Properties and performance of two wide field of view Cherenkov/fluorescence telescope array prototypes

Abstract

A wide field of view Cherenkov/fluorescence telescope array is one of the main components of the Large High Altitude Air Shower Observatory project. To serve as Cherenkov and fluorescence detectors, a flexible and mobile design is adopted for easy reconfiguring of the telescope array. Two prototype telescopes have been constructed and successfully run at the site of the ARGO-YBJ experiment in Tibet. The features and performance of the telescopes are presented.

Full text

Page 1

arXiv:1112.1463v1 [physics.ins-det] 7 Dec 2011
Properties and Performance of Two Wide Field of View Cherenkov/Fluorescence
Telescope Array Prototypes
S. S. Zhanga,∗, Y. X. Baia, Z. Caoa, S. Z. Chena, M. J. Chena, Y. Chena, L. H. Chenb, K. Q. Dinga, H. H. Hea, J. L.
Liua, X. X. Lia, J. Liua, L. L. Maa, X. H. Maa, X. D. Shenga, B. Zhoua, Y. Zhanga, J. Zhaoa, M. Zhaa, G. Xiaoa
aInstitute of High Energy Physics, CAS, Beijing 100049.
bHeibei Normal University, China, Heibei 050016.
Abstract
A wide field of view Cherenkov/fluorescence telescope array is one of the main components of the Large High
Altitude Air Shower Observatory project. To serve as Cherenkov and fluorescence detectors, a flexible and mobile
design is adopted for easy reconfiguring of the telescope array. Two prototype telescopes have been constructed and
successfully run at the site of the ARGO-YBJ experiment in Tibet. The features and performance of the telescopes
are presented.
Keywords: WFCTA, Cherenkov telescope, fluorescence telescope, Cosmic ray detector.
1. Introduction
1
The energy spectrum of primary cosmic rays spans
almost 12 orders of magnitude, from 109eV to 1021eV,
and can be well fitted by a simple power law except
in several small energy regions. A region called the
“knee”ofthespectrumexistingat around1015eV is one
of these regions where the spectrum becomes steeper
at higher energy side.Many experiments have ob-
served this phenomenon; however, controversial argu-
ments on its origin persist because of limited discrim-
ination power on the primary cosmic ray composition
and ambiguities in nucleus-nucleus interaction model-
ing. These two aspects are closely related to each other.
Modern balloon borne experiments, such as ATIC [1]
and CREAM [2], have efficiently measured the energy
spectra of individual elements at the top of the atmo-
sphere. The energy spectra for all nuclei are measured
up to ∼100 TeV which is not far from the “knee”.
Because the detector area is constrained by the pay-
load, the spectrum measurement has to be extended to
a higher energy using a ground based air shower detec-
tor array. The spectrum should initially be measured
well below 100 TeV to create an overlap with the bal-
loon experiments which serve as absolute calibrations
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
∗Corresponding author. Institute of High Energy Physics, CAS,
Beijing 100049. Tel.: +86-010-88236035
Email address: zhangss@ihep.ac.cn (S. S. Zhang)
for the ground-based techniques. Identifying the in-
dividual components of cosmic rays continues to be a
major challenge in ground-based experiments. Mul-
tiple parameter measurements on an air shower seem
to be a plausible approach. The ultimate goal is to
separate individual species out of the total observed-
event samples and measure a clear individual “knee”
for every single species, enabling the discovery of the
origin of the “knee”. As one of the major scientific
goals of the Large High Altitude Air Shower Obser-
vatory (LHAASO) project [3, 4], the energy spectrum
for a separated composition will be measured at ener-
gies above dozens of TeV. To tag each primary par-
ticle that causes an air shower, the atmospheric depth
of the shower maximum should be measured as one
of the important parameters. The wide field of view
Cherenkov/fluorescencetelescope array (WFCTA), one
of the components of the LHAASO project, is designed
to accomplish this goal.
A portable design of WFCTA telescopes is adopted
to maximize the flexibility of changing the configura-
tion of the array of telescopes. The elevations, point-
ing directions, and locations of the telescopes are then
easily reconfigured. This is one way of using the same
telescopes to serve as both fluorescence and Cherenkov
detectors. In the fluorescence detector, the telescopes
are tilted down to a horizontal position. In such an
operational mode, which is analogous to the HiRes
experiment [13], most of the Cherenkov photons are
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
Preprint submitted to Nuclear Instruments and Methods in Physics Research Section ADecember 8, 2011

