Detection Techniques of Radio Emission from Ultra High Energy Cosmic Rays

Transcription

1 Detection Techniques of Radio Emission from Ultra High Energy Cosmic Rays DISSERTATION Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Chad M. Morris, B.S. Graduate Program in Physics The Ohio State University 2009 Dissertation Committee: Prof. James J. Beatty, Adviser Prof. John F. Beacom Prof. Richard J. Furnstahl Prof. Richard E. Hughes

2 c Copyright by Chad M. Morris 2009

3 Abstract We discuss recent and future efforts to detect radio signals from extended air showers at the Pierre Auger Observatory in Malargüe, Argentina. With the advent of low-cost, high-performance digitizers and robust digital signal processing software techniques, radio detection of cosmic rays has resurfaced as a promising measurement system. The inexpensive nature of the detector media (metallic wires, rods or parabolic dishes) and economies of scale working in our favor (inexpensive highquality C-band amplifiers and receivers) make an array of radio antennas an appealing alternative to the expense of deploying an array of Cherenkov detector water tanks or fly s eye optical telescopes for fluorescence detection. The calorimetric nature of the detection and the near 100% duty cycle gives the best of both traditional detection techniques. The history of cosmic ray detection detection will be discussed. A short review on the astrophysical properties of cosmic rays and atmospheric interactions will lead into a discussion of two radio emission channels that are currently being investigated. ii

4 This dissertation is dedicated to my wife, Nichole, who has tolerated and even encouraged my ramblings for longer than I can remember. To Eva who has quickly become a central part of my life. Finally, to my parents, Mike and Beci, without whom none of this could have happened. iii

5 ACKNOWLEDGMENTS Many thanks go out to my advisor, Jim Beatty, to the members of the Pierre Auger Observatory collaboration, members of the Auger Radio Collaboration, and the technical staff of PAO in Malargüe who went out of their way to help out any time there was a problem. iv

6 Contents Page Abstract ii Dedication iii Acknowledgments iv List of Tables viii List of Figures ix Chapters: 1. Introduction Current Detection Methods Surface Detection Fluorescence Detection Pierre Auger Observatory Radio Detection v

7 2. Basic Physics of Cosmic Rays Sources of Cosmic Rays Acceleration Mechanisms Energy Limitations Interactions in the Atmosphere OSU/Leeds Geosynchrotron Detector Radio Free Malargue Log Periodic Dipole Antennas OSU/Leeds LPDA Deployment Amplifier/Receiver Onboard Electronics Solar Power Communications Radio Central Data Aquisition System Trigger Logic Hardware Triggers Software Trigger Results Issues, Problems, Success Analysis Methods Suggestions for Improvement vi

8 3.9.4 Contributions Results Future Deployment AMBER: A New Cosmic Ray Detector Introduction MBR: Molecular Bremsstrahlung Radiation Accelerator Results AWA INCOBREMS / SLAC T AMBER Hardware Contributions vii

9 List of Tables Table Page 3.1 A quick measurement of arm lengths using a Leuthold as a reference length viii

10 List of Figures Figure Page 1.1 A Cartoon Schematic of the Pierre Auger Observatory Picture of Fly s Eye Telescope Layout of the Pierre Auger Observatory Photo of First Solar Powered Radio Detector at the PAO Photo of typical Surface Detector Water Tank The ANITA Payload Prior to Launch Cosmic Ray Energy Spectrum The Hillas Plot of physical size vs magnetic field Proton propagation distance with GZK interactions Schematic detailing an extended air shower development Satellite view of Radio test site at the Balloon Launch Facility Log Periodic Dipole Array Schematic LPDA Schematic showing phase reversal of smaller elements Example of Log Periodic Behavior Gain Contour Plot for LPDA antennas ix

11 3.6 Smith chart of the LPDA antenna E-Plane gain plot for the LPDA over multiple frequencies E-Plane gain plot of LPDA over multiple surfaces Photo of installation of first LPDA deployed in Malagüe Photo of the Leeds Amplifier Schematic of the radio frequency sliders of the Leeds Amplifier Schematic of entire chain for self-contained radio detection system Full schematic of an OSU/Leeds Station Sample gain plot of the amplifier A Unified Board Field test of solar power, communications and electronics Message packet format for internal radio communications Detailed breakdown of format of message payload Balloon launch facility FastLook software display High gain antenna and SD spy antenna A missing LPDA arm Plot of battery failure Trigger multiplicity bug Radio trigger rates before the bug fix Radio trigger rates after bug fix x

12 3.27 Rapidly varying trigger rates station Nearby station 2 with identical settings Radio Background scan with LPDA at BLF First possible radio event T. Huege REAS simulation of radio events FFT of simulated pulse Damaged station New LPDA design using wires and wood frame Aerial view of layout of MAXIMA stations A MAXIMA station Long term RMS noise level with galactic passage overhead Skymap of radio events found with CLF/CODALEMA radio detectors Test site for many Auger Observatory enhancements Aerial view of AERA site Cartoon schematic of Auger Detectors The Faraday box used in beamline tests at AWA and SLAC Schematics of AWA and SLAC tests AWA beam test results AWA beam test results with background Results from antenna polarized parallel to SLAC beamline xi

13 4.7 Results from antenna polarized perpendicular to SLAC beamline Scaling of energies at SLAC test Initial test setup on roof of physics building at Univesity of Hawaii m dish on Ohio State University West Campus dish farm C and Ku-band feed horns arranged in diamond configuration for OSU test xii

14 Chapter 1 Introduction In the 1960s, it was found that Extensive Air Showers (EAS) from cosmic rays produce pulses of radio frequency emission in the range of less than 100 MHz. Following numerous experimental advances through the mid-1960s, the field came to a standstill by the early 1970s due to the development of more promising experimental techniques. Recently, the development of low-cost digitization and powerful digital signal processing techniques has brought about a revival of radio detection of cosmic rays [1, 2] The Pierre Auger Observatory, a cosmic ray telescope located in Argentina, consists of 1600 Cherenkov water tank detectors with a 1.5 km spacing. Co-located with these water tanks are 24 fluorescence detector fly s eye telescopes at 4 locations surrounding the tank array. By placing an array of new radio detectors onsite, we can utilize the Auger data to assist in our prototype testing. We hope to eliminate transient false-positive triggers of the radio detectors caused by man-made and naturally occurring noise by tapping into the real-time trigger communications of the Observatory. This will improve our own triggering logic in giving a radio event that can be verified against the Auger data. Analyzing event data that has triggered both 1

15 detectors provides a valuable energy calibration for the prototype radio detection system [3]. There are two likely methods for radio detection of cosmic rays: geosynchrotron and molecular brehmsstrahlung radiation or MBR. The first method, geosynchrotron, is the traditional radio method [4]. Charged particles, electrons or positrons, are bent via the geomagnetic field and synchrotron radiation is produced. This is coherent and highly beamed in the direction of travel. A detector, here an antenna, must be in the path of the shower in order for a signal to occur. This places these antennas as analagous to the surface array tanks described above. The latter detection method, MBR, occurs isotropically as the electrons in the tenuous plasma left in the wake of the shower front cool back to thermal equilibrium. This would make detectors of this type analogous to the fluoresence detectors. Figure 1.1 illustrates the various detection methods employed at the Pierre Auger Observatory in Malargüe, Argentina. Over the past 4 years, we have attempted to deploy detectors of both kinds on-site at the Auger Observatory with differing levels of success. These methods and their results will be discussed in detail later. 1.1 Current Detection Methods Over the past several decades there have been vastly different methods employed in order to collect some data from Extended Air Showers, EAS from cosmic rays. Going back to Hess flying in his balloon all the way to the latest cosmic ray telescope, the Fermi Satellite, and in between with the current generation hybrid detectors, such as the Pierre Auger Observatory. The extreme breadth of cosmic ray energies 2

16 (over ten orders of magnitude) ensures that we must have several different detectors probing the different energies. Most detection schemes resolve down to two competing forms: direct surface based detectors and indirect, or diffuse, detection methods. The direct based involves placing some form of material in the direct path of an Extended Air Shower and watching for the Cherenkov radiation from the passage of the particles. The indirect methods (or method) involves using some form of photon counter looking for the ionization of the atmosphere from the passing particles Surface Detection Surface detection tends to be the simplest of detectors. It can be as simple as a piece of scintillator material, a photomultiplier tube and an oscilliscope or the Pierre Auger Observatory with 1600 tanks of water. The communications and underlying framework of PAO is extremely complex but the detection itself is simple: direct detection of shower particles via Cherenkov radiation in a tank of water and a few photomultiplier tubes. AGASA The Akeno Giant Air Shower Array (AGASA) covered an area of 100km 2 with 111 surface detectors and 27 muon detectors. The array was operated by the Institute for Cosmic Ray Research, University of Tokyo at the Akeno Observatory. The results from AGASA hinted at the existence of ultra high energy cosmic rays, cosmic rays with energies beyond ev. Along with the results from the Fly s Eye this experiment led directly to the modern Pierre Auger Observatory with its hybrid detection [6]. 3

