IPC/JEDEC-9703

Mechanical Shock Test Guidelines

for Solder Joint Reliability
March 2009

A standard developed by IPC/JEDEC

g
gl

Association Connecting Electronics Industries

Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT

The Principles of In May 1995 the IPC’s Technical Activities Executive Committee (TAEC) adopted Principles of
Standardization Standardization as a guiding principle of IPC’s standardization efforts.
Standards Should: Standards Should Not:
• Show relationship to Design for Manufacturability • Inhibit innovation
(DFM) and Design for the Environment (DFE) • Increase time-to-market
• Minimize time to market • Keep people out
• Contain simple (simplified) language • Increase cycle time
• Just include spec information • Tell you how to make something
• Focus on end product performance • Contain anything that cannot
• Include a feedback system on use and be defended with data
problems for future improvement

Notice IPC Standards and Publications are designed to serve the public interest through eliminating mis-
understandings between manufacturers and purchasers, facilitating interchangeability and improve-
ment of products, and assisting the purchaser in selecting and obtaining with minimum delay the
proper product for his particular need. Existence of such Standards and Publications shall not in
any respect preclude any member or nonmember of IPC from manufacturing or selling products
not conforming to such Standards and Publication, nor shall the existence of such Standards and
Publications preclude their voluntary use by those other than IPC members, whether the standard
is to be used either domestically or internationally.
Recommended Standards and Publications are adopted by IPC without regard to whether their adop-
tion may involve patents on articles, materials, or processes. By such action, IPC does not assume
any liability to any patent owner, nor do they assume any obligation whatever to parties adopting
the Recommended Standard or Publication. Users are also wholly responsible for protecting them-
selves against all claims of liabilities for patent infringement.
g
gl

IPC Position It is the position of IPC’s Technical Activities Executive Committee that the use and implementation
Statement on of IPC publications is voluntary and is part of a relationship entered into by customer and supplier.
Specification When an IPC publication is updated and a new revision is published, it is the opinion of the TAEC
Revision Change that the use of the new revision as part of an existing relationship is not automatic unless required
by the contract. The TAEC recommends the use of the latest revision. Adopted October 6, 1998

Why is there Your purchase of this document contributes to the ongoing development of new and updated industry
a charge for standards and publications. Standards allow manufacturers, customers, and suppliers to understand
this document? one another better. Standards allow manufacturers greater efficiencies when they can set up their
processes to meet industry standards, allowing them to offer their customers lower costs.
IPC spends hundreds of thousands of dollars annually to support IPC’s volunteers in the standards
and publications development process. There are many rounds of drafts sent out for review and
the committees spend hundreds of hours in review and development. IPC’s staff attends and par-
ticipates in committee activities, typesets and circulates document drafts, and follows all necessary
procedures to qualify for ANSI approval.
IPC’s membership dues have been kept low to allow as many companies as possible to participate.
Therefore, the standards and publications revenue is necessary to complement dues revenue. The
price schedule offers a 50% discount to IPC members. If your company buys IPC standards and
publications, why not take advantage of this and the many other benefits of IPC membership as
well? For more information on membership in IPC, please visit www.ipc.org or call 847/597-2872.

Thank you for your continued support.

©Copyright 2009. JEDEC, Arlington, Virginia, and IPC, Bannockburn, Illinois, USA. All rights reserved under both international and Pan-
American copyright conventions. Any copying, scanning or other reproduction of these materials without the prior written consent of the
copyright holder is strictly prohibited and constitutes infringement under the Copyright Law of the United States.

Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT

 IPC/JEDEC-9703
ASSOCIATION CONNECTING
ELECTRONICS INDUSTRIES ®

Mechanical Shock Test

Guidelines for Solder
Joint Reliability

Developed by the JEDEC Reliability Test Methods for Packaged Devices

g

Committee (JC-14.1) and the SMT Attachment Reliability Test Methods

Task Group (6-10d) of the Product Reliability Committee (6-10) of IPC
gl

Users of this publication are encouraged to participate in the

development of future revisions.

Contact:

IPC
3000 Lakeside Drive, Suite 309S
Bannockburn, Illinois
60015-1219
Tel 847 615.7100
Fax 847 615.7105

Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT

 This Page Intentionally Left Blank
g
gl

Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT

March 2009 IPC/JEDEC-9703

Table of Contents
1 SCOPE ...................................................................... 1 6.6 Failure Analysis .................................................. 12
6.7 Reporting Recommendations .............................. 12
2 APPLICABLE DOCUMENTS .................................... 1
2.1 IPC ......................................................................... 1 7 COMPONENT MECHANICAL SHOCK
ASSESSMENT ........................................................ 12
2.2 American Society for Testing and Materials ....... 1
7.1 Component Assessment – General
3 TERMS AND DEFINITIONS ...................................... 1 Considerations ..................................................... 12
4 USE CONDITIONS OVERVIEW ............................... 2 7.2 Component Board Testing Recommendations ... 12
4.1 Scope and Application of UCs ............................. 2 7.3 Correlation Criteria and Validity of
Test Setup ............................................................ 13
4.2 Capturing UC Data ............................................... 2
7.4 Equipment Recommendations ............................ 13
4.3 Methods for Developing UCs .............................. 2
7.5 Testing Flows ...................................................... 13
4.3.1 Approach 1 – UC Derivation Approach .............. 2
7.5.1 Characterization Testing ..................................... 13
4.3.2 Approach 2 – Leveraging Existing Data to
Usage Model ......................................................... 3 7.5.2 Qualification Testing ........................................... 14
4.3.3 Torture Tests .......................................................... 3 7.6 Failure Analysis .................................................. 15
7.7 Reporting Recommendations .............................. 15
5 SYSTEM TESTS ....................................................... 4
5.1 Shock Test Equipment .......................................... 4 8 METRICS FOR MATCHING TEST ......................... 15

5.1.1 Drop Testers .......................................................... 4 8.1.1 Acceleration Based Method ................................ 15

5.1.2 Shock Machines .................................................... 5 8.1.2 Board Strain Based Method ............................... 16
5.1.3 Inclined Impact ..................................................... 5 Annex A Sample Data Reporting Format ............... 20
5.1.4 Fixturing ................................................................ 6
g

Annex B Metrologies .............................................. 22

5.2 Correlation Criteria ............................................... 6
gl

5.2.1 Unpackaged System Input Correlated to Annex C Shock Failure Analysis of

Electronic Components ......................... 25
Packaged System Environment ............................ 6
5.2.2 Unpackaged System Input Correlated to Annex D Suggestions for Selecting
End User Use-Conditions ..................................... 7 Sample Size ............................................. 27
5.3 Test Recommendations ......................................... 7 Annex E Suggestions for Finite Element
5.3.1 Testing Sample Recommendations ....................... 7 Analysis in Mechanical Shock .............. 29
5.4 Test Flow ............................................................... 8
5.4.1 Packaged System Testing ..................................... 8 Figures
5.4.2 Unpackaged System Testing ................................. 9 Figure 4-1 Typical UCs where Mechanical Shock
May be Expected for a Few Representative
5.5 Post Test Analysis ................................................. 9 Products .............................................................. 3
5.6 Testing Output and Report Recommendations .. 10 Figure 5-1 Example of the Time History Response
5.6.1 General Considerations ....................................... 10 Showing the Input from the Shock Table
(trapezoidal shape) and the Response at
5.6.2 Electrical Testing ................................................. 10 Two Locations on the PCB ................................. 7
5.6.3 Failure Analysis Considerations ......................... 10 Figure 5-2 Example of Shock Response Spectra (left)
5.6.4 Reporting Recommendations for Test Derived from the Time History ............................ 8
Development ....................................................... 10 Figure 6-1 Illustration of a Board Level Test on a
Hard Plate ......................................................... 11
6 SYSTEM BOARD LEVEL TESTING ...................... 11 Figure 6-2 Illustration of a Board Level Test on a Rigid
6.1 Board Testing Background ................................. 11 Fixture ............................................................... 12
6.2 System Board Testing Recommendations .......... 11 Figure 6-3 Illustration of a Boundary Condition Simulator
(BCS) ................................................................. 12
6.3 Correlation Criteria and Validity of
Test Setup ............................................................ 11 Figure 7-1 Flow Chart for Characterization Test ................ 14
Figure 8-1 Basic Model for Board Level Testing
6.4 Equipment Recommendations ............................ 11
(Assumes Worst Case Amplification
6.5 Test Flows ........................................................... 11 and no other Interaction with Chassis) ............. 16

i
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Figure 8-2 Example of Correlation of a Board Test Figure E-6 Comparison of Displacement-Time History
to Three Test Systems ...................................... 17 for Experimental Data and FEA Result ............. 37
Figure 8-3 Definition of Board Strain States Plot ............... 17 Figure E-7 Comparison of Board Strain versus Time
Figure 8-4 Comparing Board Bend Modes of Board History for Experimental Data and FEA
in System Chassis and Mounted to a Result ................................................................ 37
Fixed Plate ........................................................ 18
Figure 8-5 Settling Time for a Typical System Level
Test with High Damping and Rapid Decay
Tables
in the Strain ....................................................... 18
Figure 8-6 Typical Time History for a Component Table 4-1 Example of a UC for a Mobile Phone,
Board Test with Minimal Damping .................... 18 (Illustrative Only) .................................................. 3
Figure B-1 Wire (a) Strain Gage Location with Respect Table 7-1 Example of Critical Strain Determination
to Solder Ball at Package Corner. (b) Results. Value indicates number of drops
Orientation of Strain Gages for every at each level. (Shadows indicates electrical
Package Corner ................................................ 22 failure.) ................................................................ 14

Figure B-2 Wire Routing should be done So as Not Table 8-1 Recommended Matching Criteria for
to Affect Results ................................................ 23 Selected Use and Test Conditions ..................... 16
Table A-1 Test Report Recommendations (Equipment
Figure B-3 Wire Artwork Showing Strain Gage Feature
and Materials) ..................................................... 20
and Internal Board Routing ............................... 23
Table A-2 Test Report Recommendations
Figure B-4 Fast Fourier Transform of an Example (Board Assembly) ............................................... 20
System .............................................................. 23
Table A-3 Test Report Recommendations
Figure C-1 BGA Solder Joint Failure Modes ...................... 26 (Test Results) ..................................................... 21
Figure C-2 Area Array Cross-Section Diagram ................... 26 Table B-1 Recommended Scan Frequencies
Figure E-1 Coarse Mesh of Plate ....................................... 29 for Various Systems ........................................... 23
Figure E-2 Fine Mesh of Plate ............................................ 29 Table D-1 Sample Size Estimate for a Single Failure
Mechanism ......................................................... 28
Figure E-3 Rack Mounted Test Board ................................ 30
Table E-1 Full, Symmetric, and Asymmetric
Figure E-4 Simplified Satellite Antenna Model ................... 31 Meshes of Plate ................................................. 30
g

Figure E-5 Results for Simplified Satellite Antenna Table E-2 Comparing Implicit and Explicit
gl

Model ................................................................. 32 Dynamic Solutions .............................................. 33

ii
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Mechanical Shock Test Guidelines

for Solder Joint Reliability

FOREWORD Board (PCB) assemblies from system to component level.

This publication has been drafted as part of a joint working The three main categories discussed within this document
group between the IPC 6-10d task group and the JEDEC follow:
JC14.1 subcommittee. It is intended to be used as a general
guideline for mechanical drop and shock testing. It is also 1. Methods to define mechanical shock use-conditions.
encouraged that future application specific mechanical 2. Methods to define system level, system board level and
shock standards and specifications for IPC and JEDEC component test board level testing that correlate to the
should apply the methods proposed herein. use-conditions.
3. Guidance on the use of experimental metrologies for
INTRODUCTION mechanical shock tests.
With the growth of electronics and the increased accessibil-
ity and portability, drop shock and other mechanical 2 APPLICABLE DOCUMENTS
impacts are increasingly a concern. This document
attempts to improve upon past mechanical shock test meth- 2.1 IPC1
ods, and tie test conditions back to the use-conditions. A
method is proposed such that regardless of what level (sys- IPC-6012 Qualification and Performance Specification for
tem, board assembly, simplified single component board Rigid Printed Boards
testing, etc.) testing is conducted, there should be a corre-
IPC-9252 Requirements for Electrical Testing of Unpopu-
lation back to the use-condition. In order to fulfill this goal,
lated Printed Boards
additional metrologies are introduced to aid in these corre-
lations.
g

2.2 American Society for Testing and Materials2

Following the requisite introductory sections, the concept
gl

ASTM D3332 Standard Test Methods for Mechanical-

of use-conditions is introduced and suggestions are made
Shock Fragility of Products, Using Shock Machines
on how use-condition data may be acquired and applied.
Next, the testing methods for fully assembled systems are 3 TERMS AND DEFINITIONS
introduced. Options for test conditions are discussed and
For the purposes of this document, the following terms and
the data that should be collected is outlined.
definitions apply:
Testing of subassemblies and components mimics actual
Accelerometer An electronic device that converts
use configurations less than testing of fully assembled sys-
tems; however, the next two document sections outline mechanical deformation of an internal structure caused by
considerations to ensure that testing done at these levels acceleration into electrical signals that can be sensed by a
remains relevant to the intended use-condition. data acquisition system.

Specific metrics to aid in correlations are outlined in Sec- Component Board Level Test A test conducted on a sim-
tion 8. The informative annexes that close the document plified test board which contains only one type of compo-
discuss the common considerations of all mechanical shock nent, although multiple samples of the same component
testing methods. These include a sample reporting format may be present. This board need not be similar to the final
for test data, use and application of strain gages, acceler- system board.
ometers, and high speed photography. A section on failure ODM Original design manufacturer.
analysis is given. Lastly, a review of finite element meth-
ods that may be applied to mechanical shock analysis is OEM Original equipment manufacturer.
given to aid in more in-depth study of shock problems. PWB Printed wiring board, also known as printed circuit
1 SCOPE
board (PCB).
This document establishes mechanical shock test guide- SRS A shock response spectrum is an analysis of the
lines for assessing solder joint reliability of Printed Circuit acceleration output signal into the frequency domain.

1. www.ipc.org
2. www.astm.org

1
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Strain Gage A sensor of mechanical deformation consist- change over time, and/or UCs may become void and/or
ing of one or more small serpentine structures whose elec- new UCs may become appropriate to capture. Periodic
trical resistance is changed by deformation of the material review is suggested.
to which the gage has been attached.
System Board Level Test Test of a system board with all 4.2 Capturing UC Data Different parameters should be
major components being represented either by the actual captured in order to completely define the use-condition. A
components or equivalent test devices. mobile phone end-user drop UC example is given in Table
4-1. Number of drops, drop height, drop surface and drop
System Level Test Test of a fully assembled system orientation together completely characterize the end-user
including chassis or a major subassembly that is received drop behavior.
by the end customer as a unit, (e.g., server modules). This
may also include the packaging used during shipment of 4.3 Methods for Developing UCs Various methods may
the system. be used to define product or product segment UCs, as listed
UC Use Conditions, a description of the expected events below. Thinking about the best approach to define UCs is
and frequency of occurrence during the life of the product. often a useful starting point.
• If the UC has a strong impact on certification require-
4 USE CONDITIONS OVERVIEW ments, it may be appropriate to collect actual data.
Historically, shock testing has been based on a few stan- • If a new product segment is similar to an existing seg-
dardized acceleration conditions. This approach, while ment, it may be appropriate to scale existing UCs to the
simple, is not optimal for the following reasons: new product segment.
• One set of reliability requirements may not apply to all • If time or budget is limited, collaboration of data sharing
market segments. with existing UCs may be appropriate.
• Testing to standards based tests, or military-based speci- • Existing industry data may be leveraged based either on
fications may be unrealistic since it may not be cost effec- expected differences to the target application or may often
tive for products with more mild requirements. be based on customer/user expectations.
g

• Accelerated tests with fixed pass/fail requirements offer Finally, if existing data offers insight into specific stages of
little insight into the likely lifetime of a product in the the UC (e.g., the human handling portion of a shipping
gl

actual end-user environment. shock study), judgment can often be used to extrapolate
This proposal employs market segment based use- those UCs to a new application.
conditions (UC). Market segment based use-condition test-
ing offers several advantages: 4.3.1 Approach 1 – UC Derivation Approach In this
approach direct data collection is used to derive the UCs.
• A better understanding of robustness and reliability for a
This may be accomplished using user surveys, field obser-
specific product segment or for products and components
vation of handling behavior, data logging tools, or other
in a specific application.
direct means. There are some limitations that should be
• Avoidance of over-engineering to meet a standardized considered in adopting this approach:
acceleration condition, which could add cost.
• Behavior will vary by distribution channel.
• Scalability across multiple product form factors; meaning,
• Uncontrollable variables may be missed or may skew the
portions of a usage behavior from one product type might
results.
be applicable to usage behaviors of a new product type.
• Cost may be significant to obtain statistically valid data.
4.1 Scope and Application of UCs For the scope of this
Example of Approach 1: For the example of shipping
document, the most valuable use-conditions are those that
shock/drop, data collection is a viable and useful study to
have direct impact on interconnection (e.g., solder joint)
undertake, while being aware of its limitations mentioned
reliability.
above. Hardware is readily available in the industry to
Applicable UCs can typically be categorized into the major enable data logging of shipping shock/drop conditions.
stages of a product’s life cycle. Those stages include: Some variations and modulators of results are to be
assembly, shipping, storage, end-user environment and pos- expected, including:
sibly returns and repairs due to issues in the field. Figure • Development level of country.
4-1 gives examples of where these events may occur for
• Road conditions, use of automated handling.
several distinct system types. It should be expected that the
types of events seen during the product life cycle may vary • Shipping method.
based on the market segment. Furthermore, UCs may • Expedited air delivery, truck, train, etc.

