Outline Reliability of electronics: HALT-testing Isabelle Vervenne TUSofia, 9 th September 20 Reliability analysis Some definitions Failure rate of electronic components HALT (highly accelerated life tes What is it? HALT test and results HALT results and MTBF HALT as starting point for HASS 2 Reliability Failure rate R( = probability that system performs correctly during [t 0,t], if OK at t 0 0.8 0.6 0.4 0.2 0 R( e t typical R(: exponential failure law Failure rate z( expected # of failures per time-unit Example: 000 components working at t 0 after 0 hours: 990 working failure rate for each component: z( = 0.00 failures / hour (0 failures / 000 componen / 0 hours z( = failure / 000 hours 3 4 Failure rate and reliability Failure rate calculation Relation z( and R(: Bathtub curve: useful life period z( = λ = constant R( = e -λt dr( z( R( dt Determined for components systems: combination of components Determine experimentally expensive many components required for significance slow e.g. failure per 00 000 hours (=.4 years) faster by accelerated testing 5 6
Standards for λ Purpose of reliability standards Telcordia Technologies Reliability Prediction Procedure for Electronic Equipment SR-332, iss. 2 (Sep 2006) ex-bellcore assuming exponential distribution (constant ) MIL-HDBK-27 F (Notice 2) Reliability Prediction of Electronic Equipment Feb 995 (not longer supported) failures per 0 6 operating hours (constant ) 7 Estimation of inherent reliability of electronic equipment, based on component failure data Two basic prediction methods: Parts-Count Analysis: uses the number of parts in each category with consideration of part quality and environments encountered Part-Stress Prediction: uses complex models composed of detailed stress (analysis information, environment, quality applications, maximum ratings, complexity, temperature, construction, ) 8 General failure model λ p : λ b : p b Q E failure rate of one component base failure rate, based on Arrhenius equation π Q,π A,π E, : factors related to component quality, application stress, environment, A IEC standard for reliability data Reliability data handbook - Universal model for reliability prediction of electronics components, PCBs and equipment IEC/TR 62380 - Ed..0 (2004/08) Models power on/off cycling, temperature cycling 9 0 MTTF MTTF: brain teaser Mean Time To Failure expected time until first failure occurs expected time until next failure occurs exponential failure law: history-independent Experimentally e.g. N components, measure time-tofailure N MTTF. MTTF t R(dt e dt N ti i example for a component: = failure per 8 year MTTF = 8 years MTTF brain teaser: probability to be functional at MTTF? R( MTTF ) R( ) e 0.3678 Hence, 63% of components fail before MTTF! 2 2
Present MTBF tools Reliability prediction methods Relatively quick, but inaccurate Uses a reliability prediction handbook Reliability demonstration test Difficult to choose the appropriate amount of stress Not too little, not too much stress Time consuming Field data No trouble founds (NTFs) Very time consuming and not timely 3 Shortcomings of traditional reliability prediction Very time consuming Often inaccurate model in comparison to field failure reports Possible reasons: Many failure modes in real world do not follow the Arrhenius equation (mechanical vibration and shock, humidity, power on/off cycling, ESD, ) Reliability of electronic components is improving 4 Predominant failure causes Outline 20% 2% 9% 9% 4% 22% 5% 9% Parts Design Manufacturing System management Wear-out No defect Induced Software Reliability analysis Some definitions Failure rate of electronic components HALT (highly accelerated life tes What is it? HALT test and results HALT results and MTBF HALT as starting point for HASS 5 6 Objective of HALT HALT: what is it? Highly Accelerated Life Testing Objective: find weak links in design and fabrication processes of a product during design phase DVT: Design Verification Testing Tests done before product release 7 Product is stressed well outside operating specifications Stress until fundamental limits of technology Stimulate failures (not simulate the environmen Find failures - fix on the fly and continue testing 8 3