Page 2

avoided except those that are scattered onto the field
of views (FOVs) of the telescope, such as in the fluo-
rescence detector of the telescope array experiment[10]
andthefluorescencedetectorofthePierreAugerexperi-
ment[8, 9]. Only thefluorescencelight fromthe shower
is collected together with the scattered Cherenkov light
to trigger the telescopes. This requires showers hav-
ing much higher energy, usually above 100 PeV, such
as in the HiRes prototype experiment [15], because
the fluorescence light by a single electron is consid-
erably weaker and isotropic. In the Cherenkov detec-
tor, the telescopes run in high elevation mode to di-
rectly measure Cherenkov light from the showers, sim-
ilar to what was done in the Dice experiment [16]. A
Cherenkov light radiation provides considerably more
photons along the shower axis that are useful for lower-
ing the shower energy.
In 2007, two prototype Cherenkov telescopes [5, 6]
were deployed at Yangbaijing (YBJ) Cosmic Ray Ob-
servatory near the ARGO-YBJ experiment [7]. More-
over, two WFCTA telescopes have been successfully
runningin CherenkovmodebeginningAugust 2008. To
date, millions of cosmic ray events that simultaneously
trigger the telescopes and the ARGO-YBJ detector car-
pet array have been collected. An analysis of these
eventsis carriedouttostudytheperformanceofthetele-
scopes. Detailed descriptions of the telescopes and the
analysis of the findings are presented in this paper.
Several details about the apparatus are presented in
Section 2. The detector calibration is then discussed in
Section 3. The test-run of the two telescopes and re-
sults are reported in Section 4 including summaries on
the detector performance. The conclusions drawn are
provided in the last section.
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
2. Apparatus
88
The two prototype telescopes are deployed near the
ARGO-YBJ carpet detector array at a longitude of
90.53◦E, and a latitude of 30.11◦N and 4300 m a.s.l.
One telescope is about 25 m away from the west side
of the ARGO-YBJ array. The other is also 25 m away
fromthe south side of the array with separation distance
between the two telescopes is 50 m. Each telescope has
an FOV of 14◦in elevation by 16◦in azimuth. The fo-
cal plane camera is made of a 16×16 photomultiplier
tube (PMT) array, and the pixel size is approximately
1◦. Because bothtelescopes are tilted up to 60◦pointing
in the same direction, they can be operated in stereo-
scopic mode, i.e., showers striking an area covered by
the telescopes will be seen simultaneously. Since the
Cherenkovlight from a shower is veryconcentratedin a
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
forward region; thus, the telescopes can be triggered by
showers coming within a cone of approximately8◦with
respect to the main optic axes of the telescopes.
The entire telescope system is composed of an op-
tic ultraviolet light collector, a focal plane camera, front
end electronics (FEE) based on 50-MHz flash analog-
to-digital-converters (FADC), data acquisition (DAQ)
based on an embedded ARM processor and PC104 bus,
power supplies for low and high voltages, and a slow
control system. Everything is installed in a shipping
containerwith dimensionsof 2.5m×2.3m×3 m (Fig.1).
Mirrors are mounted at one end of the container and
the camera is located at the other end where the focal
planes of the mirrors are. The FEE and DAQ are placed
at the back plane of the PMT camera. A glass window
is installed at the entrance aperture to keep dust from
entering the apparatus. The container is mounted on a
dump-truck frame with a hydraulic lift that allows the
container to be lifted up from 0◦to 60◦. The mobility
of the entire telescope allows for freely switch between
configurationsof the telescope array for different obser-
vational modes. The architecture of the electronic data
acquisition and the slow control system are shown in
Fig.2, whereas that of a sub-cluster is shown in Fig.3.
ThePMTsignalsareprocessedusingananalogprocess-
ing board (AB) and then a digitization board (DB). The
first level trigger (FLT) is generated in the DB on a sub-
cluster. After the FLT is determined, 256 FLTs are then
sent to the trigger board (TB). The second level trigger
(SLT) and the third level trigger (TLT) are then deter-
mined in the TB. After this, the event trigger from the
TB is fanned out by a bus driver board (BDB) and sent
back to each of the DB and GPS boards. The data are
initially stored in the buffer of the DB when the DB has
received the event trigger, after which the data that in-
clude the GPS time are read using TS7200. Finally, the
data are stored in a PC in the laboratory via Ethernet.
A detailed description of the detector, divided into the
following 8 subsystems, is provided: 1) optics, 2) cam-
era, 3) FEE, 4) trigger system, 5) DAQ, 6) power supply
system, 7) slow control system that includes monitoring
of everything, and 8) calibration.
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
2.1. Optics
A 4.7 m2spherical aluminized mirror, composed of
20 hexagon-shaped segments, is used as an ultraviolet
lightcollector. Eachsegmentis subjectedtoa strictcon-
trol of the surface quality and their geometrical and op-
tical properties. The reflectivity is greater than 82% for
light having a wavelength ≥300 nm. The radius of the
curvature of the segments is 4740 mm with a tolerance
of ±20 mm.
146
147
148
149
150
151
152
153
154
2

Page 3

Figure 1: Photograph of the telescope with the doors open.
Figure 2: Communications diagram of one telescope; for details of
the sub-cluster see Fig.3.
The size of a light spot on the focal plane where the
PMT camera is located is designed to be similar to the
size of a pixel. The sensitivity of a PMT is non-uniform
across the surface of the photocathode [17], and small
gaps exist between PMTs in the camera. For a spot
that is exceedinglylargerthan the dimensionof the gaps
andthetypicalspatialscaleofthenon-uniformityacross
the photocathode, light from a specific direction will be
shared by a few adjacent PMTs. Thus, effects stemming
from the overall non-uniformity of response across the
entire camera are reduced. The direction of incident
155
156
157
158
159
160
161
162
163
164
165
photons can be efficiently determined by simply taking
the average of the directions of the registered pixels as
weighted by the measured charge in a pixel. Contrast-
ingly, the size of a spot is optimized to be similar to
pixel size to avoid sharing of incident photons by an ex-
cessive number of pixels.
Because of the abbreviation of the spherical reflector,
the spot size changes across the focal plane in a rather
large FOV. Large comas also occur at large off-axis an-
gles. The uniformity of the spot size over the camera
is also optimized by locating the camera slightly away
from the focus. A distance of 2305 mm between the
mirror and the camera is eventually set according to a
detailed ray-tracing calculation.
Each mirror segment is mounted on a spherical steel
frame with three adjustable screws. The pointing ori-
entations of all segments are adjusted toward the geo-
metric center of the curvature. Using a laser beam, the
orientationsarecalibratedtobelessthan7.6arcseconds
from the nominal direction.
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
2.2. PMT Camera
186
Photons in a spot at the focal plane are recorded by a
camera composed of 256 pixels which are 40 mm Pho-
tonis hexagonal PMTs (XP3062/FL). The camera is en-
closed in a box and mounted on a frame at a distance of
2305 mm from the spherical mirror.
The PMTs in a telescope are operated at a gain of
6×105. Toachievethebestuniformityathardwarelevel,
a resistor is placed between the high voltage power sup-
ply and each PMT base to compensate for the gain dif-
ference between PMTs. The gains (G) for all PMTs and
their responses to the supplied voltage, i.e., G ∝ V−β,
are calibrated. All β values are measured and recorded
in a database in the laboratoryfor futureuse. The PMTs
are thensortedaccordingto theirgainsandgroupedinto
two classes. For instance, the working voltages of the
PMTs of one of the telescopes distribute 1088 V to the
telescope in a common power supply with a variance
of 73.97 V. The average β of these tubes is 5.9 with a
variance of 0.6.
The 16 PMTs, as an integrated unit, are soldered on
a high voltage board (HVB) that distributes a negative
high voltage to all cathodes and dynodes of the PMTs.
The voltage division scheme is recommended by Pho-
tonis to yield the maximum gain of the tubes. The max-
imum output current of the PMTs allows a range of 3.5
orders of magnitude in which the non-linearity of all
tubes is less than 8% according to a calibration in the
laboratory [17]. The anode signals are finally DC cou-
pled to the FEE.
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
3

