Rome Laboratory Old
Rome Laboratory Old
Rome Laboratory Old
RL-TR-92-197
Final Technical Report
July 1992
RELIABILITY ASSESSMENT OF
CRITICAL ELECTRONIC
COMPONENTS
I IT Research Institute
William K. Denson
Rome Laboratory
Air Force Systems Command
Griffiss Air Force Base, NY 13441-5700
i
This report has been reviewed by the Rome Laboratory Public Affairs
Office (PA) and is releasable to the National Technical Information Service
(NTIS). At NTIS it will be releasable to the general public, including
foreign nations.
APPROVED:
liM^f
RICHARD A. HYLE, Jr.
Project Engineer
JOHN J . BART
Chief Scientist
Electromagnetics & Reliability Directorate
If your address has changed or if you wish to be removed from the Rome
Laboratory mailing list, or if the addressee is no longer employed by
your organization, please notify RL (ERSD) Griffiss AFB NY 13441-5700.
This will assist us in maintaining a current mailing list.
1. AGENCY USE ONLY (Leave Blank) Z REPORT DATE 3. REPORT TYPE AND DATES COVERED
July 1992 Final Mar 89 - Jun 92
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
RELIABILITY ASSESSMENT OF CRITICAL ELECTRONIC COMPONENTS C - F306O2-89-C-006;
PR - • i >«
TA - ' - 2
6. AUTHOR(S)
WU - -,E
William K. Denson
J This document presents failure rate prediction procedures for resistors, capacitors,
switches, inductive devices, relays, connectors, interconnection assemblies and
rotating devices. Data were collected from military maintenance records, warranty
records, published information and field operations to support the model development.
The existing failure rate models were examined to identify areas that were deficient
and needed to be updated/revised. The objectives were: 1) Be reflective of current
state-of-the-art in part manufacturing technology; 2) Include all part types being used
in military systems; 3) Be based only on information that is available during design
phases; 4) Be as accurate and precise as possible. The goal of the new model develop-
ment was to simplify the models in a manner that made their complexity consistent with
their precision and accuracy, while at the same time including provisions to account
for the primary variable affecting reliability. A new prediction methodology was
developed to model the failure rate of devices that exhibit wearout failure mechanisms
(i.e., switches, relays, etc.). Additional new part types have been added (i.e.,
ceramic chip capacitors, tantalum chip capacitors, etc.) which may be included in
MIL-HDBK-217E.~7
17. SECURITY CLASSIFICATION 1 a SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LIMITATION OF ABSTRACT
OF REPORT OF THIS PAGE OF ABSTRACT
UNCLASSIFIED UNCLASSIFIED UNCLASSIFIED U/L
NSN 7540-01-280-5500 Standard Farm 298 (Rev 2-89)
Presetted by ANSI Std Z39-18
2SB-102
EXECUTIVE SUMMARY
The objective of this study was to update the MIL-HDBK-217 failure rate prediction models
for Capacitors, Resistors, Inductive Devices, Switches, Relays, Connectors, Interconnection
Assemblies and Rotating Devices. These models were developed or modified primarily from the
statistical analysis of field failure rate data collected during this study. This data was collected
mainly from military maintenance records with additional information and data collected from
warranty records, published information and laboratory test results. Particular attention was given
to the requirement that all data used in support of the models be of high quality. To address this,
IITRI used only that data for which there existed confidence that it indeed was accurate.
An objective of this model development exercise was also to simplify the models in a manner
that made their complexity consistent with their precision and accuracy, while at the same time
including provisions to account for the primary variables affecting reliability.
Each part type was studied to determine their primary modes and mechanisms of failure.
This information was used to structure a hypothetical model whose factors were then quantified
from analysis of failure rate data. All reliability models relied on field data except for
interconnection assemblies which used laboratory test data. Laboratory test data was used because
the model for interconnection assemblies predicts the number of thermal cycles to failure and its
development thus relied on cycle to failure data which is only available through laboratory tests.
A new prediction methodology was also developed to model the failure rate of devices that
exhibit wearout failure mechanisms. Devices exhibiting these mechanisms, and those modeled
accordingly, are; switches, relays and interconnection assemblies (which include Plated Through
Holes (PTH) and Surface Mount Technology (SMT)). This methodology essentially converts a
time to failure statistic such as Mean-Time-to-Failure (MTTF) or characteristic Life (a) to an
average failure rate over the design Life Cycle or preventative maintenance interval. Since a closed
form solution for the calculation of this average failure rate is not possible, it was accomplished by
means of Monte-Carlo simulations.
The change in predicted failure rate between the models proposed herein and the existing
MIL-HDBK-217 models varied significantly from part type to part type. However, from the
comparison of the proposed models to the existing models, the following conclusions can be made:
(1) Capacitor failure rates are generally lower than MIL-HDBK-217E models, although
they exhibited a higher dependence on environment.
(2) Film resistors and resistor networks were approximately consistent with MIL-HDBK-
217E, and composition were slightly higher.
(3) Predicted failure rates for inductive devices are generally consistent with MIL-HDBK-
217E models.
(4) The predicted failure rates for switches and relays are generally much higher, have a
much higher dependence on environment, and lower dependence on quality than MIL-
HDBK-217E models.
(5) The predicted failure rate for connectors is generally lower than MIL-HDBK-217E
models.
(6) The predicted failure rate of interconnection assemblies/printed wiring boards depend
much more on specific design attributes, and therefore can be either higher or lower
than MIL-HDBK-217E model.
(7) The electric motor predicted failure rates are generally consistent with MIL-HDBK-
217E.
The above comparisons are qualitative since the actual ratio of new model to the MIL-HDBK-
217E model can vary significantly depending on the specific variables used in the prediction.
ii
Acronyms/Symbols
111
TABLE OF CONTENTS
IV *
TABLE OF CONTENTS (CONT'D)
Page
4.5 RELAYS 4-65
4.5.1 Relay Failure Modes/Mechanisms 4-68
4.5.2 MIL-HDBK-217E Relay Models Review 4-68
4.5.3 Relay Model Development 4-69
4.5.3.1 Hypothesized Relay Model 4-69
4.5.3.2 Relay Data Analysis 4-70
4.6 CONNECTORS 4-74
4.6.1 Connector Failure Modes/Mechanisms 4-76
4.6.2 MIL-HDBK-217E Connector Model Review 4-80
4.6.3 Connector Model Development 4-81
4.6.3.1 Hypothesized Connector Model 4-81
4.6.3.2 Connector Model Development 4-82
4.6.3.2.1 Connectors 4-82
4.6.3.2.2 Connections 4-84
4.6.3.2.3 Sockets 4-85
4.7 INTERCONNECTION ASSEMBLIES/PRINTED WIRING BOARDS 4-88
4.7.1 Interconnection Assembly Failure Modes and Mechanisms 4-88
4.7.2 Interconnection Assembly/Printed Wiring Board MIL-HDBK-217E
Model Review 4-98
4.7.3 Interconnection Assembly Model Development 4-98
4.8 ROTATING DEVICES 4-122
4.8.1 Rotating Device Failure Modes and Mechanisms 4-122
4.8.2 Current MIL-HDBK-217E Motor Model Review 4-124
4.8.3 Rotating Device Model Development 4-126
4.8.3.1 Hypothesized Motor Model 4-126
4.8.3.2 Motor Data Analysis 4-127
5.0 MODEL SUMMARY AND SAMPLE CALCULATIONS 5-1
5.1 MODEL SUMMARY 5-1
5.2 SAMPLE CALCULATIONS 5-42
6.0 MODEL COMPARISON 6-1
6.1 MODEL COMPARISON OBSERVATIONS 6-3
7.0 CONCLUSIONS AND RECOMMENDATIONS 7-1
8.0 REFERENCES 8-1
APPENDICES
APPENDIX A: DETAILED DATA A-l
APPENDIX B: PART PARAMETERS B-l
APPENDIX C: MONTE-CARLO SIMULATIONS C-l
APPENDIX D: REGRESSION RESULTS D-l
v
LIST OF TABLES
Page
TABLE 2.3-1: APPROXIMATE TIMES AT WHICH ASYMPTOTIC
FAILURE RATES WERE REACHED 2-14
TABLE 2.3-2: CUMULATIVE FAILURE RATE SUMMARY 2-15
TABLE 2.3-3: PERCENT FAILED AT MTTF AS A FUNCTION OF (3 2-17
TABLE 2.3-4: a/MTTF RATIO AS A FUNCTION OF (3 2-18
TABLE 3.0-1: DATA SOURCES 3-2
TABLE 3.0-2: DATA SUMMARIZATION PROCEDURE 3-4
TABLE 3.0-3: ADDITIONAL DATA SOURCES USED 3-4
TABLE3.2-1: SUMMARY OF DATA COLLECTED 3-6
TABLE 3.2-2: PART SPECIFICATIONS 3-8
TABLE 4.1-1: VARIABLE CAPACITOR FAILURE MODES 4-4
TABLE 4.1-2: AL ELECTROLYTIC FAILURE MODES 4-5
TABLE 4.1-3: TANTALUM WET SLUG FAILURE MODES 4-6
TABLE 4.1-4: SOLID TANTALUM FAILURE MODES 4-7
TABLE 4.1-5: TANTALUM FAILURE MODES 4-8
TABLE 4.1-6: MICA AND GLASS FAILURE MODES 4-9
TABLE 4.1-7: CERAMIC FAILURE MODES 4-10
TABLE 4.1-8: PLASTIC AND PAPER FAILURE MODES 4-11
TABLE 4.1-9: CAPACITOR QUALITY FACTOR 4-15
TABLE 4.1-10: CAPACITOR ACTIVATION ENERGIES 4-17
TABLE 4.1-11: OPERATING TEMPERATURES 4-18
TABLE 4.1-12: OBSERVED ENVIRONMENT FACTORS 4-20
TABLE 4.1-13: FIXED VS. VARIABLE FACTOR 4-21
TABLE 4.1-14: DIELECTRIC FACTOR 4-21
TABLE 4.1-15: BASE FAILURE RATE 4-22
TABLE 4.1-16: SOLID TANTALUM LIFE DATA 4-23
TABLE 4.1-17: VALUES OF n FOR VARIOUS CAPACITOR TYPES 4-24
TABLE 4.1-18: PROPOSED n VALUE 4-25
TABLE 4.2-1: COMPOSITION RESISTOR FAILURE MECHANISMS 4-31
TABLE 4.2-2: FILM RESISTOR FAILURE MECHANISMS 4-32
TABLE 4.2-3: WIREWOUND RESISTOR FAILURE MECHANISMS 4-33
TABLE 4.2-4: VARIABLE COMPOSITION RESISTOR FAILURE
MECHANISMS 4-34
TABLE 4.2-5: VARIABLE WIREWOUND RESISTOR FAILURE
MECHANISMS 4-35
TABLE 4.2-6: THERMISTOR FAILURE MECHANISMS 4-36
TABLE 4.2-7: OBSERVED RESISTOR ENVIRONMENT FACTORS 4-39
TABLE 4.2-8: CURRENT ENVIRONMENT FACTORS 4-39
TABLE 4.2-9: 217E/DERIVED ENVIRONMENT COMPARISON 4-40
TABLE 4.2-10: RESISTOR ENVIRONMENT FACTORS 4-40
TABLE 4.2-11: RESISTOR ACTIVATION ENERGIES 4-41
TABLE 4.2-12: RESISTOR BASE FAILURE RATES 4-42
TABLE 4.2-13: RESISTOR NETWORK DATA 4-43
TABLE 4.3-1: INDUCTOR FAILURE MODES AND MECHANISMS 4-45
TABLE 4.3-2: TRANSFORMER FAILURE MECHANISMS 4-45
TABLE 4.3-3: RF COIL FAILURE MECHANISM DISTRIBUTION 4-45
TABLE 4.3-4: OBSERVED ENVIRONMENT FACTORS 4-48
TABLE 4.3-5: TRANSFORMER BASE FAILURE RATES 4-49
TABLE 4.3-6: OBSERVED INDUCTOR ENVIRONMENT FACTORS 4-50
TABLE 4.3-7: INDUCTOR BASE FAILURE RATES 4-50
VI
LIST OF TABLES (CONT'D)
Page
TABLE 4.4-1: CONTACT MATERIAL PROPERTIES IMPACT SWITCH
RELIABILITY 4-52
TABLE 4.4-2: SWITCHES, GENERAL FAILURE MODES 4-54
TABLE 4.4-3: FLOAT SWITCH FAILURE MODES 4-55
TABLE4.4-4: REED SWITCHES FAILURE MODES 4-55
TABLE 4.4-5: TOGGLE SWITCHES FAILURE MODES 4-55
TABLE 4.4-6: SWITCH BASE FAILURE RATES 4-58
TABLE 4.4-7: ROTARY SWITCH BASE FAILURE RATES 4-59
TABLE 4.4-8: CIRCUIT BREAKER ENVIRONMENT FACTOR 4-60
TABLE 4.4-9: CIRCUIT BREAKER QUALITY FACTOR 4-60
TABLE 4.4-10: CONTACT CONFIGURATION FACTOR 4-60
TABLE 4.4-11: CIRCUIT BREAKER BASE FAILURE RATES 4-61
TABLE 4.4-12: CONTACT LIFE EXPECTANCY (106 ACTUATIONS) 4-62
TABLE 4.4-13: DRY REED CONTACT DATA 4-63
TABLE 4.5-1: TESTS PERFORMED TO ASSURE RELAY RELIABILITY 4-67
TABLE 4.5-2: ARMATURE RELAY FAILURE MECHANISMS 4-68
TABLE 4.5-3: EFFECTS ON RELAY QUALITY ON CYCLING FACTOR 4-69
TABLE 4.5-4: CURRENT 217E ENVIRONMENT FACTOR 4-70
TABLE4.5-5: REGRESSION ANALYSIS 4-70
TABLE 4.5-6: COMPARISON OF NEW/OLD ENVIRONMENT FACTORS 4-71
TABLE 4.5-7: PROPOSED RELAY ENVIRONMENT FACTOR 4-72
TABLE 4.5-8: OBSERVED QUALITY FACTOR 4-72
TABLE 4.5-9: RELAY BASE FAILURE RATES 4-73
TABLE 4.6-1: CONNECTOR FAILURE MODES/MECHANISMS 4-77
TABLE 4.6-2: OBSERVED ENVIRONMENT FACTOR 4-82
TABLE 4.6-3: CONNECTOR BASE FAILURE RATES 4-83
TABLE4.6-4: CONNECTION BASE FAILURE RATES 4-85
TABLE 4.6-5: OBSERVED FAILURE RATES FOR SOCKETS 4-86
TABLE4.6-6: DIP SOCKET DATA 4-86
TABLE 4.7-1: ENVIRONMENT AT VALUES 4-106
TABLE4.7-2: X-Y TCE VALUES 4-108
TABLE 4.7-3: TCE'S OF PACKAGE MATERIALS 4-109
TABLE 4.7-4: PTH/VIA MATERIAL TCE VALUES 4-109
TABLE 4.7-5: Z AXIS TCE VALUES 4-110
TABLE 4.7-6: LEAD CONFIGURATION N f (REF. #66) 4-110
TABLE 4.7-7: LEAD CONFIGURATION FACTOR 4-111
TABLE 4.7-8: CYCLING RATE VALUES 4-112
TABLE 4.7-9: PTHDATA 4-114
TABLE 4.7-10: SUMMARY OF LCC WEAROUT DATA 4-116
TABLE 4.7-11: BASE FAILURE RATE Xh 4-120
TABLE4.7-12: QUALITY FACTOR 7TQ 4-120
TABLE4.7-13: COMPLEXITY FACTOR KQ 4-121
TABLE 4.7-14: ENVIRONMENT MODE FACTORS 4-121
TABLE 4.8-1: MOTOR DATA ANALYSIS 4-129
a(observed)
TABLE 4.8-2: —^ • 4-130
a(217E)
vn
LIST OF TABLES (CONT'D)
Page
TABLE 4.8-3: CUMULATIVE AVERAGE FAILURE RATE 4-132
TABLE4.8-4: A,B CONSTANT 4-132
TABLE 4.8-5: BEARING & WINDING CHARACTERISTICS
LIFE, a B & a w , vs. AMBIENT TEMPERATURE, T 4-133
TABLE 6.0-1: MODEL COMPARISON 6-2
LIST OF FIGURES
vm
1.0 INTRODUCTION
The purpose of this study effort was to update the failure rate prediction models contained in
MIL-HDBK-217E, "Reliability Prediction of Electronic Equipment" for:
Resistors
Capacitors
Inductive Devices
Switches
Relays
Connectors
Interconnection Assemblies
Rotating Devices
This was accomplished for each part type by reviewing the existing models, identifying areas
needing updating or revising, studying the failure physics, collecting failure rate data,
hypothesizing a model, statistically analyzing the data, and using all information and data available
to construct new or revised models. Objectives of these models are that they:
(3) Be based only on information that is available during equipment design phases,
In failure rate modeling of components, defect related failure mechanisms (special cause) and
inherent failure mechanisms (common cause) must be treated separately. With a few exceptions,
the predominant failure mechanisms of the parts being modeled herein are special cause. For parts
that exhibit these mechanisms as being predominant, the best model that can be derived is a
statistical regression model from field experience data. To accomplish the above modeling
objectives, field failure rate data collected from a wide variety of sources was statistically analyzed.
Since the data was collected from a variety of sources and from various manufacturers, the models
1-1
will be representative of industry average failure rates and will predict the failure rate for the
"average" manufacturer. They will also be indicative of how well the part manufacturers, as a
whole, are able to control their processes, and how defect free they are able to manufacture them.
Since the majority of failures in the early and mid life of electronic parts are related to some
form of defect and are highly process related, the observed failure rates can vary significantly as a
function of manufacturer. It would intuitively seem logical that the variability of military parts
manufactured and screened in accordance with the applicable specifications would exhibit a smaller
degree of variation than commercial quality parts. However, this decreased variability typically
cannot be observed from field data, possibly due to the fact that there is inherently greater variation
in military environmental stresses, thus masking any decreased variability that may be present.
One way to account for variability and increase the precision of the model is to require detailed
process specific information as an input to the prediction model. It is typically not feasible to
require such information as an input to the model, since such information is only available to the
part manufacturer. Examples of this information are defect density, contamination levels, material
compositions, and statistical process control information. These inherent limitations in the type of
data that can be used as input to the failure rate models such as those in MTL-HDBK-217 highlight
the fact that such models are generic, industry average models and not manufacturer specific.
Other objectives of this study were to simplify the models, make them more consistent with
other models in MIL-HDBK-217, and to make their complexity consistent with their accuracy and
precision. For example, there is currently a separate set of environment factors for individual types
of resistors. Most other models in the handbook, including microcircuits, have only one set of
environment factors. Given the precision and accuracy of the prediction model expected, and the
fact that it is generally impossible to distinguish the difference in environmental effects for each
individual resistor type from field data, it is proposed that a single quality and environment factor
be used for a generic part type (such as resistors, capacitors, switches, etc.). The exception to this
is that if, within a generic component category, there exist part types exhibiting different
predominant failure mechanisms.
1-2
2.0 FAILURE RATE MODELING
A general failure rate modeling approach was defined to provide the basic structure for the
failure rate prediction model development process. Figure 2-1-1 presents the model development
approach and the following paragraphs briefly describe the primary tasks in this approach.
START
COLLECT
DEVELOP THEORETICAL DATA
MODEL
DATA
QC
2-1
2.1.1 Identify Potential Variables
The first step of the model development process was to identify variables which could
potentially have an effect on failure rate. These variables were limited to information available to
engineers during equipment design phases. Determination of these variables was based on physics
of failure information. Appendix B lists the variables tracked (if available) for each part type being
modeled. All variables listed are potential model parameters and are either a function of device
construction/design, circuit application, application environment, or a combination thereof. The
identification of these parameters early in the data collection phase served to focus the data
collection efforts and refine the theoretical models.
Effective data collection was critical to the successful completion of this effort. Details of this
portion of the effort are presented in Section 3.0 of this report.
A series of theoretical failure rate prediction models was hypothesized to provide the resultant
models with a sound theoretical/engineering backing. Basically, theoretical model development
involved evaluation of the effects of the parameters identified in the "Identify Potential Variables"
phase. In addition, the optimal model form (i.e., additive, multiplicative, combination) was
determined and the time dependency of each part types failure rate was studied.
The failure rate models proposed consist of two additive failure rate terms, of which one or
both are applicable to each part type. The first is a constant failure rate term associated with
random failures due to defects or event related failure mechanisms. This contribution term cannot
be modeled with a physics -of- failure approach and therefore is generally a multiplicative model in
which the factors represent the predominant failure accelerating variables. Since it is primarily a
defect related failure rate, it is an industry average failure rate and represents the capabilities of
current manufacturing technologies. The second term models wearout failure mechanisms. These
are usually referred to as common cause and are inherent mechanisms. Physics of failure
approaches are applicable to these failure mechanisms since they are generally more understood
than defect related mechanisms.
2-2
These two terms are additive since they are typically separate failure mechanisms for which
different modelling approaches are taken. For example, wearout failure mechanisms are modeled
with time-to-failure distributions (Lognormal or Weibull with beta > 1) whereas defect related
failure mechanisms are typically modeled with a constant failure rate.
If the failure mechanisms being modeled are independent, the failure rates associated with
each can be added. An example of this is relays in which one potential failure mode is binding of
the moving mechanism. This most likely is due to a combination of part defect and
environmental/use conditions. Since it is primarily a defect related failure mechanism, it can be
modeled with a constant failure rate. An example of a potential common cause failure mode is the
arcing and resulting high resistance material formed between the contacts during the switching
operation. This mechanism is a result of the use and load conditions to which the relay has been
subjected. It is a wearout failure mechanism for which an increasing failure rate (such as Weibull
with [3 > 1) is appropriate. Since these two mechanisms are statistically independent, the failure
rates associated with each can be added to derive the total failure rate.
Several current MIL-HDBK-217E models include provisions for the failure rate to increase
dramatically when the maximum electrical or temperature stress is approached. Examples of this
are capacitors which have these provisions for voltage and temperature, and resistors which have it
for temperature. Although stresses of these levels will undoubtedly adversely affect the failure
rate, it is very difficult to quantify the failure rate under these stress conditions, particularly because
different failure mechanisms are predominant than in the case where the part is used within its rated
stresses. This difficulty, coupled with the fact that most other models in MIL-HDBK-217 do not
include these provisions, has led IITRI to propose that the new models do not include these
provisions. Therefore, it must be understood and clearly noted that the models are valid only for
situations in which the parts are applied in a manner which stresses them below their rated values.
Additionally, the models are only valid within the range of stresses of the data on which the model
is based.
A general rule that IITRI followed in development of the constant failure rate defect portion
of the models was to include only those factors that were observed to significantly affect reliability.
Model factors unsubstantiated by empirical data were only included in cases where parameters are
known to effect reliability. Example of these types of factors include temperature, environment,
and quality.
2-3
Development of the theoretical models relied heavily on published literature. The literature
included many instances of mathematical models relating failure rate (or mean-time-to-failure) to
temperature, power, derating and other factors. Many other technical articles or documents
provided a qualitative assessment of reliability influences. These were useful to define the relative
effect of numerous variables. In very general terms, the theoretical models (constant failure rate
portion) were of the following form.
1=1
where
= exp (-AC= - j - ) )
where
The development of theoretical device failure rate prediction models was an integral part of
the overall model development process. Information collected through the literature review and
vendor surveys was reviewed and evaluated to aid in the development of theoretical models for
each component type. The theoretical models serve the following functions:
2-4
2.1.4 Data Analysis
The next phase of the modeling approach was data analysis of the failure rate data collected
through an intensive data collection effort (described in Section 3.0). Techniques used were
correlation coefficient analysis, regression analysis, goodness-of-fit testing and others. These are
described in the following paragraphs.
The first data analysis task was correlation coefficient analysis. The objective of this analysis
was to identify highly correlated variables. As part of this task, correlation coefficients were
computed for each pair of independent variables. The correlation coefficient is a measure of the
relation between two variables and varies between -1 and 1 (from perfect negative to perfect
positive correlation). Regression analysis requires that all independent variables are uncorrelated;
therefore, the effects of correlated variables could not be simultaneously quantified. If the
variables were correlated inherently (e.g., temperature and power), a decision was made to include
only the most significant variable in the regression analysis. If the variables were correlated due to
chance (e.g., quality vs. temperature), then several options were considered. If a valid theoretical
or empirical relationship was found for one of the correlated variables, then the effect of that
variable was removed from the data by assuming the relationship to be correct. If this assumption
was correct, then the effect of the remaining correlated variable could be accurately assessed by
data analysis.
The next step in the model development process was to apply stepwise multiple regression
analysis. Regression analysis is described in detail in Draper and Smith (Reference 2). This
technique was used to compute the coefficients of an assumed model form in a least squares fit to
the data. Regression solutions were found for decreasing confidence limits beginning with 90%.
In addition, standard error statistics were computed for each significant variable to obtain an
indication of the accuracy of coefficient estimates. Additionally, upper and lower 90% confidence
interval values were determined for each coefficient. In general, variables were not included in the
proposed model if they did not significantly affect failure rate with at least 70% confidence.
However, if a variable such as device quality was known to influence failure rate from an
engineering perspective, then coefficients were computed with less than 70% confidence and a
corresponding factor was proposed. In these instances, the resultant factor should be considered
approximate.
Generally, transformations were performed on the data to yield multiplicative model forms.
To accomplish this, a logarithmic transformation of the failure rate was made so that a linear
2-5
regression could be accomplished. For example, multiple linear regression analysis assumes a
model of the following form;
Y = b0 + b 1 X 1 + b2X2 + ...bnXn + E
where Y is the dependent variable (in this case failure rate), Xj's are the independent variables, bj's
are the coefficients to be estimated by the analysis, and E is the residual error. Since a
multiplicative model was generally used for the models herein, a logarithmic transformation of the
failure rate was required before the regression analysis was performed. Once the coefficients were
derived from the analysis, the antilogarithm was taken to yield the final model. As another
example, the effect of junction temperature is often modeled by use of the equivalent Arrhenius
relationship, which indicates that the failure rate is a function of temperature, and takes the form,
X = Aexp(-B/T)
where T is the temperature, X is the failure rate and A and B are constants. By taking the natural
logarithm of each side, the equation becomes
\nX = InA - 7y
which can be solved by regression analysis with 1/T the independent variable and \nX the
dependent variable.
In addition to quantitative regression that was used to relate failure rate to continuous
variables such as temperature and rated power, qualitative regression techniques were also
employed. Qualitative regression (often termed covariance analysis) is used to model the effect of
variables which cannot be measured on a numerical scale (e.g., screen class). A matrix of
indicator variables (0 or 1) is defined and used as the independent variables to represent the
qualitative variable.
The F-ratio and Critical F are parameters which are used in conjunction with regression
analysis to determine significance of independent variables. The Critical F value corresponds to
the degrees of freedom of the model (equal to the number of data points minus the number of
coefficients minus one) and a specified confidence limit. This number may be used to test the
significance of each variable as it is considered for addition to or deletion from the model. The F-
2-6
ratio value for a regression is the quotient of the mean square due to regression and the mean
square due to residual variation. If the F-ratio value for any independent variable is greater than the
Critical F value, then it was considered a significant factor influencing failure rate and was included
in the regression solution.
The original data records were combined by adding the number of failures and dividing by
the total number of part hours for those records having the same variables being analyzed. In this
analysis, a record is generated for a specific part in a specific system. For each of these records,
there can be zero, one or more observed failures. A regression analysis was then performed on the
combined records that had one or more observed failures. This was done on failure records only
since it is impossible to run regressions on failure rates of zero. Observances of no failures does
not imply a failure rate of zero, but rather enough part hours have not been accrued to experience
failures. To address the problem of analyzing zero failure data points the following options were
considered:
(2) Use the lower 60% confidence level for zero failure data records, providing a minimum
number of operating hours have been observed. This translates to the assumption that
.9 failures have occurred in the given number of part hours.
(3) Use of a very low failure rate (i.e., several orders of magnitude lower than the lowest
observed failure rate) for zero failure records.
(4) Use of only those records with failures for model development and multiplication of the
derived base failure rates by the ratio: [observed hours without failures/total observed
hours]. For example, if 70 percent of the total part hours correspond to records with
failures, the failure rates derived from the regression analysis of the data records with
failures would be multiplied by .7.
Option 1 is not desirable since it ignores observed part hours with no failures and will result
in pessimistic prediction models. Option 2 is also not desirable since it, in essence, assumes
failures have occurred that in fact, have not. Option 3 alleviates the concerns of pessimistic
prediction models, but confounds the derivation of specific model factors. Option 4 is the best
2-7
available option since it 1) allows accurate quantification of relative model factors and 2) results in
an overall accurate model. This occurs since it is scaled in a manner that allows accurate prediction
of the entire population of parts regardless if there have been enough hours to observe failures in
the particular data set used to derive the model.
It is necessary to modify the predicted failure rate by the percentage of zero failure hours to
account for all observed hours after the regression results are obtained. The regression analysis
can only utilize non-zero failure rates and therefore only the failure records can be used to quantify
model variables. The zero failure records are only used to scale the predicted failure rates in
accordance with the behavior of the entire population. Therefore, the hours and failures of the
entire dataset cannot be used since only a subset (those with failures) are used to derive the model
variables.
A danger in developing models with multiple regression techniques is that the resulting
models can yield unrealistically high or low results if the extremes of model input variables are
used. The next phase of the model development process was therefore to perform an extreme case
analysis. Predictions were performed using the proposed model for parameters beyond the ranges
found in the data. The intent of the extreme case analysis was two-fold: (1) to identify any set of
conditions which cause the proposed model to numerically "blow up," (2) to identify any set of
conditions which predict a failure rate which is intuitively incorrect. For instance, a model that
predicted an unscreened device with a lower failure rate than a similar screened device or that
predicted a negative failure rate would be examples of an intuitively incorrect model. IITRI was
very sensitive to this effect and included models that have such extreme values only in cases where
it is justified from theoretical or empirical considerations. Reasons for failing the extreme case
analysis primarily involve an incorrect choice of model form. If the extreme case analysis indicated
that the proposed model was unacceptable, then the entire model development process was begun
again.
It is very important that the resulting models predict failure rates that are credible to practicing
reliability engineers. For this reason, the developed models were reviewed to ensure that they
yield results that are both reasonable and intuitively correct. To accomplish this, predicted failure
rates were calculated using typical parameter values. Level 2 derating requirements of Reference
76 will be used to define typical values and are used to normalize the models since the derating
values in that document represent typical and realistic values being used. The actual derating
2-8
values to be used for this purpose are not important, only that they are representative of current
design practices. The predicted failure rates were then analyzed to verify that they yield reasonable
results that are representative of typical observed values. If the model factors resulting from the
analysis are not reasonable from an engineering perspective, the factors causing the inconsistency
were deleted and the regression analysis was performed again. A portion of this effort was also to
identify and remove outlier data points that may not have been considered as such by the statistical
analysis. While such outliers were often obvious and discarded in the original dataset, there were
instances where selected data point(s) that were not considered outliers by the statistical analysis
were severely impacting the results.
Particular attention was given to the models that appear to be yielding excessively high or low
failure rates. If this was the case, each model exhibiting these characteristics was reevaluated and
corrected until reasonable and intuitive results are obtained.
Additionally, the models were analyzed relative to the existing MIL-HDBK-217E models.
For mature technologies, or cases where there is no obvious reason for failure rates to be getting
worse, the models were scrutinized to determine if the pessimistic failure rate is justified or
whether it is merely a statistical anomaly of the modeling process. It should also be noted that in
cases where the new models differ substantially from the old, it could be due to a lack of data in the
original dataset used to derive the MIL-HDBK-217E models.
The goodness-of-fit of the regression solution was then measured using the R-squared
statistic. The R 2 coefficient or multiple coefficient of determination is equal to the ratio of the sum
of squares of the deviations explained by the regression to the sum of the squares of the deviations
of the observed data. The R value was used as a means to determine the ability of the regression
model to predict the observed results. The coefficient ranges from 0 to 1.0. A coefficient value of
1.0 indicates a perfect fit between the model and the observed data. While there is no minimum
acceptable coefficient, higher values indicate better correlation between predicted and observed
failure rates. The range of R values in this analysis was from .30 to .78.