2
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Mechanical Drop / Shock Use Conditions Flow

OEM/ODM End-user environment,
Assembly Shipping Storage Assembly including field returns

Shipping Shock Handling Drop / Shock

Desktop

Shipping Shock Handling Drop / Shock User Drop

& Shock
Mobile PC
g
gl

Shipping Shock Handling Drop / Shock

User Drop
Handheld & Shock
IPC-9703-4-1

Figure 4-1 Typical UCs where Mechanical Shock May be Expected for a Few Representative Products

Table 4-1 Example of a UC for onto the docking station), product design (e.g., form fac-
a Mobile Phone, (Illustrative Only)
tors) and the environment (concrete floor vs. laminate desk
UC Value Justification surface).
Number of Drops Three drops Based on expected use
An understanding of specific usage scenarios will lead to
Drop height 1.50 meters Based on human
factors information identifying critical use-condition events of interest. Exist-
ing literature data (from publications) can often be lever-
Drop surface Concrete Worst-case assumption
aged to define the UCs that are relevant to a given market
Drop orientation Face-down Worst-case assumption
segment.
Using data loggers attached to accelerometers inside the Example of Approach 2: Portable music player or barcode
shipping box, one can determine the height and number of scanner manufacturers may have published certain drop
drops experienced by a single package during shipment. By heights, e.g., 1 meter, corresponding to waist height (rel-
repeating this shipping study with multiple routes and con- evant to their expected usage model). This data can be used
ditions, it is possible to develop a model of the shipping to define a drop height of a mobile phone based on human
use-conditions. figure proportions (e.g., ear height is ∼1 .5X waist height).

4.3.2 Approach 2 – Leveraging Existing Data to Usage 4.3.3 Torture Tests ‘‘Torture tests’’ such as a 762.0 mm
Model A usage model describes the interactions between [30.0 in] drop or ‘‘coffee spill’’ test for Mobile PCs, are
a user and a product. It is based on the user needs (e.g., a sometimes used to evaluate performance against perceived
FedEx® package handler vs. a corporate employee), user or observed severe behavior and usage models. In most
behaviors (e.g., drop on the floor, hard dock of a laptop instances, torture tests may not be derived from common

3
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

UC model or behavior. For torture tests, or other extreme supplied with a motor hoist to automatically raise and
usage models, care must be taken to balance reasonable lower the drop arm with package in place, a micro switch
expectations of performance against device survivability to preset the drop height for repeatable drops, a drop height
and acceptable costs necessary to enable increased robust- indicator, and a remote hand switch to actuate the system.
ness and reliability. The tester should incorporate a small vertical drop prior to
drop arm swing to ensure that the package is not subjected
5 SYSTEM TESTS to rotation. Packages may be released from any angle but
The purpose of system level shock testing can be to either special procedures may be required to assure landing in the
demonstrate the ability of the product to survive relatively same position. The advantages of a swing arm tester is that
infrequent, nonrepetitive shocks encountered during the it is fast and easy to use, provides for easy test article set
intended UCs or to determine the products fragility level so up, and is relatively inexpensive. They are easy to install
that appropriate packaging can be designed. The goal of and require no facility modification. Their disadvantage is
system testing is to replicate the UC environment utilizing that test articles are limited in size and weight and there is
laboratory equipment. These tests should be correlated to a minimum drop height restriction.
the UC environment.
Drop Platform Testers: Drop platform testers are typically
The definition of a system may be a fully assembled sys-
tem as well as sub-systems. (e.g., server modules, board used when the package is of moderate size and weight
assembly handled out of system, etc.) accommodating large products and packages weighing up
to 227 Kg [500 Lbs]. They are capable of performing drops
System tests may be performed on either packaged or as low as 2.5 cm [1.0 in] and as high as 1.8 m [6 ft]. They
unpackaged systems. Packaged system tests are usually are designed to ensure the flattest impacts possible but, in
performed to demonstrate that the system and packaging addition, they allow drop testing to be performed on the
can withstand shocks that may be encountered in the ship- package’s edges and corners for complete performance
ping and distribution environment. Unpackaged system evaluation. The drop platform is raised and lowered by an
tests are performed to either ensure the system has enough electric hoist and released via an air actuator. During the
robustness to withstand handling shocks in a cost effective drop test stage, the drop platform is accelerated downward,
shipping package or to ensure the system can survive shock
g

thereby allowing the package to free fall while the drop

events experienced in a typical end-user environment.
gl

platform descends into slots below the impact surface.

The major factors to consider when defining a system Rebound of the drop platform is prevented, usually by a
shock test are as follows: hydraulic damper at the bottom of the slots. The advan-
• Equipment that provides the appropriate environment. tages of the drop platform testers are that they can accom-
modate large packages, perform over a wide range of drop
• Input profiles that are correlated to use-condition profiles.
heights, and provide better guidance of edges and corners
• Requirements that ensure the proper test article configu- than other types of drop testers. The disadvantages are that
ration and desired confidence level. they are relatively expensive and often require facility
• Procedures that ensure proper test conditions and order of modifications to allow for the platform to drop below the
testing. impact surface. Because of the large size and heavy
• Post test analysis that provides relevant and meaningful weights, they can pose a personnel safety risk and often
information. require protective enclosures with safety interlocks.
• Reporting recommendations that include data necessary
for follow-on testing or condition evaluation. Large Package Quick Release: The quick release drop
testers are designed to drop test larger, bulkier packages
5.1 Shock Test Equipment
than those possible with drop platform testers. They can
perform repeatable free-fall drop tests (flat, edge and
comer) on packages of virtually any size and shape and can
5.1.1 Drop Testers Package drop testers are simple
accommodate packages weighing up to 1360 Kg [3,000
machines that allow for repeatable dropping of packages in
lbs]. Drop test height is limited only by the lifting device
terms of height and orientation. Package drop testers can be
either swing arm testers, drop platform testers, or large used (forklift, winch, etc.). The quick release drop testers
package quick release testers. consist of a quick release mechanism, a lifting ring, and a
foot switch. Straps are wrapped around the test item from
Swing Arm Testers: The swing arm tester is typically used a sling that is attached to the lifting ring, which is placed
when the package is small or light weight and the drop in the jaws of the quick release mechanism. The quick
height is relatively large. Drop heights typically range from release mechanism and package are then raised to a prede-
30 to 150 cm [12 to 60 in]. The swing arm tester is usually termined height by a forklift, winch, or other lifting device.

4
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

The foot switch activates a solenoid in the mechanism, free fall and the pressure in the plunger impact cylinder(s)
causing the latch jaws to open and the test package to be control the amplitude and duration of the square wave
released. pulse. The disadvantages of the large format shock test
system are the cost and the limitations of acceleration pro-
The advantages of the quick release drop tester is that it is
files. The acceleration level is limited by the height of drop
relatively inexpensive compared to the other methods, pro-
and free fall acceleration. Higher acceleration levels can be
vided the lab already has the lifting device. If the lab does
reached by the use of special adapter kits or by the use of
not have a heavy duty lifting device then the cost increases
High G shock machines.
considerably. The major disadvantage of this method is that
it is difficult to maintain orientation of the object during the High Acceleration (G) Shock Testers: The high G shock
drop. testers are similar to large format shock testers but are out-
fitted with a bungee cord acceleration system which more
5.1.2 Shock Machines Shock tests are used to accu- than doubles the velocity change capability. When using
rately measure the fragility of products and to evaluate them in free fall mode-without the acceleration system-the
protective packaging by applying a controlled acceleration maximum payload can be as high as 1,360 kg [3,000 lbs].
profile. This test data is key information necessary to While in the accelerated mode, the maximum payload is
ensure that the product is capable of withstanding its ‘‘real limited to 113 kg [250 lbs]. They have a smaller test plat-
world’’ environment. Shock machines are used to accu- form and can only accommodate test items with a footprint
rately and repeatedly produce shock profiles that are corre- dimension up to 46 cm [18 in] square. A maximum veloc-
lated to the environment that a system experiences in its ity change of 25 m/sec [82 ft. /sec] and a peak acceleration
shipping and in-use environments. Shock testing with con- of 7,500 G can be obtained with specimens weighing up to
trolled shock inputs is essential to discovering and under- 45 kg [100 lbs].
standing the risk area of a product.
High G Shock Amplifiers: Another method of producing
Shock Test Systems – Large Format Shock Testers: Large high G shocks is the use of a shock amplifier. Shock ampli-
format shock test systems have the ability to test items fiers produce high acceleration shocks on a spring loaded
weighing up to 1,134 kg [2,500 lbs] with a footprint table. One such system, the mousetrap amplifier, consists
dimension up to 95 x 115 cm [37.4 x 45.3 in]. They can of a base plate mounted on the top surface of the shock
g

perform a wide variety of shock tests to a maximum accel- machine carriage, a specimen mounting carriage which is
gl

eration of 600 G. The system consists of a shock table that supported by soft springs and a resilient decelerating pad
is guided by two solid guide rods. The two guide rods are between the shock machine and its carriage. As the main
rigidly attached to a massive, isolated base that helps to carriage accelerates down, impacts upon its resilient pad
attenuate any shock energy that would otherwise be trans- and decelerates up, the specimen mounting carriage contin-
mitted to the surrounding building through the laboratory ues to travel downward (or in the opposite direction of the
floor. Also mounted to the base are two damage boundary main carriage) as they collide. The resultant acceleration at
programmers, which allow the operator to generate both impact is much greater than would be achievable if only
half sine and trapezoidal shock pulse waveforms. the main carriage impacted against a stationary base. The
At the start of a shock test the shock table is raised until it mousetrap shock amplifier can be used for two primary
reaches the programmed drop height. Upon initiation of the purposes: as an amplifier to produce increased velocities or,
drop sequence, the lifting mechanism is retracted clear of because the mousetrap amplifier is supported on springs,
the shock table and the brake system releases, allowing the the springs act as a ‘‘mechanical filter’’ which serve to
shock table to free fall. The shock table impacts the dam- reduce the ringing associated with short time duration
age boundary programmer(s), creating the desired shock pulses.
pulse waveform. As the shock table rebounds the brakes Vibration Machines: Electrodynamic, hydraulic, and
reengage to prevent any secondary impacts. The param- pneumatic vibration machines can also be used to provide
eters that can be controlled in the operation of the shock a source of shock pulses as long as the pulse requirements
system are drop height and programmer stiffness. The two do not exceed the force and motion capabilities of the
types of damage boundary programmers commonly used machine. These pulses are typically half sine pulses and
are elastomeric impact cylinders and gas actuated plunger can be on the order of a few milliseconds in duration and
assemblies. The elastomeric impact cylinders are used to less than 100G in amplitude. The major drawback to this
generate half sine profile shock pulses. The amplitude and method is the pre- and post-pulse conditioning required to
duration of the half sine pulse can be controlled by the drop generate the shock pulse.
height of the table free fall and the geometry and durom-
eter hardness of the elastomeric cylinders. The gas actuated 5.1.3 Inclined Impact The purpose of the inclined
plunger assemblies are used to generate square wave or impact test is to demonstrate that the packaging and pallet-
trapezoidal wave shock pulses. The drop height of the table izing methods, wrapping and strapping, used on the

5
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

product will protect them from damage due to side impacts. tests that are used to validate the performance of the unit
Inclined impact simulates pallet collision with walls, truck in end user drop would be fixtured directly to the table so
bulkheads, or other loads experienced during loading of a that the entire surface is exposed to the shock input.
truck, airplane, or transport of product in a storage area.
The inclined impact test apparatus consists of a wheeled 5.2 Correlation Criteria System test input profiles are a
carriage, guided by steel rails, upon which the test item is critical component of test procedures. These test profiles
placed. The rails are inclined at 10 degrees and there is an must accurately represent or be correlated to the use-
impact ‘‘backstop’’ at the low end, with its surface perpen- condition that is being represented. These profiles are
dicular to the carriage’s direction of travel. When the car- straightforward in the case of packaged system test condi-
riage is pulled up the rails and released, it freely rolls down tions that are representative of transportation and distribu-
the track and the specimen impacts the backstop. tion handling. The test profiles are specified as a number of
drops in different orientations from specified heights that
5.1.4 Fixturing Test article fixturing refers to the method are representative of measured or observed events for pack-
of holding the test article while subjecting it to the shock ages of similar size and weight and can be performed using
environment. It needs to simulate the boundary conditions drop test equipment.
of the use environment as much as possible. Care must be The test profiles are more complex for unpackaged system
taken to not suppress the failure mode or vibration mode of testing. The unpackaged system test profile must be corre-
interest. lated to the environment that is being represented. In the
Drop Test Fixturing: The fixturing for drop testing of case of the test that is conducted in order to ensure the sys-
packaged systems is the simplest since it only requires tem is sufficiently robust to withstand shocks in a cost
holding the package prior to the drop. The fixturing should effective shipping package, the test profile must be corre-
not interfere with the motion of the package at impact, as lated to the response that the system experiences in its
this will impart unwanted loads into the test article. The packaging during the transportation and distribution envi-
standard practice for orienting the package before free fall ronment. If the purpose of the test is to ensure the system
drop is to ensure the desired impact location is aligned with can survive shock events experienced in a typical end user
the center of gravity of the packaged item. This is usually environment, the test profile must be correlated to the
easy to achieve in flat edge drops but may require careful response that the system experiences due to drops, kicks,
g

balancing for edge or corner drops. and bumps associated with the end user use-conditions.
gl

The fixturing for drop testing of unpackaged systems is Several parameters are of interest when matching test con-
considerably more difficult. The system must be con- ditions to use-conditions. Most real world shock events,
strained to impact at the desired location while preventing i.e., use-conditions, are quantifiable with acceleration pro-
any rotational motion of the test article. In addition, the files. Drop events can be characterized by system sizes,
fixturing must not impede the free fall motion of the sys- drop heights, and impact surface. These events can be
tem. It must also release the test article before impact so as quantified by the acceleration profile of the system and its
to avoid any attenuation of the shock input upon impact. characteristic parameters, e.g., amplitude, pulse shape,
duration, frequency content, etc.
Shock Table Fixturing: The method of fixturing the
unpackaged system to the shock table should represent the 5.2.1 Unpackaged System Input Correlated to Packaged
use-condition that the test is simulating. Unpackaged sys- System Environment The acceleration profile experi-
tem tests that are used to validate robustness to survive enced by a package in a drop environment can typically be
packaged drops should be fixtured to represent the struc- represented by a half sine pulse. However, the acceleration
tural support that the system would have in the packaging. profile experienced by the system inside the packaging is
System packaging typically supports the unit only at the modified by the crushing of the packing material intended
corners or along the edges, and so the fixturing should only to protect the product. This profile is commonly repre-
support the system at the same locations. Blocking may be sented by a trapezoidal or square wave. The trapezoidal
required to hold the system off of the table so that it only pulse is characterized by the velocity change and the faired
contacts at these support areas. The blocking material acceleration. Faired acceleration is defined as the average
should be selected to minimize the amount of damping that of the oscillation amplitudes between the first time that the
is imposed between the table and the system. Typical mate- response goes negative and the last time that the response
rials are metal (usually aluminum or steel) but sometimes goes positive. The velocity change is the area under the
wood or a plastic material is required. The advantage that time history curve and is typically the faired acceleration
wood or plastic offers is that is can be easily machined or level times the pulse width. The trapezoidal input pulse can
molded to the profile of the test item but these materials are be correlated to measured responses pulses either by using
much more compliant, and therefore would provide consid- time domain comparisons or frequency domain shock
erable damping and signal attenuation. Unpackaged system response spectrums.

