2007-01-2372

Improving Diesel Sound Quality on Engine Level and

Vehicle Level a Holistic Approach
Philipp Sellerbeck, Christian Nettelbeck
HEAD acoustics GmbH, Herzogenrath, Germany

Ralf Heinrichs, Thomas Abels

Ford Werke GmbH, Cologne, Germany
Copyright 2007 SAE International

ABSTRACT
Diesel impulsiveness (so called Diesel knocking) present
in the cabin of diesel vehicles is perceived as unpleasant
because of its impulsive time structure. JD Power data
clearly show the customers preference of vehicles with
little Diesel knocking over those with severe knocking.
Corresponding objective descriptors that reflect the
customers perception are introduced.
The occurrence of such noise patterns is influenced by
the combustion process itself as well as by all excited
mechanical components within the power train. Further
the transfer characteristics of the engine structure and
various vehicle noise paths do contribute to a poor
Diesel Sound Quality. It is essential that all these factors
have to be considered in combination. This paper
provides an overview about suitable methods and
technologies, including Binaural Transfer Path Analysis
and Synthesis. The potential of the approach is
demonstrated by an example. Investigations are carried
out on the complete vehicle and on the engine test
bench, and principle improvements on engine level and
on vehicle level are derived.

INTRODUCTION
Turbocharged direct-injection (DI) Diesel engines in
passenger cars are gaining popularity due to their good
fuel economy and high output torque. Ambitious CO2
emission reduction targets are supported by growing
market shares of Diesel-powered vehicles especially in
Europe.
However, usually both interior and exterior sound quality
of Diesel cars doesnt reach levels of refinement
competitive with their gasoline-powered counterparts. A
dominating factor is the mid- and high-frequency noise
generated by the DI combustion process. The excitation
is caused by a steep rise of cylinder pressure after selfignition of the mixture. In combination with the inherent
modulation by cylinder firing this leads to an impulsive

noise character. The so called Diesel knocking is

mainly covering the frequency range of 500 Hz to 6 kHz.
Human hearing is highly sensitive to broad-band
impulsive noise, thus annoyance is caused even if the
absolute level is comparably low. Beside the
impulsiveness itself the cylinder-to-cylinder variation is
perceived as unpleasant too. Therefore it is important to
apply proper analysis methods to first get sufficient
insights into the physical properties of a Diesel power
train and second to be able to reflect the customer
perception regarding Diesel Sound Quality.
Previous investigations [1] show that the 2 dominating
factors 'Pleasantness' and 'Powerfulness' do describe
the Diesel Sound Quality. Further it is shown that Diesel
Knocking is the major driver for a poor Diesel Sound
Quality. In particular the operating condition Idle did
turn out to be important for the Customer.
JD Power is the standard customer questionnaire in US
and Japan and is becoming more and more popular in
Europe too. The 2005 JD Power APEAL question
'Sound of engine while idling' is chosen to verify the
Diesel Knocking importance for Sound Quality and to
define proper customer driven objective targets.
Modern common-rail Diesel injection systems offer the
flexibility for NVH tuning, but this is often antagonized by
emission and performance calibration. Especially the
trade-off between emission and noise targets is
tightened with EU5 and US bin5 regulations coming.
Thus possibly even higher combustion excitation levels
are expected.
With the combustion process offering less acoustic
optimization potential, the transfer characteristics of the
engine structure and the vehicle noise paths become
more and more important factors for sound quality
improvements. Especially airborne and structure borne
transfer paths including engine mounts and sound
package offer various parameters for noise control and
sound design. Matching the acoustic properties of the

vehicle to the noise characteristics of the engine

requires mutual support and holistic efforts of involved
departments.
Thus an effective NVH optimization of the complete
vehicle is appropriate in order to achieve high levels of
refinement and good customer acceptance.

modulations can be detected. These patterns are

created by a single cylinder and cause a disturbing
noise.
Engine Orders
0.5
1
1.5

Engine Orders
0.5
1
1.5

COMBUSTION NOISE ANALYSIS AND

PERCEPTION
SUITABLE ANALYSIS METHODS
The noticeable time structure of Diesel combustion noise
and the human ears sensitivity to modulation call for
dedicated time-based analysis methods suitable for the
detection of combustion noise patterns. The common
methods for target setting based on frequency spectra
turned out to be insufficient. As directivity plays an
important role in case of high-frequency noise binaural
recording and signal processing technology is used
throughout.
High-Resolution Spectrogram

Figure 2: Modulation spectrogram of interior noise at

idle condition, binaural recording (left and right channel)
In particular a high frequency resolution modulation
analysis such as NBMA (Narrow Band Modulation
Analysis) is capable to differentiate between different
types of knocking contribution e.g. combustion noise and
injector ticking [2]. In Figure 3 the modulation patterns
from the abovementioned recording can be recognized
as peaks at respective frequencies.