2-9
2.2 TEMPERATURE EFFECTS
An investigation into the effects of temperature was a crucial part of this failure rate modeling
effort. Based on the published literature, the impact of device temperature was determined to be an
important variable affecting the failure rate of most part types being modeled.
It was concluded in this study that, of the devices studied, the reliability of capacitors,
variable resistors, inductors, transformers, and motors exhibit a strong dependency on
temperature. It will be shown in Section 4.1.3.2 that for capacitors, the acceleration rates predicted
from analysis of accelerated life tests are much higher than those used historically in MEL-HDBK-
217. This could be due to higher acceleration rates at the highly accelerated test conditions relative
to field usage. With the exception of resistors, the other components types listed above have
similar reliability concerns to capacitors due to the similar nature of the insulating material.
Nevertheless, it is obvious that, for these part types, temperature must be accounted for in the
model. In general there was no evidence that, at field use conditions, the current MIL-HDBK-217
acceleration rates are erroneous. Therefore, for most of the applicable part types, current MIL-
HDBK-217 temperature acceleration factors will be used as a baseline to derive the new models.
While, in general, quality and/or environment were derived from analysis of the empirical
dataset, in no case during this effort could a temperature factor be derived from the empirical field
data due to the fact that an accurate operating temperature of the components was rarely known.
Although this uncertainty in temperature precludes derivation of a temperature factor from field
data, temperature is known from laboratory data to heavily influence the reliability of most part
types being modeled and must be accounted for. Alternative methods of deriving a temperature
factor were therefore used, such as; life test data, knowledge of temperature effects of failure
mechanism similar to those being modeled, results reported in the literature, and existing reliability
models.
Based on historical data, the Arrhenius relationship adequately models the reaction rate of
many failure mechanisms within a specific temperature range. The Arrhenius model is based on
empirical data and predicts that the rate of a given chemical or physical reaction, in this case a
failure mechanism, will be exponential with the inverse of temperature. Conceptually, the
Arrhenius model is given by:
2-10
where
Every chemical reaction has a unique activation energy associated with it. Most components
have several such reactions proceeding simultaneously, each capable either individually and/or
jointly of causing a part failure. However, consideration of each reaction separately would be too
complex to analyze with the available data. It has been found, however, that for general classes of
components with similar failure mechanism distributions the cumulative effects of the various
reactions can be approximated by an Arrhenius model for a specified temperature range. This
relationship has been designated as the "equivalent Arrhenius relationship." Because of the
documented accuracy of this approach and the limitations of the available data, it was decided to
investigate the effects of temperature using the equivalent Arrhenius relationship. It must be
emphasized that beyond the range of normal usage temperatures, this relationship will no longer be
applicable. It must also be noted that while the Arrhenius relationship was originally derived to
model chemical reaction rates, it is used herein as an empirical model describing the temperature
dependence of failure rate.
Several part types being modeled can exhibit wearout failure mechanisms. These part types
include: motors, switches, relays, surface mounted devices, connectors and Aluminum electrolytic
capacitors. If wearout failure mechanisms are the predominant reliability drivers for a particular
part type, a constant failure rate model clearly is not applicable.
IITRI has analyzed several alternative methods of modeling these device types, including:
2-11
Since it is desirable that the models to be developed be independent of time and based on a
constant failure rate, the use of Number 4 above is proposed. In this approach, an average failure
rate is calculated over the life cycle of the equipment in which the part is operating. The average
hazard rate over the life cycle cannot be used because it is a measure of the instantaneous failure
rate of a part under the condition that it has not yet failed. The condition of interest in this
modeling effort is the failure rate after a portion of the population has failed. This model is based
on the premise that parts are replaced upon failure and that an effective constant failure rate is
achieved after a given time due to the fact that the effective "time zero" of replaced parts become
random after a significant portion of the population is replaced.
Since this failure rate cannot be derived in a closed form, Monte Carlo simulations were
performed to estimate the failure rate of the Weibull distribution as a function of time, assuming
that parts are replaced upon failure, and assuming the Weibull distribution is valid. Since the term
failure rate implies a constant hazard rate from the exponential distribution, its use as a time varying
function is not entirely accurate. Therefore, some have referred to this time dependent failure rate
as the "Rate-of-Occurrence-of-Failure".
f(t), £ r if-i;(«)
a [a)
The time to failure of a given component that follows this pdf is:
I
TTFjj = 04 [- In (1 - RND)]^
where;
TTFj; = Time to failure of the i component which has been replaced j times. TTFjj
is relative to the "time zero" of the it'1 component
OCJ - Weibull characteristic life, time at which 63.2% of the population will have
failed (without replacements)
2-12
Weibull Shape Parameter
The "Rate of Occurrence of Failure" was calculated for each (3 value in the following manner:
1. The TTF for component i = 1 was calculated. This process was repeated for 100
failures of the i = 1 component 0 = 1- 100)
The following figure illustrates this concept (dots represent failure in time);
IOOO , . , , , , 11 i r i
0 la time
2-13
Ten simulations were performed, using a = 1 and varying the beta value from 1 to 10.
Appendix C presents the actual results of these simulations. Since the simulations were performed
with a = 1, the results can be converted to an actual situation by using an a in absolute time units.
It can be seen from these results that the failure rate of a's greater than one starts out very
low, increases when the hazard rate of the initial population starts to increase, oscillates as parts are
being replaced, and reaches an asymptotic value after some period of time. The actual failure rate
unit of these simulations is failures per 1000 components per .1 a. Therefore, dividing by 100
yields the unit failures per alpha.
The asymptotic failure rate, regardless of beta, is very close to one. The times at which the
asymptote is reached, however, is dependent on beta. These values are illustrated in Table 2.3-1.
TABLE 2.3-1:
APPROXIMATE TIMES AT WHICH ASYMPTOTIC FAILURE
RATES ARE REACHED
beta asymptote
2 la
4 2.4a
6 4.2a
8 7.0a
10 11a
An average cumulative failure rate was then calculated as a function of beta and the Life Cycle
(LC)/alpha ratio. These average failure rates are summarized in Table 2.3-2 and in Figure 2.3-1.
The values summarized in this table are average failure rates from time 0 to time LC/a and were
computed by dividing the total simulated number of failures by the time (in units of a).
The units of the average cumulative failure rate are in failures per alpha. Dividing the
cumulative failure rate by a (in 10" hours) yields a failure rate of F/10" hrs.
2-14
TABLE 2.3-2:
CUMULATIVE FAILURE RATE SUMMARY
P
LC
a 2 3 4 5 6 7 8 9 10
.1 .41 .13 .02 0.0 0.0 0.0 0.0 0.0 0.0
.2 .43 .15 .05 .01 0.0 0.0 0.0 0.0 0.0
.3 .50 .23 .10 .03 .02 .01 0.0 0.0 0.0
.4 .57 .31 .20 .09 .04 .02 .02 .01 0.0
.5 .62 .41 .25 .17 .10 .06 .04 .02 .01
.6 .68 .51 .34 .26 .20 .12 .09 .08 .04
.7 .74 .61 .46 .39 .36 .27 .25 .20 .15
.8 .78 .68 .59 .58 .53 .46 .50 .42 .40
.9 .84 .76 .71 .71 .71 .71 .74 .72 .73
1.0 .90 .82 .80 .82 .85 .86 .91 .93 .94
1.5 .97 .92 .84 .82 .89 .75 .74 .72 .70
2.0 1.01 .98 .94 .94 .95 .94 .96 .96 .96
2.5 1.04 1.01 .97 .94 .94 .93 .89 .88 .86
3.0 1.06 1.03 1.00 .98 .98 .99 .99 .98 .98
3.5 1.08 1.05 1.01 .99 .98 .99 .97 .94 .93
4.0 1.08 1.07 1.03 1.02 1.01 1.01 1.01 1.00 1.00
4.5 1.09 1.07 1.04 1.03 1.01 1.02 .99 .98 .97
5.0 1.09 1.08 1.05 1.03 1.02 1.03 1.02 1.01 1.01
The time is normalized to the ratio: Life Cycle (LC)/alpha. Components not using hours as
the independent variable (i.e., switches which use actuations) can either equate # cycles to time or
can use total number of cycles expected as LC. Life cycle in the context of this model is the design
life of the equipment in which the part is operating.
2-15
p=l
.1 .2 .3 .4 .5 .6 .7 .8 .9 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 t ( a
la
LC time
FIGURE 2.3-1:
CUMULATIVE AVERAGE FAILURE RATE AS A FUNCTION
OF LIFE CYCLE, a, AND p
For example, if the Life Cycle of a motor is .5 a and [3=3 for the motor,
2-16
If a = 100,000 hrs:
Additionally, if preventative maintenance (PM) is performed, the PM interval can be used for
LC, thus yielding the average failure rate in the PM interval.
The methodology developed herein allows a constant average failure rate to be predicted over
the life cycle (or preventative maintenance interval) if the a and p of a part are known. This allows
modeling of wearout items providing these values can be determined.
Many reliability models yield the MTTF. Since the proposed model uses the characteristic
life (a) as the variable to predict the failure rate, a must be derived from the MTTF. The ratio
MTTF
is not constant but depends on p. The following relates the p value to the percent failed at
a
the mean life (MTTF) (from Reference 51).
TABLE 2.3-3:
PERCENT FAILED AT MTTF AS A FUNCTION OF p
p Percentile
.5 75%
1.0 64
2.0 54.5
3.0 51
4.0 50
5.0 50
6 50
7 50
8 50
9 50
10 50
2-17
Using Weibull probability paper, the ratio of a/MTTF can be calculated. This data is
summarized in Table 2.3-4. For typical [3's of 2-4, this ratio is modest, on the order of 1.06 to
1.15. This indicates that there will be a negligible error if the MTTF is used instead of a. In fact,
several models to be presented later use the mean number of cycles to failure.
TABLE 2.3-4:
a/MTTF RATIO AS A FUNCTION OF [3
p a/MTTF
1 1
2 1.15
2.5 1.12
3.0 1.10
4.0 1.06
These simulation results illustrate that the failure rates associated with wearout failure
mechanisms are very close to zero, provided that the characteristic life of a given component is
much greater than the design life of the equipment in which it operates. This should occur if the
components wearout characteristics are understood and the proper design precautions have been
taken to ensure a robust design. The ultimate objective of design and reliability engineers is to
achieve a design robust enough to operate reliably in a given application for a given life cycle. This
methodology provides a tool to ensure this robustness has been achieved.
2-18
3.0 DATA COLLECTION
An aggressive data collection effort was undertaken to collect failure rate data on the part
types being modeled. The objectives of this data collection effort were as follows:
(1) To obtain data on relatively new components. Although collection of data on recently
manufactured components was given priority, the general methodology used was to
accept data of parts manufactured since 1980. (The last time most of the models were
updated was 1977).
(2) To collect as much data on all part types in as many environments and as many quality
levels as possible.
(3) To insure the data is high quality from reputable data sources.
(4) To collect data from maintenance activities which repair and report data to the piece part
level.
Collection of data from military equipments was the most important to the successful
completion of this effort. It is also, by far, the most tedious and time consuming. For these
reasons, it will be described in more detail.
Table 3.0-1 presents the military systems from which data was obtained in this effort, their
application environment, and the source of maintenance/reliability data used. The following
paragraphs provides a more detailed discussion on these data sources.
3-1
TABLE 3.0-1:
DATA SOURCES
GRC-171: This is a ground mobile, trailer mounted, communication system used in the Air
Force. This system provided IITRI with failure rate data on connectors, resistors, capacitors,
switches, relays and inductors. One reason this system was selected was to correlate failures in
ground communications equipment and airborne communications equipment.
ARN-118: This is a tactical navigation unit used in a variety of aircraft. IITRI has collected
recent information on this equipment from F-4C/D/E/G, F-15A/B/C/D, and A-10 aircraft. This
system provided IITRI with information on connectors, resistors, capacitors, switches, relays and
inductors. Failures from the F-4s, F-15s, and A-10s are based on 1635 aircraft and 582,745
flying hours. These figures are based on a 12 month period from June 1989 to May 1990. IITRI
collected all of the D056 part replacement records pertaining to this equipment on those selected
aircraft. This system was chosen due to it's versatility in use with a variety of aircraft. In addition
to D056 data, the original RIW data was also used for this system.
ARC-164: This is an airborne communication unit used in a variety of aircraft. IITRI has
collected recent information on this equipment from F-4C/D/E/G, F-15A/B/C/D, and A-10 aircraft.
This system provided IITRI with information on connectors, resistors, capacitors, switches, relays
and inductors. Failures from the F-4s, F-15s, and A-10s are based on 1635 aircraft and 582,745
flying hours. A K factor was then applied to these operating hours to account for on-hours while
the aircraft is not in flight. These figures are based on a 12 month period from June 1989 to May
3-2
1990. IITRI collected all of the D056 part replacement records pertaining to this equipment on
those selected aircraft. This system was chosen because of its use in a variety of aircraft and to
draw any correlations that can be made against ground communication equipments. In addition to
D056 data, the original RIW data from the equipment manufacturer was also used for this system.
ALQ-172: This is an airborne electronic countermeasures (ECM) pod used in the B-52 aircraft.
This system provided IITRI with information on connectors, resistors, capacitors, switches,
relays, inductors, and transformers. There were approximately 600 part failures from 80 installed
equipments with 60,288 operational hours. The failures are based on 2 years of warranty
information from ITT. This system was chosen because all of the data was reported to the USAF
through a verifiable warranty program from ITT.
Flight Control Computer: This is the main computer in the F-16. IITRI has collected recent
information on this equipment. This system provided IITRI with information on connectors,
resistors, capacitors, switches, relays, and inductors. Data collected is based on 400,048 flying
hours from 1089 aircraft. IITRI collected all of the D056 part replacement records pertaining to
this equipment on those selected aircraft. RIW data was also used for this system.
Reliability Improvement Warranty (RIW) programs typically yield very high quality piece
part data since it is generally taken by a single maintenance activity and accurately reported. Data
reported at the piece part level from maintenance systems such as D056 and MODAS is generally
suspect, but for the systems for which these sources were used, IITRI confirmed that the data was
indeed accurate, complete, and could be used to obtain the appropriate data. This assurance was
obtained by contacting the maintenance activities to verify that all maintenance actions are recorded
and reported to MODAS faithfully.
Table 3.0-2 summarizes the procedures required to obtain piece part failure rate data from
military systems and Table 3.0-3 summarizes additional data sources used. The additional sources
are primarily from manufacturers life data, published data, or data solicited during this study as a
result of a survey.
3-3
TABLE 3.0-2:
DATA SUMMARIZATION PROCEDURE
Environments/Quality
- Age
- Component Types
- Availability of Quality Data
(2) Build Parts List:
TABLE 3.0-3:
ADDITIONAL DATA SOURCES USED
3-4
3.1 DATABASE
The database used to store and manipulate the reliability data obtained in this study has been
implemented in Informix 4GL running on a MIPS 2460 platform and consists of three records
types as follows:
DEVICE
STRESS
RESULT
The device record holds component characteristic data on the specific part, the stress record is
information regarding the test (stresses, environment, duration, etc.) and the result record is
information regarding the results of the test (number tested, number failed, failure mechanism,
time/cycles to failure, etc.). The stress and result records are common to all part types but the
device record is unique to a particular class of part. The specific parameters of the device record
for the part types being addressed are given in Appendix B.
Table 3.2-1 presents a high level summary of the total part operating hours (including hours
from zero failure records) from field data and number of failures for each generic component type.
Interconnect assembly (PWB) data is not included in this table since that model is based on
temperature cycling laboratory data and not on field data.
The general approach taken in this effort was not to collect data on specific part styles and
spec, numbers, but rather to collect as much data as possible from as many different sources as
possible in the hopes that data on the predominant device types and specs, are collected.
3-5
TABLE 3.2-1:
SUMMARY OF DATA COLLECTED
3-6
TABLE 3.2-1 :
SUMMARY OF DATA COLLECTED (CONTD)
Table 3.2-2 summarizes the applicable specifications of parts for which data was collected.
Although there is some data on every specification listed, in some cases there is a limited amount
3-7
of data on parts of some specs. This is not a major obstacle to model development since the data
was pooled together with data from other parts of the same generic category. In the majority of
cases, this pooling yielded a sufficient amount of data on which to derive a model.
TABLE 3.2-2:
PART SPECIFICATIONS
3-8
TABLE 3.2-2:
PART SPECIFICATION (CONTD)
This section of the report presents the derivation of the failure rate model of each component
type. The component types for which models were developed are:
Capacitors
Resistors
Inductive Devices
Transformers
Inductors
Switches
Standard Switches
Rotary Switches
Circuit Breakers
Thermal Switches
Relays
Connectors
Connectors
Connections
Sockets
Interconnection Assemblies/
Printed Wiring Boards
Rotating Devices
For each of the above component types, this section of the report contains; a discussion of
reliability issues, failure modes and mechanisms, a review and critique of the current MIL-HDBK-
217E model, and the model derivation. The proposed MIL-HDBK-217 models are presented in
Section 5.0.
4-1
4.1 CAPACITORS
Circuit designers will typically select a capacitor based on factors such as frequency range,
volumetric efficiency, series resistance, stability, noise, voltage capability, capacitance range and
cost. Since an ideal capacitor is purely reactive with zero equivalent series resistance, there is no
power dissipation and associated temperature rise. Since all capacitors may not exhibit this ideal
characteristic, there may be some temperature rise associated with operation. Reference 52 defines
the temperature rise (AT) associated with AC power dissipation to be the following for aluminum
electrolytic capacitors;
where
The power dissipation for DC leakage is negligible. Additionally, in the majority of cases the
temperature rise from the ESD is also negligible. Therefore the capacitor operating temperature can
be considered to be the ambient temperature. In addition to temperature, the applicable stress
influencing reliability is applied voltage relative to the voltage capability of the capacitor.
The manufacturing process strongly influences the reliability of capacitors. For example,
capacitors with dielectrics deposited on the electrode, or the electrode deposited on the dielectric
typically have greater stability characteristics. The internal connections are always reliability
4-2
concerns with capacitors, particularly when exposed to high vibration environments or
environments with extreme temperature cycling. Capacitor hermeticity is also a concern if it is to
be used in an uncontrolled environment, due to the possible absorption of moisture into the
dielectric. This can cause a change in capacitance, reduction of the voltage capability or a direct
short.
The following pages summarize the various capacitor types, their reliability characteristics,
potential failure modes/mechanisms, approximate probability of occurrence if available,
accelerating stresses, whether it is a wearout or defect mechanism, potential screening stresses, and
expected screening effectiveness.
While the percentages listed are based on the best available data, it is understood that these
values can and will vary greatly as a function of the manufacturing process and the actual use
environment. Therefore, this information is only used in this study to identify predominant failure
mechanisms that must be accounted for in the model and their relative rate of occurrence.
Capacitor. Variable
Variations:
Dielectric: Ceramic
Air
4-3
Unique Characteristics:
• Many failures are of a mechanical nature due to the more complex mechanical
configuration relative to fixed capacitors.
TABLE 4.1-1:
VARIABLE CAPACITOR FAILURE MODES
Variations: Polarized
Non-Polarized
Unique Characteristics:
-B
W - AeT
w= Weight Loss
A - Magnitude Constant
B = Constant
T = Temperature
4-4
• Shorts can result due to dissolving of the electrolyte in a storage environment or
in a lightly stressed use environment (per MIL-STD-1131). Current processing
techniques have significantly reduced the probability of occurrence of this failure
mechanism.
TABLE 4.1-2:
AL ELECTROLYTIC FAILURE MODES
Variations:
Seal: Hermetic
Non Hermetic
Unique Characteristics:
4-5
• Loss of Electrolyte is the predominant wearout mechanism.
TABLE 4.1-3:
TANTALUM WET SLUG FAILURE MODES
Unique Characteristics:
• Solid tantalum capacitors have a unique current related failure mechanism that is
highly dependent on series resistance used in the circuit. This is due to intrinsic
faults in the oxide that continuously heal themselves upon application of current.
However, some faults are too large to heal themselves and can result in a thermal
4-6
runaway condition if sufficient current limiting series resistance is not present.
Current processing techniques have significantly reduced the probability of
occurrence from this mechanism.
TABLE 4.1-4:
SOLID TANTALUM FAILURE MODES
Variations:
Hermeticity: Hermetic
Non Hermetic
Polarization: Polarized
Non Polarized
Unique Characteristics:
4-7
TABLE 4.1-5:
TANTALUM FAILURE MODES
Variations:
Dielectric: Glass
Mica
4-8
TABLE 4.1-6:
MICA AND GLASS FAILURE MODES
%
Failure Occurrence Accelerating Wearout Screening
Mode/Mech (*) Stress(es) or Defect Screen Effectiveness
Variations:
Dielectric: Barium titenate
Calcium titenate
Stroutium titenate
Lead niobate
Form: Tubular
Feed through
Disks
Monolithic Multi-layer
4-9
TABLE 4.1-7:
CERAMIC FAILURE MODES
Variations:
Dielectric: Paper-Foil
Metallized Paper
Mylar Foil
Metallized Mylar
Polystyrene
Teflon
Polycarbonate
4-10
TABLE 4.1-8:
PLASTIC AND PAPER FAILURE MODES
The following items summarize the findings after reviewing the current MIL-HDBK-217E
capacitor models. These items were then addressed more specifically in the model development
phase of this effort. It should also be noted that only those items determined to be feasible are
explicitly included in the models developed.
(1) The base failure rate expression is complex and statistically unjustified. It includes
provisions to make the predicted failure rate extremely high for stresses close to or over
the rated stress. It also makes the predicted failure rate very low at stresses below the
rated value. While it may be applicable for voltage stress, it does not follow the well
accepted Arrhenius relationship for temperature acceleration.
(2) The package type is only used in the case of tantalum capacitors. It may be desirable to
include package type directly in the failure rate model for other types of capacitors.
(3) The time dependent properties of capacitor failures are not addressed. If wearout
mechanisms are predominant for a particular capacitor type, then the data collected in
the early life of that part is not representative of the reliability in the later portion of the
parts life. An example of this is dielectric breakdown, which typically will exhibit a
decreasing failure rate in early life. On the other hand some electrolytic types will
predominantly fail in a wearout manner, especially if not under a sufficient voltage
stress.
4-11
(4) Chip and surface mount capacitors, such as CDR (MIL-C-55681), CWR (MIL-C-
55365), CRL (MIL-C-83500) types are not adequately addressed.
(5) Some capacitor specifications have been canceled or classified as inactive for new
designs, such as MIL-C-14157, 18312, 11272, 3965, and 92.
(6) There are several base failure rate tables presented for each capacitor type as a function
of rated temperature. Typically the differences in the predicted failure rate between
capacitors of different rated temperatures is insignificant relative to prediction model
accuracy.
^p = ^ E ^ T ^ V ^ V R ^ C nSR+ X
E (0
TCQ = Quality Factor, function of screens and of the control the manufacturer has on
the manufacturing process (QPL status)
-Ea ,\ 1 o
K | Tj 298
Ea = Activation energy
4-12
TtV Voltage stress factor
VR = Rated voltage
71 =
VR Rated voltage factor
The premise of including a rated voltage factor in the theoretical model is that the thicker
dielectrics of higher voltage capacitors are easier to make defect free than the thinner
dielectrics of low-voltage capacitors. Since failures are usually precipitated at a defect site,
the probability of failure is proportional to the inverse of dielectric thickness.
Using a derivation methodology similar to that used to model the reliability of oxides in
integrated circuits, it can be shown that the defect density (D) is inversely proportional to the
square of the dielectric thickness (X) (Ref. 35):
D
X2
From extreme value statistics (Ref. 35), it can be shown that the defect density is directly
proportional to the failure rate (X):
D»cl
Since the rated voltage of a capacitor is directly proportional to its dielectric thickness (X
VR):
X °c D « —~ « — ~
Xz VR
4-13
X
VR 2
Since only a percentage of all failures are precipitated by defects, the above relationship
must be scaled accordingly. A and B are constants dependent on the percentage of failures
that are defect related.
X B
A +
VR 2
Whether or not the factor is important depends on the defect density for capacitors as a
function of dielectric thickness. It may be true that the dielectric thickness of capacitors are
large enough so that the premise of this model (D °= ^—) is not valid. As with the other
^C Capacitance factor
AJ+BJC
where A J , B J = Constants
C Capacitance
The rationale for this factor is that physics dictates that the probability of failure due to a
defect is directly proportional to the dielectric area and hence capacitance. Proportionality
constants A^ and B^ will compensate for the percentage of failure modes susceptible to
dielectric defects.
*SR = Series resistance factor, applicable to solid tantalum electrolytic capacitors only.
^g(t) = Failure rate of certain types of electrolytics due to the wearout mechanism
of electrolyte loss.
4-14
4.1.3.2 Summary of Capacitor Data Analysis
Initial analysis of the capacitor failure rate data consisted of analysis of variance and
correlations coefficient of the following variables:
The correlation coefficients indicated that there were several highly correlated variables,
making it difficult to devise certain factors. The most significant of these was, as expected, the
correlation between quality and environment. To alleviate this, the Quality Factors in Table 4.1-9
from MIL-HDBK-217E were assumed to be correct. This relative ranking of quality factors is also
consistent with the MIL-SPEC requirements.
TABLE 4.1-9:
CAPACITOR QUALITY FACTOR
Quality *Q
D .001
C .01
S, B .03
R .1
P .3
M 1
L 3
NonER 3
Lower 10
4-15
Although quality was correlated to environment, to the extent possible the initial regression
results suggested the above relative factors were consistent with the collected data. These factors
were then used in the regression so that valid environment factors could be derived. It was also
determined that the above quality factors should be used for all capacitor types and not a function
of capacitor type.
Additionally, certain factors considered necessary for inclusion into the model could not be
quantified from the field data collected due to lack of details available in the data. These factors
were voltage stress and temperature. As an alternative to field data analysis, these factors were
derived from life test data, published information, or current MIL-HDBK-217 factors.
To address the temperature factor, the literature was reviewed to determine the applicable
form for a temperature acceleration factor and to determine the applicable constants in that factor.
The following lists information regarding the Arrhenius activation energies found in the literature.
Included are the capacitor type, equivalent Arrhenius activation energy, the model cited (Arrhenius
or other) and the reference from which the information was extracted.
4-16
From this information, it can be seen that the Arrhenius model is the most predominant model
used in the capacitor industry to model temperature acceleration rates. The activation energies cited
are much higher than the current values in MIL-HDBK-217E. This could possibly be due to the
fact that the values were derived primarily from accelerated life test results (temperature and/or
voltage acceleration) which may inherently accelerate the temperature related failure mechanisms
more than the other non-temperature related mechanisms that would be experienced in the field.
The conclusion of this analysis is that reliability is a strong function of temperature and that
temperature must be accounted for in the reliability model. Therefore, since the temperature
acceleration rates would be enormous if the activation energies derived from the high temperature
life tests were used, and since the current MIL-HDBK-217 acceleration rates are reasonable for
field use conditions, factors consistent with the current models will be kept.
Although the current models are not based on the Arrhenius relationship, an equivalent
activation energy was calculated and used in the temperature factor. The activation energy for each
capacitor type was first calculated using the current 217 models. To accomplish this, the
equivalent activation energy was derived by calculating the acceleration due to temperature between
0°C and the maximum rated operating temperature for each specific capacitor type. The general
assumption on which the temperature factor is based is that the activation energy is solely a
function of dielectric material. These activation energies are given in Table 4.1-10:
TABLE 4.1-10:
CAPACITOR ACTIVATION ENERGIES
Dielectric Material E
a
4-17
The temperature acceleration factors were then calculated for each data record by using the
Arrhenius equation with the activation energies in Table 4.1-10 and the ambient temperatures in
Table 4.1-11. The default temperature in Table 4.1-11 were taken from MIL-HDBK-217E. The
failure rate was then compensated (divided) by the temperature acceleration factor and the
regressions were run.
TABLE 4.1-11:
OPERATING TEMPERATURES
The initial regressions used both capacitance and rated voltage as variables. The
hypothesized model was that the failure rate should be proportional to capacitance and voltage in
the following relationship;
k A + By
4-18
Since the rated voltage and capacitance were highly correlated in the dataset used, the effects
of both could not simultaneously be quantified. Given this situation, the fact that physics dictates
that capacitance should be a more dominant reliability driver of capacitors, and the fact that
capacitance was a significant factor in the initial regression analysis, voltage was discarded as a
model variable and capacitance was analyzed separately. It should be noted however that while the
rated voltage was discarded as a variable, the voltage stress ratio (actual/rated) is considered
essential to the model and will be discussed further later in this section.
The capacitance factor was calculated in a separate regression and was significantly different
between electrolytic and nonelectrolytic capacitor types. These KC factors were determined to be:
Electrolytic: ^a O^
All others: A. a O 0 9
A separate regression was performed for Electrolytics and Nonelectrolytics due to the unique
physics of failure of each. Once the above relationships were established, the regression was
performed again by normalizing the failure rate to these relationships (i.e., dividing the observed
failure rate by these factors). It is necessary to perform these regressions again since continuous
variables such as capacitance have a different model form relative to discrete variables and must be
analyzed separately.
As expected, environment was a significant variable. The factors derived for the
environments for which there existed data are summarized in Table 4.1-12. The environment
G g £ , although not defined in MIL-HDBK-217, is used here to denote commercial quality
components operating in a ground benign environment. A u refers to the uninhabited portion of an
aircraft, although the specific type of aircraft was not known. All other environments are defined
in MIL-HDBK-217E.
4-19
TABLE 4.1-12:
OBSERVED ENVIRONMENT FACTORS
Environment K
E
G
BC 1
A
UA 202
A
UF 530
A
IC 1540
A
U 15
GF 69
AIF 3400
This data suggests that the current environment factor is not stringent enough. However,
after reviewing the models developed with an extreme value analysis, it was concluded that the
resultant failure rates were unrealistically high, indicating that the results were an aberration of the
statistical modeling process. It is however clear that the current environment factor should be
increased to reflect the large observed dependence of environment on failure rate. This was
accomplished by using the relative rankings of the MIL-HDBK-217E models, calculating a
weighted average of the factor (for A J J ^ , Ayp, and AJJ) and recalculating the factor based on this
ratio. Section 4.2.3.2 presents a more detailed description of a similar process that was used for
resistors. The modified factors are presented in the model summary section of this report.
Variable capacitors were analyzed relative to fixed capacitors and the relative failure rate was
determined to be 8.03 times higher for variable. Therefore, the correction factor for capacitor type
is given in Table 4.1-13. This factor is not explicitly included in the model but rather is inherent in
the base failure rates. Although it may appear to be intuitive to have a separate set of environment
factors for fixed and variable capacitors, there was not enough data on variable types to justify a
separate factor. Therefore, the environment factor, while derived predominantly from fixed
capacitors, is also used for variable types.
4-20
TABLE 4.1-13:
FIXED VS. VARIABLE FACTOR
Fixed 1
Variable 8
Although not explicitly presented in the model, analysis of operating vs. nonoperating data
yielded an average nonoperating factor of .009 over all capacitor types, indicating that capacitors
on the average have a 110 times lower failure rate in a nonoperating environment. However, the
model is normalized to the operating environment.
The dielectric type factor was the last factor to be quantified and was determined to be the
following:
TABLE 4.1-14:
DIELECTRIC FACTOR
% of Hours
from Records
Dielectric Multiplying Factor with Failures
The right column of the above table presents percentage of hours associated with failure
records, per the discussion in Section 2.0. The base failure rate from the regression analysis was
determined to be .00637 F/10" and therefore multiplying this by the above dielectric multiplying
4-21
factors and the percent of hours corresponding to failure records yields the following base failure
rates:
TABLE 4.1-15:
BASE FAILURE RATE
Dielectric Xh F/10 6
Paper .00074
Ta Elec. .00081
Al Elec. .00024
Plastic .00102
Mica .00152
Ceramic .00198
Air .0000018
An important part type studied was chip capacitors, both tantalum and ceramic. A failure
rate for these could not be modeled with field data like the other capacitor styles since there were no
observed failures for these types. This indicates that they are either highly reliable, that there were
not enough hours observed, or both. There were, for ceramic chip capacitors, a total of 17.1 x
10" observed part hours in air inhabited cargo and attack environments. Using the application
environment factors derived from the data to multiply the observed part hours, it indicates that the
equivalent number of part hours was as high as 256 x 106 with no failures. This indicates a failure
rate less than .0039 is appropriate. The available life test data for ceramic chip capacitors
(Reference 15) indicated that an average failure rate, after accounting for voltage and temperature,
is approximately .0034 F/10". This agrees well with the worst case value of .0039 derived from
field data. Therefore, .0039 will be the base failure rate for ceramic chip capacitors.