6
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Time Domain Correlation: The trapezoidal input pulse pulses. This technique accounts for the acceleration ampli-
should envelope the majority of the pulse peaks (80% or tude and velocity change but does not consider the fre-
more) measured at several locations in the system near the quency content of the pulse.
contact points with the packaging material. This technique
accounts for the acceleration amplitude and velocity Frequency Domain Correlation (SRS): The SRS of the
change but does not consider the frequency content of the half sine input pulse should envelope the SRS of the
pulse. See Figure 5-1. responses measured at several locations in the system in
the direction of the input (drop or bump direction). The
spectra used for the measured responses should be the
100 composite spectra for both positive and negative directions.
The analyses should be performed for a Q = 10 at a
80
Aceleration (in g’s)

sequence of natural frequencies at intervals of 1⁄6 octave or

60 smaller to span at least 5 to 500 Hz. This method accounts
for the amplitude, frequency, and energy in the measured
40
input that is likely to cause damage in the system.
20
5.3 Test Recommendations
0
0 2 4 6 8 10 12 14 16
-20 5.3.1 Testing Sample Recommendations
Time (milli-seconds)
IPC-9703-5-1 Test Sample Condition/Configuration: Samples need to
Figure 5-1 Example of the Time History Response represent final system configuration in terms of mass, stiff-
Showing the Input from the Shock Table (trapezoidal ness, geometry and peripheral hardware. The peripheral
shape) and the Response at Two Locations on the PCB hardware that is normally supplied in the system must be
included in system tests because the mass, stiffness, and
Frequency Domain Correlation: The shock response spec-
structural characteristics of this hardware can have a sig-
trum (SRS) of the trapezoidal input pulse should envelope
nificant impact on the dynamic performance of the system
the SRS of the responses measured at several locations in
in shock testing. As an example, any add-in cards that plug
g

the system near the contact points with the packaging

into card guides on the processor mother board and also
gl

material. The spectra used for the measured responses

interface to the chassis side wall should be included in the
should be the composite spectra for both positive and nega-
system test, as they will change the dynamic response of
tive directions. The analyses should be performed for a Q
the mother board to the shock environment and the result-
= 10 at a sequence of natural frequencies at intervals of 1/6
ing stresses on the component solder ball attachments.
octave or smaller to span at least 5 to 500 Hz. This method
accounts for the amplitude, frequency, and energy in the Test samples should be randomly selected from production
measured input that is likely to cause damage in the sys- materials that make up the test system. Selecting all mate-
tem. This is illustrated by the examples shown in Figure rials from the same production lot will not take into
5-2. account production variability that may have a significant
effect on the dynamic performance of any structural hard-
5.2.2 Unpackaged System Input Correlated to End User ware in the system. For example, differences in weld qual-
Use-Conditions The acceleration profile experienced by ity in chassis manufacturing will alter the dynamic
an unprotected system in a drop, kick or bump environment response of the motherboard mounted in the chassis. Not
can typically be represented by a half sine pulse. The half all of the hardware used in system testing need be of opera-
sine pulse is characterized by the pulse amplitude and the tional quality. Mechanical surrogate components can be
velocity change. The velocity change is the area under the used in place of those not being electrically tested, but the
time history curve and is calculated as twice the amplitude configuration, layout, and attachment methods must be
times the period divided by π. The half sine input pulse can consistent with their operational counterparts. As an
also be correlated to measured response pulses either by example, daisy chain connectivity of CPU sockets may be
using time domain comparisons or frequency domain required for in-situ electrical monitoring of solder integrity
shock response spectrums. Methods similar to those during shock testing. However, the geometry, mass, and
described in the previous section can be used. stiffness of the interposer used in the CPU socket must be
identical to the processor that it is replacing.
Time Domain Correlation: The half sine input pulse
should envelope the majority of the pulse peaks measured Some systems may be available in multiple configurations
at several locations in the system in the direction of the and the cost required to test all configurations may be pro-
input (drop or bump direction). The duration of the pulse hibitive. In this case it is acceptable to test a smaller num-
should envelope the maximum duration of the response ber of systems deemed to be worst case configurations. The

7
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

ps modules_8ms In System Time History

50

40
ps modules_8ms In System Shock Spectra
120
30
100
Aceleration (in g’s)

80 20

60
10
40

20 0

0
-10
10 100 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055
SDOF Frequency (hz)
gilbertplngr_BottomDown_1.csv
gilbertplngr_FrontDown_1.csv
gilbertplngr_LeftDown_1.csv
gilbertplngr_RearDown_1.csv
gilbertplngr_RightDown_2.csv
gilbertplngr_TopDown_1.csv
mccar_bladrun_BottomDown25G1.csv
mccar_bladrun_Front25G1.csv
mccar_bladrun_Leftside25G1.csv
mccar_bladrun_Topdown25G1.csv
g

spec25g
gl

spec40g
spec50g IPC-9703-5-2

Figure 5-2 Example of Shock Response Spectra (left) Derived from the Time History

designation of worst case configuration is an engineering packaging configurations. Unpackaged system tests are
call that must be addressed on a case by case basis and will performed to determine the fragility level of the product, to
depend on the purpose of the test and the anticipated fail- ensure the system has enough robustness to withstand han-
ure modes. dling shocks in a cost effective shipping package, or to
ensure the system can survive shock events experienced in
All test articles used in system testing, including peripher-
a typical end-user environment.
als, should be labeled for ease of identification and report-
ing. This is especially important for investigations into 5.4.1 Packaged System Testing Shock tests for these
causes of failures. systems are performed by dropping the fully packaged sys-
Sample Size Recommendations: Not all defects are tem from a prescribed height that is based on the weight
intrinsic to the product, so multiple units should be tested. and configuration of the system. Large systems that are
The number of units included in the test series, the sample mounted on pallets for shipping are normally handled using
size, should balance the cost of the test articles and the fork lifts or pallet handling trucks. These packages are
confidence level with which one desires to know the true rarely dropped on any surface other than the palletized sur-
failure rate of the product. The preferred sample size for face.
x% failure rate at y% confidence level can be computed Unpalletized systems are manually handled and may be
using the statistical approach included in Annex D. dropped on any surface. The test flow for packaged system
testing is therefore dependent on whether the system is
5.4 Test Flow The recommended test flow for system palletized or unpalletized for shipment. Packaged product
testing will vary depending on the type of testing being will normally not be dropped more than once from the
performed. Packaged system tests, performed to demon- required test height, so it is usually acceptable to change
strate the system and packaging can withstand shocks in package components, cushioning material, cardboard, or
shipping and distribution, are performed in representative test sample during the test.

8
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Palletized Packages: The test flow for palletized systems manner which administers acceptable doses of acceleration
is recommended to drop the systems on the palletized sur- to the product, the velocity change can be very large with-
face twice from the full height for the particular package out causing damage. If the critical velocity and critical
weight and then dropping them an additional number of acceleration are both exceeded then damage occurs.
times (usually ten) from half that height. Systems weighing
Damage boundary is assessed by two different shock tests:
less than 45kgs [100 lbs] are usually dropped from a height
a critical velocity change test and a critical acceleration
of 30.5 cm [12.0 in]. Larger systems are dropped from a
test. The product critical velocity change is determined
height of 23.0 cm [9.0 in]. The flat surface drops are then
using shock pulse of any wave form with duration of less
followed by a number of rotational edge drops (usually
than 3 ms. The critical acceleration limit of a product is
four) on each of the four bottom edges. This is performed
determined using a trapezoidal shock pulse with rise and
with one edge resting on an elevated surface and the oppo-
fall time of 1.8 ms, or less. The details of these tests meth-
site end released from the normal drop height for that size
ods are described in ASTM D-3332.
of system.
Reliability Testing: Reliability testing is performed to
Unpalletized Packages: The test flow for unpalletized sys-
demonstrate the robustness of a product and its ability to
tems is recommended to drop the system from a height that
withstand the use-conditions it will be exposed to in its
is determined by the package weight. The required height
expected lifetime. The testing flow for reliability testing
is based on the defined use-condition for typical handling
starts with adjusting the shock table settings to produce the
of systems of that particular weight. The system is dropped
desired acceleration level and velocity change for the test.
once on each of the six surfaces. This is followed by a drop
This is accomplished by mounting a weight comparable to
on at least two corners, one being the manufacturer’s joint
the system weight to the shock machine table and adjust-
corner and the other one the diagonally opposing corner or
ing the drop height and programmer pressure until the
any corner that may be previously determined to be espe-
desired acceleration and velocity change are achieved.
cially vulnerable. The corner drops are then followed by
These table parameters are then used for performing the
drops on the three edges radiating from the manufacturer’s
testing for all orientations of the product under test. The
joint corner. The manufacturer’s joint corner is defined as
shock profile used to simulate the input that an unpackaged
one of the corners of the edge of the packaging where the
system would experience when dropped in its packaging is
g

cardboard is glued or stapled together. Corner drops are

represented by the trapezoidal pulse. This pulse is defined
performed with the package balanced on the corner and the
gl

by velocity change and the faired acceleration (the average

center of gravity of the package directly over the corner.
of the oscillations between the first time that the response
Edge drops are performed with the package balanced on
goes negative and the last time the wave goes positive).
the edge and the center of gravity directly over that edge.
The shock profile used to simulate the input that an
unpackaged system would experience when dropped with-
5.4.2 Unpackaged System Testing Shock tests for
out packaging is the half sine pulse. This pulse is defined
unpackaged tests are performed with the system supported
by its amplitude and pulse width. A typical unpackaged
in a manner that simulates the chassis dynamics as experi-
system test consists of shocking the test unit in each of the
enced in the normal shipping package configuration.
six orthogonal directions two or three times in each direc-
Fragility Testing: Fragility tests, or damage boundary tion. The system is inspected for mechanical damage after
assessment, are performed to choose optimum-cushioning each drop. A typical sample size is a minimum of three
materials for shipping containers or for product designs in units.
the applications such as automobile or handheld. Product
damage is cumulative due to acceleration during transpor- 5.5 Post Test Analysis The primary purpose of system
tation and the velocity change due to free fall drops with- testing is to determine the robustness of the system during
out adequate protection. The velocity change is represented handling operations. The first step is to inspect the system
by the area under the acceleration time history of the shock for mechanical damage caused by the test condition. Pack-
and can be thought of as the energy contained in the shock aged system testing analysis consists of first inspecting for
pulse. The higher the velocity change, the higher the visible damage to the packaging or exterior of the test item.
energy content and therefore the more potential that the The test item is then removed from the packaging and
shock will cause damage to the product. There is a mini- inspected for any hidden damage. The system should then
mum velocity change which must be achieved before dam- be opened and the interior inspected for mechanical dam-
age to the product can occur. This level is called the criti- age. All of this inspection will indicate the potential for
cal velocity change and below this level no damage occurs damage to the electrical components and potential solder
regardless of the input acceleration level. Exceeding the joint damage. Functional test analysis for system testing
critical velocity change does not necessarily imply that typically consists of either operational testing of the test
damage will result. If the change in velocity occurs in a item in the case of a functional system or at least electrical

9
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

continuity testing of the critical components. Functional events were recorded, the time from start of testing in
testing consists of applying power to the system and moni- which any events were detected, and the length of events
toring for any anomaly in the normal start up procedure. detected (how many counts). For further information see
Electrical continuity testing is normally performed at the Annex A.
board level.
The results of shock testing performed with the test article
Failure analysis should be performed on any items showing in its operational mode should be reported also. Any abnor-
degradation during the functional testing or continuity test- malities in performance or start up performance should be
ing. Failure analysis is typically performed at the board reported in order to aid in failure mode analysis and
level or component level and is described in more detail in follow-on failure analysis investigation.
those sections and in Annex A.
5.6.3 Failure Analysis Considerations Failure analysis
5.6 Testing Output and Report Recommendations Re- at the system level mainly involves inspection for mechani-
porting of results from system level testing is critical in cal damage that may be an indicator of damage to the
developing criteria and parameters for follow on activities, electronics or interconnects. The methods used in failure
either for board level tests or package design criteria. The analysis at the board or component level are included in
method of testing as well as the test flow is important in Annex C.
interpreting the results of the test as well as the applicabil-
ity to these follow-on activities. Output response of system 5.6.4 Reporting Recommendations for Test Develop-
testing that is important for specifying board test inputs ment The results from system testing that are necessary
include the shock input profile, board acceleration for test development at follow-on levels are acceleration
response, board strain response, board bend mode at criti- responses of critical components and hardware attachment
cal components and frequency content. points as well as strain measurements made on the printed
circuit boards. Since strain measurements are taken on the
5.6.1 General Considerations The general information
circuit boards, the analysis and reporting of strain
that should be included in all test reports are the configu- responses are identical to the methods used in board test-
ration of the hardware used in the test, all peripheral hard- ing. This information is included in Annex A. Acceleration
responses of the critical components or hardware attach-
g

ware installed in the system, and the method of fixturing

ment points are necessary for development of follow-on
gl

and materials used in the test set up. Photographs of the

test item in both pre and post test conditions should be testing at the board and component level, or to supply the
included. Photographs of the test item fixtured to the test packaging engineer information on the level of protection
equipment should also be included. For further information that should be designed into the packaging. The aspects of
see Annex A. acceleration measurements that should be included in the
report are the magnitude versus time response and the
Instrumentation descriptions and parameters are required in amplitude versus frequency of the signal which is repre-
order to properly interpret the test results. The make, model sented by the shock response spectrum of the signal. For
and, if available, the serial number of all test equipment further information see Annex A.
and instrumentation should be reported. The parameters
used in the test set up are also required for proper analysis Acceleration Magnitude: The peak magnitude of the
of the results. The transducer types and gains are important acceleration response from the system level test must be
as well as the sampling rate used to capture the data. In reported for use in design validation investigations as well
addition to sampling rate the use of any frequency filtering as for developing follow-on testing. The magnitude of the
is required in the report. The conditions under which the acceleration aids in determining the force that will be
test was performed must be reported such as drop height, applied to the circuit board during dynamic loading. The
impact material, programmer pressure in the case of a peak acceleration is used to determine an amplification fac-
shock machine, and orientation of the test article. All of tor to the weight of any large components on the board as
this information is important in interpreting the results of well as to the board itself in order to properly design the
the test or in the ability to replicate the test if the need hardware to survive this loading. For further information
arises. see Annex A.
Frequency Response: The frequency response of the
5.6.2 Electrical Testing The results of any electrical acceleration history is needed in order to assess the damage
testing performed on the test article should be reported. In potential contained in the shock event. The most damaging
the event that in-situ monitoring of the solder ball connec- energy is low frequency energy since it is associated with
tions was performed during the test, the results should the largest displacement of the hardware. A typical cut off
include the location of resistance paths (daisy chains) frequency for shock analysis is 500 Hz but the most dam-
monitored, the result of that monitoring, i.e., how many aging energy is usually contained below 200 Hz. Knowing

