You are on page 1of 8

Concept, construction and first results of a test bench for automotive power nets

1 Introduction
The requirements to the electrical power supply in vehicles increased highly through
the last years and this trend will continue over the next few years without a great saturation. First the main problem was to guarantee a balanced charge of the main battery
despite the massive increase in the demand of electrical energy. This problem was
brought to heel through different kinds of energy management systems [1,2] using intelligent battery sensors (IBS) [3] to monitor the charge.

Concept, construction and first results of a

test bench for automotive power nets

Today dynamic deviations in the power net voltage are in the focus [4-9]. In addition
the requirements to the quality of the supply voltage increased. There are different
reasons for this development:

Dipl.-Ing. Johannes Kltzl

The number of driver and driving assistance systems, like electronic stability programs (ESP), active steering, reversible belt pretensioner (RBP) and so on, increased highly and the end of this trend has not been reached yet (toe and camber
adjustment systems, x-by-wire systems). This group of loads contains important
high-power-loads that show a very high and especially dynamic power demand in
critical driving situations.

85579 Neubiberg, Werner-Heisenberg-Weg 39, 089/6004-4726

In the context of efforts to reduce emissions more and more auxiliary components
get electrified. The most widespread example for this is the electric power steering
(EPS) [10]. The electrical vacuum pump and the air-conditioning compressor are
fractional in production or scheduling. The electrification of these components has
a great potential, because the power consumption can be decoupled from the speed
of the combustion engine. It is also needed for almost all kinds of hybrid drive
trains. However, these components have also a high maximum power demand.

Prof. Dr.-Ing. Dieter Gerling

Lehrstuhl fr elektrische Antriebstechnik und Aktorik (EAA)
Universitt der Bundeswehr Mnchen
85577 Neubiberg, Werner-Heisenberg-Weg 39, 089/6004-3708

As well it is essential for the reduction of emissions to reduce the weight of the vehicles. Because of that, reserves in the wire cross sections and the battery size are
reduced. Therefore the voltage drops across the feeding lines and the internal resistance of the battery increase.
Especially because of the first two reasons the difference between the dynamic of
power consumption and power generation is enlarged. An extension of the power
generation is not possible because of two reasons: On the one hand, the actual construction type of automotive generators is at the boundary of its power generation potential and its control is limited in its dynamic. On the other hand the engines get more
and more downsized for the reduction of emissions and the enhancement in efficiency. The increasing dynamic differences in power result in interaction with the
shown reasons in massive power net voltage deviations at the terminals of different

Concept, construction and first results of a test bench for automotive power nets
The worst case in terms of deviations in the supply voltage is the temporal superposition of two or more dynamic peaks in power consumption of high power loads in
combination with a high base load (BL), e.g. heating systems and engine fan, and a
low engine speed. [11]. Especially for chassis-, driver- and safety-systems a superposition in critical driving situations is very probable and not avoidable in every case.
The testing of the power net voltage in the case of such superpositions is essential, but
in test runs hard to provoke and first and foremost not repeatable.
Because of that, FEAAM GmbH built a test bench, which is able to test the interaction
of high power loads in real power nets in a simple and repeatable way under varying
basic conditions.
In this paper the concept of this test bench is shown, just as the construction and
planed upgrades. With some exemplary results the possibilities of the test bench are
presented and a comparison with vehicle measurements is made.

2 Concept

Concept, construction and first results of a test bench for automotive power nets
car body would be too big and expensive. The ground is alternatively routed by an
aluminium frame made of special section tubes.
To get a most realistic behaviour of the terminal sizes of the loads, the mechanical
loads of the components are simulated with realistic load profiles via adjusted actuators. For a correct function of the components and for the same behaviour as in the
corresponding driving situations, the correct communication environment is simulated
for every component. For this, extensive rest bus simulations are made on a real time
simulator. Furthermore on this real time simulator the features of the not implemented
communication gateway are simulated, to guarantee the correct communication between the different components. The simulation of the gateway features is also
needed, to get diagnostic links (OBDII) to the components, as well for diagnostic inspections as for training the components and software updates of the ECUs (electronic
control units). As the heart of the test bench the real time simulator is used for the test
sequence control and the control of the loading actuators. It is controlled and monitored itself via a control-PC.
A first overview over the test bench concept is shown in the following figure:







LIN, CAN, FlexRay






measuring system




Figure 1: Overview of the test bench concept



V /I




The ground routing is not designed originally, because the test bench should be useable for different vehicles, i.e. in prototype state. The effort for change of the whole






To limit the number of possible loading variants and not to redirect from the main
topic, the base loads like entertainment, heating and similar systems are not implemented in original, but are represented by a simple ohmic load. It is directly connected
to the central fuse box of the wiring harness.



