TRANSFER PATH ANALYSIS:

SOURCE CHARACTERIZATION
Theory and example cases on modular source descriptions
using blocked forces
TRANSFER PATH ANALYSIS:
SOURCE CHARACTERIZATION
Modular source description using blocked forces

Transfer Path Analysis (TPA) has been a valuable

THEORY
engineering tool for as long as noise and vibrations of SOURCE CHARACTERIZATION FUNDAMENTALS 03
products have been of interest. Classically, a TPA
CLASSIC TPA 04
looks into a product's actively vibrating components
and the transmission of these vibrations into the COMPONENT-BASED TPA 06
connected passive structures. VIBES takes this a step USING TPA FOR ENGINEERING INSIGHTS 08
further by incorporating Source Characterization – to
describe source systems independent from their
environment – together with test-based virtual point CASES
modelling and Dynamic Substructuring techniques.
These combine into a complete and powerful ZF FRIEDRICHSHAFEN ERC CASE 11
modular design strategy, ready to face the NVH and E-COMPRESSOR CASE 14
vibration challenges of the future.

In general the VIBES Methodology is comprised of

the following steps: SOFTWARE
1. Source Characterization, for instance by blocked SOFTWARE PROCESS 18 DIRAC 19 SOURCE 20
forces;
2. Transfer path modelling of the source-receiver
structure with the advantages of Dynamic
ABOUT VIBES
Substructuring; VIBES is the all-round expert in modular sound &
3. Noise predictions and assessment of the vibration engineering for the automotive and (high)tech
performance of the source receiver assembly; industries. Using the VIBES methodology, our clients
4. Product enhancement studies by virtual solve potential vibration issues before physical
modification and virtual exchange of the prototypes are built, saving time and resources, and
components. resulting in better products. VIBES offers engineering
This in-depth whitepaper provides the theoretical services and user-friendly software (DIRAC, SOURCE
background to our most-used techniques, illustrated and VIBES Toolbox for MATLAB) that speed-up your
by engineering examples. work and give reliable, traceable test results.

2
SOURCE CHARACTERIZATION
FUNDAMENTALS
The transfer path problem so translations and rotations. Obtaining the same
The typical transfer path problem is depicted in Figure 1, information from a test has traditionally been much more
where an actively vibrating subsystem (e.g. steering gear, challenging. The interfaces often consist of holes (see Fig. 2)
compressor, etc.) transmits energy into a passive subsystem and bolts, where measurement equipment cannot be easily
(e.g. the vehicle body). For simplicity, the excitation within placed. The Virtual Point Transformation has been
active subsystem A is depicted by a discrete point (node 1), developed and implemented in our DIRAC software to
transmitted through the interface at node 2, and causes the overcome these challenges. The Virtual Point methodology
response of interest on the passive subsystem B at node 3. enables full 6-DoF models (translations and rotations) to be
This response could be, for example, the sound level at the obtained from tests, similar to their numerical FEM
driver’s ear or perhaps the structural vibration on the seat equivalent, and is one of the key technologies that enables
rail. While this example only shows three Degrees of success of the TPA described above. Some notable benefits
Freedom (DoF), the general framework is developed which of the Virtual Point Transformation include:
may represent a larger set of DoFs including: 1. Virtual Points are aligned and positioned between
1. Source: internal DoFs belonging to the active different tests (e.g. full vehicle, test bench,
component that cause the operational excitation but components) in DIRAC. As such, results can easily be
are unmeasurable in practice; transferred from a test bench to the full vehicle.

2. Interface: coupling DoFs residing on the interface 2. With six DoF per interface, vehicle/component
between the active and the passive components; models can be built accurately in the mid and high
frequency range where rotational DoF become more
3. Receiver: response DoFs at locations of interest of significant.
the passive component.
3. The test-based FRF measurements are transformed
The example of Fig. 1 is illustrative for a wide range of to the format of a computer-based model. This gives
practical problems, provided that the structure of interest us the compatibility to easily combine test-based
can be decomposed into an active and a passive part. In models with FE-models for hybrid system
what follows, all methods assume that the operational predictions.
excitation at node 1 is unmeasurable in practice but As the interfaces grow in complexity (bolts, welds, etc.), we
transmits vibrations through the interfaces at node 2 to can use these Virtual Points to reduce the interface problem
receiving locations at node 3. to the essential connecting nodes. In practice, this means
performing FRF testing around the interface, as shown in
Fig. 2, and using the Virtual Point Transformation in DIRAC
to obtain the full 6 DoF FRF matrix at the Virtual Point. This
experimental model can then be used for highly accurate
transfer path analysis or frequency-based substructuring to
other components.

Figure 1: Overview of the source receiver problem for transfer path

analysis with the representative nodes 1-3.

This problem can be studied using several families of TPA

methods, where the differences in the methods lie in the
fundamental choices for how to represent the vibrations at
the interfaces.

Virtual point transformation

Figure 2: Screenshot of DIRAC showing the design of experiment for
Within Fig. 1, the interfaces are simply defined as “node 2”.
one connection point. The orange bubble represents the virtual
Using a Finite Element Model (FEM), information can easily point, the blue arrows show the impacts and the gray cubes give the
be extracted from this discrete interface node with 6 DoF, sensor position of the measurement set-up.