Page 4

Figure 3: Photograph of a sub-cluster (left) and schematic of the sub-cluster (right).
2.3. FEE and Digitization
The FEE is located behind the HVB, which is com-
posed of an AB and two DBs, to avoid a long distance
transmission of the analog signals. Such a module is re-
ferred to as a sub-cluster (see the left figure of Fig.3) in
this paper. A block diagram on the procession of sig-
nals from the PMTs to the on-board data storage chip is
shown in the right figure of Fig.3. The signal of each
PMT is transmitted to the first amplifier on the AB for
noise filtering and pulse stretching. Upon division into
high or low gain channels, the signal is then sent to the
DB for digitization and further processing. The rest of
this sectiondescribesthe processingprocedureindetail.
216
217
218
219
220
221
222
223
224
225
226
227
228
2.3.1. Analog Processing Board
Each AB has 16 channels for the signals from 16
PMTs on HVB alone. Each channel has a shaping cir-
cuit and amplifiers for low or high gains. The board has
four main functions, namely:
229
230
231
232
233
• low-pass filter of 20 MHz,
• expanding narrow pulses,
• dual-gain system for covering a dynamic range
over 3.5 orders of magnitude,
234
235
236
237
• and receiving PMT signals from HVB and per-
forming a single-ended-to-differentialconversion.
238
239
The anode signals are fed to a four-pole low-pass
filter based on an AD8039 through DC coupling.
Cherenkov photons generated by all shower particles
move at almost the same speed as the charged particles;
thus, all photons generated over the entire shower de-
velopmentbeginningfrom the top of the atmosphere hit
the cathodes at almost the same time. The duration of
240
241
242
243
244
245
246
Input pulse duration (ns)
0 204060 80 100120140 160180 200 220
Stretching rate
0
0.5
1
1.5
2
2.5
3
Figure 4: Stretching ratio as a function of the input pulse duration.
the pulse lasts only a few nanoseconds, which is shorter
than the typical response time of the PMT, e.g., a pulse
duration generated by a single-photo-electron (SPE) is
approximately 12 ns on average. Therefore, such a nar-
row pulse has to be stretched (e.g., at > 50 ns) to be
able to measure the charge at a sampling rate of 50
MHz, as preselected for the FADCs on DBs. Taking
into account the optimized bandwidthof the noise filter,
the stretching ratio is selected in such a manner that the
narrower pulses are stretched further, and pulses wider
than 120 ns are essentially not stretched. In Fig.4, the
stretching ratio is plotted as a function of the pulse du-
ration. According to this, the original waveform is re-
constructible, and the timing of the pulse can be cor-
rected with an uncertainty less than the duration of the
stretched pulses.
The amplification of the PMT signals with a gain of
2.67 is another feature of the shaping circuit. The value
of the gain is also optimized together with the filter and
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
4

