The Analog Signal Processing System for the Auger Fluorescence Detector Prototype

Size: px

Start display at page:

Download "The Analog Signal Processing System for the Auger Fluorescence Detector Prototype"

Sherman McGee
5 years ago
Views:

1 444 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 48, NO. 3, JUNE 2001 The Analog Signal Processing System for the Auger Fluorescence Detector Prototype S. Argirò, D. V. Camin, P. Cattaneo, M. Cuautle, M. Destro, P. Facal, R. Gariboldi, V. Grassi, M. Lapolla, P. Manfredi, E. Menichetti, P. Privitera, L. Ratti, V. Re, V. Speziali, P. Trapani, and E. Tusi Abstract The fluorescence detector of the Pierre Auger Cosmic Ray Observatory will provide a measurement of the parameters of extended air showers in the range from to ev. An array of photomultiplier tubes (PMTs) is placed at the focal surface of a large-aperture telescope thus forming one of the 30 detector modules. The shape of the signal generated by each PMT is variable, depending mostly on the geometry of the air shower as seen by the detector; after analog processing the waveforms will be sampled at a rate of 10 MHz with 12-bit resolution. We have developed an analog signal processor to achieve the best compromise between energy and time resolution, low noise, and low cost. The head electronics provides an active bias network for the PMTs, which keeps the gain constant even in the presence of large dc background light from the night sky. This dc level is measured by means of a built-in optocoupled linear circuit. The pulse signal is sent through a twisted pair to the analog front-end board. At this stage a compression of the 15-bit dynamic range of the signal into the 12-bit range of the FADC is performed. Antialiasing is provided by a Bessel filter. Index Terms Analog signal processing, fluoresecence detectors, ultra high energy cosmic rays. I. INTRODUCTION THE Pierre Auger Observatory [1] for the highest energy cosmic rays was designed to disentangle the puzzle of the origin of cosmic rays above the GZK cutoff ( ev). Such a goal is pursued using a hybrid detector technique: a surface detector comprising 1500 water Cherenkov detectors spread over a 3000-km area for detection of ground particles and a fluorescence detector composed of four eyes able to track fluorescence light produced from the interaction of secondary particles with the atmosphere. Each eye is composed Manuscript received October 29, 2000; revised January 27, S. Argirò was with the University of Milan, Italy, and the INFN. He is now with the Physics Department, University of Turin 10125, Italy, and the INFN, Sezione Torino, Italy ( stefano.argiro@to.infn.it). D. V. Camin, M. Destro, R. Gariboldi, V. Grassi, and M. Lapolla are with the University of Milan, Italy, and the INFN, Sezione Torino, Italy. P. Cattaneo is with the INFN, Sezione Torino, Italy. M. Cuautle is the University of Milan, Italy, the INFN, and the University of Puebla. P. Facal is with the University of Roma II, Tor Vergata, Italy, and the INFN and also with the University of Santiago de Compostela, Spain. P. Manfredi, L. Ratti, and V. Speziali are with the University of Pavia, Italy, and the INFN, Sezione Torino, Italy. P. Privitera and E. Tusi are with the University of Roma II, Tor Vergata, Italy, and the INFN. E. Menichetti and P. Trapani are with the University of Torino, Italy, and the INFN, Sezione Torino, Italy. V. Re is with the University of Bergamo, Italy, and the INFN, Sezione Torino, Italy. Publisher Item Identifier S (01) Fig. 1. Schematic view of a fluorescence detector mirror module. (Fig. 1) of a series of segmented spherical mirrors, having focal length of 1.74 m, field of view, and overall dimensions of about m. At the focal surface of the mirror there is a camera of photomultiplier tubes (PMTs), each with a field of view of The camera will track the longitudinal development of the extended air shower, i.e., will see a light point moving, and the associated electronics will sample the light track at a rate of 10 MHz. As the number of fluorescence photons emitted per meter of shower development is nearly constant [2], this technique allows a calorimetric measurement of the primary energy where the atmosphere is used as a radiator. II. SIGNAL AND NOISE CHARACTERISTICS The signal produced by each PMT (pixel) has a trapezoidal shape, with durations varying from a few hundreds of nanoseconds to several microseconds. This particular shape is a consequence of the fact that light coming from the shower is focused on a spot of a certain size that moves along the camera (Fig. 2). The hexagonal photomultiplier Philips XP3062 was selected. The signal intensity depends both on the shower size (i.e., number of electrons) at that particular point being observed and on the distance of observation, as both dependence and atmospheric light scattering contribute to attenuation of the signal. Simulations show that the dynamic range, for showers in the range 10 to 10 ev and at a distance from 7 to 30 km, will be about 15 bits [3], [4]. Rise and fall times depend on the ratio of the spot size to the PMT field of view. With the former around 0.5 due to optical aberrations and the latter chosen to be of 1.5 due to geometrical /01$ IEEE