The best available life data for Solid Tantalum chip capacitors is from Reference 18 and is
summarized in Table 4.1-16.
4-22
TABLE 4.1-16:
SOLID TANTALUM LIFE DATA
Solid Ta
(3.3 mF, 85°C 50 Volts 9,000 hrs 18
20 V)
Although test conditions were at a highly accelerated voltage and temperature, calculating a
base failure rate after accounting for these variables yields a value of .00010. This value was
derived by dividing the observed failure rate of 2000 (18/.009 x 10") by the acceleration due to
voltage and temperature. The commonly accepted form for the voltage acceleration factor is:
Xa(^
where
V = operating voltage
VR = rated voltage
n = constant
_ .19 / 1 1
KT = 3.4
~ 8.617 x 10-5 1,85 + 273 " 298
4-23
Therefore the base failure rate for tantalum chip capacitors is:
^h = ^ = -0001 F/10 6
D 6
(5.8 x 10 )(3.4)
To derive a voltage acceleration factor for capacitors, the relationship given above is used.
Table 4.1-17 summarizes the values of n reported in the literature for various capacitor types.
TABLE 4.1-17:
VALUES OF n FOR VARIOUS CAPACITOR TYPES
Tantalum 17 1
Solid Tantalum 23 83
Tantalum Chip 18 8
Mica 10-12 6
Multilayer Ceramic Chip 2.7 14
Multilayer Ceramic Chip 3 37
Polystyrene 6 68
Multilayer Ceramic Chip 2-4 70
Multilayer Ceramic Chip 2.04 71
Ceramic 3.1-3.6 72
Paper/Paper Film 4.5 73
Aluminum Electrolytics 5 52
Solid Tantalums 17 75
To implement a voltage stress factor for capacitors, there must be a normalization factor on
which to base the equation. This factor is normalized to the Level II derating guidelines in
Reference 76. These derating values are:
V
y— - .6 for all fixed capacitors
V
y— = .5 for all variable capacitors
4-24
If the actual applied voltage was not known it was assumed they were derated in accordance
with the above criteria. Since the data was derived from the field, the vast majority of data records
did not have known voltages and therefore the derating criteria was assumed for most data.
At the Level II derating voltage, the voltage factor must equal 1. For fixed capacitors, this
factor is:
7tv = v
v - XV R
V
"v - xv R
The proposed values of n are summarized in Table 4.1-18.
TABLE 4.1-18:
PROPOSED n VALUE
Capacitor Type/Dielectric n
Paper 4.5
Tantalum 17
Aluminum Electrolytic 5
Plastic/Polystyrene 6
Mica 10
Ceramic 3
A boundary condition necessary in this model is to not have the failure rate approach zero as
the voltage approaches zero. Therefore, voltage acceleration factor must take the form:
V
7tV +1
6 VR
4-25
Since S = y^, 7Cy = f-gj + 1
The base failure rate must be compensated accordingly by dividing by 2 since 7ty = 2 at the
nominal voltage stress condition. Therefore, the base failure rates in the final models in Section
5.0 are half that of those in Table 4.1-15.
Tantalum electrolytic capacitors are known to exhibit a unique failure mechanism which is a
function of the available current. The model for tantalums must therefore include provisions for
this failure mechanism. Several references (Reference 4, 5) have suggested that the lowest circuit
impedance above which the failure rate does not worsen should be lower than the MIL-HDBK-
217E value of 3Q./V due to improved manufacturing processes relative to those of the time the
current factor was derived. It has since been changed to 1H/V. Moynihan has suggested that the
correction factor should be a function of circuit resistance (fW) and temperature as illustrated in
Figure 4.1-1 from Reference 5. While this relationship suggests the use of a more modest function
of circuit resistance, and also suggests that its value is a function of temperature, there is no
quantitative data presented in Ref. 5 to define the value above which the failure rate does not
worsen. Therefore, since there is no data available to support changing the current value, the
factor will be left intact without change.
Many references on capacitor reliability report that wearout characteristics are prevalent under
highly accelerated stress conditions. Several also report that under these conditions, infant
mortality failures are observed which exhibit Weibull P's < 1 (Reference 8). Infant mortality
failures are generally indicative of defect related failure mechanisms which normally affect only a
small percentage of a part population. If wearout type mechanisms were prevalent for capacitors
used in fielded systems, the observed failure rate would be much higher than it is since wearout
mechanisms generally affect a large portion of the population. Normal use conditions are typically
much less severe (thus dramatically increasing wearout times) than the highly accelerated
conditions for which wearout mechanisms are observed. This, coupled with the fact that the
observed data for capacitors generally implies high levels of reliability and very small cumulative
percent failure, indicates that failures observed in the field are primarily random defect related and
not wearout. This also implies that a wearout term is not applicable for capacitors.
4-26
100
80 All case sizes and voltage
ratings are represented
60 125°C
40
30
^ 20
o
O
CO
«•-
J 10
8
fi 85 °C
O 6
4
3 25 °C
Bias: 3fi/v
0.1 0.2 0.3 0.4 0.6 1.0
Circuit resistance, fi/v
4-27
4.2 RESISTORS
Resistors can be grouped into three primary types; composition, film and wirewound.
Composition types, usually made from a carbon composition material, are widely used due to their
availability in a wide range of values and power ratings, along with their low cost. They consist of
a solid resistive element encased in a molded body with leads imbedded into the ends of the
resistive element.
Film resistors can be manufactured using thick or thin film technology. Thin film resistors
are usually made by vacuum depositing a film on a ceramic substrate. Various film materials are
used including tin, metal glaze (powdered glass, palladium and silver), cermet (precious metals and
a binder material), and carbon.
Wirewound resistors are made by winding a special alloy resistive wire around an insulating
core. Since the resistance can be tightly controlled by carefully controlling the length of wire used,
very high precision values can be obtained. They are also available in high power values. Because
they are made by winding wire around a core, they are inherently inductive and thus their
properties deviate from a pure resistance at high frequencies.
Variable resistors are made from a resistive element which is contacted by a wiper arm
thereby varying the resistance between one end of the element and the wiper. They are made from
a variety of materials similar to those used in fixed resistors and are available in a wide range of
power ratings, ranging from small PC mountable trimmer potentiometers to high power
wirewound rheostats.
Resistors generally are highly reliable if properly designed and applied into a circuit. The
power is the variable that is derated during the part derating exercises, and also is the one that
heavily influences reliability. Some resistors are also very intolerant to over-voltage or over-
current conditions, even for brief periods of time. In fact some film resistors are highly susceptible
to high amplitude, short duration pulses such as ESD and EMP, especially the high resistance, low
power type of resistor. Some other types such as carbon compositions, are not susceptible to these
conditions. Some resistor types also exhibit change in resistance when simultaneously exposed to
long periods of temperature and humidity, and of course this susceptibility is a strong function of
the packaging of the resistor.
4-28
For most resistor types, the predominant failure mode is change in resistance, although
shorts and opens also occur. Typically the resistance will change (and the resistor will eventually
fail) as a function of temperature, electrical stress and humidity. For the resistance to change, there
is generally a migration of the resistive material or a change in the physical composition of the
resistive element under the applied stresses.
Since the reliability of resistors is very high, life testing that has historically been performed
on electronic components is generally not applicable. Instead, tests used for resistors are a
resistance value test and possibly a temperature humidity test.
Due to their wide spread use and inadequate failure rate models, special attention has been
given in this effort to resistor networks. After studying the reliability issues of these networks, the
following conclusions were drawn.
• Essentially the same materials have been used over the last 15 years, and resistor
networks are generally a mature technology, although there are still considerable
variations in the quality of materials. SPC programs implemented by manufacturers
have proven to be very successful in assuring reliability and quality.
• TCR (Temperature Coefficient of Resistance) is very important and can vary widely
depending on the mix in the resistance material (i.e., one mix is good at the upper end of
the temperature range and another mix may be good at the lower range). This makes it
difficult to find a mix good for entire range of temperatures, and illustrates the fact that
there can be large variations in the reliability properties as a function of manufacturer and
within a manufacturer.
• The major change in resistance occurs in the first 100 hours, and then levels off.
4-29
• Low value resistors (i.e., <100Kf2) are susceptible to current related failure mechanisms
and high (>100KD) value resistors are susceptible to voltage (overstress conditions).
• Primary failure mode is drift and if enough power is applied, open (almost never short).
• There is typically more variability in axial leaded devices due to the fact that the screening
process for networks yields a high degree of repeatability.
• There is a strong correlation in the value of resistance (in relation to the population) and
its reliability. Therefore, variability reduction (SPC) is key in the delivery of reliable
products.
This section summarizes, for each generic resistor type, predominant failure mechanisms,
accelerating stresses, and approximate percentage of occurrence.
4-30
Resistors. Composition
Variations: None
Unique Characteristics:
(1) Moisture intrusion can cause shifts in resistance values, especially if in an uncontrolled
storage condition or with < 10% power applied.
TABLE 4.2-1:
COMPOSITION RESISTOR FAILURE MECHANISMS
4-31
Resistor. Fixed Film
Hermeticity
Unique Characteristics:
TABLE 4.2-2:
FILM RESISTOR FAILURE MECHANISMS
4-32
Resistor. Wirewound
Unique Characteristics:
TABLE 4.2-3:
WIREWOUND RESISTOR FAILURE MECHANISMS
4-33
Resistor. Variable Non-wirewound
Unique Characteristics:
(1) Many failures are due to the mechanical elements of the resistor.
(2) Corrosion, oxidation, and wear of the contact are reliability concerns.
TABLE 4.2-4:
VARIABLE COMPOSITION RESISTOR FAILURE MECHANISMS
4-34
Resistor. Variable Wirewound
Unique Characteristics:
TABLE 4.2-5:
VARIABLE WIREWOUND RESISTOR FAILURE MECHANISMS
4-35
Resistor. Networks
Resistor. Thermistor
Unique Characteristics:
(1) Prone to thermal runaway conditions (with negative temp, coefficient devices).
TABLE 4.2-6:
THERMISTOR FAILURE MECHANISMS
4-36
4.2.2 Current MIL-HDBK-217E Resistor Model Review
The following items summarize the findings after review of the current MIL-HDBK-217E
resistor models.
(1) The base failure rate equations are complex, not statistically justified, and include provisions
for the failure rate to increase dramatically for stresses close to the rated stress. Base failure
rates are very low for stresses well below the rated maximum.
(2) Some of the base failure rate tables indicate indistinguishable differences. For example, the
differences between MIL-R-22684 and MIL-R-39017, and between MiL-R-55182 and MIL-
R-10509 indicate an approximate 2% difference in failure rate.
(3) The format for the resistor network model is not consistent with the others.
(4) The resistance range factor for power wirewound resistors has a range of 1 to 1.6, which is
insignificant relative to the expected model precision.
(5) There is no adequate means to calculate the failure rate of non-plated through hole technology
parts, such as surface mount and chip devices.
(6) The complexity of the thermistor model is not consistent with the other models.
(7) A primary failure mechanism of variable resistors is corrosion of the wiper contact which
results in an intermittent or open condition.
(8) The range of the voltage factor for variable resistors is 1 to 1.2, which is insignificant relative
to the expected precision of the model.
(9) The resistor network model indicates that there is a linear relationship between failure rate and
the number of resistive elements. This seems illogical because the resistor network failure
rate contribution of the package is not expected to be proportional to the number of resistive
elements.
4-37
4.2.3 Resistor Model Development
=
^p ^Q^T^PR^TO
=
^PR R ate d Power Factor
TC =
TO Tolerance Factor
Initial analysis of the resistor failure rate data indicated that there was a high correlation
between environment and quality, which was expected. Due to this correlation the initial
regression analysis with unmodified quality and environments yielded inconsistent and intuitively
incorrect results. For this reason, either a quality or environment factor had to be derived off-line
and the regression re-run with the new factor. It was determined that the factor for quality would
be derived since there are standard procedures for quantifying the reliability differences between
various quality levels of military part types. The vast majority of the data was either of commercial
quality level or of the standard MIL quality level M, between which the current MIL-HDBK-217
models indicate there is an approximately 10:1 difference in failure rates. This is the quality factor
therefore that will be used. The observed failure rates were modified in accordance with this
quality factor and the regression was re-run.
4-38
The following environment factors in Table 4.2-7 were derived from the regression:
TABLE 4.2-7:
OBSERVED RESISTOR ENVIRONMENT FACTORS
A 25.0
u
A
I 43.4
G 1.0
Although these are more generic environments than those currently in the handbook, it is the
most detailed level at which the regression analysis indicated statistically significant results. These
results indicated that, in general, the airborne applications were more severe than current models.
An average of the current 217E environment factors for all resistors are given in Table 4.2-8.
TABLE 4.2-8:
CURRENT ENVIRONMENT FACTORS
GB 1
GF 2.0
G
M 8.28
A
IC 9.0
A
UC 15.3
Ajp 11.6
A
UF 21.9
A
RW 31.3
N
U 20.8
N
S 6.14
ML 43.7
MF 18.7
cL 868
4-39
Average of the generic categories of current factors and factors derived in this effort are given
in Table 4.2-9.
TABLE 4.2-9:
217E/DERIVED ENVIRONMENT COMPARISON
A
U 18.6 25
A
I 14.8 43
GB 1 1
These values indicate that on average, the current models are 2.03 times optimistic ((25 +
43)/(18.6 + 14.8)). Thus, adjusting the current models in accordance with these factors and
adjusting all other environment categories proportionally yields the factors in Table 4.2-10.
TABLE 4.2-10:
RESISTOR ENVIRONMENT FACTORS
GB 1
GF 4
G
M 16
A
IC 18
A 31
uc
Arp 23
A
UF 43
A
RW 63
Nu 42
NS 12
ML 87
MF 37
cL 1728
SF .5
4-40
Also analyzed as a variable was whether the resistor was of a fixed or variable type. Variable
types are mechanically more complex and therefore typically exhibit higher failure rates than their
fixed counterparts. Additionally, they have a unique failure mechanism that fixed resistors do not;
corrosion and contamination of the contact. These mechanisms are temperature dependent and thus
should follow the Arrhenius relationship with a specific activation energy. Initial regressions did
not account for temperature and indicated that variable resistors exhibited a 2.5 times higher failure
rate than fixed. Temperature was then accounted for by calculating an activation energy using the
current models and default temperatures for each environment, and using the Arrhenius
relationship to determine the temperature acceleration factor. This was accomplished only for
variable resistors. Since adequate data was not available describing the temperature dependence of
fixed resistors, it will not be a factor for fixed types. The activation energies used are listed in
Table 4.2-11.
TABLE 4.2-11:
RESISTOR ACTIVATION ENERGIES
Composition .31
Non-wire Wound .09
Film .22
Cermet or Carbon Film .09
Wire wound .23
After modifying the observed failure rate for temperature, the regression was re-run and it
indicated that there was no statistical difference between fixed and variable. Therefore, since
temperature is only used as a factor for variable types, it is the only factor distinguishing fixed and
variable resistors.
M
k R (R in ohms)
4-41
Due to its relative insignificance, it will not be included in the model. Power was however a
significant variable and was determined to be:
The observed failure rate was then divided by this factor for each data record and the
regression was re-run. The base failure rates in the right column of Table 4.2-12 were derived by
multiplying the base failure rate from the regression analysis by the percentage of hours from
failure records.
Another variable analyzed separately was the number of sections in resistor networks.
Inconsistent results and a low correlation to failure rate indicate that, based on the data, the number
of sections does not influence of resistor network reliability.
TABLE 4.2-12:
RESISTOR BASE FAILURE RATES
*The values for resistor networks were derived separately from the regression analysis using
the data in Table 4.2-13.
4-42
TABLE 4.2-13:
RESISTOR NETWORK DATA
r. Failures/hours
A
b ~ K Tip 71 Q
E
Resistor networks were analyzed separately since there were only a few datapoints available
which resulted in anomalous results from the regression analysis. Table 4.2-13 summarizes this
analysis. The observed failure rate was divided by the appropriate values of the environment,
power rating, and quality factors. The geometric mean of these values were then taken which
yielded the resistor network base failure rate of .0019.
4-43
4.3 INDUCTIVE DEVICES
General classes of inductive devices are coils and transformers. Coils are reactive devices
made by winding an insulated wire around a ferrous or non-ferrous core. Use of a ferrous core
dramatically increases the inductance. Transformers are basically two coils wound on a common
iron core which closes the magnetic circuit and allows the conversion of voltage (up or down)
when an alternating current is applied to one of the coils.
Inductive devices are relatively simple and have proven to be reliable if used properly. The
failure modes/mechanisms that occur are insulation breakdown, open circuit, and a change in the
magnetic core characteristics (if applicable). The occurrence of open circuits is application
sensitive and can result from extreme current or mechanical damage. Changes in core
characteristics can result from exposure to extreme temperatures. However, the predominant
failure mode is insulation breakdown between windings for heavy wire windings and open circuit
for fine wire windings.
Inductive device design dictates the rate of insulation breakdown. This mechanism depends
on the type of insulation (type of material, thickness and purity) and is accelerated by temperature,
current and humidity.
Since ideal inductors are also a purely reactive device, they dissipate very little power under
operating conditions. However, since there is a resistance associated with the wire, there will be
some dissipation that must be accounted for when calculating its operating temperature. Therefore,
the hot spot temperature should be calculated and used in the failure rate equation. The
methodology for calculating this temperature will not be changed from the current 217E
methodology.
Tests used for inductors include (if applicable) current, winding-to-winding breakdown,
winding-to-core breakdown, and winding-to-case breakdown, all with or without accelerating
temperature and humidity.
Tables 4.3-1 through 4.3-3 summarize the predominant failure modes/mechanisms along
with their accelerating stresses and approximate percentage of occurrence.
4-44
TABLE 4.3-1:
INDUCTOR FAILURE MODES AND MECHANISMS*
TABLE 4.3-2:
TRANSFORMER FAILURE MECHANISMS
TABLE 4.3-3:
RF COIL FAILURE MECHANISM DISTRIBUTION
4-45
4.3.2 Current MIL-HDBK-217E Inductive Devices Model Review
The MIL-HDBK-217E inductive device models have been analyzed to determine areas of
deficiency and possible areas of improvement. The following summarizes these findings:
(1) The base failure rate equation does not appear to be based on the Arrhenius relationship.
(2) The construction type given in the models may have insufficient detail. Some inductive
devices have a more complex mechanical construction and thus are more susceptible to failure
when exposed to environments with high levels of shock and vibration.
(4) There is a non-linear relationship between failure rate and temperature rating.
(5) The weight vs. temp, rise needs to account for transformers less than one pound.
KQ - Quality Factor
4-46
4.3.3.2 Summary of Inductive Device Data Analysis
4.3.3.2.1 Transformers
Since both quality and environment are important reliability factors and should be included in
the model, one must be derived off-line and used in subsequent regressions to quantify the other.
It was therefore chosen to derive the quality factor using the current MIL-HDBK-217E KQ values
as a baseline. The current MIL-HDBK-217E KQ factors range from 2.5 to 3.75 and since this
range is insignificant relative to the precision of the prediction model, an average of 3:1 will be
used for the ratio of commercial to military quality.
An equivalent activation energy in the Arrhenius equation could not be derived since precise
temperatures for each observed failure rate was not known. The current MIL-HDBK-217E model
for transformers does not use an Arrhenius relationship, but rather the following equation:
T
G
HS + 273
"
Xu = A e NT
This relationship was apparently structured to allow the failure rate to increase dramatically as
the maximum rated temperature (Ny) is exceeded. Calculating an equivalent activation energy
between 0°C and the rated temperature of the transformer yields an average value of .11 eV. This
was derived by calculating the base failure rate at 0°C and at the maximum rated temperature, and
calculating the activation energy necessary to derive the same ratio of failure rate between these two
temperature extremes when using the Arrhenius relationship. .1 leV is a relatively low activation
energy, but typical of dielectric breakdown mechanisms.
4-47
Therefore, the temperature acceleration factor was calculated using a .11 activation energy
(and normalized to 25°C, as are the current MIL-HDBK-217 models) and default temperatures for
each environment defined in Reference 64.
The observed failure rates were then adjusted for quality and temperature and the regression
was re-run to quantify, if possible, the effects of:
Environment
Transformer Type
- Frequency of Operation
Secondary Voltage and Current
TABLE 4.3-4:
OBSERVED ENVIRONMENT FACTORS
Env. *E
G
BC 1
AU 5.27
A
UA 11.2
Gp and Gjyj environments were observed not to be significantly different from GgQ. These
environment factors are not significantly different from the existing MIL-HDBK-217E factors and
therefore the environment factor will remain unchanged.
The base failure rates in the right column of Table 4.3-5 were obtained by multiplying the
base failure rate (k^ regression) by the % hours from failure records.
4-48
TABLE 4.3-5:
TRANSFORMER BASE FAILURE RATES
% Hours from
Type X^ Regression Failure Records ^b
Also analyzed were the effects of operating frequency and secondary voltage or current.
Although no discernable affects due to these variables could be identified while analyzed as a
continuous variable, operating frequency is partially accounted for in the base failure rate since
there are separate failure rates for power, audio, and RF types.
4.3.3.2.2 Inductors
The initial analysis of the data indicated several correlations; between quality and
environment, inductor type and environment, and RF types and choke types.
Due to the correlation between quality and environment, the current quality factor ratio of
20:1 was chosen for commercial quality and military quality M. The observed failure rates were
then normalized to this factor and the regression was re-run.
4-49
The observed environment factors are given in Table 4.3-6 as follows:
TABLE 4.3-6:
OBSERVED INDUCTOR ENVIRONMENT FACTORS
GB 1
GF 55
G
M 23
Al 175
Au 225
From this data, it appears as though the current environment are not severe enough.
However, since the results were more significant for transformers, and coils have similar reliability
characteristics, the environment factor for transformers will be used.
There also was not enough data to quantify the difference between fixed and variable types.
Therefore, the current 1:2 ratio will be kept. The data set was modified (divided by) for this factor
and the regression was re-run.
The base failure rates obtained from the regression analysis are given in Table 4.3-7:
TABLE 4.3-7:
INDUCTOR BASE FAILURE RATES
% Hours from
Type X\j Regression Failure Records ^b
4-50
4.4 SWITCHES
Switches electrically transfer power or function from one circuit to another, resulting in the
completion of the circuit. The actuation is manually applied, differentiating them from relays. The
achievement of the transfer function is accomplished in two basic methods:
• Snap Action
• Wiping
• Cross Bar
Each type of contact style and actuation is configured in relation to their application. Lamp or
inductive loads require snap action configurations to reduce the contact degrading during arcing.
Dry circuit applications require the cross bar configuration to eliminate corrosion build up creating
a resistive connection.
Electronic circuit transfer devices do not employ contacts to perform its function but instead
transfer power through transistor like saturation of a semiconductor layer. Mainly utilized in low
power applications, their unique clean transfer and isolation properties make them popular in
microwave electronic circuits. With the exception of Solid State Relays, these electronic switches
are not addressed in this study.
The majority of switch failure modes and mechanisms relate to the contacts. Under ideal
conditions the resistance at the contact interface is zero but in reality resistance is present. Design
and application factors which influence contact failure are:
(1) Contacting Materials - different materials exhibit varying degrees of resistance to oxidation.
An oxide film causes increased contact resistance and heat. Table 4.4-1 shows various
physical properties for different contact materials.
4-51
TABLE 4.4-1:
CONTACT MATERIAL PROPERTIES IMPACT SWITCH RELIABILITY
Temp. Thermal
Melting Coefficient of Conductivity
Contact Point Resistivity Resistivity (cal-cm Oxidation Arcing
2
Material (°C) (uQ-cm) (per°C) scc-°C-cm ) Resistance Effects
Platinum 1,773 10.60 0.0030 1.17 Very Good Resists arc erosion
(2) Operating Environment - the presence of foreign particles in the environment and the
formation of surface film increases the contact resistance and adversely affects the failure rate
of the contacts.
(3) Contact Pressure - the higher the contact pressure the greater the contact area due to the
yielding of contact asperities (microscopic peaks and valleys). It also can degrade the contact
faster due to wear.
The predominant accelerating stresses in manually actuated switches is temperature and load
during switching. Temperature is generated by the natural transfer of power, occurring during the
mating of contacts. The resulting effects of this increased temperature include contact material
fatigue, oxidation, and contact contamination. All of the above conditions result in increased
contact resistance, resulting in even higher temperature increases.
4-52
Typical failure mechanisms associated with switches are contact pitting due to arcing on
break, contact material transfer, contact weldment on make (resulting from excessive resistance and
heat generation), and mechanical failures resulting from the construction or packaging of the
particular switch. Application factors affecting the failure rate of switches are:
• Switching voltage - for source voltages less than 14V, arcing typically does not cause
serious problems but for source voltage greater than 14V, arcing can occur causing
contact pitting.
• Altitude - the dielectric strength of air is less at higher altitudes causing arcing to occur for
longer durations.
Switches which are not currently covered in MIL-HDBK-217E but should be added are:
• Centrifugal switches
• Capacitive-touch switches
• Membrane switches
• Circuit breakers with hydraulic-magnetic trip mechanism
• Ground fault interrupters (part of circuit breakers)
• Slide switches
A centrifugal switch is actuated by rotational velocity. The simplest type consists of a speed-
sensing unit that mounts directly on a rotating shaft, and a stationary contact switch assembly. The
basic control element is a conical-spring steel disc that has centrifugal weights fastened to the outer
edge of its circular base.
Inductive switches, mainly used where high cyclic rates are required, are classified in the
electronic category but rely on magnetics for their functionality. As the switch is actuated, an iron
core is slid through a coil creating a frequency change resulting in a signal transfer.
4-53
Membrane switches are devices in which conductive leads on the underside of a flexible
membrane are pushed through a hole in a spacer to make contact with conductive leads on a base.
Optional overlays are provided for user interface.
The hydraulic-magnetic construction circuit breaker consists of a solenoid with a dash pot
time-delay element (i.e., iron core). The dash pot time-delay tube contains a silicone fluid and a
return core spring. Operation depends on changes in the magnetic flux. Changes in flux are
caused by changes in coil current, which in turn cause changes in the position of the iron core
within the coil. The speed at which the core moves is controlled by the damping effect of the
silicone liquid in the tube.
The following tables summarize the failure modes of switches along with their approximate
relative rate of occurrence.
TABLE 4.4-2:
SWITCHES, GENERAL FAILURE MODES
Open 15%
Shorted 8%
Intermittent 19%
Out of Spec. 14%
Other 18%
Unstable 10%
Drift 9%
Leaking 7%
4-54
TABLE 4.4-3:
FLOAT SWITCH FAILURE MODES
Cracked/Fr ac tured 8%
False Response 23%
Leaking 8%
No Operation 23%
Out of Adjustment 15%
Seized 8%
Stuck Closed 8%
Stuck Open 8%
TABLE 4.4-4:
REED SWITCHES FAILURE MODES
Intermittent 10%
No Operation 30%
Open 10%
Out of Spec. 30%
False Response 20%
TABLE 4.4-5:
TOGGLE SWITCHES FAILURE MODES
Open 24%
Short 16%
Intermittent 25%
Mechanical 35%
As with the other part types, the data listed in the previous tables are based on the best
available data and will clearly be a function of device type, manufacturer, application, etc.
Therefore, the distributions given were only used to identify predominant failure modes and to test
the reasonableness of the hypothesized model.
4-55
4.4.2 Review of MIL-HDBK-217E Switch Models
The MIL-HDBK-217E switch section was analyzed for completeness and adequacy. The
findings of this investigation are listed below. The items that were included as factors in the final
model were a function of data availability and of the findings of this study. Therefore, not all
factors discussed were included in the final model.
(1) Part types that should be addressed for addition are centrifugal switches, capacitive-touch
switches, membrane switches, hydraulic-magnetic circuit breakers, ground fault interrupters
and slide switches.
(2) Contact material should be considered for inclusion in the model because of their varying
resistance to failure.
(3) AC versus DC application should be included in the model because arcing is more prevalent
in DC operation.
(4) The difference in failure rate between thermal and thermal-magnetic circuit breakers should be
included.
(5) The switch failure rate is currently proportional to actuation frequency when cycling
frequency is greater than 1 cycle/hour and independent of cycling frequency when cycling
frequency is less than 1 cycle/hour. This approach to switch failure rate prediction is too
simplistic. If the failure rate is directly proportional to the cycling frequency, then all failure
mechanisms should relate to actuation cycles. In practice, there are mechanisms relating to
the switch package which are independent of cycling frequency.
4-56
4.4.3 Switch Model Development
Xp = ?i b 7iQK E K C + %
%Q = Quality Factor
^-TJ = Usage Factor, function of load, cycling rate, contact material, and whether the load
is applied during switching. This is a wearout failure mechanism modeled per
Section 2.3.
Initial regression analysis of the data indicated that the environment factors derived for those
environments for which there existed data were consistent with the current MIL-HDBK-217
environment factors. The environment factor will therefore be kept unchanged. The regression
was run again with the current MIL-HDBK-217 environment factors and quality was specifically
analyzed. This regression analysis indicated that, based on the available data, that there was no
significant difference between quality levels. This is not intuitively correct and does not imply that
there is no difference in the failure rates, merely that the difference is smaller than that which can be
quantified based on the database and statistical techniques used. Typically, failure rate differences
of greater than 2:1 can be identified with the techniques used. Differences less than this are
difficult to identify given the inherent amount of noise in field failure rates. Therefore, since a
4-57
difference of less than 2:1 cannot be distinguished from the data, and quality intuitively makes a
difference, a 2:1 ratio in military vs. lower qualities will be used.
The next variable analyzed was the switch current rating. Unfortunately, there was
insufficient data available to quantify the effects of contact current rating based on the data
available. The stress (both rated and actual) however will be addressed in the utilization factor to
be discussed in Section 4.4.4. There was also insufficient data to quantify the effect of contact
material.
The base failure rates of various types and styles of switches were derived after
compensating for the above described quality and environment factors. These base failure rates are
given in Table 4.4-6:
TABLE 4.4-6:
SWITCH BASE FAILURE RATES
% Hours from
Type Xfo (Regression) Failure Records ^b
The column on the right is compensated for the zero failure hours observed.
4-58
The last variable analyzed was the number of active contacts. For example, the number of
active contacts in a DPDT switch is 4, a SPST is 1, and a 3PST is 3. The relationship between
failure rate and number of contacts is:
nc = (# contacts)-•"
There was insufficient data available to quantify the effects of either quality or environment
for rotary switches. Due to similar failure mechanisms to standard switches, the quality and
environment factors previously described for switches will be used.
A regression was run normalized to these factors and the base failure rates in Table 4.4-7
were derived.
TABLE 4.4-7:
ROTARY SWITCH BASE FAILURE RATES
% Hours from
Type ^b (Regression) Failure Records h
Rotary Switch 1.13 19.5 .22
Thumbwheel 3.59 9.9 .36
Since there was insufficient data to derive a factor for number of active contacts specifically
for Rotary Switches, the factor derived for standard switches will be used.
Although Circuit Breakers are considered in the general category of switches, they were
analyzed separately due to their inherently different construction characteristics. The data set was
first analyzed to determine if there were correlations within the data or outliers which would
prevent a valid derivation of model parameters. Several outliers were excluded, one of which
implied that a naval unsheltered environment was much more reliable than a ground fixed
environment.