17 KASCADE and KASCADE-Grande KASCADE is a German surface detector array started in 1996 at Forschungszentrum Karlsruhe, Germany tuned to detect showers in the energy range of ev and measured the electronic, muonic and hadronic components of the shower simultaneously. KASCADE-Grande was a further enhancement to the original array located on and around the Karsruhe campus and still takes data [7] Fluorescence Detection As a shower passes through the atmosphere, the charged particles can excite molecular nitrogen, N 2, which then emits a photon in UV as it de-excites. The High Resolution Fly s Eye or HiRes detector was an ultra-high energy detector that collected these UV photons using large mirrors and photomultiplier tubes, see Figure 1.2 for an illustration as to the name. HiRes ran in the Utah desert for almost 10 years from 1997 to The HiRes experiment made the first observation that there may be an energy at which the cosmic ray energy spectrum begins to cutoff [8]. This cut off is called the GZK effect and is the energy at which cosmic rays begin to interact with the photons left behind from the Big Bang and is discussed in detail in a later chapter Pierre Auger Observatory The Pierre Auger Observatory is a hybrid style cosmic ray observatory designed to detect ultra-high-energy cosmic rays or above an energy of approximately ev or the equivalent of a well served tennis ball packed into a single proton or nucleus. These high energy particles have an estimated arrival rate of just 1 per km 2 per century, therefore the Auger Observatory has created a huge detection area of around 4

18 3000 km 2 or approximately the size of Rhode Island in order to record a statistically significant number of these events. It is located on the Pampa Amarilla (yellow prairie) in western Argentina s Mendoza Province. The main campus is located is the small town of Malargüe. The observatory was named after the French physicist Pierre Victor Auger. The project was proposed by Jim Cronin and Alan Watson in Today, more than 200 physicists from 55 institutions around the world are collaborating to build the southern site. The 15 participating countries are sharing the $50 million construction budget, each providing a minor part of the total cost Radio Detection The recent efforts at the Pierre Auger Observatory are a collaborative effort of groups from France, Germany, Netherlands and the United States combining knowledge and experience from LOPES/LOFAR and the French radio site CODALEMA [10, 11]. The challenge at PAO is primarily in deployment and self-triggering. The deployment will necessarily require the stations to be self contained which means solar power and wireless communications which limits the bandwidth of measurements. This will be covered in great detail in a later chapter. 5

19 Figure 1.1: A cartoon schematic of the detection methods being used at the Pierre Auger Observatory in Malargüe, Argentina [5] 6

20 Figure 1.2: Using a grid of focusing mirrors to collect the light, cameras can view the air shower up to 15 kilometers away [9]. 7

21 Figure 1.3: The layout of 1600 surface detectors with 4 telescopes each with 6 Eyes surrounding the water tanks. The total area is 3000 km 2 or approximately the size of Rhode Island [3] 8

22 Figure 1.4: The installation of the first solar powered radio detector using radio communications located near the Balloon Launch Facility. 9

23 Figure 1.5: Auger Surface Detectors (SD) are all self contained systems. Solar power and radio communications allow a near 100% duty cycle. Each tank consists of 3000 gallons of purified water, 3 photo-multipler tubes (PMT).The system detects the passing of muons and electrons from cosmic ray showers due to the Cherenkov light produced from a particle passing. The Pierre Auger Observatory has approx 1600 active tanks layed out in a 1.5 kilometer hexagonal grid (see Fig. 1.4) [9] 10

24 Figure 1.6: ANITA Launch from the 2006 flight over Antarctica. By utilizing the rotational nature of the air patterns over the pole every summer, an experimentalist can launch a balloon from McMurdo and have it return to her in 10 days. [12] 11

25 Chapter 2 Basic Physics of Cosmic Rays 2.1 Sources of Cosmic Rays The energy spectrum of cosmic rays spans 10 orders of magnitude and follows a roughly uniform power law. Two features of note, seen in figure 2.1, are the knee at ev and the ankle around ev. A change in the spectral index, such as that seen at the knee and ankle region, suggests a changing of the source of flux as one mechanism overtakes another. The cause of these changes in spectral index has been the subject of much speculation. The flux of these regions varies dramatically from 1 particle per square meter per second at ev to 1 particle per square meter per year at the knee. This flux further decreases to 1 particle per square kilometer per year at the ankle. Finally, at energies in excess of ev a flux of less than 1 particle per square kilometer per century is found. The extremely low flux of Ultra High Energy Cosmic Rays (UHECR) presents a detection problem. A detector needs a large detection area, high duty cycle, and/or a long run time in order to build a sufficient statistical database for study [3]. 12

26 Figure 2.1: Cosmic Ray Spectrum with Experimental data from AGASA, Akeno, Fly s Eye, Haverah Park, Leap, Proton, Yakutsk experiments. An E 3 power-law is shown by the dotted line. Energy in ev is shown plotted against Flux in (m 2 steradian second GeV) 1.[3, 13] 13

27 2.2 Acceleration Mechanisms The commonly accepted acceleration model, postulated by Fermi, involves the interaction of a cosmic ray, a charged particle, with a magnetized shock front [14]. The magnetic field of the shock confines the particle, and each interaction with the shock front imparts energy to the particle. This continues until the Larmor radius of the particle exceeds that of the shock radius at which the particle is no longer contained by the shock s magnetic field. One possible source for such a shock is that of a supernova remnant. By comparing the observed energy density of cosmic rays to an estimated energy output from supernovae, it can be seen whether there is sufficient cause to consider supernovae as a candidate source. We know the local energy density is ρ 1 ev/cm 3. The cosmic rays are assumed to be confined to the galactic disk with a volume of V = πr 2 d, where R is the galactic radius, 15 kpc, and d is the thickness of the galactic plane, 300 pc, V cm 3. The confinement time of the cosmic rays in the galaxy has been measured via isotope analysis and is approximately τ R = yr. This yields a cosmic ray luminosity of L CR = ρ CRV τ R erg/sec (2.1) The supernovae luminosity is known to be approximately erg/sec. Thus, only a few percent of the total supernovae energy density would be required to explain the observed cosmic ray energy density [15]. To further evaluate the possibility of supernovae as the source of the cosmic rays, we note that the maximum energy attainable via the shock acceleration mechanism is [16] 14

28 E max Z B R 1µG 1kpc 1018 ev (2.2) where Z corresponds to the charge of the cosmic ray particle, B is magnetic field of the shock, and R is the size of the shock region. The magnetic field strength is plotted against the size of the shock region of the supernova and other possible sources in the plot in 2.2, named the Hillas plot [17]. A diagonal line corresponding to a given energy, ev, of the cosmic ray particle is also shown on the plot. Thus, it is easy to see that supernovae shock acceleration could account for the observed lower energy cosmic ray flux but would not be able to accelerate the particles to the energies above the ankle or E > ev. These Ultra High Energy Cosmic Rays (UHECR) can be shown to be of extra-galactic origin due to the fact that the observed galactic magnetic field is insufficient to constrain particles at UHE. 15

29 Figure 2.2: The physical size and magnetic field of several possible candidate sources of high energy cosmic rays. The diagonal lines correspond to energy of ev for proton (solid) and iron(dashed.) Objects below the line will not accelerate particles to ev [17]. 16

30 2.2.1 Energy Limitations There is another factor we must take into account in considering the intergalactic sources postulated on the Hillas plot. At sufficiently high energies, cosmic rays begin to interact with the Cosmic Microwave Background (CMB) photons [18]. With each interaction, this process, known as the GZK process after being independently reported by Greisen [19] and Zatsepin and Kuzmin [20], reduces the incident cosmic ray s energy. This would have two results: placing an upper limit on travel distance of the cosmic rays and a pile-up of cosmic rays with energies just below the GZK threshold energy. The dominant channel for this GZK process for protons is: p + γ CMB + p(n) + π 0 (π + ) (2.3) This process can be easily verified in a lab setting whereby a proton at rest is bombarded with photons. An approximate photon energy threshold of E T = 200 MeV and a cross section of approximately σ = 250µb is found using this technique [21, 22]. The UHECRs will begin to interact with the CMB photons when the photons reach this threshold in the rest frame of the UHECR. The CMB photons have an average energy of E CMB = ev which in the rest frame of the UHECR is approximately E = γ(1 + v c cosθ)e CMB γe CMB (2.4) Therefore, at threshold energy, E T, we can find our relativistic correction of γ E T E CMB = (2.5) 17

31 Figure 2.3: Detailing the energy loss of three different initial energies of proton as a function of distance traveled with the 2.7K cosmic microwave background photons[23, 24]. 18

32 The energy of the UHECR at this threshold energy is E T = γmc ev. When this calculation is integrated over the blackbody spectrum of the CMB rather than performed at a single temperature, we arrive at E = ev. This places the threshold energy right at the ankle of our cosmic ray spectrum. The GZK process would account for the steepening of spectral index at this energy [25]. Following a simple calculation from Longair, we can find an approximate GZK horizon for the UHECRs. The UHECRs will lose approximately 1/10 of their energy for each interaction with the CMB. The mean free path of this interaction is: λ = (N CMB σ) 1 where σ = 250µb = m 2 and N CMB = m 3. We arrive at λ = m or an interaction length of approximately 3 Mpc. For a particle to lose the entirety of its energy would therefore limit its travel distance to approximately 30 Mpc. According to GZK, if we are observing particles in the UHE regime they must be of local origin [25] Interactions in the Atmosphere Upon encountering the Earth s atmosphere, the UHECR particle will collide with an atmospheric molecule setting off a cascade of particle interactions resulting in what is known as an Extensive Air Shower. An EAS consists of three main components: the hadronic core, an electromagnetic portion, and a muonic component. The incident UHECR particle is known here as the primary particle and all other particles generated in the air shower are denoted as secondary particles, see figure 3. A proton or charged nucleon is the most likely candidate for the primary particle, but an EAS can also be generated by a high energy photon. These photon generated showers will have different observational qualities as they will consist almost primarily of an 19

