Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
Performance of ACCOS, an Automatic Crystal quality Control
System for the PWO crystals of the CMS calorimeter
E. Au!ray *, G. Chevenier , M. Freire , P. Lecoq , J.M. Le Go! ,
R. Marcos , G. Drobychev, O. MisseH vitch, A. Oskine, R. Zouevsky,
J.P. Peigneux, M. Schneegans
CERN, DIV.PPE/130, Bat. 27, CH-1211 Geneva 23, Switzerland
INP, Minsk, Belarus
LAPP, Annecy, France
Received 15 December 1999; accepted 11 March 2000
Abstract
Nearly 80 000 PWO crystals for the CMS electromagnetic calorimeter will arrive at CERN/Geneva and INFNENEA/Rome between now and year 2004. The stringent speci"cations on their dimensions and optical quality have to be
veri"ed prior to their formal acceptation. Automatic systems for measuring the critical parameters of each crystal and
recording them in a database have been designed and constructed. The "rst machine is now in stable operation at CERN.
In this note, the performance of each instrument, based on the measurements on &1000 pre-production crystals, is
analysed in terms of stability and compared to the results of conventional benches. 2001 Elsevier Science B.V. All
rights reserved.
1. Introduction
The CMS collaboration [1] plans to use 76 200
tapered lead tungstate crystals (PWO) for its highresolution electromagnetic calorimeter [2]. These
crystals, produced in Russia (Bogoroditsk) and in
China (BGRI and SIC) will be delivered to two
CMS Regional Centres: CERN/Geneva and
INFN-ENEA/Rome, where the calorimeter will be
assembled. In order to guarantee that all the delivered crystals meet the speci"cations [3] on the
* Corresponding author. Tel.: #41-22-767-58-44; fax: #4122-7678930.
E-mail address: [email protected] (E. Au!ray).
parameters considered as crucial for the calorimeter performance, several measurements must be
performed.
First, the dimensions of each crystal must be
measured to a precision of &10 lm, in order to
guarantee that the crystal can be mounted in its
individual glass "ber alveolus with the necessary
clearance. Then, the optical quality of the crystals
must be checked, namely its optical transmission.
The transversal transmission allows to check the
homogeneity of the transmission along the crystal,
whereas the longitudinal transmission guarantees
a long absorption length and reveals possible absorption bands or poor radiation resistance [3].
The light yield of the crystals, in fact the light
collected on a photodetector seeing their largest
0168-9002/01/$ - see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 9 0 0 2 ( 0 0 ) 0 0 6 6 5 - 3
326
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
section, is an essential parameter for the energy
resolution (stochastic term) of the future calorimeter, as well as the uniformity of the light collection along the crystal (constant term of the
resolution). Moreover, this light must be collected
in a short time (in 4100 ns), with absence of slow
components (in the range of a few 100 ns) which can
a!ect the pulse shape, and as little as possible
contribution of longer components (afterglow) to
avoid pile-up e!ects.
All these parameters are measured currently on
classical benches (Light Yield benches, spectrometers), but the duration of the measurements and of
their analysis, as well as all the crystal manipulations needed, would make the control of several
ten-thousands of crystals fastidious, costly in manpower and risky for the crystals. This is why systems of benches (named ACCoS, Automatic
Crystal Control System) were designed [4}6]
allowing measurement of these parameters by appropriate methods in a fully automatic mode, including transfer of all the data to a database.
The "rst machine was built in LAPP/Annecy in
collaboration with INP/ Minsk, and installed at
CERN (ACCoCE1, Fig. 1) in summer 1998 and
extensively tested since. It is now in steady production mode: in fact, more than 1000 pre-production
PWO crystals from Bogoroditsk Technical Chemical Plant in Russia were recently measured. In this
note, based on these measurements, the level of
stability reached by each instrument and its performance compared to classical benches are shown
and discussed after a brief description of the system
and its operation.
Fig. 1. Photograph of ACCoCE 1.
Fig. 2. ACCoCE 1 design.
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
327
Fig. 3. View of the 3D machine (Johansson Topaz 7).