3
CLASSIC
TPA
The classic TPA approach of source characterization utilizes inconvenience of mounting transducers between the active
interface forces 𝐠 𝐁𝟐 to describe the source. These forces component and the receiving structure. Lack of space,
describe the excitations that source A excites on receiver B, distortion of the original mounting situation, and the
as depicted in Fig. 3. Their use for partial contribution incapability to measure all desired degrees of freedom at a
analyses (discussed later) makes interface forces a valuable connection point render this method generally impractical,
tool for determining which paths have the greatest especially for typical automotive applications. However,
contribution to the responses of interest. When the force transducers can be used effectively when built into a
interfaces forces 𝐠 𝐁𝟐 are known, they can be used to applied test bench setup to measure blocked forces. This is
to the interfaces of subsystem B to calculate the receiver discussed later on.
responses on the passive side with:

𝐮𝐁𝟑 = 𝐘𝟑𝟐
𝐁
𝐠 𝐁𝟐 (1)

𝐁
where 𝐘𝟑𝟐 is the transfer function matrix of receiver B alone
(decoupled from the source), with inputs at the interfaces
and responses at the locations of interest. It is assumed that
the receiver B is a purely passive structure with no external
forces. The interface forces themselves can be determined Figure 4: Schematic display of the mount stiffness method.
using a variety of methods.
Interface forces: mount stiffness method
In many automotive applications, the actively vibrating
parts are attached to the vehicle body using resilient
mounts in order to reduce vibration transmission, as
illustrated in Fig. 4. With proper tuning of the mount
flexibility (stiffness) and absorption (damping) properties, a
high level of vibration suppression can be achieved. The
mount stiffness method takes advantage of the fact that the
amount of vibration on either side of the mount will not be
Figure 3: Overview of the description of the vibration transmission the same: 𝐮𝐁𝟐 − 𝐮𝟐𝐀 is no longer zero. By combining the
through the interface by interface forces g B2 measured displacement on either side of the mount
(usually integrated from accelerometer data) with the
The variants of classical TPA are defined according to how stiffness of the mount 𝐙𝐦𝐭 , we can calculate the interface
the interface forces 𝐠 𝐁𝟐 are obtained from operational tests. forces:
One way to obtain these forces is through a direct
measurement, using force transducers mounted between 𝐠 𝐁𝟐 = 𝐙𝐦𝐭 (𝐮𝟐𝐀 − 𝐮𝐁𝟐 ) (2)
the active and the passive side. In some cases it can be very
difficult or impossible to directly measure at the interface. Although the mount stiffness method can be powerful and
Therefore a number of indirect methods have been easy to conduct, it can be challenging to get an accurate
developed to circumvent the need for a direct force characterization of the mount stiffness properties 𝐙𝐦𝐭 ,
measurement. The most commonly seen methods are particularly at the higher frequencies and amplitudes. The
summarized here. static stiffness of the mount provided from the supplier or
measured on a test-bench is likely only appropriate at the
Interface forces: direct measurement very low frequencies, above which a more sophisticated
The most straightforward technique to obtain the interface characterization is warranted. This can for instance be
forces is from force transducers mounted directly between inverse substructuring using a virtual-point FRF model of
the active and the passive side. This approach requires that the mount with measurement adapters, or a full decoupling
the transducers at the interface are stiff enough in the approach.
frequency range of interest, such that the assumption of
𝐮𝐁𝟐 = 𝐮𝟐𝐀 is valid. The main drawback of this method is the
4
 With proper selection of indicator sensors, the matrix
inverse technique is usually quite accurate and relatively
straightforward to execute. Like with any method where a
matrix-inversion is solved, proper conditioning of the FRF
matrix and the operational data is crucial. Using virtual
points for the description of the interface DoFs 𝐠 𝐁𝟐 already
Figure 5: Schematic display of indicator points 𝐮𝟒 in the proximity of helps in achieving a better conditioning, as one has more
the interface. control over de actual DoFs that get inverted. For example,
the XYZ force vectors are naturally orthogonal using virtual
Interface forces: matrix inverse method points and can be put nicely in the center of the connection
The perhaps most popular method for determining points, which can be troublesome otherwise.
interface forces is the matrix-inverse method. It was
observed in Eq. 1 that responses at the passive side are SUMMARY
obtained from the application of interface forces to the
FRFs of the passive subsystem. This equation can also be With the classic TPA approach, interface forces can
inverted to determine the interface forces. To perform this be determined. These forces are not independent
inversion, it is necessary that the FRFs contain sufficient of the source and are only valid for the assembly
independent responses to describe all interface forces and they were calculated in. The most known approach
moments. The set of receiver responses 𝐮𝟑 is typically too is the matrix-inverse method.
small in number and too distant from the interfaces to be
suitable for inversion. In practice, to improve the
conditioning of this FRF matrix, we equip the passive
structure with additional indicator sensors 𝐮𝟒 , as shown in
Fig. 5. The 𝑛 indicator sensors are positioned in close
proximity to the interfaces such that the full set of 𝑚
interface forces is properly observable, using 𝑛 > 𝑚
sensors. As these indicator DoFs merely assist in the
determination of the interface forces, the sensor type can
be chosen rather freely, although (tri-axial) accelerometers
are the most common choice. The indicator sensors are
used to calculate interface forces as:

𝐠 𝐁𝟐 = (𝐘𝟒𝟐
𝐁 + 𝐀𝐁
) 𝐮𝟒 (3)

Two sets of measurements are now required to execute

equation 3:
• operational indicator responses 𝐮𝟒𝐀𝐁 , obtained
from test cycles on the full assembly;
• 𝐁
an FRF test of the passive component 𝐘𝟒𝟐 , so after
dismounting of the active component.
First, the indicator sensor responses 𝐮𝟒𝐀𝐁 are measured on
the assembled system AB during operational tests. Next,
FRFs are measured for the passive subsystem B relating the
motion at these indicator points to the forces at the
𝐁
interface, namely 𝐘𝟒𝟐 . Note that this requires dismounting
of the active component(s) from the assembly.
𝐁
Nevertheless, the 𝐘𝟒𝟐 FRFs (needed for the matrix-inverse
𝐁
force determination) and the 𝐘𝟑𝟐 FRFs (to calculate
responses at the target locations, Eq. 2) can be obtained
from the same FRF measurement campaign, as it only
involves mounting of additional sensors.

