the cluster sizes mea-
sured in source tests (between 1.2 and 1.6 at 1 fC threshold) are higher than the values
measured in beam tests (around 1.06 at 1 fC threshold for both the irradiated and unir-
radiated modules [10]).The comparison of measured noise occupancies of the modules
results in good agreement in the orders.The values could hardly be in more precise agree-
ment,because the noise occupancy is very sensitive to the used conguration especially
correct grounding,cable shielding,temperature,calibration,...This fact can be also seen
on the dierences between the pairs of beam tests results.
6.5 Source tests simulation
To fully understand the dierences between the results of source and beam tests,sim-
ulation of the source tests setup was performed.Because the multiple scattering is more
important for electrons than for 180 GeV pions passing through the detector,the step
in Geant4 simulation was lowered down to 20 m,but the settings in the digitization was
used the same as for the simulation of the -plots at incidence angle 16
o
:step 10 m and
48
1 2 3 4 5
Median [fC]
2.2
2.4
2.6
2.8
3
3.2
3.4
3.6
3.8
4
1 2 3 4 5
Median [fC]
2.2
2.4
2.6
2.8
3
3.2
3.4
3.6
3.8
4
K5-310
S3
S2,S3
S2,S3
K5-308
S3
S10,S11
S3,S4
K5-308
S10
S2,S3
S2,S3
K5-312
S2
S10,S11
K5-303
S3
M0,S1
S4,E5
S2,S3
Figure 33:The comparison of the beam and source tests medians.
single drifting charge carrier per track segment.3 setups were simulated:with and with-
out the 1 mm thick aluminum plate and with source shifted 3 cm further from the mod-
ule.The position of the scintillator,module and source in the basic geometry followed
the real geometry of the measurements:both the distances of the scintillator and source
from the module were approximately 2 cm.In spite of the fact that forward modules were
tested,the simulation was performed using barrel module description.The only dierence
in the digitization software is the strip pitch.Because both the beam and source tests
results showed low sensitivity of the medians on the strip pitch and simulation itself is able
to describe trends of measured dependencies only,there should be no critical in uence
on the results of using the barrel module layout instead of the forward one.
The results are summed in table 6.The simulation conrmed the dierence of medians
between beamand source tests,when the simulation gave the ratio of beamtests to source
tests medians 1.1170.020.The value is within the errors in agreement with the measured
ratio:1.1090.070.The shapes of the eciency and cluster size versus threshold curves
(see gures 34,35) are in good agreement as well.The simulation also conrmed higher
cluster sizes,measured during source tests,compared to the beam tests results and higher
cluster sizes at the plane further from the radioactive source compared to the nearer one.
The absolute values of cluster sizes and medians were not confronted with the simulation
because the simulation was not able to reproduce absolute values of beam tests.
Because the simulation software provides the simulated data in the same format as
the real data are stored,it was possible to validate the eciency,cluster size and noise oc-
cupancy calculation methods described in section 6.3.All dierences were within the cal-
culated errors,but systematical:the cluster sizes from the digitization are approximately
3 % lower compared to the calculation of the analysis software using the map of hits,
while the eciency is higher about the same systematical dierence.
The trends of the characteristics across the 3 simulated geometries are in agreement
with the measurements:
 Minor change of the median on the nearer detector plane,while the median on plane
further from the source is more sensitive.
 Lower cluster size,especially on the further plane,in the setup with aluminumplate.
49
Bias [V]
Plane
Median [fC]
Noise
Cluster
Comment
occupancy
size
Source tests:
200
nearer
2.790.08
(1.390.18)10
4
1.380.03
further
2.890.08
(2.560.24)10
4
1.520.03
200
nearer
2.770.06
(1.610.17)10
4
1.390.02
source shifted
further
2.820.07
(2.840.23)10
4
1.660.03
3 cm away
200
nearer
2.680.07
(8.700.13)10
4
1.300.02
aluminum plate
further
2.620.07
(1.130.15)10
4
1.360.03
Beam test:
200
nearer
3.070.02
(1.070.06)10
4
1.0560.003
further
3.050.02
(1.360.06)10
4
1.0740.004
Table 6:The source and beam tests simulation results.The noise occupancy and cluster
size are taken at 1 fC threshold.The presented errors of beamtests simulation correspond
to dierence between the rst and second detector plane.
Threshold [fC]
0 1 2 3 4 5 6 7
Efficiency [%]
0
20
40
60
80
100
Threshold [fC]
0 1 2 3 4 5 6 7
Efficiency [%]
0
20
40
60
80
100
Threshold [fC]
0 1 2 3 4 5 6 7
Efficiency [%]
0
20
40
60
80
100
Threshold [fC]
0 1 2 3 4 5 6 7
Efficiency [%]
0
20
40
60
80
100
Figure 34:The eciency dependence on the threshold for both the source (circles) and
beam tests (triangles).Simulation in the basic geometry is shown in the left gure,source
measurement of K5-503

