1 / 76

Signal formation for energy, time and position measur

Advanced FEE solutions for large arrays of semiconductor detectors. Signal formation for energy, time and position measurements Segmented detectors; - advanced FEE for Ge Detectors Briefly, some specific issues and cases:

chandler
Télécharger la présentation

Signal formation for energy, time and position measur

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Advanced FEE solutions for large arrays of semiconductor detectors • Signal formation for energy, time and position • measurements • Segmented detectors; - advanced FEE for Ge Detectors • Briefly, some specific issues and cases: • ◦ MINIBALL & AGATA (& GRETINA) FEE for gamma rays • (CERN-Isolde & EU Tracking Array -LNL; GSI; Ganil) • ◦ LYCCA & TASISpec FEE for particles • (GSI -Calorimeter & Superheavy Element Spectroscopy) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  2. a) Signal formation for energy, time and position • measurements, • (we’ll limit our attention to capacitive & segmented detectors) • b) Related issues in segmented detectors • - dynamic range • - high counting rates • - induced signals & crosstalk - pros vs. conts • c) AGATA & MINIBALL – advanced FEE solutions • - Dual Gain CSP - for the central contact • - ToT method ( - combined dynamic range ~100 dB, up to 170 MeV) • - Transfer function, Induced signals, Crosstalk • - Applications: - Impurities concentration measurement; • - Cosmic ray direct measurement up to 170MeV equiv. gamma • LYCCA & TASISpec - FEE for DSSSD G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  3. A typical structure of a segmented, tapered and encapsulated, HP-GeDetector [- HV] (GND) + HV (~ kV/cm) Rp Ci (-) Central contact (Core) (- e ~ mm) Exterior contacts (N Segments) (+) - Qi N = 6; 12; 18; 28; 36 • Standard n-type • Intrinsic HP-Ge (P-I-N) • Closed end • Coaxial structure • Io ~ < 100 [pA] • Cdet ~ 30 - 45 pF • Collection time ~ 30 - 1000 ns

  4. FFEFEE [HP-Ge + CSP] +Analog Nuclear Electronics Spectroscopic Chain is used in order to extract the: E, t, position (r, azimuth) Fast pipeline ADC [DGF] FEE Fast Pipe line ADC [DGF] Analog E+T Filter Amplifier Chain Collected charge pulses (+ &-) Qd - delta t UCSP – exponential Pile-up of pulses t Digital Filters (Fast, Slow) UFA ~ Gaussian t Baseline restorer

  5. [HP-Ge + CSP] +Digital Nuclear Electronics Spectroscopic Chain is used in order to extract the: E, t, position (r, azimuth) Fast pipeline ADC + PSA FEE Fast pipeline ADC & [DGF] Digital Filters [for Trigger, Timing, Energy, Position] Collected charge pulses (+ &-) Qd - delta t UCSP – exponential Pile-up of pulses t Digital Filters (Fast, Slow) UFA ~ Gaussian t Baseline restorer

  6. + Rp - Detector Detector Signal Collection • a gamma ray crossing the Ge • detector generates electron-hole pairs • charges are collected on electrode • plates (as a capacitor) building up • a voltage or a current pulse Z(ω) • Final objectives: • amplitude measurement(E) • time measurement (t) • position(radius, azimuth) Electronic Circuit Which kind of electronic circuit ; Z(ω)?

  7. Z(ω) Rp + - Electronic Circuit Detector Detector Signal Collection ifZ(ω) is high, • charge is kept on capacitor nodes and a voltage builds up (until capacitor is discharged) • Advantages: • Disadvantages: if Z(ω) is low, • charge flows as a current through the impedance in a short time. • Advantages: • Disadvantages: • limited signal pile up (easy BLR) • limited channel-to-channel crosstalk • low sensitivity to EMI • good time and position resolution • excellent energy resolution • friendly pulse shape analysis position • channel-to-channel crosstalk • pile up above 40 k c.p.s. • larger sensitivity to EMI • signal/noise ratio to low worse resolution

