1、29th International Cosmic Ray Conference Pune (2005) 00, 101106Calibration of the MAGIC TelescopeM. Gauga, H. Bartkob, J. Cortinaa, J. Ricoa(a) Institut de Fsica dAltes Energies (IFAE), 08193 Bellaterra, Barcelona, Spain(b) Max-Planck Institute fur Physics, 80805 Munich, GermanyPresenter: J. Rico (j
2、ricoifae.es), spa-jrico-J-abs2-og27-posterThe MAGIC Telescope has a 577 pixel photo-multiplier tube (PMT) camera which requires precise and regularcalibration over a large dynamic range. A system for the optical calibration consisting of a number of ultra-fast and powerful LED pulsers is used. We ca
3、librate each pixel using the F-Factor Method with signals inthree different wavelengths. The light intensity is variable in the range of 4 to 700 photo-electrons per PMT.We achieve an absolute calibration by comparing the signal of the pixels with the one obtained from a 1 cm2PIN diode. This device
4、is calibrated with the emission lines of two different gamma-emitters (241Am and133Ba) which produce a precise reference signal in the active region of the PIN diode. The time resolutionof the entire MAGIC read-out system has been measured to about 700 ps at intensities of 10 photo-electronsreaching
5、 200 ps at 100 photo-electrons. With an external calibration trigger, it is possible to take calibrationevents interlaced with normal data at a rate of 50 Hz. The entire system has been used on-site for one year.1. IntroductionThe MAGIC Telescope 1 houses a camera of 397 inner pixels (0.1 ) and 180
6、outer pixels (0.2 ), each readout with 300 MSample/s flash-ADCs 2 and an optical link of 260 MHz bandwidth to transfer the electronicsignal over 160 m to the counting house. The quantum efficiency (QE) of the MAGIC PMTs strongly depend onthe incident wavelength. Moreover, differences in the exact sh
7、ape of QE( ) between PMTs had been observed.It is therefore desirable to calibrate the PMT response at different wavelengths.We use a system of very fast (34 ns FWHM) and powerful (1081010 photons/sr) light emitting diodes 3(NISHIA, single quantum well) in three different wavelengths (370 nm, 460 nm
8、 and 520 nm) and differentintensities (up to 700 photo-electrons per inner pixel and pulse) so that we are able to check the linearity of thewhole readout chain.2. Excess noise factor methodIn the last year, the camera was calibrated using the F-Factor. Assuming that the number of photons impingingo
9、n the photo-cathode has a Poisson variance, that the photon detection efficieny is independent from the placewhere and under which angle the photo-electron is released and that the excess noise introduced by the readoutchain does not depend on the signal amplitude, one can derive 4: Nphe = F 2 2=( 2
10、1 20)Here, 0 describes the signal extractor resolution (mostly due to noise of the night sky background), 1 themeasured standard deviation of the signal peak and is mean reconstructed (and pedestal-subtracted) signal.F stands for the excess noise factor, previously measured in the lab.The method yie
11、lds one value of Nphe per pixel the average of which is used to extract an averaged photo-electron fluence per light pulse and inner pixel . All reconstructed signals from data are then multi-plied with a conversion factor ciphe = Riarea= i where i stands for the pixel index and Riarea for theratio
12、of covered areas. Riarea is 1 for all inner pixels and 4 for all outer pixels.Figure 1 left shows the evolution of the mean number of photo-electrons over one night. One can see that,2 M. Gaug et al.despite some long-term dependency, remains stable to about 1% on minute scales. More preciseinvestiga
13、tions revealed a dependency of the light output with ambient temperature of about 2 %=K.Figure 1 right shows , calculated at different intensities and with different light colours.Time min.0 50 100 150 200 250 300 350 1pheFADC cntsMean Charge 1pheF-Factor Method PIN Diode Methode10 Leds Blue 291 3 2
