Design and realization of a facility for the characterization of Silicon Avalanche PhotoDiodes
Andrea Celentano, Luca Colaneri, Raffaella De Vita, Stuart Fegan, Giuseppe Mini, Gianni Nobili, Giacomo Ottonello, Franco Parodi, Alessandro Rizzo, Irene Zonta
aa r X i v : . [ phy s i c s . i n s - d e t ] A p r Design and realization of a facility for thecharacterization of Silicon AvalanchePhotoDiodes
Andrea Celentano ∗ , Luca Colaneri , Raffaella De Vita , StuartFegan , Giuseppe Min`ı , Gianni Nobili , Giacomo Ottonello ,Franco Parodi , Alessandro Rizzo , and Irene Zonta INFN, Sezione di Genova, Via Dodecaneso 33, Genova, Italy INFN, Sezione di Roma Tor Vergata, Via della RicercaScientifica 1, Roma, ItalyPublished in: JINST 9, T09002, 2014
Abstract
We present the design, construction, and performance of a facil-ity for the characterization of Silicon Avalanche Photodiodes in theoperating temperature range between -2 ◦ C and 25 ◦ C. The systemcan simultaneously measure up to 24 photo-detectors, in a completelyautomatic way, within one day of operations. The measured data foreach sensor are: the internal gain as a function of the bias voltage andtemperature, the gain variation with respect to the bias voltage, andthe dark current as a function of the gain. The systematic uncertain-ties have been evaluated during the commissioning of the system tobe of the order of 1%.This paper describes in detail the facility design and layout, andthe procedure employed to characterize the sensors. The results ob-tained from the measurement of the 380 Avalanche Photodiodes ofthe CLAS12-Forward Tagger calorimeter detector are then reported,as the first example of the massive usage of the facility. ∗ Corresponding author: [email protected] Introduction
Silicon Avalanche PhotoDiodes (APDs) have become in recent years a conve-nient and common photo-detector choice because of the compact size, goodradiation hardness, and insensitivity to intense magnetic fields. AvalanchePhotoDiodes replaced traditional photo-multipliers tubes in many particlephysics detectors. APDs have been successfully used in the CMS detectorat the CERN LHC, employing more than 150k APDs in the electromagneticcalorimeter [1], and in the ALICE Photon Spectrometer (PHOS), equippedwith 18k sensors [2]. The PANDA detector at FAIR foresees to use more than15k Large Area APDs for the electromagnetic calorimeter readout, operat-ing them at − ◦ C [3]. Other significant examples of large-scale detectorsemploying APDs for light readout include the Forward Tagger Calorimeterin Hall-B at Jefferson Laboratory [4] and the HPS-Ecal Calorimeter [5], bothcurrently under development.The intrinsic gain and the dark current of Avalanche PhotoDiodes stronglydepends on the bias voltage and on the operating temperature. It is thereforenecessary to characterize them before installation in any experimental setupto be able to work at the selected working point. This is even more criticalfor detectors employing Avalanche PhotoDiodes for multiple channel read-out, as for electromagnetic calorimeters, since differences between workingpoints could induce degradation of the full detector performances. Mappingthe gain of each sensor also permits to group together those with close work-ing point and supply them with the same bias voltage, thus reducing thenumber of independent high voltage channels.In this paper we present our realization of a new a facility for the au-tomatic characterization of Avalanche PhotoDiodes in the − ◦ C - 25 ◦ Ctemperature range. The facility was first developed to measure the 380Large Area APDs (model Hamamatsu S8664-1010) employed in the CLAS12-Forward Tagger Calorimeter. This was the first massive usage of the system.The good performances obtained validated the design of the facility, whichwas used again later to characterize the 516 Large Area APDs employed inthe HPS-Ecal detector.