2. Main features of ACCoCE1
The machine is composed of a circular server
presenting the crystals in turn to a dimension
measuring machine and to optical benches, either
"xed or mounted on a linear movement bolted to
a rectangular table (see Fig. 2). The 110 cm diameter server for 20 crystals arranged radially is bolted on a turntable driven by a step by step motor
yielding a crystal positioning precision of a few
tenths of mm. The crystals have their section tilted
by 60 and are supported horizontally by two thin
vertical pillars allowing all measurements with
minimum limitations. The small section is always
placed facing the centre of the circle and carries
a label with barcode. A "xed barcode reader
identi"es each crystal as it stops facing it during an
initial turn of the server. Each crystal number is
then associated to a position on the server in
a look-up table.
Hanni"n, Motion & Control, Digiplan, Poole, Dorset, UK;
or Groupe Jeambrun, Lyon, France.
Barcode reader Symbol LS 1220.
To measure the dimensions of the crystals, a classical three-dimensional (3D) machine, capable of
a precision of &5 lm per point and with an open
geometry (Fig. 3) for easy crystal presentation by
a circular server, was chosen. Each crystal in turn is
presented to the dimension measurement with
end-faces vertical or nearly and side faces at 30 or
603 from vertical. Fifteen points are measured on
each lateral face of the crystals and nine points on
the two end faces, allowing precise reconstruction
of the six planes. In order not to lose precision by
probing slanting faces, the 2 mm diameter balls,
ending the "ve stylets of the star-shaped probe,
approach the crystal surface slowly and perpendicularly on the last millimetres. A calibration on
a reference ball (&30 mm diameter) allows to rede"ne the origin of the coordinate system and to
recalibrate the diameter of the probe balls for corrections. The geometry of the crystals for each of
the di!erent types is determined by a set of seven
variables (Fig. 4) for left and right parity (mirror
symmetry). Side P is always placed on the support
Johansson Topaz 7, Eskilstuna, Sweden; or Marne-laValleH e, France.
328
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
pillars with the same orientation (at 60 from vertical, whether left or right). Side D is depolished by
the producer to a speci"ed [3] degree of roughness,
in order to obtain light collection uniformity.
Two compact (7;8;22 cm) spectrometers
measure the optical transmission through the crystal length or width of a light beam produced by
a 20 W halogen lamp. The eleven discrete
wavelength values are de"ned by interference "lters
chosen for best determination of the band edge
slope and of possible absorption bands. A large
area UV-extended pin photodiode (PD) [4] detects
the transmitted light.
For Transversal Transmission Optics (TTO), the
light beam always passes through the optically
polished sides N and W and the PD is located at
&5 cm from the light emitting optics. The spectrometer, "xed on a chariot mounted on a linear
motorised movement , (see foot note 1) moves along
the crystal and stops every 2 cm in 11 points start-
ing at 1.5 cm from the small end. The diameter of
the light beam is adjusted to &10 mm. A calibration measurement in air is taken at a position
outside of the crystal.
For Longitudinal Transmission Optics (LTO),
the PD is located at &27 cm from the light emitting optics which was optimised for such a distance:
quasi-parallel beam of &7 mm diameter entering
the crystal and &9 mm exiting. This diameter provides adequate sensitivity to clouds of scatter
points sometimes present near the central axis of
the crystal. The Longitudinal Transmission (LT)
optics and its PD are in a "xed position and the
light beam is centred on the crystal when it stops in
its measuring place. A calibration measurement is
performed in air between each crystal.
To measure kinetic parameters, Light Yield (LY)
and LY Non-Uniformity (NU), a method allowing
to derive the three parameters from a single time
spectrum was chosen. One reason for preferring
a time measurement is the relatively low light yield
of PWO crystals, which makes direct light yield
measurement with a source and PM di$cult unless
good optical contact is performed, which is rather
unpractical for a fully automatic measuring device.