33 electromagnetic portion [26]. At the highest energies, the photon induced shower comprises less than 10% of the total flux of incident showers [27]. Figure 2.4: Schematic detailing an Extended Air Shower development. The shower is divided into 3 catagories: the hadronic core which includes the surviving nucleons from the incident cosmic ray, the muonic cascade where the charged pions decay to muons, and the electromagnetic cascade fed by neutral pion decays[3]. 20

34 In considering hadronic initiated showers, the first interaction high in the atmosphere will likely be an inelastic collision with an atmospheric molecule, likely nitrogen. This produces large number high energy pions, kaons, and a high energy nucleon. The secondary particles then proceed to either interact further with the atmospheric molecules or decay. This hadronic core of the shower feeds the shower progression as photons are generated via decay of the pions. The muonic component arrives from the decay of lower energy pions and kaons. The high energy photons are generated mostly from the decay of neutral pions and go on to feed the electromagnetic portion via pair production and then bremsstrahlung. These interactions continue until the constituent particle energies are reduced below the interaction energy or the interaction length is greater than their relativistically adjusted decay times. Most of the hadronic secondaries undergo further interactions with approximately one third of the energy going into the electromagnetic component at each interaction. Therefore, an estimated 90% of the primary s energy will result in the EM portion of the shower. [3] To an observer, the air shower has a thin disk or pancake appearance as it traverses the atmosphere at nearly the speed of light. The direction of travel, or shower axis, is given by the incident direction of the primary particle. The number of particles in the pancake reaches a maximum, N max, after which the shower development slows as many of the secondary energies fall below the critical interaction energy. This N max occurs in the atmosphere at a location denoted as X max which is related to the primary particle s mass and energy. As the cross-section of the primary particle increases, the location of the first interaction occurs earlier in the atmosphere. For example, an iron nucleus will have an interaction with the atmosphere earlier than 21

35 a proton of the same energy. As the shower traverses the atmosphere, the passing secondary particles will excite the atmospheric nitrogen, causing it to fluoresce. This fluorescent light is detectable and is directly related to the number of particles present in the shower. This in turn relates to the shower development and to the energy of the shower. As the shower progresses, some secondary particles will scatter away from the shower axis via Coulomb scattering and transverse momentum yielded via decay products. This lateral development was characterized experimentally for both the muonic [28] and the electromagnetic portions of the shower [29, 30, 31]. By detecting these components on the surface at different radii to the shower core, the lateral development of the shower can be measured. By comparing this lateral development to various models and Monte Carlo simulations of showers, the primary energy can be estimated. 22

36 Chapter 3 OSU/Leeds Geosynchrotron Detector 3.1 Radio Free Malargue The OSU/Leeds radio detection effort consists of 3 stations co-located with the Pierre Auger Observatory (PAO) on the Pampa Amarilla near Malargue, Argentina. The OSU/Leeds detector is part of an wider effort to enhance the PAO with a radio detection system. Teams from The Netherlands, France and Germany are also taking part in this collaborative effort. Each OSU/Leeds station consists of a log-periodic dipole antenna, electronics package, wireless communications package, and solar power with batteries and a charge controller. The overall layout of the system is shown in Figure 3.1. Note that the Pierre Auger Observatory has also installed a Surface Detector tank named Olaia near our antenna array to aid in verification of a radio detection. The Germans and Dutch utilized poles 1, 2, 3 in Figure 3.1 which are connected via cables to digitizers and oscilliscopes inside the Balloon Launch Facility. The Dutch also employed scintillator paddles at the Balloon Launch Facility and used them as triggering devices for the radio array [32]. 23

37 Figure 3.1: Overhead Google Earth view of the layout of the poles for the LPDA antennas. Poles 4,5,6 are laid out with an approximate 300 m baseline intended for the inital testing of the OSU/Leeds detectors with solar power and wireless communications. Poles 1,2,3 are laid out with an approximate 100 m baseline are were used by German and Dutch teams and are wired for signal transmission with all processing done in the Balloon Launch Facility (also pictured) [33]. 24

38 3.2 Log Periodic Dipole Antennas The Log Periodic Dipole Array (LPDA) antenna is a directional antenna covering a wide frequency range along with a reasonably constant antenna response. Originally designed in the 1950s by researches at the University of Illinois, the LPDA can be designed to cover a very broad range of frequencies. The name derives from the property that a repeating periodic antenna response (gain, impedance, etc) can be seen going with the log of the frequency, see Figure 3.4. These flucuations can be minimized over one period and thereby extend to an overall consistent antenna response. An octave is defined as the frequency harmonics associated by powers of 2, i.e. 15MHz to 30 MHz as a single octave and 15 MHz to MHz= 120 MHz as three octaves. While designs of up to 3 octaves have been achieved, most designs span a single octave (i.e. 40 Mhz to 80 MHz). The design for our purpose fits into the latter catagory with just over one octave covered (40 85 MHz). A high degree of directivity is another characteristic of the LPDA design. 25

39 Figure 3.2: Log Periodic Dipole Arrary(LPDA) Antenna. The geometry is defined by a constant ratio τ where successive elements retain this proportionality [34] Figure 3.3: LPDA schematic illustrating the phase reversal added at each antenna arm. This is intended to minimize interference created by the smaller arms radiating for large λ s [34] 26

40 The basic design characteristics of an LPDA antenna are shown in Figure 3.2. The individual arms of the antenna can be thought of as simply dipole antennas connected via a central waveguide also called a phase-line. As such, the frequency range of the antenna is determined by the length of the arms. The longest arms of the LPDA are designed to be equal to or longer than 1λ 2 low where λ low is defined as the wavelength of lowest end of frequency range. Similarly the shortest arms are equal to 1λ 2 high where λ high corresponds to the wavelength of the upper extreme of the desired frequency range [34]. The three basic design parameters of an LPDA antenna are: α, σ, and τ. The geometric ratio τ is defined as the ratio of consecutive spacings, consecutive rods lengths and in ideal situations consecutive rod diameters. τ = R n+1 R n = l n+1 l n = D n+1 D n (3.1) The angle 2α can be easily seen in Figure 3.2 as the opening angle created by extending a line through the ends of the dipole arms of the array. The final parameter σ or relative spacing parameter is defined as: σ = D n 2l n (3.2) Once two of the parameters are known the third can be found using the geometric relationship σ = 1 τ 4 tanα (3.3) From Figure 3.3, there is a crisscross design to the feed line of the array and plays a crucial role in the performance of an LPDA. The current and phase between 27

41 two closely spaced adjacent arms are very similar when λ/2 is large relative to the length (l n ) of an arm. This crisscrossing adds a 180 phase at the terminal of each element. This causes very little energy to radiate from them and interference is kept to a minimum [34]. This effect allows us to treat this region as a transmission line for larger λ. The first active region of the antenna for a given wavelength occurs when the dipole lengths are approaching one-half of a wavelength. The phase reversal of the elements and feeding the antenna from the small end leads to a phase progression towards the small end of the antenna. The energy will radiate end-fire in the direction from the small end of the antenna and a high degree of directionality is achieved. The name of the antenna derives from the repetitive nature of several parameters. For example, when plotted against frequency the input impedance will be repetitive, but if plotted against the logarithm of the frequency that repetitive nature is revealed as periodic, seen in Figure 3.4. Parameters showing similar log-periodic behavior are pattern, directivity, beamwidth, and sideband levels. The span of each cycle is determined from the geometric ratio, τ, defined by equation

42 Figure 3.4: A hypothetical example of the periodicity of input impedance against the log of the frequency [34]. Figure 3.5: Computed contours of constant gain versus τ and σ for an LPDA. Originally computed by R.L. Carrel at Univserity of Illinois in 1961 and later found to need a slight modification and reduced by an average of 1dB. [34, 35, 36]. 29

43 3.2.1 OSU/Leeds LPDA Deployment The OSU/Leeds geosynchrotron radio detector utilizes 3 stations spaced approximately 300m apart. Each station consists of an LPDA for the N-S polarization and one for E-W polarization. The LPDA antennas were designed by collaborators from Aachen University [37]. As noted previously, LPDA antenna are wideband antenna with a high degree of directionality. These properties are confirmed from field testing performed at Karlruhe Institute of Technology (KIT). The E-plane diagram, Figure 3.7, shows a wide main lobe with a 3dB beamwidth of 100 with an average gain of 4dBi. The dramatic sidelobe attenuation reduces the effects of man-made radio frequency interference (RFI) and the negligible backlobe minimizes variable surface conditions from our calibration process. With a 50Ω system the return loss, shown by the Smith Chart in Figure 3.6, is nearly a constant -15dB or less [38] 30

44 Figure 3.6: The Smith chart of the LPDA antenna deployed in Malargüe. The return loss in a 50Ω system over 40 80MHz rotates near the matching point in the center of the diagram [38]. 31

