Instrumentation for Colliding Beam Physics 2017 Laboratory tests of the response stability of the ATLAS Tile Calorimeter photomultipliers High performance of the ATLAS Tile Calorimeter response is achieved with a multi-stage calibration. One step of the calibration is based on measurements of the response to laser pulse excitation of the PMTs used to read out the calorimeter cells. A facility to study the PMT stability response is operating in the PISA-INFN laboratories since 2015. Goals of the tests are to study the time evolution of the PMT response as a function of the integrated anode charge and to compare test bench results with the observed response drifts of the Tile Calorimeter PMTs during LHC Run I and Run II. A new statistical approach was used to measure the drift of the absolute PMT gain. A new procedure which combines studies of the time evolution of the global PMT responses and of the individual PMT gains was adopted to derive the evolution of the cathode quantum efficiency. The experimental setup of the Pisa facility is described and the first results obtained by testing about 30 PMTs Hamamatsu model R7877 (a special evolution for Tile of the R5900 model,10 stages, 20 mm x 20 mm active cathode surface) are presented. The statistical method for measuring the PMT absolute gain The daily PMT data acquisition and excitation loop The experimental setup 6 x 10000 laser pulses @ 2KHz, variable intensity 10000 laser pulses @ 2KHz, maximum intensity 8500 seconds laser or diode pulsing @ 200 kHz, maximum intensity Each daily measurements consists of 9 cycles. In each cycle: - 10k events with laser pulsing are acquired at maximum intensity; - 1k laser events are acquired with 6 different OD filters in the wheel for the intensity scan; - 10k events with LED pulsing are acquired; - The cycle is completed with about 2,5 hours laser or LED pulsing for integrating anode charge without data storing. 10000 diode pulses @ 2KHz, maximum intensity The simplified model: - all electronic noise contributions to the signal fluctuations can be neglected; - only the photo statistics and the laser intensity fluctuations contribute to the signal variance: <q> is the average anode charge; <N> is the average number of p.e.; <I> is the average light intensity of a pulsed source on the cathode; f is the noise excess factor. If G is the PMT gain at a given voltage (e is the electron charge) In the Poisson statistics and where is the source coherence factor. The factor k can be statistically evaluated with two methods: Here n and m refer to the same PMT response at two different light intensities. Alternatively: Here qi and qj are the anode charges of any PMT pairs receiving similar fractions of the same light pulse at fixed intensity. Fiber bundle Patch panel Ref. PMT0 PMT signals are acquired PMT signals are not acquired, just anode charge integration Beam expander 1 cycle = 2.5 hours x 9 = 1 day loop = 22.5 hours Ref. PMT1 1 𝑽𝒂𝒓(𝒒) <𝒒> 𝟐 =𝒇∙ 𝝈 𝑵 𝟐 <𝑵> 𝟐 + 𝑽𝒂𝒓 𝑰 <𝑰> 𝟐 . Figure 3. PMT excitation cycle and daily loop. PMT absolute gain measurements Time stability of the PMT response & gain Filter Wheel <𝒒> = 𝑵∙𝑮∙𝒆 𝝈 𝑵 𝟐 = <𝑵> Figure 4. PMT response in intensity scans Figure 6. PMT integrated anode charge (C). Laser head 2 𝑽𝒂𝒓(𝒒) <𝒒> =𝒇∙𝑮∙𝒆+𝒇∙ 𝑽𝒂𝒓 𝑰 <𝑰> 𝟐 ∙<𝒒> Examples of the variance divided by the average value of the pulse height distribu- tion of a tested PMT as a function of its average signal <q> in: a laser intensity scan (red points); a diode intensity scan (blue points). A linear fit is superimposed assuming the simplifies model of equation (2); The coherence parameter  = Var(I)/I2 is expected to be less than 0.1% for our laser model and to vanish for an incoherent source. In this case k(diode) ~10-5 +/- 10-5; The value of Var(q) / <q> at <q> = 0 is proportional to the PMT gain; The gain values obtained with the two different gains are in agreement within 5%; Error bars include statistical and systematic errors. 3 𝒇∙𝑮∙𝒆= 𝑽𝒂𝒓 𝒒 <𝒒> −𝒇∙∙<𝒒> Figure 1. The optics box. Since 2015 in the Pisa-INFN labs the response of the Tile PMTs is studied with the set-up in Fig.2 - Light sources are (alternate operation): a) a 437 nm laser, 80 ps pulse width b) a 532 nm LED, 150 ns pulse width -Laser beam intensity is varied with a remote controlled filter wheel and monitored with two reference PMTs. - Laser pulses are used to measure the response of the PMTs under test. -The monitor PMT signals are used for normalization purposes - LED pulses are used to excite and to integrate large amounts of anode charge for the PMTs under test in the PMT box - The light from the sources is expanded and fed to the PMT box through a white fiber bundle. - All PMT signals are digitized with 12 bit charge integration ADCs 𝑽𝒂𝒓(𝑰) <𝑰> 𝟐 = Figure 7. Variation of the PMT response normalized to the reference PMT0 and to the first observation day. The measured average down drift is about -0.2% / C, consistent with the drift observed with PMTs on the detector. 4 = 𝑽𝒂𝒓( 𝒒 𝒏 ) < 𝒒 𝒏 > − 𝑽𝒂𝒓( 𝒒 𝒎 ) < 𝒒 𝒎 > < 𝒒 𝒏 >− < 𝒒 𝒎 > . (5)  = 𝑪𝒐𝒗( 𝒒 𝒊 , 𝒒 𝒋 ) < 𝒒 𝒊 >< 𝒒 𝒋 > Figure 8. Variation of the PMT absolute gain normalized to the first observation day. Central values are stable in time within 1%. Conclusions & future plans Figure 5. PMT gain at different HV values. A statistical approach can be used to measure absolute gain of PMTs; Good agreement between two different methods for PMT absolute gain measurements, the covariation method is more accurate; A down-drift of PMTs response is seen and it seems due to a cathode Q.E. rather than to a gain loss. Next steps: Repeat measurements with new green pulsed laser Compare PMT linearity response with passive and active HV dividers; Repeat measurements with PMT dismounted from the detector and having received different doses during LHC run I and run II The PMT gain (3) is calculated with: - the intensity scan method (open circles) (equation (4) ), - the covariance method (full circles) (equation (5) ); On day 20/01/2017 the PMT HV was increased from 700 V to 830 V; The expected increase of the gain by a factor about 3 is measured in all cases; The covariance method appears to be more precise, but a very good general agreement between the two methods is observed; No measurable gain drift is seen in the observation period of PMTs excitation. Figure 2. Experimental setup layout. Vassili Kazanin1, Sandra Leone2, Fabrizio Scuri2 for the ATLAS Tile hadron calorimeter group 1) Budker Institute of Nuclear Physics, Novosibirsk, Russian Federation 2) INFN sezione di Pisa, Italia