The decay time of the PWO crystal's scintillation
light is measured using the detection by photomultipliers (PMT) of the two collinear 511 keV
gamma-rays from a Na source. In fact, c detec
ted in a small BaF crystal produces the Start
signal and the detection of c in the PWO yields the
Stop signal. The time di!erences are analysed in
a TDC, capable of multi-hit event rejection [7]. For
the time spectrum shape and the LY measurement
to be correct, the detected light of PWO must not
be larger than a few tenths of photoelectrons per
Na event, which is readily obtained with a PM
(square 3;3 cm) seeing the crystal through its
largest section at a distance of 1}2 cm. The
BaF /Na/PWO geometry is very compact in or
der to maximise the c solid angles and thus the
rates. The Start telescope, comprising PM with
divider, small crystal and source, is moved along
the PWO crystal on the same chariot as the TTO
Radiation Instruments & New Components Ltd., (RI&NC),
Minsk, Belarus.
PMTs for Kinetics 5900U, Hamatmatsu UV-extended
photomultipliers (28;28;20 mm).
Fig. 4. Crystal geometry.
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
spectrometer, with 21 positions of measurement,
each cm, starting at 1.5 cm from small end.
Normally, the crystals are "rst measured in dimensions, then LT followed by decay time and
"nally TT is measured on each crystal. The sequence
is fully automatic with always indication of what is
being measured and online presentation of the measured spectra. For each crystal number, ACCoCE
communicates with the C.R.I.S.T.A.L. [8] system
and its &Objectivity' database, in order to know what
has to be measured next. After each type of measurement, some calculations take place and all the data
are transferred to the database. The present total
time is about 7 h for 20 crystals. One crystal is in fact
a reference crystal, which is never moved and allows
to check all the measurements stability.
3. Dimension measurement
For each face of a measured crystal, a plane is
"tted to the measured points. The planarity of this
face is de"ned as the distance between the two
planes parallel to the "tted plane and passing by
the points most distant from it, outside and inside.
With these planes, three volumes can be de"ned:
the average volume corresponding to the "tted
planes, the outside volume and the inside volume.
The corners are then de"ned as intersection of three
planes and the dimensions are derived according to
their de"nition (see Fig. 4). The average, outside
and inside volumes yield respectively the average,
the maximum (relevant for penetration in alveolus)
and minimum dimensions.
Four values of the length are de"ned: one is along
the side line perpendicular to the end faces (PW
intersect); the three others are de"ned as the length
between the three other small section corners and
their projections on the large section, parallel to the
PW intersect line. Four length values can also be
de"ned for the outer and inner crystal volume. For
fast comparisons or for physics, the average of the
four average length values can be taken.
A steel standard (SSD) was manufactured to the
dimensions of a type six left crystal to a precision of
the order of 10 lm. The SSD was measured three
times in each of the 19 server positions, i.e. 57 times
with SSD removed and repositionned each time. The
329
Table 1
Planarities and dimensions of the Steel Standard measured by
ACCoCE1 and CERN Metrology
Planarities
Nominal
*Accos 57
*MetroCERN
F
R
P
D
N
W
0
0
0
0
0
0
3
2.5
4.5
7
5
12
3
5
4
2
5
9
Dimensions
Nominal
*Accos 57
*MetroCERN
AF
BF
CF
AR
BR
CR
L
21.831
22.601
21.999
25.435
26.182
25.630
230
!20
#15
!17
!7
!17
!3
!9
!7
#6
#1
#8
!12
#16
#10
distributions of front and rear dimensions (Fig. 5)
show a very good reproducibility (p"0.8}1.3 lm).
The planarity of each of the 6 faces is rather small
and in good agreement with SSD measurements at
the CERN metrology service (Table 1). The dimensions, compared to CERN metrology and to the
nominal values for Type 6, show agreement well
within 20 lm (Table 1). More precise knowledge of
the SSD's e!ective dimensions is needed to establish the absolute precision of ACCoCe1.
Long-term stability of dimension measurements
can be seen (Figs. 6 and 7) on 61 measurements of
the Reference crystal: R0488 (Type 6L), spread over
a period of 6 months. A small shift can be observed
on L2, L3 (&6 lm) and B (&12 lm) dimensions
after February 20, which came back to the previous
level after a calibration on March 5 and stayed very
stable ever since. In fact, after March 5, a 3D
calibration is systematically performed at the beginning of each run. The dispersions of all dimensions show sigmas below 4 lm.
