Calibration of photomultiplier tubes for the large-angle beamstrahlung detector at CESR

Transcription

1 WSU-REU2002/West Calibration of photomultiplier tubes for the large-angle beamstrahlung detector at CESR M. West Wayne State University, Detroit, MI ABSTRACT This project is to prepare for the upcoming installation of a large-angle beamstrahlung detector at the Cornell Electron Storage Ring (CESR). The goal is to thoroughly characterize a set of sixteen photomultiplier tubes (PMTs) that will be used in the final detector. The detector will provide a wealth of information about various parameters of beam collisions. I. INTRODUCTION Beamstrahlung is the electromagentic radiation produced when electron and positron beams are bent at the interaction point. [1] This radiation can be detected directly using photomultiplier tubes without the need for scintillators. By carefully observing this radiation, researchers are able to closely monitor the beam conditions and diagnose beam pathologies that lead to wasted luminosity due to incomplete overlap, as illustrated in Fig. 1. The characterization process involves looking at noise level and photoelectron count rate as a function of high voltage input, discriminator threshold, and temperature, as well as recording the Gaussian single-photon spectra for all sixteen tubes. Proper calibration of the PMTs will allow superior precision once the final detector is constructed. II. EQUIPMENT USED Apparatus involved in the calibration included sixteen Hamamatsu photomultiplier tubes, eight of model R6095 (visible range) and eight of model R (infrared range). These were mounted on a set of plates provided by the Wayne State machine shop, along with a set of Hamamatsu 14-pin voltage dividers. This array was placed inside a black-painted aluminum box nine inches tall by nine inches wide by 42 inches long to shield it from ambient light. Two light sources were provided at the end of the box opposite the phototubes: a 1-16 inch hole with a paper diffuser to admit ambient light, and a breadboard with red, yellow, and green LEDs mounted to provide an internal light source that can be powered by a pulse generator. A negative high voltage power source was needed to power the tubes, and a wide range of NIM electronics and two oscilloscopes were used to analyze the PMT output. We also prepared two sets of long signal cables: one 10 feet in length for use in the lab, and one 85 feet in length to interface with the CESR controls if time allowed. Finally, we made provisions to chill the tubes, since the infrared models are intended to be operated at up to -30 Celsius. An insulating box was quickly fashioned from polystyrene and chilled with liquid nitrogen. Temperatures inside the box were monitored with a Fluke digital temperature probe. The specific models of electronics modules used are listed in Table I, while Table II lists the phototubes according to the arbitrary numbering (1 through 8) that occurred during installation in our test apparatus. III. ELECTRONICS MODULE SCHEMES Several electronics schemes were attempted, with varying results. The first successful arrangement is illustrated in Fig. 5. It puts the analog PMT output signal through a discriminator to convert it to a logic signal and then sends it to a scalar which displays the count rate in counts per second, 0.1 second, or per 10 seconds. The ten second setting was typically used to minimize the effect of small time fluctuations. This scheme may be operated either with the pulsed LED as pictured or with a point light source created by admitting ambient light through a pinhole. (A pinhole proved difficult with our aluminum box, so a 1/16 inch hole was drilled and covered with layers of paper to provide a sufficiently low intensity so that the scalar would not count beyond its capacity.) Since the amount of light the phototube sees is in no way taken into account in this scheme, it is necessary to take two sets of data for each run, one with the light source, and on without, so that counts due to background radiation may be subtracted. A NIM-to-TTL converter, not pictured, was also employed because our scalar required a TTL input signal, and all discriminator thresholds were set at 50 millivolts. The second working scheme is illustrated in Fig. 6. It adds a strobed coincidence module, which serves to correlate PMT output with the pulse rate of the LED light source, and in so doing rejects most of the counts due to background, so that the awkward and time-consuming subtraction process need not be employed. (Some noise counts are still counted, but this number turns out to be a fraction of a percent of the total count rate under most circumstances.) The gate signal for the coincidence is provided by the crate pulse generator, which ensures that the gate is firing at the same rate as the LED, and also