  8. Charge Sensitive Preamplifier • Active Integrator(Charge Sensitive Preamplifier -CSP) • Input impedance very high ( i.e. ~ no signal current flows into amplifier), • Cf/Rffeedback capacitor /resistor between output and input, • very large equivalent input dynamic capacitance, • sensitivityor~(conversion factor) A(q) ~ - Qi/ Cf • large open loop gain Ao ~ 10,000 - 150,000 • clean transfer function (no over-shoots, no under-shoots, no ringing) (Rf.Cf ~ 1ms) Ci ~ “dynamic” input capacitance tr~ 30-1000ns) - Qi Step function R f o Invert ing - Ao • Ci ~ 10 - 20,000 pF • ( up to 100,000) “GND” Non- Inv. + jFET GND Charge Sensitive Stage (it is a converter not an amplifier)

  9. Pole - Zero cancellation technique Rf . Cf ~ 1 ms Cf~ 1pF (0.5pF-1.5pF), Rf ~ 1GOhm Rd . Cd ~ 50 µs simpledifferentiation without Rpz Rpz~ 20 k Ohm Baseline shifts if (RfCf)= (Rpz .Cd) and RdCd ~ 50 µs differentiation with P/Z adj.  no baseline shifts with Rpz Cd~ 47 nF, Rd~1.1 kOhm Baseline restored

  10. Pole - Zero cancellation technique Rf . Cf ~ 1 ms Cf~ 1pF (0.5pF-1.5pF), Rf ~ 1GOhm Rd . Cd ~ 50 µs simpledifferentiation without Rpz Rpz~ 20 k Ohm Baseline shifts if (RfCf)= (Rpz .Cd) and RdCd ~ 50 µs differentiation with P/Z adj.  no baseline shifts with Rpz Cd~ 47 nF, Rd~1.1 kOhm Baseline restored

  11. Pole - Zero cancellation technique Rf . Cf ~ 1 ms Cf~ 1pF (0.5pF-1.5pF), Rf ~ 1GOhm CSP Rd . Cd ~ 50 µs simpledifferentiation without Rpz R pz ~ 21 k Ohm Baseline shifts if (RfCf)= (Rpz .Cd) and RdCd ~ 50 µs - clean differentiation with P/Z adj.  no baseline shifts with Rpz Cd ~ 47 nF, Rd ~1.1 kOhm Baseline restored

  12. This is only the ‘hard core’ of the CSP stage • (ChargeSensitivePreamplifier) but the FEE • must provide additional features: • a P/Z cancellation (moderate and high counting rate) • a local drive stage (to be able to drive even an unfriendly • detector wiring !) • (opt.) an additional amplifier (but with Gmax.~ 5) • (N.B. a “free advice”: … never install an additional gain • in front of the ADC ! -namely, after the transmission cable !) • a cable driver (either single ended –coax. cable or • differential output - twisted pair cable) Any free advice is very suspicious ( anonymous quote ) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  13. Block diagram of a standard CSP (discrete components and integrated solution… - what they have in common ) (alternatives) (alternatives) (+) Optionally with cold jFET (-) Warm part (outside cryostat) Cold part (cryostat) (alternatives) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  14. Block diagram of a standard CSP (discrete components and integrated solution… - what they have in common ) (alternatives) (alternatives) (+) Optionally with cold jFET (-) Warm part (outside cryostat) Cold part (cryostat) • tr 25 ns ( 1 - 200 ) ns • tf 50 μs ( 10 - 100 ) μs • CSP- ‘gain’  50 mV / MeV (Ge) • (10-500 mV / MeV) (alternatives) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  15. tr~ 30-40 ns Ch.1 @ 800 mV - no over & under_shoot IF1320 (IF1331) (5V; 10mA)& 1pF; 1 GΩ also GRETINA Eurysis warm • Warm & cold jFET • DGF-4C(Rev.C) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  16. AGATA τopt~ 3-6 µs J.-F. Loude, Energy Resolution in Nuclear Spectroscopy, PHE 2000-22, Univ. of Lausanne • the equivalent noise • charges Qn assumes • a minimum when the • current and voltage • contributions are equal • current noise ~ (RC) • voltage noise ~ 1/(RC) • ~ Cd 2 • 1 /fnoise ~ Cd2