10
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

the frequency content of the signal is also important adjacent to the corners. If noncritical-to-function solder
because the structural system responds to energy that is at joints are used, these should be routed in a separate chain
the resonant frequencies of the system. Knowing the reso- from critical-to-function joints.
nant frequencies of the system and the energy in the input Preconditioning: Assembled boards should be precondi-
shock at these frequencies can help in designing system tioned as necessary to the expected use-conditions. This
structural elements that can survive these loads. The fre- may include multiple reflow cycles, moisture soaking and
quency content of the shock input to the board measured at thermal cycling.
the input points can be used in designing test fixtures and
profiles that will stress the board in a representative man- 6.3 Correlation Criteria and Validity of Test Setup It is
ner. For further information see Annex B. expected that the system board test should reasonably cor-
relate to system use conditions. There are two general
6 SYSTEM BOARD LEVEL TESTING methods of matching acceleration and strain. If accelera-
tion based methods are used then the strain should at least
6.1 Board Testing Background The primary goal of a be monitored to ensure the test is not excessive.
system board level shock test is to evaluate a system board
design without the chassis or other system components. 6.4 Equipment Recommendations The board level tests
This may represent two distinct UCs. First, boards may be depend on the board design and its interaction with a chas-
dropped without protection during system assembly. In this sis. The basic concept is to mimic the chassis mounting.
case, the device tested may just be a board or a sub assem- The board can be mounted with the mounting screws onto
bly of the board with some system components. The sec- the standoff on a hard plate which is fixed on the shock
ond, a board level shock test, may reproduce the system table, if the board is a through-hole mount, as shown in
board response in the fully assembled system shock. The Figure 6-1. For rack mounted systems a rigid fixture such
components of the system will affect the motion of the as the one shown in Figure 6-2 may be useful. The board
board. It is important to ensure that the impact of these may be mounted on a boundary condition simulator (BCS)
effects is accounted for in the system board test. which is mounted on the shock table (see Figure 6-3), if the
6.2 System Board Testing Recommendations
board assembly is an integral part of the system and a bet-
ter board response match is desired.
g

Sample Recommendations: It is assumed that the system

board used for testing is either the final shipping board or
gl

is the equivalent board using daisy chain or mechanical

dummy components. If an equivalent board is used it
should match the system board in all essential features,
including:
• Utilize the same BGA pad surface finish and quality.
• Utilize the same PCB material and thickness.
• Be assembled with a process that is similar to that
expected for the system board in terms of the board flux,
reflow profile and board handling.
• Use components that represent the final product configu-
ration materials in terms of mechanical properties, solder
materials, and surface finish; if testing a range of prod-
ucts, a worse case configuration may be used to reduce
testing costs. Figure 6-1 Illustration of a Board Level Test on a Hard
Plate
Quality: The assembled boards should be free from qual-
ity issues that would adversely affect the results. Boards A BCS can be a modified version of the fixtures listed
should be inspected to ensure that the lands are within above or a simplified chassis. A BCS should be tunable for
dimensional tolerances and are free from plating defects. matching tests described in the following sub-sections.
Electrical test and X-ray inspection should be conducted on Adjusting the stiffness of the mountings or the patterns of
all assemblies to verify that no opens/shorts exist, and to the mountings, a board response on the BCS can be tuned.
establish that typical solder joint assembly criteria for Changing the edge bending to change the local stiffness of
voids, bridging and abnormalities are met. the BCS can tune the board response too.
Daisy Chain Design: If electrical monitoring is to be used, 6.5 Test Flows
the daisy chain design should provide coverage for all cor- Qualification Testing: In this test flow, the correlated sys-
ner joints. It is advisable to provide coverage of other joints tem test is performed to replicate system level tests. The

11
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

selected test samples should also be subjected to failure

analysis to confirm results.

6.6 Failure Analysis Normal electrical testing is not

always able to reliably detect failure in solder joints. This
is especially true in shock testing where the joint may open
momentarily but remain in contact after the event passes.
Failure analysis should be used to verify all testing results.
Cross sections of the corner solder joints and stain tech-
niques are both commonly used for these tests. See Annex
C for details.

6.7 Reporting Recommendations Detailed reporting

recommendations are found in Annex A. Typically the
report will contain the strain levels tested, the acceleration
inputs (pulse type, duration, and amplitude), the electrical
Figure 6-2 Illustration of a Board Level Test on a Rigid test results, and failure analysis results. The report or
Fixture attached document should also describe the test setup
including the fixtures used, the board bend mode produced,
frequency and the board strain rate if practical.

7 COMPONENT MECHANICAL SHOCK ASSESSMENT

7.1 Component Assessment – General Considerations

The primary goal of the component mechanical shock test
method described below is to create a procedure to evalu-
ate component reliability using a simplified test board
g

assembly, rather than requiring more complex system level

gl

tests. The mechanical shock response of the test boards are

correlated to the intended system application.

Figure 6-3 Illustration of a Boundary Condition Simulator This document specifies general recommendations for
(BCS) mechanical shock evaluation of a test board. Derivative or
daughter standards may exist for specific applications and
number of drops should be at least as great as that required
should be consulted, where applicable.
for the system level. In some cases the number of drops in
the worst case orientation may be raised to ensure robust- 7.2 Component Board Testing Recommendations
ness. Orientations that show little strain may be neglected
if desired. System boards using daisy chains should be Material Recommendations: Use a test board that reason-
checked for electrical opens before and after the tests. ably replicates the expected system board in terms of
Functional system boards should be boot tested in a simi- mechanical properties, thickness, BGA pad surface finish
lar manner. A minimum of three randomly selected test and quality. The board should be assembled with a process
samples should also be subjected to failure analysis to con- that is similar to that expected for the system board in
firm results. terms of the board flux, reflow profile and board handling.
Use components that represent the final product configura-
Fragility Testing: In this test flow, the correlated test is tion materials in terms of mechanical properties, solder
performed by starting with a selected input level and then materials, and surface finish; if testing a range of products,
checking for electrical failure. If no failure is found the a worse case configuration may be used to reduce testing
input level is raised and the test repeated in this pattern costs.
until failure is detected. Only the worst case orientation is Quality: The assembled boards should be free from qual-
typically used. The number of drops at each level should be ity issues that would adversely affect the results. Boards
kept to a minimum to avoid confounding results with should be inspected to ensure that the lands are within
fatigue damage. This test adds the advantage of being able dimensional tolerances and are free from plating defects.
to assess reliability margins, but is more complicated and Electrical test and X-ray inspection should be conducted on
fatigue damage will accumulate during the test making all assemblies to verify that no opens/shorts exist, and to
interpretation more difficult. A minimum of three randomly establish that typical solder joint assembly criteria for

12
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

voids, bridging and abnormalities are met. See IPC-6012 towers as described in the system test section above,
and IPC-9252 for further information on bare board quali- although any equipment capable of producing repeatable
fication and electrical test. results that meet the correlation criteria is acceptable.
Daisy Chain Design: If electrical monitoring is to be used, 7.5 Testing Flows There are two general testing flows
the daisy chain design should provide coverage for all cor- for component board testing: characterization tests and
ner joints. It is advisable to provide coverage of other joints qualification tests. In the former, the goal is to establish the
adjacent to the corners. If noncritical-to-function solder maximum extent of the component’s capability. For quali-
joints are used, these should be routed in a separate chain fication testing, however, the component’s ability to meet a
from critical-to-function joints. set performance criteria is determined.
Preconditioning: Assembled boards should be precondi- 7.5.1 Characterization Testing A characterization test
tioned as necessary to the expected use-conditions. This flow consists of two testing rounds. The first round is used
may include multiple reflow cycles, moisture soaking and to find the critical strain level at which the component fails
thermal cycling. as measured by electrical monitoring. Subsequent testing
There are other conditions that may also impact the test can be used to establish the component strain capability
results, but at present there is insufficient data to standard- limits in terms of allowable damage. An example test flow
ize their use. Time between reflow and test may have sig- is shown in Figure 7-1.
nificant impact on the reliability of the solder and should As shown in Figure 7-1, the test flow has two distinct
be recorded and reported. Enabling hardware used to con- stages, critical strain determination and validation. In the
strain heat sinks may apply significant load. The strain cliff-finding stage the goal is to determine the maximum
induced by this load should be recorded. The component strain value at which the component is still electrically
board test should represent the impact of both the heat sink functional. As illustrated in Figure 7-1, a test board is
mass and compression loads. If supplemental adhesives tested at increasing level until it fails. Other boards are then
such as underfill or corner-glue are to be used on the sys- tested to confirm the strain level for failure.
tem, then similar materials and processes should be used in Table 7-1 summarizes an example of this process. In this
the component test. Moisture sensitivity level should be example the component is required to survive six drops to
considered if supplemental adhesives are used.
g

correlate to the expected use-condition. Board number 1 is

dropped six times each at 1,000 µStrain and 1,200 µStrain
gl

Elapsed Time after Reflow: Strain-to-failure and/or failure

modes may be sensitive to elapsed time between the before it failed on the first drop at 1,400 µStrain. The sec-
mechanical shock test and the last solder ball reflow. Con- ond board is then dropped six times at 1,300 µStrain and
sequently, shock acceptance criteria may differ, depending five times at 1,400 µStrain before failing. Since board num-
upon elapsed time after reflow. For data comparison, the bers 1 & 2 both failed at 1,400 t µStrain, but only after
same elapsed time after reflow and total number of reflow many drop cycles, the engineer choose to run board 3 at
exposures should be used. 1,400 µStrain. After passing six drops, the board then failed
on the first drop at 1,500 µStrain. To confirm the capability
Post-Stress Test: Depending on the specific production
at 1,400 µStrain, board number 4 is also dropped at 1,400
flow or reliability stress test condition, e.g., burn-in, pre-
µStrain, the required six drops, and fails after two drops at
conditioning, high temperature storage, temperature
1,500 µStrain. This confirms the maximum strain limit of
cycling, etc., the tested devices may produce lower or
1,400 µStrain.
higher strain-to-failure values than the time-zero test
results, and the values expected must be adjusted appropri- It is prudent to define the capability limits of the compo-
ately. Coarsening of solder grain structure following nent based on failure analysis results and electrical data. In
extended temperature exposure may result in an increased order to demonstrate the true capability of the component,
bulk solder failure rate. this test flow includes a validation stage. In this stage, two
strain levels are tested. These have been reduced from the
7.3 Correlation Criteria and Validity of Test Setup Cor- maximum strain limit found in the first stage. Each sample
relation of the component test should be based on the cri- should be dropped the number of times required to meet
teria listed in the correlation criteria section of this docu- the use-condition. After testing, each unit should be electri-
ment. The correlation study should be documented. cally tested and undergo failure analysis to determine the
extent of solder joint cracking. If the extent of cracking
It is expected that the component board test should reason-
exceeds the allowable crack size, then retesting may be
ably correlate to system board in terms of the strain range,
required. Otherwise the strain limit should be set to the
bend mode, frequency and board strain rate if practical.
highest strain level that meets the crack criteria.
7.4 Equipment Recommendations The test equipment One of the advantages of this test flow is that it gives fun-
used for component board testing will generally be drop damental understanding of the failure mechanisms and

13
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Critical Strain
Determination

Set Initial
Input

Shock with
Electrical Monitor

No Increase
Fail? Input

Validation Test validation

samples at strain
level 1
Strain to failure Yes Strain = Maximum
Determined? Functional Strain
Test validation
Samples at strain
level 2
g

Remove board from

test fixture,
gl

install new board Failure Analysis of

all to determine
crack size
Select new
input level

Reduce Strain Both Select level

And Retest with acceptable
Fail crack

IPC-9703-7-1

Figure 7-1 Flow Chart for Characterization Test

Table 7-1 Example of Critical Strain Determination understanding the capability provides some protection
Results. Value indicates number of drops at each
level. (Shadows indicates electrical failure.) against being forced to retest if the system requirements
should change. This test flow, however, is more expensive
Board
# 1000ue 1100ue 1200ue 1300ue 1400ue 1500ue and requires more material, testing and failure analysis
1 6 6 1 time.
2 6 5 7.5.2 Qualification Testing When the user only desires
3 6 1 to know whether a component will meet a particular
4 6 2 requirement, an abbreviated testing flow can be used. This
testing flow consists of testing the component in a pre-
mode. It also allows understanding of capability and mar- scribed manner and at a prescribed level.
gins compared to system requirements to better optimize The testing consists of instrumenting a small number of
component and system design. It can be used to quantify test boards to ensure that the shock table is properly con-
the impact of process and material options. It can also be figured to produce the required response in the component
used to verify Finite Element simulation results. Lastly, board. Once the setup has been validated, then the samples

14
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

are tested and set for failure analysis. If a large sample size mitted to the board, the factor is approximately two and
is to be tested, then test boards with strain gages should be reduces for stiffer chassis. This method has been widely
interspersed among the samples to verify that the setup has used for simply mounted boards. For boards that have more
not drifted. In this test flow, electrical testing should be complex interactions with the system or chassis, other
used to demonstrate that the component meets the require- methods should be used. In this method the accelerations at
ments. A minimum of three randomly selected test samples key components and the board mounting points are the ini-
should also be subjected to failure analysis to confirm tial matching criteria. These should be matched in terms of
results. both acceleration amplitude and Shock Response Spectra
This test flow has the advantage of being very quick and (SRS). Finally, board strains should also be matched, as
limited only to the critical testing levels. It can also be per- board acceleration does not necessarily replicate solder
formed on live product samples rather than requiring daisy joint loading even when matched. Care should also be
chain test vehicles. It does not provide for margin assess- taken to ensure that the natural frequency of the board test
ment. It requires a very clear correlation to the end use- sample reasonably matches that of the system test. This is
condition. The test flow may also require multiple tests if a best ensured by matching the SRS. Lastly, as the board
range of applications are to be supported or if the require- must be mounted to the shock table, limitations in the
ments change after testing has been completed. shock table inputs may also limit the ability to match sys-
tem tests.
7.6 Failure Analysis Normal electrical testing is not
always able to reliably detect failure in solder joints. This Complex System Matching: In many cases the amplifica-
is especially true in shock testing where the joint may open tion method will not be able to sufficiently match the sys-
momentarily, but remain in contact after the event passes. tem response. These include cases where there is signifi-
As such, failure analysis should be used to verify all test- cant interaction with the chassis and board, such as a laptop
ing results. Cross sections of the corner solder joints and or server board with chassis mounted thermal solutions. In
stain techniques are both commonly used for these tests. these cases, the user must decide the extent of matching
See Annex C for details. desired. The same board may be used in multiple systems.
Within the board there are multiple components. It may not
be practical to attempt to develop a test to match the full
g

7.7 Reporting Recommendations Detailed reporting

recommendations are found in Annex A. Typically the range of systems and the response at every component on
gl

report will contain the strain levels tested, the acceleration the board. Matching can vary from the simple matching of
inputs (pulse type, duration, and amplitude), the electrical the worst case corner of the worst case component (worst
test results, and failure analysis results. The report or case as determined by system testing and failure analysis),
attached document should also describe the test setup to matching the entire board response in a single drop ori-
including the fixtures used, the board bend mode produced, entation. To match a system board level test to that for sys-
frequency and the board strain rate if practical. tem level test, it is required that both test board boundary
condition and the shock input be tailored. In cases with
8 METRICS FOR MATCHING TEST
complex interaction, a boundary condition simulator or
chassis surrogate may be used.
It is very important to choose the right metrics for demon-
strating correlation between tests. The available metrics SRS Correlation: One can also improve matching by
include acceleration, board strain, and board deflection. matching the distribution of acceleration as a function of
Each is discussed below. Table 8-1 may be useful provid- frequency, through the use of SRS. In brief the SRS repre-
ing guidance on selecting a method based on the intended sents the acceleration response of the system in discrete
use and test condition. frequency bands. This method is described in detail in a
number of shock testing handbooks and textbooks. These
8.1.1 Acceleration Based Method
books describe many criteria for matching SRS, but in
Amplification Factor Based Matching: In simple systems practice the most conservative methods are used which
it may be possible to use amplification factors to simulate envelope all of the resonant peaks. In Figure 8-2, the SRS
the loading of the system on the board. This method is for three systems are shown for an accelerometer mounted
illustrated in Figure 8-1. In the system, the acceleration (as) near the component of interest. The board test shows an
measured at the mount point of the board can be related to attempt to envelope the system responses. As can be seen
the input acceleration by a simple amplification factor. The in the figure, the response on the board test is larger in
acceleration at the board is usually measured at the heavi- magnitude at nearly every frequency. Importantly, it is
est component (ab). The amplification is related to the natu- larger at the resonance peaks indicated. However, since the
ral frequency ratio and the mass ratio between the board resonant response of the test board is very much larger than
and the chassis. In cases where the chassis bending is trans- the systems, this test may be excessively conservative.