4. Transmission measurements
The transversal transmission (TT) spectrum was
measured many times along the reference crystal
330
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
Fig. 5. Distribution of the 57 steel standard measurements for dimensions front and rear, respectively.
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
Fig. 6. Stability of length (a), front (b), and rear dimensions
(c) for the reference crystal measured from February to July 1999
(61 measurements).
(R0488). Fig. 8 shows the very good superposition
observed for sampled transmission curves measured from April to July at positions D"15, 115,
210 mm from the crystal small end. Fig. 9 presents
the distribution at three wavelengths of the 62 measured TT values at the three positions along the
crystals. The observed stability: p/mean &0.2%
(&1% at 350 nm) is excellent.
TT measurements on ACCoCE1 of &20 crystals were compared to measurements on the
331
CERN/Lab27 Reference spectrometer [9]. Fig. 10
shows correlation and di!erence in transmission
between the two instruments at three wavelengths,
using the measurements at all positions along the
20 crystals. The observed correlation is restricted
by the small variation in transmission values from
crystal to crystal. The larger dispersion and
the shift of &2% observed for transmissions at
350 nm can be attributed to uncertainties in the
wavelength de"nition in a spectrum region where
the wavelength dependence of the transmission is
very large.
Longitudinal transmission was measured 76
times on crystal R0488, from April to July 99.
Fig. 11 illustrates the very good superposition of
some sampled curves. Fig. 12 shows the distributions at three wavelengths of the 76 LT measurements. A very small dispersion of the order of 0.5%
is observed. At 350 nm, the e!ect of a temporary
"lter drift, later cured, can be observed.
Fig. 13 represents the correlation and di!erence
at three wavelengths between LT measured on the
reference spectrophotometer and ACCoCE1 for
104 crystals. The observed dispersions are small
with again a &3% systematic shift for 350 nm.
Similar to our reference spectrophotometer in
B27, the ACCoCE1 spectrophotometer has been
designed to provide adequate sensitivity to the possible presence of cloudy cores of bubbles or impurities along the axis of the crystal on part or totality
of its length. When the density of such a core is high
enough, it may a!ect the light yield of the crystal
and in particular the uniformity of light collection,
thus making the uniformisation process hazardous.
High densities of scatters in crystals also a!ect the
response of the monitoring system in uncontrolled
ways and must therefore be avoided. For spectrophotometers not using integration spheres, the level
of the longitudinal transmission is a!ected, which
in principle permits rejection of the crystal. But the
diameter of the light beam must be optimised for
a moderate sensitivity to these clouds but not to
miss them if they are o!-axis. We found that a diameter of 7 mm at crystal entrance resulting in 9 mm
at crystal exit was optimum. Fig. 14 illustrates the
very good agreement observed between ACCoCE1
and B27 reference spectrometer for LT measurements of two good crystals and a still satisfactory
332
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
Fig. 7. (a) Histogram of the length dimensions for the reference crystal measured from February to July 1999. Histogram of the (b) front
and (c) rear dimensions for the reference crystal measured from February to July 1999.
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
Fig. 8. Superposition of TT spectra measured at three positions along the reference crystal from April to July.
333
334
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
Fig. 9. Distribution of the 62 TT measurements at three positions along the reference crystal for 350, 420 and 620 nm.
Fig. 10. Correlation & di!erence between TT at three wavelengths values for each position of 20 crystals measured on ACCOS & B27
Reference spectrophotometer.
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
335
Fig. 11. Superposition of LT spectra measured from April to July on the reference crystal.
agreement when the LT level is near or below
speci"cation.
A good stability of the band-edge slope, as de"ned
by a linear "t to three points on transmission rise, can
be observed in Fig. 15(a). Good correlation and small
di!erences between slopes measured on our reference
spectrophotometer in B27 and on ACCoCE1 spectrophotometer can be seen in Fig. 15(b and c).
336
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
the parameters of the exponential "tting of the
decay time spectrum. In fact, for all pre-production
crystals measured, the ratio LY(100)/LY(1000) was
found to be above 99%.
The relative LY value (in % of e$ciency) in each
position along a crystal is derived from the Poissonian probability of detecting zero light photon
produced in PWO by c from a Na source event,
when c is detected at 1803 in the Start telescope.
