6. Reliability Handbooks (Stress Analysis):
United States Department of Defense MIL-HDBK-217 establishes uniform methods for predicting the reliability of electronic parts and systems. It provides a common basis for reliability predictions during acquisition programs for military electronic systems and equipment. It also establishes a common basis for comparing and evaluating reliability predictions of related or competitive designs.
The handbook contains two methods of reliability prediction: â€œ Part Stress Analysisâ€ (PSA) and â€œPart Count Analysisâ€ (PCA). These methods vary in degree of information needed to apply them. PSA requires greater amount of information and is used later in the design process when circuit structures are known. PCA only requires part quantities, quality levels and the application environment and is therefore less accurate but useful during the early design phase and during the proposal formulation or â€œtentative device specificationâ€.
A similarity between PCA and PSA is that both prediction techniques use the same formulas. For PCA only estimated values are used where PSA uses calculated or measured values. There is a similarity between most part models. Generally speaking a failure rate formula will look like:
lp=Part failure rate
lb=base failure rate, dependent on temperature and applied stress
p..=acceleration factors for the used environmental application and other parameters that will affect the part reliability
pe= Environmental acceleration factor
pq= quality acceleration factor
Environment and quality are used for most parts, other p factors are part dependent. The base failure rate is usually expressed by a model relating the influence of electrical and temperature stresses on a part. One of the most characteristic parts of many models is the relation between temperature and failure rate. These models use thermal stresses in a form related to the Arrhenius Law.
Other acceleration factors are modeled in terms of acceleration factor p. The data used to model these acceleration factors are mostly obtained from manufacturers data and field returned data. Using this method, it is possible to model the effects of using a component under certain environmental conditions, the effect of using certain methods of component quality screening, etc.
British Telecom HRD-4:
The British Telecom Handbook of Reliability Data is quite similar in approach to the MIL-HDBK-217. In case of the British Telecom handbook, the term Part Stress Count analysis replaces the term Part Stress Analysis used in MIL-HDBK-217. The British Telecom Handbook of Reliability Data uses formulas of the form:
lp= part failure rate
lb= base failure rate
pt= Thermal acceleration factor
pq= quality correction factor
pe= Environmental acceleration factor
For some components not all these parameters are used. For those components where pt is used, it takes the form of the Arrhenius law. For most components the lb is used as a constant factor, independent of external stress factors. The table below gives an overview of the most common BT-HRD-4 parameters with their meaning, their influence factors and their origin.
||Environmental Acceleration factor
||Quality correction factor
||Base failure rate, depending on number of bits/gates
||Thermal acceleration factor
||thermal, device structure
|Discrete semiconductors and passive components
||Base failure rate
Generally speaking the BT-HRD-4 is less detailed than the MIL-HDBK-217. The MIL handbook aims at the military world and needs, for that purpose, to cover a wider range of environmental and application influence factors compared to the British Telecom handbook which only covers equipment for telecommunication purposes. Most other handbooks use models for most components equal or similar to either the MIL or the BT handbook.
Although the individual formulas describing the failure rate are quite different among the various reliability analysis handbooks they have several aspects in common. First of all the relation between part failure rate and effective device temperature is expressed in various forms of the Arrhenius law. Activation energies used in this expression are often very different for similar components. Most other influence factors are modeled in the form of acceleration factors. The acceleration factors are nearly always presented in the form of tables divided in certain classes. Most times these classification tables are based on practical experience and do not use an underlying physical model.
The majority of the acceleration factors is either related to the effective device temperature, the device structure or is environment bound.
Standard failure rate prediction handbooks use models in which effective device temperature is the dominant circuit related stress influence factor. If other circuit related stress parameters are used they are used in the form of correction factors.