15
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Table 8-1 Recommended Matching Criteria for Selected Use and Test Conditions
Test Condition or
Use Condition Packaged System Unpackaged System System Board Test Component Board Test
Shipping Drop Drop Height Acceleration at contact Acceleration (input Board Strain (amplitude,
Impact Orientation points acceleration profile and mode, and frequency)
Acceleration response at SRS amplitude response)
Package Weight
large components Board Strain
End User Drop N/A Expected drop height
Impact angle
Impact surface
Handling Shock/ N/A Acceleration at contact
Bump points
Orientation
Board Handling N/A N/A Drop Height Board strain (amplitude,
Shock Orientation mode, and frequency)
Input acceleration profile
SRS response

far field motion. Finite element analysis has shown the cor-
relation of the local strain to the stress in the solder joint.
mb ab=FbFsa
Strain gages can also be placed in highly confined spaces
inside systems and do not require a clear line of sight.
Using these advantages, strain gages can be used to corre-
kb late shock tests in several ways. These include: comparison
of strain amplitude, frequency or strain rate matching, bend
mode determination and settling time analysis. Each of
ms as=Fsa mb ab=FbFsa these is discussed below.
g

Board Strain Amplitude Correlation: Board strain ampli-

gl

tude is the most significant correlation metric. It requires a

fine tuning after the other metrics are correlated. For strain
ks kb gages placed on the opposite side of the board from the
a Fsa component, the minimum principal strain is typically used
for correlation, being an indication of tensile loading of the
IPC-9703-8-1 solder joint. Often a strain time history will show multiple
peaks in the strain signal. The highest should be used.
Figure 8-1 Basic Model for Board Level Testing
(Assumes Worst Case Amplification and no other To accomplish matching, other metrics such as bend mode
Interaction with Chassis) and frequency are usually matched first. The table input
can then be increased or decreased to match the strain
Implications of acceleration based matching: Acceleration
amplitude. Care should be taken to ensure the setup
is a popular approach due to the ease of setup and wide
achieves the desired strain at a reasonable input accelera-
spread availability of the data acquisition systems; how-
tion. A setup requiring excessive acceleration inputs com-
ever, for developing tests for solder joints at the system
pared to the system level tests may indicate a poor setup
board or component board level, acceleration poses a
design. Excessively large inputs can lead to issues with
couple of challenges. First, acceleration is a global measure
board mounts and/or heat sink attachment failures and/or
of motion. One cannot determine if a resonance response is
other undesirable and artificial failure modes.
coming from the board response or from the chassis, with-
out more complex modal analysis. Secondly, recent litera- At fine tuning, single peak values in maximum principal or
ture has shown that the flexure of the board is the primary minimum principal strain may be used for direct compari-
cause of solder joint failure rather than inertial loading of son for convenience. Fine tuning board strain amplitude
the device. Matching the acceleration does not indicate that normally can be achieved by changing the shock table
the local flexure is equivalent. As such, strain is the pre- input amplitude.
ferred matching criteria for board level test.
When a circuit board is installed to a chassis, static (or
8.1.2 Board Strain Based Method The use of strain preload) board strains may result. Typically, thermal solu-
gages has several advantages for analyzing solder joint reli- tions, board mounting points and system interface structure
ability. The strain is independent of rigid body motion and can cause the board to deform during installation. These

16
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

800

700 System 1
System 2 Board Test
600 System 3
Board Test
Acceleration (g)

500
System 3
System 1
400

300 System 2

200

100

0
10 100 1000

Frequency (Hz)
IPC-9703-8-2

Figure 8-2 Example of Correlation of a Board Test to Three Test Systems

g

preloads may affect solder joint reliability and the strain by Figure 8-3. A comparison of the angle of the points
measurement. If the time duration and thermal exposure yields the bend mode. Ideally all the points would fall
gl

between static strain events are minimal, the creep effects along the critical angles shown (45° and - 135° for spheri-
during board installation may be ignored, depending on cal, 0° and -90° for planar, and -45° for saddle or twist). In
specific solder alloy, surface finish, solder joint geometry, practice, the data will scatter and the nearest ideal line
etc. Assuming minimal elapsed time between static (from should be selected as illustrated by the colored zones in the
assembly/installation) and dynamic loading, the static and figure.
dynamic strains may need to be combined for proper
matching. The total board strains can be calculated using
the equations below in two steps: In the first step, the com- Min Principal
bined board strains are computed for each grid of the strain
gauge measurement: Spherical

e i (t) = e Si + e Di (t), i = 1, 2, 3
In the second step, principal strains are computed using the Planar
combined grid strains: Max Principal

e1 + e3 1
ε1x2 =
2
± √(e1 − e2)2 + (e2 − e3)2 Saddle (Twist)
√2
Where ei (i=1, 2, 3) represents the combined grid strain,
the superscripts S and D represent static and dynamic,
respectively and ε1 and ε2 are maximum principal strain
IPC-9703-8-3

and minimum principal strain respectively. Figure 8-3 Definition of Board Strain States Plot

Bend Mode Correlation: Board bend mode can be visual- The board strain states plot presents the board bend curva-
ized by using board strain states plot, as shown in Figure ture and bend mode that are directly related to solder joint
8-3. At each strain data point, the minimum and maximum risk. Finite element modeling has shown that, for the same
principal strain are calculated and then plotted as illustrated strain magnitude, the stress is higher in the solder joint for

17
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

spherical and saddle modes than for planar modes. Match- Rise Time: Figure 8-5 shows the time history of strain in
ing in the strain state or bend mode is therefore important a system being dropped. In this system there is a single
in developing a testing method. If a variety of modes are large strain event. In such cases it can be convenient to
observed, then it is advisable to use a conservative mode determine the time from the beginning of the event to the
when matching. strain peak which is called the rise time (tr). If in two sys-
tems the strain magnitude and rise times are matched, then
When comparing the board bend modes, the X and Y
the strain rate is also matched assuming that the response
scales for the plot should be the same. An unequal scaling
can be treated in a linear fashion.
can cause skew of the plot thus skew the bend mode. Over-
lay plot is suggested for better visualization to compare
board test and system test and/or component test. Figure Board Strain Settling Time
8-4 shows an example strain state plot comparing the sys-
500
tem level test data to a proposed system board level test. In tr
this case, the system test is more planar. The proposed sys-

Board Strain (ue)

0
tem board level test matches well in the compressive/ 2.22 2.24 2.26 2.28 2.3 2.32 2.34 2.36
positive portion of the plot and is more conservative (more -500
spherical) in tensile/negative portion of the plot.
-1000

Board Strain States -1500

500 ts
-2000
Time (sec)
0 IPC-9703-8-5
Min Principal Strain (ue)

-500 0 500 1000 1500 2000

Figure 8-5 Settling Time for a Typical System Level Test
-500 with High Damping and Rapid Decay in the Strain

Frequency Matching: In many cases the shock input will

g

-1000 create an oscillatory motion in the PCB, as shown in Fig-

ure 8-5 and Figure 8-6. In these cases it is simple to calcu-
gl

Chassis late the resonant frequency of the response. This is com-

-1500 monly done using a Fast-Fourier Transform (FFT)
Fixed Plate
calculation which is implemented in many data acquisition
-2000 systems. From the FFT one can determine the primary fre-
quency components of a response. If the amplitude and
Max Principal Strain (ue) frequency responses of two systems are matched, then
IPC-9703-8-4
strain rate will also have been matched.
Figure 8-4 Comparing Board Bend Modes of Board in
System Chassis and Mounted to a Fixed Plate
Board Strain Settling Time
Frequency/Strain Rate Correlation: It is known that the
properties of solder materials are rate dependent. This
Board Strain (ue)

dependency can affect the reliability of the solder joint and 500
can shift the failure mode. It is therefore important to con-
0
sider the strain rate in correlating tests. As there is no 6.5 6.6 6.7 6.8 6.9 7 7.1 7.2
means to directly measure the complex strain state in the -500
solder ball, board strain is substituted.
-1000
Strain Rate Calculation: One can directly compute the ts
strain rate by taking a derivative of the strain time history. -1500
Care should be taken however to not introduce errors Time (sec) IPC-9703-8-6
caused by measurement noise. All experimentally acquired
Figure 8-6 Typical Time History for a Component Board
data is subject to many kinds of noise. Noise prevents use
Test with Minimal Damping
of a simple forward or backward difference, (εi+1 - εi)/Δt,
method of computing the strain rate. The data must either Dynamic Event Settling Time Evaluation (Relating to
be smoothed first via filtering or using a regression fit to Dynamic Damping): A board strain response can reveal
the desired portion of the time history. the system damping effect. More free vibration cycles are

18
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

expected to occur for low damping system after a shock or oscillations for a single input pulse as compared to the first
drop impact. These can lead to fatigue damage. Obviously, test. The settling time can be estimated by assuming the
lower damping system means higher solder joint risk, if steady state zone is ± 10% of the peak value, in this case,
they have the same board strain amplitude, board bend ± 250 µStrain. The settling time is the time from the start
mode and vibration frequency. of the event to the point at which the strain remains within
As such, it is also important to have a reasonably correlated the steady-state zone. The zone should be greater than pos-
settling time. Figure 8-5 and Figure 8-6 compare the set- sible error and variance. An alternative way to define the
tling times of two tests. The second test may lead to overly zone is to use the effective board strain value within which
conservative results due to the excessive number of board the solder joint damage is not continued.

g
gl

19
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Annex A – Sample Data Reporting Format

Test Report Recommendations A test report including comprehensive documentation of the bend test is recommended
(see Tables A-1 through A-3). Disclosure of the test report contents will depend on specific customer/supplier agreements.

Table A-1 Test Report Recommendations (Equipment and Materials)

Category Description
Equipment • Manufacturer, model number and operational conditions for the following: drop or shock equipment used
• strain measurement equipment
• continuity monitoring equipment
• spectral analysis/accelerometer equipment
Fixture • type of fixture used and photograph
• support locations used
• fixture materials
Component • package outline drawing or reference to JEDEC outline
• die dimensions (width, length and thickness) and orientation (rectangular die)
• daisy-chain connection map and/or net list
• measured solder ball or lead coplanarity
• solder ball shear values or lead pull strength, and associated failure modes (parts from same production lot
as tested devices)
• solder-wetted pad dimensions, if applicable
• solder ball land pad type, if applicable (solder mask defined, etc.)
• lead finish/pad metallization, including thicknesses of all layers and composition of solder, if applicable
Test Board • width, length and thickness
• dielectric material
• surface finish
• external trace, pad and solder resist opening dimensions measured board coplanarity
Strain Gage • manufacturer and part number resistance
• strain gage element size
g

• strain gage locations used

gl

Table A-2 Test Report Recommendations (Board Assembly)

Category Description
Board Assembly • preheat temperature, ramp rate, critical peak temperatures
• (solder, package surface, board, etc.) duration above solder liquidus temperature and cooling rate
• solder composition
• solder paste metal percentage, particle mesh size type and flux type
• reflow atmosphere
• nominal solder paste volume
• nominal solder joint geometry
• component/PCB storage and/or bake-out conditions prior to board assembly
• storage conditions and duration following board assembly

20
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Table A-3 Test Report Recommendations (Test Results)

Category Description
Set-up data • drop height or input acceleration history
• torque setting used to screw attachments
Strain data (each • strain vs. time (max & min principal, gage component should be retained)*
test board)
• maximum strain-rate or strain-rate vs. time **
• acceleration (both input and near the component) vs. time
Resistance data • resistance vs. time
(each test board)
Failure distribution • test condition/cycle count at which failure occurred
• cumulative failure percentage for test level(s)
• failure map by location for dye-and-pry data, showing percent crack and mode
Failure mode • failure mode histogram (all test packages)
• time-zero cross-section of a single pkg/board test assembly
• failure analysis of representative sample of each observed failure mode
*Note: measurements at each strain gage location should be reported
**It is recognized that the strain rate vs. time calculation is subject to factors which may influence its accuracy and relevance. This calculation is a second
derivative and thus can magnify any noise in the strain vs. time measurement. Although strain rate vs. time information can be valuable, one must use
caution in interpreting the results. If strain rate vs. time is reported, the method used to smooth the data (regression, filter, etc.) should accompany the
report. See 8.1.2 for guidance on performing the calculation.

g
gl

21
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Annex B – Metrologies

B.1 Strain Gage Use and Techniques age. Default location would be opposite the corner most
solder joint. For comparisons between system and compo-
B.1.1 Strain Gage Selection There are a multitude of
nent testing, precise gage location should be documented
strain gages available; however, only small, stacked and held consistent. In the example below, the peak strain
rosettes are required. Coefficient of thermal expansion location was offset to the interior from the corner most ball.
compensation and high temperature dielectric are not The gage was placed accordingly.
required for room temperature testing. Pre-attached lead
wires simplify instrumentation of the test, but may not be Orientation of strain gages should be recorded. Recom-
available or be the at lowest cost. Gage element size should mended strain gage orientations at all corners are shown in
be on the same order of the package solder ball pitch, not the example in Figure B-1. This arrangement keeps the
to exceed 2.0 mm [0.079 in]. For comparisons between orthogonal sensing elements aligned to the edge of the
system and component testing, gage selection should be package.
documented and held consistent. Gage resistance is usually In situations with high component density, gage lead-wire
120 Ohms or 350 Ohms, and either is acceptable. Strain routing or interference with other components may con-
gages with pre-attached lead wires are available in two and strain gage placement. Small, adjacent, interfering compo-
three wire arrangements. A three-wire configuration is pre- nents may be removed without affecting results. It is not
ferred for improved accuracy and noise reduction. A two- recommended that large adjacent components be removed.
wire configuration may be used to ease wire routing; how-
ever, it is recommended that a transition to a three-wire Silk Screen: It is advisable to print a silk screen on the
configuration be placed as near as possible to the gage. board to guide the placement of strain gage within accept-
able visual tolerance.
B.1.2 Strain Gage Placement Strain gages are typically
Attachment Method: Follow manufacturer’s recommenda-
located on the secondary side of the board at the highest
tion and process. Cyanoacrylate adhesives are acceptable
strain location. Positioning of the strain gage is in reference
for room temperature testing.
g

to the center of the gage resistive element grid. Determina-

tion of location may be based on finite element analysis, Wire Routing: Care should be taken in routing gage lead-
gl

failure analysis, or experimental measurements. In most wires so that they are not to interfere with the dynamic
applications, this location will be at the corner of the pack- behavior of the system or test board. This is especially a

1 PIN-1 3
2 2

3 1

1 3

2 2
3 1
IPC-9703-B-1

Figure B-1 Wire (a) Strain Gage Location with Respect to Solder Ball at Package Corner. (b) Orientation of Strain Gages
for every Package Corner

22
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

concern for testing of small form factor and hand held Gage Lifetime: Gages are designed for single installation.
devices. Figure B-2 shows an example of wire routing in a The adhesive may degrade due to time, temperature and
small form factor laptop. Small diameter wiring is used to moisture exposure. Follow gage manufacturer recommen-
route strain signals away from the components. On the dations.
exterior of the device, it is advised that the wiring is scaled
up to a larger diameter wire which is less fragile and easier B.1.3 Data Acquisition Recommended scan frequencies
to connect to the data acquisition system. for typical applications are shown in Table B-1.
Table B-1 Recommended Scan
Frequencies for Various Systems
Typical Maximum
Frequency Minimum Hardware
Response Sampling Anti-aliasing
System Type (Hz)* Rate (Hz) Filter (Hz)
Server/
20-300 3,000 1 ,200
Communications
Desktop 40-500 5,000 2,000
Laptops 50-600 10,000 4,000
Cell Phone
500-4000 50,000 20,000
and Handheld
*Note: Shipping shock will typically produce a lower frequency response
while user drop of an unpackaged system may excite a higher frequency
Figure B-2 Wire Routing should be done So as Not to response.
Affect Results
Table B-1 follows the rule that the sampling rate should be
In the case of test boards, this problem may be eased by greater than 10 times the highest frequency of interest, to
including wiring for the strain gage directly in the PCB accurately capture strain response.
circuitry. The gage itself is then connected to the PCB by
soldering the gage leads into plated through holes or sur- The frequency range of the system response can be deter-
face mount pad. This is illustrated in Figure B-3. mined via a Fast Fourier Transform (FFT) analysis of the
g

strain data. An example is shown in Figure B-4. For this

gl

system there is a significant response at approximately 400

G1- G1+ Hz; the minimum sampling rate should be at least 4000Hz.

