Skip to main content

Full text of "Testrun results from prototype fiber detectors for high rate particle tracking"

See other formats

Testrun results from prototype fiber detectors 
for high rate particle tracking 

E.G. Aschenauer, J. Bahr, V. Gapienko \ B. Hoffmann^, 
H. Liidecke, A. Menchikov^, G. Mertens^, R. Nahnhauer ' 

R. Shanidze^ 

DESY Zeuthen, 15738 Zeuthen, Germany 


A fiber detector concept has been realized allowing to registrate particles within 
less than 100 nsec with a space point precision of about 100 fim at low occupancy. 
Three full size prototypes have been build by different producers and tested at a 
3 GeV electron beam at DESY. After 3 m of light guides 8-10 photoelectrons were 
registrated by multichannel photomultipliers providing an efficiency of more than 
99 %. Using all available data a resolution of 86 fim was measured. 

1 Introduction 

The advantageous use of fiber detectors for particle tracking has been demon- 
strated for very different conditions e.g. in the UA2-Experiment [1], CHORUS 
[2], for DO [3] and the Hl-Forward Proton Spectrometer [4]. Due to the differ- 
ent experimental situations in these applications three types of optoelectronic 
read out techniques are applied - Image Intensifiers plus CCD's, Visible Light 
Photon Counters (VLPC) and Position Sensitive Photomultipliers (PSPM). 
However, all have in common that the precision of space point measurements 
is given by hits of overlapping fibers of several staggered fiber layers. For high 
rate experiments demanding online tracking of several hundred particles per 

* corresponding author, phone: -^49 33762 77346, Fax: -^49 33762 77330, 

^ on leave from IHEP Protvino, Russia 
^ now at Esser Networks GmbH, Berlin 
^ on leave from JINR Dubna 

^ Summerstudent from University of Clausthal Zellerfeld 

on leave from High Energy Physics Institute, Tbilisi State University 


Preprint submitted to Elsevier Preprint 

2 Eebruary 2008 

100 nsec bunch crossing such a concept may not work due to too high occu- 
pancy of single fiber channels. 

We propose in the following to use overlapping fiber roads composed of several 
thin scintillating fibers read out with one clear optical fiber. The demands and 
the solutions presented below match to a possible application of the detector 
as the inner tracker in the HERA-B project at DESY [5]. Similar ideas have 
been used by others [6] to build a fiber detector for the DIRAC experiment at 

2 Detector Principle 

The fiber detector under discussion is aimed to detect throughgoing particles 
with more than 90 % efficiency within less than 100 nsec and a precision of 
better than 100 ^m. The fibers should not change their characteristics signif- 
icantly after an irradiation of 1 - 2 Mrad. The sensitive detector part should 
have a size of about 25 x 25 cm^. The scintillating fibers should be coupled 
to clear optical fibers of about 3 m length guiding the light to photosensors 
placed outside the experimental area. 

It is assumed that most particles of interest hit the detector perpendicular, i.e. 
with angles less than five degrees with respect to the beam axis. In this case 
low occupancy and high light yield are guaranteed by using overlapping fiber 
roads like schematically drawn in fig. 1. One fiber road consists of several thin 

LG (1.7 mm) 


Fig. 1. Schematic view of the proposed fiber detector cross section and coupling 
principle (LG: light guide fiber) 


scintillating fibers arranged precisely behind each other and coupled to one 
thick light guide fiber. The scintillating fiber diameter determines the space 
point resolution of the detector. The number of fibers per road is fixed by the 
scattering angle of particles and the allowed amount of multiple scattering. 
It will also infiuence the factor of background suppression for tracks with 
larger inclination or curvature. The pitch between fiber roads is defined by 
demanding a homogeneous amount of fiber material across the detector width. 