  17. Dynamic range issue (DC - coupled) • Factors contributing to saturation: • Conversion factor – ( step amplitude / energy unit [mV/MeV] ); • Counting rate [c. p. s.] and fall time; • The allowed Rail-to-Rail area [LV-PS] {(+Vc - Vc) – 2xΔf -2δFilt.} +Vc (+ Rail ) DC – unipolar (-) Saturation (+Vc) δFilter A(q) ~ - Qi/ Cf Δf+ ( forbidden region ) Linear range DC - bipolar DC coupled channel Δf- Saturation (-Vc) DC – unipolar (+) -Vc (- Rail) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest δFilter

  18. Dynamic range issue (AC - coupled) • Factors contributing to saturation: • Conversion factor – ( step amplitude / energy unit [mV/MeV] ); • Counting rate [c. p. s.] and fall time; • The allowed Rail-to-Rail area [LV-PS] {(+Vc - Vc) – 2xΔf -2δFilt.} +Vc (+ Rail ) Saturation (+Vc) δFilt A(q) ~ - Qi/ Cf Δf+ ( forbidden region ) AC -Unipolar (negative) Linear range AC -Unipolar (positive) BL shift Δf- AC coupled channel Saturation (-Vc) -Vc (- Rail) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  19. What to do to avoid saturation? Conts(“price”) • to reduce the “gain” Resolution ( Cf larger ) • to fix the base line asymmetric if DC coupled (expand:F~2), • but for AC ? (expand only: F~ 1.5)! • to reduce the fall time  Resolution ( Rf smaller ) • (OK only for high counting rate limitation) • to reduce the fall time, how ? • passively(smaller tf) Resolution ( Rfsmaller ) • linear active fast reset • in the 2. stage  ToT 2.nd stage ( <10 -3) • (GP et al, AGATA- FEE solution) • in the first stage ToT 1.st stage ( <10 -3 ??) • (not yet tested for high spectroscopy) • (G. De Geronimo et al, FEE for imaging detectors solution • A. Pullia, F. Zocca, Proposal for HP-Ge detectors) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  20. Potential solutions for active reset @1st stage a) & b)  for sequential reset c) through g)  for continuous reset G. De Geronimo, P. O’Connor, V. Radeka, B.Yu; FEE for imaging detectors, BNL-67700

  21. a) Custom designed vs. Commercial FEE ? • b) Discrete components vs. ASIC FEE ? • (Application Specific Integrated Circuits) • - Pros vs. Cons - • (price, performance, size, quantity, price/performance • ratio, R&D and production time, maintenance • manpower … but generally, it is more a • project management problem ! ) • - personally, I am trying to avoid generalization ! G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  22. - the dominant pole compensation technique GDC~30,000 Zo~ 66 Ohm NINO, an ultra-fast, low-power, front-end amplifier discriminator for the Time-Of-Flight detector in ALICE experiment F. Anghinolfi et al, ALICE Collab. ANALOGUE CIRCUITS TECHNIQUES, April , 2002; F. ANGHINOLFI ; CERN

  23. “ A Large Ion Collider Experiment, ALICE-TPC -TDR”, ISBN 92-9083-153-3, (1999), CERN

  24. 1. Charge Sensitive Preamplifier ( Low Noise, Fast, Single & Dual Gain ~ 100 dB extended range with ToT ) 2. Programmable Spectroscopic Pulser (as a tool for self-calibrating) 3. Updated frequency compensations to reduce the crosstalk between participants(-from adverse cryostat wiring and up to - electronic crosstalk in the trans. line) C. Chaplin, Modern Times (1936) crosstalk between participants  transfer function issue GSI-2012 8 Clusters (Hole 11.5cm, beam line 11cm) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  25. Best performance: Majorana dedicated FEE (PTFE~0.4mm; Cu~0.2mm;C~0.6pF; R ~2GΩAmorphous Ge (Mini Systems) ~ 55 eV(FWHM) @ ~ 50 µs (FWHM) BAT17 diode (GERDA) BF862 (2V; 10mA) 1pF; 1 GΩ Test Pulser ? -yes-not & how ? G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  26. Dual Gain Core Structure Ch1 (fast reset)-Pulser @ ~19 MeV Ch2 (linear mode) Ch 1 ~200 mV / MeV C-Ch1 /C-Ch1 INH1 SDHN1 Pole /Zero Adj. Fast Reset (Ch1) Differential Buffer (Ch1) Segments (linear mode) Common Charge Sensitive Loop + Pulser + Wiring Ch 2 ~ 50mV / MeV one MDR 10m cable 36_fold segmented HP-Ge detector + cold jFET C-Ch2 /C-Ch2 INH2 SDHN2 Pole /Zero Adj. Fast Reset (Ch2) Differential Buffer (Ch2) Ch1 ( tr ~ 25.5 ns) Programmable Spectroscopic Pulser Pulser CNTRL Ch2 ( tr ~ 27.0 ns) 2keV -170 MeV @ +/- 12V in two modes & four sub-ranges of operations: a) Amplitude and b) TOT