The procedure employed in the facility to measure the APDs internal gain( G ) at fixed temperature ( T ) as a function of the bias voltage ( V ) is as2 (Volt)0 50 100 150 200 250 300 350 400 I ( n A ) -1 G a i n -1 off I on IGain
Figure 1:
Typical LA-APD photo-current I on (closed black markers) and darkcurrent I off (open black markers) behaviour, as a function of the bias voltage V (these data were obtained measuring a LA-APD, model Hamamatsu S8664-1010). The corresponding internal gain G is also reported (red markers). Thismeasurement was performed at 18 . ± . ◦ C, in the voltage range between 0 and400 V. follows. The dark current I off and the photo-current I on due to a continuousillumination are measured at different values of the reverse bias voltage witha picoammeter. The gain is derived using the following equation: G ( V ) = I on ( V ) − I off ( V ) I on ( G = 1) − I off ( G = 1) , (1)where I on ( G = 1) and I off ( G = 1) are, respectively, the values of the photo-current and the dark current when the internal gain of the APD is equalto one, i.e. when the bias voltage is sufficiently low so that the avalanchemechanism is not active. Since the procedure employed to calculate the gaininvolves the normalization to the unitary gain current, the only requirementsfor the light source are the stability during the measurement time and anintense light emission to provide a significant I on ( G = 1) current, not lowerthan ≃
10 nA, to guarantee a good signal over noise ratio in the currentmeasurement. Being the spectral response of the sensor typically peaked at420 nm, a standard blue LED can be employed.The typical behavior of an APD photo-current I on and dark current I off is shown in Figure 1. After an initial plateau, which corresponds to the uni-tary gain, I on grows rapidly for increasing bias voltage. I off remains smallerthan I on by a factor ≃ − ain20 40 60 80 100 120 140 160 180 200 220 240 D a r k C u rr e n t ( n A ) Bias voltage (V)360 365 370 375 380 385 390 395 400 / G d G / d V ( % / V ) Figure 2:
Left: LA-APD dark current I off as a function of the LA-APD gain G . Right: LA-APD intrinsic gain variation as a function of the bias voltage. Thecorresponding sensor gain is also reported on the top axis. (1) thus corresponds to the initial plateau of I on . To reduce the measurementtime, the voltage scan is performed with non-uniform steps in the above men-tioned bias range, with denser measurements in the two regions of interest:the unitary-gain plateau and the neighborhood of the foreseen working point( G = 150 for the specific example here reported). The corresponding inter-nal gain G is also shown in Figure 1. The unitary gain plateau ( G = 1) isvisible for bias values lower than ≃
100 V, while for bigger values G increasesrapidly. The measured dark current I off is reported in Figure 2, left panel,as a function of the intrinsic gain G . It shows a linear behavior, as expectedfrom the most common interpretation of a LA-APD dark current as a sumof two terms: the first (“surface current”) being independent of G , and thesecond (“bulk current”) being directly proportional to it [3].The relative APD gain dependence on the bias voltage is evaluated nu-merically, by computing the corresponding incremental ratio:1 G ∂G∂V ≃ G ( V ) G ( V + ε ) − G ( V ) ε , (2)with ε = 0 . G ( V + ε ) is evaluated from the measured G ( V ) data points,employing a spline interpolation. A typical result is shown in Figure 2, rightpanel. The temperature dependence of the APD intrinsic gain and dark currentis investigated by comparing measurements performed at different tempera-4
Bias voltage (V)360 365 370 375 380 385 390 395 400 G a i n C ) ° T e m p e r a t u r e ( Figure 3:
Left: LA-APD intrinsic gain G for different temperatures. + : − . ◦ C, ∗ : 2 . ◦ C, ◦ : 6 . ◦ C, × : 10 . ◦ C, • : 14 . ◦ C, N : 18 . ◦ C, H : 22 . ◦ C. Right: LA-APD gain (color-scale), as a function of the bias voltage V andoperating temperature T . The isogain curves are reported in black. tures. Figure 3, left panel, shows an example for a sensor characterized inthe temperature range between − . ◦ C and +22 . ◦ C. The same data arereported in Figure 3, right panel, linearly extrapolated in the whole temper-ature range. We observed that the isogain curves in the V − T plane areapproximately straight lines (in the region where G >
1, i.e. outside the uni-tary gain plateua). Therefore, the APD gain can be actually parametrizedas a linear combination of these two variables: G ( V, T ) = G ( αV − βT ) ≡ G ( x ) , (3)with α and β two proper coefficients to be measured experimentally, and x = αV − βT .We exploited this behavior to derive the bias voltage V correspondingto a given gain G , at temperature T , as follows. First, the photo-sensorgain is measured at a few different temperatures T i , and the bias voltage V i corresponding to a certain fixed gain G is evaluated. Then, the ratio k ≡ α/β is calculated by performing a best-fit with a linear function to the T i vs V i data points. The procedure is repeated for different gain values, to check thestability of the parameter k . An example is reported in Figure 4, left panel.The corresponding value of k is 1 . ± . V is finally calculated as: V = V i + k − · ( T − T i ) , (4)where V i is the bias voltage corresponding to G at any of the temperatures T i . 5he intrinsic gain relative variation with respect to the operating tem-perature follows from Eq. 3: (cid:26) ∂G∂V = αG ′ ( x ) ∂G∂T = βG ′ ( x ) ⇒ G ∂G∂T = k − G ∂G∂V (5)
We identified a set of quality-checks to ultimately decide if a measured sensoris properly operational or not. Any APD not fullfilling any one of theserequirements has to be characterized again and, if still presenting anomalies,discarded. • The measured dark-current I on and the photo-current I off as a functionof the bias voltage V must be checked for all the measured tempera-tures. These data are required to have, qualitatively, the same behavioras reported in Figure 1, i.e. the photo-current must show a plateau for V .