45 Doing a short back of the envelope calculation using Fig. 3.9 as a reference for the Aachen LPDA design we can sanity check the reported antenna gain/directionality. Arm # l (leutholds) τ Table 3.1: A quick measurement of arm lengths using a Leuthold as a reference length. From Table 3.1 we can infer much about the Aachen design. The shortest arm of the LPDA (arm 1) corresponds to λ high 2 = l leut 1leuthold 1.89m = 1.335m (3.4) and a frequency of approximately 112 MHz. Similarly, on the other end of the array we have λ low = 4.56 leutholds or 8.62 m and a lower frequency limit of 35 MHz. Both of these frequencies agree well with the both the quoted specs and the lab measured performance. The τ ratio for the antenna is approximately 0.85 ± With a bit of geometry α can be found to be approximately rad. Having two of the constants yields the third using 3.3, σ Referring to Figure 3.5, the gain found from the LPDA contour plot from the approximated LPDA parameters yields G = 7.5 8dB. This is slightly higher than the actual forward looking gain of the antennas of 6 db [38] 32

46 Figure 3.7: The E-plane directionality plot for an LPDA antenna at 2.5m above ground. Multiple frequencies are shown from 30 80MHz [38]. 33

47 Figure 3.8: The E-plane directionality plot for an LPDA antenna at 2.5m above ground. Multiple surface conditions are shown for a given frequency. Little differentiation is seen among the surfaces [38]. 34

48 Figure 3.9:. Matthias Leuthold of Aachen University installs the LPDA on-site at the Balloon Launch Facility. Each antenna consists of a N-S polarization and an E-W polarization. 35

49 The tubing used in our LPDA is of constant diameter. Ideally, the ratio of tube (or wire) diameters are determined by the geometric constant of the array, τ. Due to the difficulty in maintaining tolerances across multiple tube diameters and finding/machining proper brass fittings for the attaching the arms to the framing, this constraint is ignored. This factor is of little effect and is not worth the cost for such a minor degradation to the overall performance [34]. 3.3 Amplifier/Receiver The Cosmic Ray Radio Unit (CRRU) receiver/amplifier, shown in figure 5, was designed by collaborators at Leeds University and is the only piece of new custom electronics used in this experiment. Designed with the view of rapid deployment and enabling a proof of concept system, the receiver is designed to utilize the Unified Board used in the Surface Detectors at PAO. The receiver has two inputs with one for each polarization and outputs 3 channels per polarization for a total of 6 channels out that lead to the Unified Board. Each channel spans a 15 Mhz frequency range and is configurable along a sliding range, see Figure 3.11, for a total of 45 MHz in the range of MHz. The amplifier has a range of approximately 75 db with a noise figure of 5.5 db at maximum gain. At maximum gain, the output is a 2 V pk-pk measurement of a -100 db signal. Both the frequency ranges and amplifier power for each individual channel are software controlled via an RS-232 port. A phantom feed of 12 V DC is available to allow the use of active antennas but was not used in this testing [39]. 36

50 Figure 3.10:. The Leeds designed Cosmic Ray Radio Unit (CRRU) receiver and amplifier. Note the 2 inputs located of the left side of the face of the device and the 6 outputs on the right. Each polarization of the LPDA antenna is fed into the device and 3 outputs per polarization are fed out to the Unified Boards [39]. 37

51 Figure 3.11:. Each channel from the Leeds receiver span 15MHz of bandwidth from 13 85Mhz. The settings are software controlled and can be one continuous band or split into inidividual bands. R-CDAS can relay new settings to the amplifier/receiver units via the radio communications [39]. 38

52 The receivers are effectively doubled up allowing a simultaneous measurement of both the E-W and N-S polarizations. Wideband receivers and amplifliers present a design and price challenge and efforts were taken to minimize these issues. The wideband, almost 3 octave range, of measurement available (15-85 MHz) is handled by splitting each polarization channel into 3 sub-bands of 15 MHz each. An intermediate frequency (IF) of 326 MHz is used and each filtering sub-band is performed using a surface acoustic wave (SAW) filter. Amplification occurs at the IF where bandwidth ratio will be minimized and reliable amplifiers are available for an affordable price. The final signals are then mixed down to digitization signals of 2-17MHz in each subband. The finite roll-off of non-ideal filters and unavoidable mixing products give an unusable band from 0-2MHz and 17-20MHz in each sub-band. The 0-20MHz range was chosen in order to allow a 40MS/s analog-digital-converter (ADC) to properly sample the antenna signals. A 40MS/s ADC yields a Nyquist frequency, the maximum sampling rae at which unambiguous signal digitization is possible, of 20 MHz. As such, the CRRU output signal of 2-17MHz fits right into the Unified Board digitizer s sweet spot. 39

53 Figure 3.12:. An entire station schematic for a self-contained wireless radio detection system [39]. 40

54 Figure 3.13:. Full schematic diagram of a CRRU receiver system [39]. 41

55 from: To calculate the minimum signal that is detectable by the receiver can be derived P mds = Bk B T(F 1) (3.5) B is the sytem noise bandwidth (15Mhz), k B is Boltzman s constant ( JK 1 ), the receiver temperature T in K which is taken to be +50 C or 323 K, and receiver noise factor F derives from the previosuly stated noise figure at maximum gain of 5.5 db, F = log 10 (5.5/10) = This yields a minimum detectable signal level of 1.7x10 23 W or 98 dbm. We can convert this minimum detectable signal to field strength, E, using the standard formula P = E2 c 2 480π 2 f 2G (3.6) The lab measured gain performance shows a typical shape to the gain curve. In Figure 3.14 we see that the amplifier has a large range of log-linear gain. Near both extremes the gain curve turns over and we lose this very important feature. 3.4 Onboard Electronics Unified Boards (UB) from the Auger Surface Detector (SD) are used as the onboard electronics for each detection site. Using pre-existing electronics allows us to save time by modifying the existing firmware to our needs as a radio detection system over the costly route of custom hardware/software design. These electronics serve as primary digitizers, triggers and communications hub for each individual detection site. 42

56 Figure 3.14:. A sample gain plot of one of the six amplifier channels. A partial plot is shown here corresponding to the upper half of the radio slider from Figure 3.11 of channel 1a from amplifier 4. The important features to note are the log-linear nature of the amplifier up to 80dB at which point the plot turns over and we lose linearity (or log-linearity in this case). 43

57 Figure 3.15:. A standard PAO Surface Detector Unified Board was used in each antenna station. Some parts of the embedded controller software were re-written and custom trigger logic flashed on the Field Programable Gate Arrays (FPGAs) in order to utilize this SD hardware for our radio task. 44

58 The boards are outfitted with 6 channels of digitized input with 3 channels allocated for each polarization. The analog signals are converted to digital using 12-bit 40 MSa/s analog-to-digital convertors (ADC). Each ADC has a 2V range and a standard error of 0.5LSB. The triggering is done using a Field Programmable Gate Array or FPGA which can be flashed with custom logic for the triggering requirements.the UB is controlled via an IBM PowerPC CPU operating at 40MHz. Each board has 32MB of onboard RAM for event storage. Microwares OS-9000 real-time operating system is stored on a 2 MB Flash memory along with the acquisition systems custom control software. Timing of events is accomplished by a combination of 40 MHz and 100 Mhz onboard clocks along with a Motorola GPS card providing a 1 pulse-per-second (1pps) at a quoted accuracy of <50ns. A counter of clock cycles for the 40 and 100 MHz clocks are stored in a buffer to be read out at the time of a trigger. Each trigger is tagged with two time-stamps, the 100 MHz cycle counter for the rising and falling edge of the trigger interrupt, plus the GPS second from the 1pps. The total number of cycles of the two clocks for the previous second are also stored for further timing corrections. 3.5 Solar Power Power was provided via Isofoton 150W 24V solar panels shown in Figure A SunSaver charge controller was used to ensure one way flow of electricity else the batteries would try to charge the panel at night. Standard 12V deep-cycle batteries were connected in series to obtain a 24V configuration. Used batteries were initally provided from the PAO engineering array but 45

59 proved problematic. Later, locally purchased deep-cycle sloar batteries were used. Battery performance will be discussed in detail later. 46

60 Figure 3.16:. Isofoton 24V 150W solar panel mounted on uni-strut framing during inital field tests. The NEMA 4X electronics crate can be seen as well as the batteries. The radio modem was tested to a range of 750m with little trouble well beyond the line of sight range of test site at BLF. 47

61 3.6 Communications Radio communication with the individual detectors is accomplished using commercial radio modems from MaxStream Technologies with the XStream 9600 model. These low-power low-cost commercial off-the-shelf units operate at 9600 bits per second at 2.4 GHz with a line of sight range of up to 20 miles using a high-gain antenna or 7 miles using a small dipole whip. With built-in transmission error detection and frequency hopping channels, these modems allow for fast field deployment with minimal overhead. Each station was outfitted with a single XStream and small 3dB dipole antenna. Power consumption of 3W was added to the total power budget of each station. A similar packet structure to the Auger Surface Detector radio communication packet structure is used to and maximize the amount of legacy software on the UB. This provides for a bit packet (see Figure 3.17)to be sent once per second to the Radio Central Data Acquisition System (R-CDAS) housed nearby. Much of the packet is dominated in transmitting the RT2 timestamps to R-CDAS. The RT2 timestamps have top priority in message payloads. An RT3 trace is broken into smaller pieces and transmitted in any gaps left in the payload after the RT2 s are inserted. As there is only one packet per second, an RT3 trace can take several second or even minutes if the RT2 trigger rate is too high. 48