5
COMPONENT-BASED
TPA
Component-based TPA basics
A fundamentally different class of TPA methods is that of
component-based TPA. Using classical TPA, the interface
forces between two subsystems will reflect the dynamics of
both subsystems (source and receiver) and are therefore
not transferable to an assembly with a different receiver. Figure 6: Representation of the vibration transmission through the
For this reason, classical TPA cannot be efficiently used to eq
interface by equivalent forces 𝐟2 .
assess subsystem modifications, as one would need to
conduct a new operational test for every change in design. Direct blocked forces
This is a well-known drawback and in fact the limiting factor Consider the following scenario. The operating source is
for applying classic TPA methods in the early phase of NVH rigidly clamped on its interface so that the interface
design, where changes to source and receiver are still likely vibration 𝐮𝟐 is zero, see Fig. 7. The reaction forces in the
to be made. clamped support (“blocked forces”) 𝐠 𝐛𝐥𝟐 ensure that:

Component-based TPA remedies this problem by

𝐮𝟏 𝐘𝐀 𝐀
𝐘𝟏𝟐 𝐟𝟏
characterizing the source excitation with a set of equivalent [ 𝐀 ] = [ 𝟏𝟏 𝐀 ] [𝐠 𝐀 = −𝐠 𝐛𝐥 ]
(6)
𝐮𝟐 = 𝟎 𝐀
𝐘𝟐𝟏 𝐘𝟐𝟐 𝟐 𝟐
forces that are an inherent property of the active
component itself, independent of the receiver it is attached
to. This framework facilitates testing of the active This yields:
component in a broader set of test environments (e.g. in
the full vehicle or on a test bench), as the equivalent forces 𝐠 𝐛𝐥 𝐀 −𝟏 𝐀
𝟐 = (𝐘𝟐𝟐 ) 𝐘𝟐𝟏 𝐟𝟏 (7)
will not be influenced the dynamics of the receiving
𝐞𝐪
component. As the equivalent forces are only a function of 𝐟𝟐 = 𝐠 𝐛𝐥
𝟐 (8)
the active component itself, we denote this family of TPA as
component-based. Using component-based TPA, the
responses at the receiving side are simulated by applying
the equivalent forces to the FRFs of the assembled system,
as illustrated in Fig. 6 and equation 4.

𝐀𝐁 𝐞𝐪
𝐮𝟑 = 𝐘𝟑𝟐 𝐟𝟐 (4)

There are two ways to think about this family of TPA. In one
𝐞𝐪
way, we desire a set of equivalent forces 𝐟𝟐 that would Figure 7: Schematic display of the source rigidly mounted on its
interface.
cancel out the vibrations from the active source 𝐟𝟏 to yield
a completely motionless interface 𝐮𝟐 when both sets of
The blocked-force method assumes the boundary to be
forces are applied concurrently in opposing directions. This
infinitely stiff in all directions, which in practice is rarely the
is where the term “blocked force” comes from, which is
case. Hence the accuracy of the blocked forces is subject to
often used interchangeably with “equivalent force”.
the stiffness of the boundary relative to the component at
𝐀𝐁 𝐞𝐪 hand. An additional difficulty is the measurement of
𝐮𝟐 − 𝐘𝟐𝟐 𝐟𝟐 = 𝟎 (5)
rotational moments, as most commonly used sensors are
Another way to think of component-based TPA is that we unable to measure collocated 6-DoF interface loads. As a
desire a set of equivalent forces that will cause the same set consequence, the blocked-force method is expected to
of responses 𝐮𝟑 as when the active component is running perform best at the low frequency end for which the rigid
in the assembly (caused by 𝐟𝟏 ). This is really the essence of boundary assumption is most valid and rotational effects
Eq. 4. are in practice least prominent.

6
Free velocity Reviewing Eq. 13, we see that the two inverted FRF matrices
The direct counterpart of the blocked-force method is the represent the dynamic stiffness of the test rig and the active
free-velocity method as depicted in Fig. 8. In this case the component at the interface DoFs. We can simply add these
component's interfaces are left free, such that all vibrations stiffnesses to get
are seen as ‘free displacements’ 𝐮𝐟𝐫𝐞𝐞
𝟐 at the interface DoFs. 𝐞𝐪 𝐑 𝐀 𝐀𝐑
𝐟𝟐 = (𝐙𝟐𝟐 + 𝐙𝟐𝟐 ) 𝐮𝟐 = 𝐙𝟐𝟐 𝐮𝟐 (14)
From here on, equivalent forces can be calculated by
inverting the admittance measured at the free subsystem's
Transforming back to admittance notation, we find that the
interfaces, which can be understood by comparing the free
blocked forces can be calculated inversely from the
displacements with the blocked force definition.
admittance of the assembly's interface. Or, more
𝐮𝟏 𝐀 𝐀 preferably, the blocked forces can be calculated in an
𝐘𝟏𝟏 𝐘𝟏𝟐 𝐟𝟏
[ 𝐀 𝐟𝐫𝐞𝐞 ] = [ 𝐀 𝐀 ] [𝐠 𝐀 = 𝟎]
(9) overdetermined fashion using a sufficient set of indicator
𝐮𝟐 = 𝐮𝟐 𝐘𝟐𝟏 𝐘𝟐𝟐 𝟐
points 𝐮𝟒 around the interface of the receiving subsystem,
similar to what was seen in the classical matrix-inverse TPA.
This yields:
𝐞𝐪 𝐞𝐪
𝐀𝐑 −𝟏
𝐟𝟐 = (𝐘𝟐𝟐 ) 𝐮𝟐 or 𝐀𝐑 −𝟏
𝐟𝟐 = (𝐘𝟒𝟐 ) 𝐮𝟒 (15)
𝐮𝐟𝐫𝐞𝐞
𝟐 = 𝐀
𝐘𝟐𝟏 𝐟𝟏 (10)
As the equivalent forces are a property of the active source
𝐞𝐪
𝐟𝟐 𝐀 −𝟏 𝐟𝐫𝐞𝐞
= (𝐘𝟐𝟐 ) 𝐮𝟐 (11) alone, these forces could be calculated in an assembly of
the source system on a test rig R, or more simply in the
target assembly AB (see Figure 9) with
In practice, it is almost impossible to measure under these
𝐞𝐪 𝐞𝐪
conditions. For one, we must settle for a “quasi-free”
𝐀𝐁 −𝟏
𝐟𝟐 = (𝐘𝟐𝟐 ) 𝐮𝟐 or 𝐀𝐁 −𝟏
𝐟𝟐 = (𝐘𝟒𝟐 ) 𝐮𝟒 (16)
condition above the decoupling frequencies. Secondly, it is
As the name implies, this enables a true in-situ calculation
often challenging to run the active component in operation
of equivalent forces in the vehicle, where it is not necessary
in a free boundary condition.
to dismantle any of the parts. All that is needed for this
calculation is:
• an operational test of the full assembly with the
indicator sensors around the interfaces;
• an assembly FRF test between those indicator
sensors and the forces at the interfaces.
Figure 8: Schematic display of the source freely suspended on its
interface.