50 V, and then grow with approximately an exponential behavior.The dark-current I off should follow the I on behavior, and always belower than it by at least one order of magnitude. • The dark-current behaviour I off as a function of the photo-detectorgain G has to be checked for all the measured temperatures. It shouldexibit a linear dependence, I off = c + c G , and the constant contribu-tion c should decrease with the operating temperature. • The relative gain variation with respect to the bias voltage G dGdV mustbe finally checked, as a function of the gain G . Results obtained atdifferent temperatures should approximately superimpose, for G & The layout of the facility for automatic APD characterization that we devel-oped is shown in Figure 5. The facility can measure up to 24 sensors at once,characterizing them in the voltage range between 0 and 500 V. The operatingtemperature can be varied from -2 ◦ C to +25 ◦ C. The measurement at eachtemperature takes approximately 2.5 hours. Therefore, one day is required6 ias Voltage (V)350 360 370 380 390 400 C ) ° T e m p e r a t u r e ( G=50 G=100 G=150 G=200
Gain20 40 60 80 100 120 140 160 180 200 ) - ( V d V d G G -3 × Figure 4:
Left: bias voltage corresponding to different LA-APD gain values, asa function of operating temperature. The curves are the results of the best fitsperformed with a linear function to the different data-sets to evaluate the LA-APD k parameter (the lowest temperature point for G = 50 was intentionallyexcluded from the fit). Right: Measured G dGdV vs G curves, for different operatingtemperatures. Same symbols as in the previous Figure. to fully characterize 24 sensors from the minimum to the maximum tem-perature, with 8 measurements at ≃ . ◦ C intervals. The facility has beenused to characterize the following APD models: Hamamatsu S8664-55 andHamamatsu S8664-1010. There are no limitations to characterize other APDmodels, providing they have the same mechanical layout for the connectionpins as the two aforementioned models.Each installed sensor is characterized using the procedure described inthe previous section. The dark current I on and the photo-current I off aremeasured for different bias voltages to obtain the internal gain. In orderto employ a single picoammeter to measure the currents, and a single high-voltage source, a custom current multiplexer was developed. The currentmultiplexer is the core of the facility (see Figure 6). During operation, theAPD cathodes are all connected to the common HV source via independentbias resistors, while the anodes are connected to the multiplexer inputs. Themultiplexer single output is connected to the picoammeter. In this way, onlyone APD at a time is physically connected to the instrument, while the other23 are coupled to ground. We designed the current multiplexer using mechan-ical relays (type SPDT, “single pole, double throw” ), because the measuredcurrents are of the order of 10-100 nA and other types of electronic switches, A SPDT relay is a 3-terminals device, with the common terminal connect to either ofthe two others, depending on the state of the magnetic coil.
Left: simplified scheme of the facility for LA-APD gain measurement.Right: picture of the facility, showing the two cold copper plates with holes forLA-APDs connection on the underlying PCB cards. One LA-APD is installed onthe left copper plate. ◦ C stability. Furthermore, during operation,the copper assembly is inserted in a vacuum- and light-tight box flushed withnitrogen to prevent moisture formation.The system, controlled via a Labview program running on the DAQ PC,works as follows. During the initial setup, the user enters the values for thetemperature scan and the IDs of the sensors under measurement. For eachprogrammed temperature, the APDs are individually enabled and charac-terized. At the end of the full cycle data are analyzed to calculate the gainof the APDs as a function of the bias voltage and temperature, the darkcurrent dependence on the gain, and the relative gain variation at the differ-ent temperatures. The raw data and other relevant parameters, such as thebias voltage corresponding to the nominal gain at room temperature and therelative variation with respect to the temperature and the bias voltage in theneighborhood of the nominal working point, are then recorded to a file. Aset of histograms is also produced, to let the user perform the quality checkpreviously described.