To be able to display the interaction of all components as precise as possible, all components are implemented in the test bench as far as possible as original parts. These
are for the power supply the generator (Gen) and the battery (Bat), for the loads all
components of the steering column, which is composed of the steering wheel angle
sensor (SAS), the super imposed steering system (SIS), the steering link and the electric power steering (EPS). Also the reversible belt pretensioners (RBP) for the driver
and the co-driver, a variable rear axle differential (VAD), an air suspension compressor (ASC) and a wiring harness (WH) are implemented.


real time simulator


Requirements to the test bench are the usage of real vehicle components, as well production run components as prototypes, and the possibility to represent components
which are not existing yet or to complex to use in original. The measurement of voltages and currents of the different components has to be possible with minimum
1 kHz, as well as temperatures, but with a much lower sampling rate. Basic conditions
like nominal voltage, basic load, engine speed, battery condition and so on have to be
adjustable, as well as the timing of maximum peak power of all implemented loads.

Concept, construction and first results of a test bench for automotive power nets

Concept, construction and first results of a test bench for automotive power nets

In contrast, the acquisition of the voltages, currents and temperatures at the components and some other system parameters like the engine speed is not implemented in
the real time system, but is done by a measuring system which is controlled via an
own PC. This measuring PC also traces in addition to the mentioned signals the whole
communication on the installed bus systems like LIN, CAN and FlexRay. The Measuring-PC is synchronised by the Control-PC at the beginning and the end of every

3 Construction


The core of the test bench is build by a multi-processor real time simulator made by
dSpace GmbH. Beside several digital and analogue interfaces it is primarily equipped
with a lot of interfaces for the different state of the art automotive bus systems. As far
as the components are integrated with their ECU, they are controlled via rest bus
simulations on the dSpace-System. All the other components are controlled via the
analogue and digital interfaces. In these cases the ECU functions are partly emulated
on the real time system.
As well analogue controlled are five electronic loads of Hcherl und Hackl. These
electronic loads are used as representation for the not implemented components. For
this they are controlled via current or power profiles from vehicle measurement. It is
also possible to implement them in a control loop with the real time simulator to implement some relevant ECU functions like e.g. power derating.
The real test bench is divided into three levels as shown in figure 2. The lower level is
a machine basement of cast iron. It is mainly used for the heavier components and for
the loading actuators. The central level is made out of acrylic glass and sustains the
main parts of the wiring harness. Furthermore it is possible for one person to stand
and work on this level for reconstruction or service. On some points it is opened, so
that the wiring can be connected to the components. The upper level carries the electronic loads and the halogen headlamp, which are not only integrated as a part of the
basic load, but to be able to evaluate the optical power net stability (subjective), too.
The three level construction allows it to connect all components and loads to the wiring harness without any changes to it. The three levels are connected through an aluminium frame of special section tubes, which also carries all measuring and control
lines. The single tubes are connected resistively low to each other and to the machine
basement. It is obvious that this fits not the ground routing of the car body, but it is
similar enough for the intended measurements.

Figure 2: The three levels of the test bench



3.2.1 Battery and generator

As mentioned in the concept, battery and generator are implemented as original parts.
The batteries are preconditioned for the measurements, if needed. Also the so called
intelligent battery sensor is implemented and addressed via a LIN-interface.
At the same interface the voltage controller of the generator is addressed to be able to
send the rated voltage and other rated values to it and to monitor e.g. the degree of
utilisation. The generator itself is driven via the original belt pulleys with a gear ratio
of 3.25 by a servomotor, which is able to cover the hole operating range of the combustion engine. It is speed-controlled and can change the (engine-)speed depending on
the loading with up to 70 s-2. An easy change of the generator is possible despite of its
placement under a containment. The containments main function is the protection in
the case of a bursting belt. But it makes it also possible to heat the environment of the
generator with its own losses up to more critical operating conditions.