4-59
Quality and environment were highly correlated, making it impossible to quantify the effects
of both. Therefore, the MIL-HDBK-217E environments were assumed to be correct, the observed
failure rate was adjusted to compensate for environment and the regression was re-run. This
analysis indicated an approximate 20:1 ratio in failure rate between commercial and military parts.
However, the significance of this factor was relatively low and therefore the available data does not
contradict the current 8.4:1 ratio in failure rates between commercial and military parts. Therefore,
the Quality and environment factors are given in Tables 4.4-8 and 4.4-9.
The contact configuration was also regressed against and the results were very consistent
with the current factor, which is equal to the number of contacts, as in Table 4.4-10. Therefore,
the contact configuration will be kept intact.
TABLE 4.4-10:
CONTACT CONFIGURATION FACTOR
Configuration *c
SPST 1
DPST 2
3PST 3
4PST 4
4-60
Tt was also attempted to quantify the failure rate as a function of the rated current of the circuit
breaker. However, since it was highly correlated to number of contacts, its effect could not
explicitly be quantified and was therefore not included in the model.
The observed failure rate was then adjusted for quality, environment, and contact
configuration, and the regression was re-run to quantify the base failure rates for each type of
circuit breaker. These base failure rates are given in Table 4.4-11:
TABLE 4.4-11:
CIRCUIT BREAKER BASE FAILURE RATES
Type ^b ( p / 1 0 6 )
Magnetic .68
Power Switch 1.74
Thermal .68
Bimetallic thermal switches were analyzed separately. Applicable specs, for these are MIL-
S-12285 and MIL-S-24236. Since all data available for thermal switches was from a Gg
environment and from commercial device types, the model was normalized to these variables.
Since there was no data available on MIL-Spec. thermal switches, a quality factor could not
be derived from the data. Therefore, the ratio of 2:1 between commercial and military derived for
basic switches will also be used for thermal switches. Similarly, the current MIL-HDBK-217
environment factors will also be used.
^p = ^b^Q^E
There was a total of 193, 879, 400 operating hours with 12 observed failures in the dataset;
yielding a base failure rate of .0619. Since this failure rate is in reference to a commercial part,
dividing by 2 yields a base failure rate normalized to a military quality part.
4-61
4.4.4 Switch Utilization Factor (k^j)
Switch and relay contacts can exhibit wearout failure mechanisms when exposed to repeated
switching operations under electrical load. This is primarily due to the arcing and subsequent
carbon generation of the contact. The variables accelerating this degradation mechanism are contact
configuration and material, voltage, current, temperature, operating interval and inductance and
capacitance of the load being switched. Although all of these variables affect wearout times for
switches and relays, the predominant variables, and those readily available to designers are current
voltage, inductance and capacitance. Therefore, these are the variables researched further for use
in the utilization factor.
References 4 and 74 present data and analysis of switching cycles to failure under various
operational conditions. The equations in Table 4.4-12 from Reference 4 relate the characteristic life
(in 10" cycles) to applied operating voltage and current for both AC resistive loads and DC loads.
TABLE 4.4-12:
CONTACT LIFE EXPECTANCY (10 6 ACTUATIONS)
Contact Current
Rating (Amps) AC Resistive Load DC Load
29.08 26.323
3
y.75^.14 v 1 . 3 3 j l . 3 e130L/R
*(0-4)
103.45 123.187
5
v.75r1.14 v 1 . 3 3 j l . 3 e 1 3 0 LR
*(>4-8)
219.74 307.94
10
v.75jl.l4 v1.33 jl.3e130L/R
*(>8)
4-62
An attempt was made to regress on the constant in these equations as a function of rated
current in an effort to derive a single equation representing the number of cycles to failure as a
function of rated current, applied voltage, and applied current. This attempt was unsuccessful due
to the fact that the linear regression implied negative cycles to failure for low rated current relays.
Therefore, the approach taken was to assume the equations in the previous table are valid for the
ranges of current ratings. The equation for 3 amp rating was assumed valid for the range 0-4
amps, for the 5 amp rating, >4-8, and for the 10 amp rating, >8.
Table 4.4-13 summarizes data available from Reference 74 on dry reed contacts made of
cobalt hardened diffused gold, containing carbon with a top layer of ruthenium. Contained in this
table is the voltage, current, Weibull a parameter (characteristic life), Weibull P parameter,
characteristic life predicted from Table 4.4-12 and the predicted/observed ratio. Several
conclusions were made from this data. First, the predicted mean cycles to failure are generally
pessimistic by an average failure of .41. Although not entirely accurate, it does err on the
conservative side which is desirable in this situation. Second, the beta values (needed for the
wearout failure rate term) observed range from 1.1 to 8.6, with a mean of 3.5. Again, a
conservative beta (lower value) is desirable since it will yield the worst case failure rates in the
early life of the component. Therefore, a beta value of 3 will be used in the model.
TABLE 4.4-13:
DRY REED CONTACT DATA
V I a a Predicted
/Predicted \
(V) (A) (106) P (106)
1 Observedl
4-63
As with the majority of electronic components, the failure rate of switch and relay contacts is
a strong function of quality, and of the manufacturing process. Reference 77 presents data
illustrating this dependence and indicates there are several orders of magnitude difference in the
times to failure between a good part and a marginal part. Since the models being developed herein
are generic models, they cannot explicitly account for specific manufacturing process variables.
The effects of marginal manufacturing processes are however partially accounted for in the quality
factor, assuming that the process controls and screens are effective in reducing defects related to
early and mid life failures. Given these limitations, the models developed herein are representative
of industry wide average failure rates.
Reference 77 also contains time to failure data on dry reed contacts. While enough data did
not exist to validate the predicted failure rate, the available data does indicate that the predicted
mean-time-to-failure is in the right range.
Since the wearout failure is being separately accounted for, the constant failure rate portion of
the predicted failure rate must be decreased so that only non-wearout failure rates are included.
From the failure mode distributions, it is apparent that approximately 50% of observed failures can
be attributed to failure mechanisms that the wearout term is intended to model. The base failure
rates must be decreased by 50% to accommodate this. Therefore, the final proposed models in
Section 5 of this report contain base failure rates which are 50% of those contained in Table 4.4-6
derived from the regression analysis. The value of 50% was derived from the data in Table 4.4-2
by assuming that "open", "intermittent", and "out of spec." failure modes are wearout related. The
percentage of these sum to 48%, or approximately 50% of the total failure rate. These failure
modes were identified as wearout related since the ultimate mode of failure for switch contacts
subjected to wear is open, intermittently open or increased contact resistance (out of spec).
4-64
4.5 RELAYS
The two main categories of relays are electromechanical relays and solid state relays (SSRs).
Electromechanical relays are magnetically-operated devices available in many different styles, each
having unique mechanical construction and electrical characteristics. Solid state relays control load
currents through solid state switches such as TRIACs, SCRs or power transistors. Unlike
electromechanical relays, solid-state relays have no moving parts and are often used in applications
where rapid on/off cycling would lead to wear out of conventional electromechanical relays.
Failure modes in SSRs are primarily associated with the TRIAC or SSR switching
characteristics. Most common failures take the form of SSR false turn-on with no turn-on signal.
For example, turn-on can occur if operating temperatures exceed the thyristor rating or transients
from the switched load or AC line momentarily exceed the thyristor breakover voltage. Other
failure modes/mechanisms include thermo-mechanical fatigue caused by cyclic temperature surges,
chemical reactions such as channeling and physical changes such as crystallization of materials.
4-65
The physical design of an electromechanical relay can be described by the contact
combination or form and the construction type. The current-carrying parts of a relay that are used
for making and breaking the electrical circuits are available in various combinations of contact
forms. Single-throw contact forms have a pair of contacts open in one armature position and
closed in another. Double-throw contact combinations have three contacts, of which one is in
contact with the second but not with the third in one relay position, and in the reverse connection in
the other relay position. Double-make and double-break contact forms have two independent
contacts that are both connected to a third contact in one position on the relay. The choice of
contact material and the shape of contacts impact relay failure rate. Contact reliability concerns for
relays are very similar to those of switches, and therefore the contact reliability discussion
presented previously are applicable.
• Class of application (e.g., military, commercial, industrial, machine tool control, etc.)
• Operational specifications
A number of test methods have been standardized to assure reliable performance of relays.
Several of the more important tests are listed in Table 4.5-1.
4-66
TABLE 4.5-1:
TESTS PERFORMED TO ASSURE RELAY RELIABILITY
Insulation Resistance Test Measures the resistance between mutually insulated members of a relay.
Values of insulation resistance can be important in the design of high
impedance circuits. Low insulation resistance may permit excessive
leakage current that can affect isolation of independent circuits.
Excessive leakage current can also be indicative of the presence of
corrosive impurities that can cause deterioration by electrolysis or
heating.
Dielectric Withstanding Voltage Test Detects flaws in materials, design, or construction of the unit which
might result in failure to withstand the specified test potential. It is a
static test, conducted without contact switching and in the absence of
contact arcing.
Winding Resistance Test Measuring the direct current resistance of a relay coil winding.
Winding Inductance Test Measuring the inductance of the coil winding. In relays, coil inductance
is a function of the number of turns of wire and the geometry and
reluctance of the magnetic circuit.
Winding Impedance Test Measuring the impedance of relay windings designed for use on
alternating current.
Contact Bounce Test Measurement of the duration of the intermittent opening or closing of
contacts caused by contact bounce.
Contact Chatter Test Monitoring contact chatter when relays are subjected to vibration,
shock, and acceleration tests.
Functioning Time Test Measure the operate and release time of relays.
Leak Test for Hermetically Scaled Relays Determine the effectiveness of the seal of a hermetically scaled relay,
which either is evacuated or contains air or gas. A defect in any portion
of the surface area of a seal part can permit the entrance of damaging
contaminants that could reduce the effective life of the relay.
4-67
4.5.1 Relay Failure Modes/Mechanisms
The failure modes and mechanisms for armature relays are summarized in Table 4.5-2.
TABLE 4.5-2:
ARMATURE RELAY FAILURE MECHANISMS
Review of the current MIL-HDBK-217 relay models resulted in the following observations:
(1) Model development activities for relays specifically addressed the impact of cyclic operation
and relay terminology. The existing relay cycling factor depends on relay quality and cycling
rate. Examples of computed cycling factors per the current model are given in Table 4.5-3.
4-68
TABLE 4.5-3:
EFFECTS OF RELAY QUALITY ON CYCLING FACTOR
Cycling Factor
Cycling Rate
(Cycles/Hour MIL-SPEC Lower Quality
1 .1 1
1 .1 1
10 1 1
100 10 10
1000 100 100
10000 1,000 10,000
Several aspects of this factor seem illogical. Initially, the difference between MIL-Spec. and
lower relays becomes smaller as the cycling rate increases (and is the same value for cycling
rates between 10 and 1,000 cycles/hr.). In practice, the opposite should be true. High
quality relays and contacts may be able to withstand repeated cycling better than the lower
quality parts.
(2) Specific characteristics of the relay (e.g., incorporate contact material, AC/DC operation,
frequency, shape, contact force, amount of wiping/sliding) should be investigated for
possible inclusion in the model.
?l 7l
^p = b E 7 l Q + ^u
where:
4-69
4.5.3.2 Relay Data Analysis
Initial regression results of the relay data were relatively consistent with expectations. This
was undoubtedly due to the fact that there were a large percentage of records (98%) which had
observed failures, thus resulting in a relatively large dataset to analyze. There were therefore
relatively few iterations required to arrive at the final results.
The results of the environment analysis are summarized in Tables 4.5-4 and 4.5-5. Table
4.5-4 summarizes the current MIL-HDBK-217 environment factors and Table 4.5-5 summarizes
those obtained from the regression analysis.
4-70
Comparison of existing factors (combined if necessary for consistency with the
environment/combination of environments in Table 4.5-5) to observed environment factors is
given in Table 4.5-6. The column labeled "current" denotes the 7tg value of MIL-HDBK-217E for
such environment for which a regression solution was obtained. The "observed" column presents
the regression solution, and the Column "Observed Normalized to Gg" presents the observed
factors normalized to a Gg environment. This was accomplished by dividing the observed factor
for each environment by the observed factor of. 12 for Gg. In this manner the 7tg for Gg is one.
It is these factors that are the proposed 7tg values for the new model. In cases where there is not
an observed environment factor for a particular environment, the ratio of proposed to current 7tg
values of similar environments were used to multiply the current values. For example, there was
not sufficient data identify an observed factor for each airborne environment. Therefore, all
airborne environments were combined for the analysis and a K£ factor of 233 was obtained. The
current average factor is 9.7 and therefore the ratio is 233/9.7 = 24. Each current airborne 7tg was
therefore multiplied by 24 to obtain the proposed 7tg values. Other environmental factors were
derived similarly.
TABLE 4.5-6:
COMPARISON OF NEW/OLD ENVIRONMENT FACTORS
Observed
Normalized
Environment Current Observed toGg
A *9.7 28 233
A
RW 46 100 833
G
M 15 7.4 64
GB 1 .12 1
GF 2 1.0 8.3
SF .5 .098 .82
NS 8 .98 8.2
* Average of all air environments.
4-71
TABLE 4.5-7:
PROPOSED RELAY ENVIRONMENT FACTOR
GB 1
GF 8.3
G
M 64
A
IC 168
A
UC 264
A
IF 216
A
UF 288
A
RW 833
Nu 27
NS 8.2
*M L 1584
**M F 600
cL N/A
sF .82
The environment factor for solid state relays is not expected to be as stringent as for
mechanical types and therefore the current MIL-HDBK-217E environment factor will be kept.
The quality factor obtained from the regression is given in Table 4.5-8. Although 1.9:1 is a
relatively modest factor, it was significant from the regression analysis.
TABLE 4.5-8:
OBSERVED QUALITY FACTOR
Quality
^Q
Military l
Lower 1.9
The base failure rates obtained for different types of relays are (after accounting for zero
failure hours) give in Table 4.5-9.
4-72
TABLE 4.5-9:
RELAY BASE FAILURE RATES
% Hours with
Type A.^ (Regression) Failure Records ^b
Additional factors analyzed were number of contacts and current rating of the contacts. There
was a very low statistical significance in the rated current factor and the regression illustrated a
negative relationship between failure rate and number of contacts. Due to these results, rated
current and number of contacts will not be included in the proposed model. Although originally
identified as potential model variables, the effects of contact shape and material could not be
obtained from the data.
The wearout failure mechanisms for relay contacts is essentially the same as for switches.
Therefore, the utilization factor for relays will be the same as that derived for switches. From the
Relay Failure Mode/Mechanism information, it is apparent that approximately 40% of observed
relay failures are due to wearout. Therefore, the base failure rates for relays in the final model in
Section 5 will be decreased 40% since this percentage will be accounted for in the XJJ failure rate.
4-73
4.6 CONNECTORS
• Ribbon
• Edge Board
• Pin
Connections:
• Terminal
• Connector Panel
• Wirewrap
• Crimp
• Clip
• Solder
• Weld
4-74
Connector failure modes include shorts, opens, high resistance, and intermittent failure.
Based on data collection from military and commercial applications, short and intermittent failures
are the predominant modes of failure (with short contributing 50%, and intermittent contributing
40%). Failure accelerating stresses contributing to failure modes of opens and intermittents are
temperature cycling, vibration, and corrosion from exposure to humidity or contaminants.
Additionally, the mating cycling rate highly influences reliability. When the cycling rate is very
low, a cleaning action takes place counteracting the formation of corrosion or oxide films without
causing excessive wear. Conversely, as the cycling rate increases, wearout failure mechanisms
become very significant.
There are two critical manufacturing aspects which must be maintained to produce a reliable
connector. For electrical and signal connectors, contact plating, contact form and physical
dimensions are critical variables. For optical connectors, physical dimensioning and alignment are
important design and manufacturing variables. For a reliable connector, there must be a consistent
connection between its male and female components. This consistent connection must be
maintained despite vibration and temperature cycling which can result in small amounts of
movement and corrosion. Without sufficient contact force and plating, corrosion can cause
increased resistance between contacts leading to failure.
There are a number of connector designs which can be used for a specific environmental
application. For example, if the application for a circular connector were in a high temperature
environment, the insert insulating material can be specified as vitreous glass or alumina ceramic
which will maintain it's mechanical integrity up to 250°C. However, as is the case with other
component types being modeled, it is assumed that the parts are operating below their maximum
ratings. If not, the models are invalid.
The failure rate and failure mechanisms for edge-board PCB connectors are distinct from pin
and socket PCB connectors. For edge board connectors, the connector mates with the edge of a
PCB to provide electrical connection. For many applications, including airborne environments, the
use of edge board connectors is restricted because of their greater frequency of failure.
4-75
Environmental contamination, vibration, temperature cycling and altitude tests are often
performed on connectors. Plating procedures and the even dispersement of plating are other
concerns resulting in the qualifications of connectors. Only military connectors are typically
subjected to formal qualification tests, but commercial grade connectors are often subjected to
functional tests to determine design integrity.
The dominant application variables affecting the failure rate of connectors are vibration,
temperature cycling, mating and unmating cycles, and contamination. To a lesser extent,
application variables affecting connector failure rates are the loads passing through the connector.
If the loads are properly specified by gauge versus current carrying capacity, this factor is of
relatively small influence.
Connectors have been a leading cause of reliability problems for many avionic electronic
systems. Due to the space constraints in high performance aircraft equipment bays, is it often
necessary to remove/replace several electronic boxes during flight-line maintenance simply for the
failed box to become accessible. As a result, many equipments are being repeatedly removed and
connectors are being stressed by mating/unmating cycles.
Table 4.6-1 summarizes failure modes/mechanisms, their accelerating stresses and percent
occurrence for connectors. This data is based on Reference 13 and is a summary of all connector
types for which data existed. It is a generic listing and will vary depending on connector type,
application, manufacturer, etc.
4-76
TABLE 4.6-1:
CONNECTOR FAILURE MODES/MECHANISMS
Short Contamination 9%
Abuse
Accelerating Stresses: Accelerating factors that degrade the reliability of electrical and fiber optic
connectors can be identified by temperature, environment, and mechanical stresses. Separately,
each causes specific degradation mechanisms and modes, but realistically they are interrelated to
induce combined acceleration of failure factors.
Temperature: Temperature cycling in some applications causes the expansion and contraction of
the mated connectors. If the temperature cycling is prolonged, then there is a possibility of the
mated connectors to loosen and separate, causing intermittent anomalies and open failures. This
condition would be further accelerated in high-vibration applications such as aircraft or with
connectors that do not have screw-type mating or mated connector support such as D-sub or DIN
connectors.
Another type of temperature accelerating factor is high contact resistance. This is caused by
increased temperatures accelerating the diffusion of inner plating materials such as silver, tin and
palladium-based metals to diffuse through the outer plating materials such as silver or gold.
4-77
Environmental: Environmental stresses are usually confined to acidic or caustic environments.
These types of environmental stresses will accelerate corrosion in all non-gold plated connectors.
The combination of temperature with environmental acceleration factors will induce the acceleration
of contact corrosion. Initially, early degradation will develop a thin film on the outer plating layer
which will require higher current potential to penetrate through to the contact. Later stages will
induce corrosion on all non-gold plated connectors.
Mechanical: Mechanical stress is confined to three areas of stress: Cyclic mating/unmating, pin
insertion stresses, and vibration stress. Cyclic mating and unmating and vibrational stresses are
the more important areas to address. Failures caused by pin plating deficiencies are directly related
to connector mating/unmating. Gold-plated connectors are standard for military applications while
commercial applications may use less expensive silver or tin plated connectors.
Gold, by definition, is a soft noble metal. Prolonged mating/unmating cycles will erode the gold
outer plating off of the connector pins, causing the tin, nickel or palladium-based inner plating to
be exposed to the temperature and environmental accelerating stresses listed above. Another
mechanism created by constant mating cycles is the loss of tension in female pin receptacles. The
results of this mechanism is a loose mating connection and high probability of an open connector
failure or intermittent anomalies in a high vibration environment.
Acceleration of connector failures due to insertion stresses are mostly human induced. Many of the
insertion stresses are caused by pin misalignment which will usually lead to broken pins or
shorting them against other pins. This type of failure would be most prevalent in high pin count
connectors.
The following outline summarizes in more detail potential failure modes and factors affecting
their prevalence.
4-78
Connector Failure Modes (Cont'd):
Moisture Intrusion
• Inadequate sealing of the internal structure
Pin/Receptacle Damage
• Use of probes
• Connector Misalignment
• Connector mismating
• Relative connector movement due to vibration
- Vibration damage
• Absence of positive screw-type couplings
• Inadequate support of cables or wire bundles
Silver Plating
High resistance/intermittent contact failure
• Silver sulfide build-up on contact surface
Wear-through of silver to contact base metal
• Silver oxidation
Rhodium plating
Hard open contact failure
Rhodium's inherent poor corrosion resistance
Galvanic corrosion caused by Rhodium to gold connector mating
• Mating/Demating
4-79
Plating Specific Failure Modes (Cont'd):
Tin plating
• Contact surface melting
• Heat generation
• Increased contact resistance
• Oxidation
• Creep
• Tin's inherent low-current capability
• Contact mating/relaxation
The current MIL-HDBK-217E failure rate model for connectors has been reviewed and the
following observations have been made:
(1) The connector factor for active contacts needs revising. The existing factor increases
somewhat gradually for pin counts up to 150 pins and then increases rapidly from 150 to 200
pins. As connector manufacturing and design becomes more advanced, the relationship
between pin count and failure rate is expected to have changed since the connector models
were last revised in the 1970s. Additionally, connectors are now available with greater than
200 pins.
(2) The current cycling rate factor should be reviewed with respect to the cycling stresses to
which many connectors are being exposed.
4-80
(3) Models need to be updated to incorporate newer technologies in connector design. There
have been advances in connector housing material and contact form, including zero insertion
force connectors.
(6) Fiber optic technology is increasing in popularity, especially where weight and reliability are
concerned. The Navy uses fiber optic technology on shipboard radar systems to effectively
reduce retrofit costs, save weight and space, and increase the performance capabilities of their
systems.
4-81
4.6.3.2 Connector Model Development
4.6.3.2.1 Connectors
Initial analysis of the connector dataset revealed several limitations. First, there was
insufficient data to quantify several variables, including quality and insert material. Quality again
could not be quantified due to the high correlation between quality and environment. Therefore,
the current quality factor of 2:1 between military and commercial connectors were assumed to be
correct and the observed failure rates were adjusted (divided by) this factor and quality was not
used in subsequent regression analysis.
The next variables analyzed were environment, connector type and connector plating
material. Since the precise temperature of all observed failure rates was not known, the
temperature factor for each was calculated using a E a = .14 and the default temperatures of MIL-
HDBK-217E. The value of .14 eV was derived from the current MIL-HDBK-217 temperature
factor and the observed failure rates were then normalized to this value.
After several iterations of combining various environment categories to obtain consistent and
intuitively logical results, the following environmental factors were obtained in Table 4.6-2:
TABLE 4.6-2:
OBSERVED ENVIRONMENT FACTOR
Environment ^E
Ground 1
Airborne 5.53
N
SB 1
Although all the specific environment categories could not be quantified from the available
data, the above factors are consistent with the current MIL-HDBK-217E factors for MIL-Spec.
connectors. Therefore the connector environment factors should be kept intact without
modification.
The factor for gold plated connectors were observed to be 1.27 times better than copper
although it was statistically a relatively insignificant factor and will not be used in the model.
4-82
The mating/unmating frequency factor was the next variable analyzed. Since the failure rate
data indicates that the reliability of connectors in general is very high, wearout failures due to
mating/unmating are not prevalent in the data set. If they were prevalent, the observed failure rates
would be much higher due to the fact that wearout mechanisms are common cause, indicating they
would effect a large percentage of the population.
The base failure rates for various connector types were then derived from the regression
analysis by compensating the observed failure rates for the quality, temperature, environment, and
mating/unmating frequency factors previously described.
The base failure rates for the various types of connectors are given in Table 4.6-3 (after
compensating for the percentage of hours associated with 0 failures).
TABLE 4.6-3:
CONNECTOR BASE FAILURE RATES
% Hours with
Type \^ (Regression) Failure Records ^b
4-83
Also analyzed in the regression was the number of pins. There was a statistically
insignificant relationship between number of pins and failure rate and when forced into the
equation, it indicated that failure rate was inversely proportional to pin count. Since this is not
intuitive, the factor for pin count was therefore discarded from the model.
4.6.3.2.2 Connections
^p = ^b ^Q^E
where X^ is the base failure rate as a function of connection type. Since the predominant failure
modes are similar for connectors and connections, the environment factors for connectors will also
be used for connections. The initial regressions also indicated that there was not a significant
difference between military and commercial connections. This is not surprising since the
technology is essentially the same.
The only connection type quality is considered important is crimp types. For these the
current factor will be kept. For all others quality is not a model variable.
Table 4.6-4 presents the results of this analysis and includes, for each connection type, the
217E X^, the observed X^ and the proposed X^. Observed failure rates were corrected for
environment and then averaged to obtain the observed X^.
4-84
TABLE 4.6-4:
CONNECTION BASE FAILURE RATES
The connection model will therefore be kept unchanged with the exception of the
modification of the base failure rates and addition of the terminal and contact spring categories. If
the new data for which there were zero failures (indicated with a "<" symbol) suggested the worst
case failure rate (calculated with assuming one failure) is lower than the current value, the new
worst case number was used. If the current number is less than the worst case assumed value, the
current number was kept. Only one failure rate, for solderless wire wrap, was increased.
4.6.3.2.3 Sockets
All data records available for which there existed observed failures on sockets were from a
ground benign commercial environment. Therefore, the quality and environment factors could not
be derived from this dataset and therefore the connector factors will be used. The models will be
normalized to ground benign environment and commercial quality level. The socket failure rate
model is:
=
^p ^b^E^Q
4-85
Since there was insufficient data to quantify the environment factor specifically for sockets,
the environment factor previously described for connectors will be used.
The observed failure rates for the socket types (for which there existed failures) are given in
Table 4.6-5.
TABLE 4.6-5:
OBSERVED FAILURE RATES FOR SOCKETS
DIP .00064
Relay .037
Transistor .0051
Tube <.011
Chip Carrier <.0024
Pin Grid Array <.014
SIP <.0030
Since all failure data was from the same environment and quality level, a regression analysis
was not necessary and the above failure rates were computed by summing the failures and hours
for all ground benign, commercial data.
The failure rates preceded by a "<" sign are of device types for which there was no observed
failures. For these, the upper limit of failure rate presented was calculated by dividing one failure
by the observed number of operating hours.
Although there was no observed failures for military sockets, there was a substantial number
of observed hours for Military DIP Sockets. Table 4.6-6 summarizes the DIP data.
TABLE 4.6-6:
DIP SOCKET DATA
Commercial Military
4-86
The number of total operating hours for the military data was calculated by adjusting for the
environment by multiplying each data records hours by 7tg. This indicates that there is at least a
.3:1 difference in military vs. commercial DIP sockets. Therefore, this ratio will be used for the
7lQ.
While there was not enough failure data to quantify the failure rate of Chip Carriers, Pin Grid
Arrays, or SIP's, there was a significant number of observed hours associated with them.
Therefore, the upper limit values in Table 4.6-4 will be used. Additionally, there was insufficient
data to quantify the effects of the number of active pins.
^p = V ^ Q
Xb = .00064 for DIP Sockets
TZQ = .3 Military
1 Commercial
4-87
4.7 INTERCONNECTION ASSEMBLIES/PRINTED WIRING BOARDS
Interconnection assemblies are the medium which provides electrical connections to the
components which collectively form an electrical circuit. The circuit board can be various
combinations of multi-layer or double-sided, printed wiring or discrete wiring and components can
be mounted to the board using either Plated Through Hole (PTH) or Surface Mount Technology
(SMT). A Surface Mount Technology (SMT) interconnection assembly typically is comprised of a
circuit board and solder connections which both physically and electrically connect the components
to the board. PTH technology uses the solder joint for electrical connection only. There are
various methods for forming solder connections including wave solder, hand or vapor phase
soldering.
Most soldering operations for military systems utilize wave soldering. Wave soldering
systems for printed wiring assemblies generally produce more reliable connections due to less
variability in the process. These systems can apply the flux, dry and preheat the board, solder
components, and clean the completed assembly. Some of the systems have special features where
the flux is applied by passing through a wave, by spraying, by rolling or by dipping. Several
systems employ oil mixed with the solder to aid in the elimination of icicles and bridging between
conductor paths. Vapor phase or IR soldering is typically used for the reliable soldering of Surface
Mount Boards.
For interconnection assemblies using plated through hole (PTH) technology, fracture of the
PTH is the primary cause of failure. For these types of circuit boards, holes are drilled through the
pads of the inner layers of a multilayer printed circuit board. Drilling exposes a rim of copper
around the entire circumference of the hole. The copper on the individual layers in the PTH is
connected by copper plating. Plated through holes are also used for interconnection on some types
of discrete wiring assemblies. The discrete wiring boards are plated in an electroless copper bath
where copper is deposited to form the component holes and make connections to the discrete
wires.
PTH barrel stresses are significantly higher in the central portion of the PTH when the
assembly is exposed to thermal cycling. Internal circuit planes which inhibit free expansion of the
PTH and additive loading on PTH lands have been considered to be the principal reasons for
higher centralized stresses. As the number of internal circuit planes increase on a printed wiring
4-88
board, the stresses in the central plated through hole region become larger and more failures are
expected.
One advantage of surface mount technology is its ability to minimize board real estate. For
surface mount devices, the component is attached directly to the surface of the printed wiring
board. Even when surface mount technology is predominantly used, it is still necessary to use via
PTHs to provide electrical connection between circuit planes. Via holes are also subject to barrel
cracking but, the physics of failure are different due to the absence of an inserted lead. The
absence of this lead changes the mechanical strength and TCE of the via. Also the integrity of the
via is a strong function of the completeness which the hole is filled.
Manufacturing difficulties can accelerate the formation of PTH barrel cracks. The formation
of barrel cracks is generally due either to imperfections in the PTH barrel wall which greatly
amplify the level of axial strains or very poor effective ductility of the copper plating. Poor drilling
or excessive acid etching during the hole wall cleaning process can lead to rough barrel walls. A
level but thin plating on the rough barrel wall may then lead to localized stress concentrations and
large plastic strains. Even if the PTH walls are smooth, variable electroplating processes may yield
copper of very low conductivity.
In addition to surface mount or plated through hole printed wiring boards, design options for
circuit boards include discrete wiring boards and flexible boards. These technologies are
sometimes used in specific instances justified by particular design requirements.
Flexible circuit boards are not restricted by a rigid substrate and are commonly used in many
electronic systems. They are sometimes used in place of interconnect cabling to connect between
moving assemblies, or when a flexible board is required for volume or enclosure shape reasons.
Since they are not rigid, their reliability concerns differ from those of rigid boards. More
specifically of concern is the integrity of the solder joints when the board is exposed to movement
or vibration. Additionally since the mechanical and thermal properties of the board substrate is
different than rigid board, their behavior under temperature cycling conditions is expected to be
different.
The most common form of discrete wiring boards are Multiwire boards (trademark name).
In this technology, small wires are imbedded in the laminate in lieu of printed wiring. For these
designs, it is possible to cross paths on a single circuit plane due to the insulation on the wire.
Two distinct failure mode areas for Multiwire assemblies are the wire crossover points and where
4-89
the PTH interfaces with the wire. The wire crossover potentially can be a source of failure. When
one wire crosses another, there is typically 0.0012 inch of polyimide insulation between them.
The typical breakdown voltage at a single crossover is 1,500 - 2,000 volts. The wire is ordinarily
tested by the manufacturer to determine its ability to maintain insulation integrity under extreme
conditions. Environmental testing at several testing laboratories has not shown degradation of the
insulation resistance between crossovers; however more detailed analyses are required. Although a
limited amount of test data that is available has indicated that the connection of the wire end to the
copper plating in the PTH is reliable, there is another reliability concern in the use of multiwire
technology that relates to the drilling and etching operation. Specifically, the wires are prone to
overetching, causing the wire to withdraw thus exposing it for potential shorting to other circuit
elements or stressing it such that opens can occur. Therefore, quality control procedures are
critical in the fabrication of these boards.