In contrast to standard third-octave or FFT analyses, a

high-resolution spectrogram based on 48th octave
filtering provides both the necessary time and frequency
resolution. Figure 1 shows the analysis of a binaural
interior noise recording of a passenger car equipped
with a four cylinder DI Diesel engine (idle condition).
Diesel knocking patterns are clearly visible at 1.1-1.2
kHz, 1.4-2 kHz and 3-4 kHz. Differences between left
and right channel can also be seen.

Figure 3: NBMA of interior noise at idle condition,

binaural recording (left channel). The dotted black line
shows the average spectrum and the colored lines show
the high resolution engine order modulation cuts.
Relative Approach
Figure 1: High-resolution spectrogram of interior noise
at idle condition, binaural recording (left and right
channel)
Modulation Spectrum
A modulation analysis reveals details about the
composition of modulations and thus about the
perceived sound quality. Moreover the Diesel Knocking
severity and characteristic can be depicted with a
modulation analysis since the modulations carry all
impulsiveness information. Figure 2 shows a modulation
spectrogram of the same recording as in Figure 1.
Besides 2nd order modulations inherently generated by
the four cylinder engine also strong 0.5th order

Based on a model of human hearing, Relative Approach

detects signal changes in the time domain and in the
frequency domain and thus highlights disturbing noise
patterns [3]. The analysis proves especially useful for
transient conditions like run-up or tip-in. For such
operating conditions, in a spectrogram combustion noise
may be masked by other power train noises. The left
diagram in Figure 4 presents the Relative Approach
analysis of the interior noise recorded in a different
vehicle during a run-up from 1500 to 2500 rpm at low
load, a transient condition during which severe Diesel
noise was found. Obviously several modulated noise
patterns can be detected at 1.2 kHz, 2.4 kHz, 3.4 kHz
and 5.8 kHz while other noises are hidden. The right

diagram shows a high-resolution spectrogram for

comparison.
Correlation : APEAL Score <> DKI
(Diesel : B, C & CD Segment)

CD Segment
B, C Segment

8.5
8.3

Vehicle 7

Vehicle 5

Vehicle 2

8.1
APEAL Score

Note:
DKI only applicable
for vehicle interior

Vehicle 8

Vehicle 11

7.9

Vehicle 6

Vehicle 3
Vehicle 5

7.7

Vehicle 1

7.5

Vehicle 1
Vehicle 9

Vehicle 10

7.3

Vehicle 2

7.1

R2 = 0.81

Vehicle 7

Vehicle 8

Vehicle 4

6.9
Vehicle 6

R2

6.7

Figure 4: Relative Approach Analysis (left) and highresolution spectrogram (right) of interior noise at run-up
condition, respectively left channel of binaural recording
DIESEL KNOCKING INDEX DKI
Based upon NBMA a single number has been
developed in order to quantify the Diesel Knocking
performance of a vehicle and to support objective target
setting. Figure 5 shows the DKI equation which
comprises a weighted summation of all relevant
modulations and level in dB(A).
.25

Carrier frequenc y f C

8
6
4

.13

2
0

10

20

30

40

DKI=

1
n
o

O(step=0.5)

w M

i=0.5

eo

i,o

Lfu fl

N / dB

Modulation frequency fM

Figure 5: Single number DKI composition for interior

noise at idle condition
Figure 6 shows that the Diesel Knocking Index DKI
delivers an excellent correlation to the JD Power
question 'Engine sound while idling'. Further it turns out
that meanwhile the customer's expectation with regard
to absence of Diesel Knocking is very high for CD class
vehicles. Moreover one may assume that this trend will
also become true for C and B class vehicles within the
next couple of years.
With that knowledge the automotive OEM is now able to
define targets according to his marketing strategy.
Developing a best in class vehicle in terms of Diesel
Knocking can be achieved with either having a perfectly
running power train without any knocking or with a
perfectly isolating power train installation, means very
well isolating transfer path. An ideal method/tool for
supporting the Power train <> Installation Trade-Off
discussion is the Binaural Transfer Path Analysis and
Synthesis. Herewith all imaginable Trade-Off scenarios
can be played through without any hardware changes
i.e. in the CAE world.