Page 5

Figure 5: A Cherenkov signal with sky background; the first 100 bins
are low gain background and the remaining 800 bins are high gain
background and the Cherenkov signal. The Cherenkov signal is in the
current frame (from the 300th bin to the 600th bin).
the stretcher. The AD8039 has a sufficiently high speed
(350 MHz), low power dissipation, low cost, low noise,
lowdistortion,andanonlinearityless than1%inarange
of the input from 0.5 mV to 800 mV, which nearly fits
the entire range of the PMT signals.
Tomaintaingoodlinearityoverawidedynamicrange
of 3.5 orders of magnitudein charge,a dual gain system
is designed. Signals comingout of the AD8039are split
into two channels and are then separately amplified by
two AD8138. With such a design, considerable flex-
ibility in obtaining different resolutions of the charge
measurements for pulses with different pulse heights is
achievable by choosing the range covered by the high
gain channel. In this paper, a ratio of the gains of 1:8 is
selected as a result of the optimization between the dy-
namic range and the resolution in the charge measure-
ment. The nonlinearity of AD8138 is less than 2% at
both gains of 1 and 8 in the entire range of the input
signals.
Another excellent feature of the AD8138, the con-
version between a single-ended input and differential
output for instance, makes it the best choice for an
amplifier. Superposing an offset to the output of the
AD8138 as a pedestal of the signal before feeding into
the positively-polarized FADC is also a convenient ap-
proach. To manage possible undershoot of the pulses
(see Fig.5), a pedestal is preset higher than the under-
shoot.
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
2.3.2. Digitization Board
Each DB has 8 pixel channels, which includes 16
FADC modules managed by two field programmable
gate array (FPGA) modules.
AD9215, has a good cost-performance ratio with a lin-
294
295
296
The 10-bit FADCs,
297
298
earity better than 2% over a range of 40 to 900 counts.
However, it is not sufficient to cover the required dy-
namic range of the detector. Two FADCs are then used
for the dual gain system. At an input range of 2 V, the
FADC continuouslydigitizes analogsignals with a sam-
pling rate of 50 MHz (20 ns/bin) and a resolution of 2
mV per count. In the dual gain system, a range of 0 to
500 photoelectrons with a resolution of 1.7 counts per
photoelectron is set for the high gain channel, whereas
a range of 0 to 4000 photoelectrons with a resolution
of 0.21 counts per photoelectron is set for the low gain
channel.
A digitized waveform is collected by an FPGA, Xil-
inx XC3S1000, and fed into a pipeline with a length
of 1500 clock cycles. Every 300 cycles is defined as a
frame in which a single channel trigger is formed. Such
a long pipeline enables an enduring waveform wait-
ing for global trigger formation and transmission. This
pipeline also allows three frames, i.e., previous, current,
and post ones, to be recorded once a global trigger is
received. This guarantees storage of a complete wave
form regardless of when the pulse starts in the ”current”
frame.
The FPGA also makes a choice between the signals
from high gain and low gain channels when the bit
stream flows in. Any signal higher than 900 counts trig-
gers a switch from the high gain channel to the low gain
channel. This way, an effective overall dynamic range
of 12 to 13 bits is achieved.
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
2.4. Trigger system
328
A final trigger among the telescopes is determined
based on the three trigger levels, namely, the single
channel trigger (the lowest level), telescope trigger (the
second level), and event trigger (the highest level). The
topology of the three-level trigger algorithm is outlined
in Fig.6, with correspondingdetails discussedin follow-
ing subsections.
329
330
331
332
333
334
335
2.4.1. First Level Trigger in a Single Channel
The FLT is formed in the FPGA located on the
DBs provided that the signal-to-noise ratio in a win-
dow is greater than a given threshold (e.g., 4 as a typi-
cal value). The width of the window is predetermined
for corresponding observation modes (e.g., 8 bins for
the Cherenkov light signals). Running over the entire
frame of 300 cycles once in a bin, 293 sums of FADC
counts in the windows (denoted as WINSUM) are pro-
duced. The average of WINSUM is calculated using
these WINSUMs avoiding the maximal WINSUM and
336
337
338
339
340
341
342
343
344
345
346
5

Page 6

Figure 6: Three-level trigger system. SM is stereo or mono enable
signal, GL is global or local enable signal, T1 is the telescope trigger
from telescope No. 1 and T2 is from telescope No. 2. See detailed
explanation of the diagram in the text.
18 WINSUMs on each side of the maximum. The stan-
dard deviation (σ) of the WINSUM around the aver-
age serves as a measure of the background fluctuation.
If the maximum of WINSUMs exceeds the average by
nσ, the FLT is formed, where n is preselected from a
list of
the run. All trigger signals from 64 FPGAs are encoded
into bitmaps and transmitted to the Trigger Board (TB)
in parallel (see Fig.6).
The frame size of 300 bins is selected for a full scale
array with more than 24 telescopes covering an area of
1 km2, over which a highly inclined air shower takes
several microseconds to cross in the fluorescence light
observational mode. Such a frame is large enough to
fully contain the entire shower; it is also large enough
for Cherenkov light. All Cherenkov photons arrive at
the telescopes at almost the same time (few ns). There-
fore, an air shower signal appears only within 5 or 6
bins after shaping and the night sky backgroundis mea-
sured in the rest of the frame (see Fig.5). Such contin-
uous measuring of the sky background is highly useful
not only in estimating the signal-to-noise ratio but also
in monitoring the transparency of the atmosphere using
well-known bright stars in the field of view.
347
348
349
350
351
√12,
√16,
√24,
√32 and
√40 before starting
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
2.4.2. Second level trigger for a single telescope
A trigger for a telescope registered by an air shower
(denotedas second level triggeror SLT) is formedwhen
a specific pattern of triggered PMTs corresponding to
a possible air shower is found in the camera of a tele-
scope. A pattern recognition technique, developed in
the Pierre Auger experiment [11] that is operated in a
FPGA located on the TB, is applied. Numerous pat-
terns,suchasafullyfilledcircleformedwithonehexag-
onal pixel surrounded by six others, or a straight line
formed by six aligned pixels, are pre-loaded into the
371
372
373
374
375
376
377
378
379
380
381
Figure 7: Two typical patterns in the second level trigger.
FPGA as a look-up table. Once all 256 FLT signals are
collected, the FPGA matches all the pre-stored patterns
with the observed one within a 6 × 6 matrix of pixels.
Moreover, it keeps such a box running throughout the
entire PMT camera with a step of one row or one col-
umn. The SLT is formed as long as any one of the pre-
stored patterns is matched. According to the simulation
for air showers, mainly two types of patterns exist. Flu-
orescence light images of air showers seen from several
kilometers away tend to form line-shaped patterns on
thecamera, whereasCherenkovlightimages ofshowers
hitting the telescope head-ontend to formround-shaped
patterns, as shown in Fig.7. There are 16 round-shaped
patterns and 729 line-shaped patterns in a 6 × 6 box.
To speed up the formation of SLT, the pattern com-
parisons in the box are conducted in parallel. All 121
bitmaps of the boxes are generated by sliding the box
are done in parallel. All 121 bitmaps of the boxes are
generated by sliding the box throughout the camera. A
pattern comparison algorithm is then carried out among
the 121 bitmaps. The telescope trigger (SLT) is formed
in 123 clock cycles, i.e., 2.46 µs.
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
2.4.3. Third level trigger for entire array of telescopes
An event trigger for the entire array of telescopes
(denoted as the third level or TLT) is generated using
one of the FPGAs used for SLT. For the two proto-
type telescopes, only two modes, namely, stereoscopic
andmonocularobservationof showers, exists if they are
configured in such a way that the two telescopes have a
maximal overlap of the FOV. In stereoscopic mode, the
two telescopes are required to simultaneously observe a
shower. In monocular mode, each of the two telescopes
can trigger the entire site. The operational mode should
be selected at the beginning of a run by assigning two
controlling parameters, SM and GL, as marked in Fig6.
The two telescopes communicate with each other
through a two 80 m coaxial cables by sending telescope
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
6