In-situ characterization
As both of the methods described above pose their own
practical limitations, another more versatile method is
often used. In practice, the active component will often be
measured on a test rig whose dynamics fall somewhere Figure 9: Schematic display of the in-situ method for the full
between the completely rigid boundary and free conditions assembly.
as described above. We can combine the direct force
(equation 8) and free velocity (equation 11) methods to
give:
SUMMARY
𝐞𝐪
𝐟𝟐 = 𝐠 𝐑𝟐 + (𝐘𝟐𝟐
𝐀 −𝟏
) 𝐮𝟐 (12) With the component-based TPA approach,
However, we typically will not have the direct force equivalent forces can be determined. These forces
measurements 𝐠 𝐑𝟐 available and instead formulate the are a property of the actively vibrating component
approach as follows: only and are thus transferable to an assembly with
another passive side.
𝐞𝐪 𝐑 −𝟏 𝐀 −𝟏
𝐟𝟐 = (𝐘𝟐𝟐 ) 𝐮𝟐 + (𝐘𝟐𝟐 ) 𝐮𝟐 (13)

7
USING TPA FOR
ENGINEERING INSIGHTS
The theory as described above is in-line with the general Combine and virtually improve
framework for TPA1, with additional remarks on practical Hybrid NVH models can be built and combined with the
application. The following section addresses more blocked force loads measured on the component test
practicalities and relevant aspects for implementation of benches. The response at the point of interest is accurately
source characterization techniques in NVH processes. computed into the kHz range. The responses calculated
enable a complete assembly optimization where certain
Blocked force process in practice properties, e.g. bushing stiffnesses, can easily be modified
As described before, blocked forces are an equivalent set of and renewed responses can be calculated. As the blocked
loads that produce the same response on the passive forces are independent of the source components
component as in the operational condition. These themselves, one may go through the (virtual) optimization
equivalent forces are a function of the active component as many times as desired, maintaining accuracy in the
alone and are independent of what it is attached to. This is prediction.
useful if a given component (e.g. engine, steering gear,
compressor) is used in multiple vehicle platforms. It also On-board validation & control inputs
eliminates test bench influences usually found on classic Blocked forces obtained in-situ can be validated using the
component test benches. No matter if one conducts the so-called on-board validation points (index: ov). These are
blocked forces with the in-situ method on a test bench or in additional sensors positioned on the same assembly (‘on-
the target assembly the process is the same. board’) that was used for the identification of the forces.
A blocked force campaign can be subdivided in transfer The validation data recorded during the operational tests is
function measurements and operational measurements. compared to the predicted results using the equivalent
For the FRF measurements we highly recommend using our forces:
software DIRAC. DIRAC provides an efficient environment to 𝐀𝐁 𝐞𝐪
𝐮𝐨𝐯 = 𝐘𝐨𝐯,𝟐 𝐟𝟐 (17)
prepare and analyze your measurements while measuring
and avoids typical miscommunication about interfaces, This requires additional sensors during the source
measurement points and directions, et cetera. The output characterization procedure. The validation points can be
from DIRAC is a virtual-point transformed FRFs model which seen as additional indicator points. However, they are not
can directly be used to calculate blocked forces with the in- included for the computation of the equivalent forces. This
situ method. allows to use them for validation purposes.
After operational measurement, one can obtain blocked Similarly, it is recommended to provide control inputs for
forces in a few simple steps with our software product the calculation, by means of additional excitation points.
SOURCE. Here you can load in the operational These can be used to assess the completeness of the
measurements performed on your DAQ system of choice equivalent force description.
through for instance ATFX files. The FRF model can be
mapped easily to the measurement to directly calculate
blocked forces for all operational states. These forces can
DIRAC & SOURCE
be combined to make predictions and (virtually) improve DIRAC and SOURCE are designed to understand
your assembly. The workflow is summarized in Fig. 10. component-based TPA principles. In DIRAC, one
can already assign sensors to be an indicator,
target or on-board validation sensor. SOURCE
understands these “roles” and configures the
source characterization automatically.
Figure 10: Schematic process of source characterization.

1
M. V. van der Seijs, D. J. Rixen en D. de Klerk, “General framework for
transfer path analysis: History, theory and classification of techniques,”
Mechanical Systems & Signal Processing, vol. 68, pp. 217-244, 2016.
8
Partial contributions The following equation governs the conversion:
Aside from the source characterization, we are often 𝐞𝐪
𝐁 )+ 𝐀𝐁
interested in understanding which paths are the most 𝐠 2 = (𝐘42 𝐘42 𝐟2 (19)
critical between each of the individual interfaces and the
This can also be expressed as:
responses of interest (e.g. the driver’s ear). These “partial
contributions” are computed by multiplying individual 𝐀 𝐁 −𝟏
𝐀 𝐞𝐪
(20)
𝐠 𝟐 = (𝐘𝟐𝟐 + 𝐘𝟐𝟐 ) 𝐘𝟐𝟐 𝐟𝟐
forces with the corresponding FRF/NTFs.
The set of receiver responses 𝐮𝟑 are in this framework often This is very useful when one for example conducted a
considered as the end result of a TPA method. However, the blocked force characterization on a test bench for a specific
actual purpose of a TPA is rather to identify the individual component. These blocked forces are an independent
contributions that sum up to these responses, because property of the source only and can thus be transferred to
these contributions reveal the amount of vibrations new assemblies.
transmitted through each path. Receiver response spectra One can also use the conversion of these blocked force to
are composed from the complex sum of the partial determine interface forces for the new receiving structure.
response spectra, i.e. the sum of all contributions over the The only requirement is to have an FRF measurement of the
considered transfer paths. To illustrate for classical TPA the receiving structure B (often called the trimmed-body). This
p receiver response spectra ui are composed from their allows to virtually determine partial contributions based on
path contributions as follows: interface forces from pure blocked forces, without the need
to conduct measurements on the new assembly.
𝑢𝑖,𝑗 = 𝑌𝑖𝑗B 𝑔𝑗B (18)

