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 5.2.2 Definition of RHA end-point limits (RHEPL). The parameter values against which the post radiation performance of a part will be compared to determine if the part has passed or failed the test have been termed RHEPL. These RHEPL values are included in the procurement specification and are used for the lot acceptance tests. If the procurement system uses a qualified manufacturers list or qualified parts list, as, for example, the present military standard specification system does, then these RHEPLs may also be used for part qualification.

The discussions which follow explain what decisions must be made before the RHEPL can be calculated and, once the required decisions are made, how they can be calculated from the characterization measurements. If the part can be purchased against any one of several radiation levels, as for example the eight ionizing dose radiation levels, M, D, P, L, R, F, G, AND H, shown for military standard parts in table I (see 4.2), then a separate RHEPL must be calculated for each radiation level. A device can cease to function as a result of a radiation exposure and will, obviously, fail the test being conducted. Such failures are not associated with the definition of an end- point limit.

Of course, the functional failure of a part will count as a failure for lot acceptance purposes. 5.2.2.1 One and two sided limits. A parameter used for determining whether a device has passed or failed a test, may be bounded by maximum and minimum values i.e., by “two-sided” limits, or by either a maximum or a minimum value i.e., by a one-sided limit. Because a one-sided limit is the one most commonly used in radiation response testing, only such limits are treated here. The mathematical formulas for the case of two-sided limits are almost identical to the one-sided case so the present discussion can easily be extended to that case if it is necessary to do so.

5.2.3 Criteria for calculating RHA end-point limits (RHEPL). 5.2.3.1 Selection of objectives for lot acceptance tests. Meaningful RHA end-point limits cannot be calculated until the objectives of the lot acceptance testing have been selected. Simply put, if the object of the tests is that at least 90 percent of future lots should pass the test, the end-point limits will be less stringent than if the object is that at least 80 percent of future lots should pass.

It should be noted, however, that the statistical uncertainties associated with the relatively small sample sizes that are typically used (tests are usually based on tens of devices but not hundreds) have the consequence that, almost regardless of where the RHEPL is set, there will be a significant probability that a good lot may fail or that a bad lot may pass the specified lot acceptance test. The real constraints on selecting an objective for the LAT and on calculating the corresponding RHEPL, therefore, are the performance of the part and the economic costs of rejecting lots.

Thus if the RHEPL is set so that, with 90 percent confidence, 10 percent of future lots may be expected to pass, then the performance of the lots that pass may be very good but the costs of rejecting such a high percentage of lots may make the part too expensive to use. Similarly, if the end-point limit is set so that 90 percent of future lots may be expected to pass, then the cost of the part may be attractive but the performance of the part may not be as good as may be desired. The objectives of the lot acceptance tests are thus seen as a trade-off between passing the largest percentage of lots and keeping end-point limits which will make the performance of the part desirable.

5.2.3.2 Lot rejection probability and choice of end-points. In the data analysis sections which follow, these guidelines show how lot rejection probability depends on the choice of end-point limits. They should, therefore, assist manufacturers or specification developers in selecting end-point limits which best meet their needs. The discussion also shows that lot to lot variability in radiation response can present serious difficulties to end-point limit selection. For this latter case, which is common, the methods recommended here may provide a better assessment of the risks associated with different values of the end-point limits than other methods which have been used.

5.2.3.3 Many parameters and environments. The most tractable case for calculating end-point limits that are to be used for lot acceptance tests is the case of a single radiation environment and a single device parameter. As soon as two or more different and independent test parameters are involved in lot acceptance, then the probabilities of passing for each parameter have to be higher by amounts such that the product of all the passing probabilities equals the probability of passing that is desired.

Thus for example, if two different and independent parameters are specified and the desired lot acceptance probability is 0.9, then the respective passing probabilities, P1 and P2, have to be such that (P1)*(P2) = 0.9. The complexities of the calculations for combined radiation environments or parameters or both are such as to put them outside the scope of this document. Unless otherwise noted, therefore, the analysis examples given in the sections which follow are for a single radiation environment and a single device parameter. 5.2.3.3.1 Many parameters.

In practice, device specifications generally do have to require that post radiation tests be made on several parameters against specified end-point limits. The recommendation here is that, whenever possible, the characterization data be used to select the most sensitive critical parameter and that the end-point limit be calculated for that parameter to give the desired lot acceptance probability. The end-point limits for the other relevant parameters then must be calculated to give sufficiently high passing probabilities so that those parameters will not cause a significant number of lot failures.

Methods are presently being developed to deal with the case when two or more independent parameters have comparable likelihood’s of causing failure but they are not yet in a form which could be included here. SUPERSEDES PAGE 28 OF MIL-HDBK-816 28