2nd Qtr 2010
(C) 2010 Design/Analysis Consultants, Inc.
Newsletter content may be copied in whole or part if attribution
to DACI and any referenced source is prominently displayed with the copied material
This Issue: NEWS BITE: Sneak Peek of “Smiley” Terminator 5! / NEWS BULLETS: Unintended Consequences Strike Again / DM V8 Scheduled for Release This Year / SPECIAL REPORT: Aging (Drift) Considerations for Electronic Components
NEWS BITE: Sneak Peek of “Smiley” Terminator 5!
Photo: Osaka University and Kokoro Company
(“Geminoid F: More Video and Photos of the Female Android” by Erico Guizzo, April 20, 2010 IEEE Spectrum)
NEWS BULLETS: Unintended Consequences Strike Again
From “The climate changers: How wind turbines make their own clouds” by Andrew Levy, April 25, 1010 Daily Mail online.
Design MasterTM V8 (Major Upgrade) is planned for release later this year, with several new features. We’ll keep you posted.
Also, based on requests from users, we have shifted from a subscription basis to a flat purchase basis, with upgrade fees for new releases. (Those who purchase V7 within a year of the release of V8 will receive V8 at no added charge.)
Sample and Part files are being updated for V8 to include the effects of aging (see the Special Report in this newsletter).
For more details on the current version, please click here: Design Master V7
Aging (Drift) Considerations for Electronic Components
A key goal for an electronics circuit is to ensure that the circuit will function properly throughout its desired lifetime. Achieving this goal requires an understanding of how electronic component parameters can shift, or drift, as they age.
DMeXpertTM Tip Accounting for Aging Is Essential
For even moderate desired lifetimes of a few thousand hours, aging may have more of an impact on total error than other factors such as initial tolerance and temperature.
End-Of-Life, Lifetime, MTBF, and Other Confusion Factors Designed to Torment the Design Engineer
Aging analysis should not be confused with “reliability prediction” efforts, which employ terms such as end-of-life, lifetime, and MTBF (mean time between failures). These efforts are intended to provide a statistical estimate of when a component will fail; i.e. suffer such a large parameter shift that it becomes inoperable. Although these predictions may be of interest to a reliability engineer, the design engineer is concerned with parameter shifts that occur as the part ages but is still operable.
For example, end-of-life generally refers to parameter drift after several years (usually not well defined, but the term typically implies a period of 7-15 years), whereas “lifetime” may refer to the time for actual part failure (wearout), which is generally exhibited as gross inoperability of one or more key functions. MTBF (mean time between failure) calculations are supposed to predict useful lifetime, but they can be highly inaccurate.
DMeXpertTM Tip MTBF and Lifetime Predictions
Don’t bother. As we’ve said before, it makes much more sense to spend resources on improving processes and designs than to spend those resources on making predictions that are based on inadequate samples and circular reasoning, predictions that often prove to be wildly inaccurate and misleading.1
Note 1: Our research has not uncovered a single example of a long-term study of failure rates or lifetimes that compares those actual values to a baseline prediction methodology. If you know of one, please let us share it with our readers.
The Design MasterTM Approach to Aging
When the desired operating hours are high, it’s important to use parts that define how their key attributes (resistance, capacitance, offset voltage, output current, etc.) shift with time, and also with applied stresses. Unfortunately, this information if often very hard or impossible to find, which leads to the following guidelines:
DMeXpertTM Tip Selecting Electronic Parts
For long-life designs, use parts that have aging effects clearly defined in the manufacturer’s data sheets or application notes.
DMeXpertTM Tip Don’t Predict, Confirm
Don’t try to predict the lifetime of a circuit.
Do specify the required lifetime of a circuit and confirm that, considering the effects of aging, the circuit will perform properly over that lifetime.
If you must use parts that have poorly-defined lifetimes (unfortunately this is often the case), then your product will need to have periodic maintenance (which can be manual, or automated via built-in self-test) that monitors performance degradation. In either case, the user will be know when the equipment should be recalibrated or receive preventive maintenance.