62 Figure 3.17:. Up to 1200 bits are sent in each packet from the remote stations to the Radio Central Data Aquisition System (R-CDAS). Each packet can be divided into multiple messages if enough room is available. Basic structure based on that of the surface array [40]. Figure 3.18:. Interior structure of the message payload that is transmitted within message packet. Multiple payloads may be transmitted in each packet. RT2 timestamps being the most common but RT3 ADC traces, UB status messages, radio communication checks, background ADC traces and others are used to pad each message to fullest [40] 49

63 3.7 Radio Central Data Aquisition System Housed at the Atmospheric Balloon Launch Facility the Radio Central Data Acquisition system consists of three radio modems, a spy radio for the larger array, uninterupted power supplies (UPS), a linux box with ample on-board storage and finally a WLAN link to the nearest Flourescence detector at Cohuico. The R-CDAS system consisted of a standard Ubuntu Linux box with 4 RS-232 ports, multiple radio modems, a PAO Surface Detector communication package with a standard SD tank Yagi antenna, a separate W-LAN radio modem with high gain antenna for a link to PAO campus and an uninterrupted power supply (UPS) for battery backup during power glitches. The core of the R-CDAS system is the custom software running on the linux box. A multithreaded application which monitors all radio communications independently and decodes/encodes message traffic. The software watches for overlapping radio triggers and requests traces from individual stations in one occurs. It also monitors PAO communications for the sector using the PAO SD radio to spy on the radio traffic for all surrounding water tanks. This allows the system to request a trace of the radio detectors for the same given triggered events. The packet structure is such that no single packet contains an entire trace to itself. The local station software will pack portions of the trace into the 1Hz message packets where space allows. The software stores the completed traces in the native binary format in which it was transmitted locally on a spacious harddrive. The user can monitor the incoming messages, data rates, trigger rates, and events from the FastLook MATLAB software written for R-CDAS, see Figure Nightly data backups are performed via an rsync to the central PAO campus. 50

64 Figure 3.19: The Balloon Launch Facility (BLF) played host to the German, Dutch and OSU/Leeds radio efforts while also performing the routine weather monitoring. The converted sea container housed the R-CDAS system while communication to stations was accomplished with small whip antennas on the roof of the container and a W-LAN link maintained via a high gain antenna aimed at the nearest FD building (Cohueico). 51

65 Each individual station has a dedicated XStream radio modem committed to it. While it is possible to utilize a point-to-multipoint transmission scheme with these radio modems it is not an optimal solution for this system. More details on this matter are discussed in a later Section. 52

66 Figure 3.20: An example of the FastLook capability of the R-CDAS. A nearly realtime display of captured events is possibly while working from the base station housed inside the BLF. The top two rows correspond to the direct ADC traces of each channel, N-S polarization row 1 and E-W polarization is row 2. The bottom 2 rows are the FFT s corresponding to the top two rows. The columns are aligned according to channel. Column Mhz, Column MHz, Column MHz. 53

67 Radio communications to PAO central campus is accomplished using a standard Commercial Off The Shelf (COTS) wireless router hooked to a high gain directional antenna to the nearest fluorescence detector building, Coihueco. At Coihueco, the data is streamed to central campus via the normal PAO radio communication towers. A UPS is a necessity when working on the Pampas, as we did. The availability of AC power at the BLF was a primary reason the site was chosen. The reliability of the AC power was rather limited. Daily outages are expected and any system stationed there needs to be robust enough to suffer repeated and extended power outages. The UPS used in this case was less than ideal but the price was great (free). The UPS provided enough power for the software to close down all aquisitions and await either a resumption of power or eventual failure of the battery in the UPS. The SD spy station is an attempt to seek out a confirmed radio event by using the surrounding SD tanks as truth. Listening for trace requests from the actual PAO CDAS the local Radio CDAS would send similar requests to the local antenna stations asking for traces along the same timestamp. 3.8 Trigger Logic The trigger logic is divided into the fast hardware triggers and a slower software trigger. The hardware trigger is by necessity a fast trigger and is usually as simple as a logical combination of ADC thresholds. The slower software trigger can use some clock cycles to impose a simple shaping cut, a 3σ cut or other similar cuts. The slowest trigger is that which is done in R-CDAS itself. This final trigger compares timestamps from nearby antennas (or SD tanks) looking for overlapping lower level triggered events. 54

68 Figure 3.21: The high gain antenna for the Wireless LAN and the Yagi antenna for the SD spy station mounted to the side of the BLF container. 55

69 3.8.1 Hardware Triggers As stated above, the hardware level trigger resides in the onboard electronics at each antenna installation. We have re-purposed the electronics package from the PAO surface detectors. There are two triggers possible at each antenna. Radio Trigger Level 1 (RT1) consist of simple boolean combinations of hardware threshold triggers of the ADCs. If the signal activates a given trheshold in a single channel of its 10bit ADC a latch is enabled at the hardware level. If within a user-designated window of cycles some user-designated logical combination of channels is noted then an RT1 is noted in the embedded control software. A Radio Trigger Level 2 is a fast software trigger at the embedded controller level. An RT2 is achieved if the traces from an RT1 follow a simple shaping guide line which is also user-determined. Once an RT2 is noted, the trace is stored onboard with an 10ns accurate timestamp along with a microsecond level timestamp which is used in the data packet transmitted to R-CDAS Software Trigger Once the radio packet arrives to the R-CDAS a comparison with timestamps of RT2s from the other stations is done. If an overlap is observed between 2 or 3 antenna a Radio Trigger Level 3 (RT3) is sent. The RT3 is transmitted abck to the UBs at the antenna and a request for any traces matching the timestamp. The traces are then sent over numerous data packets (dependent upon trigger rates) back to R-CDAS for storage and later transmission via the WLAN to central campus. 56

70 3.9 Results The initial results were not entirely encouraging. Shipping problems, power problems, hardware trigger glitches, communication problems and other hardware and software issues limited the total data runs. The radio frequency interference in a radio quiet area such as the Pampa was surprising and only enhanced the need for a proper trigger Issues, Problems, Success While not entirely successful, the OSU/Leeds project gave valuable experience with the deployment of radio detectors and unique insights to the problems the Pampa provides. This knowledge has been put to use in furthering the PAO Radio Collaboration efforts. The next step in the process, MAXIMA, is a combination of technologies used by the French, Dutch, German and the US/Leeds groups. MAX- IMA is currently taking data with 4 stations with the intent of moving to 10 in the near future. AERA, the Auger Engineering Radio Array, has the intent of deploying a dense array of 150 radio antennas over a 20km 2 area co-located with other R&D projects, HEAT and AMIGA, over the next two years [41, 42, 43, 44]. Shipment Shipment of our initial gear presented a tremendous delay in our deployment. The shipment was lost for 3 weeks before arriving during the fourth week of the first on-site service trip to Argentina. The gear arrived on our last day on site after we had left for the airport. The initial installation took place on the following trip six 57

71 weeks later. The moral of the story here is not to travel to Argentina until you gear is confirmed onsite. Antenna Issues The LPDA antennas designed in Aachen turned out to be less then ideal for deployment on the Pampa. During the frequent high winds of area the lower longer arms of the antenna would hit a resonant pitch and snap off the mast at the point of attachment. To replace an antenna arm solo the antenna must remain in place on the mast as it is not possible to remove then remount the antenna alone. In this case, the technician must climb a ladder to reach the antenna, which are a minimum of 3m above the sand to protect the local gauchos. So while leaning the ladder against the round pole and resting the ladder in the soft sand of the Pampa the technician must gently apply an 8lb sledge and chisel to knock loose the portion of the arm still lodged inside the mast. Then with the mast cleared, the new 2m λ/4 length aluminum pole can be installed. On the surface when installing these arms, a rubber mallet is used to push the aluminum pole into the mast where it is attached via bolts. This presents a bit of a problem while the mast is still mounted to the pole and 3m up. I will leave that as an exercise for the reader to deduce a solution. 58

72 Figure 3.22:. An LPDA Antenna missing an arm after a particularly strong windstorm. This problem was mostly solved by attaching rubber tubing visible as the verticle attachments between arms. While breakages still occured they were greatly reduced from the rate of 1/day to 1/month. 59

73 Radio Modem Problems The commercial radio modem from MaxStream seemed to be an ideal fit for our purposes in Argentina. These low-cost (< $200) modems perform the tedious issues of handshaking and delivery confirmation for transmitted data packets. The modems also guarantee packet delivery and retransmit the data packet packet if a collision occurs. A packet collision occurs when two or more stations attempt to transmit at the same time. When a collision occurs, each station will wait a random number of user specified microseconds before retransmitting and hopefully get a clear transmission window. This relieves R-CDAS from having to incorporate and debug these issues. However, problems presented themselves after installing the modems onsite and attempting to use mutliple modems sending communication packets to the base staion at R-CDAS. While it is true that the radio modems prevent packet collisions what the manufacturer does not tell you is that the modems will slice your data (which already in a fully formed packet) into packets prior to transmission. It is possible to set a maximum packet size but no minimum requirements exist. The sets up a problem of interleaving of messages where each individual packet is received from the multiple stations but the end receiver is unable to identify which packet is which and the communications are garbled. R-CDAS Power Interruptions Power supply problems were not strictly encountered at the remote stations. The Balloon Launch Facility has an external power supply which made it an ideal site for our testing. This supply is not especially reliable, however. A quality Uninterupted Power Supply (UPS) was needed for the R-CDAS equipment inside the BLF. Initially 60