  27. Segment CSP  Negative Output AGATA CSPs – the versions with large open loop gain ( INFN-Milan – IKP-Cologne) Segment Non-Inverting DC coupled P/Z cancellation Cv R1 R1 Core CSP  Positive Output R1 Core Inverting from Active Reset Cv * (Cv) stability adj. whylarge Ao > 100,000 ?  frequency compensation, slope & crosstalk AC coupled

  28. Fast Reset as tool to implement the “TOT” method Core Active Reset OFF one of the segments Core -recovery from saturation (but base line …) Fast Reset circuitry G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  29. Fast Reset as tool to implement the “TOT” method Core Active Reset – OFF one of the segments Core -recovery from saturation Active Reset – ON Fast Reset circuitry ToT Normal analog spectroscopy one of the segments • very fast recovery from TOT mode of operation • fast comparator LT1719 (+/- 6V) • factory adj. threshold + zero crossing • LV-CMOS (opt) • LVDS by default > 220 MeV @ +/-15V G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  30. Fast Reset as tool to implement the “TOT” method Core Active Reset – OFF one of the segments Core -recovery from saturation Active Reset – ON Fast Reset circuitry ToT Normal analog spectroscopy one of the segments INH-C • very fast recovery from TOT mode of operation • fast comparator LT1719 (+/- 6V) • factory adj. threshold + zero crossing • LV-CMOS (opt) • LVDS by default > 220 MeV @ +/-15V G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  31. see Francesca Zocca PhD Thesis, INFN, Milan A. Pullia at al, Extending the dynamic range of nuclear pulse spectrometers, Rev. Sci. Instr. 79, 036105 (2008) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  32. Comparison between “reset” mode (ToT) vs. “pulse-height” mode (ADC) A. Pullia at al, Extending the dynamic range of nuclear pulse spectrometers, Rev. Sci. Instr. 79, 036105 (2008)

  33. Due to FADC; G=3 range  ! X-talk ! with CMOS  10 MeV

  34. AGATA Dual-Core LVDS transmission of digital signals: - INH-C1 and INH-C2 (Out) and Pulser Trigger (In) signals AGATA Dual Core crosstalk test measurements Ch2 (analog signal) vs. LVDS-INH-C1 (bellow & above threshold) Core amplitude just below the INH threshold Core amplitude just above the INH threshold Ch1@ INH_Threshold - (~ 4mV) Ch1 @ INH_Threshold + (~ 4mV) Ch2 @ INH_Threshold Vp-Vp(~ 1mV) Ch2@INH_Threshold + (- 1mV) LV_CMOS LV_CMOS INH_Ch1/+/ INH_Ch1/-/ tr ~ 1.65 ns INH_Ch1/+/ tf ~ 2.45 ns INH_Ch1/-/ (1) Core_Ch1, (2) Core_Ch2, (3) INH_Ch1(LVDS/-/, (4) INH_Ch1(LVDS/+/) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  35. If we have developed a FEE solution with: • Dual gain for the central contact (Core); • ToT for both Core channels and all Segments; • Saturation of the CSP at 170 MeV @ +/-12V … • ( and ~ 220 MeV @ +/- 15V ) • … then why not to perform a direct spectroscopic • measurement up to 170 MeV equivalent gammas ? • … were to find them ? … in cosmic rays! G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  36. To extend the comparison between active “reset” mode (ToT) vs. “pulse-height” mode (ADC) well above 100 MeV measuring directly cosmic rays (i.e. equivalent with inter- action of gamma rays above 100 MeV) • Interaction of muons with matter • Low energy correction: • excitation and ionization ‘density effect’ • High energy corrections: • bremsstrahlung, pair production • and photo-nuclear interaction MUON STOPPING POWER AND RANGE TABLES - 10 MeV|100 TeV D. E. GROOM, N. V. MOKHOV, and S. STRIGANOV David Schneiders, Cosmic radiation analysis by a segmented HPGe detector, IKP-Cologne, Bachelor thesis, 03.11.2011