Keeping in mind the conditions at HERA-B, the following choices are made: 

= 480/im 


= 7 

= 30 cm 


= 340 fim 

= 1.7 mm 

= 640 


= 300 cm 


= 217.6 mm 

with $ and L: diameter and length of scintillating and light guide fibers, 
^fib/road- uumbcr of fibers per road, Proad- distance between neighboured road 
centers, ^road- number of roads per detector, Wdet- detector width. 

The light guide fibers are read out with the new HamamatsuQ 64 channel 
PSPM R5900-M64 with a pixel size of 2 x 2 mm^ [7]. To diminish optical 
cross talk the thickness of the entrance window of the device was decreased 
to 0.8 mm. 

The coupling between scintillating and light guide fibers is done by loose 
plastic connectors. The light guides are coupled to the PSPM using a plastic 
mask fitting the corresponding pixel pattern. 

3 Material Studies 

Double clad fibers of three different producers^ Q were tested concerning 
light output, light attenuation and radiation hardness for several fiber diam- 
eters and wavelengths of the emitted light. Details of these measurements are 
given in [8]. A few results are summarized below. 

The light output of fibers of 500 fim diameter is shown in fig. 2. Generally it 

^ Hamamatsu Photonics K.K., Electron tube division, 314-5, Shimokanzo, Toky- 

ooka Village. Iwatagun, Shizuoka-ken, Japan 

^ BICRON, 12345 Kinsman Road, Newbury, Ohio, USA 

^ pol. hi. tech., s.r.l., S.P. Turanense, 67061 Carsoli(AQ), Italy 

9 KURARAY Co. LTD., Nikonbashi, Chuo-ku, Tokyo 103, Japan 






£ 2 

i B 











] mirrored 



; Kur 




• 12' 





: Fo r. 




! * 



Fig. 2. Light output from 500 /xm diameter fibers for several fiber materials of three 

can be seen, that the light yield decreases with increasing scintillator emission 
wavelength because the PM sensitivity curve is not unfolded. There is no 
remarkable difference between the best materials of the three producers. A 
mirror at the end of the fiber increases the light output by a factor 1.7. 

Several tests were performed to couple scintillating and light guide fibers. 
Finally the coupling efficiency became better than 95 %, independent of the 
medium between both parts (air, glue, grease). 

The light attenuation of clear fibers was measured coupling them to single 
scintillating fiber roads excited by a Ruthenium source. The clear fibers were 
cutted back to the length under investigation. Results for two producers are 
given in fig. 3. 

Radiation hardness tests of fibers were made using an intense 70 MeV proton 
beam at the Hahn-Meitner-Institute Berlin. 1 Mrad radiation was deposited 
within a few minutes. For all materials investigated we observed a damage of 
the scintillator and the transparency of the fiber which was followed by a long 
time recovery of up to 600 h. An example is shown in fig. 4. More detailed 
studies using glued and nonglued fibers and irradiate them in air and nitrogen 
atmosphere are still ongoing. 

Summarizing all results of our material studies we decided to use the KU- 



^ f-, 

;50 iti 

' .. ( 



( ■ ■ . 

- -1 





■■ ■) 



- • 


..m " 

• (I 


H) kur< 
3) Pol,l 


irdy T. 

. , , , 


, , . . 







50 100 150 200 250 300 350 400 

Length of light guide (cm) 

Fig. 3. Light attenuation in clear fibers of 1.7 mm diameter produced by Kuraray 

A - 

SCSF 78 

■ (20cm) 
□ (25cm) 



1 1 1 1 

1 1 1 1 

1 1 1 1 

1 1 1 1 


Fig. 4. Time evolution of light output for KURARAY SCSF 78M fibers irradiated 
with 0.2 and 1.0 Mr ad at 10 and 20 cm respectively. The solid and dotted curves 
correspond to measurements with a source placed at 20 and 25 cm. 

RARAY fibers SCSF-78M with a diameter of 480 fim for the scintillating part 
of our detector prototypes. For clear fibers still two choices seem to be possible: 
1.7 mm fibers from KURARAY or pol. hi. tech.. 