HALT: what is it? Process of discovery and optimization Operating limits and destruct limits: points where the system ceases to work as specified but will return to operation if the stress is removed (operating limi and ceases to operate even if the stress is removed (destruct limi Try to maximize the operating and destruct margins of the product Not a pass/fail test HALT parameters Prior to HALT After HALT 9 20 HALT tests Example of test pattern HALT= always a series of tests Single environments AND combined environments Cold testing Heat testing Vibration testing Heat and vibration testing Cold and vibration testing Thermal swings Don t forget productspecific stresses: voltage variation, frequency variation, power on/off, 2 22 HALT chamber HALT test 23 23 24 24 4
HALT test chamber FMEC Temperature Range: -00 C to +200 C Variation: up to ±60 C/min (air temperature) Air speed: up to 8 m/sec Air flow: distributed evenly over the surface of the table HALT test chamber FMEC Vibration Segmented table with 6 pneumatic hammers 6 degrees of freedom (omni-axial) quasi-random At least 30 Grms in the 6 axes of the frequency spectrum 5-2500 Hz and 50 Grms in frequency range 5-5000 Hz 6-channel accelerometer feedback 3-axes closed loop control 25 26 HALT test results What to consider before testing? Be thoroughly familiar with the product to be tested Final environment Need for constantly monitoring of the product Intermittent failures What will be considered as a failure? First intermittent failure? All the way to a hard failure? 27 28 Why do HALT? Why do HALT? Rule of 4: Every new step in product design costs 4 times more than the previous step Example: Cost design change: 00 Cost change in the field: 00 x 4 x 4 x 4 = 6400 29 30 5
HALT benefits Examples of detected failures Faster time to market Increased reliability more robust products Greater customer satisfaction Lowered warranty cost through higher MTBF Minimized chance of product recalls 3 32 Examples of detected failures HALT versus traditional testing HALT 33 Stresses product beyond specification Gathers information on product limitations Focus on design weakness & failures 6 DoF Vibration High thermal rate of change Loosely defined Modified on the fly Not a Pass/Fail Test 34 Traditional Verifies that a product meets specification Simulates a lifetime of use Focus on finding failures Single axis vibration Moderate thermal rate of change Narrowly defined Rigidly followed Pass/Fail Test HALT results and MTBF Difficulties: Number of samples is low (typically 4) Acceleration factors are not only high, but varying Yields reliability figures with poor confidence bounds Possibilities:. Use HALT data to choose aggressive acceleration factors to run an RDT (reliability demonstration tes 2. Compare HALT results of a new product with field results of earlier HALT tested products HALT as a start for HASS HASS: screening of all products Use of HALT results to develop an effective production screen Ability to find defects in the product without removing significant life from the product 3 steps: - HASS development - Proof-of-screen - Production HASS 35 36 6
HASS development Choose stress types Take a percentage of the thermal/ vibration response levels discovered during HALT Thermal: 20% of operational limit Vibration: half the response destruct limit Map the production fixture with a product Precipitation/detection screen Modulated vibration and tickle vibration HALT limits by environment Environment Thermal Data, C Vibration Data, Grms LOL LDL UOL UDL VOL VDL Office -62-80 92 8 46 52 Office with user -2-50 67 76 32 36 Vehicle -69-78 6 23 2 24 Field -66-8 06 24 66 69 Field with user research M. Silverman -49-68 8 06 62 62 Airplane -60-90 0 0 8 29 37 38 Proof of screen Initial HASS-profile applied to one chamber load of test articles for 30-50 times Proof-of-screen POS : safety of screen Hass not too destructive? Production HASS Each product is screened to prevent early field failures Profile may need to be changed based upon manufacturing data and field data, but with careful consideration Escape of defects => modifying the screen POS 2: efficiency of screen HASS effective? 39 HASS yields about 20% of undetected failures during HALT => effectiveness of HALT? 40 HALT research Questions? HALT and traditional reliability modeling techniques cover different areas Is there any correlation between them? Can we use HALT to predict the lifetime of electronic products? Contact: isabelle.vervenne@khbo.be davy.pissoort@khbo.be 4 42 42 7