Freescale Semiconductor, Inc.
Reliability Issues for Silicon Pressure Sensors
combinations. Once the testing has been completed predic-
by Theresa Maudie and Bob Tucker
Sensor Products Division
tions can be made for the intended product customer base.
If a failure would be detected during reliability testing, the
cause of the failure can be categorized into one of the
following: design, manufacturing, materials, or user. The
possible impact on the improvements that may need to be
made for a product is influenced by the stage of product
development. If a product undergoes reliability testing early
in its development phase, the corrective action process can
generally occur in an expedient manner and at minimum
cost. This would be true whether the cause of failure was
attributed to the design, manufacturing, or materials. If a
reliability failure is detected once the product is in full
production, changes can be very difficult to make and
generally are very costly. This scenario would sometimes
result in a total redesign.
The potential cause for a reliability failure can also be
user induced. This is generally the area that the least
information is known, especially for a commodity type
manufacturer that achieves sales through a global distribu-
tion network. It is the task of the reliability engineer to best
anticipate the multitudes of environments that a particular
product might see, and determine the robustness of the
product by measuring the reliability lifetime parameters.
The areas of design, manufacturing, and materials are
generally well understood by the reliability engineer, but
without the correct environmental usage, customer satis-
faction can suffer from lack of optimization.
Revised June 9, 1997
ABSTRACT
Reliability testing for silicon pressure sensors is of
greater importance than ever before with the dramatic
increase in sensor usage. This growth is seen in applica-
tions replacing mechanical systems, as well as new
designs. Across all market segments, the expectation for
the highest reliability exists. While sensor demand has
grown across all of these segments, the substantial increase
of sensing applications in the automotive arena is driving
the need for improved reliability and test capability. The
purpose of this paper is to take a closer look at these reli-
ability issues for silicon pressure sensors.
INTRODUCTION
Discussing reliability as it pertains to semiconductor elec-
tronics is certainly not a new subject. However, when devel-
oping new technologies like sensors how reliability testing
will be performed is not always obvious. Pressure sensors
are an intriguing dilemma. Since they are electromechanical
devices, different types of stresses should be considered to
insure the different elements are exercised as they would be
in an actual application. In addition, the very different
package outlines relative to other standard semiconductor
packages require special fixtures and test set-ups. However,
as the sensor marketplace continues to grow, reliability
testing becomes more important than ever to insure that
products being used across all market segments will meet
reliability lifetime expectations.
RELIABILITY STATISTICS
Without standardization of the semiconductor sensor stan-
dards, the end customer is placed in a situation of possible
jeopardy. If non-standard reliability data is generated and
published by manufacturers, the information can be
perplexing to disseminate and compare. Reliability lifetime
statistics can be confusing for the novice user of the informa-
tion, “let the buyer beware”.
The reporting of reliability statistics is generally in terms of
failure rate, measured in FITs, or failure rate for one billion
device hours. In most cases, the underlying assumption
used in reporting either the failure rate or the MTBF is that
the failures occurring during the reliability test follow an expo-
nential life distribution. The inverse of the failure rate is the
MTBF, or mean time between failure. The details on the
various life distributions will not be explored here but the key
concern about the exponential distribution is that the failure
rate over time is constant. Other life distributions, such as the
lognormal or Weibull can take on different failure rates over
time, in particular, both distributions can represent a wear out
or increasing failure rate that might be seen on a product
reaching the limitations on its lifetime or for certain types of
failure mechanisms.
RELIABILITY DEFINITION
Reliability is [1] the probability of a product performing its
intended function over its intended lifetime and under the
operating conditions encountered. The four key elements of
the definition are probability, performance, lifetime, and
operating conditions. Probability implies that the reliability
lifetime estimates will be made based on statistical tech-
niques where samples are tested to predict the lifetime of
the manufactured products. Performance is a key in that the
sample predicts the performance of the product at a given
point in time but the variability in manufacturing must be
controlled so that all devices perform to the same functional
level. Lifetime is the period of time over which the product is
intended to perform. This lifetime could be as small as one
week in the case of a disposable blood pressure transducer
or as long as 15 years for automotive applications. Environ-
ment is the area that also plays a key role since the oper-
ating conditions of the product can greatly influence the
reliability of the product.
The time duration use for the prediction of most reliability
statistics is of relatively short duration with respect to the
product’s lifetime ability and failures are usually not
observed. When a test is terminated after a set number of
hours is achieved, or time censored, and no failures are
observed, the failure rate can be estimated by use of the chi-
square distribution which relates observed and expected
Environmental factors that can be seen during the lifetime
of any semiconductor product include temperature, humidity,
electric field, magnetic field, current density, pressure differ-
ential, vibration, and/or a chemical interaction. Reliability
testing is generally formulated to take into account all of
these potential factors either individually or in multiple
Motorola Sensor Device Data
www.motorola.com/semiconductors
1–3
For More Information On This Product,
Go to: www.freescale.com
MOTOROLA [ MOTOROLA ]