Fig. 5. Photograph of part of the pohshed endface of the Zeuthen prototype detector 

Similar investigations of the same materials have been done by our colleagues 
from Heidelberg University [9]. For their irradiation tests they used a Co^° 

4 Detector Production 

Using winding technology as developed for the CHORUS experiment [10] we 
built a detector production chain at our institute. A drum of 80 cm diameter 
allows to produce five detectors at once. The production time for winding one 
drum is about 14 h. Sticking the fibers to the connector holes was still done 
by hand and rather time consuming. A part of the polished endface of one of 
our detectors is shown in fig. 5. 

Two other detector prototypes are ordered from industry. GMS-Berlirf^ fol- 
lowed a technology proposed by the university of Heidelberg [9] mounting 
single layers on top of each other using epoxy glue. Each layer is prepared on 
a v-grooved vacuum table. One layer per day can be produced in this case. 
The connector is here also added by hand. The production technology used 
by KURARAY is unknown to us. 

To get the precision of the detector geometry quantified we measured the 
coordinates of all fibers of the polished endface of the three detectors. In fig. 
6 the deviation from the ideal position is given per fiber road. Some stronger 
local effects are visible. Averaging these results characteristic accuracies of 20 
/im, 50 fim and 10 /zm are calculated for the Zeuthen, GMS and KURARAY 
detectors respectively. 

" GMS - Gesellschaft fiir Mess- und Systemtechnik mbH, Rudower Chaussee 5, 
12489 Berhn, Germany 


Q L I I I I I I I I I I I I I I I I I I I I I I I I I I I I L 

1DD 2DD 3DD 4DD 5DD 

Q l_I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I L 

5D 1DD 15D 2DD 250 3DD 35D 4DD 450 

Road number 

Fig. 6. Deviation of fibers from ideal position per fiber road for the prototype 
detectors from Zeuthen, GMS and KURARAY 

5 Opto-Electronic Readout 

The 640 channel scintillating fiber detectors are coupled via plastic connectors 
to ten 3 m long light guide cables of 64 clear fibers of 1.7 mm diameter. Each 
cable is coupled to a Hamamatsu 64 channel position sensitive photomultiplier 
R5900 - M64. The coupling is done using a plastic mask, fitting the anode 
pattern of the PSPM with 2 • 2 mm^ pixel size. A detailed description of the 
photomultiplier is given by the producer in [11]. 

The pulse rise time of the PSPM under study is a few nanoseconds. Therefore 
the time behaviour of the fiber detector will be mainly restricted by the readout 
electronics. For a high rate experiment the signals have to be digitized and 
pipelined in an early stage. We will not treat this problem here instead we try 
to measure the total light amplitude as seen by the PSPM. For this purpose 
we used the readout boards originally developed for the L3 fiber detector 
[12]. Like in previous beam tests [13], [14] the serial output of these boards is 
digitized by a SIROCCO-11 fiash-ADC using a VME 0S9 online data taking 
system with CAMAC interface. 


We got the first prototypes of tlie PSPM just before the testrun started. Finally 
we used only one exemplar for the beam measurements. The corresponding 
light guide cable could be moved to all ten connector places across the de- 
tectors. Mechanical precision was guaranteed using precise pins and holes for 
each position. Using 20 nsec long hght pulses of different intensity produced 
by a LED coupled to a fiber we made an extensive check of the behaviour of 
the PSPM in combination with our readout electronics. In the meantime five 
other exemplars of the photomultiplier have been studied with similar results. 

