Reliability Analysis of a Low Voltage Power Supply

Reliability Analysis of a Low Voltage Power Supply Design for the Front-End Electronics of the ATLAS Tile Calorimeter Gary Drake, Member IEEE, James Proudfoot Abhirami Senthilkumaran, Bruce Mellado, Anusha Gopalakrishnan, Sanish Mahadik University of Wisconsin-Madison, WI USA Argonne National Laboratory, Lemont, IL USA On Behalf of the ATLAS Tile Calorimeter System The Tile. CAL Low Voltage System LVPS Brick • Power for Tile. CAL Front-End Electronics • Novel Switching DC-DC Power Supply LVPS Box 8 bricks per Box – Custom, Compact, High-Efficiency, 250 Watt – 8 Different Voltages Customized Bricks – Water Cooled; System Interface & Monitoring – Environment: Magnetic Field, Radiation Tolerant • 256 boxes on detector, 2048 bricks, + spares ð Reliability is Important Infrequent Access LVPS Access on Detector Section End of Long Barrel Drawer Electronics We have performed a reliability analysis on the new upgraded supplies 2048 Bricks in the detector system Reliability Analysis Methodology • Failures in Electronics – Failures generally described by the Bathtub Curve – Failures over useful life are random but have an average rate: Poisson Process – Probability of k failures between time t and t+t: Poisson Distribution • Mean Time Between Failures MTBF – Expected time between failures – MTBF = 1 / l – This is not “useful lifetime” • Probability of Failure-Free Operation – Interested in region of Constant Failure Rate – l = Average number of units failing per unit time – Measured in Failures In Time FITS (# / 109 hrs) – Probability of no failures at time t – R(t) = e-lt ð Must calculate l for the entire unit Calculations 1. Series Reliability Model – Any single failure can cause brick to fail – Use only critical parts in model – – – § Assume – Part failures are independent & random ð Start with single tantalum cap § Parts Count Method – Use FITS values for each part Part IR 2110 S FET Driver IRFS 9 N 60 MOSFET Inductor 470 u. H Capacitor 47 u. F HCPL 7800 Opto-Isolator LM 6142 Op-Amp LT 1681 Controller Chip Result #1: 0. 992992 R(t = 5 yrs) 0. 982573 § – Rated voltage of capacitor is 20 V – Calculated failure rate: 12. 4 bricks/year Tantalum capacitors most critical Rated for 35 V, used at <= 15 V Higher voltage rating reduces failure When a capacitor fails – Probability of short = 0. 75; Probability of open = 0. 25 ð Also include 4 caps in parallel + diodes + LC filter Failures/109 hr 43. 6 0. 544 10 288 52 2 6. 04 # Failures/yr = #Units * S li = 7. 20 R(t = 2 yrs) 3. Comparison with Previous Design 2. Series-Parallel Model and Voltage De-rating for Capacitor R(t = 10 yrs) 0. 965450 R(t = 20 yrs) 0. 932094 ðDominated by Tantalum Reliability – Observed failure rate: 5. 2 bricks/year – From 3 years of operation § Statistical Analysis – Poisson distribution; Neyman procedure ð ð Use observations as a correction Result #2: # Failures/yr = #Units * S [li * wi] = 5. 03 R(t = 2 yrs) 0. 995097 R(t = 5 yrs) 0. 987788 R(t = 10 yrs) 0. 975727 R(t = 20 yrs) 0. 952049 ðStill Dominated by Tantalum Reliability Result #3: # Failures/yr = # Units * S [li * wi] X x R(t = 2 yrs) 0. 997938 R(t = 5 yrs) 0. 994853 R(t = 10 yrs) 0. 989732 R(t = 20 yrs) 0. 979569 ðExpect 2. 11 failures per year in the system Argonne National Laboratory is a U. S. Department of Energy laboratory managed by U Chicago Argonne, LLC.