module and test beam of K5-312

module in the right.
Threshold [fC]
0 1 2 3 4 5 6 7
Cluster size
1
1.1
1.2
1.3
1.4
1.5
1.6
Threshold [fC]
0 1 2 3 4 5 6 7
Cluster size
1
1.1
1.2
1.3
1.4
1.5
1.6
Threshold [fC]
0 1 2 3 4 5 6 7
Cluster size
1
1.1
1.2
1.3
1.4
1.5
1.6
Threshold [fC]
0 1 2 3 4 5 6 7
Cluster size
1
1.1
1.2
1.3
1.4
1.5
1.6
Figure 35:The cluster size dependence on the threshold for both the source (circles)
and beam tests (circles) and for both the nearer (open marks) and further (full marks)
detector plane.Simulation in the basic geometry is shown in the left gure and source
measurement of K5-503

module in the right.
50
7 Conclusion
A method using radioactive 

source for detection capability check of ATLAS silicon
microstrip detectors has been developed.This method provides relatively fast measure-
ment of the electrons median energy deposit in 285 m thick silicon wafers.Compared
to the beam tests,the advantage of source tests is the possibility to build the setup at any
working place equipped with standard tools needed for QAtests of the SCT modules.Due
to the simplicity of the source tests setup,module response in various geometrical cong-
urations,like set of incidence angles,can be studied.Great advantage of the source tests
is their availability during the whole year without complicated preparation compared to
the beam tests,that take place few times a year.The easy reproduction of electrical and
geometrical setting makes the source tests applicable for repeated tests of modules for ex-
ample before and after irradiation or any other intervention to the module functionality.
But the simplicity of the setup also leads to the main disadvantages of the source tests.
Firstly compared to the beam tests,the track positions of the passing electrons are known
only with the precision and limitations (depending on the set threshold etc.) of the tested
module.Secondly,as was described in chapter 6.3,the eciency calculation method
doesn't provide correct values at low thresholds,where the requirements on module ef-
ciency and noise occupancy are dened.And so it is needed to judge on the eciency
from the measured median signal as there must be xed relation between these 2 char-
acteristics.The next disadvantages are connected with the used source:the particles
passing through the detector are not monoenergetic,their initial tracks are not parallel
and due to low energy and mass of the electrons,multiple scattering has more signicant
in uence compared to the beam tests,which was observed in the source tests data,and
conrmed by simulation,as wider signal clusters.
Analysis and data acquisition software for the source tests was written by myself.
Based on the DAQ software used for standard module tests,it assures necessary mea-
surement features and calculation and visualization of basic tested module characteristics
as the eciency,noise occupancy,width of hit clusters,etc.Development of the soft-
ware provided me great experience in using ROOT [17] tools for the analysis and DAQ
purposes.
Simulation of the module response in both the beam and source tests was performed.
The main aim was to describe the dierent results from the 2 setups.In spite of the fact
that the simulation involves both the detector and readout electronics functionality,it was
not possible to reach agreement with the beam tests data on the level of absolute values
of measured signal's height and width.But trends of angular,bias,etc.dependencies
of the resulting signal from the simulated readout electronics follow the data trends.I
tried to nd mechanism causing the discrepancies rstly by simplifying the model to pure
geometry and than in the opposite by involving in the current induction on the strips
from the drifting free carriers.But the results stayed close to what the original model
predicted.
To verify the usability of source tests,it was needed to nd the relation of their results
to the beam tests ones.Having taken into account the diculties of the simulation soft-
ware,I compared the ratio of the simulated median signals from the 2 tests to the ratio
obtained from the real data analysis.Both the measurements and simulations resulted
in higher signal in the beam tests compared to the source ones,and the measured and
simulated average ratios 1.1090.070 and 1.1170.020 are in agreement within the writ-
ten standard deviations.The simulation also conrmed lower sensitivity to geometrical
settings of the signal measured at detector plane nearer to the source compared to the fur-
ther one,and so to get stable results,it is suggested to test every detector plane separately
51
by proper placement of the source.The desired result is to have dened relation between
the source and beam measurements,but the presented values show quite large spread
of the beam-to-source tests ratios of signals.There are several eects that cause devia-
tions of the measurement results.Firstly,the calibration process,that is made separately
for every chip,has nite precision,that was observed on the beamtests results as dierent
measured signals depending on from which chip were the data analyzed.These standard
deviations are around 0.2 fC [19] and are re ected in the errors of ratios by approximately
0.026.Secondly,the sensitivity of the median signal,measured is source tests on the de-
tector plane nearer to the source,is lower than in the further-from-source plane,but some
dependence within 0.1 fC is still there and is re ected in the error of beam to source tests
signals ratio by around 0.013.And at last contribution of eciency calculation statistical
errors has to be considered.
The discussion above shows that source tests are able to provide median signal mea-
surement in ATLAS SCT detectors almost as precisely as the beamtests.Some of the pre-
sented source tests results were involved in the Endcap Module Final Design Review
(CERN August 2002).The software used for the source tests became a part of the stan-
dard DAQ package used for QA module tests.
52
References
[1] ATLAS Collaboration (1994):ATLAS Technical Proposal for a General-Purpose pp
Experiment at the Large Hadron Collider at CERN,CERN,Geneva
[2] ATLAS Inner Detector Community (1997):ATLAS Inner Detector Technical Design
Report,Volume I,CERN,Geneva
[3] Kittel C.(1996):Introduction to Solid State Physics,John Wiley and Sons,New
York
[4] Lutz G.(1999):Semiconductor Radiation Detectors,Springer,Berlin
[5] Sauli F.(1993):Instrumentation In High Energy Physics,World Scientic Publishing
Co.Pte.Ltd.,Singapore
[6] CERN Program Library documentation of Geant (detector description and simula-
tion tool)
http://wwwinfo.cern.ch/asdoc/geantold/H2GEANTPHYS430.html
cited 15.04.2003
[7] Wagner W.et.al.(1998):Characterization of Silicon Microstrip Detectors Using
an Infrared Laser System,Werner-Heisenberg-Institut,MPI-PhE/98-13
[8] Hagiwara K.el.al.(2002):Particle Physics Booklet,CERN,Geneva
[9] Knoll G.F.(1979):Radiation detection and measurement,John Wiley and sons,
New York
[10] Barr A.J.et.al.(2001):Beamtests of ATLAS SCT Modules in August and October
2001,ATL-INDET-2002-024,CERN,Geneva
[11] Documentation of ATLAS SCT readout electronics:ABCD3T chip specication ver-
sion V1.2,24/7/2000
http://scipp.ucsc.edu/groups/atlas/elect-doc/abcd3t_spec.pdf
cited 15.04.2003
[12] Documentation of ATLAS SCT Low Voltage VME module providing power supply
to the readout electronics
http://www-hep2.fzu.cz/Atlas/WorkingGroups/Projects/MSGC.html
cited 15.04.2003
[13] Documentation of ATLAS SCT High Voltage VME module providing bias voltage
on detectors
http://chall.ifj.edu.pl/~atlas/SCTHV/SCTHV_PAJ.html
cited 15.04.2003
[14] Documentation of Multichannel Semiconductor Tracker ABC(D) Readout VME
module used in ATLAS SCT module tests
http://hepwww.rl.ac.uk/atlas-sct-mm/Mustard/
cited 15.04.2003
[15] Documentation of Clock and Control VME module used in ATLAS SCT module
tests
http://www.hep.ucl.ac.uk/atlas/sct/#CLOAC
cited 15.04.2003
53
[16] Documentation of Slow Command Generator VME module used in ATLAS SCT
module tests
http://hepwww.rl.ac.uk/atlas-sct-mm/Slog/
cited 15.04.2003
[17] Home page of an object-oriented data analysis framework ROOT
http://root.cern.ch
cited 15.04.2003
[18] Description of the ATLAS SCT module tests data acquisition system
http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/SCT/testdaq/
cited 15.04.2003
[19] Home page of the ATLAS SCT beam tests
http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/SCT/testbeam/
cited 15.04.2003
[20] Horejs