In fig. 7 we show the amplitude distribution for one pixel illuminated with 
weak light signals. Beside a narrow pedestal a clear one photoelectron peak 
is seen. The gain of the PSPM comes out to be independent of the number 
of incoming photons up to more than 10 photoelectrons and linear in the 
high voltage range between 800 and 980 V. This allows to calibrate the light 
output measured in FADC-channels to the corresponding average number of 
photoelectrons. What has to be taken into account in addition in this case 
is the different sensitivity of the PSPM pixels. We measured it, putting the 
same amount of light to each channel and found a variation by about a factor 
of two. 

The applicability of a PSPM for fiber detector readout is limited by its cross 
talk properties, i.e. the response of other pixels than the illuminated one to the 
incoming hght signal. In case of the R5900-M64, Hamamatsu has decreased 
the cross talk already considerably by decreasing the thickness of the entrance 
window from 1.3 mm to 0.8 mm. Nevertheless the cross talk depends strongly 
on the threshold cut applied and is still about 10 % for a one photoelectron 
cut and typical light signals as can be seen from fig. 8. From the figure one 
observes also that most of the cross talk comes from the nearest neighbours 
of the directly illuminated pixel. Using an appropriate hit selection procedure 
this cross talk can be strongly suppressed. If for instance a local maximum 
selection is used it vanishes completely [13], [14]. 

Whereas the amount of cross talk depends only weakly on the PSPM high 
voltage, it is strongly correlated to the amount of incoming light to a pixel 
(see fig. 9). The diameter of the light guide fibers connected to the pixels starts 
to influence the cross talk behaviour above 1.5 mm. 

Our fiber detector data taking in the testbeam has been synchronized with re- 
spect to data taking with a four plane Microstrip Detector Hodoscope (MSD). 
To increase the final data rate for the two systems with free running clocks 
a long trigger gate of 2 fisec was chosen. As found later, that leads already 
to discharge effects for early PSPM signals. In fig. 10 the flash ADC response 
to a constant light signal arriving at different trigger times is shown. For our 
testrun data the arrival time of a trigger is unknown but uniform. Correspond- 
ingly we had to unfold all flash ADC spectra using the behaviour measured 
in flg. 10 to determine the correct average number of photoelectrons. 


Characteristics of PSPM R5900-M64 

Cross — talk vs cut ( signa 

- 8 PE) 


bC 1G€ I zbO JGC bbO 40G 

FADC counts 

Fig. 7. Flash ADC spectrum for one 
pixel of the 64 channel photomulti- 
plier R5900-M64 illuminated with weak 
light signals from a LED 

, I. Ill • • III fii'l 1; I II 

Mean PE 

Fig. 9. Detection probability for sig- 
nal (circles) , near (squares) and far (tri- 
angles) cross talk for incoming light of 
different intensity given in number of 
photoelectrons. An one photoelectron 
cut has been applied to the data. 

si - signa 

H- Crosss talk(near) 
A- Crosss lolk(far) 

0,2b O.b 0.7b 1 '.zb I .b 1.7b 2 2.2b 2.b 

ADC cut (In P.E.) 

Fig. 8. Detection probability for light 
signals corresponding to about 8 pho- 
toelectrons in dependence of a thresh- 
old cut given in numbers of photoelec- 
trons for the illuminated pixel (circles), 
cross talk from direct neighbouring pix- 
els (squares) and cross talk from dis- 
tant ones (triangles) 

=: ■ 

■ ■ ■ 














0.25 0.5 0.75 

1.25 1 

5 1 

75 2 

Delay (/^s) 

Fig. 10. Average number of FADC - 
channels in dependence of the arrival 
time of the corresponding trigger in a 
2 usee gate for constant light signals 
of two different intensities (squares and 


6 Testrun Setup 

In a two weeks testrun in October 1997 the three available detector prototypes 
have been studied in the 3 GeV electron beamline T21 at DESY. More than 
5*10^ triggers have been recorded under several conditions in various runs. 
The setup is schematically drawn in fig. 11. 

