1
Conventional Electronic System Reliability Prediction

The history of reliability engineering goes back to 1950s when electronics played a major role for the first time. At that time, there was great concern within the US military establishment for the reliability and maintainability of the current electronic systems. Many meetings and ad hoc groups were created to cope with the problems. Developing better parts, finding quantitative reliability for parts, and collecting field data on actual part failures to determine the root cause of problems were three major fields of research in those days.

When the complexity of electronic equipment began to increase significantly, and new demands were placed on system reliability, a permanent committee (AGREE) was established to identify the actions that could be taken to provide more reliable electronic equipment (1952). The reliability era began when the first Radio Corporation of America (RCA) report on reliability of electronic parts was released in 1956, the first time when reliability was defined as a probability. On the other hand, one of the first reliability handbooks titled Reliability Factors for Ground Electronic Equipment was published in 1956 by McGraw-Hill under the sponsorship of the Rome Air Development Center (RADC); while the McGraw-Hill handbook gave information on design considerations, human engineering, interference reduction, and a section on reliability mathematics, failure prediction was only mentioned as a topic under development.

Reliability prediction and assessment are traced to November 1956 with publication of the RCA release TR-1100, titled "Reliability Stress Analysis for Electronic Equipment," which presented models for computing rates of component failures. It was the first time that the concepts of activation energy and the Arrhenius relationship were used in modeling component failure rates. However, in 1960s, the first version of a military handbook for the reliability prediction of electronic equipment (MIL-HDBK-217) was published by the US Navy [1]. It covered a broad range of part types, and since then, it has been widely used for military and commercial electronics systems.

In July 1973, RCA proposed a new prediction model for microcircuits, based on previous work by the Boeing Aircraft Company. In the early 1970s, RADC further updated the military handbook and revision B was published in 1974. The advent of more complex microelectronic devices pushed the application of MIL-HDBK-2 17B beyond reason. This decade is known for development of new innovative models for reliability predictions. Then, RCA developed the physics-of-failure model, which was initially rejected because of the lack of availability of essential data.

To keep pace with the accelerating and ever-changing technology base, MIL-HDBK-217C was updated to MIL-HDBK-217D on January 15, 1982 and to MIL-HDBK-217E on October 27, 1986. In December 1991, MIL-HDBK-217F became a prescribed US military reliability prediction document. Two teams were responsible for providing guidelines for the last update. Both teams suggested:

1. that the constant failure rate (CFR) model could not be used;
2. that some of the individual wear-out failure mechanisms (like electromigration and time-dependent dielectric breakdown) could be modeled with a lognormal distribution;
3. that the Arrhenius-type formulation of the failure rate in terms of temperature should not be included in the package failure model; and
4. that stresses such as temperature change and humidity should be considered.

Both groups noticed that temperature cycling is more detrimental to component reliability than the steady state temperature at which the device is operating, so long as the temperature is below a critical value. This conclusion has been further supported by a National Institute of Standards and Technology (NIST), and an Army Fort Monmouth study which stated that the influence of steady-state temperature on microelectronic reliability under typical operating changes is inappropriately modeled by an Arrhenius relationship [2-4]. However, considering the ability to separate failure mechanisms by separate Arrhenius activation energies, it may be possible to return to the physics of failure (PoF) assumption that each mechanism will have a unique activation energy.

1.1 Electronic Reliability Prediction Methods

There are several different approaches to the reliability prediction of electronic systems and equipment. Each approach has unique advantages and disadvantages; several papers have been published on the comparison of reliability assessment approaches. However, there are two distinguishable approaches to reliability prediction, traditional/empirical, and PoF approach.

Traditional, empirical models are those that have been developed from historical reliability databases either from fielded applications or from laboratory tests [5].

Handbook prediction methods are appropriate only for predicting the reliability of electronic and electrical components and systems that exhibit CFRs. All handbook prediction methods contain one or more of the following types of prediction:

• Tables of operating and/or non-operating CFR values arranged by part type,
• Multiplicative factors for different environmental parameters to calculate the operating or non-operating CFR, and
• Multiplicative factors that are applied to a base operating CFR to obtain non-operating CFR [6].

MIL-HDBK-217 reliability prediction methodology which was developed under the activity of the RADC (now Rome Laboratory) and its last version released in February 1995 intended to "establish and maintain consistent and uniform methods for estimating the inherent reliability (i.e. the reliability of a mature design) of military electronic equipment and systems. The methodology provided a common basis for reliability predictions during acquisition programs for military electronic systems and equipment. It also established a common basis for comparing and evaluating reliability predictions or related competitive designs. The handbook was intended to be used as a tool to increase the reliability of the equipment being designed."

In 2001, the office of the US Secretary of Defense stated that ".. the Defense Standards Improvement Council (DSIC) decided several years ago to let MIL-HDBK-217 'die the death.' This is still the current OSD position, i.e. we will not support any updates/revisions to MIL-HDBK-217" [6].

Two basic methods for performing the prediction based on the data observation include the parts count and the parts stress analysis. The parts count reliability prediction method is used for the early design phases when not enough data is available, but the numbers of component parts are known. The information for parts count method includes generic part types (complexity for microelectronics), part quantity, part quality levels (when known or can be assumed), and environmental factors. Since equipment consists of the parts operating in more than one environment, the "parts count" equation is applied to each portion of the equipment in a distinct environment. The overall equipment failure rate is obtained by summing the failure rate for each component over its expected operating condition.

A part stress model is based on the effect of mechanical, electrical and environmental stress and duty cycles such as temperature, humidity, and vibration on the part failure rate. The part failure rate varies with applied stress and the strength-stress interaction determines the part failure rate. This method is used when most of the design is complete, and the detailed part stress is available. It is applicable during later design phases as well. Since more information is available at this stage, the result is more accurate than the parts count method.

The environmental factor gives the influence of environmental stress on the device. Different prediction methods have their own list of environmental factors suitable for their device conditions. For instance, the environmental factor of MIL-HDBK-217F covers almost all the environmental stresses suitable for military electronic devices except for ionizing radiation. The learning factor shows the maturity of the device; it suggests that the first productions are less reliable than the next generations [7, 8]. The parts stress model is applied at component level to obtain part failure rate (?p) estimation with stress analysis. A typical part failure rate can be estimated as:

where ?b is the base failure rate obtained from statistical analysis of empirical data, the adjustment factors include: pT (temperature factor), pA (application factor), pV (voltage stress factor), pQ (quality factor), and pE (environmental factor). The equipment failure rate (?EQUIP) can be further predicted through parts count method:

where ?g is the generic failure rate for the ith generic part, pQ is the quality factor of the ith generic part, Ni is the quantity of ith generic part and n is the number of...