The sum over 𝑗 for the partial response results in the full INSIGHTS FOR EVERY STAGE
prediction at this specific point.
VIBES.technology is constantly innovating ways to
The partial contributions are typically computed using the generate valuable engineering insights in all stages of
interface forces. This allows us to focus on how the forces the design process, and to connect these along the way.
travel across the passive component only as we multiply the This focuses on:
𝐁
interface forces with 𝐘𝟑𝟐 . This is usually what we are
interested in. One could technically compute partial • Component test bench strategies. The best
strategy for independent and cost-effective source
contributions using the blocked forces with:
characterization really varies per component.
𝐀𝐁 bl VIBES specializes in finding the best approach and
𝐮3 = 𝐘32 𝐟2
providing the right test bench design and software
These path contributions would include dynamics of both
to implement into your development process.
the active and passive components as we multiply the
𝐀B
blocked forces with Y32 . This is not what we are typically • Cross-validating TPA results. Oftentimes, results
interested in. However, blocked forces have other essential from just one analysis method may not provide
qualities and use cases for vehicle simulation and can be enough certainty to inform critical design changes.
By cross-validating several approaches, the
“converted” to interface forces!
required certainty can be gained for a robust
choice.
Conversion from blocked to interface forces
A very practical conversion allows to use blocked forces and • New ways to study path contributions. Partial
an FRF measurement of the receiving structure B responses are just one way to study the relative
importance of transfer paths. Other breakdowns
decoupled from the source to calculate interface forces at
can be added based on complex energy flow,
the same interface points. This combines the strength of
balance between translations and rotations,
blocked force and interface force based methods. sensitivity to changes in bushing stiffness and
Note that, in order to warrant any conversion, virtual point more.
transformations are strictly required to guarantee Get in touch to find out what more VIBES has to offer
compatibility between the DoFs in the various datasets. for your process!
Using DIRAC, this is of course easily satisfied.

9
11 ZF FRIEDRICHSHAFEN
ELECTRIC ROLL CONTROL
CASE
In this project, an electromechanics roll controller (ERC) is the vibration
source of interest. The ERC is built into high-class vehicles to increase
the driving performance.

E-COMPRESSOR 14
SOURCE CHARACTERIZATION
CASE
The goal of this study is to show the successful application of independent source
characterization for an electric air cooling compressor.

10
ELECTRIC ROLL CONTROL
FOR ZF FRIEDRICHSHAFEN
An insightful and successful project has been conducted with The baseline set of forces includes three translational
the German automotive supplier ZF from Friedrichshafen. In Degrees of Freedom (DoF) per coupling point, and the
this project, an electro-mechanic roll-control (ERC) is the necessity of including rotational DoF was also investigated.
vibration source of interest. The ERC is built into high-class The Virtual Point Transformation (VPT) is used to calculate
vehicles to increase the driving performance and comfort. In the moments at the rotational DoF, while the blocked forces
the future, we expect OEMs to be interested in the blocked at the translational DoF are calculated both with and
forces of components like the ERC. It was therefore chosen without the use of VPT. Thus, three approaches were
to use this component as an example for general considered:
investigations regarding the applicability of component-
1. Six DoF interface with VPT
based TPA approaches in practice.
2. Three DoF interface with VPT
Component 3. Three DoF interface without VPT
The ERC enhances the driver’s comfort and safety through
The primary goal of this project was to validate the blocked
agile cornering and steering. It stabilizes the vehicle around
force methodology for the ERC. Additionally, we wanted to
the X-axis using an electric engine and a high-torque gear.
provide the customer with insights into the quality of the
As seen in Fig 11, the ERC is mounted to the vehicle at four
obtained data from the vehicle or test bench. The final goal
connection points: two at the wheel hubs and two at the
was to demonstrate the transferability of the blocked forces
subframe. The ERC is connected directly to the wheel hubs
and prove that they are indeed an independent property of
while the subframe connections are isolated with rubber
the source itself, as stated in the theory.
bushings.
From an NVH perspective, it is important that the driver Blocked force determination
does not experience any noise originating from this As no test bench was available, the blocked forces were
component. In some vehicles the ERC has caused a high- determined using the in-situ method in the full vehicle. Two
pitched sound from the electric engine, and in other sets of measurements are needed to calculate the in-situ
vehicles rattling upon load changes from gear backlash. By blocked forces:
characterizing the loads of the ERC with blocked forces, we
1. FRF measurements of the full vehicle 𝐘𝐀𝐁 (vehicle
should be able to predict this type of noise in new
body + ERC) and (2) operational measurements
assemblies before actually building it into the assemblies.
with the same set-up of sensors.
2. Vehicle modifications were also made to
demonstrate the transferability of the blocked
forces, and thus the FRF and operational
measurements were also repeated on the
modified vehicle.

FRF measurements
Full-vehicle FRF tests were performed with sensors in close
proximity to each of the coupling points, along with some
additional validation sensors and microphones inside the
Figure 11: Picture of the electric roll control by ZF. One can see the
car. As mentioned, the ERC connects to the vehicle at two
electric engine gear assembly in the center of the component. The locations on the subframe and to both wheel hubs. As the
four connections points are depicted by orange circles. wheel hub connections are ball joints which cannot transfer
moments, it was decided to capture the forces at the wheel
Goal hub interfaces using only three DoF, while both three DoF
The main goal of this project is the blocked force and six DoF scenarios were considered at the subframe
characterization of the ERC at the four connection points. connection points.

