• A Cosmic Ray Trigger for LOFAR

29th International Cosmic Ray Conference Pune (2005) 8, 245–248
A Cosmic Ray Trigger for 
S. Lafebrea , A. Nigla , W.D. Apelb , F. Badeab , L. B¨ hrenc , K. Bekkb , A. Bercucid ,
a
M. Bertainae , P.L. Biermann f , J. Bl¨ merb,g , H. Bozdogb , I.M. Brancusd , S. Buitinka ,
u
M. Br¨ ggemannh , P. Buchholzh , H. Butcherc , A. Chiavassae , K. Daumillerb , A.G. de Bruync ,
u
C.M. de Vosc , F. Di Pierroe , P. Dollb , R. Engelb , H. Falckea,c, f , H. Gemmekei , P.L. Ghia j ,
R. Glasstetterk , C. Grupenh , A. Haungsb , D. Heckb , J.R. H¨ randelg , A. Horneffera, f ,
o
T. Huegeb, f , K.-H. Kampertk , G.W. Kantc , U. Kleinl , Y. Kolotaevh , Y. Koopmanc ,
O. Kr¨ meri , J. Kuijpersa , G. Maierb , H.J. Mathesb , H.J. Mayerb , J. Milkeb , B. Mitricad ,
o
C. Morello j , G. Navarrae , S. Nehlsb , R. Obenlandb , J. Oehlschl¨ gerb , S. Ostapchenkob ,
a
S. Over , H.J. Pepping , M. Petcu , J. Petrovic , T. Pierog , S. Plewniab , H. Rebelb ,
h c d a b
A. Rissem , M. Rothg , H. Schielerb , G. Schoonderbeekc , O. Simad , M. St¨ mpertg , u
G. Toma , G.C. Trinchero , H. Ulrich , S. Valchierotti , J. van Buren , W. van Capellene ,
d j b e b
W. Walkowiakh , A. Weindlb , S. Wijnholdsc, J. Wocheleb , J. Zabierowskim , J.A. Zensus f and
D. Zimmermannh
(a) Dept. of Astrophysics, IMAPP, Radboud University, 6525 ED Nijmegen, The Netherlands
(b) Institut f¨ r Kernphysik, Forschungszentrum Karlsruhe, 76021 Karlsruhe, Germany
u
(c) ASTRON, 7990 AA Dwingeloo, The Netherlands
(d) National Institute of Physics and Nuclear Engineering, 7690 Bucharest, Romania
(e) Dipartimento di Fisica Generale dell’Universit` , 10125 Torino, Italy
a
(f) Max-Planck-Institut f¨ r Radioastronomie, 53121 Bonn, Germany
u
(g) Institut f¨ r Experimentelle Kernphysik, Universit¨ t Karlsruhe, 76021 Karlsruhe, Germany
u a
(h) Fachbereich Physik, Universit¨ t Siegen, 57068 Siegen, Germany
a
(i) Inst. Prozessdatenverarbeitung und Elektronik, Forschungszentrum Karlsruhe, 76021 Karlsruhe, Germany
(j) Istituto di Fisica dello Spazio Interplanetario, INAF, 10133 Torino, Italy
(k) Fachbereich Physik, Universit¨ t Wuppertal, 42097 Wuppertal, Germany
a
(l) Radioastronomisches Institut der Universit¨ t Bonn, 53010 Bonn, Germany
a
(m) Soltan Institute for Nuclear Studies, 90950 Lodz, Poland
Presenter: Andreas Nigl (a.nigl@astro.ru.nl), net-nigl-abs2-he15-poster
We present the development and first results of an independent cosmic ray trigger for the multiple dipole an-
tenna radiotelescope  (LOw Frequency ARray).  (LOfar PrototypE Station), at the  (KArls-
ruhe Shower Core and Array DEtector) site in Germany, has been initiated as a test case for , and designed
to detect air showers through their coherent radiation pulses upon external triggers by particle detectors. To
fully exploit the capabilities in detecting s with the final  telescope, however, an independent, radio-
only trigger is needed. We are searching for pulse coincidences in the antenna digital data stream. The limiting
factor for applying real-time detection is the capacity of digital processing of the 8 · 10 7 samples per second per
antenna. Here, we discuss constraints on basic criteria for the detection, like number of antennas, pulse height,
pulse width, direction, distance of shower maximum and polarisation. We also present the first results of the
application of such a trigger, and discuss optimization of the different parameters.
1. Introduction
A Dutch consortium, headed by  (The Netherlands Foundation for Research in Astronomy), is building
a new radio telescope which is based on an array consisting of simple dipole antennas (see Figure 1a) with
bandwiths of 30 to 80 MHz and 120 to 240 MHz. This telescope is called  (Low Frequency ARray), and
246 S. Lafebre et al
 
