Reliability Prediction - Parts Count MTBF

A parts count MTBF (Mean Time Between Failures) is a first approximation of the calculated reliability of a device or system. The MTBF is a calculated metric, based on one of two industry standards. Mil-Hdbk-217 is a military standard for the calculation of reliability of a device or system. Bellcore is a commercial standard (derived from the military standard) which is often used in commercial and industrial equipment. The MTBF analysis accounts for the pedigree of components used, improvements relating to burn-in to eliminate infant mortality failures, operating environment and temperature.

The parts count analysis method uses a generic default for the stress applied to each component type. A stress based MTBF analysis is a better approximation of the MTBF than a parts count analysis. The stress based analysis accounts for the actual stress applied to each component rather than relying on a generic default.

The results of the MTBF analysis provide indications as to the elements within the device or system that are most likely to fail, allowing improvements to be made to the most critical elements.

This analysis is typically performed using specialized computer software for the task.

Stress and Derating Analysis
A stress and derating analysis is a detailed percentage accounting of the applicable Voltage, Current and Power stress on each individual element within a device or system with regards to the component rating. Most failures in a device or system are a direct result of overstress of the element compared with its rating. This analysis is typically performed by hand, and summarized in a spreadsheet format. SPICE models or other mathematical models may be used to determine the applicable stress.

Interface Analysis
An interface analysis assures that each interface meets the output and input requirements to perform a given function. In digital circuits, a good example is to assure that TTL outputs are joined with TTL compatible inputs, and that the fanout requirements are met. This also applies to some analog functions. As an example an interface analysis might be used to assure that a Mosfet switch is fully enhanced and also fully depleted in order to assure full circuit performance.

Failure Simulation and FMECA Analysis
This analysis looks at the failure of each element within a device or system to determine the effect on the end performance. The failure of an element may be simple or complex. As an example, a resistor has three failure modes. The resistor could be open, shorted or outside of it’s specified tolerance. An integrated circuit is obviously much more complex. The goal is to eliminate any single point failures that would cause the device or system to fail to meet its performance requirements. Another way to look at this is to assure “graceful” degradation of a device or system. As a further example, suppose that the input to a device or system is couple to the outside world, via a resistor. If the resistor were to open, there would be no output from the device or system. The effect is therefore non-performance, the criticality is very high, and this constitutes a single point failure. The goal is to minimize the likelihood of single point failures and to provide the greatest amount of graceful degradation.

This analysis is performed using specialized computer software, along with SPICE and other mathematical models of the system.

WCA (Worst Case Analysis)
Worst case analysis is the analysis of a device or system that assures that the device meets its performance specifications. There are typically accountings for tolerances that are due to initial component tolerance, temperature tolerance, age tolerance and environmental exposures (such as radiation for a space device). The beginning of life analysis comprises the initial tolerance and provides the datasheet limits for the manufacturing test cycle. The end of life analysis provides the additional degradation resulting from the aging and temperature effects on the elements within the device or system.

This analysis is performed by determining the sensitivity of the output function to each element of the device or system and then applying the tolerances to the sensitivity to determine the deviance of the output function due to each element. The elements can then be added to determine the total variance of the output function. This allows a visual determination of the effects from each element, so that the worst offenders can be improved in order to improve the end result.

This analysis is performed using SPICE and mathematical models of individual circuits within the device or system to determine the sensitivities. A spreadsheet program is used to total and summarize the results.