Fig. 11. Schematic view of the testrun setup. M1-M4: silicon microstrip detectors, 
RDl and RD2: fiber reference detectors, T1-T4: trigger paddles, FD: fiber detector 
to be tested 

A trigger system of four scintillator paddles Tl - T4 selects by 4-fold coinci- 
dence a beam region of 2*2 cm^. The corresponding trigger rate was 50 kHz, 
leading to a final data acquisition rate of about 5 Hz. 

Two fiber reference detectors RDl and RD2 fixed to the same support as the 
MSD hodoscope paddles are used mainly to perform an independent alignment 
of these components. Each fiber reference detector is made out of two double 
layers of 32 fibers of 1 mm diameter. The total intrinsic resolution of the 
system is 72 /xm. 

The MSD hodoscope consists of four paddles with 320 or 640 strips of 100 /im 
pitch. The intrinsic resolution for one hit clusters is about 30 /im. 

The distribution of differences of coordinates measured by the fiber reference 
detector and the corresponding MSD paddle has been used to ahgn the mea- 
suring system. Alignment errors of 58 /xm, 70 fim, 13 fim, and 10 fim have 
been found for the paddles 1 to 4. The result refiects problems of the geomet- 
rical arrangement with respect to the beam divergence. The MSD prediction 
precision at the fiber detector position has been calculated from fits of four 


hit tracks to be 75 //m. 

The fiber detector FD is movable in x and y direction perpendicular to the 
beam. About 30 different positions have been used for measurements. The 
alignment of the fiber detector with respect to the MSD-coordinate system 
has been done separately for every position with a precision of about 10 /im. 

7 Results from Testrun Data 

AH testrun data have been sensitivity corrected corresponding to the correc- 
tion table measured before for the PSPM operating with a high voltage of 
980 V. A noise threshold of 35 FADC channels has been applied afterwards 
leaving 1 °/oo efficiency for a noise hit per beam trigger. The threshold to se- 
lect events with more than one photoelectron seen is chosen to be 150 FADC 
channels at this high voltage with the one photoelectron peak observed at 100 
FADC channels. 

For the selection of hitted fiber roads in principle a simple cut excluding single 
photoelectron hits would be sufficient to reduce background from cross talk 
and other sources. Results can be improved, if a local maximum search is 
possible removing all nearby cross talk hits [13], [14]. In the following we will 
use both selection criteria in common. 

After selecting hits, a cluster algorithm combines neighbouring ones to a track 
cluster. As can be seen from fig. 12 from geometry one or two fiber roads could 

A B2 

Fig. 12. Illustration of the different zones A and B for overlapping fiber roads 
leading to one or two hit events 


Fig. 13. Unfolded light output spectra as measured by the FADC for the detectors 
from a.) Zeuthen, b.) GMS and c.) KURARAY 

be hitted by a particle crossing perpendicular to the detector plane . 

In figures 13 a-c the unfolded FADC spectra are shown for the detectors pro- 
duced by Zeuthen, GMS and KURARAY respectively. Using the described 
calibration leads to an average number of photoelectrons between eight and 
ten and correspondingly efficiencies above 99 % if the one photoelectron cut 
is applied (see table 1). The larger values for the GMS detector seem to be 
due to the very good quality of the polished and mirrored detector endface. 

As can be seen from fig. 14 the efficiency varies only weakly for different 
Table 1 

Average number of photoelectrons < Ape >, efficiency e and resolution a for the 
fiber detectors produced by Zeuthen, GMS and KURARAY 


< Ape > 
















Efficiency vs FDET vs pes (unfolded dctc) 

Efficiency 1 

or I F 
















, , , , 

1 23456789 10 

Position in FDET 

Eft vs. Position ( 1 PE cut for unfolded ddto) 

Fig. 14. Efficiency at different positions of the fiber detector prototypes from 
Zeuthen (squares), GMS (triangles) and KURARAY (circles) averaged for 64 fiber 
roads. Particles hit 5 cm from the near (open symbols) or far (filled symbols) end of 
the ordered detector part. An one photoelectron cut has been applied to all data. 

positions of the light guide cable, i.e. different fiber roads across the detector. 
The variation of the efficiency along the fibers is also marginal. However the 
distance between near and far measurements is only about 15 cm. 