14、94 320 Leds Blue 605 6 613 6Table 1. Average number of photo-electrons for one pixel, calculated with both calibration methods for two different lightintensities. For the PIN Diode Method, an additional systematic error of +8 6% and for the F-Factor Method, 5% pluspossible degradations of the overal
15、l PMT quantum efficiency and transmission coefficients have to be added.channel0 50 100 150 200 250counts10102103104 20 Leds Blue10 Leds BlueAm (60 keV)241Ba (81 keV)133Ba (Comp. edge 207 keV)133Figure 2. Spectra taken with the PIN Diode and various sources: bottom: 241Am with a 60 keV photon emissi
16、on line,center: 133Ba with a photon emission line at 81 keV and a Compton egde at 207 keV, top: the spectrum obtained from lightpulses with two combinations of LEDs.Using the light pulser at different intensities, we measured the time offsets and time spreads of the readout anddetection chain. Event
17、 by event, the reconstructed arrival time difference of every channel with respect to areference channel was measured and its mean and RMS calculated. The former yields the measured relativetime offset while the latter is the convolution of the arrival time resolution of the measured and the referen
18、cechannel.Figure 3 (left) shows the time offset versus the applied HV for each PMT. Like expected, one can see a clearanti-correlation. The smaller the applied HV, the longer the signal takes to travel from the first dynode to theanode.Figure 3 (right) shows the time resolution (RMS of the arrival t
19、ime differences histogram, divided by the squareroot of 2), measured at different intensities. The measurements have been fitted by the following function:T =sT 21=phe +T 222 =phe2 + T20 : (1)where T1 parameterizes the combination of intrinsic arrival time spread of the photons from the light pulser
20、and the transit time spread of the PMT, T2 parameterizes the reconstruction error due to background noise andlimited extractor resolution and T0 is a constant offset which might be due to the residual FADC clock jitter.4 M. Gaug et al.HV V950 1000 1050 1100 1150 1200 1250 1300 1350 1400T ns (arb. re
21、f.)-4-3-2-101234mean signal ph.el.10210 310Time Resolution ns-1101/ ndf 2 49.92 / 10ns: 1T 0.16 1.69 ns: 2T 0.78 5.68 ns: 0T 0.02 0.21 Green Pulses Inner PixelsBlue Pulses Inner PixelsUV Pulses Inner PixelsFigure 3. Left: Calibrated arrival time offsets T vs. applied HV. Right: Calibrated time resol
22、ution for various intensities.5. ConclusionsThe LEDs pulser calibration system has been used to calibrate the MAGIC camera for about one year. If runin standard mode, it fires UV-pulses of 370 nm in dedicated calibration runs and additionally as interlacedcalibration events with a frequency of 50 Hz
23、. This frequency allows to accumulate enough statistics beforereaching the typical time scales of residual short-term fluctuations of the optical transmission gains. Thecamera is such continuously monitored and re-calibrated 6 using the F-Factor method.In dedicated calibration runs with different co
24、lours and intensities, the response of the system and signal ex-traction methods 7 have been tested.Using a calibrated PIN Diode, the light flux of the pulser has been measured with an independent methodyielding consistent results with the F-Factor method. The systematic error of both methods is sti
25、ll above 5%(8% for the PIN Diode method) and will be reduced in the future.Using the light pulses to calibrate the arrival times extracted from each channel, we find an upper limit to thetime resolution of the MAGIC telescope for cosmics pulses of:Tcosmics s4 ns2=phe +20 ns22 =phe2 + 0:04 ns2: (2)Re
26、ferences1 R. Mirzoyan for the MAGIC coll., these proceedings2 F. Goebel et al., Proc. of the 28th ICRC, Tokyo (2003)3 T. Schweizer et al., IEEE Transactions on Nuclear Science, San Diego (2001)4 R. Mirzoyan and E. Lorenz, Proc. of the 25th ICRC, Durban (1997)5 L. Holl, E. Lorenz, G. Mageras, IEEE Transactions on Nuclear Science, Vol.35, Nr. 1 (1988)6 J. Cortina, these proceedings7 H. Bartko, M. Gaug, A. Moralejo, N. Sidro, astro-ph/0506459