The advent of surface mount technology has had a dramatic impact on the reliability of
interconnection assemblies. The printed circuit board design and manufacturing process of SMD
boards require much greater attention to produce reliable solder connections. To produce a reliable
surface mount solder connection, it is necessary to tailor the thermal coefficient of expansion
(TCE) of the printed circuit boards substrate to the TCE of the device in order to minimize thermal
fatigue in the solder connection. The distinction between "tailoring" and "matching" TCEs is
important because of the localized heating of the electrical component when power is applied.
Solder Joint Fatigue: A prime reliability issue associated with SMT assemblies involves the solder
joint integrity between the surface-mounted component and the printed wiring board.
Thermal stress results when materials with different TCEs (Printed wiring board and chip carrier)
are joined and exposed to variations in temperature. When the materials respond to fluctuations,
each at their own rate, the bond which ties them together (the solder joint) restricts their
independent movement. The resulting damage to the solder joint is cumulative in nature; that is, as
the number of temperature fluctuations increases, the solder joint progressively weakens and the
probability of failure increases. A worst-case scenario for solder joint fatigue is power cycling
with large temperature fluctuations. The substantial changes in temperature coupled with materials
which have widely differing thermal coefficients of expansion produce an extreme fatigue
environment.
When such stress is applied to the assembly, both the substrate and the component deviate from
their original shape concurrent with their individual TCEs. The difference in TCE between the
4-90
substrate and device results in stress on the solder joint. Solder cracking problems become
significantly worse as the number of solder joints increases with package size and the power
dissipation increases with die size and function. As a leadless chip carrier increases in size from 18
to 64 pins, the allowable TCE difference between the substrate and the chip carrier must decrease
from the typical 7 ppm/degree C to 2 ppm/degree C in order to achieve the same solder joint cycles
to failure.
Printed wiring board substrate designs can be produced from a variety of materials.
Historically, epoxy glass boards have been the most popular for PTH technology. Other board
materials are necessary for SMD technology since the TCE of glass epoxy is too high to produce
reliable SMD boards surface mount technology. However, the use of the polyimide boards has
long been proposed as an alternative for epoxy glass for PTH boards as well. Each board material
has different TCE, drilling characteristics and other parameters which impact failure rate. A
summary of various substrate materials and their TCE characteristics are given in Table 4.7-2 later
in this section.
Electrically, active and passive elements are designed and fabricated with similar technology,
reliability standards and manufacturing processes for both SMD applications and PTH
applications. Therefore, the failure mechanisms of the active elements are also similar. The
connections and packaging of these two device types, are however very different. Surface-mount
components (SMCs) are not afforded the inherent internal board heat sink that PTH devices are,
whose leads penetrate the board surface and thermally connect to internal metalization. SMCs
often rely on thermal vias to transport heat away from the chip. Heat transfer by this mechanism
can be efficient if the vias are located where heat concentration occurs. The heat sinking properties
of the mounting technique along with the thermal properties of the package are important factors
since the failure rate and reliability are heavily dependent on device operating temperature.
The poor solderability of printed wiring boards is estimated to cause 50% of the solder
defects and approximately 20% are caused by the component lead solderability problems. The
other 30% are possibly due to solder composition or processing methods but more likely due to the
application of operating stresses.
4-91
Improper or defective solder joints may occur in response to a large variety of factors, including:
• Mechanical Considerations
Solder joint fatigue
Solder joint formation anomalous effects
• Metallurgical Considerations
Solder composition
Wettability of metallizations
Solder contamination
• Chemical Considerations
Oxide formation effects
Cleaning of flux residues
Solder Joint Formation Anomalous Effects: The formation of the solder joint is also an important
factor in the reliability of the assembly. The alignment, location, the degree of parallelism between
the package and the substrate as well as the amount and shape of the solder contained at each joint
location all have a dramatic effect on how the solder joint reacts to stress.
Solder Composition: The solder alloys themselves have fatigue properties which are inherently
characteristic of the alloy composition. Their behavior, therefore, is largely dependent upon how
that composition reacts to the thermal-mechanical stresses to which it is exposed. Solder alloy
selection is based on its strength characteristic and its metallurgical compatibility with the base
metal with which it will form a bond. Over 90% of the solder used in the electronics industry is of
a tin-lead composition. The tin-lead solders typically used in the soldering of surface mount
assemblies are considered to be soft solders due to their physical behavior under stress conditions.
Soft solders react to the mechanical tension by absorbing some of the stress; however, some
deformation occurs with each stress load. After repeated load applications, the solder becomes
permanendy deformed which allows cracks to develop and propagate into failures.
4-92
solder in a solder joint is responsible for solder bridges that develop between adjacent leads. This
solder bridging creates a conduction path between leads which should be isolated.
Increasing the clearance or stand-off heights between the component and the board allows the
strain which develops during cycling to be absorbed by the main body of the solder connection. A
small stand-off height limits the area through which the strain can be absorbed which results in
solder joint cracking.
Solder Contamination: Surface mount terminations are generally formed from or coated with
precious metals such as gold, silver, platinum, palladium, etc. These terminations are readily
soluble in solder, and if left unprotected the terminations become contaminated when placed in
contact with solder. The intermetallic compound formations which result from the interaction
between the active solder components (tin) and the soluble metallization (precious metals) produce
weak solder joints at elevated temperatures. The process of intermetallic compound formation can
be controlled by proper heat treatment, choice of solder alloy or the use of an underlying film
(nickel) as a barrier to inhibit the dissolution of materials. The use of barrier materials has been
widely accepted as a means of providing an interface between the terminations and the solder,
thereby protecting each from contamination.
The intermetallic compound formations produced by the dissolution of the component lead
material into the solder is responsible for the contamination of the solder joint. Any precious metal
which dissolves into the joint becomes a problem which is aggravated as the concentration of the
metal increases. This is typically expressed as a solder joint which becomes consumed by the
process of diffusion between the precious metal and the tin in the solder. This consumption
process is initiated as the molten solder comes into contact with the surfaces to be joined but may
also continue throughout the life of the joint.
4-93
This contamination process is responsible for producing rough or gritty surfaces which
reduce the ductility of the solder joint. This loss in the plastic response behavior of typical solder
can be influenced by a relatively small amount of contamination. The contamination reduces the
yield point (i.e., the point on the stress-strain curve which separates elastic and inelastic
deformation) and causes the solder connection to be sensitive to even smaller temperature
fluctuations which negatively impacts the life of the solder joint.
This contamination is also responsible for the formation of brittle solder joints which fail
characteristically at much lower temperatures than would ordinarily be expected. Additionally, the
dissolution of these metals decreases the melting point of the solder itself, which makes assembly
and rework difficult.
Oxide Formation Effects: Surface mounting relies on the component being supported during
solder reflow by the surface tension forces of the solder. When molten solder is exposed to air it
quickly forms and oxide skin which can reduce the surface tension plays an important part in
successful soldering operations. Careful monitoring of the soldering process is required to ensure
the application of quality solder. Reduced exposure to oxidizing agents and other contaminants is a
must in the formation of reliable solder connections.
Cleaning of Flux Residues: The criticality of removing flux residues prior to performing the
soldering process is evidenced by the number of voids formed in the solder. Trapped air and flux
forcefully escape from the solder, leaving behind harmful voids. Defects such as voids in a solder
joint have a large effect on the fatigue resistance of a solder joint. Voids become stress-
concentration sites which alter the typical stress patterns.
Substrate Reliability: The primary failure mechanism plaguing substrate reliability have
traditionally been due to the plated through holes required to accommodate inserted package leads.
With the elimination of hole drilling for surface mount packaging and the size reduction in the holes
drilled for thermal/electrical vias, surface mounted substrates have the potential for a corresponding
increase in reliability.
The problems of mating materials with unlike thermal coefficient properties have been
addressed at the board level, by manipulating substrate materials and constructions, the magnitude
of the stress which develops in the solder joint has been substantially reduced.
4-94
The operation of the component mounted generates heat in the component package at a
greater rate than the substrate during powered operation, and, therefore, the lag time of the
substrate heating causes stress to develop in the solder bond which connects the component to the
substrate.
A summary of the failure modes and mechanisms of Printed Circuit Boards, Multi-wire
Boards and Wire Wrap Boards that have been reported in the literature are as follows:
Single sided
• Open
• Open Run
• Delamination
• Lifted Pad
• Excessive acid etching during cleaning
• Thermal expansion of different materials
• Cracked solder joint
• Cracked board
• Short
• Delamination
• Thermal expansion of different materials
• Excessive solder
• Intermittent
• Thermal expansion of different materials
• Delamination
• Cracked solder joint
• Cracked board
4-95
Printed Circuit Boards (Cont'd):
• Short
• Delamination
• Thermal expansion of different materials
• Excessive solder
• Intermittent
• Thermal expansion of different materials
• Delamination
• Cracked solder j oi n t
• Cracked board
Multi-wire Boards
Short
Shorted run @ crossover
• Wire insulation & wire deformation
• Vibration
• Thermal cycling
4-96
Multi-wire Boards (Cont'd)
Open
• Open Run
Delami nation
Lifted Pad
Excessive acid etching during cleaning
Thermal expansion of different materials
Cracked solder joint
Vibration
Thermal cycling
Cracked board
• Intermittent
• Thermal expansion of different materials
• Delamination
• Cracked solder joint
• Cracked board
• Intermittent
• Poor connection between wire & wire post
• Insufficient tension of wire
• High vibration environment
• Cracked board
• Short
• Wire insulation cold flow
4-97
4.7.2 Interconnection Assembly/Printed Wiring Board MIL-HDBK-217E Model Review
The existing MIL-HDBK-217E model has been reviewed and the following deficiencies have
been noted:
(1) Board materials other than epoxy-glass (FR-4, G-10) need to be included.
(2) Via holes used to provide interconnection between circuit planes need to be handled
differently than plated through holes.
(4) It must be clearly and definitively stated that the interconnection assembly model pertains to
the failure rates of both the printed wiring board and the solder connections.
(6) Wearout failure mechanisms including solder fatigue from temperature cycling needs to be
addressed for both Surface Mount Devices and Plated through Holes.
(7) The various lead configurations including leadless, Gull Wing and S-lead need to be
accounted for.
(8) Temperature cycling effects from the various environments need to be defined and accounted
for.
Most of the models developed in this effort were derived primarily from field failure
experience. There are several problems however in deriving a circuit board model in this manner.
First, it is almost impossible to collect meaningful field data on circuit boards due to the fact that
most maintenance activities will trace the failure of a populated board to a specific component and
rarely attribute the failure to the board itself or the solder connection. Secondly, the model being
developed herein is extremely sensitive to the temperature variations and cycling rates of a
particular application. Since this data is not available for any data collected in this effort, the
resulting data is of limited value. For these reasons, and the fact that many researchers have been
4-98
studying and modeling SMT and PTH, the model for circuit board developed herein was
developed from theoretical considerations and from laboratory test data. The single exception to
this is the fact that part of the existing MIL-HDBK-217 model is used to model defect related PTH
failures.
From the research conducted in this model development effort, it was concluded that the
primary failure mechanism of surface mount devices is a common cause wearout mechanism due to
solder joint fatigue. Plated through holes on the other hand exhibit both wearout from temperature
cycling of the PTH barrel and defect related early and mid life failures due to incomplete filling of
the hole and subsequent mechanical stresses. This is not to say that SMD assemblies are not prone
to failure from defects, only that the predominant failure mechanism is wearout related.
Additionally while there is data to support a defect related failure rate for PTH assemblies, the
field data necessary to accomplish this is not available for SMD assemblies. It will be shown later
in this section that the wearout term is based on the Weibull distribution whose parameters have
been empirically derived from test data. The shape parameter therefore will be representative of the
observed values and will include the effects of defects.
where ?ii (o^) = Average Life Cycle Failure Rate due to Surface Mounted wearout, function
of oq (characteristic life) and Design Life Cycle, oq is a function of:
^2(0:2) = Average Life Cycle Failure Rate due to PTH wearout, functions of a2 and Life
Cycle. <x2 is a function of:
4-99
- P T H material
Substrate thickness
Temperature change
Temperature cycling rate
The premise of this model is that there are basically two types of failures possible for PWB's:
(1) Common Cause - i.e., as a result of X-Y expansion mismatch resulting in fatigue (and hence
wearout).
(2) Special Cause - i.e., defects in plated through holes that result in early and mid-life failures.
Special Cause (defect related) failures tend to have P's (from the Weibull distribution) close
to 1 and therefore can be modeled with a constant failure rate. The probability of defect related
failure mechanisms occurring is strongly a function of the quality of the fabrication process.
Additionally, the screens for defect related failure mechanisms are typically very effective,
indicating that the field failure rate is a strong function of both quality of the fabrication process and
the screening to which the board is subjected.
Another premise of this model is that temperature cycling is the primary failure accelerating
stress. While shock and vibration can also accelerate some failure mechanisms, it typically is only
an issue in cases where the board is exposed to severe conditions of shock and vibration. These
conditions can occur if the board is not damped enough or rigid enough and the applied stresses
causes a resonance. While these are important reliability considerations, they are unpredictable due
to the fact that they are special cause design problems and not related to the inherent reliability of
the board itself. For this reason, the only stress the wearout failure mechanisms are a function of
is temperature cycling. The environmental effects are, however, accounted for in the environment
factor for defect related PTH failure rate.
4-100
Based on the assumption that PTH and via cracks are a function of defects, the failure rate
contribution is treated in this model as exponential, corroborated by the conclusions in Reference
54. The solder joint fatigue contribution to the failure rate is a function of X-Y plane TCE
matching and is treated in the model as a wearout item. The factors for this portion of the model
are based on the Coffin-Manson model.
(1) Identify the device on the board exhibiting the worst characteristic life. This will be a
function of material (substrate and device) device dimensions, and solder height.
(2) Predict the characteristic life for this component and translate to a failure rate per the
methodology in Section 2.3.
The wearout failure rate is only calculated for the part exhibiting the lowest predicted number
of cycles to failure. This occurs for the largest device exhibiting the largest mismatch in TCE.
This was done simply for usability and to expedite the performance of reliability predictions using
the model, and to avoid calculations which have little or no impact on the final predicted result.
Reference 62 confirms this by stating that there is little risk from small passive devices and the
predominant reliability risk comes from large ceramic chip carriers.
The via and PTH are separated since their reliability characteristics vary due to the fact that
the PTH typically has a component lead through it and the via does not (solder only). This results
in different thermal response characteristics.
Iannuyzelli (Reference 53) has shown that the manufacturing process can impact field
reliability. This is based on the fact that damage is cumulative and that the manufacturing process
4-101
exposes the assembly to the highest level of stress that will ever be seen. He concludes that the
least to most damaging method is as follows:
Wave Soldering
- SMT Repair
- Vapor Phase Soldering
Quantification of how these processes affect the field reliability of assemblies is not possible
and therefore they will not explicitly be accounted for in the model.
The characteristic life a ] and (X2 are based on the unmodified Coffin Manson model:
f &y\
N
f=2 2 Ef
V V
where:
The fatigue ductility exponent, c, is a constant in the unmodified version of the Coffin
Manson model. Englemaier (Reference 62) has proposed a modified version of the Coffin-
Manson model in which c, instead of being a constant, takes the following form:
where:
f = Cycling frequency
4-102
After reviewing this model and consulting with various industry experts, it was concluded
that although the modified Coffin Manson model appears to be valid under some conditions, the
unmodified version appears to be more universally accepted and applicable to a wider range of
situations. For this reason, and to keep the models as simple as possible, the unmodified version
is used in these models.
Generally accepted values of 2Ef and 1/c are .65 and -2.26, respectively. Using these
values, the mean number of cycles to failure can then be rewritten as:
M 1 MyV2.26 M MyY2.26
N = N
f = 2 U5j b(SMT) ^65J
Here, Nwgfyi j \ has been included in place of the constant ^ since, as will be discussed later, it
A
T= ^[aS(TSS-To)-aCC(TCC-To)]
a
CC ~ TCE of chip (device)
=
^CC Upper device temperature (Pwr. on)
4-103
To use this model for failure rate predictions, values for TCE's (ag, CI.QQ) and temperatures
(T 0 , TQQ, T^g) must be derived as a function of operating environment. Ideally, the prediction
would be performed based on knowledge of the actual values of a given application. Since this is
rarely the case, however, default values must be available. The following discussion summarizes
the derivation of these default values.
Temperature
1
ambient
8 (case-amb.)
LCC CA '
I
T
LCC (CaSe)
8 Air (for thickness h)
T s - (substrate)
FIGURE 4.7-1:
THERMAL MODEL
The thermal resistance between the junction and case (QJQ) is much lower than the thermal
resistance of the case to ambient (i.e., QJQ « QQA), which is obvious by examining typical QJQ
and 6JA values ( 6 j ^ « 8JQ). This indicates that the case temperature (TQQ) will be higher than
the substrate temperature by an amount of temperature rise due to power dissipation. This
temperature rise can be calculated in two ways, as is currently done in MIL-HDBK-217 models:
T
RISE = p e JC
where:
4-104
or:
T
RISE = ( AT ) (S)
where:
S = The electrical stress on the device divided by its maximum rated stress.
Figure 4.7-2 illustrates the thermal profile for this situation as a function of time.
The worst case difference between the case and substrate temperature is BJQ P:
T
CC " T SS = e
JCp
T T
cc - A + eJCP
T T
SS = A
4-105
The strain range can therefore be rewritten as:
AY = H [ a s ( T A " T 0 ) - « C C ( T A + 6 j C P " T o ) ]
TABLE 4.7-1:
ENVIRONMENT AT VALUES
AT
Proposed MIL-HDBK-217E (Ref. #55
Environment Environment Recommended) TA T AT
GB G 30 30 23 7
B> G MS
GF GF 55 40 14 26
G
M GM,Mp 55 35 14 11
A
IC A
IC> AIB> A IT 30 55 14 31
A
UC A
UC> A UT' A UB 55 71 14 57
A
IF A
IA'AIF 30 55 14 31
A
UF A
UF> A UA 55 71 14 57
A
RW A
RW 30 55 14 31
Nu N
U- NUU> N H 55 75 14 61
NS N
BS- N
S 50 40 14 26
ML USL' ML 50 55 14 31
MF Mpp, M F A 50 45 14 31
cL cL 50 40 14 26
4-106
T^, obtained from MIL-HDBK-217E, defines default ambient temperatures as a function of
application environment. These are worst case values and the actual ambient operating temperature
should be used to calculated AT if possible. Also temperature rise from a nearby heat source must
be accounted for.
T Q is the ambient temperature when the equipment is not in operation. Ref. #65 has
determined that the average outdoor ambient temperature in the continental U.S. is 14°C.
Therefore, 14°C will be used for TQ in uncontrolled outdoor environments. With the exception of
ground benign , all environments are considered in this category. Ground benign is a controlled
environment for which an ambient temperature is typically 23°C.
With the exception of Gg and Gp environments, the AT values arrived at agree very well
with the recommended AT values published in Reference 55, thus lending a degree of confidence
in the values.
-2.26
Nf - Nb(SMT)^|as(Tss-T0)-acc(Tcc-T0)|xlO-6j
or:
-2.26
Nf = N b ( S M T ) ^|as(AT)-a c c (AT + TRISE)|xlO-6j
where:
AT = Environmental AT
Although the above equation is specifically applicable to SMT solder joints, it will be
extended to model PTH wearout failures.
4-107
Table 4.7-2 summarizes the X-Y thermal expansion coefficient for various circuit board
substrate materials (extracted from References 56-63). Table 4.7-3 summarizes the TCE's of
package material, Table 4.7-4 summarizes the PTH/via material TCE's, and Table 4.7-5
summarizes the TCE values of the Z axis.
4-108
TABLE 4.7-2: X-Y TCE VALUES (CONT'D)
TCEfPPM,
Substrate Material Reference Average Value
TABLE 4.7-3:
TCE'S OF PACKAGE MATERIALS
TABLE 4.7-4:
PTH/VIA MATERIAL TCE VALUES
Solder 27
Copper 17
4-109
TABLE 4.7-5:
Z AXIS TCE VALUES
TABLE 4.7-6:
LEAD CONFIGURATION N f (REF. #66)
The geometric mean of these ranges can be used in the model developed herein as a relative
figure of merit between lead configurations. This factor is normalized to the leadless configuration
since the model developed herein is normalized to the leadless configuration. Therefore, the lead
configuration modification factor is given in Table 4.7-7.
4-110
TABLE 4.7-7:
LEAD CONFIGURATION FACTOR
Leadless 1
S Lead 150
Gull Wing 5,000
The study producing these values (Reference 66) used an 8 mil solder joint height for the
leadless configuration. Since the model is normalized to the leadless configuration, predictions for
S Lead and Gull Wing Configurations should use h = 8 in the equations.
The PWB model yields a failure rate in failures per calendar time since the accelerating
stresses are power cycling related and not related to operational time. Therefore, the mean cycles
to failure predicted must be converted to mean hours to failure. This is done first by identifying the
number of temperature cycles per calendar hour for a given application. The conversion is
therefore:
where:
Cycling Period = Average calendar time per temperature cycle (in 10" hrs./cycles)
If the actual cycling period is not known, the default periods listed in Table 4.7-8 should be
used. These values are obtained from Reference 55.
4-111
TABLE 4.7-8: CYCLING RATE VALUES
Consumer 4200
Computers 170,000
Telecommunications 4200
Commercial Aircraft 340,000
Industrial 21,000
Military Ground Applications 30,000
Military Aircraft 115,000
It would be desirable to define the absolute values of MTTF and P based on empirical data
for a given process since there can be large degrees of variability as a function of the manufacturing
process. Theoretical models, such as the Coffin Manson model, although based on sound physics
of failure principals, do not necessarily offer an accurate absolute measure of the number of cycles
to failure. Additionally, they provide only MTTF information and do not estimate the variance or
Weibull shape parameter (P) in a given process. For situations in which the circuit board design is
robust enough to function reliably in a given application for long periods of time, the failure rate is
highly dependent on the value of p. Although the P is highly process dependent, and can indeed
vary significantly within a given process, a worst case value should be used unless it can be shown
through empirical data that another p value is appropriate for a given process. Using a
conservative P will also serve to account for some of the early life defect related failure
mechanisms.
4-112
PTH Wearout Modeling
PTH wearout modeling is accomplished in essentially the same manner as surface mount
devices since the predominant failure mechanism is also mechanical fatigue due to TCE
mismatches. The differences is that instead of the fatigue occurring in the solder joint, the fatigue
occurs in the Z-axis between the board material and PTH material. Therefore, for this situation the
number of cycles to failure model becomes:
-2.26
N |[a AT a AT + T
f = (T SZ( )" 2( RISE)]l)
where:
a =
SZ T n e Z axis TCE of the substrate
Table 4.7-9 summarizes the data set for PTH wearout. Detailed cycles-to-failure data was
available for a variety of conditions. This data was plotted on Weibull paper to derive the
characteristic life and (3. Contained in this table is the board thickness in mils, T 0 (-55°C), T§
(125°C), AT, TCE of the board, TCE of the PTH material, observed MCTF (Mean Cycles to
Failure), the characteristic life (Weibull a), Weibull shape parameter (P), the strain gauge
(excluding d, h), and the calculated value of X. This value of X was derived such that the
observed MCTF is equal to the predicted. The geometric mean of these values of X is .0061.
Therefore the predicted PTH wearout number of cycles to failure is:
-2.26
61
N
f(PTH) = ^ ° |(asz(AT)-a2(AT + TRISE))|
4-113
.0046
.0046
.0049
.0046
.0071
.0057
.0062
.0040
.0038
.0035
.0058
.0027
.0054
.0039
.0025
.0035
.013
.018
.019
X
.—.
^o
I N
IN
r-
<l
102
u ^ 369 CN
369
369
VO ^o ^o r- r» r- ^ H ^ H ^ H
V> V> V> CN r- r» r-
a o •<r •<r •<r V~l VI V~l V~l VI VI r- r- r» o o VI i n V~l
1 v_^
,—s
H
<l
v
~' t/1
Sa
'
' '
o o o o o o
VO V> VO VO V> V> VI VO VI o r- 00 ON VI VI
CO.
cn •<r V> CN CN CN cn •<r cn •<r •<r •<r CN cn cn ^r A CN cn
erisLic
les)
o o
e
oo IN o VI o o o o o o
o J ^
3
^ H
CN
^ H
00
IN
ON ^ H
V>
cn
CN
^ H
•<r o
O
^ H o
1
o
1
CN V>
V~l VO
O
cn CN CN
ON
V~l o
VI
^r VI
r •<r cn cn •<r ^ H ^ H V~l IN CN CN
A A
.c
s y.
u
"3 u,
£ f- cn cn cn o o o r» o o cn r» o
CN
^ H V~l 00
00
ON
^ H •<r
CN
VO ^o
IN
^ H
cn 1 1 CN r»
CN
o
VO
IN
M"
CN
00
CN
CN
CN
VI
^O
o 5 VI
^r ON
5 cn cn cn cn CN
a*
copper
TCE
r» r- r- r- r» r- r» r- r» r» r- r- r- r» r» r» r- r- r» r- r- r- r- r-
_
<*^, S CN CN cs ON ON ON
^_, ^_, ^_, r- r» r» f_^ ^_, ^^ •-i •-i
— ,_ ^_, ^^
a r- r-- r- o o o oo oo 00 t— r» r» ] 1 1 ON ON ON o o o oo oo oo
H ^ CN CN CN
o o o o o o o o o o o o o o o o o o o o o o o
S?
e-5
oo oo 00 00 oo oo 00 00 oo 00 00 00 00 00 oc oo oo oo 00 00 oo oo oc oc
VI V> V> V> i/n V> V> V> V~i i/n V> V> i/n V> V> i/n VI >T) VI •vn
r " CT CN CN CN CN CN CN CN
V~i
CN CN
V~i
CN CN CN CN CN
V~i
CN CN CN CN CN CN
V~i
CN cs CN CN
V> VI V> V~i V> V~i i/n V> VI V~i V~i V> i/n i/n VI V> VI i/n i/n VI »r^ in VI "vn
H°£ "? "? "? "? VI i/n
"? VI i/n
"? "? "? V~l VI
"? "? i/n i/n
^ ^ »r^ in
"? •vn
C/5
Thickne
(Mils)
Board
4-114
SMT Wearout Modelling
N
f(SMT) = Nb(SMT) (J^5h K a S ( A T ) " a C C ( A T + T
RISE)) I ] x 10 6
~
nas
N^ SMT ^een a(
lded as a replacement to the 1.32 constant to adjust the model in
accordance with the best data available. The 10"" factor has also been included to account for units
used. Table 4.7-10 summarizes the data used and includes d (in mils), h (in mils), AT, TCE of the
substrate (oc$), TCE of the ceramic package (OLQQ), observed mean cycles to failure (MCTF),
Weibull characteristic life (a), Weibull shape parameter (3, and the N^/gjyjj\ calculated such that
the observed MCTF equals the predicted for each data point.
The characteristic life differs from the mean cycles to failure primarily due to the fact that in
some cases there were large variances in the data, and the best fit Weibull line often yields a
characteristic life which differs from the true MCTF.
As can be seen from this data that there is a large degree of variation between the predicted
MCTF and the observed. Part of this variation is a result of the uncertainty in the TCE of both the
substrate and device and part is due to the inherent variation in the observed MCTF. As can be
seen in Table 4.7-10, there are several values of N^s^-p) that are significantly higher than the rest
of the population. Therefore, the model may be more sensitive to the input variables than is
indicated by the data. Since these outlier datapoints significantly increased the calculated N^/g^T)
value, they were discarded from the dataset and the geometric mean was calculated. This resulted
in a N ^ S M T ) value of 3.5, which will be used in the model. This effort also highlights the fact
that the model is extremely sensitive to the TCE values and suggests that, to obtain accurate results,
accurate data must be supplied.
-2.26
N 6
f(SMT) = 3-5 ^ | ( a s ( A T ) - a c c (AT + TRISE))|xl0-
To use the wearout modeling methodology proposed in this study, a representative Weibull
shape parameter (3 must be derived. The histograms in Figures 4.7-3 and 4.7-4 summarize the
distribution of observed [3's from the data presented previously for both Fl'H's and SMT's.
4-115
c
V~) OO Tf
O ^ \£) (S o\ \c t o \ h rV s^ C^Or O^ Oi r^^ iCr Oi O' ^- }o- —
o "t Cs O •—< m c-J r^ cs ^o•—'
^o
CO C-J •—< C-J •—' OO T f V~) C-J V~) C-J ( S O
g 00 M
r^i—«^DO^D - ' *n o oo (S h o co o V~J o C-J
^- ,-H i-H ,-H C N O ^ ^ ^ ^ - j ^ O -< O r n O r n
z <~>
a
XH o" oo" -rt"
S r~ i*- oo
M
cJ
I 1
SO
o
'^ "W
o
2 r- r- r~ r- r-
H \D \D \0 ^D ^D
«-« T f O
O —< T f r-« o
0 0o^ 3 X ^ 0 0 ^ 5 oO ^ Do
O ^ D Oo ^ D
H
P
o"
\*_y
\JS
' S
o § o o o o o S o o o o
Life
(a)
r^ — r^ r^ TT \c v~> ^ CO *T v-ir^
-c
a
U
r ^ o ^ t ^ c o o o ^ c ^ ^ r ^ O Q o o o o S r ^ ^ c ^ o o ^ MS c ^ r ^ r ^ ^ i r ) 0 ^ r ^ ^ c o ^ o o o o c o < N « - i C O
U T t h ( N ^ > n h ' n ^ o ^ r H O « i o \ n ^ H n o o o ^ i n ^ O h ^ ^ ^ ^ i n ( N T t ( N r N O N ^ r H
^ ( ^ ^ o ^ T f ^ o ^ n n ^ M ^ ( > o ^ c * o o ^ T f T f ^ ^ ^ n ^ n ^ < ! l n ^ > n ^ T ^ c n T t ^ c J ^
2
U
\ O ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ' » O ^ 0 ^ 0 ^ 0 ^ 0 ^ 0 ^ 0 ^ 0 ^ 0 \ 0 ' v O ^ 0 \ 0 ^ 0 ^ 0 ^ 0
I
u V~)V^V~)V~)V~)V~1 V~)V~)V~)V~)V~)lT)V~)V~) V~)ini/~)V~)V~)V~1V~)
o
c^<Nc^<Nr^r^t^r^r^r^<Ncsc^<Nc^c^c^r^r^r^r^r^r^^
v^v^i/^v^^J^^o^o^o^ov^v^inv^v^v^v^^o^^o^o^J^o
v - ) V - ) W 0 V - ) V - ) V ^ v - ) V - ) V - ) V - ) O O O O O O O O - — ' O O O O O O O O O O O O O O O O O O O
invr)vri»n^»n^»n^inooxooooxoooocooooociocoi>Dcoooc«coc«coxco
t/1
"1 cococococococococococococococococococococococococ^
VI O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O
"1 T f o ^ O T r c ^ O T f O T f ^ O T r o T r o T f O T t o ^ O T f o ^ r o ^ c ^ O T f o ^ ^ O T r c
t^^r^^t^^t^^t^^t^t^^t^^t^^or^^or^^or^^r^^or^^or^or^^or^^or^r^^or^^o
4-116
7
6-
<0
o
c 5
o
tl
3
o 4-
o
o 3-
.8 2-
B
3
2 1 -
7-
6-
g 5-
4~
o
3-
<0
2-
1 -
4-117
The mean value of PTH p's is 3.3 and the mean value for SMT p's is 3.7. The fact that
there is such a wide variation in values for a single manufacturing process indicates the variability
inherent in this modeling process. However, conservative p's of 3 for both cases will be used as
representative values.
The defect related failure rate term is modeled as a constant failure rate. For these failure
mechanisms, the screening effectiveness tends to be very high, indicating that a quality factor is
applicable. The model currently contained in MIL-HDBK-217E contains provisions for all
necessary model variables associated with early and mid life failures. It also indicates that there is
a linear relationship between failure rate and number of PTH's. Reference 67 presents data
indicating that the reject rate of both double sided and multilayer boards is directly proportional to
both the board area and the number of holes. This indicates that the number of defects are also
directly related to the number of holes. This observations lends an additional degree of confidence
in the current model to be used herein for modeling defects.