74 the system was not able to restart itself after a hard crash of the pc. This problem was later solved by enabling the BIOS to restart after power failure and a cronjob to detect if the aquisition system is operational. Battery Problems Reliable battery power was a consistent problem all through our testing phases. Initially we were given 6 12V batteries that had been retired from the PAO Engineering Array(EA). As such, their reliability varied dramatically from battery to battery. During the inital testing of electronics it was found that when the batteries approached fully discharged they would spike (as voltage dropped away the current would max out) and blow a fuse on the UBs. After replacing all of the fuses on the boards with higher current fuses this no longer was a problem. Of the original six EA batteries we were given, two were kept for use at the antenna. The other 4 batteries, later to become 8, were purchased locally and were advertised as solar power deep cycle batteries. Reliability problems remained and can be seen in Figure A voltage report can be obtained from the UB by monitoring the regular status messages relayed from each station to R-CDAS. An example of a consistent battery failure is displayed in Figure 3.23, a snow storm covered the solar panels around day 6. From day 6 through day 10, a steady downward trend of voltages are reported from the station. These days corresponded to overcast wintry days. On days 11 and 12, a more typical daily power cycling of the battery can be seen. The voltages drop throughout the night and are replenished via the solar panels during the day. With a power budget of 24W, the station is walking a fine line with available solar power. Even the good batteries would require recharging after being fully discharged 61

75 Figure 3.23: This example displays the voltage read out at the power input of the Unified Board in mv. Follows the progression to battery failure during winter. Battery fails on Day 5 due to extended period of overcast skies and is returned to campus for a fast recharge. Days 7-10 show the slow decline of voltage during a snow storm. Days 11 and 12 are more typical of full diurnal power cycle with blue skies all day. 62

76 during an extended snow storm. This required returning the batteries to the main PAO campus and using a fast recharge system in the SD assembly room. Over a period of repeated snow storms all 12 batteries needed this treatment. Moving 12 lead acid batteries across 100m from the station to the access road became a dreaded routine during an especially snowy period in July. Hardware Trigger Problems The first deployment of complete stations was marred by an error in the flash programmable gate array (FPGA) which controls the hardware trigger logic. It was found that if channel 2a triggered for any reason that would prompt a full RT1 response no matter the logical combination that may be set. This effectively shut down data taking until a work-around was found. The simplest work-around possible was simply disengaging channel 2a from the antenna. This problem was later remedied by David Nitz at Michigan Tech with a new flash for the FPGA. 63

77 Figure 3.24: An example of the trigger logic multiplicity bug. The logic should force a 3-fold multiplicity for this trigger setting. Instead the station triggers from a single channel crossing the threshold (red dashed line) Figure 3.25: An example of the extremely high RT1 rates while the trigger logic multiplicity bug existed. 64

78 Figure 3.26: A similar look at the RT1 rates after the trigger logic was updated. 65

79 A problem related to the batteries required constant checking of station status. After a power failure, the UB returns to a basic state when power resumes. The R- CDAS must constantly be alert for battery failures and instruct the UB to set a basic trigger state and renew data collection operations. The early versions of the R-CDAS software had no option for loading a specified set of trigger logic to the local stations. This had to accomplished by hand after every battery failure. This problem was fixed but while it lasted little to no reliability could be counted on when an operator was not on-site. A greater problem arose from the large variation of trigger rates due to external sources. For the SD, a constant 20Hz T2 rate, corresponding to a known flux of muons, is desired and voltages on the PMTs are adjusted accordingly to yield this result. The radio trigger rates have no known baseline for calibration, yet. Trigger rates were seen to vary from 3 Hz to 90 Hz back to 5 Hz in the space of a few seconds. Above 40 Hz T2 rates the communication system breaks down as the excess triggers cannot fit into a 1200 bit message and a backlog of triggers begins to mount up. Excessive trigger rates cause traces to be lost as onboard storage for triggered traces is designed to exist in a rotating buffer for up to 10 seconds. The storage gets swamped if the trigger is not calibrated properly. Should we adjust the trigger logic every second or simply check periodically throughout the day. There seemed to be great variation between stations using identical trigger and receiver settings, see Figures 3.27 and In trying to maximize trigger rates hand tuning was required in order to mow the lawn of the noise floor of the receivers. Some of this will be discussed in the RFI section later. 66

80 Figure 3.27: The rapid change of triggers seem to be dependent upon the stations themselves. Here is a track of Radio Trigger level 2 (RT2) over a set period of time. Figure 3.28: Here is another radio station with identical gain and receiver settings. 67

81 RFI Problems The expected radio quiet area was not quite as quiet as was hoped. The power lines into the Balloon Launch Facility and passing cars and trucks were a large portion of false triggers. For example, the daily 9am bus to Malargue could be seen passing along the nearby road, or rather the spark plugs in the bus can be seen. In Figure 3.20, other interference could be seen from the stations electronics itself. In the first column of the bottom two rows are the Fast Fourier Transforms (FFTs) of the first channel (40-55 MHz) of each polarization after the amplification and mixing have shifted the signals to the range of 2-17 MHz. The large spike in each plot at 2 Mhz corresponds to a large 40 MHz signal at the antenna. The electronics package at the base of each antenna features six 40MS/s ADCs and a 40MHz PowerPC chip. All of these completed overwhelmed any signal in channel 1 of either polarization. Digital filtering or a bandpass filter can remove this RFI from the data Analysis Methods The initial data were studied for overlap with the events recorded by the surrounding tanks during post processing. The spy station was not installed until the third data run. Combined with the inability to run three stations and the rapidly varying trigger rates most if not all of the inital 50k events were from RFI. By only having 2 stations operational a true direction can not be determined merely an overlap in timing with the larger SD array. In Figure 3.30, the recorded trace occured within the same microsecond as a locally triggered SD event. This leaves it as a possible dection but further analysis led to a less than ideal timing match as it was off by 10s of ns from the expected time stamp. 68

82 Figure 3.29: Radio Background of the Balloon Launch Facility. Figure displays the mean value of data for a three day run on the LPDA in October, The feature at 40 MHz is the LPDA antenna turning on as the longest arms of the antenna correspond to a λ/2 of 40 MHz. The spikes below 20 MHz are nearby AM radio stations and similar spikes are seen in the 88 MHz FM band radio [37]. 69

83 A comparison of Monte Carlo simulated events and an end-to-end MATLAB simulation of electronics gives an example of a correctly identified, noise-free event. REAS Monte Carlos Tim Huege has modified the Monte Carlo simulation CORSIKA which simulates the passage of particles in an extended airshower to obtain a valuable radio-centric Monte Carlo simulation by which we can get a basic order of magnitude idea of the signal strengths and sanity check our results [45, 46]. MATLAB Simulations and Data Analysis Most if not all of the analysis code is written for a MATLAB environment. A simple GUI is available for the FastLook semi-realtime access to the data at R-CDAS. Several undergraduate students have written assorted pieces of an end-to-end MAT- LAB simulation which allows us to take a REAS generated shower and look at the predicted results Suggestions for Improvement While the overall results were discouraging, the valuable experience gained from the problems we encountered will ease future radio deployment at PAO. Trigger Logic The trigger logic problems have been chronicled elsewhere but it is worth repeating. Self-triggering, even in a relatively radio quiet region, is difficult but it is key to this effort. All of the PAO-Radio collaboration teams encountered similar problems. With the rapidly fluctuating RFI levels and numerous unidentified short burst transients, a quality trigger becomes crucial. Future systems will require faster digitizers, 70

84 more memory for storage of traces and faster processors to allow for a digital signal shape analysis. Batteries In light of the problems we encountered with both the aged PAO Engineering Array batteries and the locally purchased solar batteries, it is worth the added cost of shipping the heavy lead-acid batteries to Malargüe from the states if a solution could not be found in Buenos Aires. Power The current deployment with 150W Isofoton solar panels should be expanded to 200W versions or possibly dual solar panels. A safety factor of two should be added to any power budget calculation. Harsh winter storms can deplete even the best of batteries and we need an excess of solar power in the available power when panels may be partially covered by snow. Station Deployment Any framing system should be securely weighted or otherwise secured to prevent wind damage, see Figure Our stations were weighted by 40lbs plus weight of the framing and panels yet still toppled over during some extended downtime. A simple cross bar on the base allowing the batteriy box to act as ballast weight would be sufficient. More Manpower A larger involvement with other institutions closer to home would be valuable. The creation of an entire system solo left some problems that might otherwise been 71

85 caught in the lab or local field testing slip through the cracks until stumbling upon them while on the Pampa. Initial installation would be aided by a larger presence as well as the numerous battery recharging trips to and from campus. LPDA Antenna The rigid aluminum antenna is too fragile for sustained deployment on the Pampa. The high winds can easily snap off the longer arms and need regular maintenance inspection. A simpler design, such as the wire LPDA shown in Figure 3.34, will be essential for long-term deployment Contributions When Ohio State became involved in this research, much of the hardware (and design work associated with it) was provided by other collaborators. The Leeds group provided us with four working CRRR (Cosmic Ray Radio Receiver) amplifier/receivers. We had a few spare Unified Boards around the lab from the Auger design work during the Engineering Array. A former post-doc with the group Matthias Leuthold had left for Aachen University where they were designing the LPDA antennas to be used by the German radio group. The Aachen group granted us the use of the LPDAs when not in use by the Germans. Treating this as largely a COTS (Commercial Off The Shelf) project we used the hardware provided as if it were a commercial product. We rewrote a good portion of the embedded software on the UB controller boards to repurpose them for our use. We tested the CRRRs in the lab for gain profiles and any glitching in the response profiles. We had been provided a pre-compiled software driver for the Leeds CR- RRs so we reverse engineered the communication protocals from bytecode and wrote 72