Expecting in principle only one incoming particle from beam triggers one 
observes in the data after the one photoelectron cut still 10 % of events with 
more than one track cluster in the fiber detector. A rough GEANT Monte 
Carlo simulation including only the detector material comes up with about 
4 % additional hits produced by matter effects. These hits increase the number 
of two hit track clusters and lead to additional "matter" tracks. 2 % of such 
tracks are observed. Fig. 15 shows the distance between tracks in multi-track 
events for data and Monte Carlo. Comparing both one can not exclude cross 
talk at the percent level. Possible sources are accidentals (noise), the PSPM 
glass window, fibers in the detector, connectors or hght guide cables. 

To calculate the intrinsic resolution of our fiber detector geometry one has to 
keep in mind again that the fiber roads have two regions A and B where one 
or two of them should give light signals for throughgoing particles (see fig. 12). 


2.5 5 7.5 10 12.5 15 17.5 20 22.5 

Distonce between tracks (mm) 

2.5 5 7.5 10 12.5 15 17.5 20 22.5 

Distance between tracks (mm) 

Fig. 15. Distribution of distances between tracks in multitrack events after an one 
photoelectron cut on the light output has been made for a.) data and b.) Monte 
Carlo simulation. 

Table 2 

Width, fraction of appearance, intrinsic geometrical resolution and resolution mea- 
sured in beam test for the two regions of a fiber road where one or two hits are 
produced by a crossing particle. 




apD /iJ-T^ 











A + B 





With a fiber diameter of 480 //m, a cladding thickness of 30 //m and a pitch of 
340 fim values of 260 /im and 80 /xm are derived for the widths of the regions 
A and B. Keeping in mind that the one photoelectron cut excludes low light 
signals one calculates the values given in table 2 for the width of the regions, 
the fraction of crossing particles and the corresponding resolution. Combining 
these numbers an overall intrinsic resolution of 69 /im could be expected. 


Fig. 16. Distribution of the difference of coordinates measured by the fiber detectors 
and the MSD hodoscope after an one photoelectron cut has been apphed. The curve 
is the result of a fit with two gaussian functions. 

The difference between the coordinates of a crossing particle measured with 
the fiber detector and with the MSD hodoscope is distributed in fig. 16 for 
all data of all runs. A fit with two gaussian functions yields a resolution of 
(7 = 114//m for the main component. Unfolding the MSD precision of 75 
yum a fiber detector resolution of 86 /im is calculated, near to the theoretical 

Separating events with one and two fiber roads hit allows to calculate the 
resolution for tracks crossing regions A and B. The values given in the last 
column of table 2 are also close to the theoretical estimates. Only the double 
road hit distribution needs two gaussian functions to be described reasonably. 
The second component with a large width is mainly due to matter effects like 
delta-electrons which lead to additional hits in neighbouring fibers. 

The resolution has also been measured for different positions in different de- 
tectors (see fig. 17 and table 1). No remarkable variations were observed. The 
statistics and precision of measurements are not sufficient to establish local 
effects due to problems in mechanical detector production. 


Fiber Detector Resolution 

Zeuthen □ 

0,1 1 



23456789 10 

Position in FDET 

Fig. 17. Resolution at different positions of the fiber detector prototypes from 
Zeuthen (squares), GMS (triangles) and KURARAY (circles) averaged for 64 fiber 
roads. Particles hit 5cm from the near (open symbols) or far (filled symbols) end of 
the ordered detector part. An one photoelectron cut has been applied to all data. 

A few additional phenomena which may influence the tracking precision in an 
experiment have also been considered during the testrun. All prototypes have 
"unordered" fiber regions of about 7cm near to the 640 channel connector. 
Throughgoing particles are registrated in this region with a precision of about 
1 mm. 

