Department of Defense MIL-HDBK-816 Notice 1 — MIL-HDBK-816 Notice 1 Radiation Hardness Assurance Guidelines Manual PDF

Downloaded from https://www.everyspec.com MIL-HDBK-816 NOTICE 1 5.2.3.4 End-points and electrical conditions. It is common also for a device specification to give values of a lot acceptance parameter as a function of the operating condition of the device. Thus, for example, end-point limits may be given for the post irradiation gain of a transistor as a function of the collector current. It can then happen, because the calculations are statistical in nature, that different mean values and standard deviations in the characterization data for the different collector currents can result in calculated end-point limits that do not behave in the expected way as function of the collector current.

The following caution is therefore advised: Caution: Data deficiencies and statistical effects or both can lead to end-point limits, calculated as a function of some device parameter, that do not behave in a reasonable way. For example, for a given neutron fluence, the calculated RHEPL for transistor gain may show a decrease with increasing collector current, instead of an increase (in the region below the collector current at which the gain is a maximum). It is important therefore, that the calculated results be subjected to a “sanity check” to make sure that ”unphysical” end-point limits are not entered into the device specifications.

In general, a RHEPL should be a smooth function of bias and radiation stress. It should be noted in this connection, however, that a gradual approach to failure may not always be monotonic; sometimes, for example, a parameter may first increase with radiation fluence and then subsequently decrease. 5.2.3.5 Abrupt failures. If abrupt failures are expected to be the dominant failure mode at the radiation levels required for the lot acceptance tests, any parameter end-point limits that are used should be calculated so that they will not cause a significant number of lot failures. This situation resembles that of the several parameters discussed in 5.2.3.3.

A method for estimating end-point limits for devices which suffer abrupt failure has been given by Namenson and Arimura (1985); this paper is listed in 2.2. In practice, even though lot acceptance tests may be conducted at a radiation level where some abrupt failures may occur, the subject part type should not be used in a system application if the specified radiation level is such that the likelihood of abrupt failure is expected to be significant; a derating factor of two on the radiation level has sometimes been used but a more accurate value should be based on actual fluence to failure data. 5.2.3.6 Use of end-point limits for system design.

Strictly speaking, the end-point limits used for lot acceptance tests do not guarantee how the parts will behave in a system exposed to a specified radiation environment. Nevertheless, systems designers frequently use the RHEPL as the starting point for derating a given parameter for a particular system application. For systems with moderate RHA requirements, the derating may be very small or the RHEPL may even be used directly in the system design. For systems with more stringent RHA requirements, the RHEPL value may be derated significantly or may be set so that only the hardest lots are accepted for system production.

Thus, if a part specification is being developed for a particular system, the needs of the design engineer should be considered. The entire topic of lot acceptance and its impact on system design and survivability, while very much deserving of detailed discussion, is beyond the scope of this document. 5.2.3.7 An assumed lot acceptance objective. The many factors involved in selecting an objective for the lot acceptance tests have been discussed in 5.2.3.1. To facilitate the data analysis discussions which follow, the lot acceptance objective is here taken to be that, with 90 percent confidence, at least 90 percent of future lots are intended to pass.

This objective means specifically that the end-point limit, which determines the probability, P S , that a single part will pass the test, must be set so that the probability that the entire test will be passed is 90 percent. This particular objective, in general, does not unduly penalize the performance specifications for a device nor does it place undue requirements on the amount or quality of characterization data. The assumption of a specific lot acceptance objective makes it possible to give actual numerical values for the examples which are to be discussed. The formulas required are given in their parametric forms so that calculations can be made for other lot acceptance objectives.

5.2.4 Data analysis. Because most radiation response results display significant and seemingly random variations, the methods needed to analyze such data are statistical in nature. In addition, most radiation testing involves the use of sample parts which are destroyed by the tests, so the analysis techniques and expectations of future performance (against which lot acceptance parameter end-point limits are calculated) must be based on statistics of sampling without replacement. The sections which follow discuss general analysis principles and their applications but do not attempt to cover all the data analysis variations that can occur in practice. 5.2.4.1 Data examination.

A necessary prerequisite for beginning any data analysis, which was mentioned previously but must be discussed in greater detail of its crucial importance, is an examination of the data to ensure that the data do not contain any faulty measurements or nonrandom effects. The data should be examined also for sufficiency. Data examination is all the more important when a significant amount REPRINTED WITHOUT CHANGE. 29