ASIC QA/QC for DUNE Carl Bromberg, Kendall Mahn, Dean Shooltz, Daniel Douglas Michigan State University July 16-18, 2018 C. Bromberg - CE Workshop 2018
Cold ASIC QA/QC before FEMB produc:on? • CE is tricky w/many constraints: e.g., seamless operaLon warm & cold • Foundry or Packaging issues can occur at any Lme. • Single channel failure rate goal is low, and failed chip rate goal is zero. • Reworked FEMB are likely to have reliability problems • Hot air removal and soldering of chips will affect their long-term reliability • Other components (R & C, Diodes & Reg.) may be weakened with long-term reliability issues • Discarding many below par FEMB could cost $M’s • If ASIC cold tests are safe, feasible & cost effec(ve , why not do it? C. Bromberg - CE Workshop 2018
What channel/chip failure rate is acceptable? • 0.1% channel failure is a few dead wires per APA • Without cold QA, ProtoDUNE FE failures would have been 40 Lmes 0.1 • With cold QA, ProtoDUNE FE ASIC beat the 0.1% dead channel rate. • Our desires are closer to a few dead channels/detector, < 0.001% rate • Note: FEMB readout mixes x, u, v: full ASIC failures are really bad. • A 2D reconstrucLon dead-spot (~ 4x4 cm 2 ) generated if FE or ADC failure • A 2D reconstrucLon dead-spot (~ 8x8 cm 2 ) generated if ColdDATA failure • For FE and ADC, QA/QC tests must select 24,000 chips in ~ 6 months C. Bromberg - CE Workshop 2018
Feasibility of Cold ASIC QA/QC • Dean and I learned a few things during the MSU CTS development • EliminaLng H2O from cold tesLng was essenLal for reliable CE R&D & QA/QC • Have to take safety reviews seriously- it takes real Lme and effort to go from prototype to approved device • ASIC test clamp works if H2O avoided. • Quad Clamp Boards did QA/QC for the FE ASIC and likely can do ADC (ColdDATA?) • CTS & Quad Board does 4 ASIC QA/QC in < 60 minutes (15 min/ASIC) • To avoid mechanical damage, use automated loading of chips into Quad Boards • To use OCR, put serial numbers on the ASIC chips when packaged • To get 24,000 good ASICs selected in < 6 months (2 min/ASIC) • More ASICs done in parallel - perhaps factor of 4 • Cycle Lme reducLons – perhaps a factor of 2 C. Bromberg - CE Workshop 2018
RTS Conceptual Design • RoboLc loading & unloading of Quad Clamp Boards • Buy Commercial SCARA robot (Epson T3 or similar for about $10k) • Robot can open/close test clamps, or operator does • Include vision based system with OCR (opLcal character recogniLon) • Robot Tasks • Vision based inspecLon of chips (bent leads?) • OCR (opLcal character recogniLon) for ASIC ID (serialized at fab house) • Load chips into Quad Clamp test boards • Unload chips from test boards into pass/fail trays • Tray-to-tray chip sorLng (if we get into cherry-picking ASICs) C. Bromberg - CE Workshop 2018
RTS Conceptual Design • RoboLc loading of Quad Clamp Boards • Buy Commercial Pick & Place Machine • Quad Clamp Boards loaded/unloaded by Pick & Place Machine • A Framework hand loaded & cabled with many verLcal Quad Clamp Boards • Two styles possible for Quad FE ASIC Board (similar design for ADC, ColdDATA) FPGA Mezz FPGA Mezz 64 single-ch. ADCs 64 single-ch. ADCs LN2 Flexible Cables Level LN2 Level ASIC ASIC ASIC ASIC ASIC ASIC ASIC ASIC Quad FE ASIC Board Quad FE ASIC Board C. Bromberg - CE Workshop 2018
RTS Conceptual Design • N2 Enclosure • CTS-like LN2 chamber at floor level with “Gate Valve” Lid • Door to N2 enclosure Door Open Frame w/mulLple N2 Enclosure Quad Boards * * * Gate Valve LN2 Level LN2 Test Chamber C. Bromberg - CE Workshop 2018
RTS Conceptual Design • Frame posiLoned in N2 enclosure, cables connected, Door closed • N2 Enclosure Purged with N2 Gas at high flow rate Door closed N2 Enclosure * * * Gate Valve LN2 Level LN2 Test Chamber C. Bromberg - CE Workshop 2018
RTS Conceptual Design • Gate Valve opened • Frame lowered into LN2 Test chamber to Cool Level • Strong Fans circulate N2 gas to cool ASIC loaded Quad Clamp Boards Door closed N2 Enclosure * * * Gate Valve LN2 Level LN2 Test Chamber C. Bromberg - CE Workshop 2018
RTS Conceptual Design • Frame lower ASIC Clamps into LN2 • ASIC tests on each Quad Clamp Board run in parallel Door closed N2 Enclosure Gate Valve * * * LN2 Level LN2 Test Chamber C. Bromberg - CE Workshop 2018
RTS Conceptual Design • Frame raised into N2 Enclosure • Gate Valve closed • Hot N2 gas circulated to bring Quad Boards rapidly to Room Temp. Door closed N2 Enclosure * * * Gate Valve LN2 Level LN2 Test Chamber C. Bromberg - CE Workshop 2018
RTS Conceptual Design • Frame removed from N2 enclosure and Cables removed • Quad Boards have ASICs removed by Pick & Place machine • New Frame loaded and cycle starts again. Door Open N2 Enclosure * * * Gate Valve LN2 Level LN2 Test Chamber C. Bromberg - CE Workshop 2018
More Elaborate RTS Conceptual Design • N2 Enclosure • Input and Output N2-Locks • Open top CTS-like LN2 chamber at floor level • Cabling sealed below floor and moves with Framework Frame w/mulLple Input N2-Lock N2 Enclosure Output N2-Lock Quad Boards Purge Purge & Warm * * * Open Top LN2 Test Chamber C. Bromberg - CE Workshop 2018
RTS Throughput & Cost • Loading and Unloading Quad Boards done in parallel with tesLng • TesLng done on each Quad Board in parallel • Full cycle of say 16 ASIC chips takes 1/2 hour ( 2 minutes/ASIC) Costs (fully loaded) • Conserva(ve esLmate based on Design/Prototype Time: 1.5 years • One (DS) Full-Lme Engineer-Fabricator-Programmer ($400k) • M&S for prototype ($75k), 3 producLon units ($150k) • ProducLon Units Assembly & TesLng ($100k) • OperaLon of 3 RTS units, 6 students for 6 months ($100k) • Total = $750k --- Compare to cost of discarding 500 FEMB ($2.5M ?) • MSU can produce RTS. We want to parLcipate in ASIC & FEMB QA/QC C. Bromberg - CE Workshop 2018
Recommend
More recommend