G2+
240
220
G2-
200
180
Response FFT

160
G3+ 140
120
100
G3- 80
60
40
IPC-9703-B-3 20
Figure B-3 Wire Artwork Showing Strain Gage Feature 0 100 200 300 400 500 600 700 800 900 1000
and Internal Board Routing
Frequency (in HZ)
Wire Strain Relief: Strain gage lead wires should be pro- IPC-9703-B-4

tected with adequate strain relief. Damage to unprotected Figure B-4 Fast Fourier Transform of an Example
lead wires is common in mechanical shock testing. In free System
fall shock tests it can be challenging to secure the wires in
light weight systems without influencing the way the sys- Table B-1 also shows the minimum setting for analog anti-
tem falls, so in these cases guided drop systems may be alias filters. These filters are used to prevent the inadvert-
advised. ent sampling of high frequency data as a lower frequency
(see Nyquist Sampling Theorem). Many modern data
Quarter Bridge Setup: Strain gages should be wired in a acquisition systems may automatically select an appropri-
quarter bridge setup due to a lack of symmetry in strain. ate filter.

23
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Digital filters may also be used to eliminate high frequency The highest possible pixel resolution of the high speed
noise from the strain signal. Filter type and cut-off fre- camera is also a large factor for total accuracy.
quency should be selected to optimize signal response. If
one suspects excessive electrical noise, it is advisable to B.3.1 Setup Recommendations
attempt to isolate these by using a three wire setup and Line of Sight: In order to track the movement of a target
braiding the lead wires where possible. Ground isolation point or feature on a board assembly, the camera must have
transformers may also be used to isolate data acquisition an unrestricted line of sight. This can be achieved through
equipment from noise in the incoming power. positioning of the camera and/or the use of creative solu-
tions including mirrors, light weight post (target) or line of
B.2 Accelerometers
sight openings in chassis or test fixtures. With a 2D camera
setup the camera needs to be perpendicular to the center-
B.2.1 Accelerometer Selection There are many avail-
line of the sample at a vertical angle that will maintain a
able accelerometers. The weight of the selected accelerom-
line of sight through the event.
eter should be minimum, internal damping well outside the
frequency of system and size small without interfering. The Sample Setup and Preparation: To prepare a sample or
frequency range and the linear range should also be consid- printed circuit board for imaging, an array of targets suit-
ered. For solder joint reliability testing, a single axis accel- able for tracking with the motion analysis software needs
erometer will be sufficient as the out of plane is of primary to be applied to the sample in known locations. These X
interest. axis and Y axis dimensions are tied to the Z axis dimen-
sions from the motion analysis software to generate surface
B.2.2 Accelerometer Placement Placement depends on profiles at any selected point in time during the event.
how the data is to be used. Mount points for test develop-
ment near the components of interest. B.3.2 Calibration The point displacements are measured
by tracking a target as it moves through a camera field of
Attachment Method: For systems that can be easily view and assigning a dimensional value for an individual
opened, simple wax attachment can be used. For systems pixel for the target being tracked. The pixel dimensions are
that will be subjected to severe shocks or cannot be easily influenced by the geometric relationship of the target rela-
g

accessed, adhesive attachment should be used. Follow tive to the camera, as well as distortions from the lens cur-
accelerometer manufacturer guidelines for adhesive and
gl

vature.
wax attach.
Geometric Introduced Errors: Geometric factors including
Wire Routing: Care should be taken in routing accelerom- distance of the target from the camera, the relative angle of
eter lead-wires so as not to interfere with the dynamic the camera and the relative angle change of the sample as
behavior of the system or test board. This is especially a it moves through the field of view will all introduce
concern for testing of small form factor and hand held changes to the value of a pixel. The tools for correcting
devices. these factors are included in the image analysis software
Wire Strain Relief: Accelerometer lead wires should be
and it is up to the user to define what methods are best fit-
protected with adequate strain relief. Damage to unpro- ted to their accuracy needs.
tected lead wires is common in mechanical shock testing. Basically the angle of the camera is measured and an
In free fall shock test it can be challenging to secure the image of a scale block with 4 points with known spacing is
wires in light weight systems without influencing the way captured at the physical point of the target to be tracked.
the system falls; in these cases guided drop systems may be These captured images are imported into the motion analy-
advised. sis software to create a motion plane that corrects for all of
these factors for the individual target point.
Instrument Lifetime and Calibration: Accelerometers
should be calibrated per the manufacturer’s recommenda- Lens Distortion: Lens distortion errors can be minimized
tion. by using longer focal length lenses or eliminated with the
use of correction factors included in the software (only
B.3 High Speed Photography and Motion Analysis Sys- required for wider angle lenses).
tems A high speed camera is a measurement tool that
requires calibration, and (like other measuring tools) is B.3.3 Tracking Software There are several software
dependent on proper setup. This includes the selection of packages available that will provide the ability to scale and
the correct lens, correcting for geometric induced errors track target points on an object in motion. Select the best
and selecting a frame rate that will eliminate image blur. package that meets your budget and accuracy requirements.

24
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Annex C – Shock Failure Analysis of Electronic Components

C.1 Scope To support the shock methodology, the use of used include x-ray, coupled scanning acoustic microscopy
failure analysis is critical and should be performed at a (CSAM), and side view optical microscopy, etc. Common
level that clearly identifies the failure mechanisms and methods and tools for destructive failure analysis include
locations involved. The following sections discuss failure cross section, dye and pry, and dye and pull. Some tech-
analysis recommendations in more detail. niques such as CSAM, x-ray, scanning electron microscopy
(SEM) and energy dispersive x-ray (EDX) can be used to
C.1.1 Failure Statistics If failure distribution is analyze samples undergoing destructive or nondestructive
unknown and of interest, the Weibull model is recom- techniques.
mended. When data is limited, true statistical representa-
The mechanical failure mode(s) of each component and
tion is extremely risky as characterization of mean, stan-
assembly should be identified. It is recommended that each
dard deviation, etc. becomes meaningless for small sample
observed failure mode be documented using an appropriate
sizes.
failure analysis technique. Before using any destructive
C.1.2 Guidance for Failure Analysis It is recommended
technique, the assembly should be inspected by optical
that failure analysis be performed on samples representing microscopy and the overall apparent damage, including
the full population. The use of only early (low strain, low those to package, solder joint, and PCB traces/vias, should
stress, low cycle count, etc.) or late (high strain, high be documented. Destructive techniques should then be used
stress, high cycle count, etc.) failures could be misleading. to determine the failures. Example failure modes for solder
The number of samples selected for failure analysis will be ball array style packages are illustrated in Figure C-1. Note
dependent on the experimental plan, previous knowledge that electrical monitoring may not identify many of the
of failure signatures and testing objectives, etc. failure modes shown in this figure. Failure modes include
solder cracking and breakage, package body cracking and
C.1.3 Failure Criteria Prior to a test start, it is necessary separation/delamination, and cracking between the various
to document a threshold increase in resistance which could lead, Inter-Metallic Compound (IMC), solder and PCB
g

be defined as an electrical ‘‘failure’’ during shock testing. metal pad interfaces.

gl

This should be accomplished via continuous monitoring on

C.1.5 Example of Failure Characterizations
the order of the strain and acceleration sampling rate, i.e.,
of at least 10x of the natural frequency. In shock testing, an Area Array Cross-Sectioning/Inspection Figure C-2
electrical resistance change, even for a brief period, may be shows typical cross-sections for an area array package.
an indication of damage. Although a 20% increase in daisy- Diagonal sections may also be desirable in the package
chain net resistance may be appropriate for transient bend corners.
loading, a different threshold may be required for a The sectioning samples should be potted or encapsulated in
dynamic loading due to test equipment capability and spe- an appropriate medium. The cross-section cut should be
cific daisy-chain design scheme. If no prior knowledge of made through the approximate center of each solder joint
the resistance increase is available, it is recommended that depending on the area and pattern geometry.
a characterization study be completed before full testing
begins. Inspection should be conducted under a microscope/SEM
and the following images generated:
The clear goal of failure analysis is the ability to detect the
location, mode and mechanism of damage. A ‘‘failure’’ can The whole cross-section (may be combined from multiple
be defined as observed mechanical damage (pad lifting/ images) Magnified individual failed solder joints (at appro-
pad cratering, interface delaminations, solder joint crack- priate magnification levels to clearly show the damage)
ing, trace cracking, etc.) or an increase in electrical resis-
C.1.6 Dye and Pull/Pry Test Description The following
tance beyond the previously defined acceptable threshold.
procedure describes the method that should be used for
For shock tests completed without an electrical failure
Dye and Pull/Pry.
being observed, it is recommended that failure analysis be
performed to ensure that damage was not missed due to Recommended Tools:
issues with test hardware or the inability to monitor partial • Low viscosity colored layout fluid (dye).
damage.
• Mechanical force testing equipment or pull gage (for dye
and pull).
C.1.4 Failure Analysis Techniques Failure analysis
techniques consist of two types: destructive and nonde- • Screwdriver or pliers (for dye and pry).
structive. Nondestructive failure analysis tools that may be • Microscope.

25
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Package Substrate

Metal Pad Fracture @ pkg

metal/IMC interface

Fracture @ pkg side Pkg pad lift/crater

IMC/solder interface
Solder Ball Fracture within bulk solder
Fracture @ PWB side
IMC/solder interface PWB pad lift/crater

Fracture @ PWB
Metal Pad metal/IMC interface
Printed Wiring Board (PWB)

IPC-9703-C-1

Figure C-1 BGA Solder Joint Failure Modes

Carefully place the container with the specimen inside

A1 Corner vacuum chamber, to allow the dye to penetrate into any
cracks or crevices.
• Remove the specimen from the dye. Excess dye is
Middle (Optional) removed by hanging the specimen and allowing it to drip.
(includes center)
g

Edge Place the specimen in oven to dry at 100°C for eight

hours. Alternatively, place the stained specimen in a ven-
gl

tilated hood in room temperature for 24 hours to allow

the dye to dry.
Diagonal
Pull/Pry Procedure:
• For Pull: mount the specimen to mechanical pull testing
IPC-9703-C-2 equipment or pull gage and remove the specimen.
Figure C-2 Area Array Cross-Section Diagram • For Pry/Peel: use a screwdriver or pliers to pry the corner
of the package. This should be done quickly with the
• Saw for isolating the package area.
intention to remove the entire package.
• Cleaning solution or acceptable PCB assembly cleaning
• Visually inspect and document the failure locations on
procedure.
both board and component side using a microscope and
• Bake chamber for drying. visual-analysis software (if available).
• Vacuum chamber. • The presence of dye on a fracture surface is an indication
Recommended Dye Procedure: that damage occurred prior to the package removal.
• Place specimen in a container and pour enough IPA to Failure Mechanism and Mode Description:
entirely submerge the specimen in solution. Leave the
• Record all failures by location within the package and
specimen in solution for about 5 -10 minutes. Remove
printed circuit board and the type of failure mode.
units from IPA use compressed air gun to dry. Let the
specimen air dry for 1-2 minutes and then discard IPA in
C.1.7 Report
a proper container. Or it is permitted to apply a user’s
standard PCB assembly cleaning procedure. An ultrasonic • Background including test setup, DOE plan, etc.
bath is not recommended because it can propagate a sol- • Summary of results including images, failure mode by
der joint crack. location, etc.
• Place the specimen inside a container and fill the con- • Identification of key trends and conclusions/
tainer with dye until the specimen is fully submerged. recommendations.

26
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Annex D – Suggestions for Selecting Sample Size

The definition of sample sizes is a decision that should bal- bution model which has the broadest use in the industry to
ance the costs of the experiment with the minimum level of calculate sample sizes. The following equation shows how
confidence necessary in the results of the test. With a larger to estimate sample size:
sample size comes better confidence in the results, but
ln β
obviously the test costs and time requirements rise accord- N=
ingly. ln (1 − p+)

During all phases of testing, multiple, similar products Where:

should be subjected to each test environment. If every fail- N = number of test specimens
ure mechanism were present in every product, testing a ln = natural log
single unit would be sufficient to detect all defects. Unfor- β = 1 - confidence level
tunately, because of component and manufacturing process p+ = upper confidence limit or failure probability
variability, every product is different from each other and Example: How many units must be tested without failure to
more than one sample must be tested to prove the perfor- demonstrate that the probability of failure is less than 0.1
mance of a product in the field. In real life applications, (10%) with 90% confidence?
some products will fail while others will not. The desired
confidence level to expose a failure mode in a product will ln (1−0.9) ln (0.1) −2.30
N= N= N= N = 21.8 = 22
determine the number of products to test. ln (1−0.1) ln (0.9) −0.105
Assuming that the failure mechanism is present in only a Based on the binomial distribution, a sample size of 22 is
small percentage of the units, testing with small sample necessary to achieve a 90% probability of detecting a
sizes is very risky. On the same lines, testing on product defect that occurs in 10% of the population. If you are
with known performance problems will warrant larger concerned with achieving very low defect densities, for
sample sizes than product proven to be defect free. example 1% or less, the sample sizes become very large.
Table D-1 summarizes this calculation for some common
g

Assuming a normal distribution, a sample size of two pro-

duces only a 20% probability of detecting a defect that confidence levels and failures rates.
gl

occurs in 10% of the products. Testing even 10 units only

For a more complete introduction to this topic, the reader
raises the probability to about 65%.
is referred to a text on reliability engineering such as Intro-
If the actual failure distribution of the product under test is duction to Reliability Engineering, by E.E. Lewis, John
not known, a recommendation is to use a binomial distri- Wiley & Sons, Inc. ISBN: 0-471-01833-3.