Page 7

triggers(SLT) out andreceivingthe eventtrigger(TLT).
Such a trigger function is also useful for receiving an
externaltrigger and deliveringa triggerto other co-sited
detectors. This makes the telescope an open and modu-
larized system.
419
420
421
422
423
2.5. DAQ system
The philosophy of the design of the Cherenkov tele-
scope DAQ is integrated and compact so that the en-
tire DAQ can be arranged on the backboard of the focal
plane camera. It allows for the maximal mobility of the
telescope for flexibility in switching between configura-
tions. Because of the limited space, a low power con-
suming and compact-embedded online computer based
on an ARM processor, industrial standard PC104 bus,
and flash disk are selected as the backbone of the DAQ
system, eliminating the need for moving parts such as
the CPU fan and hard drives. All of the components are
integrated on a bus driver board, which bridges the data
storage disk and the 32 DBs that are connected by flat
cables. Because the event rate is not extremely high,
i.e., around 1 Hz, a band width of 14 MHz for a PC104
busis sufficientfortransferringeverybitofthe6µs long
waveformsto thedisk forall 256channels. Itis also ex-
tremely advantageous for system debugging. Each tele-
scope is an independent detector with a complete DAQ
system on board. The communication with the rest of
the experiment is through a 10/100 Mbps Ethernet.
Each telescope has its own independent DAQ, devel-
oped using C++ language under Linux. Operating in
a polling mode instead of an interrupt mode, the DAQ
checksthe interfacestatus regularlyforwhetherthedata
are ready or not. If so, the data in the buffer are read
and immediately stored in the disk. Before a night shift
ends, the data in the hard disk are moved to the Institute
of High Energy Physics (IHEP) in Beijing for further
analysis.
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
2.6. Power supply, slow control, and status monitoring
Thetwopowersupplieswith+7V(maximumcurrent
80 A) and -7 V (maximumcurrent 20 A) are installed in
the container. Regulators are used at each AB and DB
to furtherstabilize the voltage. An adjustableHV power
supply with a maximal output of -2000 V and 100 mA
is used for each telescope.
One of the difficulties at high altitudes is the heat dis-
sipation of the power supplies and electronics enclosed
in a metal box. The total power consumption is about
50 A at +7 V and 16 A at -7 V. A forced cooling sys-
tem is necessary in the prototypeexperimentto improve
the heat dissipation of the entire camera and the power
supplies.
455
456
457
458
459
460
461
462
463
464
465
466
467
468
The entire telescope system is powered by an unin-
terrupted power supply with a sufficient battery backup.
It protects the telescopes from damage when a blackout
occurs.
The detector is designed to work in remote control
mode, including the opening and closing of the doors,
turning the power supplies on/off, enabling/disabling
the high voltage (HV) power supply and low voltage
(LV)powersupplies,andswitchingtheUVLEDson/off
for detector calibration. All kinds of controls are real-
ized using the on-board computer through a COM port.
Carefullymonitoringthe status of eachtelescope is nec-
essary, including the door status, voltages, and tem-
peratures at different places (e.g., enclosed areas inside
the camera, backboard of the PMT camera, and inside
the housing of the UV LED). All parameters are mea-
sured and recorded through an 8-channel 12-bit analog-
to-digital-converteron the board of the embedded com-
puter, TS7200.
All of the controls are performed through a user in-
terface running at the embedded computer connected
through Ethernet from an operational center, 3000 km
away in IHEP in Beijing. A small portion of the data
can be copied to display the event during the operation.
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
3. Calibration
493
3.1. Method of calibration
494
An accurate shower reconstruction requires a con-
verting factor from a pulse area in terms of FADC
counts to the number of photons for each pixel. There-
fore, the absolute calibration of the detector response is
essential. To achieve this objective, having an accurate
knowledge about PMT cathode effective areas, cath-
ode quantum efficiency, PMT gains, amplifier gains,
and digital converting factors, is necessary. Measur-
ing all these effects item by item is difficult. In this
paper, a method similar to the HiRes experiment [14]
andPierreAugerexperiment[12]calibrationprocedure,
which considers the entire effect, is applied to measure
the overall response of each pixel. The procedureis dis-
cussed in the paragraphs that follow.
Being mountedat the center of the mirror, a UV-LED
(375 nm) light source with a diffuser is used for the cal-
ibration of PMTs in the camera by beaming nearly uni-
form light to every pixel. The LED light density on the
camerasurfaceis calibratedusinga pre-calibratedprobe
detector located beside the camera. The calibration of
all pixels in the cameras is performed twice a day, i.e.,
before and after the daily operation. The crucial part is
the measurementof the absolute numberof UV photons
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
7