This probability is proportional to the crystal light
yield. To obtain absolute LY values in photoelectrons/MeV, average normalisation coe$cients
were determined for typically 10 crystals measured
both on ACCoCE1 and on the CERN/B27 pulse
height analysis bench: B3.
For LY pro"les and NU Factors, corrections for
several di!erences in measuring conditions should
be applied:
E Di!erent geometry causing di!erent Compton
contributions,
E For ACCoCE1: absence of wrapping, of optical
contact; distance to PM of &1 cm,
E Di!erent PM quantum e$ciency spectra causing
di!erent sensitivity to the emitted light, waveshifting along the crystals due to light absorption.
Fig. 12. Distribution of the transmission values for 76 measurements from April to July 1999 at three wavelengths on the
Reference crystal.
5. Decay time, light and uniformity measurements
On ACCoCE1, a full crystal decay time spectrum
in the range 0}1000 ns is recorded in 21 points
along the crystal, allowing an e$cient detection of
possible slow and medium-slow (the most dangerous for pulse shape determination) components.
For a quick check of the crystals kinetic parameters, the LY fraction collected in the "rst 100 ns
of the spectrum relative to the light in 1000 ns is
derived from the recorded time spectra, as well as
These corrections being relatively small and di$cult to calculate with su$cient accuracy, it was
preferred to include an empirical correction in the
normalisation coe$cients 1C 2 of the ACCoCE1
G
LY, for each position i along crystals.
For all crystals, each position on the ACCoCE1
LY curves was then multiplied by the relevant
coe$cient
LY (pe/MeV)"(C ';LY (ACCoCe1).
G
G
G
Typical values of these coe$cients are shown in
Table 2. This set of coe$cients is checked for each
batch and for each type of crystals.
A good stability of the normalised and corrected
LY curves, measured 43 times on the reference
crystal R0488 from April to July 99 is observed on
17 sampled curves in Fig. 16(a).
The distribution of the LY measured at 8X
depth in the crystal (shower maximum) can be seen
on Fig. 16(b) for the 43 measurements on crystal
R0488. Fig. 16(c) shows the distribution of the NU
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
337
Fig. 13. Correlation and di!erence between LT values at three wavelength of 104 crystals measured on ACCoCE1 and B27 Reference
spectrophotometer.
Fig. 14. LT spectra for two good clear crystals and for two crystals near speci"cation limit due to cloudy cores.
338
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
Fig. 15. (a) Distribution of the band edge slope on 76 measurements of the reference crystal from April to July 99. (b) Correlation and di!erence between band-edge slope values of 104
crystals measured on ACCOS and (c) B27 Reference spectrophotometer.
Factor (NUF "relative slope of the linear "t to
the LY values in "rst half of crystal) for the 43
measurements of the reference crystal.
Fig. 16. (a) Superposition of the LY non-uniformity curvesmeasured from April to July on the reference crystal. (b) Distribution of respectively the normalised LY at 8X values and (c)
FNUF for 43 measurements from April to July 1999 of the
reference crystal.
Fig. 17 shows the correlations and di!erences
between B3 and normalised ACCoCE1 measurements for LY and NUF for 54 crystals not used in
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
339
Table 2
Normalisation coe$cients (including corrections) for 21 positions along crystals
Pos. 15
(mm)
Ci
25
35
45
36.6 34.4 33.9 33.5
55
65
33.3 33.1
75
85
33.1 32.9
95
105
115 125
135 145
155 165
175 185
195 205
215
32.8 32.9
32.9 33.0
32.8 33.1
33.1 33.4
33.8 34.6
35.4 36.2
37.0
Fig. 17. Correlation and di!erence between the LY and NUF values of 54 crystals measured on ACCoCE1 and B27 Reference LY
Bench (B3).
Fig. 18. Photograph of ACCoCE machine for Bogoroditsk.