During the facility commissioning we investigated the relevance of possiblesystematic effects in the measurement, such as those due to long term fluc-tuations introduced by temperature drifts, residual moisture formation, andvariation in the LED emission. The gain of the same LA-APD has been9igure 6:
Left: Schematic of the gain measurement circuit (just three channelsare reported). “HV” is the common APDs high voltage source, while “A” is thepicoammeter. Right: the current multiplexer. anufacturer Bias Voltage (V)402 404 406 408 410 412 414 416 418 420 422 C ( V ) ° B i as V o l t a g e G = @ Bias Voltage (V) ∆ Figure 7:
Left: correlation between the measured bias voltage for G = 150 at +20 ◦ C ( y axis) and the value reported by the manufacturer for G = 50 at +25 ◦ C ( x axis). Right: absolute bias voltage difference for G = 150 between the measureperformed at 0 ◦ C and the extrapolated result from +20 ◦ C. repeatedly measured over a time period of 15 days, and the bias voltage cor-responding to a fixed gain at room temperature ( G = 150 @ T = +18 ◦ C forthe specific sensor) has then been calculated for each measurement. Resultsshowed that the working point was constant within less than 0.1 V for allmeasurements. This corresponds to a relative error on the measured gainapproximately equal to 0 . G ∂G∂V ≃ /V in the neighborhood ofthe LA-APD selected working point ( G = 150). The same systematic effectwas found on the other measured quantities, such as the dark current andthe k parameter. The first large-scale use of the APD characterization facility was to measurethe LA-APD employed in the FT-Cal detector of the CLAS12 experimentat Jefferson Laboratory. The 380 sensors were characterized using the abovefacility within one month. Results were analyzed to identify and select theproperly working sensors. As a result, 17 samples were discarded, since theydid not fulfill the previously described data-quality checks, either after a sec-ond measurement. Figure 7, left panel, shows the comparison between themeasured bias voltage corresponding to APD gain G = 150 at +20 ◦ C, andthe value declared by the manufacturer for the working point G = 50 at +25 ◦ C, for the accepted samples. The good correlation demonstrates the relia-bility of the system. The right panel, instead, shows the absolute difference11etween the bias voltage measured for G = 150 at 0 ◦ C and the correspondingvalue extrapolated from the measure at 20 ◦ C, using the procedure previouslydescribed. The average difference of ≃ ≃
2% (see Figure 4). This is a worst-case scenario,since in reality APDs are characterized at temperatures close to the foreseenworking point, thus reducing the extrapolation error. The result, however,proves the reliability of the gain extrapolation procedure in the whole (
V, T )parameter space.
We developed a facility to automatically characterize Avalanche PhotoDiodesin the -2 ◦ C - 25 ◦ C temperature range. Such a facility can characterize upto 24 sensors at once, within one day of operation, measuring APDs fromthe minimum to the maximum temperature in ≃ . ◦ C intervals . Foreach sensor, the bias voltages corresponding to the nominal gain of 150 atdifferent temperatures, and the relative gain variation with respect to thevoltage and temperature are recorded. Particular care was taken to checkpossible systematics on the measured gain, that have been evaluated to beof the order of 0 . G = 150. The authors want to acknowledge the electronic group of INFN-Genova sec-tion for the support given during the design of this facility and A. Balbi forhelp during the construction.
References [1] CMS Collaboration,
Performance and operation of the CMS electromag-netic calorimeter , Journal of Instrumentation (2013) T03010[2] D.C. Zhou (for the ALICE Collaboration), PHOS, the ALICE-PHOtonspectrometer , Journal Of Physics G: Nuclear and Particle Physics (2007) S719[3] W. Novotny et al. (PANDA collaboration), The ElectromagneticCalorimetry of the PANDA Detector at FAIR , Journal of Physics: Con-ference Series (2012) 012063124] A. Celentano et al. (CLAS Collaboration),