2 ARGIRÒ et al.: ANALOG SIGNAL PROCESSING SYSTEM 445 Fig. 2. Signal formation in the PMT. reconstruction and signal-to-noise (S/N) requirements, the rise and fall times turn out to be about 1/3 of the pulse fullwidth at half-maximum (FWHM). The dominant source of noise is coming from the night sky background light and it is caused by its fluctuations. The mean photoelectron rate due to the background is estimated to be 2.7 /100 ns, or, assuming a PMT gain of 5 10, 250 na at the anode. The noise spectral power density, calculated at the preamplifier input, is where relative variance in the gain of the PMT dynodes; PMT gain; and cathode currents associated to background and signal, respectively. For, the spectral power density of the noise coming from sky background is of the order of 70 pa/ Hz and 92 pa/ Hz when the minimum expected signal (of 2 /100 ns) is present. The design requirement of the front-end electronics is that it should not raise the total noise by more than 10%. III. THE ANALOG SIGNAL PROCESSING SYSTEM The main functionalities of the analog system are the following [5]: readout fluorescence signals from the PMT with high linearity and S/N ratio; fulfill the requirement of a 15-bit dynamic range; prepare the signal for sampling at a rate of 10 MHz; readout with high resolution and low sampling rate the mean sky background current; provide tools suitable for testing the entire chain (test pulse network); ensure high reliability throughout the experiment s lifetime of 20 years. The electronic system is composed mainly of two pieces, the head electronics (HE) and the analog front-end board (AB). A (1) distribution board provides routing for signals and power supplies. A schematic representation of the system is given in Fig. 3. Signals processed by the analog system are output to the first level trigger board, to which the AB is connected. Here the waveforms are sampled at 10 MHz [12]. In the following we will review these building blocks in more detail. A. The HE The HE is directly connected to the PMT. It consists of two circular PCBs of 32 mm in diameter, interconnected and positioned on the same axis with the PMT. The first PCB contains the PMT bias network. The second PCB houses the signal driver and the circuit that performs the readout of dc anode current. Note that since the cathode is grounded to avoid dust deposition due to electrostatics, a novel solution had to be devised. In fact, when the anode is grounded, a current meter in series with the load resistor is sufficient, but, when the anode is at high potential with respect to ground, the measurement is not trivial. 1) PMT Bias Network: The HE provides an active bias network, in which the potential of the last three dynodes is stabilized by high-voltage bipolar transistors. This helps keep the PMT linear even in the presence of large dc currents flowing through the anode. While a standard passive circuit shows an increase of the gain by 10% for an anode current of 20 A, the gain variation for an active bias network is only 2% under the same conditions. In addition, the active circuit dissipates nearly one-half of the power with respect to the standard passive solution [6]. 2) Signal Driving: A fully differential configuration has been adopted in order to achieve the best rejection of commonmode noise. The current from the PMT anode flows, after ac coupling, through a load resistor and reaches one of the differential input of the line driver, while the other input only picks up the same common-mode noise. A common mode rejection ratio (CMMR) of 28 db in the range from 1 to 100 khz was measured [7]. A monolithic integrated circuit drives a twisted-pair cable that brings the signal to the AB. 3) DC Anode Current Measurement (Current Monitor): Recording dc current for every pixel will give valuable information to keep a record of the charge accumulated by