Page 8

at each PMT cathode from a pulse emitted by the LED,
which is performed in two steps.
First, we move one PMT over all the places on the
frame of the PMT camera to measure the uniformity of
the LED light density on the camera surface. This has
to be accomplished before the PMT camera is installed.
The light density is a function of the polar angle, θ, in a
form of cos4θ, where θ is the angle between the connec-
tion from the LED to the PMT location and the perpen-
dicular connection from the LED to the PMT camera
surface. The variation from the center to the corner of
the PMT camera is within 7%.
Inthe secondstep, thelight densityis measuredusing
a pre-calibrated probe consisting of two XP3062 PMTs
with the same FEE and DAQ as the two telescopes. The
only difference is that the two PMTs in the probe are
operatedat a veryhighgainso that the single photoelec-
tron can be measured. Therefore, the gain of the probe,
Gprobe, can be calibrated at any time. Then, the absolute
calibration of the probe is performed at the HiRes lab
at the University of Utah, USA, by comparing the re-
sponse of a hybrid photo diode (HPD) pre-calibrated at
NIST [18] to the same light source. The probeand HPD
arelocatedside by side in frontof a UV-LEDat 355nm.
Using the HPD, we measure light density IU(numberof
photons per square millimeter) from the LED. We also
measure pulse area Fprobe
U
ously. At the operational site in Tibet, the light density
from the LED mounted on the center of the mirror is
calibratedto be IT= IU
Fprobe
U
for Utah and Tibet, respectively.
Applyingthe knowledgeobtainedfromthe two steps,
we have determined both the light density in front of
the cathode and the pulse area in terms of the FADC
counts for each PMT in the camera. Finally, the over-
all converting factor for a pixel between the number of
photons reached to the surface of the camera and cor-
responding pulse area measured by the FADC counts
behind the pixel is
Fcamera
T
ITAPMT
where APMTis the geometric area of the PMT cathode,
and Ccamera
375
is the calibration constant for a pixel. The
subscript 375indicates that the calibration is done using
UV light at 375 nm. The unit of the calibration constant
is FADC counts per photon. θprobeand θcamerarepresent
the angular locations of the probe and the pixel in the
camera, respectively. A further correction according to
the wavelength dependence of the quantum efficiency
of PMTs is applied in the operation for the cosmic ray
observation.
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
using the probe simultane-
544
545
546
Fprobe
T
Gprobe
U
Gprobe
T
, where U and T stand
547
548
549
550
551
552
553
554
555
556
Ccamera
375
=
cos4θprobe
cos4θcamera,
(1)
557
558
559
560
561
562
563
564
565
566
567
Modified Julian Day
54820 54840 54860 54880 54900 54920 54940 54960 54980
FADC counts/photon
0.25
0.3
0.35
0.4
0.45
0.5
Figure 8: Absolute gain of one telescope from Dec. 2008 to May
2009, as a function of time.
3.2. Result of the calibration
The probe was calibrated at the HiRes lab. The cali-
brationresultsareshownintable.1; thegainoftheprobe
is 85.7±0.3 FADC count per SPE. The second column
shows the lower LED photon density and the third col-
umnshows the higherLED photondensitythat coincide
with one another.
568
569
570
571
572
573
574
Table 1: The calibration results of the probe.
IU(photons/mm2)
Fprobe
U
(FADC count)
0.359 ±4.8%
8959 ±4
0.538 ±4.8%
13650 ±5
The correspondingcalibration constant for one of the
telescopes is presented in Fig.8. An average of the cal-
ibration constants of all pixels in the camera is plotted
with respect to time from December 2008 to May 2009.
As mentioned above, the calibration constant is mon-
itored every observational day. The systematic uncer-
tainty of the calibration constant is estimated to be 7%.
The downtrend of the constant shown in the figure indi-
cates obvious decreases of gains in all newly produced
of PMTs. The transmission of the glass window and re-
flectivity of the mirrors are not take into account in the
above calibration. These two effects will be monitored
in so called end-to-end calibration using nitrogen laser
in future.
575
576
577
578
579
580
581
582
583
584
585
586
587
588
4. Performances and Results from Test Run
589
4.1. Test Run Information
The two telescopes began recording cosmic ray data
in August 2008.Furthermore, both monocular and
stereoscopic modes have been tested. About 500,000
590
591
592
593
8