Since inadequate field data was collected during this study, the current model is used as a
baseline. The derivation methodology was to assume that a percentage of the current MIL-HDBK-
217 failure rate are actually failures accounted for by the wearout modeling discussed previously.
This percentage was derived by calculating a PTH wearout failure rate for a typical printed wiring
board used in a typical application. The parameters for this calculation is as follows:
4-118
(*5 years is used for the life cycle since it is the approximate time period over which the original
data was collected in support of the current MIL-HDBK-217 model).
DOA1 -2.26
Nf = f ^ - l [ 1 4 ( 3 0 ) - 17 (30 + 10)] \)
=
NfPTH 11,676 cycles
The expected cycling rate in the use environment is 360 cycles per year. Equating the MCTF
to mean-time-to-failure yields:
=
X] = .15 (using the table in Section 2.3 with P = 3 and LC/a = ^2 -2
(rounded up))
Therefore, an average of 42% f T^TQ | of the current models failure rate is accounted for in the
PTH wearout failure rate. The current models base failure rate is therefore scaled in accordance
with this percentage and, with this exception, is left largely intact. The primary assumption made
in this model is that the defect rates have not changed dramatically since the current model was
4-119
developed. While this may not be entirely true for conventional low complexity board types,
newer boards of higher complexity can have higher defect rates. There was no evidence however
to refute the fact that, on the average, board defect rates have stayed relatively constant. A
summary of the defect (PTH) model is as follows:
The failure rate model for plated through hole (PTH) assemblies is:
n + n + #
^p = ^b^O^E l ^C 2 (^C 13) DC (failures/10" calendar hours/assembly)
where:
Quality Grade *Q
4-120
TABLE 4.7-13:
COMPLEXITY FACTOR nc TABLE 4.7-14:
Number of Circuit Planes ENVIRONMENTAL MODE FACTORS
*C
<2 1 Environment ^E
3 1.3
4 1.5 GB 1
5 1.8 GF 2.0
6 2.0
7 2.2 G
M 7.0
8 2.4 N 13
S
9 2.6 5.0
10 2.7 %
11 2.9 A
IC 5.0
12 3.1 Ajp 8.0
13 3.2 A 16
14 3.4 uc
15 3.5 A
UF 28
16 3.7 sF .5
Discrete Wiring w/PTH 1 Mp 10
For greater than 16 circuit planes, ML 27
KC = .65 C 6 3 cL 500
4-121
4.8 ROTATING DEVICES
Motors: • Induction
• Direct current
• Single-phase
• Poly-phase
Generators: • Single-phase
• Poly-phase
• Externally excited
• Internally excited
The devices are generic categories of rotating devices. Within each category there are a
variety of device styles and types which have specific operating characteristics for a given
application. For example, the use of poly-phase motors has the widest general application of any
type of motor because of its characteristics of good speed regulation and high starting torque.
More importantly the simplicity of the poly-phase motor construction results in less maintenance
and higher reliability.
The life limiting components affecting the failure rate of rotating devices are bearings,
windings and brushes. The primary failure accelerating stress acting on these components is
temperature. Sources of the damaging temperature are the environment and the load requirements
of the driven device in the case of a motor, or the required electrical load in the case of a generator.
Temperature cycling stresses degrade the insulation material on the field windings and armature
windings resulting in the reduction of magnetic efficiency and increase of temperature rise.
Temperature affects the viscosity of the lubrication necessary for long bearing life. As temperature
cycling occurs at an increasing rate the reliability of the bearings will decrease. Brush wear
4-122
increases as a function of armature speed, temperature, and electrical power transfer which is the
most dominant of these stresses.
There are several manufacturing procedures which must be monitored to ensure an efficient
and reliable rotating device. Bearing alignment, and armature and field (or permanent magnet)
matching is critical to the efficiency because of the lines of flux being cut at precise distances
through the rotational area. Clearances between the armature and fields correlate to the efficiency
of the rotating device. The closer the tolerance, the more efficiently the flux lines are cut resulting
in the higher output levels. Misalignment of the bearings or non-parallelism of the armature and
fields can cause internal heat build-up amplified with additional load requirements and resulting in
acceleration of the degradation process.
Device variations for rotating devices are based on the load requirements. The design
variations which primarily affect reliability are complexity and size. Full horsepower vs. fractional
horsepower motors require a completely different approach to design. Full horsepower motors,
designed for higher loads, tend to experience additional bearing loads and generate more internal
heat. Complexity of the rotating devices directly affects reliability. Motors needing assistance in
initial start-up (including capacitive start motors) are more complex and have a higher failure rate.
DC or AC rotating devices with brushes have additional design complexities which affect failure
rate.
If properly designed, rotating devices are selected for specific applications and should
provide reliable service. There are, however, application variables which do have a negative effect
on reliability. On-off cycling or cyclic loads create internal heat generation resulting in accelerated
degradation of starting components and windings. Environmental effects of contamination and
ambient temperature including temperature cycling also have a negative effect on reliability.
The primary failure mechanisms for all types of motors are a function of the electrical or
mechanical stresses that the windings and bearings experience. Windings experience degradation
of their insulation and hence their ability to produce a sufficient magnetic field. The primary
accelerating factor for insulation degradation is temperature. More specifically, the temperature rise
4-123
in the winding during motor operation. According to Reference 80, "If two motors are running
with a 10°C differential in temperature, the hotter motor's service life expectancy is reduced by
one-half."
The class of insulation (A through F) designates the operating temperature limit the insulation
can operate at and still maintain its integrity. Reference 80 indicates, "A motor operating within
Class B temperature limitations and having a Class F insulation system that has a higher
temperature rating is operating below its temperature rating. The cooler motor's insulation will be
subject to a much lower degradation than that of the hotter running motor and will experience a
longer life." Therefore, the conclusion derived from information collected is that the primary
accelerating factor for windings in motors is ambient temperature and temperature rise. This is
entirely consistent with the current MIL-HDBK-217E model.
Bearing failure mechanisms, such as galling or branell hardening are caused by the lack of
lubrication. Lubrication loss can be traced to two operating characteristics, load and speed. These
characteristics generate heat which increases the failure acceleration process. Load and speed
influence reliability, but are normally designed for a specific application. Temperature again is the
primary failure accelerating stress which results in the loss of the protective film on the bearing
surface. Most susceptible to this occurrence are motors with heavy loads requiring frequent starts
and stops. As stated by Lincoln manufacturing, "Bearings fail primarily because of heat.
Contamination from a minute particle of dust, dirt or even cigarette ash will cause the bearing to
run hot enough to melt the grease that will then run out... grease that is moisture resistant and has
a operating range from -35° to 350°F ... and bearing sized for 40 to 50,000 hours of life is the
standard design criteria for most motors."
In summary, temperature, reducing the motor life by as much as 1/2 per 10°C rise, is the
dominant accelerating factor for motors.
Shaker Research (Reference 79) had developed the current MIL-HDBK-217E model. In that
study the failure data collected are predominantly comprised of life test results. It was analyzed by
means of a Weibull cumulative distribution analysis of each individual test population. The results
provided a linear regression best fit Weibull slope and characteristic life for each test group of
motors. Additional regression techniques are applied to determine the influence of parameters such
as temperature, speed, bearing lubricant, motor type, etc., on characteristic life.
4-124
The current model considers bearing and winding to be the dominant factors in motor failure.
These failure mechanisms are predominantly accelerated by temperature. The data collected during
this study indicates that there are three major failure modes, they are:
It is apparent that bearing failures are the dominant failure mode. This finding also explains
why the Reference 79 model emphasizes the bearings and windings only for their model.
Although temperature is the primary failure accelerating stress, additional variables include:
bearing size, quality code and grease type. Among these variables the most dominant is grease.
Additional observations from review of the current MLL-HDBK-217 model are as follows:
(1) Full Horsepower (FLHP) rotating devices should be considered as an addition to the present
reliability model. Brushes, as an additional failure mechanism, should also be considered.
(2) When considering FLHP motors, a distinction must be made based on the loading
characteristics and power consumption affecting temperature life limiting characteristics.
(3) Technology has changed in the form of newer materials, resulting in increased efficiency of
rotating devices. These changes should be accounted for in the models. These newer
materials include:
(4) A major flaw in the current 217 motor model is that it uses a hazard rate for the failure rate.
This is accurate if the total cumulative percent fail of a given population, for a given life
cycle, is relatively low. If it is not, then it is very inaccurate (and pessimistic) since the
hazard rate provides the instantaneous failure rate on the condition that the part has not yet
failed. This results in predicted failure rates approaching infinity where in reality it reaches
4-125
an asymptotic value. Figure 4.8-1 illustrates this concept. The point the two curves begin to
depart are approximately at a time equal to one a.
Old Model
X proportional to t
X
Proposed Model (Asymptote
proportional to — )
Time
FIGURE 4.8-1:
FAILURE RATE FOR NEW AND EXISTING MOTOR MODEL
Since both bearing and winding failures are normally wearout failures, they will be modeled
in accordance with the methodology outlined in Section 2.3. The hypothesized model is therefore:
Xr + a
VttBPy v WPy
XQQ = Cumulative average failure rate for bearings as a function of LC/a and P
a =
BP Weibull characteristic life predicted for bearings
= c^BB^L^HP^R
4-126
a =
BB ^ase characteristic life, function of generic motor type
=
^"CW Cumulative average failure rate for windings as a function of LC/a and (3
a =
WP Weibull characteristic life predicted for windings
Kj = Temperature factor
=
7tL Load (mechanical) factor
The collected motor data was analyzed in an attempt to quantify the motor life times and
failure rates as a function of the parameters outlined in the hypothesized model. Unfortunately, the
effects of actual mechanical load stress, rated horsepower, and rotation rate factor could not be
quantified due to the fact that these quantities were not known for most of the observed data points.
The bearing characteristic life (a) and failure rate is therefore a function of only generic motor type
and operating temperature.
The model developed in Reference 79 was based on thorough research and a good set of
data and therefore the temperature dependence of the model should be accurate. The approach
therefore was to use the current base failure rate as a function of temperature and scale the model
for each generic type of motor for which data existed.
4-127
The following items summarize the assumptions and methodologies used:
- The LC/a ratio was assumed to be <.l for commercial data (since it is from 1st year warranty
and the fact that the observed failure rates were low). In this case the LC is the time period
over which the data is collected.
The LC/a ratio was assumed to be >2 for military data since it is generally data from systems
that have been fielded for years and the fact that the observed failure rates are generally high.
This assumes the failure rate has reached its asymptotic value (see Section 2.3).
- The calculations assume that 20% of the observed motor failures are due to windings and 80%
bearings.
The observed (5 values from Reference 79 are generally between 2 and 3. A value of 3 will be
used in this model.
Table 4.8-1 summarizes the data and analysis for motors. The a was calculated in the
following manner:
h.
*°bs = a
a =
^obs
A. j = Cumulative average failure rate over time period from which data was taken
(from table in Section 2.3)
4-128
;-?
alio)
fl UJ Tf i^ Tt t~ Os \n Os rsi <N
XI OS SO SO r^ SO 00 O u-i Tt r^ SO Tt en en en r» Tt en
OS
o 5 (S so r^
^- •*t
\n Tt
Tt r^ " ^* Tt
ts
^™ OS SO r^ ^- 00
ts ^ \n
Tt ts SO CN
rsi r^ ^- Os
"8 "8 ?J
Observ
Predict
w
XI
SJ
r^
u-i
t~ t~
SO o
SO en OS en
Os Tt
en o o o
Os
u-i \n ^t en Os en SO
UJ
t~ C<1 Os r^ Tt r^ u-i o ts 00 >n r^ •^ " r^ ts Tt Tt ^H »n
^ FS SO
•—
"N en en ts so r^
«B(
8 " 00
SO SO so so SO
O so o © o o o o o o o
00 o o o o o"
008
590
048
32, 360
O O o o so o r^ O
o oo ^ X 00 00 X Os o
r^ en en X so o
o X "l X
40,
'££
158
344
416
325
625
377
235
X
366
00 00 <s" 00
sq en so
"2 ^ o(S
en tN
i '8 ^" ^-
"3 u
Calc
Bas
SO so
,300
156 ,250
304 ,800
518 ,000
294 ,980
obs)
909 ,000
39, 840
94, 340
86, 207
91, 740
45, 200
O
300
200
59, 000
o
en (S 00 <=> 2
X Os en
00* \n o" o 00 en r^
00
8 en en o Os so
so
r^ so 00
t~ 00 Tt (S Os r^ u-i en Tt r^ r^ en
so .5 o OS o
OS Tt CN u-i u-i o ^- 00 tN SO r^ 00 Tt o
SO ts sO O
O o
o Tt Tt u-i en 00
o
en ts (S ts
"5 <•* en
(S
(S 00 ts
so (S ^H
i | 3
CO <-<
•a
u
Obs erv
O 8 X X X X X X X X X X X X X X X X X X X X
H .— en OS o o 00 en .— en o Os ^H en o en o .— q en Os
O •-< (S 00 \ri ir\ 00 ts — ts \ri 00
•"" ts >o ts \ri — \ri oi oo
"5 G
'SB
y 3
gu o o o o o o c o o o o o o o o o o o o o
Cha
from
o o o o o o o o o o
00 CO t~ 00 en (S <N o
en 00 o 00 o
(S o
en r^ o
00 o
CN 00 o
(N o o en
r^ ts oo o
•s* 8 en 00 r^ en
(~ en en o en en oo
^H
(S rt r^
o
00 o
00 Tt
^H
(S Tt
o 00
00 r^
^H
ts Tt 00 Tt
o
00 ts o
T—
00 •* r^
W
m
< ci. u-i o o o u-i \n u-i o o \n o u-i o o \n o
E t~ \n en Tt•*t t~- u-i
p u-i Tt en r^ u-i Tt u-i Tt
f~ Tt u-i en
H1>
1150
457
785
985
702
en
00
so o so Tt OS vn SO o
SO
00 ts O ts
en * s
u-i rt Os u-i Os SO \n o
00 OS 00 (S SO Tt Tt r^ Tt m OS en ^t en in in \n o
o SO OS u-i r
r^
o ^- (S
^ en OS ^- en u-i o 00 TT t~- Tt en Os
\ri Tt Tt
OS
OS 00
00
SO r-< Tt ts ts 00
in
en ,—< SO
en
^*
nvironment
£ > >
&S a s a s
< CO u. t/1 D < < O,
D s D ^ u. LU D ^ u. s on D o, 2 co
< < O o z Z o < < o CO < < o o z < O o o
w
Motor Type
Electric
Stepper
1 Sensor
Servo
4-129
The ^- 0 b s A (217E) ratio was also calculated and is summarized in Table 4.8-1. The
geometric means of this ratio as a function of motor type and failure mode (bearings, windings) is
given in Table 4.8-2:
a(observed)
TABLE 4.8-2:
a (217E)
These values can therefore be used as multipliers to adjust the current 217E model a's in
accordance with observed field data and as a function of motor type.
The next analysis conducted on motors was in an attempt to determine the relationship
between failure rate and horsepower rating. For this analysis, data was extracted from the same
generic environment (Ground), in an attempt to minimize uncontrolled variables. The data in
Figure 4.8-2 summarizes this data.
i i
0 - •
9 - •
8 -
7 - m
6 -
X(F/10°)
5 -
4 -
3 -
• •
2 - •
1 -i •
•
i i i i i i i i i i i i i i i i i i i *
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Rated Horse Power
FIGURE 4.8-2:
FAILURE RATES VS. HORSE POWER RATING
4-130
The arrows in this figure are indicative of datapoints with zero failures. For these, one
failure was assumed to establish an upper bound on the failure rate. This graph indicates that a
horse power rating cannot be derived from this dataset, and therefore will not be included in the
model.
( XA Xn ^
1 2
\ + 1 (x 106) (F/106)
AocB Baw
where:
X^ is a function of Design Life Cycle (operating hours) and characteristic life for bearings and is
summarized in Table 4.8-3 (ag must be calculated first)
^2 is a function of Design Life Cycle and characteristic life for windings and is in Table 4.8-3
(a^/ must be calculated first)
4-131
TABLE 4.8-3:
CUMULATIVE AVERAGE FAILURE RATE
LC/ocB, L C / a w M'^2
0-.10 .13
.11-.20 .15
.21-.30 .23
.31-.40 .31
.41-.50 .41
.51-.60 .51
.61-.70 .61
.71-.80 .68
.81-.90 .76
.91-1.0 .82
>1.0 1.0
TABLE 4.8-4:
A,B CONSTANTS
Motor Type A B
4-132
TABLE 4.8-5:
BEARING & WINDING CHARACTERISTICS
LIFE, a B & a w , vs. AMBIENT TEMPERATURE, T
2357
1.83
T+273
**a w = 10
4-133
5.0 MODEL SUMMARY AND SAMPLE CALCULATIONS
This section of the report summarizes the complete models being proposed for inclusion
into MIL-HDBK-217.
5-1
CAPACITORS
^p = ^ C / E ^ T W ^ S R
5-2
TEMPERATURE (TCT), CAPACITANCE ( K C ) , VOLTAGE (Ky), AND
SERIES RESISTANCE (rcSR) FACTORS
Paper . . 1 ^r~r~f\( A * 1 1
e x ^ - 2 5 5 0 ( T A + 2 7 3 ) " 298
1"
c.09
(r-
Plastic ....
exp
[-
OCCfV
2550(
*•
TA+273 )
\
" 298
c.09
®6-«
1" 0
Mica, Glass . ,1 dOAA^
exp^-4290( TA+273 ) " 298
v
> c.09 csV
u) +1,
G)3*'
1"
Ceramic •i 11 i n i i v ^ c.09
CXP
[- 3940(
TA+273 ) " 298
1"
Ceramic Chip I on '\f\( ^i c.09
exp
[- 3940(
TA+273 ) " 298 ( ! ) 3
-
1"
Al Electrolytic
1 r~r\ i e-/ * \ c.23
exp^-5215( T A + 2 7 3 ) " 298 (!)5+'
Solid
1 "
c.23
Ta Electrolytic, i 11 oonr^f
exp^-2200( TA+273 ) " 298
^
®"« 1
(Non-Solid)
1"
Tantalum Chip, , Ll ,-J OOA/V
exr|-2200( T A + 2 7 3 ) " 298
\ c.23
(I)' 7 - ^SR
(Solid)
1"
Variable, Air -,, ,J oonn/ \ c.09 1
exr|-2900( T A + 2 7 3 ) " 298 (!)3+'
11
Variable, Ceramic ,. ,-j on i n / \
exr|-3940( T A + 2 7 3 ) " 298
c.09
©'- 1
5-3
CAPACITORS (CONT'D)
Quality TCQ
D .001
C .01
S,B .03
R .1
P .3
M 1
L 3
NonER 3
Lower 10
Environment 7t E
GB 1
GF 10
G
M 20
N
S 7
% 15
A
IC 12
Ajp 15
A
UC 25
A
UF 30
A
RW 40
sF .5
MF 20
ML 50
cL 570
CAPACITORS (CONT'D')
>0.8 .66
0 to 0.1 3.3
5-5
RESISTORS
^p = V o ^ E ^ p
Varistor .0023 1
Variable Non-
Wirewound MIL-R-94 .0037 exp ">Cf,C\( * \
-^660( T + 2 ? 3 - 298)
MIL-R-23285
T = Resistor operating
Temp = T A + Gj^P
5-6
RESISTORS (CONT'D')
Quality *Q
S .03
R .1
P .3
M 1
Lower 10
Environment n
E
GB 1
GF 4.0
G
M 16
A
IC 18
A 31
uc
A 1F 23
A
UF 43
A
RW 63
NTJ 42
NS 12
ML 87
MF 37
cL 1728
sF .5
5-7
TRANSFORMERS
Specification Description
^p = ^b^Q^E^T
Switching .00057
Flyback .0054
Audio .0137
Power .0486
RF .133
Quality ^Q
MIL-Spec. 1
Lower 3
TRANSFORMERS (CONT'D)
ENVIRONMENT - rcE
Environment K
E
GB 1.0
GF 6.0
G
M 12
Ns 5.0
16
A
IC 6.0
AIF 8.0
A
UC 7.0
A
UF 9.0
A
RW 24
.50
Mp 13
ML 34
cL 610
1 1
K-p = exp -12750Tpjs+273 298
5-9
INDUCTORS
Specification Description
Ap = XbKQKEKT
Quality XQ
MIL-Spec. 1
Lower 3
5-10
INDUCTORS (CONT'D)
ENVIRONMENT - nE
Environment *E
GB 1.0
GF 6.0
G
M 12
Ns 5.0
Nu 16
A
1C 6.0
A
IF 8.0
A
UC 7.0
A
UF 9.0
A
RW 24
.50
MF 13
ML 34
cL 610
5-11
Hot Spot temperature can be estimated as follows:
T H S = T A + U(AT)
where:
T HS Hot Spot Temperature (°C)
Inductive Device Ambient Operating Temperature (°C)
TA
AT Average Temperature Rise Above Ambient (°C)
DT can either be determined by the appropriate "Temperature Rise" Test Method paragraph in the
device base specification (e.g., paragraph 4.8.12 for MIL-T-27E), or by approximation using one
of the procedures described below.
AT Approximation
Information Known AT Approximation
NOTE: Methods are listed in preferred order (i.e., most to least accurate). MIL-C-39010 are
microminiature devices with surface areas less than 1 in2. Equations 2-4 are applicable to devices
with surface areas from 3 in2 to 150 in2. Do not include the mounting surface when determining
radiating surface area.
Xp = V Q 7 t E 7 t c + ^U
BASE FAILURE RATE - Xb
Rocker .023
Slide .003
Centrifugal 3.4
Microwave (Waveguide) 1.7
Liquid Level 2.3
Rotary MIL-S-3786 .11
MIL-S-15743
MIL-S-21604
MIL-S-22710
5-13
SWITCHES (CONT'D^)
Quality k>
MIL-Spec. 1
Lower 2
ENVIRONMENT - rcE
Environment *E
GB 1.0
GF 3.0
G
M 18
Ns 8.0
29
A
IC 10
A
IF 18
A
UC 13
A
UF 22
A
RW 46
SF .50
Mp 25
ML 67
cL 1200
5-14
SWITCHES (CONT'D)
KC = (N c )-33
N C =- Number of Contacts
Ex: SPST = 1
DPDT = 4
3PST = 3
L
U
ar
X-[ = Cumulative average base failure rate over the life cycle (LC) time (desired life
expectancy or preventative maintenance interval) as a function of a
a _ a
c a I SR
SR = Switching rate in actuations per 10" calendar hours (necessary to convert a to a time
scale)
5-15
SWITCHES (CONT'D)
Contact Current
Rating (Amps) a a (AC Resistive Load) a a (DC Load)
29.08 26.323
0-4
y.75^.14 v 1 . 3 3 I .3e130L/R
1
103.45 123.187
>4-8
y.75^.14 v1.33T1.3e130LR
219.74 307.94
>8
V.75J1.14 v 1 . 3 3 jl.3 e 1 3 0 L / R
L = Load inductance
R = Load resistance
5-16
SWITCHES (CONT'D)
LC
*1
0-.1 .13
.11-.20 .15
.21-.30 .23
.31-.40 .31
.41-.50 .41
.51-.60 .51
.61-.70 .61
.71-.80 .68
.81-.90 .76
>.9 1.0
5-17
CIRCUIT BREAKERS
APPLICABLE SPECIFICATIONS
MIL-C-55629
MIL-C-83383
MIL-C-39018
MS-24510
MS-25244
Magnetic .34
Thermal .34
Quality n
Q
MIL-Spec. 1.0
Lower 8.4
5-18
CIRCUIT BREAKERS (CONT'D)
ENVIRONMENT - 7tE
Environment ^E
GB 1.0
GF 2.0
G
M 15
Ns 8.0
Nu 27
A
IC 7.0
A
IF 9.0
A
UC 11
A
UF 12
A
RW 46
S
F .50
MF 25
ML 66
N/A
Configuration *C
SPST 1.0
DPST 2.0
3PST 3.0
4PST 4.0
5-19
THERMAL SWITCHES
Specifications
MIL-S-12285
MIL-S-24236
Quality *Q
Military l
Lower 2
ENVIRONMENT - nE
Environment *E
GB 1.0
GF 3.0
G
M 18
Ns 8.0
Nu 29
A
IC 10
Arp 18
A
UC 13
A
UF 22
A
RW 46
Sp .50
Mp 25
ML 67
cL 1200
5-20
RELAYS. ELECTROMECHANICAL
Specifications
MIL-R-27745
MIL-R-39016
MIL-R-5757
MIL-R-6106
MIL-R-83726
+
^p = ^b^Q^E ^U
Quality K
Q
MIL-Spec. 1
Lower 1.9
5-21
RELAYS. ELECTROMECHANICAL (CONT'D)
ENVIRONMENT - nE
Environment *E
GB 1
GF 8.3
G
M 64
A
IC 168
A
UC 264
Ajp 216
A
UF 288
A
RW 833
Nu 27
NS 8.2
ML 1584
MF 600
cL N/A
Sp .82
=
^U Wearout failure rate due to relay utilization.
Xi = Cumulative average base failure rate over the life cycle time (desired life expectancy or
preventative maintenance interval) as a function of a c
a = a
c a (§RJ
a =
a Weibull Characteristic life (in 10" actuations) as a function of load
SR = Switching rate in actuations per 10" hours, (necessary to convert a to a time scale)
5-22
RELAYS. ELECTROMECHANICAL f CONT'D!
Contact Current
Rating (Amps) a a (AC Resistive Load) a a (DC Load)
29.08 26.323
0-4
V.75J1.14 vl.SSjl.SeBOL/R
103.45 123.187
>4-8
y.75^.14 v 1 . 3 3 ! l . 3 e 1 3 0 LR
219.74 307.94
>8 V-75 jl.14 v1.33 jl.3e130L/R
5-23
RELAYS. ELECTROMECHANICAL (CONT'D)
LC
ac
h
0-.1 .13
.11-.20 .15
.21-.30 .23
.31-.40 .31
.41-.50 .41
.51-.60 .51
.61-.70 .61
.71-.80 .68
.81-.90 .76
>.9 1.0
5-24
RELAYS. SOLID STATE
Specifications
MIL-R-28750
Quality K
Q
MIL-Spec. l
Lower 1.9
Environment *E
GB 1.0
GF 3.0
G
M 12
Ns 6.0
NTJ 17
A
IC 12
A
IF 19
A
UC 21
A
UF 32
A
RW 23
sF .40
Mp 12
ML 33
cL 590
5-25
CONNECTORS
SPECIFICATIONS
CONNECTORS
MIL-C-21097 MIL-C-21907
MIL-C-22857 MIL-C-23353
MIL-C-24308 MIL-C-26482
MIL-C-28748 MIL-C-3643
MIL-C-3767 MIL-C-38999
MIL-C-39012 MIL-C-39024
MIL-C-5015 MIL-C-55302
MIL-C-81511 MIL-C-83723
MIL-C-83733
^p = ^ b ^ E ^ T ^ K
Signal .0000044
Rectangular .046
Elastomeric .0071
Edge Card .040
Cylindrical .0010
RF .00041
Hexagonal .146
Rack and Panel .021
D-Subminiature .66
Telephone .0075
5-26
CONNECTORS (CONT'D)
Quality *Q
MIL-Spec. l
Lower 2
Environment *E
GB 1.0
GF 1.0
G
M 8.0
NS 5.0
Nu 13
A
IC 3.0
AIF 5.0
A
UC 8.0
A
UF 12
A
RW 19
Sp .50
Mp 10
ML 27
cL 490
5-27
CONNECTORS (CONT'D)
RF Coaxial Connectors
(High Power Applications) AT = 50°C
0 to .05 1.0
> .05 to .5 1.5
> .5 to 5 2.0
> 5 to 50 3.0
>50 4.0
5-28
SOCKETS
Specifications
MIL-S-83734
MS-25328
MS-27400
?ip = ?ib7tQ7tE(F/106hrs.)
DIP .00064
Chip Carrier .0024
Pin Grid Array .014
SIP .0030
Relay .037
Transistor .0051
Tube .011
QUALITY FACTOR - KQ
Quality %
Q
MIL-Spec. .3
Lower 1
5-29
SOCKETS (CONT'D)
ENVIRONMENT FACTOR - Kg
Environment MIL-SPEC
GB 1.0
GF 1.0
G
M 8.0
NS 5.0
Nu 13
A
IC 3.0
A
1F 5.0
A
UC 8.0
A
UF 12
A
RW 19
sF .50
Mp 10
ML 27
490
5-30
CONNECTIONS
DESCRIPTION
Connections Used on All Assemblies Except Those
Using Plated Through Holes (PTH) or Surface
Mounted Technology (SMTs)
APPLICATION NOTE: The failure rate model in this section applies to connections used on all assemblies
except those using plated through holes or Surface Mounted Technology. Use the Interconnection Assembly Model
to account for connections to a circuit board using PTH or SMT. The failure rate of the structure which supports
the connections and parts, e.g., non-plated-through hole boards and terminal straps, is considered to be zero.
Solderless wrap connections are characterized by solid wire wrapped under tension around a post, whereas hand
soldering with wrapping does not depend on a tension induced connection.
n
X = ^-jjito^E Fai'ures/lO Hours
n = number of connections
5-31
INTERCONNECTION ASSEMBLIES WITH PLATED THROUGH
HOLES AND/OR SURFACE MOUNT CONNECTIONS
=
^p ^SMT + Vmi +
^PTH2
=
^•SMT Average failure rate over the expected equipment life cycle due to surface mount
device wearout. This failure rate may be calculated only for the Surface Mount
Device exhibiting the highest value of the strain range;
=
^•PTH1 Average failure rate over the expected equipment life cycle due to plated through
hole wearout (F/10" hrs.)
=
^•PTH2 Failure rate from PTH defects (F/10" hrs.)
^SMT a
SMT
.-2.26" K
6 LC
a SMT 3.5 ^ | (a s AT - a c c ( A T + T R I S E ) ) | x 10" CR
where:
h = Solder joint height for leadless devices, use h=8 for compliant lead configurations
AT = Environmental AT
=
TRISE Temperature rise due to power dissipation = 9j£ P
5-32
INTERCONNECTION ASSEMBLIES WITH PLATED THROUGH
HOLES AND/OR SURFACE MOUNT CONNECTIONS (CONT'D)
\\ = Cumulative average base failure rate over the life cycle time (desired life
expectancy or preventative maintenance interval) as a function of a. This value
is:
LC
a
*1
SMT
0-.1 .13
.11-.20 .15
.21-.30 .23
.31-.40 .31
.41-.50 .41
.51-.60 .51
.61-.70 .61
.71-.80 .68
.81-.90 .76
>.9 1.0
Leadless 1
S Lead 150
Gull Wing 5,000
5-33
INTERCONNECTION ASSEMBLIES WITH PLATED THROUGH
HOLES AND/OR SURFACE MOUNT CONNECTIONS (CONT'D)
X
L
PTH 1 a
PTH
n-2.26
.0061 | ( a ( A T ) - a ( A T + T
a
PTH sz 2 RISE))
CR
where:
*1 Cumulative average base failure rate over the life cycle time (desired life
expectancy or preventative maintenance interval) as a function of a. This value
is as follows:
LC
a
*1
PTH
0-.1 .13
.11-.20 .15
.21-.30 .23
.31-.40 .31
.41-.50 .41
.51-.60 .51
.61-.70 .61
.71-.80 .68
.81-.90 .76
>.9 1.0
5-34
INTERCONNECTION ASSEMBLIES WITH PLATED THROUGH
HOLES AND/OR SURFACE MOUNT CONNECTIONS (CONT'D)
The failure rate model for plated through holes (PTH) assemblies is:
^PTH2 = ^b^O^E I n l ^C + n
2 ( K C "+- 13)1 • DC (failures/10"calendar hours/assembly)
where:
^E Environment factor
n
l Quantity of wave soldered functional PTH's
n
2 Quantity of hand soldered PTH's
^C Complexity factor
DC Duty cycle, % of calendar time the circuit is operating
5-35
INTERCONNECTION ASSEMBLIES WITH PLATED THROUGH
HOLES AND/OR SURFACE MOUNT CONNECTIONS (CONT'D)
5-36
INTERCONNECTION ASSEMBLIES WITH PLATED THROUGH
HOLES AND/OR SURFACE MOUNT CONNECTIONS (CONT'D)
If actual values of a$, OLQQ, AT, or CR cannot be determined use the following:
AT VALUES
Env. AT
GB 7
GF 26
G
M 11
A
IC 31
A
UC 57
A
IF 31
A
UF 57
A
RW 31
Nu 61
N
S 26
ML 31
MF 31
cL 26
a s VALUES
Substrate Material «S
FR-4 Laminate 18
FR-4 MLB 20
FR-4 MLB w/Copper Clad Invar 11.3
Ceramic MLB 7.15
Copper Clad Invar 5.1
Copper Clad Molybdenum 5
Carbon-Fiber/Epoxy Composite .75
Kevlar Fiber -3
Quartz Fiber .54
Glass Fiber 4.5
Epoxy/Glass Laminate 15.17
Polyimid/Glass Laminate 13.25
Polyimid/Kevlar Laminate 5.5
Polyimid/Quartz Laminate 7.8
Epoxy/Kevlar Laminate 6.75
Aluminum (Ceramic) 6.5
Epoxy Aramid Fiber 7
Polyimid Aramid Fiber 5.75
Epoxy-Quartz 9
Fiberglass Teflon Laminates 20
Porcelainized Copper Clad Invar 6.5
Fiberglass Ceramid Fiber 6.5
5-37
INTERCONNECTION ASSEMBLIES WITH PLATED THROUGH
HOLES AND/OR SURFACE MOUNT CONNECTIONS f CONT'D)
Plastic 6.5
Ceramic 5.6
————^———————
Equipment Type Number of Cycles per 10" hrs.
