Reliability Analysis

OVERVIEW

The purpose of a reliability program is to ensure reliable products and systems. Such a program is intended to help a product manager or designer face the many reliability issues. Reliability is a performance characteristic concerned with the probability of success and frequency of failures. It is the probability that a device will perform under established conditions for a specified time or over its useful life usually stated in the specification sheet.

Often performance is thought of in terms of speed, capacity or other “in spec” measures. On the other hand, if a device fails so often it’s seldom usable, the “in spec” measurements are not relevant. As well, Reliability is critical to safety and liability.

Most semiconductor manufacturers assess the reliability of finished die/wafers thru the establishment of a reliability monitoring program. A matrix of products representing key wafer fabrication processes (as well as subcontractor supplied products) is periodically sampled and subjected to a full range of accelerated reliability stress tests. Typically, the primary stress tests utilized are Operating Life Tests (OPL), Temperature Humidity Bias Tests (THBT), Autoclave (ACLV) and Temperature Cycling (TMCL). The Data from testing is utilized to track the Early Failure Rate (EFR) and Long Term Failure Rate (FITS) for specific processes. Early Failure Rates are listed in parts per million (PPM) and Long Term Failure Rates in units per billion device hours (FITS) for the different wafer processing and fabrication operations.

Packaged parts are discrete circuit elements that lend themselves to such analysis by the device manufacturer since the manufacturer controls the fabrication process of the subject device. However, from the customer’s standpoint, the issue of reliability analysis is focused on next-level assemblies. The principles of reliability analysis apply to assembled products just as they apply to individual devices.

Unpackaged die are not easily isolated as discrete circuit elements for reliability analysis by the customer. Therefore, in the case of bare die mounted in multi-chip-modules (MCMs), the MCM assembly as a whole is the focus of reliability analysis rather than the individual die.

For a system, there are two major types of reliability predictions: parts count and stress. During the system’s design, the reliability prediction will be used to make design trade-offs in the areas of device derating. Wear-out mechanisms must also be looked at with respect to preventive maintenance and product warranties. It is also essential for safety purposes to understand how the product behaves in a faulted condition. The reliability engineer uses a fault tree analysis tool to understand a product's fault behavior.

Availability is the probability that the product will perform its function when needed. Using fault tree data with the results of an availability model, the design engineer can weigh the use of various redundancy techniques in order to meet customer requirements.


		Reliability Program Basics				Reliability Standards & Handbooks