Page 9

coincidental events with the ARGO-YBJ experiment
in stereoscopic mode and 700,000 coincidence events
in monocular mode have been collected up to January
2010. The average trigger rate is about 0.5 Hz in stereo-
scopic mode and 0.7 Hz in monocular mode.
594
595
596
597
598
4.2. Off-line Coincidence with the ARGO-YBJ Experi-
ment
All WFCTA telescopes and the ARGO-YBJ experi-
ment recorded the cosmic ray arrival time based on a
GPS. A time window of 8 µs containing a Cherenkov
event is searched for coincidence with the ARGO-YBJ
event stream, which is about 4 kHz. A difference be-
tween the recorded event time by the two experiments
for a matched event is typically less than 100 ns (Fig.9).
Fora coincidenceevent,the shower geometryis mea-
sured using the ARGO-YBJ detector. The distribution
of shower arrival directions is shown in a 2-dimensional
map, i.e., zenith angles versus azimuth angle, as in
Fig.10. Approximately 85% of coincident events occur
inside the FOV of the Cherenkov telescopes marked by
the trapezium. The rest of the 15% of events have their
images partially seen by the telescopes and sufficiently
trigger the telescopes. The shower core distribution is
shownin Fig.11. TheARGO-YBJ experimentarrayand
the two Cherenkov telescopes are marked as the rectan-
gle and two dots in the figure, respectively. For events
that have the reconstructedcores inside the ARGO-YBJ
array,the shower parameters, such as the numberof hits
on the carpet detector (Nhit) and shower geometry, are
well measured. A distribution of Nhitof these events is
plotted in Fig.12. According to the number of hits as a
function of the primary energy of proton from ARGO-
YBJ detector [19], the mode energy of protons is about
40 TeV. This estimates the threshold of two Cherenkov
telescopes in stereoscopic mode.
In Fig.13, two Cherenkov images are shown in the
event display. The top image is caused by an event at
a 10 m Rp(the impact parameter of the shower to the
telescope) and the bottom image stems from an event at
a 173 m Rp. The image is clearly moreelongatedforthe
farther event. Using the well-defined image parameters
created by Hillas [21], length and width, this effect is
more quantitatively presented in Fig.14. The widths of
the images seem to be no longer shrinking once show-
ers are sufficiently far from the telescopes (e.g., farther
than 100 m). The showers essentially resemble linear
patterns.
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
4.3. Electronic noisy and sky background
For a triggering system completely based on the
signal-to-noise ratio, such as the WFCTA telescopes,
641
642
643
/ ndf = 97.04 / 74 / ndf = 97.04 / 74
22
χχ
Constant Constant 8.3 8.3
±±
926.5 926.5
Mean Mean 0.60 0.60
±±
1.31 1.31
Sigma Sigma
0.45 0.45
±±
83.17 83.17
dt (ns)
-400-300-200-1000100200300400
Entries
0
100
200
300
400
500
600
700
800
900
Figure 9: The difference time between WFCTA and ARGO in coinci-
dence event.
azimuth (deg.)
200210220 230240250260
zenith (deg.)
15
20
25
30
35
40
45
0
500
1000
1500
2000
2500
Figure 10: Distribution of coincidence events are seen by ARGO and
Cherenkov telescopes simultaneously over zenith and azimuth angle.
About 85% of coincidence events are located in the Cherenkov tele-
scope FOV (trapezium). The image of the rest of the 15% of coinci-
dence events is partially in the Cherenkov telescope FOV.
the sensitivity is constrained by the noise level. For a
most optimized system, the electronic noise must be
negligible compared with the night sky background.
Both of them are measured during the operation as the
door is closed and opened, respectively.
The sources of electronic noise include thermal
noises in the PMTs, noises from the HV power supply
and LV power supplies, noises of amplifiers, and finally
from the counting error of FADCs. The PMT Photonis
XP3062/FL has a very low dark current; therefore, the
thermal noise level is sufficiently low so that it can be
ignored compared with the other sources of electronic
noises. The HV power supply has a ripple less than
0.02% in terms of RMS of rated voltage. This con-
tributes a variation within 0.12% in terms of the gain
of the PMTs. Voltage regulators are used to block pos-
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
9

Page 10

x (m)
-200 -150 -100-50050100150 200
y (m)
-200
-150
-100
-50
0
50
100
150
200
0
500
1000
1500
2000
2500
3000
ARGO
Two WFCTA Prototypes
Figure 11: Distribution of shower core. Two Cherenkov telescopes
(two dots) and the ARGO-YBJ experiment array (rectangle) are also
marked in the figure.
log10(number of hits)
2.533.54 4.5
Entries
0
200
400
600
800
1000
1200
Figure 12: Distribution of ARGO nhit of these coincidence events,
whose reconstructed core is located in the ARGO-YBJ cluster array.
sible noises from the LV power supplies. All ampli-
fiers, AD8138 and AD8039, are selected to generate
a very low noise level. The noise from the FADC is
about 0.5 FADC counts per tube. Taking into account
all additional sources such as the distributing capacity
on boards and connectors, the total electronic noise, in-
cluding PMT and high voltage power supply, is typi-
cally about 1.0 FADC counts per tube. It is measured
during the calibration with the LED (Fig.15) as long as
the signal is avoided.
The night sky background is measured similarly as
a shower trigger is formed when the doors are open
(Fig.5). The fluctuation in the night sky back-ground is
typically more than 2.2 photon electrons per 20 ns per
tube on a clear moonless night. This suggests a much
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
Figure 13: Two Cherenkov events; the Rp of the tope event is 10 m
and that of the bottom event is 173 m; the image of the bottom event
has a longer tail than the top event.
stronger noise than electronic noises. One of the impor-
tant sourcesoftheskybackgroundis lightfromthe stars
(Fig.16), which shows the sky background at night, in
which each pulse denotes stars passing thoughthe FOV.
A bright individual star can be traced when it passes
through the telescopes. Stars provide numerous stable
point-like sources that typically sweep across a tube in
about 4 minutes. Those ideal point-like sources can be
used for multiple calibration purposes. For instance,
well-known bright stars can be used as light houses to
establish the pointing direction of the telescope itself,
with accuracy proved to be better than 0.05◦[20]. A
bright star can be seen as a very stable point-like source
at infinity, which is a perfect tool for measuring the spot
size produced by the optical system. The background
of a tube signal steadily increases when the light spot
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
10