340
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
the normalisation procedure. The observed LY correlation is good with a dispersion: p&7% and no
signi"cant residual di!erence, taking into account
that crystals of four types are normalised using the
same coe$cients. The observed NUF correlation is
excellent with a dispersion: p&0.12%/X and no
signi"cant residual di!erence showing that 10 crystals for determining the normalisation coe$cients
are su$cient in statistics.
for the delivery of 11 000 BGO crystals, has never
been used at a scale as large as for CMS: 80 000
PWO crystals. Moreover, the installation of similar
machines at the producer plants will mean double
control of each crystal, not only reducing the percentage of rejected crystals at reception, but also
giving best chance of high-quality crystals with
well-known and recorded characteristics.
Acknowledgements
6. Conclusion
The reproducibility and long-term stability of
ACCoCE1 measurements having been established
and con"rmed by the measurement since April 99
of more than 1000 pre-production crystals, we can
state that the machine is now ready to face control
of the crystal production at a rate of currently
2;20 crystals/day (&60 in 3 shifts, if necessary).
ACCoR (Rome) will soon be operating at a similar
capacity.
To reduce the fraction of crystals to be rejected
because they do not satisfy to speci"cations (at
present ranging from 16% to 20%), it was agreed
to equip the producers with similar systems. A
second ACCoCE machine, with increased capacity
(30 crystals) and assisted loading (no manual
handling of the crystals) (Fig 18), was constructed
and commissioned at CERN, before being transported and installed in Bogoroditsk (Russia) end
of July.
To face deliveries at CERN, which will constantly be near the 1000/month level in years 2000}2005,
means to be capable of controlling up to 100 crystals/day during peak periods, if one takes into
account remeasurements, maintenance, special
tests and breakdowns. This is why a third machine
with same capacity and loading as ACCoCE2 will
be assembled in fall at CERN and should be in
operation beginning of year 2000. ACCoCE1 will
later be upgraded, so that CERN Regional Centre
will have a capacity of &120 crystals/day in two
shifts.
The successful start of ACCoS machines also
highlights the success of a concept: the full characterisation of each delivered crystal. This concept,
already applied with success in L3 experiment
We are deeply indebted to A. Federov and M.
Korjic (INP Minsk) for their essential contribution
to ACCoS conceptual studies and for their constant
support.
We are very grateful to G. Dromby, P. Letournel
and A. Oriboni for their excellent work on the
prototype construction at LAPP and also to B.
Buisson, H. Cabel, A. Conde-Garcia, A. Deforni, R.
Morino and V. Panov for their invaluable help in
installing, tuning and running ACCoCE1 at
CERN. We are also indebted to our Software team
for providing the very powerful C.R.I.S.T.A.L system, particularly to Z. Kovacs and M. Zsenei.
We would like to thank warmly all our collaborators of the CMS/ECAL community and in
particular the ECAL/Rome group for many illuminating and stimulating discussions.
References
[1] The Compact Muon Solenoid, CMS Technical Proposal,
CERN/LHCC 94-38, LHCC/P1, December 15, 1994.
[2] The Electromagnetic Calorimeter Project, Technical
Design Report, CERN/LHCC 97-33, CMS TDR 4, December 15, 1997.
[3] E. Au!ray, M. Lebeau, P. Lecoq, M. Schneegans, Speci"cations for lead tungstate crystals preproduction, CMS-Note
98/038, May 19, 1998.
[4] G. Drobychev, et al., Studies and proposals for an automatic crystal Control system (ACCOS), CMS-Note
1997/036.
[5] E. Au!ray, et al., Certifying procedures for lead tungstate
crystal parameters during mass production for the CMS
ECAL, Proceedings of the 1998 IEEE Nuclear Science
Symposium Conference Record. Toronto, Canada, November, 1998, Vol.1, pp. 508}13
[6] G. Basti, et al., A proposal for an automatic crystal control
system, CMS-IN 97/033, 1997.
E. Auwray et al. / Nuclear Instruments and Methods in Physics Research A 456 (2001) 325}341
[7] W.W. Moses, A method to increase optical timing spectra
measurement rates using a multi-hit TDC, Nucl. Instr. and
Meth. A 336 (1993) 253.
[8] J.-M. Le Go! et al., C. R. I. S. T. A. L./ concurrent repository and information system for tracking assembly and
341
production lifecycles } a data capture and production
management tool for the assembly and construction of the
CMS ECAL detector, CMS Note/1996 } 003.
[9] C. Laviron, P. Lecoq, private communication.