3 446 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 48, NO. 3, JUNE 2001 Fig. 3. Schematic representation of the analog system. on the action path is such that almost all of the current to be measured circulates through the first photodiode, and it is replicated at ground level at the second photodiode. This current is then used to produce a voltage, where is the dc level at the signal driver s output, is the dc level for total darkness, V, and is a conversion factor of about 60 mv/ A of anode current. The fluorescence signal is sent differentially to the AB via twisted pair, whereas the sky background signal is sent as a linearly related dc level-toground using the same wires. 4) Test Pulse: The HE provides a test-pulse input that emulates a fluorescence signal by injecting a current at the same node as the PMT does. A bipolar transistor in common base configuration converts an input voltage signal generated by an external pulser into a current. At present, the testing of 11 channels at the same time is possible; for the next version of the system, a gating to allow individual channel testing and pattern generation has been foreseen. Fig. 4. The optocoupled current mirror. the PMT, trigger protecting actions when the current goes too high, track stars and other objects, and therefore check alignment and its stability and get complementary informations for monitoring atmospheric conditions. To obtain a direct measurement of this quantity for a PMT with its cathode set to ground, a novel optocoupled circuit has been devised [8]. A combination of a linear optocoupler illuminating two diodes and a standard optocoupler followed by a high current-gain stage form a feedback loop in which both the action and the feedback paths are optically coupled (Fig. 4). The current gain B. The Distribution Board The distribution board provides the link between the camera and the front-end crate. It distributes high voltage, low voltage, signal, and test signal, arranged in a modular way. It also provides protection against power surges. C. The Front-End Analog Board The front-end analog board (FEB-AB) is connected to the digital front-end board (first level trigger board, FLT); together they form a VME euro-size module that is placed in a crate. Each board accommodates 22 channels; therefore 20 combinations AB FLT are needed in order to readout the camera. The AB