11
Thus, three impacts were used to excite the three
orthogonal directions at each wheel-hub connection point,
and a single triaxial accelerometer was used to capture the
response (see Figure 12). Where more DoF were desired at
the subframe interfaces, additional impacts and sensors
were needed. Six sensors were placed in close proximity to
each of the connection points on the subframe: three on
Figure 14: The first picture shows where the steel plate was
the active side (on the ERC) and three on the passive side mounted to the subframe. The right graph shows the difference in
(on the subframe). Transmission simulators were attached the FRFs caused by the modification of the vehicle.
close to the connection points to allow proper sensor and
impact positioning near these interfaces, with increased Operational measurements
accuracy and accessibility of all six DoF. At each subframe The operational measurements are conducted with the
connection point, 13 impacts were used to excite the same set of sensors to obtain 𝐮𝟒 and 𝐮𝟑 . In total, we differ
system. As discussed in a following section, this set of between three different operational cycles:
impacts was then used in three different ways to calculate
the various sets of blocked forces and moments. 1. Impacts on the housing of the ERC
2. Manual actuation of the ERC
3. Actuation of the ERC on a test track
The first cycle is a very simple operational state obtained by
exciting the housing of the ERC with an impact hammer.
With this type of “operational” condition, the repeatability
is high and you don’t have to activate the ERC itself. The
disadvantage is that this operational state is not
representative of an actual operational state. The second
Figure 12: The picture on the left shows a photo of the measurement operational state was the manual actuation of the ERC
set-up for one subframe connection point with a mount. The right component. Using a remote control, one is able to drive the
picture shows the same set-up in DIRAC, including the VP defined ERC from one extreme (minimum angle) to the other
for this connection point.
(maximum angle). The last and most realistic operational
measurement was conducted on the test track by driving
At each of the four connection points, a Virtual Point is
over rubber mats with different thicknesses. During these
defined in DIRAC. The Virtual Point is placed exactly at the
measurements, only the left wheels drove over the plates
ERC/vehicle interfaces where we will calculate the blocked
while the right wheels stayed on the road. This caused
forces. Figure 13 shows the position of these VPs.
rolling of the vehicle around the X-axis and therefore an
activation of the ERC. Cycles 1 and 2 were repeated with the
modified vehicle to obtain 𝐮𝟒,𝐦𝐨𝐝 and 𝐮𝟑,𝐦𝐨𝐝 .

Figure 13: The ERC mounted to the subframe. The four virtual
points are shown indicated by the blue circles.
Figure 15: The photo shows the test track for the one sided drive
over of the rubber mats with different thicknesses.
To demonstrate blocked force transferability to a different
vehicle, the vehicle was modified by mounting a seven- Blocked force calculation
kilogram steel plate close to the subframe. With this
From the operational and FRF measurements, one gets the
addition, the dynamics in this region change drastically.
𝐮𝟑,𝐦𝐨𝐝 validation signal and the 𝐮𝟒 indicator responses in
Figure 14 shows a comparison between the FRFs of the
the frequency domain, and the FRFs from the connection
original and modified vehicles. From this set of tests, one 𝐀𝐁 𝐀𝐁 𝐀𝐁
points 𝐘𝟒𝟐 and 𝐘𝟑𝟐 , 𝐘𝟑𝟐,𝐦𝐨𝐝 to the indicator and validation
gets 𝐘𝐀𝐁 and 𝐘𝐀𝐁,𝐦𝐨𝐝 for the full vehicle.
points for both, the modified and unmodified vehicle. Now
all the required data is collected to calculate blocked forces

12
and to make a transfer validation. The blocked forces in Results of application
general are calculated as follows (without any Now we can apply all three sets of blocked forces obtained
regularization): in the original vehicle to the FRFs of the modified vehicle
and compare these predictions with the measured
𝐟2bl = (𝐘42
AB +
) 𝐮4 (1) responses. Figure 17 shows a comparison of the measured
and predicted responses at the seat rail during the second
As these blocked forces should be transferable to other
operational state. It is immediately apparent that the third
assemblies, they can now be used to predict the responses
pred case (yellow), where the Virtual Point Transformation was
𝐮3,mod of the validation points in the modified vehicle using: not used, performs much worse than the other two cases.
pred AB
At the lower frequencies, below 800 Hz, both the three- and
𝐮3,mod = 𝐘32,mod 𝐟2bl (2)
six-DoF cases match the validation measurement well.
Above 800 Hz we see the six-DoF case outperforming the
These predictions are then compared to the measured
three-DoF case, as rotations become more critical.
signals. In this project, the validation points included a
microphone at the driver’s ear and a sensor on the seat rail.
As mentioned above, three different approaches were used
to obtain blocked forces. For all cases, the wheel hub
interfaces were described using 3-DoF Virtual Points on
each side; this is reasonable as the ERC connects to the
wheel hubs with a ball joint. The three cases instead focus
on different descriptions of the forces/moments entering
the body at the subframe interfaces.

Figure 17: Application of the ERC blocked forces (calculated using

three methods) to the modified vehicle, showing accurate prediction
of the response at the seat rail using Virtual Points with six degrees
of freedom per interface.

Takeaways
Figure 16: Screenshots from DIRAC. The left shows the virtual point
transformed approaches for 3-DoF and 6-DoF calculation. The right This study demonstrated several useful properties inherent
shows the approach where no virtual point transformation was used. to the blocked force methodology.