If you are building a product that cannot be manually maintained (e.g. a space probe) then the design will require not only the best (lowest aging) components, but will also likely need auto-calibration circuits to maintain performance.
CAUTION: Use “Useful Life” ONLY for Maximum Allowable Drift Points
The “useful life” figure provided by many part vendors is, well, not too useful, because it is not expressed in designer-friendly terms. Life testing and related data presentation seem to be controlled by statistically-oriented reliability engineers, but the dominate Big Gorilla factor in determining real-world useful life is the allowable parameter shift that a designer can tolerate, not an end-of-life guesstimate embedded in a generic test standard. For example, a failure definition for a capacitor life test, such as a capacitance change of -30%, or a change in ESR of +300%, will typically far exceed what is acceptable for a long-life design application.
In addition, “useful life,” for valid economic reasons, is based on very small samples, which means that a lot of assumptions and guesswork are involved.1 Also, the limits of the useful life value are not usually provided in the data sheet (no minimum values or standard deviations).
For these reasons, “useful life” or similar lifetime data provided by vendors should not be used for predicting life, but used only for defining maximum allowable parameter shifts.
Note 1: Part vendors make educated guesses as to the life of their product. They don’t say “educated guesses”; that would make some of their customers nervous. If anyone doubts the validity of this, here’s a challenge: scour the literature and send in a data sheet that provides a designer-friendly specification for a component’s useful life; e.g., “This component has a useful life, as defined by a tolerance shift of not more than +/- 5%, of 5000 hours minimum when operated within its maximum ratings.” In all cases we’ve seen, when you look at the fine print you find that useful life figures are based on small-sample statistical testing; i.e. the stated hours will typically be a mean value for a defined failure rate, as further defined by arbitrary large parameter shifts, with a 60% confidence level based upon an assumed exponential distribution — got a headache yet?
Our View A New Paradigm for Part Vendors
Wouldn’t it be great if part vendors defined the min/max drift rates of all key part parameters, and dispensed with “useful life” mumbo-jumbo and related reliability “predictions”?
Modern ICs = Shorter Lifetimes
Due to decreasing geometries in ICs, wearout lifetimes (the time at which the part becomes inoperable) have been decreasing, and have become a significant reliability concern for electronics that are expected to operate for many years. Since shorter wearout lifetimes translate into a higher rate of aging versus time, aging analysis becomes even more critical when using modern ICs. The corollary is the following non-intuitive rule-of-thumb:
DMeXpertTM Tip Designing for Long Life
For designs with a desired life of many years, use older-technology ICs whenever possible.
With shorter wearout times and correspondingly quicker aging, it is becoming more and more necessary to employ auto-calibration so that circuits can periodically readjust themselves to compensate for drift errors.
DMeXpertTM Tip Ensuring Proper Auto-Cal
The required auto-cal dynamic range must be sufficient to compensate for the maximum accumulated drift over time.
Accounting for Aging Effects
Consider a simple resistor. A resistor’s deviation from its ideal nominal value is defined by its initial tolerance, and by reversible and irreversible variables. For a resistor,
Reversible: Temperature coefficient (static variable)
Irreversible: Aging (or drift), a time-dependent permanent deviation that is a function of exposure to elevated temperatures, voltages, humidity, etc. (dynamic variable)
Aging address irreversible deviations versus time, and should be factored with the initial tolerance and temperature coefficient to obtain the total tolerance:
TOTAL TOLERANCE =
(1 + TOLi) * (1 + TOLtc*(Tap-Tref)) * (1 + TOLage*Life)
TOLi = initial tolerance; e.g. -0.01/+0.01 for a 1% part
TOLtc = temperature coefficient, /C; e.g. -1.0E-4/+1.0E-4 for a +/-100ppm/C part
Tap = application temperature, C (typically ambient temperature Ta or case temperature Tc)
Tref = data sheet reference temperature, C (typically 25C)
TOLage = aging tolerance, /hour for a reference operating time; e.g. +/-20% for 2000 hours, or -1E-4/+1E-4
Life = desired operating time, hours
(For severe applications (e.g. circuits exposed to a radiation environment), additional tolerance terms need to be included in Total Tolerance.)