4 ARGIRÒ et al.: ANALOG SIGNAL PROCESSING SYSTEM 447 receives its input from the HE through the distribution board and outputs signals ready to be digitized by the FADCs that are located on the FLT. The main functionalities are: receiving the signal from the HE, performing a differential to single-ended conversion; controlling the gain of the channel; adapt the dynamic range of 15 bit to the 12-bit FADC; provide antialiasing filtering in order to avoid reconstruction errors; provide test pulse generation; perform the measurement of the dc level on the twisted pair, which is linearly related to dc anode current. The dc level measurement, i.e., the current monitor readout, is performed by a six-channel 16-bit ADC. Four such devices are connected in cascade and communicate with the FLT board through a serial link. The samples continuously fill dedicated registers on the FLT at 1 KHz. A software routine extracts this information at a programmable rate, typically of 50 to 100 Hz, adequate to the relatively slow variations of the night sky (i.e., star motion). The 16-bit system allows a resolution of 0.4 na in anode current or 0.2% of the minimum expected background current. The test pulse system uses an external pulser to emulate fluorescence pulses of arbitrary shape which are injected into the HE. The generator enters all 20 AB in parallel; on each board, two differential line drivers of the same kind as the ones used on the HE distribute the pulse to a group of 11 channels. 1) Channel Architecture: The channel on the AB can be divided in the following logical blocks: receiver, gain stage, antialiasing filter, and dynamic range adapting. The signal receiving and conversion from differential to single-ended is done by a transformer. This has the advantage of adding no noise and dissipating no power. In addition, it provides galvanic decoupling, ensuring the absence of ground loops. On the other hand it introduces a time constant of about 250 s; a wide pulse ( 10 s) will have a drop of 4% and a similar undershoot. The signal dynamic range is limited in the upper side by the maximum signal the driver can handle, 2.5 V at its input. The lowest signal is limited by the total noise. The noise of a monolithic differential receiver increases the total noise in such a way that it becomes impossible to reach a higher dynamic range than 33K : 1 (15 bit). With the pulse transformer, which introduces no noise, a 64K : 1 (16 bit) can be reached, also because it has an intrinsic high rejection to common-mode noise. The receiver allows dc coupling and pixel selection through its enable pin. The channel gain is adjusted by means of digital potentiometers, in general connected in series with a resistor on the feedback loop of an operational amplifier. These devices allow fine tuning of the gain in 256 positions, allowing compensation for gain drift in the PMTs without having to control their HV individually which would be much more expensive. 2) The Antialiasing Filter: A study of the optimal antialiasing filter has been carried out [10]. A compromise has been reached between reconstruction error and circuit complexity. A third-order Bessel filter has been implemented, with a cutoff frequency of 1.5 MHz. The Bessel filter has Fig. 5. Dynamic range adapting: (a) the virtual channel and (b) the compressor. a linear dependence of the transfer function phase with the frequency and therefore adds no distortion to the signal, but only a constant delay. 3) Dynamic Range Compression: Two solutions have been implemented to provide dynamic range compression: the virtual channel and the compressor solution. The virtual channel, represented in Fig. 5(a), exploits the feature that the signal does not appear on all pixels at the same time, but rather in a sequence. Every channel is configured with a high gain of about 30, tuned on small pulses, but the signal before the gain stage is also routed, together with ten other channels, to a sum stage. The sum is then processed by a virtual channel with a low gain of about unity. If one channel out of the group of 11 with high gain goes into saturation, the signal can be recovered from the virtual channel, which has lower gain. The compressor solution [Fig. 5(b)], on the other hand, provides a dynamic compression of the signal through a bilinear transfer characteristic [9]. Signals below a threshold undergo a large amplification, while signals above that threshold undergo a small amplification. The threshold value is controlled by a digital potentiometer and so is the slope of the two gain branches. IV. TEST RESULTS A preliminary test has been carried out with a small portion of the camera, consisting of seven PMTs [13]. The setup is shown in Fig. 6; a pulsed blue LED was simulating the fluorescence pulse of the desired duration, while another LED was fed with a dc current to simulate the sky background at different intensities. Between the LEDs and the photomultipliers, a filter wheel was placed, carrying five calibrated filters, for unbiased linearity measurements. The test yielded the following results. Noise Performance: An rms noise of 0.6 least significant bit (LSB) was registered for the AB and FLT. The full chain, including PMTs with high voltage on, showed an rms noise of 1 LSB in the virtual channel option. The compressor option showed a slightly worse noise performance ( LSB), due to the larger number of components needed. To compare the noise produced by the electronics to that coming from the night sky background, a special

5 448 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 48, NO. 3, JUNE 2001 (a) Fig. 6. Test setup. test run was set up. The PMTs were illuminated with a blue LED that was fed with a dc current to provide a dc anode current of the order of 250 na, roughly corresponding to the minimum expected background current. Comparing fluctuations of the baseline for this setup to those recorded in dark conditions, thereby corresponding to noise coming from the system alone, the contribution of the electronics noise was found below 20% of that coming from the simulated background. Crosstalk: Crosstalk measurements were performed by use of a plastic mask which allowed blinding of all PMTs but one. While a given PMT was illuminated with a light pulse, the neighboring channel showed a maximum crosstalk of 0.3%. Timing Resolution: The timing resolution was tested by calculating the centroid of each pulse, defined as (2) (b) Fig. 7. (a) Charge integrals (in ADC count) versus filter attenuation. (b) Deviation from linearity. filter attenuation and the plot was fit to a straight line. Results showed a maximum deviation from linearity, defined as (3) (where is the time slot and value of the corresponding sample), then plotting the histogram of the centroid position with respect to the time the trigger occurred. The rms of that distribution was defined as the timing resolution. The dependence of this quantity on the signal width and amplitude was studied. The typical resolution was of the order of 50 ns. It has been proved that by applying more sophisticated reconstruction algorithms currently under development, it can be further improved. Linearity: The system linearity was tested by use of a set of calibrated filters, whose attenuation is known to an accuracy of 1%. The transmission factor ranged from unity to The pulsed LED produced signals of different shapes. For a given light signal, waveforms were recorded at different attenuations. The measurements were repeated using a LED producing dc light to simulate night sky background noise. An analysis was performed to evaluate the charge carried by each pulse. The charge integral was plotted versus where is the charge and the corresponding value of the fit, of less than 2% without dc light and less than 4% in presence of dc light. Results are presented in Fig. 7. V. CONCLUSION The fluorescence detector for the Pierre Auger Cosmic Ray Observatory, with its innovative design, puts particular requirements on the electronics performing its readout. In particular, the analog readout electronics must feature low noise and a 15-bit dynamic range, while keeping the reconstruction error on the sampled signal well below 10%. In addition, it has to perform a precise measurement of the dc anode current and provide self-testing capabilities. Its engineering must cope with simplicity and low cost, due to the large total number of channels required (15 000) and the long experiment s lifetime (ten or more years). We put our efforts into realization of such a system and tested the first prototype. The tests we have performed yielded very