27
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

Table D-1 Sample Size Estimate for a Single Failure Mechanism

Max Failure Rate Confidence Level
% DPM 25% 50% 60% 70% 80% 90% 95% 99%
0.01 100 2877 6931 9162 12039 16094 23025 29956 46049
0.05 500 575 1386 1832 2407 3218 4604 5990 9208
0.10 1000 288 693 916 1203 1609 2301 2994 4603
0.50 5000 57 138 183 240 321 459 598 919
1 10000 29 69 91 120 160 229 298 458
2 20000 14 34 45 60 80 114 148 228
3 30000 9 23 30 40 53 76 98 151
4 40000 7 17 22 29 39 56 73 113
5 50000 6 14 18 23 31 45 58 90
6 60000 5 11 15 19 26 37 48 74
7 70000 4 10 13 17 22 32 41 63
8 80000 3 8 11 14 19 28 36 55
9 90000 3 7 10 13 17 24 32 49
10 100000 3 7 9 11 15 22 28 44
15 150000 2 4 6 7 10 14 18 28
20 200000 1 3 4 5 7 10 13 21
25 250000 1 2 3 4 6 8 10 16
50 500000 0 1 1 2 2 3 4 7
75 750000 0 1 1 1 1 2 2 3
g
gl

28
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Annex E – Suggestions for Finite Element Analysis in Mechanical Shock

E.1 Linear Dynamics A problem should have the follow-

ing characteristics for it to be suitable for linear transient
dynamic analysis:
• The dynamic response of the system should be linear:
linear material behavior, no changing contact conditions,
and small displacement only.
• The response should be dominated by relatively few fre-
quencies if modal superposition is to be used. As the fre-
(a) Mode 1: 31.1 Hz
quency content of the response increases, the modal
superposition technique becomes less effective.
• The dominant loading frequencies should be in the range
of the extracted frequencies to ensure that the loads can
be described accurately.
• The initial accelerations generated by any suddenly
applied loads should be described accurately by the
eigenmodes.
• The system should not be heavily damped. (b) Mode 6: 140 Hz
IPC-9703-E-1

All mode-based, transient, linear dynamic calculations start Figure E-1 Coarse Mesh of Plate
by extracting the natural frequencies of the system, and
these are then used as the basis of the linear dynamic solu- plate is such that there is significant excitation of a mode,
tion. Nonlinear preloading effects can be accounted for in a mesh refined sufficiently to model this mode must be
linear dynamics, in that the stiffness used for the natural used; the results from a coarse mesh will not be accurate.
g

frequency extraction can be the current stiffness after any

preload has been applied.
gl

E.1.1 Natural Frequency Analysis

Element Selection: In general, the rules for selecting ele-

ments for linear dynamics are the same as those for static
simulations. Second-order elements should generally be
used unless there is a particular reason to use a different
element type.
(a) Mode 1: 30.2 HZ
Mesh Design for Dynamics: When designing meshes for
dynamic simulations, the user needs to consider the modes
that will be excited in the response and use a mesh that is
able to represent those mode shapes adequately. This
means that a mesh that is adequate for a static simulation
may be unsuitable for calculating the dynamic response to
loading that excites high frequency modes. Good practice
would be to run a mesh convergence study.
Consider, for example, the plate shown in Figure E-1. The (b) Mode 6: 124 HZ
IPC-9703-E-2
mesh of first-order shell elements is adequate for a static
analysis of the plate under a uniform load and is also suit- Figure E-2 Fine Mesh of Plate
able for the prediction of the first mode shape. However,
the mesh is clearly too coarse to be able to model the sixth It is generally recommended that at least six to eight ele-
mode accurately. ments are used per wavelength of the highest frequency
mode to be modeled.
Figure E-2 shows the same plate modeled with a more
refined mesh. The displaced shape for the sixth mode is For linear dynamic analysis, care needs to be taken when
modeled more accurately, and the frequency predicted for introducing symmetry into a model. A symmetric model
this mode is more accurate. If the dynamic loading on the assumes that geometry, constraints, loads, and the response

29
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

will all be symmetric. The assumption of symmetry will

prevent the analysis from extracting non symmetric modes, Exciter The Test Board
if they exist.
For example, consider a plate with all edges simply sup-
ported. Table E-1 shows the first three modes extracted
from a full model, a half symmetric model and a half
asymmetric model. This clearly shows that modes are miss-
ing from the symmetric and asymmetric models.
Materials: The material response during a natural fre-
quency extraction is linear. If nonlinear materials are
defined, then, the linearized elastic stiffness will be used. If
there is a preload, then the current, linearized elastic stiff-
ness about the preloaded state of the materials should be
used.
Boundary Conditions: Natural frequency extractions can
be performed on either an unconstrained or constrained Force An Array of
Transducer Accelerometer
body. When the eigenmodes are to be used as the basis for IPC-9703-E-3
a transient modal dynamic solution, the boundary condi-
Figure E-3 Rack Mounted Test Board
tions for the natural frequency analysis should reflect those
needed for the dynamic analysis. Prescribed motion in the motion. So, two base motion sets will be required, one for
transient dynamic analysis is applied using boundary con- each constraint. These base motion sets are defined in the
ditions called base motions. All points that will either be eigenmode extraction step.
fixed or have a prescribed motion in the transient dynamic
For the primary base motion the constraints are applied
analysis will need to be constrained in the natural fre-
directly. This is not possible for secondary base motion;
quency analysis. If different constraints are to be applied at
instead, big masses with suitable accelerations are applied
g

different points in the model, then multiple base motions

(generally this process is automatic and internal to the
will have to be defined.
gl

solver). There will be additional low frequency modes

For example, if a test board is to be rack mounted into a associated with the big masses; one per degree of freedom.
chassis and then have an excitation applied, as shown in So, the solver should automatically increase the number of
Figure E-3, both the edges of the board and the point of modes extracted to account for this and maintain the same
excitation would be fixed for the natural frequency extrac- number of ‘‘real’’ modes, associated with the actual struc-
tion. ture.
In the transient dynamic analysis, the edge of the board is Contact and Other Constraints: All constraints (such as
to be fixed, while the excitation point has a prescribed multi-point constraints and couplings) will be linearized for

Table E-1 Full, Symmetric, and Asymmetric Meshes of Plate

Full Model Symmetric Model Asymmetric Model

Mode 1: 13.18 Hz Mode 1: 13.18 Hz Mode 1: 17.41Hz

Mode 2: 17.41 Hz Mode 2: 24.65 Hz Mode 2: 34.89 Hz

Mode 3: 24.65 Hz Mode 3: 47.75 Hz Mode 3: 51.37 Hz

30
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

a natural frequency analysis. Any point that was in contact

before the natural frequency will remain in contact for the
duration of the linear analysis; similarly, any point that was
not in contact will remain open for the duration of the lin-
ear analysis.
Preloading: Some FE packages allow for the inclusion of
preloading effects, before the natural frequency extraction.
For many structures, assembly and active loading will
change the frequency content; the frequencies will vary
depending on the previous load history. Generally, struc-
tures under tension will stiffen, while structures under com-
pression will soften.
Solver: Different solution methods exist for eigenmode
extraction; two such methods are subspace iteration and
Lanczos. Both of these techniques extract undamped eigen-
modes. Complex eigenmode solvers are available which
will account for material damping, frictional contact, etc.
These are generally not required for linear dynamics analy-
sis of shock for electronic components.
The Lanczos solver is better suited for larger models, or
when large numbers of modes are to be extracted. It also
has the advantage of being able to make use of the restart
technique such that additional modes can be extracted
without the need to repeat the extraction of the existing
modes. The subspace iteration method may be faster when
g

only a few eigenmodes (generally less than twenty) are

needed.
gl

Typically, the number of eigenmodes to be extracted can be

specified directly, or a frequency range of interest can be
given. It is generally recommended to use the frequency
range method, as this then avoids any potential problem
with insufficient modes being calculated, when secondary
bases are used with their additional required modes.
Residual Modes: Some FE packages support the inclusion
of residual modes in the modal basis for linear dynamics
calculations. These are pseudo eigenmodes used to aug-
ment the true eigenmodes in a mode-based linear dynamics
analysis.
The static solution to a set of load cases (typically is the
excitation loading) forms the basis for the residual modes:
• They minimize the number of eigenmodes necessary to
achieve a desired accuracy.
• They are typically calculated based on excitation loading.
IPC-9703-E-4
Typical application: Linear dynamic analyses where the
excitation loads have high frequency content; i.e., in order Figure E-4 Simplified Satellite Antenna Model
to capture the loading behavior, a large number of eigen-
The results of a linear dynamic analysis on this model are
modes would be required to obtain an accurate solution
shown in Figure E-5. These show that 10 eigenmodes are
without residual modes.
insufficient to accurately capture the dynamic response, but
For example, consider the simplified model of a satellite 10 eigenmodes and 1 residual mode give a much more
antenna shown in Figure E-4. This model is subjected to accurate solution and compares well with an analysis using
various loadings. many more eigenmodes.

31
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

For the user, no special consideration needs to be made

when applying either loads or specifying initial conditions
SF1 at Element 802 Int Point 1 in ELSET OUTEL

2.80
for a modal dynamics analysis provided adequate eigen-
modes ( and residual modes) are used. These quantities are
2.40 applied in the same way as for any other analysis type.
Prescribed motion is applied through base motions. The
2.00 location for these bases are defined in the natural frequency
analysis, as described above. The type, magnitude, and
1.60 form of the excitation are defined in the modal dynamic
analysis.
fiftyModes
1.20 Damping: As previously stated, most natural frequency
tenAndResidualMode
solvers extract the modes based on the undamped system.
tenModes
0.80 For small values of critical damping the eigenfrequencies
(and eigenvectors) of the damped system are very close to
0.40 the corresponding quantities for the undamped system
(fraction of critical damping, ξ <0.1). As ξ increases, the
undamped eigenfrequencies become less accurate; and as ξ
0.00
0.00 2.00 4.00 6.00 8.00
approaches 1, the use of undamped eigenfrequencies
becomes invalid.
Time IPC-9703-E-5 In most linear dynamic problems the proper specification
Figure E-5 Results for Simplified Satellite Antenna Model of damping is important to obtain accurate results. How-
ever, damping is approximate in the sense that it models
Output: Results of interest from an eigenmode extraction the energy absorbing characteristics of the structure with-
are the shapes of the extracted modes and their associated out attempting to model the physical mechanisms that
frequencies. Other results that should be checked before cause them. Therefore, it is difficult to determine the damp-
continuing to the transient modal dynamic analysis: ing data required for a simulation. Data may be available
g

Modal Effective Mass: This reports how the total model

from dynamic tests, but often the user will have to work
gl

mass is distributed to the different directions represented by with data obtained from references or experience. In such
a mode shape. For translational modes, the sum (over all cases, care should be taken in interpreting the results, and
extracted modes) of the modal effective mass should be parametric studies should be used to assess the sensitivity
close to the body’s total structural mass. If not, additional of the simulation to damping values.
modes are needed. Time Incrementation: The transient modal dynamic proce-
dure requires the user to specify the total time period for
Modal Participation Factor: This indicates the predomi-
the analysis and a fixed time increment. If the required time
nant direction (degree of freedom) in which a mode acts.
increment is known, then the user can specify a frequency
This information can aid modal identification and help to
cut-off for the eigenmode extraction; there is no point in
choose modes that will best represent the response to a
extracting modes whose period is substantially smaller than
given loading. It is then possible to select specific eigen-
the time increment used. Generally the time increment is
modes to use as the basis for the modal dynamic solution.
not known. It is more common for the user to extract suf-
ficient eigenmodes to satisfy the modal effective mass
E.1.2 Modal Dynamics From an analysis standpoint,
check described above. In this case the given time incre-
once the frequency extraction is complete, a transient
ment for the transient modal dynamic analysis must be
modal dynamic analysis is performed in a similar manner
capable of resolving the highest frequencies of interest.
to a nonlinear transient dynamic analysis, but it is compu-
tationally much faster. The user must ensure that the subset Output: Processing the output requests for a transient
of modes used as the basis of the solution is sufficient to modal dynamic analysis may be more expensive than
accurately represent the response of the structure as it is actual analysis. Analysis codes will generally, therefore,
loaded. save time by calculating output only for points where it is
requested. It is recommended to request only the necessary
Excitation: There are three possible forms of excitation output to reduce the cost of analysis.
for modal dynamic problems:
Care needs to be taken when interpreting results from a
• Concentrated and/or distributed loads.
transient modal dynamic analysis; most results are given as
• Initial displacement and/or velocity. perturbations relative to the motion of the primary base.
• Base motion. The sum of the relative motion and the base motion yields

32
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

the total motion; separate variables giving this total motion E.2.1 Explicit Nonlinear Dynamic Analysis
may also be available. In the absence of primary base
Element Selection: For the explicit dynamic procedure we
motions, the relative and total motions are identical.
need elements with lumped mass matrices. This require-
E.2 Nonlinear Dynamics A dynamics problem is consid- ment arises from the need to perform nodal calculations
ered nonlinear when: that invert the mass. Generally this limits the element
• The motion or deformation produced by the dynamic selection to first-order elements. Another consideration is
behavior of the structure is large enough that we must the cost associated with the element calculations, as these
account for the changes in geometry. are often the most expensive part of the explicit dynamic
solution. The use of reduced-integration elements mini-
• Deformation large enough to change the stiffness of the
mizes the cost of these calculations. So, for explicit dynam-
structure, thereby changing the natural frequency.
ics, the user is typically limited to first-order, reduced-
• The problem involves nonlinear material response. integration elements. Additional simplifications can be
• The problem involves contact. made to further reduce the cost of element calculations
It is not uncommon for a problem to contain all four such as the use of small-strain formulation elements. The
sources of nonlinearity. user needs to check whether any such simplifications are
being used (they may be default options) and ensure that
A nonlinear dynamic analysis requires the time integration they are appropriate for the analysis being performed.
of the full, coupled equations of motion. The superposition
of eigenmodes cannot be used and, therefore, a frequency Some FE packages allow the use of special formulation
extraction analysis is not required. The loads and boundary second-order tetrahedral elements for explicit dynamics.
conditions on the structure may be arbitrary because no These elements offer the advantage of being robust for
projection onto the eigenmodes is necessary. Procedures large deformations and contact, while not demonstrating
for solving nonlinear dynamic problems are much more shear or volumetric locking problems. They also have the
computationally intensive than those for linear dynamics. obvious advantage of allowing tetrahedral meshing to be
The equations of motion can either be integrated implicitly used.
or explicitly. There are advantages and disadvantages to
For solid meshes the use of hexahedral elements is recom-
both methods, and these are briefly listed in Table E-2.
g

mended when such meshing is possible and reasonable. If

gl

The explicit approach is more suited to impact and shock mesh generation dictates that tetrahedral elements must be
type applications, and so that will be the focus of this sec- used, then second-order tetrahedral elements are recom-
tion. mended. For example, many electronic components are
Table E-2 Comparing Implicit and Explicit Dynamic Solutions
Implicit dynamics Explicit dynamics
The implicit procedure is unconditionally stable. The time The explicit procedure is conditionally stable. The time incre-
increment size is not limited by stability requirements; generally, ment size is limited by the stability of the algorithm; generally,
fewer time increments are required to complete a given simula- many more time increments are required to complete a given
tion. simulation.
Each time increment is expensive since each requires the solu- Each time increment is relatively inexpensive because solving
tion for a set of simultaneous equations. a set of simultaneous equations is not required.
Iteration is required to meet convergence criteria. Iterations may No iterations or convergence criteria. Nonconvergence is not
not converge. an issue. Since accuracy is not controlled by convergence,
care must be taken in the time integration algorithms.
Ideal for problems where the response period of interest is long Ideal for high-speed dynamic simulations.
compared to the vibration frequency of the model, for example, Require very small time increments; implicit dynamics inefficient
earthquake shock. (especially for large models).
Slow, long time duration dynamics problems difficult to solve
effectively using explicit dynamics because of the limit on the
time increment size.
Use for problems that are mildly nonlinear and where the non- Usually more reliable for problems involving discontinuous
linearities are smooth (e.g., plasticity). nonlinearities.
With a smooth nonlinear response, implicit solution will need Contact behavior is discontinuous and involves impacts, both
very few iterations to find a converged solution. of which are difficult to solve using implicit time integration.
In impact problems, implicit solution has to perform very expen- Other sources of discontinuous behavior include buckling and
sive momentum transfer calculations for each impact. material failure.