Figure 1. : A prototype  antenna. The dipoles run down the black  tubes. :  layout showing the core and
logarithmically spaced remote stations on five spiral arms in the border region of the Netherlands and Germany.
will consist of a core of a few square kilometers (3 200 antennas), and 61 outer stations with a diameter of
100–200 m (100 detectors) each, spread out over an area 150 km in diameter in total (see Figure 1b). The total
bandwidth from 30 to 240 MHz is covered by two different antenna types. The low frequency antennas are
tuned to 30 to 80 MHz, the high frequency antennas to 120 to 240 MHz.
Of the five key projects, four propose to observe specific astronomical objects, and one addresses the study of
Ultra High Energy Cosmic Rays (). In this latter project, we aim to detect air showers through radiosyn-
chrotron radiation. The charged particles in the shower (mainly electrons and positrons) are deflected in the
Earth’s magnetic field and emit coherent geosynchrotron radiation [1, 2].
Currently, three test stations for  exist, two of which are optimized for cosmic ray air shower measure-
ments in the radio regime from 40 to 80 MHz.  [3], the LOfar PrototypE Station consists of 30 dipoles.
For simultaneous data acquisition of particles and radio emission,  is triggered by the KArlsruhe Shower
Core and Array DEtector (), situated at the Forschungzentrum Karlsruhe in Germany [4]. Since the
beginning of 2004 we have recorded more than a million triggered events, more than a thousand of which are
detected in radio. The second prototype is called  (LOfar @ Radboud University Nijmegen) consisting
of four crossed dipole antennas on top of the university building.  is triggered by two particle detectors
of the Nijmegen Area High School Array (). The use of this prototype station lies mainly in the fields of
outreach and education. The third test station is  (Initial Test Station), but currently no cosmic ray research
has been carried out with this instrument.
From confirmed detections of cosmic ray air showers with  [5], we have learned about basic properties
of the antenna response to the radio signal. Based on this experience, we are developing a method to trigger
the antennas without using particle detectors. In this way, all kinds of transient events in the radio data can
be studied, and detecting air showers is not limited to the field of view and sensitivity of an external particle
detector.
In this article we explain the self-trigger method and discuss its application for the  project.
 A Cosmic Ray Trigger for  247
2. Method
For the detection of Coscmic Rays () with  dipole antennas, we develop an algorithm to find short
pulses in the digitized data. The  design suggests to divide the trigger into three hardware levels. In the
first step, the signal will be monitored on a per-antenna level. A dedicated transient buffer board, consisting of
Field Programmable Gate Arrays (s), performs real-time data analysis according to the algorithm
|x(t)| > µn (t − t1 ) + kσn (t − t1 ), (1)
where x(t) is the received signal, and µn (t − t1 ) is the time averaged signal, running over a time interval tn
(containing n samples) ending at time t − t1 . This is done to avoid ‘contamination’ of the average. σn (t − t1 )
is the standard deviation over the same sample block, and k is the threshold factor for the standard deviation.
When the condition is fulfilled and the signal is above an overall power level, the station core processor is
notified.
At station level, the core processor monitors the antenna messages. A station trigger is generated, when the
number of antenna messages exceeds a decision limit na within a certain time window t2 , e.g. the light travel
time between the antennas. When the decision for a trigger is made, the relevant part of the data buffer for
all antennas in the station is downloaded and sent to the  core processor (∼25 MB for 1 ms of data per
antenna). Since the data production is not continuous the data transfer can be performed without disturbing the
data acquisition of other observations.
Finally, at full  level, the obtained event is roughly analyzed to obtain estimates for shower properties
like direction of arrival, lateral particle distribution, and primary energy. If the event is found to be a ‘random’
occurence it is deleted, else it is transferred to the data center at the Radboud University Nijmegen, where
further offline analysis is performed. We estimate to receive up to a few GB of data per day.
3. Parameter optimization
The above trigger introduces a handful parameters to adjust. First of all, we have to rely on absolute gain
calibration of a single antenna element to set a minimum absolute signal height for the peak search algorithm.
This level also sets a lower limit on the minimum cosmic ray energy that can be detected at single antenna
level. Since no absolute calibration is available yet for  or , we will not discuss this parameter here.
To test dependencies on the other parameters, a collection of bright events was used, which is a subset of the
selection by A. Horneffer [3]. Steps 1 and 2 of the trigger algorithm were run over 156 event files, using
various combinations of parameters. Because the exact time of the occurence of a cosmic ray pulse is known,
the returned triggers could be marked ‘correct’ or ‘incorrect’.
Changing na between 4 and 7 (out of 8 available dipole signals) does not influence the amount of correct
detections much. Changing the value for t2 also hardly has an effect on the trigger quality; a value of 0.63 µs,
slightly larger than the light travel time through the detector, does the trick quite well, and this value has been
used throughout.
The block size tn and threshold value k are probably the most important selector for the type of event one
wants to trigger on: larger block sizes will trigger broader events, higher threshold values will produce less
triggers in general. In Figure 2a, we tested the performance of our algorithm for certain values of k and t n . This
performance ratio, given on the vertical axis, is the fraction of correctly delivered triggers over the total amount
of triggers. The error bars also provide some information on the total amount of triggers given: the longer the
bars, the lower the number of triggers. Therefore, good data points lie in the top of the diagram and have a
248 S. Lafebre et al
0.9
0.8
!
0.7 ¢£
+ ¤¥
+
23
+
+
Fraction of correct triggers
+
¦§ ¨©
0.6  ¡
 