Consumer 4200
Computers 170,000
Telecommunications 4200
Commercial Aircraft 340,000
Industrial 21,000
Military Ground Applications 30,000
Military Aircraft 115,000
5-38
ROTATING DEVICES. ELECTRIC MOTORS
(x 10 6 )
A aB B a w 10 6 hrs
LC
*1
aB
0-.10 .13
.11-.20 .15
.21-.30 .23
.31-.40 .31
.41-.50 .41
.51-.60 .51
.61-.70 .61
.71-.80 .68
.81-.90 .76
>1.0 1.0
5-39
Design life cycle of the equipment in which the motor is operating (or preventative
maintenance interval) divided by the winding characteristic life (a-^y)
LC
X2
aw
0-.10 .13
.11-.20 .15
.21-.30 .23
.31-.40 .31
.41-.50 .41
.51-.60 .51
.61-.70 .61
.71-.80 .68
.81-.90 .76
>1.0 1.0
5-40
ROTATING DEVICES. ELECTRIC MOTORS (CONT'D)
Motor Type A B
T ccB* oc w ** T ccB* cc w *
(°C.) (Hr.) (Hr.) (°C.) (Hr.) (Hr.)
C
B = { l O ^ 2 - 5 3 4 - ! ^ ^ ^ ! / 10(20"T+273) + 300j }
135L - 1.83
T+273
** = 10
«vv
where T is ambient temperature in °C.
5-41
5.2 SAMPLE CALCULATIONS
Capacitors
a, b 7tQ7t E 7t T 7C C 7c V TC SR
.0004 (F/106)
1
10
1 1
exp " 2 2 0 0 ['35 + 273 " 298 1.27
7t c = (100)- 2 3 =2.88
>50 17
100
Tty + 1 = 1.045
7t SR = 1.3
5-42
Resistors
Conditions: Fixed resistor network (MIL-R-83401)
Mil. quality M
Ground Benign Environment
Power rating (per resistor) = .25W
= Xb7lQ7CE7tT7Cp
^P
^b = .0019 (F/106)
= 1
*Q
*£ = 1
7ly = 1
Tip = (.25)- 39 = .58
Transformers
XbKQKEKj
^P
^b .0137 (F/106)
3
*E 8.0
1 1
7tj = exp -1275(:56.5 + 273 298
= 1.5
( T H S = T A + 1.1(AT) = 56.5)
5-43
Toggle switch, 5 amp rating
Mil. quality
Gp environment
DPDT configuration
Design life (LC) for the equipment in which the switch is
operating = .1752 x 10" hrs. (20 years) (no preventive maintenance)
AC resistive load, 24 volts, 2 amps.
Switching rate (SR) = 100,000 per 106 calendar hours
Xb7tQ7rE7tc + ^u
.102
1
3.0
(4)- 33 = 1.58
cc c
a
a[sR
= 103.45 = 103.45 , .. in6
—T<—TTT — T <1— T1 T, iJ4 = 4-33 x 10u actuations
y . / 5 jl.14 24-'- 2
10 6 hrs.
= .004
— = ^ t = .003 F/10 6
ac 43.3 (10 6 hrs.)
5-44
Circuit Breakers
Conditions: - Magnetic type
Mil. quality
A\JQ environment
SPST configuration
^p = hnQnE*C
7C E = 11
7C C = 1
Thermal Switches
rcQ = 1
TCE = .5
5-45
Relays
Conditions: - General purpose electromagnetic relay (2 amp. Rating)
Commercial quality
Gg environment
- Equipment design life (LC) = 5 years = .0438 x 10" hrs.
- AC resistive load, 120 Volts, 1.5 amps applied
Switching rate = 10 x 10" actuations per 10" hrs.
^p = V ^ E + ^u
Xb = .020 F/10 6
T: Q = 1.9
7l E = 1
""C
a a
c - a ^SR
a 364 l o 6 hr 0364 ! 6 hr
«c•c =
" "aa (<n?l
SR = ~ -- ^ v( i u "-)W(jfi\
ho = - ( ° -)
LC _ .0438 (lO 6 )
ac .0364 (lO 6 )
1.2
M 10 <
lu = = -~-r- = 27.5 F/10 6
6
ac .0364 (lO )
X? = (.020)(1.9)(1) +27.5 = 27.54 F/10 6
5-46
Relav. Solid State
Conditions: Solid state relay (MIL-R-28750)
Mil. spec.
; Ajjp env.
7CQ = 1
TCE = 32
K = ^ Q ^ T ^ K
Xb = .040 (F/10 6 )
*Q = 1
K
E = 1
exp lei?/ 1
7l r r — " 1 D i : ) l T 0 + 273 " 298j
l
exp r i^^r
L" " 35 + 273 ' 298 J. = 1.19
16 5
(To = T
A + AT
= 35°C + . 6 4 ( . 0 5 ) L 8 5 - 35°c)
TCK = 1.5
5-47
Sockets
Conditions: 144 pin grid array isocket
Socket commercial quality
Gg environment
=
^p V^E
.014 F/10 6
= 1
= 1
Connections
K = 350Xb7iQ7tE
Xb = .0000068 F/10 6
= 1
*E = 6.0
Interconnect Assemblies
K = ^SMT + ^PHTl +
VrH2
5-48
^SMT
Since all surface mounted devices are plastic encapsulated, the one exhibiting the largest
value of strain gauge is the largest package, with d = 740 mils. Therefore the calculation of ^.$MT
will be based on this device.
n -2.26
a 3 5( (a AT a AT+T xl 6
SMT - ^5h I s " cc( RISE) I » )) CR
d 740 mils.
h 5 mils.
a 15.17
s
AT 31°C (default for Ajpenv.)
a,'CC 6.5
T 6 J C P = 20(.5) = 10°C
RISE
1 (leadless)
CR 115,000 (cycles/106 hr.)
740 -2.26
6
a 3 5
SMT = - ,(.65)(5) 1(15.17(31) -6.5(31 + 10)) | xlO"
C
115,000( P )
10 6 hrs.
.0314 x 10 6 hrs. (calendar time)
LC (.1752 hrs.)
a
SMT .0314 (lO 6 ) hrs.
5.58
= 1 (from table)
.03 Y4 = 31.8F/10*
l
SMT a
SMT
5-49
Vrm
-, -2.26
a AT + T
a PTH ^ M ( S Z AT - a 2 ( RISE)) I
T = 50 Mils.
a =
SZ ^0 (TCE of Epoxy - Glass Z axis)
a2 = 17 (TCE of Copper PTH)
(all other factors as calculated for ^ S M T )
-2.26
.0061
50 | (20 (31) - 17 (31 + 10)) |
a
PTH - .115
= .33 x 10 6 hrs.
LC = (.1752) (lQ 6 ) = 53
6
"PTH .33 (lO )
Xi = . 5 1 (from table)
^1
^PHTl = a
— = ^ 6 ~ = 1-55 F/10 6
PTH -33(l0 hr.)
XPTH2
^b = .000025
K = 1
Q
K
E = 8.0
n
l = 500
*c = 1.5
n
2 = 0
DC = .04
^p = ^SMT + V r H l + ? l PTH2
Xp = 31.8+1.55+ .0043 = 33.35 F/10 6
5-50
Rotating Devices
p
AocB Bav/
LC 87,600
80
= >1
aB " >200
^ = 87,600/5xl0 5 = .175
aw
— 5
= . 3 x l 0 " 6 h—
r
Baw 1.12 (5xl0 ) ' -
= .3 F/10 6
5-51
6.0 MODEL COMPARISON
This section compares a sampling of the models developed in this effort to the existing MEL-
HDBK-217E, Notice 1, models. Table 6.0-1 summarizes this comparison and presents the
predicted failure rates for each and the ratio under both benign conditions and severe conditions.
Benign conditions used in these calculations are:
Environment = Gg
Stress = .5
Quality = MIL-Spec.
T A = 25°C
Environment = Arjp
Stress = .9
Quality = MIL-Spec.
T A = 70°C
6-1
TABLE 6.0-1:
MODEL COMPARISON
Capacitors
Paper .0011 .0114 .10 .59 15.3 .039 .l|iF
Plastic .0020 .0093 .21 .15 7.8 .019 .l^F
Mica .0011 .0042 .27 11.0 3.19 3.45 lOOpF
Ceramic .0008 .0080 .10 .018 .045 .43 lOOpF
Al Elec. .00029 .037 .008 ..46 30.8 .015 10|iF
Ta Elec. .000031 .0017 .018 1.76 .011 160 10|iF
Resistors
Film .0037 .0014 2.6 .16 .048 3.3
Network .0019 .0066 .29 .082 .039 2.1 NR=10
Transformers
Audio .013 .0072 1.8 .21 .22 .95
Power .048 .019 2.7 .76 .60 1.3
Pulse/Switching .00057 .0036 .16 .009 .112 .09
Switches (Resistive
Toggle .102 .00045 226 2.2 .0098 224 Load)
6-2
6.1 MODEL COMPARISON OBSERVATIONS
From this analysis, several conclusions can be drawn relative to the current MIL-HDBK-
217E models:
(2) Tantalum capacitor failure rates exhibit a very high dependency on applied voltage,
making their predicted failure rate lower at low voltages and higher at higher voltages.
(5) Switches and relay failure rates in general are very much higher and have a much higher
dependence on environment.
6-3
7.0 CONCLUSIONS AND RECOMMENDATIONS
The objective of this effort was to develop or modify the MIL-HDBK-217 failure rate
models for Capacitors, Resistors, Inductive Devices, Switches, Relays, Connectors,
Interconnection Assemblies/Printed Wiring Boards, and Rotating Devices. This was accomplished
with the statistical analysis of field failure rate data or from laboratory test results. A new
methodology was also developed to predict failure rates of items exhibiting wearout characteristics.
More specifically the objectives of these models are that:
(2) They be based on data available to design engineers during equipment design phases.
(3) They are inclusive of all part types used in military systems.
(4) They be as accurate as possible and be based on sound physics of failure principals.
The failure rate models developed in this effort and summarized in Section 5.0 of this report
meet all objectives listed above.
It was also apparent after developing these models that the failure rates predicted with them in
some cases differed significantly from existing MIL-HDBK-217E models being either higher or
lower. Additionally, new part types not included in MIL-HDBK-217E are included in the
proposed models. Examples of these include:
7-1
• Various Socket Types
• Surface Mount Technology
• Full Horse Power Motors
7-2
8.0 REFERENCES
(1) Morrison, J.D., "Reliability Modelling of Tantalum Capacitors," 1988 CART Proceedings,
p. 128.
(2) Draper, N., Smith H., "Applied Regression Analysis," Second Edition, Wiley.
(3) Priore, Mary G., "Discrete Semiconductor Device Reliability," RAC publication number
DSR-4, 1988.
(4) Yudewitz N. "Predicting Relay Life Mathematically," 26th Relay Conferences, 1978.
(6) B.D. Hatch, et al., "Technical Information Series No. R6250 196," PT Missile and Space
Div., G.E. Co.
(7) Word P., "Considerations In Design and Use of Multilayer Ceramic Capacitors in Surface
Mounting," 1987 CARTS Proceedings, p. 79.
(8) Burnham, J., "Weibull Life Tests of KEMET Solid Tantalum Chip Capacitors at Highly
Accelerated Voltages," 22nd Electronic Components Conference.
(9) Campbell and Hayes, "An Analysis of the Field Failure of Passive and Active
Components," Loughborough University of Technology, June 1989.
(11) "MTBF and Stability Data for Hy-Cal Engineering Platinum Resistance Temperature
Sensors," Hy-Cal Engineering Report No. CF-393.2.
(12) Coit, David W., and Mary G. Priore, "Impact of Nonoperating Periods on Equipment
Reliability," RADC-TR-85-91.
(13) Rossi, Michael, "Nonelectronics Parts Reliability Data," RAC publication number NPRD-
3, 1985.
(14) Minford, W.J., "Accelerated Life Testing of High K Multilayer Ceramic Capacitors,"
NASA Conference Publication 2186.
(16) Bora, "Long Term Performance Studies of Electronic Component at Rated Electrical
Stress," Microelectronics and Reliability, Vol. 26, N5, pp. 989-991.
(17) "Expressing Capacitor Reliability Accurately," Union Carbide Electronics, Report No. F-
1884.
(18) Burnham, "Weibull Life Tests of Kemet Solid Ta Chip Capacitors at High Accelerated
Voltages," Hughes Aircraft, 22nd Electronics Components Conference, 1982, pp. 439-
455.
8-1
(19) Brodsky, et al., "Reliability and Application of Leaded Plastic Chip Carriers," Electronic
Packaging and Production, Nov. 1981.
(20) Smeby, "Solder Joint Behavior in HCC/PWB Interconnects," 1984 Proceedings 34th
Electronic Components Conference.
(21) Chappen and Banas, "Integrating Surface Mounted Technology into Manufacturing," 1984
Electronic Components Conference.
(22) Reynolds, "SMDs Invade Military and Commercial Equipment," Electronic Packaging and
Production, Feb. 1985.
(23) Wright, "Thermal Shock Testing of Ceramic Thick Film Substrate," 1984 International
Symposium on Microelectronics.
(24) Baron, R.V., Weingarten, D.H., "A Reliability Assessment of Polycarbonate Film
Capacitors," 1986 CARTS Symposium, p. 117.
(25) Danielson, "Chip Carriers on Ceramic Thick Film Multilayer Boards in High Reliability
Applications," 1984 International Symposium on Microelectronics.
(26) Isaacson, et. al., "Implementing Copper Thick Film Substrates for Surface Mount
Technology," April 1985.
(27) Wright and Wolverton, "The Effect of the Solder Reflow Method and Joint Design on the
Thermal Fatigue Life of Leadless Chip Carrier Solder Joint," 1984 Electronic Components
Conference, 1984.
(28) Gray, "Substrates for Chip Carrier Interconnections," Surface Mount Technology: ISHM
Technical Monograph series 6984-002, 1984.
(29) Caswell, "Vapor Phase Soldering of SMDs: Reliability Characteristics," Circuit World,
1985.
(30) Foldman, "An Unsymmetric Metal Base PWB: Design Parameters and Thermal
Considerations," Circuit World, 1985.
(31) Walker, "Surface Mounted Device: Hybrid Module Technology Trade-offs in
Communications Equipment," 1983 International Microelectronics Symposium, 1983.
(32) Brice-Heames, "Direct Attachment of Leadless Chip Carriers to Various PWB Material,"
1984 International Symposium on Microelectronics, 1984.
(33) Chen and Verulle, "Direct Mounting of Ceramic Leadless Chip Carrier into Glass Epoxy
Circuit Board," 1983 International Microelectronics Symposium, 1983.
(34) Greer, "Post Molded Leaded Chip Carrier," Texas Instruments, 1983.
(36) Rivard, W.H., "Performance Characteristics of Gold vs. Tin Plated Relay and Socket
Terminals."
8-2
(37) Maher, G.H., Hofmaier, R., Neumun, J., Love, G.R., "Accelerated Life Testing and
Reliability of X7R Multilayer Ceramic Capacitor with a PLET Dielectric" 1986 CARTS
Symposium, p. 1.
(38) Stockman, S. and D. Coit, "Surface Mount Technology: A Reliability Review." Reliability
Analysis Center, 1986.
(43) "Standard Reliability Table for Semiconductor Devices," Nippon Telegraphanel Telephone
Public Corp., 1982.
(45) Spencer, J., "The High and Lows of Reliability Predictions," 1986 R&M Symposium
Proceedings, 1986, pp. 156-162.
(46) Denson, W.K.., et. al., "Development of Reliability Prediction Models for Electronic
Components in Automotive Applications," SAE paper 840486, 1984.
(48) Priore, M., and W. Denson, "Automative Electronic Reliability Prediction," SAE Paper
870050, 1987.
(52) Lauber, A., "Aluminum Electrolytic Capacitors Reliability Expected Life, and Shelf
Capability," Sprague Technical Paper TP83-9, Dec. 1985.
(54) Coit, D., "Printed Wiring Assembly and Interconnect Reliability" Sept. 1981, RADC-TR-
81-318.
8-3
(55) Whelan, S., Pecht, M., Dasgupta, A., "Operational Temperature Cycle Values for
Application Environment Categories," Internation Journal for Hybrid Microcircuit Vol. 13
Num 1, March 90.
Reimer, D.E., Saulsburry, Boeing Aerospace "Power Cycling of Ceramic Chip Carriers on
Ceramic Substrates".
Lau, J.H., Rice, D.W., "Solder Joint Fatigue in Surface Mount Technology: State of the
Art" Solid State Technology, Oct. 1985.
Anderson, R., et. al., "Manufacturing Technology for High Reliability Packaging Using
Hermetic Chip Carriers (HCC) on Compatible Printed Wiring Boards (PWBs)," Interim
Technical Report, Jan 88 through April 88.
Coit, D.W., Priore, M.P., "Reliability Prediction Models for Discrete Semiconductor
Devices" RADC-TR-88-97, April 1988.
Bivens, G.A., "Predicting Time to Failure Using Finite Element Analysis," 1990 R/M
Symposium Proceedings.
Weyde, B., "High Failure Rates of Printed Circuit Boards Demonstrate the Need for Bare
Board Testing," IPC Report Number IPC-TP-493, 1983.
Prince, M.D.H., Hayes, J.A., "A Reliability Analysis of Multilayer Ceramic Capacitors
Under Voltage and Temperature Acceleration," 1987 CARTS Europe, p. 195.
Waser, K., "Accelerated Life Testing and Long Term Reliability of Multilayer Ceramic
Capacitors," 1987 CARTS Europe Symposium, p. 189.
Mogilevsky, B.M., Shirm, G.A. "Accelerated Life Tests of Ceramic Capacitors," 1988
Electronic Components Conference, p. 362.
8-4
(74) Tieman, B.M., "The Characteristic Life of a Dry Reed Contact - The influence of
Switching Phenomena," 29th Relay Conference, 1981.
(77) Sutherland, E.F., "Quality and Reliability Considerations for Dry Reed Relays,"
Evaluation Engineering, November 1989.
(78) Kimura, K., Ishino, M., Matsui, K., and Mitani, S., "Resistance Increase of Gold-Plated
Silver Contacts by Carbon and its Acceleration Factor," 27th Relay Conference, 1979.
(79) Wilson, D.S., Smith, R., "Electric Motor Reliability Model," RADC-TR-77-408 (AD
A050179) December 1977.
(82) Munikoti, R. and Dhar, P., "Highly Accelerated Life Testing (HALT) For Multilayer
Ceramic Capacitor Qualification."
(83) Cozzolino, M.J., and R.C. Straessle, "Design, Characteristics and Failure Mechanisms of
Tantalum Capacitors," 1988 CARTS Symposium, p. 98.
8-5
APPENDIX A:
DETAILED DATA
Reliability Modeling of Critical Components Appendix A
FLD G 0 40000.OH 0
Capacitor, Unknown Mica (Metallised) Unk 1
FLD G 1 200000000.OH 0
Capacitor, Unknown Paper (Metallised) Unk 1
FLD G 0 40000.OH 0
Capacitor, Unknown Paper Plastic 30.OOd 1
FLD G 0 3000.OH 0
Capacitor, Unknown Paper Plastic Metal Unk 1
FLD G 0 700000.OH 0
Capacitor, Unknown Polycarbonate Foil Unk 1
FLD G 0 20000000.OH 0
Capacitor, Unknown Polycarbonate Metal Unk 1
FLD G 0 2000000000.OH 0
Capacitor, Unknown Polyester Metallise Unk 1
FLD G 4 2000000000.OH 0
Capacitor, Unknown 1
Polystyrene Foil Unk
FLD G 10 300000000.OH 0
Capacitor, Unknown Preset Unk
FLD G 0 8000000.OH 0
Capacitor, Unknown Ta Electrolytic Unk
11T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-2
Reliability Modeling of Critical Components Appendix A
IIT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-3
Reliability Modeling of Critical Components Appendix A
IIT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-4
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-6
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-7
Reliability Modeling of Critical Components Append'.x A
I IT Research Institute « Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-8
Reliability Modeling of Critical Components Appendix A
3632712000.OH 254544 29
M NOP GF 0 906554000.OH 38456 17
Capacitor, Fixed Mica O.OOv
FLD 0
(_)
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-9
ReliabiIity Modeling of Critical Components Appenaix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-10
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Canpus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-11
Reliability Modeling of Critical Components Append''x A
I IT Research Institute * Beeches Technical Campus * Rte. 26H * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-12
Reliability Modeling of Critical Conponents Appendix A
IIT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A- 13
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-14
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-15
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-16
Reliability Modeling of Critical Components Append'A A
11T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, MY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-17
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-18
Reliability Modeling of Critical Components ftppencn < A
IIT Research Institute * Beeches Technical Canpus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-19
Reliability Modeling of Critical Components Appendix A
M FLD GF 0 114048.OH 4
Circuit Breaker, Magnetic DPST 5.00a
M FLD GM 0 3690000.OH 210
Circuit Breaker, Magnetic DPST 10.00a
M FLD GM 0 123000.OH 7
Circuit Breaker, Magnetic SPST 0.20a
M FLD GF 0 142560.OH 5
Circuit Breaker, Magnetic SPST 1.00a
M FLD GM 0 123000.OH 7
Circuit Breaker, Magnetic SPST 2.00a
M FLD GF 0 355902.OH 0
M FLD GM 0 246000.OH
Circuit Breaker, Magnetic SPST 3.00a 14
M FLD GM 0 123000.OH
Circuit Breaker, Magnetic SPST 4.00a 7
M FLD GM 0 246000.OH
Circuit Breaker, Magnetic SPST 5.00a 14
M FLD GM 0 123000.OH
Circuit Breaker, Magnetic SPST 8.00a 7
M FLD GM 0 123000.OH
Circuit Breaker, Magnetic SPST 10.00a 7
M FLD GF 0 355902.OH
M FLD GM 0 123000.OH 0
Circuit Breaker, Magnetic SPST 20.00a 7
M FLD GF 0 355902.OH 0
M FLD GM 0 123000.OH
Circuit Breaker, Magnetic SPST 30.00a 7
M FLD GF 0 711804.OH
Circuit Breaker, Magnetic SPST 50.00a 0
M FLD GF 0 1067706.OH
Circuit Breaker, Molded Case 3PST 15.00a 0
M FLD GF 4 6341952.OH
Circuit Breaker, Molded Case 3PST 70.00a 1322
M FLD GF 0 1477520.OH
Circuit Breaker, Molded Case 3PST 125.00a 80
M FLD GF 6 4944480.OH
Circuit Breaker, Molded Case DPST 15.00a .280
M FLD GF 8 6392880.OH
Circuit Breaker, Molded Case SPST 15.00a 1010
M FLD GF 11 7029216.OH
Circuit Breaker, Power Switch Unknown Unk 1172
U FLD GF 70 43219000.OH
Circuit Breaker, Power Switch 3PST 200.00a 3888
M FLD GF 6 2083968.OH
Circuit Breaker, Thermal unknown Unk 216
U FLD GF 3 8944000.OH
Circuit Breaker, Thermal SPST 7.50a 675
M FLD GM 0 26116.OH
Circuit Breaker, Thermal SPST 15.00a 69
M FLD GM 0 52232.OH
Circuit Breaker, Thermal SPST 20.00a 138
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-20
Reliability Modeling of Critical Components Appendix A
M FLD GM 0 26116.OH 69 1
Circuit Breaker, Under Voltage Unknown Unk
M FLD GF 8 4278000.OH 350 2
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-21
ReliabiIity Modeling of Critical Components Appendix A
Part Type
Qual DType Env Tot. Fail otal Duration Total Pop. No. Rec
Connector, Unknown,
M FLD A 162 2480955.OH 430 74
M FLO AIA 0 1653296.OH 3552 5
M FLO AIT 0 902400.OH 1504 5
M FLO GF 12 6404060.OH 258 2
M FLO GM 0 2503749.OH 9 1
M FLO NH 8 9265.OH 0 1
U FLO A 0 0.0H 0 1
U FLO ARW 0 0.0H 0 1
U FLD G 0 0.0H 0 1
U FLO GF 0 0.0H 0 1
U FLD N 0 0.0H 0 1
U FLD NSB 0 0.0H 0 1
U FLD SF 0 0.0H 0 1
Connector, Electrical, ,
C FLD A 0 1028595000.OH 9 4
C FLD AI 0 238000.OH 0 15
C FLD AUT 0 2368000.OH 0 5
C FLO GBC 0 213870800.OH 164516 23
C FLD GF 0 1451388.OH 384 1
C FLD GM 0 7000.OH 0 1
C FLD GMW 0 3380000.OH 1 1
M FLD A 32 603853.OH 40 17
M FLD AI 0 328000.OH 0 5
M FLD AIA 0 1653296.OH 3552 5
H FLD AIF 0 65574875.OH 860826 923
M FLD AIT 0 2170042.OH 7510 6
H FLD AU 1 45812040.OH 108690 25
H FLD AUA 2 5166550.OH 13320 25
M FLD AUF 1 2925800.OH 13950 25
M FLD DOR 0 11624755000.OH 135063 13
M FLD G 46 420000000.OH 0 4
M FLD GF 1 5109130350.OH 23486 61
M FLD GM 3 39901283.OH 77590 294
M FLD HEL 0 1850000.OH 0 1
M FLD MP 0 3889520.OH 64492 38
M FLD KBS 0 240973400.OH 0 63
M FLD NS 0 79339190.OH 19552 52
M FLD NSB 2 2842055378.OH 66413 305
M FLD SF 0 40633000.OH 0 2
M LAB N/R 0 40000.OH 20 1
Connector, Electrical, AC,
C FLD GBC 0 834766400.OH 642128 32
Connector, Electrical, AMP,
C FLD GBC 0 2366000.OH 1820 1
Connector, Electrical, Adapter 1
C FLD GBC 40 842041200.OH 647724 68
Connector, Electrical, Ampheno I,
C FLD GBC 0 275600. OH 212 1
Connector, Electrical, Anode,
C FLD GBC 0 14138800.OH 10876 2
I IT Research Institute * Beeches Technical Campus * Rte. 26« * Rome, MY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-22
Reliability Modeling of Critical Components Appendix A
Part Type
Qua I DType Env Tot Fail Total Duration Total Pop. No. Rec.
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, MY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-23
Reliability Modeling of Critical Components Appendix A
Part Type
Qual DType Env Tot. Fail otal Duration Total Pop No. Rec.
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-24
Reliability Modeling of Critical Components Append i x A
Part Type
Qual DType Env Tot. Fail Total Duration Total Pop No. Rec.
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-25
Reliability Modeling of Critical Components Appendix A
Part Type
Qual DType Env Tot. Fail Total Duration Total Pop. No. Rec.
M FLD AI 0 1707000.OH 0 20
M FLD AIA 5 1446634.OH 3108 6
M FLD AIF 0 1242320.OH 279 8
M FLD AIT 1 789600.OH 1316 6
M FLD AU 0 71343204.OH 10869 3
M FLD AUA 0 413324.OH 1332 2
M FLD AUF 0 234064.OH 1395 2
M FLD GF 2 67721358.OH 530 18
M FLD GM 0 765665.OH 0 7
M FLD NBS 0 310000.OH 0 18
M FLD SF 0 829000.OH 0 1
Connector, Electrical, Receptacle,
C FLD GBC 0 114192000.OH 87840
Connector, Electrical, Receptacle, Blue Ribbon
C FLD GBC 0 17222400.OH 13248
Connector, Electrical, Receptacle, D-Microminiature
C FLD GBC 0 665600.OH 512
Connector, Electrical, Receptacle, D-Subminiature
C FLD GBC 4 2225995200.OH 1712304 173
Connector, Electrical, Receptacle, Microribbon
C FLD GBC 0 922937600.OH 709952 76
Connector, Electrical, Rectangular,
C FLD A 0 68699000.OH 0 2
C FLD GBC 0 136224400.OH 104788 27
C FLD GF 0 140018000.OH 0 1
M FLD AIA 0 206662.OH 444 1
M FLD AIT 0 112800.OH 188 1
H FLD G 139 2000000000.OH 0 1
H FLD SF 0 1450400.OH 0 2
U FLD A 0 0.0H 0 1
U FLD G 0 0.0H 0 1
Connector, Electrical, Round,
C FLD GBC 176800.OH 136 1
Connector, Electrical, Signal,
C FLD GBC 72148185200.OH 55498604 150
Connector, Electrical, Signal, QDISC
C FLD GBC 3943966000.OH 3033*20 90
Connector, Electrical, Special Purpose,
C FLD GBC 119532400.OH 91948 9
Connector, Electrical, Telephone,
C FLD GBC 242216000.OH 186320 29
M FLD GF 1954560.OH 509 2
H FLD MP 1093560.OH 18226 7
M FLD NS 460340.OH 10 1
Connector, Electrical, Test Adapter,
C FLD GBC 1523600.OH 1172
Connector, Electrical, Test Point,
M FLD AIF 7715400.OH 148148
M FLD GF 4515304772.OH 301931
H FLD NS 844486168.OH 18440
Connector, Electrical, Utitly,
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-26
Reliability Modeling of Critical Components Appendix A
Part Type
Qual DType Env Tot. Fail Total Duration Total Pop. No. Zee.
11T Research Institute * Beeches Technical Campus * Rte. 26M * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-27
Re Li ability Modeling of Critical Components Appendix A
Part Type
Qua I DType Env Tot. Fail Total Duration Total Pop. No. Rec.