86 some new controller code to go onto the embedded UB software. This controller code could in principle adjust the filter, gain and trigger settings of itself based upon the received noise over the last second, minute or hour. We wrote the Radio Central Data Aquisition System (R-CDAS) as a multi-threaded system capable of coordinating communications with the individual stations, adjusting gain and filters remotely, triggering a local radio event by processing the Auger Surface Detector radio chatter or as a radio only event from trigger timing overlaps, storage of data and remote synchronizing with central Auger campus, recovering from power outages both internal to R-CDAS and with the individual stations, and various other maintenance tasks. It was possible to adjust the filters and gain settings remotely from the main Auger campus through the radio link to the R-CDAS software and out to the individual stations. We wrote several software tools for testing the individual UB stations while connected via RS-232. We were responsible for the overall system design and purchasing of equipment necessary to complete the task: solar panels, charge controllers, uni-strut framing, tools, NEMA 4X containers, and various other smaller parts. We were responsible for the assembly and testing of the individual stations in Ohio and onsite in Malargüe. This led to a roughly six week rotation of travel to Argentina for approximately 2 years. While onsite, the day to day maintenance, testing and data runs of the equipment took the majority of time. We, also, wrote the MAT- LAB source code involved with the data processing and analysis for onsite work, at R-CDAS, or back home. We were constantly looking for a better trigger method to try for a few days or weeks. The declining performance of the batteries meant many round trips to and from the site while recharging on campus. Maintenance of the LPDAs also fell to we were usually the only people on-site. 73

87 3.9.5 Results While the overall results from this particular design were discouraging. The experience and knowledge gained from deploying a wireless, solar powered standalone cosmic ray detector are extremely valuable as the PAO Radio R&D effort continues. The Radio group has combined the working knowledge of all the participating groups into a detector design for the next pahse of deployment. The MAXIMA, Multi Antenna experiement In Malargüe Argentina, [44] effort deployed 10 antennas to the BLS area starting 2008 and is still running, see Figure

88 Figure 3.30: A raw ADC trace of one polarization of one pole of the first possible candidate event. The event overlapped a Surface Detector recorded event on the same microsecond. Further timing analysis revealed a less than ideal match but it still resides as the first event not easily rejected as RFI. 75

89 Figure 3.31: A Monte Carlo simulation from Tim Huege s REAS code [45] of an shower recorded by the Dutch team while triggering from muon counters. The black line is the first channel of the CRRU receiver MHz results after being filtered at MHz shifted to the IF of 315Mhz, amplified and shifted down to the 2-17 Mhz recorded at the UB. The black diamonds are 25ns samples from the 40MS/s ADC s. 76

90 Figure 3.32: The FFT of a Monte Carlo simulation from Tim Huege s REAS code [45] of an shower recorded by the Dutch team while triggering from muon counters. The red portion is the first channel of the CRRU receiver MHz. 77

91 Figure 3.33: A station showing wind damage after an extended period without a maintenance visitation. All power cables were severed and the electronics crate showed signs of internal water damage. Solar panels on a different station had detached from the frame and toppled forward, slicing the comm and solar cables. 78

92 Figure 3.34: A new design for the LPDA using a simple, inexpensive wood or plastic frame and wires. The longer base wire allows for extending to lower frequencies. This LPDA extends to 25 MHz from the initial 40 MHz [47] 79

93 Figure 3.35: The layout for the 10 station MAXIMA deployment. Poles P4, P5, P6 were used in the OSU/Leeds effort. Four MAXIMA stations lay out a triangular pattern surrounding a central station at M1, M2, M3, M4 for the initial test data run. Planned expansion towards 10 stations with antennas at B2, C4, D3, E3, E4, P6 seen in yellow. The infill SD station Olaia is seen as the red dot with the regular SD deployed stations Apolo, Arbolito, and Sandra seen in yellow [44] 80

94 Figure 3.36: A MAXIMA station. MAXIMA addresses many if not all of the shortcomings of the OSU/Leeds stations. The stations are outfitted with 400MS/s digitizers, 2 Isofoton 150W solar panels, quality batteries, ample local memory, and a comms system reliant upon the WiFi g standard [44]. 81

95 Of the three designs that were tested on-site in Malargüe, the total number of radio events in coincidence is still relatively small with few confirmed events mainly due to the small scale of these R & D setups. The CODALEMA style of active fat antennas were tested near the Central Laser Facility (CLF) laid out in triangular fashion with 139m baseline using a self-trigger. These events were then compared against the timestamps of the SD Array events in the area. This system has registered 36 confirmed events that coincide with SD Array events [48]. Another test set-up consisted of LPDA antennas of both designs (tubular metal frame and wire) along with inverted-v dipoles similar to that of LOFAR. These antennas were laid out on a 100m baseline and were triggered from 2 scintillator paddles located nearby the BLF. These events were then compared to the official SD Array events. This setup has confirmed 27+ events that coincide with with the SD Array events. However, most events consisted of 3 and 4 tank events and would not be listed in the official Trigger level 6 published event list [32]. The OSU/Leeds test set-up registered 10k+ events coincident between two antenna with only a small fraction coincident with SD Array. Of those, none were coincident to an accuracy beyond 1µs. Due to the communication problems previously discussed only 2 of 3 stations could run at any one time and no angular agreement could be tested. Using a random background trigger, a long-term noise study has been completed. The root mean square value of a time trace band-limited to MHz sampled at the rate of 400MS/s, see Figure The east-west polarization is shown and the data display a rise in noise near 18:00 local sidereal time corresponding to a radio 82

96 bright feature near the galactic center. This sky noise could be used as a calibration tool for the radio detector stations [49]. Of the 36 events registered near the CLF a sky map was generated, see Figure The map displays a highly asymmetric distribution with roughly 70% of events coming from a southerly direction. The angle of the shower axis relative to the local magnetic field plays a large part in the generation of these pulses. This provides further support to the geomagnetic nature of these signals [48, 50, 4]. 83

97 Figure 3.37: RMS of noise level in arbitrary units measured by one test design at the Balloon Launch Facility (BLF) triggered via scintillator paddles plotted against local sidereal time. The increase in noise around 1800 LST is caused by the passage of the galactic center overhead. This noise presents the ability to calibrate the antenna, cable, amplifier, receiver chain. Data were collected from May 2007 to April 2009 [49] 84

98 Figure 3.38: A skymap of 36 events logged at by the CODALEMA design near the CLF in coincidence with the Auger SD events. The plot is in local spherical coordinates with east to the right at 0 degrees, north up at 90 with the zenith being the center of the plot. The geomagnetc field direction at Malargüe is denoted by the red circle. The asymmetry suggests a geomagnetic dependence on signal generation [50, 48] 85

99 3.9.6 Future Deployment Based upon the promising results of the MAXIMA, CODALEMA, and OSU/Leeds, the Pierre Auger Collaboration has approved the deployment of a large scale radio array. The Auger Engineering Radio Array, AERA, is scheduled to deploye 100+ antennas in the Auger Engineering region in the northwest corner near the Coihueco Fluoresence Detector, see Figure Along with AERA, there will be AMIGA, Auger Muons and Infill Ground Array, which will place infill tanks at 750m rather than the usual 1.5km distance between tanks and also install muon counters in this area. The FD enhancement HEAT, High Elevation Auger Telescopes, has been installed overlooking this section of the SD which allows for tracing the shower development at higher elevations, 30 to 60. The infill tanks effectively lower the triggering threshold in this region to which will allow probing of the knee region with SD, FD, and Radio simultaneously [41, 42, 43]. AERA will cover 20 km 2 with 156 radio detector stations. The stations will deployed in a staggered triangular fashion. The first 24 will be a densely packed triangle with 150m spacings surrounded by 60 antennas spaced at 250m which then extends westward to 72 antennas at 375m spacing, see Figure Each station will have a 30 to 80 MHz frequency band and self-trigger using an FPGA-based smart trigger with pulse shaping. The data will be read out at 12 bit ADCs with a sample rate of 2000MS/s. Based upon extensions of the CODALEMA data, LOPES data [51] and the REAS2 Monte Carlo simulations [45] AERA expects 5000 events per year once the array is complete [43]. Of these events 1000 events will have energies exceeding ev. The lower energy limit of the array is estimated to be around the infill lower limit of ev. 86

100 Figure 3.39: Location of the different radio testing locations. Existing stations are in operation at the BLS, Balloon Launch Site, and CLF, Central Laser Facility. AERA will be located near the Coihueco FD telescope on the northwest edge of the array [43] 87

101 Figure 3.40: An aerial view of the layout of the Auger Engineering Radio Array, AERA. Surface detectors are denoted by triangles. The radio stations are shown here as dots. The dashed line shows the upper portion of the hexagon where the infill AMIGA tanks are located [43]. 88