2 2 allows the user to delay the gate signal to compensate for delay in the PMT output with respect to the original firing of the trigger signal. Since the crate generator s output is analog, and additional discriminator had to be placed between it and the gate input in the coincidence. IV. TESTING The very first step of the testing process was to use a single tube and divider, boxed and wrapped in aluminum foil, to observe a noise signal similar to that shown in Fig. 2. This verified two things: one, that at least one of our tubes was working as expected, and two, that the voltage divider scheme copied from the Hamamatsu catalog was appropriate to power the dynode scheme properly. A 50 Ω termination was required to observe the PMT signals on the oscilloscope. In subsequent trials we took advantage of the 2430A s internal 50 Ω termination. One important factor with the visible tubes turned out to be their extreme instability when waiting time between switching on or adjusting the high voltage and taking data. The first plateau attempts had revealed a very large standard deviation, ranging from 16 to 30 percent, for each set of 15 readings that were averaged to produce the data points. A number of trials revealed a one-percent stability level when using a wait time of about 75 minutes; a 90-minute wait time was used from then on in the visible trials at room temperature. TABLE II: Correlation between numerical labels and Hamamatsu serial numbers. Model PMT No. Serial No. R6095 Visible 1 CA2945 R6095 Visible 2 CA3560 R6095 Visible 3 RA3966 R6095 Visible 4 RB0286 R6095 Visible 5 RB0270 R6095 Visible 6 RB0268 R6095 Visible 7 RB0273 R6095 Visible 8 RB0278 R Infrared 1 GA3758 R Infrared 2 GA3756 R Infrared 3 GA3760 R Infrared 4 GA3727 R Infrared 5 GA3733 R Infrared 6 GA3753 R Infrared 7 GA3754 R Infrared 8 GA3724 Model R6095: Visible phototubes. Model R316-02: Infrared phototubes. V. ACKNOWLEDGEMENTS I wish to thank Professor Giovanni Bonvicini and postdoc Mikhail Dubrovin of Wayne State University for their extensive help and advice on this project. The staff of Wilson Laboratory for Nuclear Studies at Cornell were also of considerable assistance. TABLE I: Electronics modules used for PMT calibration. Manufacturer Model Function BNC BH-1 Tail pulse generator Global Specialties 4001 External pulse generator Hewlett-Packard 6516A DC power supply LeCroy/LRS 222 Dual Gate Generator LeCroy/LRS 334 Quad Amplifier LeCroy/LRS 370 Strobed coincidence LeCroy/LRS 621AL Quad discriminator LeCroy/LRS 621BL Quad Discriminator LeCroy/LRS 3001 qvt multichannel analyzer ORTEC 474 Timing Filter Amplifier ORTEC 935 Quad discriminator Tektronix 466 Analog storage oscilloscope Tektronix 2430A Digitial oscilloscope Tenellec TC 453 CF Discriminator Failed to work properly FIG. 1: Illustration of beam-beam pathologies, taken from Ref. [1]. The ellipses represent the electron and positron beams as viewed on end; incomplete overlap is said to waste luminosity since not all of the beam particles have the opportunity to collide and annihilate to produce interesting events. APPENDIX A I worked closely with WSU postdoc Mikhail Dubrovin and professor Giovanni Bonvicini for the extent of my project. When work started this summer the task at hand was clear, but we needed to acquire equipment to make it work. What we did have was a set of sixteen PMTs and caps from Hamamatsu and a set of drilled plates from the Wayne machine shop designed to hold