6 ARGIRÒ et al.: ANALOG SIGNAL PROCESSING SYSTEM 449 satisfactory results in terms of crosstalk, timing accuracy, and linearity. The increase of the total noise due to the electronics was shown to be still larger than the stringent requirement of 10%. Further investigation is underway. ACKNOWLEDGMENT The authors would like to thank the Auger Group at the Forschungzentrum Karlsruhe, and especially H. Gemmeke, M. Kleifges, A. Menschikov, and D. Tcherniakhovski for useful discussions and for building an excellent trigger and readout system [12]. They also would like to thank them for hosting the sunflower test, in which they and their system played an irrenounceable role. REFERENCES [1] Auger Collaboration. (1977, Mar.) The Pierre Auger Observatory Design Rep.. [Online]. Available: [2] P. Sokolsky, Introduction to Ultra High Energy Cosmic Ray Physics. New York: Addison Wesley, [3] B. Dawson. (1997) Amplitude dynamic range in Auger fluorescence electronics. Pierre Auger Observatory. [Online]. Available: [4], (1999) Amplitude dynamic range in Auger fluorescence electronics: Update. Pierre Auger Observatory. [Online]. Available: [5] D. Camin and V. Re. (1999) Analog processing of signals from the fluorescence detector. Pierre Auger Observatory. [Online]. Available: [6] S. Argirò, D. V. Camin, M. Destro, and C. K. Guerard. (1999) Passive and active PMT biasing networks II. Pierre Auger Observatory. [Online]. Available: [7] D. V. Camin, M. Cuautle, M. Destro, and R. Gariboldi. (1999) Fabrication of the first 150 head electronics units Results of the acceptance tests. Pierre Auger Observatory. [Online]. Available: [8] S. Argirò, D. V. Camin, M. Destro, and C. K. Guerard, Monitoring DC anode current from a grounded-cathode PMT, Nucl. Instrum. Methods, vol. A435, pp , [9] P. Manfredi, L. Ratti, V. Re, and V. Speziali, A bilinear analog compressor for the Auger fluorescence detector, in 8th Pisa Meeting Advanced Detector: Isola d Elba, May [10] P. Cattaneo and L. Ratti. (1999) The anti-aliasing requirements for the FD read-out channel. Pierre Auger Observatory. [Online]. Available: [11] S. Argirò, D. V. Camin, P. Cattaneo, M. Destro, R. Fonte, and R. Gariboldi et al., The analog signal processor of the Auger fluorescence detector prototype, Nucl. Instrum. Methods, vol. A461, pp , Mar [12] H. Gemmeke, A. Grindler, H. Keim, M. Kleifges, N. Kunka, D. Chernyakhovsky, and Z. Szadkowski, Design of the trigger system for the Auger fluorescence detector, IEEE Trans. Nucl. Sci., vol. 47, pp , [13] S. Argirò, H. Gemmeke, M. Kleifges, A. Menschikov, P. Privitera, and D. Tcherniakhovski et al., The sunflower test: Results, Pierre Auger Observatory,

Monitoring DC anode current of a grounded-cathode photomultiplier tube

Nuclear Instruments and Methods in Physics Research A 435 (1999) 484}489 Monitoring DC anode current of a grounded-cathode photomultiplier tube S. Argirò, D.V. Camin*, M. Destro, C.K. GueH rard Dipartimento