33
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

housed in complex-shaped molded plastic parts, and these Mass scaling can be used to artificially increase the stable
are generally modeled with second-order tetrahedral ele- time increment for a given element. Some FE packages
ments. These elements offer reasonable accuracy, even with offer semi-automatic mass scaling, whereby only the
a single element through the section and can save signifi- masses of elements whose stable time increment is below
cant time in model generation. The use of first-order tetra- some criteria will be scaled, and that scaling can be
hedral elements is not recommended; a very fine mesh repeated periodically during the solution to account for
would be required to achieve reasonable accuracy. large deformations of the elements. If accurate surface
stresses are required from a solid mesh, skins of membrane
For shell meshes the use of quadrilateral elements is rec- elements can be used. These elements can be used at strain
ommended, with triangular elements being used where gage locations to give the necessary output.
mesh generation dictates. Continuum shell elements have
the topology of a hexahedral continuum element, but use a Materials: Most FE packages contain constitutive models
shell-type formulation. These elements allow for accurate for all materials commonly found in electronics compo-
modeling of bending with only a single element through nents and assemblies: metal, plastic, rubber, foam, glass,
the thickness, and they have advantages over regular shell etc. These materials may require direct specification of
elements in terms of contact modeling and visualization. parameters or they may accept test data with which the FE
These elements are therefore ideal for thin components code will perform the parameter evaluation.
whose main response is bending; for example, circuit
For a nonlinear analysis, many of the available material
boards, ICs, LCDs, etc.
models can be made strain-rate dependant, but obtaining
Hourglassing must be considered for any first-order, the necessary test data may be difficult.
reduced-integration element mesh. There are a variety of Material properties may vary with direction, and so ortho-
hourglass suppression methods available and so the user tropic or fully anisotropic material models may be
needs to understand the implication of the method being required. If this is the case, the user needs to take care
used to ensure that accurate results are obtained. Some FE when specifying the local directions for the material prop-
packages also include options which can help to stop erties.
elements turning inside out under extreme deformation.
These formulations are not physically motivated and are Damping is another factor that could affect analysis, but for
g

really used to allow an analysis to progress when it would which test data may be difficult to acquire. It may be
gl

ordinarily fail with element distortion problems. This can possible to use damping to tune the response of a model to
be a useful option for soft materials under high compres- match test data. The resulting damping values could then
sion, for example gaskets and seals. be used in other similar analyses.
Some FE packages contain progressive failure models that
Mesh Design for Dynamics: For complex, system-level
can be used to model ductile or brittle failure of materials.
models, good practice is to build the model up part by part,
Application areas for this type of material model might be
or by sub-assemblies. It is much easier to debug a part-
in detailed solder joint failure models, or LCD failure in a
level model than a full system-level model. Some FE pack-
system-level analysis.
ages allow models to be defined in terms of parts, instances
of parts, and assemblies. This approach fits well with the Contact: Contact definition depends on the FE package
methodology of building models up from parts and means being used; it may be necessary to define individual sur-
that assemblies of those parts can easily be added to. faces, and then identify which of those can interact, or an
Mesh design for nonlinear dynamics is similar to that for automatic general contact algorithm may be available.
linear dynamic simulations, in that the user needs to con- General contact obviously makes the model generation
sider the modes that will be excited in the response and use very much simpler.
a mesh that is able to represent those mode shapes There are a number of different methods available for
adequately. enforcing contact constraints. Two of the more common are
the kinematic and penalty methods:
For explicit dynamics an additional consideration is the
element sizes. For a mesh made up of a homogeneous Kinematic enforcement of contact is done by calculating
material, the smallest element in the model will determine corrections to accelerations, distributing masses and forces
the stable time increment and, the smaller the element, the associated with contacting nodes on to opposing surfaces,
smaller the time increment. So, the user needs to try to bal- such that no penetration occurs. The advantage of this
ance the element size required for an accurate solution with method is that the contact condition is exact; no penetration
the element size required for a reasonable runtime. The use occurs. The disadvantage of this method is the potential for
of small elements where not specifically required should be conflicts with other constraints in the model, and certain
avoided. contact types (for example rigid-rigid) cannot be modeled.

34
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

Penalty enforcement is done by applying resisting forces to transfer of data between these solvers. This makes it pos-
nodes that have been penetrated. This method allows for sible to model the preloading operations with an implicit
more general contact cases, for example contact between solution method, and then import the deformed shape and
rigid bodies. Also, this method is less likely to result in associated stresses into the explicit dynamic solution as the
problems when contact nodes are common to other types of start point for the shock test.
constraint. The disadvantage of this method is that small
penetrations may occur and so the contact condition is not Analysis: The user generally has the choice of precision
enforced precisely. If possible, using general automatic level for the explicit dynamic solver. A double-precision
contact with the penalty enforcement method would be the solution is normally preferred for shock and impact type
recommended approach, unless there is good reason to use load cases. Most FE packages allow the use of parallel
an alternative. execution and may allow for the use of both shared- and
distributed-memory machines. Depending on model size,
Initial overclosures can cause problems with the analysis. runtime requirements, and available hardware, this option
These overclosures may be caused by inaccurate geometry, should be explored.
or by geometric simplifications and assumptions made for
the analysis. For example, if a small fillet is removed for Output: Typical results of interest from a nonlinear
the analysis, parts may overlap. Most FE packages will explicit dynamic shock analysis include the deformed
attempt to resolve any initial overclosures and some offer shape and animations of the same. This allows for an
tools for the user to manually make these adjustments. The assessment of which parts come into contact during the
user should be aware of this potential problem and it is analysis. If high speed video footage of a physical test is
something that should be checked. available, then some FE packages allow that video to be
synchronized with the animation from the analysis to allow
Constraints: Various types of constraint are available for
for simple side-by-side or overlay comparison.
either modeling joints and interfaces, or just bonding
regions of mesh together. Tie constraints can be used to Stresses and strains in critical components. This allows for
join incompatible meshes. In this way, they can also be an assessment of whether material yielding or failure may
used for very simple representations of package intercon- occur.
nects, in that the package mesh can be tied to the circuit
Interface or joint loads and stresses. This allows for an
g

board mesh. This type of constraint is rigid, but some FE

assessment of whether joints or bonds might break. If fail-
gl

packages may allow the joint to fail.

ure models are included in the analysis, then, parts can
Some FE packages have spot weld representation, which physically separate during the analysis. In this case, failure
are mesh independent constraints that can either behave could be detected by looking at the deformed shape.
rigidly or have material properties associated with them.
Care needs to be taken when requesting results from an
The latter option allows for elastic, plastic, and/or failure
explicit dynamic analysis. In the absence of runtime filters,
response in the joint. These connection types can be used
data for xy-plots should be written every increment to
to represent solder joints and allow for quite complex phys-
avoid decimation and aliasing of the data. An explicit
ics to be modeled using a very simple joint representation.
dynamic analysis can take tens of thousands to millions of
Boundary Conditions: Most shock tests involve some increments to complete, and writing results every incre-
amount of free-fall of the part under test, before impact. ment will generate an extremely large output file. Most
There is no need to model this rigid body motion in the post-processors have filtering tools to reduce the noise in
finite element model. Rather, the test part should be mod- an xy trace but, for the results to be truly valid, these filters
eled in the position just prior to impact and be given the need a full set of raw data to operate on. Some FE pack-
correct initial velocity to account for the free-fall. ages have runtime filters built into the analysis itself. This
Preloading: Most system-level components include some means that the analysis code will temporarily store results,
type of preloading - snap fit of plastic joints, compression filter it, and then write the filtered data to the output file.
of gaskets and seals, tightening of bolts or screws, com- This drastically reduces the amount of information that
pression of electrical connections, etc. It is not generally needs to be stored.
efficient to model these types of preloading operations with
the explicit dynamic method. The preloads are really quasi- E.3 Model Validation Techniques
static operations which would therefore be required to run
Introduction: Reality has a lot of challenges for modeling,
for a long time period, making them computationally
such as nonlinearity, uncertainty, variability and complex-
expensive given the small stable time increment required
ity. Modeling simulation is limited by engineering knowl-
when using explicit dynamics.
edge, understanding of physics and limited resources
Some FE packages also have implicit solvers which are (trained analysts, computers, software etc.). The growing
compatible with their explicit solver that allow for the need for virtual prototyping is demanding ‘‘smarter’’ (not

35
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
IPC/JEDEC-9703 March 2009

bigger) models that are adequate for its intended purpose Step 2: Correlate measurement points and FEM nodes
with the required level of accuracy for all necessary fea- Normally the degrees of freedom in FE model are much
tures. This requires model validation. higher than the number of test measuring points. Mode
shape truncation in the FE model is recommended since it
In this section we will suggest some commonly used model
is more conservative than mode shape expansion in the test
validation methods both in frequency and time domains.
model.
The modal method has been applied to dynamic FE analy-
sis for electronic boards. Modal test and analysis is an Step 3: Compare the natural frequencies
important and effective step in model validation. However, Step 4: Visual comparison of mode shapes
we believe there are two concerns if model validation is
Step 5: Correlate mode shapes
stopped here. First, the mode shapes and the frequencies
The modes may not be in the same order or in the same
are global measurements and may not account for localized
numbers between FEM and a modal test. Mode shapes par-
board deflection or bending which is directly related to the
ing is required before computing Modal Assurance Crite-
solder joint stress state. Secondly, modal tests are normally
rion (MAC).
conducted at a very low excitation input which may not
provide an accurate simulation for large inputs like a shock Step 6: Compute MAC
event. To address these concerns, a shock simulation model Mode shape validation can be done not only by visualiza-
should be rechecked using time domain measurements. tion through animation, but more importantly by comput-
ing the MAC, which is a correlation coefficient between the
Direct integration analysis has the advantage in handling
two mode shapes in Eq. (E-1) for real (normal) modes or
nonlinearity thanks to the FEA software capability and the
Eq. (E-2) for complex modes. If the MAC coefficient is
growing workstation’s computing power. The direct inte-
equal to 1.0, then the two shapes are perfectly correlated.
gration method solely relies on comparing model simula-
tion with direct board response measurements in the time For real (or normal) modes:
domain to validate the model.
Acceleration responses may be used for comparing in the MACjk = ?φ mj φak
T
?2

(φakT φak)(φmjT φmj) (E-1)

time domain, between model simulation and empirical
g

measured board level acceleration measurements. How- ϕak - kth eigenvector from analytical model
ever, as revealed in Section 8, using acceleration alone to ϕmj - jth eigenvector from measurement
gl

validate a model can lead to incorrect load states of solder

joints, impacting reliability assessment. Thus using either For complex modes:
modal or direct integration analyses, one should addition-
ally obtain a direct measurement of the board deflection
MACjk = ?φmj φak
H
?
2

and/or board strain to provide an accurate validation of a (φak

H
φak)(φmj
H
φmj) (E-2)
model.
where superscript H is Hermitian transpose, [A]H = ([A]T)*,
E.3.1 Modal Validation Two sets of data in frequency
*denotes the complex conjugate.
domain are available for a modal model validation. One set
is global parameters: natural frequency, mode shape and The computing results are in a matrix. The ideal MAC
damping ratio corresponding to each mode. Another set of matrix is an identity matrix where the coefficients on the
data is FRF (Frequency Response Function) at each mea- diagonal are 100% and all off the diagonal are 0%. Model
suring point. FRF has more data available without errors updating is required to improve MAC, increasing the
induced in extracting modal parameters and better for high diagonal and reducing those off the diagonal. Double check
modal density cases. However, validation by using modal the mode pairs and recomputed MAC before updating the
parameters is more efficient thus is most commonly used. model.
The following are suggested steps in using modal param-
The equations can also be used for checking the test data
eters to update models. Consult the literature for more on
or FEA data to confirm the validity of the data. All modes
other approaches.
should be orthogonal to each other.
Step 1: Pretest analysis
Local mode shape correlation can be computed by using
Before a modal test, use the finite element model (FEM) to
Mode Shape Difference (MSD):
help determine the locations for sensors and excitations to
avoid mode aliasing and help determine how many and MSD = {φa − φm}
which modes should be captured. Pretest analysis is recom-
mended to reduce the test effort and the validation effort. Model updating is required to minimize the differences.

36
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
March 2009 IPC/JEDEC-9703

E3.2 Time History Validation No matter which method

is used for FE modeling, a typical shock FE model valida- 1.0
Experiment

Displacement (normalized)
tion will compare board response time history measured
FE Model
across the board. The length of the time history should be 0.5
long enough to show the dynamic event and the validation
should be across the board. Peak value only or single point
comparison is not sufficient to validate a circuit board FE 0.0
model.
-0.5
When comparing the empirical measurements, make sure
the frame of reference for the measurement is the same as
that in the model. Board acceleration measurement from an -1.0
accelerometer is referenced to the earth. Board displace- 0 0.01 0.02 0.03 0.04 0.05
ment can be measured relative to the shock table when the Time (seconds) IPC-9703-E-6
reference is fixed on the table. If the board is mounted on
a rigid fixture, the displacement measured can be the board Figure E-6 Comparison of Displacement-Time History for
Experimental Data and FEA Result
deflection. If the board is mounted in a chassis, the board
deflection and the chassis deformation are coupled in the
measurement. Board strain measurements are referenced to 1.0

Principal Strain (normalized)

the existing board local flatness at the time zero before tak- Experiment
ing measurement. Thus board strain measurement can be FE Model
0.5
freely applied to compare across board level or system
level of shock tests. 0.0
An overlay plot is commonly used for time history com-
parison. Figure E-6 shows a comparison of the displace- -0.5
ment from the model and captured by using the high speed
camera system. Further confirmation was gained by com- -1.0
g

paring the board strains derived from the model with mea- 0 0.01 0.02 0.03 0.04 0.05
gl

surements using strain gages as shown in Figure E-7. As

Time (seconds)
can be seen from these figures the model captures the IPC-9703-E-7

behavior very well. Figure E-7 Comparison of Board Strain versus Time
History for Experimental Data and FEA Result
E3.3 Field Comparison (Contour Plots) Displacement or
strain contour plots have been useful for comparing Where:
between the simulation and the empirical measurements. Ua – Displacement from analytical model
When doing so, make sure not only the frame of references Um – Displacement from measurement
is matched, but also the measuring target grid density is
enough to derive accurate board displacement or strain The ideal DAC is 1.0, when two displacement shapes are
contour map. Similar to mode shape correlation, displace- perfectly matched.
ment shape (map) at a specified time can be validated by Local correlation can be computed by using Displacement
applying the Displacement Assurance Criterion (DAC): Shape Difference (DSD):

?U T
m Ua ?2 DSD (Ua ,Um) = {Ua − Um}
DAC(Ua ,Um) =
(UaT Ua)(UmT Um) Model updating is required to minimize the differences.

37
Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT
 ANSI/IPC-T-50 Terms and Definitions for
Interconnecting and Packaging Electronic Circuits
Definition Submission/Approval Sheet
The purpose of this form is to keep SUBMITTOR INFORMATION:
current with terms routinely used in Name:
the industry and their definitions.
Individuals or companies are Company:
invited to comment. Please
City:
complete this form and return to:
IPC State/Zip:
3000 Lakeside Drive, Suite 309S Telephone:
Bannockburn, IL 60015-1249
Fax: 847 615.7105 Date:
❑ This is a NEW term and definition being submitted.
❑ This is an ADDITION to an existing term and definition(s).
❑ This is a CHANGE to an existing definition.

Term Definition
g
gl

If space not adequate, use reverse side or attach additional sheet(s).

Artwork: ❑ Not Applicable ❑ Required ❑ To be supplied

❑ Included: Electronic File Name:
Document(s) to which this term applies:

Committees affected by this term:

Office Use
IPC Office Committee 2-30
Date Received: Date of Initial Review:
Comments Collated: Comment Resolution:
Returned for Action: Committee Action: ❑ Accepted ❑ Rejected
Revision Inclusion: ❑ Accept Modify

IEC Classification
Classification Code • Serial Number
Terms and Definition Committee Final Approval Authorization:
Committee 2-30 has approved the above term for release in the next revision.
Name: Committee: IPC 2-30 Date:

Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT

 ®

Standard Improvement Form IPC/JEDEC-9703

The purpose of this form is to provide the Individuals or companies are invited to If you can provide input, please complete
Technical Committee of IPC with input submit comments to IPC. All comments this form and return to:
from the industry regarding usage of will be collected and dispersed to the IPC
the subject standard. appropriate committee(s). 3000 Lakeside Drive, Suite 309S
Bannockburn, IL 60015-1249
Fax 847 615.7105
E-mail: [email protected]

1. I recommend changes to the following:

Requirement, paragraph number
Test Method number , paragraph number

The referenced paragraph number has proven to be:

Unclear Too Rigid In Error
Other

2. Recommendations for correction:

g
gl

3. Other suggestions for document improvement:

Submitted by:

Name Telephone

Company E-mail

Address

City/State/Zip Date

Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT

 g
gl

Association Connecting Electronics Industries

3000 Lakeside Drive, Suite 309 S

Bannockburn, IL 60015
847-615-7100 tel
847-615-7105 fax
www.ipc.org

Downloaded by tahereh mohammadpour ([email protected]) on Sep 15, 2018, 9:21 pm PDT