102 Chapter 4 AMBER: A New Cosmic Ray Detector 4.1 Introduction As noted earlier, the two general techniques employed in the current generation of cosmic ray detector systems each have their own strengths and weaknesses. A surface detection system relies upon sampling particles from the extended air shower. While this technique yields a very strong directional solution through timing analysis of the various tanks or scintillators, the energy estimate produced is reliant upon models and Monte Carlos. The optical fluorescence detectors can follow the longitudinal development of the shower and yield an accurate energy estimate based upon direct observation. However, the fluorescence detectors operate only with about a 10% duty cycle with only being active on clear moonless nights [cite Corbato]. The energy estimate also requires a detailed knowledge of atmospheric content through weather balloons, laser shots for calibration and other methods. The air shower from an ultra-high energy cosmic ray dissipates most of its energy into ionization but only a small fraction of that is applied to molecular nitrogen ionization that is observed by the fluorescence detectors. The cooling process of the tenuous plasma generated from the ionization leaves many other observation channels open for study. Several studies, including two accelerator beam tests, were done 89

103 in the microwave region [52]. Further studies are underway attempting to observe the mircrowave bremsstrahlung radiation (MBR) [53] from a high energy extended air shower. The Airshower Microwave Brehmsstrahlung Experimental Radiometer (AMBER) experiment is underway with prototypes being tested at the University of Hawaii at Manoa (UHM) [52] and the Ohio State University with an expected deployment at the Auger Observatory in the near future. 4.2 MBR: Molecular Bremsstrahlung Radiation Free electrons accelerating through collisions with the fields of neutral air molecules is the basic definition of MBR. MBR is isotropic and unpolarized which differes greatly from traditional bremsstrahlung seen in high energy physics. This isotropic behavior can be thought of as analagous to the fluorescence detector while the geosynchrotron emission discussed previously is highly beamed and analagous to traditional surface detection. 4.3 Accelerator Results Two accelerator tests were designed to test this MBR property of rapidly cooling plasma from a lab created air shower for coherence effects. The results for these tests are summarized here AWA INCOBREMS / SLAC T471 The Argonne Wakefield Accelerator (AWA) INCOBREMS experiment from June 2003 used a beamline energy of ev. After colliding into a tungsten target, the resultant gamma rays passed through the chamber shown in Figure 4.2. Inside the 1 m 3 copper Faraday anechoic air chamber, see Figure 4.3, the simulated shower 90

104 Figure 4.1: A schematic sketch of different detection techniques employed at the Pierre Auger Observatory [52]. Figure 4.2: The interior and exterior of the Faraday Anechoic chamber utilized in the beam tests at Argonne Wakefield Accelerator and SLAC. The cube is approximately 1m 3 [52]. 91

105 Figure 4.3: A schematic view of the AWA INCOBREMS (top) and SLAC T471 (bottom) tests [52]. 92

106 typically deposited about 1 PeV of ionization energy. The received power if from incoherent sources should scale linearly with beam energy. The observed signal, from a circularly polarized Ka band (20-21 GHz) antenna, seemed to indicate coherence as the observed signal scaled with the square of the beam energy. The phase stability of the coherent portion of the signal allowed the subtraction of any incoherent background and hinted at a partially coherent emission 50 ns after the beam passage as shown in Figure 4.4. However, there was significant background radiation resulting from the collimation of the AWA electron beam prior to the beam striking the tungsten target. This collimation resulted in a high background radiation inside and out of the target box. Even with the beam blocked by lead the background radiation approached the level of the apparent signal, see Figure 4.5. In 2004, the experiment was taken to SLAC where a more controlled beam was available. The antenna used inside the chamber was changed from circularly polarized antenna to a log-periodic design with one antenna s polarization aligned to the direction of the beamline and the other perpendicular to it. A 28 GeV electron beam collided with with a target of 90% Al 2 O 3 and 10% SiO 2 of varying thickness to create shower energies were of order ev which is similar in total shower energy at the AWA test. The results from the SLAC test are shown in Figures 4.6, and 4.7. The beam bunch takes about 3 ns to traverse the box. A strong initial signal is seen in the co-polarized plot, Figure 4.6, but this was to be expected from transition and radio Cherenkov radiation. The anechoic absorber material limits the reflections ( 30 db per reflection) and the strong initial pulse dies out quickly after which a longer exponential decay is seen. The noise level in this co-polarized antenna is limited by the dynamic range of the digitizer and the strength of that initial pulse. The 93

107 Figure 4.4: The background subtracted signal in a 1Ghz band at 20Ghz observed emission when using a 5mm tungsten target [52]. Figure 4.5: The total emission observed at from a 1GHz band at 20GHz. The blue line observed when the beam was on target and the red dashed line observed when the beam was blocked. This background radiation due to the beam scraping done during collimation approached the signal levels even with the beam blocked with lead from the target [52]. 94

108 cross-polarized, Figure 4.7, antenna was mostly insensitive to the transient signals from the electron beam. A decaying signal can be seen out to 60 ns with a decay constant of 7 ns and the noise floor is now the thermal limit rather than instrument limited. Care was taken to ensure there was no beam background noise in this test to be subtracted out. Figure 4.8 plots the integrated power from ns after the beam pulse against the externally measure beam energy. The observed signal grows quadratically with the beam energy implying at least a partially coherent process dominating the incoherent signal. If we assume that the results scale from Figures 4.8 and 4.7, the minimum detectable signal from a GZK energy shower, ev should be seen from 20km. 95

109 Figure 4.6: Average amplitude from 100 beam shots at near shower-maximum during the SLAC T471 experiment. Using an antenna that was polarized along the direction of travel and therefore sensitive to partially coherent radiation directly from the eletron bunch as it passed through the container. A strong sharp pulse is seen then dies away quickly followed by a slower exponential decay tail [52]. 96

110 Figure 4.7: A plot similar to the previous graph but with the polarization perpendicular to direction of travel and insensitive to the crossing of the electron bunch. The decay extends to 60 ns with a decay time of 7 ns. The upper and lower red dashed lines are the minimum detectable signal for single shot and 100 shot average. The diagonal dotted lines correspond to the two extreme cases of partial coherence. The lower being fully suppressed and the upper being zero suppression present [52]. 97

111 Figure 4.8: Relative microwave energy in the tail of the observed air plasma emission vs. the transistion radiation readout which is directly proportional to the beam energy. The observed power goes as the square of the beam energy [52]. 98

112 4.4 AMBER Hardware Following the promising results of the SLAC T471 test it was determined that a prototype system would be assembled from satellite television components. Economies of scale work in our favor at the frequencies tested during the SLAC run. C-band and Ku-band frequencies are primarily satellite communication frequencies and high quality and inexpensive low noise amplifiers are available in these frequencies. At the University of Hawaii, a rooftop prototype detection system, shown in Figure 4.9, consisting of a 1.8m off-axis parabolic dish with a feed of 4 individual C (4-8 GHz) and Ku-band (12-18 GHz) dual band feed horns arranged in a diamond pattern. Figure 4.9: Prototype AMBER reflector and feed array on the roof of the University of Hawaii physics building rooftop [52]. At Ohio State University, we have started assembling a new prototype station on West Campus with the intent of having a sister station in Argentina as soon as 99

113 possible. The dish size in the new system has increased to 2.4m, as seen in Figure The horn assembly is similar in structure to the Hawaii assembly. While we are using similar horns, ones that have C- and Ku-band pickups, we are using primarily C-band. C-band, 4 8GHz, response is sufficient for our needs, i.e. we do not want to pay for Ku-band receivers when they add very little to our signal strength. The horns purchased for deployment in Argentina are C-band only and are properly tuned to a 2.4m dish. From the dish site on West Campus, 8 lines of signal (2 polarizations per horn) are transmitted to an electronics crate at the base of the dish with low-loss 75Ω cable and then passed along into a nearby communications building with low loss 50Ω cable. We have arranged a rack of electronics in the nearby communications building for our digitizers and data aquisition linux box. An electronics crate will be located at the satellite dish which will allow for local triggering and possibly power detection. At 3Ghz, the LMR-400 low-loss cable we installed 6inches under the turf will lose 7 db per 100 ft and we have approximately 200 ft of cable running from the dish to the digitizing rack. This eliminates any realistic possibility of transmitting the raw signals directly from the dish to the comms building. In one design the electronics package at the dish would have fast triggering logic and upon a trigger convert the signal to a baseband frequency for transmission to the communication building. We have 50 MS/s digitizing boards which yield 25 Mhz of bandwidth. If we take 20 Mhz worth of signal to allow for bandpass filter roll-off, in our example we will take 3Ghz to 3.020GHz and down convert it with a mixer to 3Mhz to 23Mhz we will lose considerably less signal during transmission. At 30Mhz, the LMR-400 cable loses considerably less signal at 0.70dB per 100ft of cable. 100

114 In a different possible design, the triggering is still done at the dish and using a series of Mini-Circuits power detectors the power is transmitted to the digitizers. This would be the simplest of options as the power detectors have a well documented response to signals. A trigger signal would be sent to the digitizers and the received voltage peaks would be directly proportional to the received power at the dish. Figure 4.10: Initial installation of a 2.4 m Prodelin dish on the OSU West Campus dish farm. 101

115 Figure 4.11: The diamond shaped C and Ku-band feed horn array for use at the OSU West Campus dish farm site. 102