i J.(2003):Fundamentals of Electroweak Theory,Karolinum,Praha
[21] Description of 

spectroscopy experiment
http://www.ph.unimelb.edu.au/studentresources/undergraduate/
part2labs/pracs/beta_v2.doc
cited 15.04.2003
[22] Home page of a detector description and simulation tool Geant4
http://geant4.web.cern.ch/geant4
cited 15.04.2003
[23] Home page of the ATLAS SCT simulation
http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/SCT/simulation/
cited 15.04.2003
[24] Documentation of the ATLAS SCT beam test digitization software written by Szy-
mon Gadomski
http://gadomski.home.cern.ch/gadomski/SCT_Digitization.ps
cited 15.04.2003
[25] Gadomski S.(2001):Model of the SCT detectors and electronics for the ATLAS
simulation using Geant4,ATL-SOFT-2001-005,CERN,Geneva
[26] Broklova Z.(2003):Vyhodnocen ucinnosti a kvality polovodicovych stripovych de-
tektoru pro detektor ATLAS (LHC CERN),Diploma thesis at Faculty of Mathemat-
ics and Physics at Charles University,Prague
[27] Eklund L.et.al.(2003):Electrical Tests of SCT Hybrids and Modules,
ATL-COM-INDET-2003-004,CERN,Geneva
[28] Home page of Prague SCT group at Institute of Particle and Nuclear Physics
at Charles University in Prague
http://www-ucjf.troja.mf.cuni.cz/~sct
cited 15.04.2003
54