The first approach, which is generally recommended as it is 1. Blocked forces are transferable between different
assemblies. While demonstrated using simple
the most complete, uses the Virtual Point Transformation
vehicle modifications here, these blocked forces
to calculate blocked forces for six degrees of freedom at
could also be applied to other vehicles that use the
each interface. The second approach also uses the Virtual ERC.
Point Transformation, but only includes the three
translational degrees of freedom per interface. Figure 16 2. The Virtual Point Transformation greatly improves
(left) shows the impacts and sensors that were used for the quality of results, even when only three DoF are
considered. This is likely because we average away
both cases. The impacts on the crosses were transformed
the measurement inaccuracies inherent to impact
to the Virtual Point described using six DoF forces and
testing.
moments (case 1) as well as three DoF forces (case 2). In
both cases, all six sensors were retained in the FRF matrix 3. Rotational DoF become important for getting
to help identify the forces. The third approach also uses accurate results at the higher frequencies. The
three DoF blocked forces to describe each interface, but Virtual Point Transformation enables us to
accurately characterize the rotational DoF for better
this time without use of the Virtual Point Transformation. In
results at the higher frequencies. Note that the
this case, figure 16 (right) shows the impacts and sensor on
frequency at which rotations become important will
the subframe used in the blocked force calculation. From
vary for other components.
figure 16, it already becomes apparent that using the Virtual
Point with many impacts (left) could help to mitigate the From these results, it is recommended to characterize the
measurement inaccuracies inherent with impact testing. ERC using blocked forces and moments with six degrees of
freedom per interface via the Virtual Point Transformation.

13
ELECTRIC COMPRESSOR
SOURCE CHARACTERIZATION
2020 has already been filled with exciting technological The goal
project at VIBES, collaborating with several key players in The main goal is the characterization of the ELAC by a set of
the automotive industry in Europa and Asia. Most of these blocked forces, projected to the center of gravity (CoG) of
projects include source characterization using blocked the ELAC. For the calculations, a six degree of freedom
forces. (6DoF) Virtual Point (VP) is placed in the center of gravity.
The goal of this study is to show the successful application Two different methods were used to calculate the blocked
of independent source characterization for an electric air forces, as will be discussed further on.
cooling compressor (ELAC). We take several approaches to The determined forces were used for the synthesis of the
compute blocked forces and demonstrate the robustness of structure-borne compressor noise at a structural response
the characterization. point and on the drivers ear. The goal of having two sets of
blocked forces for the same operational states is to prove
The component the transferability and overall robustness of the
Due to the lack of “masking noise” from a combustion characterization.
engine in electric vehicles, the noise of auxiliary
components is becoming a relevant topic. One of these
components is the electric air cooling compressor (ELAC). In
addition to temperature control of the vehicle interior, the
ELAC also performs tasks such as cooling the battery and
power electronics during operation and especially fast
charging.
Throughout industry, one can observe different mounting
concepts for e-compressors. In the presented case, the
component is rigidly mounted with three connection points
to a bracket. This bracket is again mounted with three
rubber bushings to the E-drive.

Technical approach
Because of the low frequency range of interest (0-500 Hz)
the source description by blocked forces is sufficient to be
derived in the CoG . The underlying assumption here is that
the ELAC behaves as a rigid body in the frequency range of
interest. This was verified by identifying the first eigenmode
of the compressor. Secondly, it was verified by looking at
the virtual point consistency. Another reason for describing
the compressor in the CoG and not at the connection points
Figure 18: DIRAC screenshot of the CAD geometries illustrating the is to avoid the amplification of measurement noise due to
mounting of the ELAC (blue) with a bracket (gray). bad matrix conditioning.

14
The following two approaches will be used to determine
blocked forces: RESULT
1. ELAC mounted: In-situ characterization using a From the operational and FRF measurements, one
matrix inversion procedure; gets the 𝐮𝟑 validation signal and the 𝐮𝟒 of the
2. ELAC freely suspended: blocked forces calculated indicator sensors at the brackets in the frequency
from ‘free vibrations’. domain, and the FRFs from the CoG 𝐘𝟒𝟐 𝐀𝐁
and 𝐘𝟑𝟐
𝐀𝐁

For both applications, the virtual point transformation (VPT) to the indicator and validation points.
will be used to express the blocked forces in the CoG. Next
to that, the VPT ensures the transferability of the blocked
forces, to the current vehicle or to another compressor /
receiver combination.
The validation of the two sets of blocked forces will be done
with a component TPA synthesis to a structural response
point and to a microphone in the vehicle’s interior.

Figure 20: Schematic drawing of the ELAC mounted condition for

the in-situ blocked force characterization.

Figure 19: The ELAC and the bracket with a virtual point in the
center of gravity.

ELAC mounted
The E-compressor is rigidly mounted onto a bracket, which
is in turn resiliently mounted to the front electric motor.
The E-motor is also resiliently mounted to the vehicle’s
body. To obtain blocked forces, operational measurements
and FRF measurements are conducted with the same set of
sensor positions. Two accelerometers per connection point
on the bracket were used as indicator sensors for the matrix
inversion. In addition to that, sensors on the seat rail and
Figure 21: The mounted condition of the ELAC in the vehicle. One
microphones at the driver’s ear were used for validation. can see the top right connection point of the ELAC to the bracket
Operational measurements were conducted for distinct which is equipped with two sensors.

constant speeds from 800 RPM up to 9000 RPM and for

With these results, one can calculate the blocked forces and
linear sweeps over the full range. The operational
conduct an on-board validation. The blocked forces are
conditions were recorded at the sensor and microphone
calculated as follows (no regularization was used):
positions and later transformed into the frequency domain.
Next, an impact measurement was conducted for impact 𝐟2bl = (𝐘42
AB +
) 𝐮4 (1)
points on the compressor housing as shown in Fig. 19. pred
These forces can now be used to predict 𝐮3 into to the
These impacts are transformed into virtual point forces and
validation points inside the car. These predictions can be
moments in the center of gravity of the compressor. This
AB compared to the measured signal. This procedure is called
yields the transfer function 𝐘32 from the center of gravity
AB an on-board validation and produces typically good results.
to the validation points and 𝐘42 to the indicator points on
the bracket.
15
The following component TPA predictions are calculated:
pred AB bl
𝐮3 = 𝐘32 𝐟2 (2)