The 640 clear fibers of 1.7 mm diameter and 3 m length give a good possibil- 
ity to produce Cerenkov light if crossed by charged particles. That depends 

strongly on the light guide arrangement: the largest effect can be expected 
near to the connector. Exposing this region to the electron beam, less than 
1 % of crossing particles produced hits after applying the one photoelectron 

Particle showers from nearby material increase background and occupancy 
and decrease the detector resolution. A small effect of this kind may be seen 
already in fig. 17 comparing the values measured for the detector boundaries 
with those in the central region. 


8 Summciry 

Three fiber detector prototypes have been tested. They are made out of 640 
overlapping roads of seven 480 fim fibers coupled to 1.7 mm diameter light 
guides of 3 m length readout with 64 channel photomultipliers. 

For all three detectors an efficiency of about 99 % and a resolution better 
than 100 /im have been measured in an exposure to a 3 GeV electron beam 
at DESY. 

These results together with radiation hardness studies of the used fiber mate- 
rials seem to make it possible to use a corresponding detector in a high rate 
experiment like HERA-B. 

The readout of such a detector - connectors, light guides, mask arrangement, 
front end electronics - has to be adapted with care to the specific demands of 
the application. 


Part of this work was done in close collaboration with groups from the uni- 
versities of Heidelberg and Siegen. We want to thank our colleagues for their 
good cooperation and many fruitful discussions. 

In particular we thank Prof. F.Eisele for providing the KURARAY detector, 
bought by Heidelberg University, for the beam tests. 

The fiber irradiation tests were possible only due to the kind support of the 
Hahn-Meitner-Institute Berlin. We are deeply indebted to the ISL accelerator 
team and want to thank in particular Dr. D. Fink, Dr. K. Maier and Dr. M. 
Miiller from HMI and Prof. H.A. Klose from GMS for a lot of practical help. 

We acknowledge the benefit from the DESY II accelerator crew and the test 
area maintainance group. 


[1] Ansorge, R., et al., 7V/M265, 33 (1988) 
[2] Annies, P., et al., 7V/MA367, 367 (1995) 


[3] Bross, A.D., Nucl. Phys. B (Proc.Suppl.) 44, 12 (1995) 
Adams,D.,et al, Nucl. Phys. B (Proc.Suppl.) 44, 332 (1995) 

[4] Bahr, J., et a.. Proceedings of the 28th Intern. Conf. on High Energy Physics, 
Warsaw, Poland, 1996, eds. Z.Ajduk,A.K.Wroblewski\. II, p. 1759 

[5] Lohse, T., et al., HERA-B Technical Proposal, DESY-PRC 94/02 (1994) 

[6] Ferro-Luzzi, M., et al., contribution presented by A.Gorin, to appear in 
Proceedings of the workshop SCIFI97, Notre Dame, USA, 1997 

[7] Yoshizawa, Y., et al., to appear in Proceedings of the workshop SCIFI97, Notre 
Dame, USA, 1997 

[8] Aschenauer, E.G., et al., preprint DESY 97-174 (1997) 

[9] Drcis, B., ct al., preprint DESY 98-049 (1998) 

[10] Nakano, T., et al., Proceedings of the workshop SCIFI93, Notre Dame, USA, 
1993, eds. A. Bross, R. Ruchti, M. Wayne, p. 525 

[11] Yoshizawa, Y., et al., Hamamatsu technical information. No. TPMH 9002E01 

[12] Akbari, H., et al., Ar/MA302, 415 (1991) 

[13] Bahr, J., et al.. Proceedings of the workshop SCIFI93, Notre Dame, USA, 1993, 
eds. A. Bross, R. Ruchti, M. Wayne, p. 578 

[14] Bahr, J., et al., AT/M A371, 380 (1996)