Connection, Assembly,
M FLD AIF 0.0H
Connection, Connector Post,
C FLD AI 53 270000.OH 0 1
C FLD GBC 8 35533383600.OH 27333372 1896
Connection, Contact, Spring
C FLD GBC 23805600.OH 18312
Connection, Solder,
M FLD A 5995553000.OH 0
M FLD AIF 6287550.OH 121693
M FLD DOR 34900000000.OH 0
M FLD GF 162329440000.OH 0
M FLD NS 1640528000.OH 0
U FLD A 0.0H 0
U FLD G 0.0H 0
U FLD GF 0.0H 0
U FLD N 0.0H 0
Connection, Solder, Hand Lap
M FLD DOR 52594180000.OH 0
M FLD SF 39610000000.OH 0
Connection, Solder, Reflow
M FLD GF 8835115000.OH 0
Connection, Solder, Wave
M FLD NS 57835239168.OH 935482
Connection, Terminal,
C FLD A 0 28000.OH 1
M FLD A 158 31948000.OH 0
M FLD AIF 0 27699516.OH 612404 165
M FLD AIT 0 2535284.OH 16107
M FLD AU 0 3054136.OH 7246
FLD AIM 0 206662.OH 888
FLD AUF 0 117032.OH 930
FLD GM 0 2629458.OH 84608 37
FLD MP 0 252360.OH 4206 2
FLD NS 0 7384888.OH 1819 27
FLD NSB 0 95089800.OH 2121 32
Connection, Terminal, Barrier Block
C FLD GBC 20 731848000.OH 562960 98
Connection, Terminal, Block
C FLD GBC 0 68411200.OH 52624 11
Connection, Terminal, Board
C FLD GBC 0 644800.OH 496 2
M FLD AU 0 13743612.OH 32607 9
M FLD AUA 0 1859958.OH 3996 9
M FLD AUF 0 1053288.OH 4185 9
u FLD A 0 0.0H 0 1
U FLD AUT 4 702784.OH 1856 1
U FLD GF 9 2340646.OH 2063 5
U FLD GM 17 7894728.OH 4371 4
U FLD NS 26 23662586.OH 6071 4
U FLD NSB 5 658800.OH 954 3
IIT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-28
Reliability Modeling of Critical Components Appendix A
Part Type
Qual DType Env Tot. Fail Total Duration Total Pop. No. Rec
U FLO NU 1 99560.OH 92 1
Connection, Terminal, Crimp
C FLD GBC 0 4249949600.OH 3269192 140
Connection, Terminal, Feed Th rough
M FLD AIA 0 1033310.OH 2220 1
M FLD AIF 0 5960959.OH 45725 31
M FLD AIT 0 564000.OH 940 1
M FLD AU 0 256547424. OH 608664 9
M FLD AUA 0 1859958.OH 74592 9
M FLD AUF 0 1053288.OH 78120 9
M FLD GM 0 5967.OH 192 2
Connection, Terminal, Lug
M FLD AIF 0 4951870.OH 86427 74
M FLD AIT 0 1267642.OH 1911 1
M FLD AU 0 85515808.OH 202888 9
M FLD AUA 0 1859958.OH 24864 9
M FLD AUF 0 1053288.OH 26040 9
M FLD GM 0 3815890.OH 122816 67
M FLD MP 0 504720.OH 8412 4
M FLD NS 0 1688068.OH 416 8
U FLD GBC 4 4108213200.OH 3160164 81
Connection, Terminal, Metal Sleeve
C FLD GBC 0 1453181600.OH 1117832 40
Connection, Terminal, Screw
C FLD GBC 0 1658800.OH 1276 3
Connection, Terminal, Stand-o:ff
M FLD AU 0 85515808.OH 202888 4
M FLD AUA 0 826648.OH 24864 4
M FLD AUF 0 468128.OH 26040 4
Connection, Terminal, Strip
C FLD GBC 0 447657600.OH 344352 27
Connection, Terminal, Stud
C FLD GBC 4 5769010000.OH 4437700 69
M FLD AIA 0 6819846.OH 14652 4
M FLD AIF 0 15458610.OH 160782 47
M FLD AIT 0 3722400.OH 6204 4
M FLD AU 0 171031616.OH 405776 5
M FLD AUA 0 1033310.OH 49728 5
M FLD AUF 0 585160.OH 52080 5
M FLD GM 0 1182726.OH 38336 30
H FLD MP 0 3364800.OH 56080 7
Connection, Terminal, Tab
C FLD GBC 0 46113600.OH 35472 5
Connection, Terminal, Test Point
C FLD GBC p 3982908800.OH 3063776 4
Connection, Weld Joint,
C FLD GF 0 490600000.OH 0 1
M FLD A 0 157063000.OH 0 1
H FLD GF 0 65259910000.OH 0 2
M FLD GM 0 529200000.OH 21168 1
U FLD A 0 O.OH 0 2
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-29
Reliability Modeling of Critical Components Apper.ai * A
Part Type
Qual DType Env Tot. Fail Total Duration Total Pop. No. Rec.
U FLD G 0 0.0H 0
Connection, Wire, Joke
C FLO GBC 0 61510800.OH 47316
Connection, Wire Wrap,
M FLO A 0 100000000.OH 0
M FLD GF 0 556809888000.OH 128
U FLD A 0 0.0H 0
U FLD G 0 0.0H 0
Connect ion, Wire Wrap, Solder
M FLD GF 0 3056630000.OH
Connection, Wire Wrap, Solderl ess
M FLD AUT 0 456105000.OH
H FLD MSB 4 32500000000.OH
11T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-30
Reliability Modeling of Critical Components Appendix A
Part Type HP
Qua I DType Env Tot. Fail Total Duration Total Pop. No. Rec.
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-31
Reliability Modeling of Critical Components Appendix A
Part Type HP
Qua I DType Env Tot. Fail Total Duration Total Pop, No. Rec.
M T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-32
Reliability Modeling of Critical Components Append i x A
Part Type HP
Qua I DType Env Tot. Fail Total Duration Total Pop No. Rec.
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-33
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-34
ReLlability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-35
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 134A0-2069 * 315/336-2359 * FAX 315/336-1371
A-36
ReliabiIity Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Canpus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A--37
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, MY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-38
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-39
Reliability Modeling of Critical Components Appendix A
M FLD GF 0 41600.OH 0 1
Relay, Electromechanical 0.10a DPDT
M FLD NSB 1 13096000.OH 299 1
Relay, Electromechanical 0.75a DPDT
M FLD NSB 1 2847000.OH 65 1
Relay, Electromechanical 1.00a 1A
C FLD GBC 52 474016400.OH 364628 7
Relay, Electromechanical 1.00a 1A DRY
C FLD GBC 0 572000.OH 440 2
Relay, Electromechanical 1.00a 18
C FLD GBC 0 130000.OH 100 1
Relay, Electromechanical 1.00a 1C
C FLD GBC 0 7077200.OH 5444 2
Relay, Electromechanical 1.00a 2A
C FLD GBC 0 31938400.OH 24568 3
Relay, Electromechanical 1.00a 4PDT
H FLD AIT 16 294000000.OH 528
Relay, Electromechanical 1.00a 6PDT
M FLD NSB 1 2835000.OH 65
Relay, Electromechanical 1.00a DPDT
M FLD AU 0 3054136.OH 7246
M FLD AIM 0 206662.OH 888
M FLD AUF 0 117032.OH 930
M FLD GF 0 741312.OH 26 3
Relay, Electromechanical 2.00a Unknown
M FLD GM 1 1989.OH 64
Relay, Electromechanical 2.00a 1A
C FLD GBC 0 13410800.OH 10316
Relay, Electromechanical 2.00a 1C
C FLD GBC 0 1861600.OH 1432
Relay, Electromechanical 2.00a 1D
C FLD GBC 0 83200.OH 64
Relay, Electromechanical 2.00a 2A
C FLD GBC 0 25584000.OH 19680
Relay, Electromechanical 2.00a 3A
C FLD GBC 0 18345600.OH 14112
Relay, Electromechanical 2.00a 3PST
H FLD GM 1 5967.OH 192
Relay, Electromechanical 2.00a DPDT
M FLD AIT 5 98000000.OH 175
M FLD GF 1 826848.OH 29
M FLD NS 0 15938496.OH 32
M FLD NSB 1 2277600.OH 52
Relay, Electromechanical 3.00m 1A
C FLD GBC 0 104000.OH 80
Relay, Electromechanical 3.00m 2A
C FLD GBC 4 36129600.OH 27792
Relay, Electromechanical 5.00a Unknown
M FLD NSB 2 4555200.OH 104
Relay, Electromechanical 5.00a 2D
C FLD GBC 4 665600.OH 512
1 IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-40
Reliability Modeling of Critical Components Appendix A
A-41
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26M * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-42
Reliability Modeling of Critical Components Append iA A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-43
Reliability Modeling of Critical Components Appendix A
M FLD GM 0 246000.OH 14 1
Relay, Time Delay Unk SPDT
M FLD AUT 1 321600.OH 0 1
Relay, Time Delay Unk SPST
H FLD NBS 0 500000.OH 0 1
I IT Research Institute * Beeches Technical Campus * Rte. 26M * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-44
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-45
ReliabiIity Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-46
ReliabiI ity Modeling of Critical Components Append i x A
I IT Research Institute * Beeches Technical Campos Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-47
Reliability Modeling of Critical Components Appendix A
11T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-48
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-49
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-50
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-51
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26M * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-52
Reliability Modeling of Critical Components Append i x A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-53
ReliabiIity Modeling of Critical Components Appendix A
Part Type
Qual OType Env Tot. Fail Total Duration Total Pop. No. Rec.
Socket, Unknown,
C FLD GBC 0 13202800.OH 10156 8
H FID AIF 0 16069404.OH 147158 56
H FLD GF 0 1954560.OH 509 2
H FLD MSB 0 384343800.OH 8775 7
Socket, Adapter,
C FLD GBC 0 199992000.OH 153840 5
Socket, Coax,
C FLD GBC 0 50585600.OH 38912 1
Socket, Crystal,
C FLD GBC 0 3920800.OH 3016 2
Socket, Crystal, HC-25/U
C FLD GBC 0 39296400.OH 30228 3
Socket, Crystal, HC-6/U
C FLD GBC 0 50819600.OH 39092 4
Socket, DIP,
C FLD GBC 8 10631020400.OH 8177708 147
C FLD GF 0 1821936000.OH 483152 2
M FLD GF 0 3285772188.OH 0 12
M FLD MS 0 200500000.OH 40744 1
Socket, Display,
C FLD GBC 0 578156800.OH 444736 17
Socket, Ground,
C FLO GBC 0 165105200.OH 127004 1
Socket, Hi-densi ty,
c FLD GBC 0 27752400.OH 21348 1
Socket, IC,
c FLD GBC 0 39150800.OH 30116 4
Socket, IC, Chip> Carrier
c FLD GBC 0 418121600.OH 321632 5
Socket, IC, PGA
C FLD GBC 0 74001200.OH 56924 18
Socket, Lamp,
M FLD GF 0 124942090.OH 7859 1
M FLD NS 0 76218231.OH 1656 1
Socket, Recpepticle,
C FLO GBC 0 127114000.OH 97780 4
Socket, Relay,
C FLD GBC 4 98498400.OH 75768 15
C FLD GM 0 52232.OH 138 1
M FLD AIF 0 118444.OH 507 1
H FLD NS 0 6343310.OH 138 1
Socket, SIP,
C FLD GBC 0 336044800.OH 258496 19
Socket, Spring,
C FLD GBC 0 2683200.OH 2064 1
Socket, Strip,
C FLD GBC 0 188130800.OH 144716 6
Socket, Strip, DIP
C FLD GBC 0 35406800.OH 27236 3
Socket, Strip, SIP
I IT Research Institute * Beeches Technical Campus Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-54
Reliability Modeling of Critical Components Appendix A
Part Type
Qual DType Env Tot. Fail Total Duration Total Pop, No. Rec.
11T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-55
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-56
Reliability Modeling of Critical Components Appendix A
I IT Research Instftute * Beeches Technical Campus Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-57
Reliability Modeling of Critical Components Appendix A
U FLD GM 0 0.0H
Switch, Flow, Unk
M FLD A 11 34679.OH 1
M FLD GF 1 342144.OH 12
M FLD GM 0 3978.OH 128
H FLD NS 0 498078.OH 1
M FLD NSB 0 3985800.OH 91
Switch, Flow, 0.500a
H FLD GF 10 2737152.OH 96
Switch, Flow, Liqu id Unk
M FLD NH 24 30252.OH 0
U FLD GM 4 535968.OH 3260
U FLD NS 7 1386117.OH 368
U FLD NU 2 20960.OH 20
Switch, Flow, Paddle Type Unk
M FLD GF 56 11612160.OH 740
Switch, Foot, Unk
C FLD GBC 0 436800.OH 336
M FLD A 13 25492.OH
Switch, Frame, Unk 1
C FLD GBC 0 561600.OH
Switch, Humidity, Unk 432
N FLD GF 4 238444.OH
Switch, Impact, Unk 54
C FLD GBC 0 24637600.OH
Switch, Inertiat, Unk 18952
M FLD DOR 9 137100000.OH 649
U FLD GF 0 0.0H 0
Switch, Interlock, Unk
M FLD GF 3 15697000.OH 0
U FLD GM 6 9500.OH 190
Switch, Interlock, 10.000a
M FLD GF 2 7584192.OH 266
H FLD NS 1 1494234.OH 3
Switch, Keyboard, Unk
C FLD GBC 0 68577600.OH 52752
M FLD GMW 0 13882.OH 0
Switch, Keylock, Unk
C FLD GBC 0 150800.OH 116
Switch, Keylock, 100.000a
C FLD GBC 0 338000.OH 260
Switch, Keyswitch, Unk
C FLD GBC 0 171600.OH 132
Switch, Lever, Unk
C FLD GBC 0 332800.OH 256
Switch, Limit, Unk
M FLD A 296 11982000.OH 0
M FLD AU 42 96000.OH 0
M FLD GF 5 711000.OC 21
M FLD GF 31 5265574.OH 305
M FLD GM 0 5967.OH 192
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-58
Reliability Modeling of Critical Components Appendix A
IIT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-59
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-60
Reliability Modeling of Critical Components Appendix A
IIT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-61
Reliability Modeling of Critical Components Appendix A
11T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-62
Reliability Modeling of Critical Components Appendix A
IIT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-63
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-64
Reliability Modeling of Critical Components Append i x A
M FLD A 34 4573000.0A 0 1
M FLD A 265 31353013.OH 79 27
M FLD AIF 69 1911430.OH 22990 26
M FLD AIT 1 1267642.OH 546 1
M FLD DOR 0 1010000.OH 0 1
M FLD GF 135 410367414.OH 11957 14
M FLD GM 23 1317960.OH 68 3
M FLD GMU 1 359000.OH 0 1
M FLD HEL 8 430000.OH 0 1
M FLD MP 0 42060.OH 701 1
M FLD NAB 0 569400.OH 13 1
M FLD NBS 18 442723000.OH 0 13
M FLD NS 60 110706010.OH 4445 48
M FLD NSB 0 37011200.OH 845 13
M FLD SF 0 5480000.OH 0 1
U FLD A 0 0.0H 0 1
U FLD ARU 0 0.0H 0 1
U FLD G 0 0.0H 0 1
U FLD GF 0 0.0H 0 1
U FLD GM 0 0.0H 0 1
U FLD N 0 0.0H 0 1
Switch, Toggle, 0,.020a
C FLD GBC 0 85472400.OH 65748 26
Switch, Toggle, 0.,400a
C FLD GBC 0 8845200.OH 6804 1
Switch, Toggle, 0.,500a
C FLD GBC 0 22984000.OH 17680 5
Switch, Toggle, 2.,000a
C FLD GBC 0 103469600.OH 79592 25
Switch, Toggle, 3..000a
C FLD GBC 4 25844000.OH 19880 5
Switch, Toggle, 4,.000a
M FLD Al 6 197800.OH 23 5
M FLD AIF 16 592220.OH 2535 2
Switch, Toggle, 5..000a
C FLD GBC 0 61594000.OH 47380 8
M FLD A! 0 4000.OH 0 1
M FLD AIF 1 236888.OH 1014 1
M FLD GF 0 3341832.OH 111 2
M FLD GMU 0 257000.OH 0 3
M FLD NBS 0 453000.OH 0 5
Switch, Toggle, 6.,000a
C FLD GBC 0 1669200.OH 1284 2
Switch, Toggle, 7.,500a
C FLD GBC 0 5200.OH 4 1
Switch, Toggle, 1CI.OOOa
C FLD GBC 0 17539600.OH 13492 6
C FLD GM 0 26116.OH 69 1
M FLD GM 0 104464.OH 276 3
M FLD NS 0 498078.OH 1 1
Switch, Toggle, 1«I.OOOa
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-6 5
ReliabiIity Modeling of Critical Components Appendix A
M FLD GH 0 26116.OH 69
Switch, Toggle, 20. 000a
M FLD GF 0 1026432.OH 36
M FLD GH 0 78348.OH 207
H FLD NS 0 2490390.OH 5
Switch, Toggle, 25. 000a
M FLD GH 0 120000.OH 0
Switch, Toggle, 28. 000a
M FLD GF 0 684288.OH 24
Switch, Toggle, 30. 000a
C FLD GBC 0 1320800.OH 1016
Switch, Toggle, Alarm Unit
M FLD AIF 0 77626.OH 1762
Switch, Toggle, Alarm 5.000a
H FLD GF 31 6404060.OH 258
Switch, Toggle, Alarm 20. 000a
M FLD GF 4 6404060.OH 258
Switch, Toggle, Sensitive Link
M FLD AIF 0 38813.OH 881
M FLD GF 1 180982000.OH 0
M FLD HP 0 84120.OH 1402
M FLD NS 2 9239000.OH 0
Switch, Toggle, Sensi tive 7.000a
M FLD AIA 0 413324.OH 888
M FLD AIT 0 225600.OH 376
Switch, Voltage 1 Unk
C FLD GBC 0 1019200.OH 784
Switch, Wave Gu ide, Unk
M FLD GF 1 580000.OC 0
M FLD GF 4 1123512.OH 1
U FLD GF 2 500000.OH 20
U FLD GM 3 59481.OH 25
U FLD NS 3 46200.OH 34
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-G6
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26M * Rome, MY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-67
Reliability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-58
ReLlability Modeling of Critical Components Appendix A
I IT Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-69
Reliability Modeling of Critical Components Appendix A
I IT Research I n s t i t u t e * Beeches Technical Campus * R t e . 26M * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-70
Reliability Modeling of Critical Components Appendix A
11T Research Institute * Beeches Technical Campus * Rte. 26N * Rome, NY 13440-2069 * 315/336-2359 * FAX 315/336-1371
A-71
APPENDIX B:
PART PARAMETERS
B-l
RESISTORS
Part number
Spec, number
Style designation
Manufacturer
Type
• Fixed
• Variable
• Potentiometers
- Single-turn
- Multi-turn
• Trimmer
• Rheostat
• Network
• Chip
• Thermistor
• Varistor
Material
• Carbon composition
• Film
• Metal
• Carbon
• Cermet
• Wirewound
Part description
Resistance value
Package Type
• Axial lead
• SIP
• DIP
• Surface mount
Package hermeticity
Rated Power (in watts)
Tolerance
Rated temperature
Quality level (failure rate level)
B-2
CAPACITORS
• Part number
• spec, number
• Style designation
• Manufacturer
• Type
• Fixed
• Variable
• Dielectric materia)
• Paper
• Mica
• Electrolytic
• Aluminum
• Tantalum
• Solid
• Non-solid
• Ceramic
• Glass
• Plastic
• Polystyrene
• Polypropelene
• Polyester
• Polycarbonate
• Package type
• Package material
• Hermetic
• Non-hermetic
• Polarization
• Polarized
• Non-polarized
• Tolerance
• Temperature range
• Capacitance value
• Voltage rating
• Quality level
• Series resistance
B-3
TRANSFORMERS
• Type
• Power
• Audio
• Isolation
• Auto
• Pulse
• Part number
• Spec, number
• Style number
• Manufacturer
• Core material
• Iron
• Nickel
• Cobalt
• Insulation material
• Operating frequency range
• Voltage rating
• Current rating
• Impedance
• Primaty
• Secondary
• Turns ratio
• Number of windings
• Case type
• Quality level
B-4
INDUCTORS
• Type
• Fixed
• Variable
• Part number
• Spec, number
• Style designation
• Manufacturer
• Core material
• Iron
• Nickel
• Cobalt
• Insulation material
• Operating frequency range
• Voltage rating
• Current rating
• Number of windings
• Case type
• Quality level
B-5
ROTATING DEVICES
• Type
• Full Horse Power
• Fractional horse power
• Part number
• Specification number
• Style designation
• Manufacturer
• Function
• Asyncronous
• Syncronous
• Description
• Single phase
• Multi-phase
• Induction
• Capacitor
• Shunt
• Series
• Compound
• Rated output
• Motors (in hp)
• Generators (in Kva)
• Brushes
• Brushless
• Commutator
• Slip ring
• Bearing type
• Roller
• Ball
• Bushing
• Lubrication
• Sealed
• Grease
• Oil
• Winding material
• Rated temperature
B-6
RELAYS
Type
• Electromechanical
• Contact type
• Armature
• Reed
• Mercury wetted
• Contact material
• Electronic (solid state)
Part number
Specification number
Style designation
Manufacturer
Voltage rating (contact)
Current rating (contact)
Mounting type
Terminal type
• solder lug
• pin
• stud
Enclosure
• Hermetic
• Non-hermetic
Temperature rating
Configuration
• SPST
• DPST
• 3PST
• etc.
Quality level
B-7
SWITCHES
Type
• Mechanical
• Toggle
• Push button
• Sensitive
• Rotary
• Thumwheel
• Circuit breakers
• Magnetic
• Thermal
• Ground fault
• Hydraulic
• Trip free
• Centrifugal
• Capacitive touch
• Membrane
• Slide
• Solid state
Part number
Specification number
Style designation
Manufacturer
Contact configuration
• SPST
• DPDT
• #PST
• etc.
Contact material
Voltage rating
Current rating
Enclosure type
Temperature rating
Quality level
B-8
CONNECTORS
Electrical
• Coaxial
• Twinaxial
• DIN
• D-subminiature
• IC sockets
• Rack and panel
• Surface mounted
• High voltage
• Edge card
• PWB
• One piece
• Two piece
• Zoro insertion force
• Mass termination
• Phone
• Multi pin circular
• Press fit
• RF
• Rectangular
Fiber optic
• Tube
• Straight sleeve
• Double eccentric
• Tapered sleeve
• Multi rod
• Couplers
Part number
Specification number
Style designation
Manufacturer
Package
• Sealed
• Non-sealed
Shield
• Shielded
• Non-shielded
Contact material
Insert material
Number of active pins
Current rating per pin
Quality level
B-9
INTERCONNECT ASSEMBLIES/PWB's
Type
• Printed wiring assembly w/PTH's
• Multiwire board
• Flexible circuit board
• Discrete wiring board w/PTH
• Printed wiring board w/surface mount
Part number
Spec, number
Style designation
Manufacturer
Interconnect type
• Wave soldered
• Hand soldered
• Reflow soldered
• Laser soldered
• Vapor phase soldered
• Wire wrapped
• Wrapped and soldered
• Discrete wiring assembly with electroless PTH's
• Weld
• Crimp
Complexity
• Number of circuit planes
• Number of plated through holes
• Cross sectional area of circuit trace
• Distance between traces
Substrate material
• Flexible board
• Teflon
• Polymide
• Polyester
• Polyvinyl
• Polypropelene
• Polyethelene
• Ceramic
• Laminant
• Glass cloth teflon
. ?
• Glass mat polyester-resin
• Rigid board
• Epoxy glass
• Polymide-glass
• Teflon-glass
• Epoxy-kevlar
• Polymide-kevlar
• Epoxy quartz
• Polymide-quartz
• Thermoplastics
• Alumina
• Copper-invar-copper
B-10
INTERCONNECT ASSEMBLIES/PWB's (CONT'D)
• Bonding adhesives
• Vinyl
• Modified epoxy
Conductor
• Copper
• Aluminum
• Steel
• Tin
• Silver
• Quality
B-ll
APPENDIX C:
MONTE CARLO SIMULATIONS
C-l
Failure Rate
1 1 1 1 1 2
0 1 2 8 9 0
0 0 0 0 0 0
0.10 .
0.20 •
0.30 •
0.40 •
0.50 .
0.60 •
0.70 •
0.80 •
0.90 .
1.00 •
1.10 •
1.20 •
1.30 •
1.40 •
1.50 •
1.60 •
1.70 .
1.80 •
1.90 •
2.00 •
2.10 •
2.20 •
T 2.30 .
I 2.40 .
M 2.50 •
E 2.60 .
2.70 •
a 2.80 .
2.90 .
3.00 .
3.10 •
3.20 •
3.30 •
3.40 .
3.50 .
3.60 •
3.70 .
3.80 .
3.90 •
4.00 .
4.10 .
4.20 .
4.30 •
4.40 .
4.50 1
4.60 •
4.70 *
4.80 •
4.90 1 •
5.00
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C-2
Failure Rate
1 1 1 1 1 1 2
9 0 1 2 3 8 9 0
0 0 0 0 0 0 0 0
0.10 ,
0.20 •
0.30 -
0.40 *
0.50 •
0.60 •
0.70 •
0.80 •
0.90 •
1.00
1.10 .
1.20 •
1.30 .
1.40 a
1.50 •
1.60 •
1.70 .
1.80 •
1.90 •
2.00 •
2.10 .
2.20 •
T 2.30 •
I 2.40 •
M 2.50 •
E 2.60 •
2.70 •
Of 2.80 •
2.90 •
3.00 .
3.10 .
3.20 •
3.30 •
3.40 •
3.50 *
3.60 •
3.70 •
3.80 •
3.90 •
4.00 •
4.10 •
4.20 •
4.30 •
4.40 *
4.50 •
4.60 *
4.70 •
4.80 •
4.90
5.00
1 1 1 1 1 1 1 1 1 1 2
9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C-4
Failure Rate
1 1 1 1 1 2
S 7 8 0 1 2 8 9 0
) 0 0 0 0 0 0 0 0
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80 •
0.90
1.00
1.10 «
1.20
1.30
1.40
1.50
1.60 a
1.70
1.80
1.90
2.00
2.10
2.20
T 2.30 •
I 2.40 m
M 2.50 •
E 2.60 •
2.70
a 2.80
2.90
3.00
3.10
3.20 m
3.30
3.40
3.50
3.60
3.70
3.80
3.90
4.00
4.10
4.20
4.30
4.40
4.50
4.60
4.70
4.80 *
4.90
5.00
1 1 2
8 9 0
0 0 0
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90 •
1.00
1.10
1.20
1.30
1.40
1.50
1.60 m
1.70
1.80
1.90 m
2.00 *
2.10
2.20
T 2.30
I 2.40 a
M 2.50 m
E 2.60 m
2.70
a 2.80
2.90
3.00
3.10
3.20 *
3.30
3.40
3.50
3.60 •
3.70
3.80
3.90
4.00
4.10
4.20
4.30
4.40
4.50
4.60
4.70
4.80
4.90
5.00
1 1 1 1 1 1 2
9 0 1 2 3 8 9 0
0 0 0 0 0 0 0 0
0.10
0.20
0.30
0.40
0.50
0.60
0.70 «
0.80 •
0.90
1.00
1.10
1.20
1.30 •
1.40
1.50
1.60
1.70
1.80
1.90
2.00
2.10
2.20
T 2.30 •
I 2.40
M 2.50 •
E 2.60
2.70 m
a 2.80
2.90
3.00
3.10
3.20
3.30
3.40
3.50
3.60
3.70
3.80
3.90 m
4.00
4.10
4.20
4.30
4.40 •
4.50
4.60
4.70
4.80 *
4.90
5.00
1 1 1 1 1 1 1 1 1 1 2
9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0
C-7
F a l l u r e Rate
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C-8
F a i L u r e Rate
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C-9
Failure Rate
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1 1 1 2
1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C-10
Failure Rate
1 1 1 1 1 1 2
9 0 1 2 3 8 9 0
0 0 0 0 0 0 0 0
0.10 |
0.20 j
0.30 j •
0.40 |
0.50 | .
0.60 |
0.70 | •
0.80 | •
0.90 |
1.00 j
1.10 j
1.20 j
1.30 | •
1.40 j •
1.50 j
1.60 j
1.70 |
1.80 |
1.90 |
2.00 |
2.10 |
2.20 |
T 2.30 |
I 2.40 |
M 2.50 | •
E 2.60 | "
2.70 | •
a 2.80 |
2.90 |
3.00 |
3.10 |
3.20 j
3.30 |
3.40 j
3.50 j
3.60 |
3.70 |
3.80 j •
3.90 j •
4.00 |
4.10 j
4.20 |
4.30 j •
4.40 |
4.50 j .
4.60 j
4.70 j
4.80 |
4.90 |
5.00 |
1 1 2
8 9 0
0 0 0
C-ll
APPENDIX D
REGRESSION RESULTS
D-l
This appendix presents the multiple regression results on which the reliability models
developed in this study have been based. Using the regression analysis described in Section 2, the
constants summarized in this appendix have formed the basis for both base failure rates and
multiplicative model parameters. The results presented here are the results of the final regression
runs, and as such may not include all factors present in the final model. The reason for this, as
described in the model development Section (4), some model parameters needed to be derived
independently from the final regression analysis. Examples of these parameters are quality and
environment. Typically, these parameters were quantified with initial repression results along with
any other information available. When the final parameters were derived, the regressions were re-
run by compensating (dividing) the observed failure rate for these "given" values. Other
parameters analyzed in this manner were typically continuous variables, such as switch current
rating. The reason for this is that the entire dataset typically will not have values for those variables
and with these "unknown" values, the regression yields erroneous results. A more efficient
method to analyze the effect of such variables is to subset the database with those data points for
which the parameter is known, then compensate the failure rate for the derived value and re-run the
regressions. The final regression results, therefore, will be inclusive of the discrete variables
which comprise the final model (see the discussions in Section 4 for the relevant initial
parameters).
Since the logarithmic transformation was taken to yield a multiplicative model, the inverse In
must be taken for the values listed in the regression results. Also listed in this appendix are various
statistics relevant to the regression analysis.
The variables listed under "variables not in the equation" are those determined by the analysis
to be not significantly different than the variables to which the models are normalized. The
normalizing variables are listed on the cover page corresponding to each part type.
D-2
Switches
Normalized to;
CoAx Switch
Gp Environment
Military Quality
D-3
VARIABLES IN THE EQUATION
D-4
VARIABLES NOT IN THE EQUATION
Multiple R .46944
R Square .22037 R Square Change .00626
Adjusted R Square .17616 F Change 1.55891
Standard Error 1.88606 Significant F Change .2133
D-5
Inductors
Normalized to;
Fixed Inductor
Gg Environment
Military/Commercial Quality of 20:1
D-6
VARIABLES IN THE EQUATION
Multiple R .88273
R Square .77921 R Square Change .02191
Adjusted R Square .70561 F Change 1.78636
Standard Error 1.89122 Significant F Change .1980
D-7
Transformers
Normalized to;
Gg Environment
Commercial Quality
Non-RF Transformers
D-8
VARIABLES IN THE EQUATION
Multiple R .82823
R Square .68596 R Square Change .07121
Adjusted R Square .55137 F Change 3.17463
Standard Error 1.27324 SignifF Change .0965
D-9
Resistors
Normalized to;
D-10
VARIABLES IN THE EQUATION
D-ll
VARIABLES NOT IN THE EQUATION
Multiple R .79436
R Square .63101 R Square Change .00618
Adjusted R Square .61213 F Change 3.59886
StandardError 1.98815 Significant F Change .0592
D-12
Capacitors
Normalized to;
D-13
VARIABLES IN THE EQUATION
Multiple R .89146
R Square .79470 R Square Change .00143
Adjusted R Square .77520 F Change 1.24349
Standard Error 1.38686 Significant F Change .2663
D-14
Connectors
Normalized to;
COAX
Military Quality
Ground Environment
D-15
VARIABLES IN THE EQUATION
Variable B
CI5 (telephone) -2.03
CI3 (signal) -9.10
C12 (rectangular) -2.66
C6 (elastometeric) -2.35
C5 (edge card) -2.96
C4 (cylindrical) -2.80
C9(RF) -4.87
C8 (PC edge) -4.64
CI (NOC) -.95
El (airborne) 1.71
(Constant) -14.00
Variable Tolerance
C3 (elect, assy) .08
C7 (hexagonal) .06
C10 (rack & panel) .08
Cll (D-subminiature) -.03
E2 (N SB ) -.03
D-16
Relays
Normalized to;
Reed Relay
Gp Environment
Military Quality
D-17
VARIABLES IN THE EQUATION
D-18
VARIABLES NOT IN THE EQUATION
Multiple R .74502
R Square .55506 R Square Change .00293
Adjusted R Square .52311 F Change 1.28497
Standard Error 1.90839 Significant F Change .2584
D-19
MISSION
OF
ROME LABORATORY