Status report of the ATLAS SCT optical links John Matheson (RAL) for the ATLAS opto-links team links architecture total dose radiation hardness investigation of SEU cooling issues mechanical issues system tests pre-production harness conclusions 1
silicon microstrip detector readout chips silicon barrel module silicon microstrip detector readout chips need timing, trigger, control in need data out optical links for low mass, crosstalk 2
40 MHz clock and control on single fibre by biphase mark encoding links architecture 40 MHz clock and control on single fibre by biphase mark encoding 2 data fibres/module, 40 MHz NRZ radiation hardness (10yr): 10 Mrad, 21014 1 MeV neutrons cm-2 3
power budget 4
DORIC and VDC barrel harness opto package (2 VCSELs, 1 p-i-n) power tapes, redundancy system furcation tube (now black !) each harness serves 6 modules 5
20 VCSELs irradiated at INER, Taiwan 30 MeV protons, 21014 cm-2 VCSEL irradiation 1 20 VCSELs irradiated at INER, Taiwan 30 MeV protons, 21014 cm-2 168 hr at 10 mA, 168 hr at 20 mA 6
VCSELs degrade i.t.o. threshold, efficiency VCSEL irradiation 2 VCSELs degrade i.t.o. threshold, efficiency injection annealing foreseen in ATLAS 60 Mrad gamma - little change 7
VCSEL ageing 20 VCSELs aged at 100°C for 2534 hr no failures R = (1/Nt) ln(1/(1-C)) AF = (I2/I1)2 exp[EA/KB (1/T1 - 1/T2)] duty cycle during data transfer 50% fraction of time sending data 50% fraction of time LHC operates 33% 8
Irradiation of other components VDC/DORIC 50 Mrad, 2.5×1014cm-2 n 1 week at 100°C no failures 0.3% max chip failure in SCT (redundancy makes OK) wafers being qualified for production p-i-n diode 1×1015cm-2 n, 0.5 0.35 A/W aged 60°C 5 failures in SCT passives 3×1014cm-2 24 GeV p, no change opto-flex assembly 3×1014cm-2 24 GeV p, behaviour as expected from component-level irradiations 9
energy deposition in p-i-n diode may cause SEU single event upset energy deposition in p-i-n diode may cause SEU window A => command bit error window B => BC clock error clock arrives up to 6.25ns early ABCD may lose synchronisation 10
single event upset in command signals Links measured in loop-back mode measure errors in command bits loop-back link DORIC VDC SEU irradiations at PSI pions, protons 300-500 MeV/c vary light power DORIC threshold varies 11 11 11 11
single event upset in bunch crossing clock BC clock SEU measured in 24 GeV PS beam forward module in beam - optical link barrel module in counting room - Cu link L1 trigger beam spill L1 triggers BC counter of ABCD is read out after each L1 trigger difference between the modules => SEU SEU error rate (s-1) / flux 12
single event upset summary minimum Ip-i-n = 75A max SCT flux 2106cm-2s-1 fraction of module data lost: 9 10-4 command bit (worst case) 7.8 10-5 BC clock error 13
defines T at opto-package component lifetime is T dependent cooling measurement 1 heat load from chips 0.4W defines T at opto-package component lifetime is T dependent need to get the heat to the cooling block 14
Lower face dT1 through flex cooling measurement 2 (Cooling block) dT3 T drop = dT1 + dT2 +dT3 minimise dT1 using vias below chips AlN backing, BN-loaded epoxy 15
total dT ~ 18°C at -7°C ambient cooling measurement 3 total dT ~ 18°C at -7°C ambient expect opto-package T ~ 0°C up to 10°C tolerable for lifetime FEA predicts total dT ~ 25°C with no heat transfer to environment, shows importance of vias 16
mechanical issues: design clearance clearance between top of opto-package/chip cover and lower side of module is most critical point compliance of flex must compensate variations in: barrel eccentricity tape stack height 17
mechanical issues: clearance measurement clearance measured using jig (Geneva) and envelope module worst case uses max. barrel eccentricity, 6 tapes to push opto-package upwards 1.4mm clearance measured using slip gauges in agreement with the design value 18
send 2 closely spaced L1 triggers CERN system test: a surprise send 2 closely spaced L1 triggers module being read out when second trigger arrives excess noise correlated with position 19
excess noise is due to light leaking from optical fibres CERN system test - light leaks excess noise is due to light leaking from optical fibres different colour furcation tubes measured at RAL - black is opaque 20
first pre-production harness harnesses to be made by Radiantech Inc, Taiwan before full-scale production, 8 pre-series harnesses will be made pre-series harnesses to be used in system test at RAL with at least 6 modules at low T on B3 sector first harness received August 02 some minor issues to be resolved optical links are fully functional 21
pre-production harness: optical measurements 22 pre-production harness: optical measurements p-i-n diode responsivity mean 0.44 A/W VCSEL LI curves (50% duty cycle)
pre-production harness: bit error rate measurements ttc link operating point data link operating point both links have been operated for 12 hr with no errors 23
pre-production harness: towards a system test at RAL barrel module on pre-series harness fully functional, 1600 e- noise pre-series tests completion will define start of production in industry 24
industrial production of scheduled to begin late 2002 conclusions SCT harnesses must be rad-hard to 10Mrad, 11014 n/cm2 fibre, ASICs, VCSELs, p-i-n diodes, optical interconnects qualified harnesses are not immune to SEU but the level is tolerable mechanical tolerances must be respected to preserve clearances cooling must be adequate to preserve device longevity on-sector tests at RAL will be completed before production begins industrial production of 352 barrel harnesses scheduled to begin late 2002 25