Concept, construction and first results of a test bench for automotive power nets

Concept, construction and first results of a test bench for automotive power nets

3.2.2 Basic load

To get no confusing number of load scenarios, but to get nevertheless the possibility
to bring the generator and the battery in a state of high utilisation the basic load is implemented as a variable ohmic load. It is connected directly to the central fuse box.

steering actuator

torque sensor

The ohmic load is realized with two water cooled cascades with 13 resistances 1,35
each. These resistances are connected with electronic relays in a way, that makes it
possible to change the basic load in steps of 10 A between 0 and 260 A at a terminal
voltage of 13.5 V.

steering wheel
angle sensor

3.2.3 Steering

super imposed

The actual most complex construction at the test bench is the test bench part for the
steering actuators. Into this structure the whole steering mechanics are built, starting
with a servo motor with a gear reduction that emulates the driver and imposes the
steering wheel angle. It can impose up to 20 Nm of steering torque. Directly after the
gear the steering wheel angle sensor is installed. It is needed, because as well the EPS
as the superimposed steering system, which is integrated in the steering link, needs its
information for correct functionality and error detection. The toothed rack of the
power steering is loaded with an electrical cylinder, which can bring up to 25 kN. The
whole construction is shown in Figure 3.
For state of the art electrical power steerings it is a little bit oversized, because the input power for these steerings is limited through the power net. But to guarantee full
usability for the next generation of EPS and power nets for e.g. heavier vehicles this
oversizing is justified and useful.
The control of the steering actuator and the electrical cylinder is not implemented in
dSpace, but on an own industrial PC with a real time operating system. It gets the
rated values for the steering wheel angle and the force on the toothed rack from
dSpace and gives the actual values back to it. The reason for this method is, that this
part of the test bench was not built in house, but by Klotz Special Machinery. In that
way the acceptance procedure was much easier and faster and the interfaces could be
defined very precisely.

3.2.4 Reversible belt pretensioners

The belt pretensioners are integrated together with their buckles. The buckles are not
connected directly with the frame, but via springs with a rigidity of about 7 N/mm.
The springs are used to emulate the yielding of the human body and the clothes. The
pretensioners are forced to action just through the bus communication via simulating a
full brake application.




Figure 3: Construction of the steering test bench

3.2.5 Others
The other components are integrated very similar to the already described ones. An integration of further components can be done relatively simple, as well as real components, as just as a simulation model in combination with an electronic load, because
there are still some interfaces in reserve.


Control system

For the test bench control there are actually two operation modes: The manual and the
automatic mode. Two more modes are planned: The driving an the replay mode.
The manual mode is mainly intended to be used for testing and for bringing into service of new components. Engine speed, rated voltage, basic load and so on can be
controlled and also monitored as well as the main communication values. Furthermore
it is possible to control the steering actuators and most of the implemented components in a very rudimental way. It is not possible to execute any kind of load profiles.

Concept, construction and first results of a test bench for automotive power nets
For this the automatic operation mode is used. It is based on a finite state machine,
which handles the whole measurement process and is shown in the following figure.

Concept, construction and first results of a test bench for automotive power nets
mode, because it is very valuable for fast testing and for working out the influence of
single factors, which is not possible in complex driving manoeuvres.
The fourth planned mode is a mode which can play back traces from driving tests, for
easy comparison. This mode e.g. can help to find out if a measured error or behaviour
was a single event or is a problem in principle.




Measurement system

The data logging is realised on an own Measurement-PC via the LabView-Software

of National Instruments. The measurement transformer is built by a rack with 50
modules for current and voltage measurement, as well as 16 channels for temperature
measurement. On demand it is possible to connect two special power meters and oscilloscopes via IEEE448-Bus or Ethernet to the Measurement-PC. All instruments are
controlled by a Lab-View program developed at FEAAM GmbH. With this program it
is possible to log all analogue channels with up to 10 kHz as far as supported by the
instruments. These data are written to file via a tdms-stream. This file gets converted
to a mat-file after measurement. For measurements with higher resolutions the oscilloscopes are used.