  37. Two set-up have been used: LeCroy Oscilloscope with only Core signals: Ch1; Ch2, INH-Ch1; INH-Ch2 from Core Diff-to-Single Converter Box 10x DGF-4C-(Rev.E) standard DAQ - complete 36x segments and 4x core signals from Diff-to-Single Converter Boxes (segments & core) David Schneiders, Cosmic radiation analysis by a segmented HPGe detector, IKP-Cologne, Bachelor thesis, 03.11.2011

  38. Experimental results for cosmic ray measurement Calibrated energy sum of all segments vs. both low & high-gain core signals (both in ToT mode of operation) Calibrated energy sum of all segments vs. both low & high-gain core signals (linear & ToT ) Determination of the High Gain Core Inhibit width directly from the trace while the low gain core operates still in linear mode up to ~22 MeV ( deviation ~0.5%) David Schneiders, Cosmic radiation analysis by a segmented HPGe detector, IKP-Cologne, Bachelor thesis, 03.11.2011

  39. Combined spectroscopy up to ~170 MeV Direct measurement of cosmic rays with a HP-Ge AGATA detector, encapsulated and 36 fold segmented • Averaged calibrated segments sum +++ • Averaged calibrated Low gain Core xxx • Scaled pulser calibration (int. & ext.) ---- R.Breier et al., Applied Radiation and Isotopes, 68, 1231-1235, 2010 David Schneiders, Cosmic radiation analysis by a segmented HPGe detector, IKP-Cologne, Bachelor thesis, 03.11.2011

  40. Transfer Function & Crosstalk Transfer function - calculation (Frequency domain, Laplace transf., time domain) - measurement  spectroscopic pulser - applications: - bulk capacities measurement - crosstalk measurements and corrections

  41. In standard way the pulser input signal is injected AC (1pF) in the gate electrode of the jFET δq(t) 1pF 50 Ω The AC coupled Pulser - classical approach ! Detector

  42. δq(t) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  43. AGATA HP-Ge Detector Front-End Electronics Cold partWarm part AGATA – 3D Dummy detector Cold partWarm part G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  44. AGATA HP-Ge Detector Front-End Electronics Cold partWarm part Cold partWarm part G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  45. Rewritten as a Laplace transform of an exp. decaying function with If τ1 is sufficiently small, the exponential function can be “δ(t)“ and than the transfer function becomes: Simple current dividing rule Miller part Cold resistance equivalent input impedance of the preamplifier

  46. to be able to measure the transfer function, • we need to build and incorporate also a clean pulser with • spectroscopic properties and rectangular pulse form … ! G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  47. Incorporated Programmable Spectroscopic Pulser (PSP) • why is needed?  self-calibration purposes • brief description • Specifications, measurements and application: - Transfer function; - Charge distribution; - Impurities concentration measurements G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  48. The use of PSP for self-calibrating ParameterPotential Use / Applications • Pulse amplitudeEnergy, Calibration, Stability • Pulse FormTransferFunction in time (rise time, fall time, structure)domain, ringing (PSA) • Pulse C/S amplitude ratio  Crosstalk input data (Detector Bulk Capacities)(Detector characterization) • Pulse FormTOT Method (PSA) • Repetition Rate (c.p.s.) Dead Time(Efficiency) (periodical or random distribution) • Time alignment  Correlated time spectra (DAQ) • Segments calibration Low energy and very high energy calibration • Detector characterization  Impurity concentration, passivation (Detector characterization) G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

  49. +/- 1ppm • 16 bit +/- 1bit • fast R-R driver CSP return GND • Analog Switches: • -t on / t off , • -Qi, • -dynamic • range (+/- 5V) • Op Amp: • -~ R to R • -bandwidth • Coarse attenuation • (4x 10 dB) (zo~150Ohm) • transmission line • to S_ jFET and • its return GND! G. Pascovici , Carpathian Summer School of Physics, Sinaia 2012 Institute of Nuclear Physics, Univ. of Cologne and NIPNE-HH, Bucharest

More Related