The predictions were validated at the driver’s ear

microphone and at a sensor at the seat rail.
The results in Fig. 22 show that the first order of the
compressor at 4000 RPM is dominant in the driver’s cabin.
One can observe that the blocked force prediction matches
the first order quite well. With the blocked forces one can
even predict below the noise floor.
As mentioned, the on-board validation typically produces
quite good results. To further assess the transferability of Figure 23: The freely-suspended condition of the ELAC. One can see
the blocked forces, we determined an alternative set of the compressor freely suspended with rubber ropes (yellow) in the
blocked force under freely-suspended conditions and engine compartment.
applied these to the vehicle FRF. The purpose is to prove
the transferability of blocked forces. With these results, one can calculate the blocked forces for
the free condition. The blocked forces are given by the
following equation:

𝐟2bl,free = (𝐘22
A + free
) 𝐮𝟒 (3)

These forces can again be used to make a prediction

pred,free
𝐮3 to the validation points inside the car. These
predictions can be compared to the measured signal and
the predictions from the previous set of blocked forces. This
procedure is called transfer validation and can prove the
source independency of the blocked forces. The prediction
is calculated similarly:

pred,free AB bl,free
𝐮3 = 𝐘32 𝐟2 (4)
Figure 22: The on-board validation for the seat rail sensor. The
results are shown for the state at 4000 RPM. The predictions were again validated at the driver’s ear
microphone and at the seat rail sensor. Fig. 25 shows the
ELAC freely suspended comparison between the measured signal and the
For the freely suspended measurement, the ELAC was prediction from the blocked forces obtained from the free
dismounted from the vehicle and suspended by rubber condition and the mounted condition.
ropes (Figure 23 & 24). This time, six sensors are placed on
the ELAC itself. Again, operational tests and FRF
measurements were conducted with the same set of
sensors. For the operational tests, the same operational
states were recorded as in the mounted approach.
For the FRF measurement the ELAC was turned off again
and the impacts were placed on the housing of the ELAC.

RESULT
Figure 24: Schematic display of the ELAC freely suspended
As a result of the operational and the FRF condition for In-Situ blocked force characterization.
measurements, one obtains the 𝐮𝐟𝐫𝐞𝐞 𝟐 of the
indicator sensors in the frequency domain and the
FRFs from the CoG 𝐘𝟐𝟐 𝐀
to the indicator points.

16
 Figure 25: Comparison between the measured signal and the
predictions from the blocked forces obtained from the freely
suspended condition (green) and the mounted condition (blue).

RESULT
Both characterizations yield the same result, which
means that the blocked forces are indeed
independent of the testing conditions. This is even
true for very different boundary conditions,
meaning that the ELAC characterization is not
affected by the mounting condition. It was also
verified that the individual blocked forces show a
high similarity between in-situ and freely
suspended configuration. This finding gives the
automotive OEM and supplier(s) freedom to
choose the most effective characterization
method.

What’s next?
One powerful way to further use the identified blocked
forces is the combination with transfer paths obtained from
dynamic substructuring (DS). DS can be used to simulate
modular transfer path models. Together with the blocked
force source description, new isolation/packaging concepts
for the compressor can be studied for new platforms and
vehicles. This allows to create virtual acoustic prototypes of
assemblies without the need to physically build them.
Variants, for example different bushing stiffnesses, can be
virtually exchanged and evaluated. This saves a lot of time
and work in the development phase and could avoid late-
phase troubleshooting.

17
SOFTWARE PROCESS
Effective blocked forced calculations

For the most effective solution for blocked force calculation, VIBES recommends using two software
applications: DIRAC and SOURCE. Both applications come with specific features perfectly suited to simplify the
process of source characterization and response prediction.

DIRAC in combination with SOURCE ensures:

• Design of Experiment using the unique 3D preparation capabilities of DIRAC.

• Virtual Point blocked forces and moments. This creates 6 degrees of freedom per coupling point,
indispensable to capture all source characteristics.

• Transferability of blocked forces between test bench and vehicle, as the virtual points guarantee true
compatibility at the connecting points.

• Higher bandwidth of validity, thanks to rotational FRFs and innovative test-based modelling features.

• More quality control in the blocked force calculation process.

DIRAC and SOURCE cover the entire workflow for blocked forces from start to end.

MORE
INFO
+31 (0) 85 744 0970
[email protected]
www.vibestechnology.com

18
DIRAC
A new way of measuring
DIRAC allows engineers to accurately measure Virtual Point transformed FRFs and NTFs. These can be combined
with simulated FRFs, to make reliable full-vehicle models available earlier in the development process. DIRAC
ensures traceability of results and indicates the quality of the measurement. DIRAC includes checks for outstanding
model quality: from the measurement preparation with DoF contribution insights, signal checks, classical FRF checks
(passivity, coherence), up to checks on the resulting test-based model.

Overall, DIRAC helps to avoid the final-phase troubleshooting – which can postpone vehicle production – and
reduces the number of prototype variants needed, thus saving valuable time and resources.

MORE
INFO
+31 (0) 85 744 0970
[email protected]
www.vibestechnology.com

19
SOURCE
Software for Blocked Force source characterization and TPA

Source Characterization (SC) and Transfer Path Analysis (TPA) are two highly popular topics in sound & vibration
engineering, as they enable a modular NVH approach. Accurate characterization of active sources, typically by
Blocked Forces, requires a clear and robust workflow with quality checks at each and every step.

SOURCE is the software application that enables you to do all of this. It is the result of years of method and software
development in the field of Transfer Path Analysis (TPA), blocked forces and substructuring. As it seamlessly reads
and writes ATFX data, it can be used in virtually any engineering environment alongside the standard data acquisition
systems. This includes common software solutions in the NVH industry such as Siemens LMS, Head Acoustics, Brüel
& Kjær and Müller-BBM. Results and insights generated by SOURCE fit perfectly in current CAE simulations, allowing
to solve critical NVH aspects in an early phase of product development.

MORE
INFO
+31 (0) 85 744 0970
[email protected]
www.vibestechnology.com

20
GET IN TOUCH FOR MORE INFO
We look forward to help you tackle your challenges!

MORE
INFO
+31 (0) 85 744 0970
[email protected]
www.vibestechnology.com

21