Figure 4: Finite state machine for automatic mode

It starts in a safe state, in which all interfaces are passive or in uncritical state. In this
state the user can set all measurement parameters, choose load profiles and set up the
load schedule. After this step the program is setting up chosen values, initializing the
steering test bench and checks the communication status. If everything works proper
the user is able to start the measurement process. After the measurement process is
completed, all components are shut down and not until this the measurement is
stopped. The same happens in the case of error detection. This procedure helps in the
error analysis, because the whole system behaviour before, during and after the occurrence of the error is logged.

After the start-up of the program, the user just has to load or create a configurationfile. The measurement itself is triggered by the real time processor.
The tracing of the bus systems communication is not implemented in LabView, even
though CAN and LIN measuring is possible with the FEAAM-program, because there
was no FlexRay-Support for LabView available at the time the test bench was constructed. So the bus tracing is realised via a CANCaseXL (LIN), a VN7600 (CAN,
FlexRay) and the CANape software by Vector Informatik. The tracing gets also synchronised from the real time simulator, so that the time stamps of the traces and data
logs are aligned. The traces are converted to mat-files after the measurement, too.

After the shutting down and an eventually needed acknowledgement of the error, the
system goes back into the safe state and is ready for a new measurement. During the
whole process the main system and component values can be monitored on the user
interface. A manual braking off is possible at any time, but always leads to a controlled shutdown, too.

4 First results

Two further operation modes are planned and will be implemented soon. First a driving mode, in which tracks and driving manoeuvres can be chosen and the whole vehicle dynamics is simulated. This mode is needed after the automatic mode, because not
all load superpositions which can be created in the automatic mode, can happen in the
same intensity during real driving. But it is not intended to replace the automatic


In this chapter some first results are shown. First, the possibility and the process of
load superposition will be demonstrated and later some comparisons between vehicle
and test bench measurements will be drawn.

Superposition measurements

As mentioned in the introduction, a main reason for building this test bench was to get
able to check worst-case situations for the power net, like the superposition of differ-


Concept, construction and first results of a test bench for automotive power nets

Concept, construction and first results of a test bench for automotive power nets

ent actuators of the chassis system and other high power loads, like driver assistance
systems etc..
To get the worst-case a calibration measurement is made first. In this measurement the
interesting loads are provoked one after the other without any superposition. This is
shown in figure 5 for two loads of the chassis system and one driver assistance system. The chassis system loads are implemented as real loads and the driver assistance
system via an electronic load in this case. This measurement gets then analyzed for
the points of occurrence of the peak current of every load. With this information the
schedule is changed in a way that the peak currents should superpose exactly.

Figure 6: Superposition of the maximum currents of different high power loads


Comparison to vehicle measurements

The best abilities to provoke worst-cases and to reproduce measurements exactly

would be worthless, if the measurement results are not comparable to vehicle measurements. To show the good match of the test bench with the real vehicle, some comparisons between measurements will be shown in this section. These measurements
are only two out of a whole row of measurements which were made to validate the
test bench.

Figure 5: Load currents of different high power loads

As it can be seen in figure 6 this process works very accurately. The peak currents are
now all in a slot of 1 ms. The difference of 1 ms is no mistake of the control system.
In fact the peak currents of the loads, except the electronic load, decrease in the moment of superposition because of the high voltage drop. So it can happen that the current in this moment is lower than some milliseconds before or after.
To provoke intentionally a superposition, like it is shown in figure 6, is nearly impossible and never exactly repeatable for different vehicle configurations. So here the test
bench shows its great potential.


First, the vehicle measurements were made, and later the measurements were repeated
at the test bench. Important values like engine speed, rated generator voltage and so
on were replayed as profiles at the test bench. Also the same battery and generator
were used to get almost the same boundary conditions. Just the voltage controller of
the generator had a different software version.
Figure 7 shows the terminal values of the battery, the generator and the hydroaggregate of the ABS during a full brake application. Also the engine speed nEng of
the full measurement is presented. What first attracts attention is the almost perfect
match of the both measurements of IABS. This is not surprising, because the ABS is
represented by an electronic load at the test bench. There is also a good match observ-


Concept, construction and first results of a test bench for automotive power nets
able in the dynamic current split between battery and generator. That there are greater
differences in the further process is caused through the differences in the voltage controllers. The same reason causes the differences in the fluctuations of the generator
voltage. The battery and the ABS voltage, in contrast, are barely influenced, and
match very well. The little deeper voltage drop at the ABS at the test bench is assumed to be caused by the smaller line inductance as a consequence of the missing car
body. The engine speed shows a good match, too.