3 3 FIG. 2: A typical PMT noise signal. These were observed with the tubes powered to a nominal high voltage and with a 50 Ohm termination at the oscilloscope. FIG. 3: A single-photon multichannel analyzer peak for infrared PMT number 1. the tubes during testing. What we lacked was voltage dividers to properly power the tubes, a power supply, a dark box to shut out ambient light, a light source, and electronic equipment necessary to analyze the PMT output. 1. A lesson in elementary electronics Before attempting any tests, it was necessary to acquire voltage dividers so that the tubes diode structure could be held at the proper potential levels. Since a set of Hamamatsu dividers that were on order had yet to arrive, we elected to attempt the production of the necessary dividers ourselves, using the bare caps supplied with the tubes. This posed a challenge to me, since I had no previous experience with electronics, but after a few attempts (and purchasing a good text on introductory electronics) I was able to construct a divider that provided the needed potential differences across the various terminals. The very first trial, aimed at discovering whether we were able to observe the proper noise signal, was conducted with a single visible photomultiplier placed inside its original packaging and wrapped with aluminum foil to shield it from ambient light. A small Tektronix digital oscilloscope was used to observe a satisfactory noise signal from this first PMT (later labelled as number 1) using one of our homemade voltage dividers. This test that the tube was working, and also that the divider scheme taken from the Hamamatsu catalog was appropriate to power the tube in an acceptable manner. With this initial test completed, Dubrovin gave the goahead to produce the remaining fifteen dividers. Three more days were spent purchasing and soldering resistors, until I learned that we had received a shipment of dividers from Hamamatsu, and so the production of dividers was abandoned. 2. Optical insulation Dubrovin and I then turned to the task of mounting all eight visible photomultipliers in a dark box, using a set of drilled plates from the Wayne State machine shop. Our first discovery was that the voltage dividers could not be mounted on the plates because the holes were too big. Therefore, I obtained some aluminum bar stock and produced a set of bars that could be screwed to the back plate, and the dividers then screwed to the bars. The question of a dark box was settled with about 16 ft 2 of 1/32 sheet aluminum, which was formed into a 9x9x42 box, sealed with liquid rubber and foam weatherstripping, and painted at black inside and out to minimize the ambient light level surrounding the phototube array. 3. NIM electronics Finally, it was necessary to accumulate the needed electronics to allow meaningful readouts of the PMT signals, but we were much plagued with difficulties in this quest. The small oscilloscope we had used previously was no longer available, and a two-channel analog oscilloscope lent to us by the electronics shop proved difficult to live with under constantly changing test conditions, so we borrowed an additional digital oscilloscope with only one working channel to ease general signal analysis tasks. The need for a discriminator, which converts the ana-log signal from the tubes to a logic signal, caused us to turn to an old NIM crate that was present on our bench. However, this proved a major source of frustration, since neither the LeCroy 621BL four-channel discriminator nor the coincidence module already mounted showed no signs of life, and sorting through a large pile of older discriminators proved a major source of frustration until we realized the crate was not in working order. With the old crate in the electronics shop, we resumed work with a working crate that Bonvinicini brought from Wayne State. We were finally able to locate a handful of discriminator modules that worked intermittently, and in the meantime, we obtained a two-channel scalar (pulse

4 4 counter) in the CHESS cabinet (after first trying four other scalars that didn t work) that would display the number of counts collected in a period of 0.1s, 1s, or 10s. The 10s setting was used exclusively for my data runs to minimize the effect of small fluctuations in counting rate. 4. Getting started With some working electronics in hand, we were finally able to begin work on taking high voltage plateaus of the photomultiplier tubes. The plateaus are simply plots of scalar count rate vs. high voltage; Fig. 7 represents my first attempt at taking a plateau, as well as the bestlooking plateau produced. However, at this stage, the more responsible experimental procedure of taking multiple readings for each data point revealed a very large standard deviation ( 30 percent) over the data set for each point. An extra plot made from the next data collection run using 1σ error bars revealed enormous uncertainty in the actual slope of the plot. Possible sources of error were hypothesized at this point to include variations in the high voltage supply and instability in the tubes themselves, since this dynode structure is known to fire considerably on its own immediately after changes in its high voltage input. Therefore, an important question at this point was how long we needed to wait for the tubes to stabilize. I conducted a series of tests in which I powered the tubes to 1450V and took readings after 20 minutes, and again after 8 hours (other intervals would have been ideal, but I was away at a picnic during this time). The initial reading, with fifteen readings for each data point, revealed an average standard deviation of about 16 percent, while the latter showed an average deviation of one percent. Later trials revealed the ideal waiting time to be between 60 and 90 minutes, and so the 90 minute wait time was used from then on. [1] G. Bonvicini, et al., Beam-beam collision monitoring for CESR: A Large Angle Beamstrahlung Monitor [2] N. Detgen, et al., CBN-99-26, 1999 [3] G. Bonvicini, D. Cinabro, and E. Luckwald, Phys. Rev. E 59: 4584, 1999 [4] G. Bonvicini and M. Dubrovin, personal communication

5 5 FIG. 4: View of eight PMTs as installed in the dark box, which is in turn resting inside our polystyrene cold-box. The light source was placed at the far end of the box, as was the liquid nitrogen for cooling. FIG. 5: Electronics scheme without coincidence module. Data were taken first with light source on, then off, and then subtracted. The crate pulse generator was used to trigger the external pulse generator because of its superior adjustability. Only two PMTs are shown because we had only two scalar channels. A small multiplex box was also employed to permit easy tube selection without mixing up or damaging the signal cables.

6 6 FIG. 6: Electronics scheme with coincidence module. An additional signal was taken from the crate generator to provide a reference pulse to the coincidence. FIG. 7: First attempt at taking a PMT plateau, using visible PMT number 1. This was the best-looking plateau produced.