Page 11

Rp (m)
050100150200250300
deg.
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
Length
Width
Figure 14: Length (filled circles) and width (open circles) as a func-
tion of Rp.
Bin No.
0 100 200 300400500 600700800 900
FADC counts
50
100
150
200
250
Figure 15: LED calibration for door closed. The low noise level of
1 LSB of the complete electronics is to be seen before and after the
LED pulse.
moves into the FOV of the tube, and steadily decreases
when it moves away from the FOV, forming a light pro-
file. The light profile can be fitted using a Gaussian
function, whose sigma denotes the spot size. The spot
size grows larger when the star moves away from the
center of camera (Fig.17).
691
692
693
694
695
696
5. Summary
697
The telescopes were successfully run at YBJ from
August 2008 up to July 2009. Millions of coincidence
events with the ARGO-YBJ experiment have been col-
lected. The performance of the telescopes was studied
using these events. The trigger rate is about 0.5 Hz in
stereo mode. Moreover, the mode energy of the tele-
scope is 40 TeV when a pure proton composition is as-
sumed.
698
699
700
701
702
703
704
705
time (10s)
5001000150020002500 3000 3500
photon electrons per us
0
200
400
600
800
1000
1200
1400
1600
1800
Figure 16: Sky background is monitored by a PMT in a clear moon-
less night; each pulse corresponds to a star passing through the FOV
of the telescope.
) °
Angular distance to mirror axis (
2345678
Relative brightness
0
0.01
0.02
0.03
0.04
0.05
0.06
°
0.29
°
0.24
°
0.21
Figure 17: A star moves away from the center of the camera, and
its light profile recorded by PMT is drawn in filled circles. The light
profile described by a Gaussian function is drawn in a solid line. The
sigma of Gaussian denotes that the spot size is also marked.
The features of the two WFCTA prototypetelescopes
are summarized as follows:
706
707
• a 4.7 m2spherical mirror,
• a 16×16 PMT array covers an FOV of 14◦× 16◦
with 1◦pixels,
708
709
710
• dual gain system for a dynamic range to 3.5 orders
of magnitude,
711
712
• DC couplingand modulizeddesign for electronics,
• three-levelonlinetriggerlogic: singlechanneltrig-
ger based on S/N ratio, telescope trigger based on
pattern recognition, and event trigger for stereo-
scopic observation,
713
714
715
716
717
11

Page 12

• maximized mobility and the telescope can be up-
lifted from 0◦to 60◦in elevation.
718
719
Theabsolutegainsofthe telescopesare calibratedus-
ing calibrated LEDs mounted at the centers of the mir-
rors. The systematic uncertainty of the calibration con-
stant is about 7%. The pixel gains are monitored on a
daily basis.
720
721
722
723
724
6. Acknowledgements
725
This work is supported by the Chinese Academy of
Sciences (0529110S13)and the Key Laboratory of Par-
ticle Astrophysics, Institute of High Energy Physics,
CAS. The Knowledge InnovationFund (H85451D0U2)
of IHEP, China and the project Y0113G005C of NSFC
also provide support to this study.
WeareverygratefultotheARGO-YBJCollaboration
for authorizing us to use the data of the ARGO-YBJ
experiment.
We also acknowledge the essential support of H. M.
Zhang, W.Y. Chen, G. Yang, X.F. Yuan, C.Y. Zhao in
the installation, debugging, and maintenance of the de-
tector.
726
727
728
729
730
731
732
733
734
735
736
737
738
References
739
[1] V.I. Zatsepin et al., Nucl. Instr. and Meth. A 524 (2004) 196
195-207
[2] H.S. Ahn et al. Nucl., Instr. and Meth. A 579 (2007) 1034-1053
[3] Zhen Cao et al., A Future Project at Tibet: The Large High Al-
titude Air Shower Observatory (LHAASO), Chinese Physics C
2010, 34 (02): 249-252
[4] Huihai He et al., LHAASO Project: detector design and proto-
type, 31st ICRC, LODZ,(2009).
[5] Z. Cao et al., j. Phys. G: Nucl. Part. Phys. 31 (2005) 571
[6] He HuiHai et al., Proc. of 30th ICRC, Vol. 5 (2007) 949
[7] G. Aielli et al., Nuclear Physics B (Proc. Suppl.) 166 (2007)
96-102.
[8] J.Abraham et al., The Fluorescence Detector ofthe Pierre Auger
Observatory, Nucl. Instr. and Meth. A620 (2010) 227-251
[9] J. Abraham et al., Measurement of the energy spectrum of cos-
mic rays above 1018eV using the Pierre Auger Observatory,
Phys. Letter B685 (2010) 239-246
[10] T.Nonaka et al., The present status of the Telescope Array ex-
periment, Nuclear Physics B (Proc. Suppl.) 190 (2009) 26-31
[11] J. Abraham et al., Nucl. Instr. and Meth. A 523 (2004) 50-95
[12] P.Bauleo et al., 29th International Cosmic Ray Conference Pune
(2005) 8, 5.8
[13] John H. Boyer et al., Nucl. Instr. and Meth. A 482 (2002) 457-
474
[14] D. Bird et al., Nucl. Instr. and Meth. A 349 (1994) 592-599
[15] T. Abu-Zayyad et al., 2001 ApJ 557 686
[16] S. P. Swordy, D. B. Kieda, arXiv:astro-ph/9909381v1
[17] G.Xiao et al., 29th International Cosmic Ray Conference Pune
(2005) 8, 21-24.
[18] L.P. Perera,Calibration of the Roving Xenon Flasher
with a Hybrid Photodetector internal HiRes report, 2003,
http : //www.cosmic − ray.org/papers/Lalith Hpd.pdf.
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
[19] GUO Yi-Qing et al, Chinese Physics C, 2010, 34 (5): 555-559
[20] Ma Ling-Ling et al, Geometry and optics calibration of WFCTA
prototype telescopes using star light, Chinese Physics C, pro-
ceeding.
[21] Hillas A. 1985, in Proc. 19nd I.C.R.C. (La Jolla), Vol. 3, p. 445.
772
773
774
775
776
12