Concept, construction and first results of a test bench for automotive power nets
generator voltage VGen. Apart from this small and easy to fix offset the two measurements match very precisely as well in the current split as in the voltage drops and the
dynamic behavior. The higher current fluctuations in the vehicle measurements are
caused by another load, which is represented by a constant ohmic load at the test
bench. The dynamic behavior of the load cannot be represented in that way.
With the two shown examples it can be demonstrated that the test bench is able to
substitute a lot of vehicle measurements with the needed precision.

Figure 8: Comparison of vehicle and test bench measurements during ASC switch off

5 Conclusion and Outlook

Figure 7: Comparison of vehicle and test bench measurements during full brake application

Figure 8 shows the terminal values of the battery, the generator and the ASC during
the switch off process of the ASC. The first thing, which is recognizable, is the difference in the battery voltage VBat. But it is just a constant offset of about 0.6 V which is
caused by the voltage controller of the generator. The nominal values are in both
measurements identical. The same offset can be seen not so clearly, because of the
higher voltage fluctuations at the generator, but with the same value in the measured


The concept and the construction of a test bench for automotive power net stability
were presented in this paper. A comparison to vehicle measurements and the ability to
arbitrary load current superposition were shown. It was demonstrated that it is able to
substitute a lot of vehicle tests. It is also qualified for early tests of new vehicle components, even before the test in prototype vehicles. It offers degrees of freedom which
cannot be provided by vehicle tests.
For further enhancements of precision it will be extended in the next months with a
battery simulator, which also helps to reduce the effort for battery conditioning and
increases the degree of reproducibility. Also the whole braking system will be integrated for a more precise representation of the driver assistance systems. For this also


Concept, construction and first results of a test bench for automotive power nets
the simulation of the whole longitudinal and lateral vehicle dynamics is needed. With
this simulation the mentioned driving mode will be realized, too. Furthermore, the integration of a car body is in discussion at the moment.

6 References

S. Bchner: Energiemanagement-Strategien fr elektrische Energiebordnetze in

Kraftfahrzeugen, Technische Universitt Dresden, 2010


R. Fabis: Beitrag zum Energiemanagement in Kfz-Bordnetzen, Technische

Universitt Berlin, 2006


H. Gering, Uwe Mhrstdt: Die Spannung halten, automotive 7-8.2008, Carl

Hanser Verlag, Mnchen, 2008


M. Hohmann: Spannungsstabilisierende Manahmen in Kfz-Bordnetzen, Visualisierung und Bewertung, ATZelektronik 01/2009, Springer Automotive Media,
Wiesbaden, Germany, 2009


R. Gehring, J. Frschl, T. Kohler, H.-G. Herzog: Modeling of the Automotive

14 V Power Net for Voltage Stability Analysis, VPPC, Dearborn, 2009


J. Pforr: Innovative elektrische Energienetze im Kraftfahrzeug, Forschungsbericht 2006, IAF, Fachhochschule Ingolstadt


H. Prbstle, D. Polenov; F. Przywecki, F. El-Dwaik: Advanced Power Net Architecture for BMW Micro-Hybrid Vehicles, Stuttgart International Symposium, Stuttgart, Germany, 2009


T. Kohler, J. Frschl, R. Gehring, T. Wagner, A. Thanheiser, C. Bertram. D.

Bcherl, H.-G. Herzog: Experimental Investigation on Voltage Stability in Vehicle Power Nets for Power Distribution Management, VPPC, Lille, 2010


L. Brabetz, M. Ayeb, P. Grbel, G. Jilwan: HiL Prfstand zur Charakterisierung

und Optimierung von Bordnetzen, Energiemanagement und Bordnetze III, Expert Verlag, Renningen, 2010

[10] A. Vhning, A. Gaedke, M. Heger, H.-C. Reuss: Investigation of Power and

Energy Consumption of Hydraulic and Electric Power Steering Systems, Stuttgart International Symposium, Stuttgart, Germany, 2009
[11] T. Kohler, J. Frschl, C. Bertram, D. Bcherl, H.-G. Herzog: Approach of a
Predictive, Cybernetic Power Distribution Management, EVS, Shenzhen, 2010