+ 01

0.5
+
+ ()
0.4

0.3
tn = 0.4 µs

+
0.2
&'
tn = 0.8 µs +
tn = 1.6 µs "#
0.1 tn = 3.2 µs
+
+
tn = 6.4 µs
$%
45
0
5 6 7 8 9
k
Figure 2. Trigger performance for the discussed trigger algorithm for various values of k and tn . In this plot, t1 = 0,
t2 = 63 µs, and na = 5. The error bars are a measure for the absolute number of detected events: the longer the error bars,
the smaller the number of events.
small error bar. From the diagram, it is clear that the performance ratio greatly varies with the parameters, and
one can say that a combination of k = 6, tn = 0.8 µs gives best results.
The time shift t1 has not been invesigated thoroughly yet, but this will be done in the course of the next few
months. It is probably best to have the sample at t just after the sampled average, so that it is not contaminated
by the peak itself or additional interference from electronics.
4. Conclusions
For the brightest events (E > 1017 eV), the proposed trigger seems to work reasonably well. One has to keep
in mind, of course, that the data set over which it was tested is not extensive, and that the final setup will be
different from the one currently in use. For example, in the final  setup there will be a lot less radio noise
than in . Further testing on more () data and simulations will provide more insight.
References
[1] H. Falcke, P. W. Gorham; Astropart. Phys. 19, 477–494 (2003).
[2] T. Huege, H. Falcke; A&A 430, 779–798 (2005).
[3] A. Horneffer et al.; Proc.  5500, 129–138 (2004).
[4] T.Antoni et al.; Nucl. Instr. Meth. A513, 490 (2003).
[5] H. Falcke et al.; Nature Phys. Lett. 435, 313–316 (2005).
[6] P. Marciniewski; PhD Thesis, Uppsala University (2001)