Vehicle 3

= 0.89

Vehicle 4

6.5
0.9

1.1

1.2

1.3

1.4

DKI

Figure 6: JD Power APEAL score correlation with Diesel

Knocking Index DKI for the question 'Engine sound while
idling'

INVESTIGATIONS ON VEHICLE LEVEL

Once Diesel knocking patterns have been defined in the
interior noise, the transfer of the noise from engine to the
ears is investigated.
BINAURAL TRANSFER PATH ANALYSIS AND
SYNTHESIS
Binaural Transfer Path Analysis and Synthesis (BTPA
and BTPS) technology has been developed in order to
find the origins of unwanted noise patterns and to predict
sound quality in vehicles, not only in terms of numbers
and graphs, but also for binaural auralization. The
results can be used to create an accurate model for
path-related noise heard anywhere inside the car, thus
yielding possibilities for troubleshooting and sound
design. Each individual path contribution or combination
of such may be listened to independently, in order to
assess their respective impact on the overall sound
quality. Paths can be modified synthetically to simulate
countermeasures and their effect on the interior noise
The method makes use of FIR-filtering of measured
excitation signals in order to create synthesized sound
contributions. The filters are based on measured transfer
functions including the phase - between source and
receiver locations, taking into account both air-borne and
structure-borne noise paths. Binaural recording and
playback technology is used in order to achieve a
realistic sound reproduction. Time domain signal
processing is mandatory for capturing transient noise
patterns like Diesel knocking. More details about the
method and related technologies can be found in [4, 5].
INTERIOR NOISE CONTRIBUTIONS
For the operating condition described in Figure 1 and
Figure 2 (idle) a binaural transfer path synthesis has
been carried out at which 39 structure borne paths and

10 airborne paths have been used. Figure 7 shows the

spectrograms of the synthesized interior noise along
with the structure borne and airborne contributions.
Several knocking patterns can be clearly identified.
While combustion noise at 1.2 and 1.6 kHz is transferred
mainly structure borne, patterns at 1.2, 1.4, 1.8-2 and 34 kHz are transferred mainly airborne.

respective interior noise contributions are shown in

Figure 9. While the 1.2 kHz noise is radiated from all
engine sides, the other noise patterns can be assigned
to certain engine surfaces.
right up/dn

left

front

top

rear up/dn

bottom

Figure 7: High-resolution spectrogram of BTPS results

for idle condition, top: left channel, bottom: right channel
Consecutively the interior noise contributions are broken
down into single transfer paths. Figure 8 contains BTPS
results of 12 transfer paths related to the left and right
engine mount featuring two attachment points each
(front, rear) analyzed in 3 directions (x, y, z). The 1.2
kHz pattern originates mainly from the front attachment
of the left engine mount, x-direction and the 1.6 kHz
noise is transferred via the right engine mount, ydirection.
right engine mount
front y rear y
front x

Figure 9: Interior noise contributions of engines

airborne radiation, top: left channel, bottom: right
channel
Also DKI values can be calculated from the synthesized
time signals, supporting the subjective listening and
making an easy ranking of transfer paths possible, e.g.
Figure 10 shows the air borne noise contributions.
Airborne noise DKI contributions
1,1

left engine mount

DKI Left Ear

1,0

DKI Right Ear

0,9

DKI

0,8
0,7
0,6
0,5
0,4
rear down

rear up

bottom

top

front

left

right front

right rear

Bin_IG_LS

0,3

Transfer Path

Figure 10: DKI of interior noise contributions

Figure 8: Interior noise contributions of left and right

engine mount paths, top: left channel, bottom: right
channel
The airborne noise paths are based on eight
microphones positioned in the engine compartment. The

Once the significant noise paths contributing to the

combustion noise heard inside the vehicle are identified,
more detailed information about the transfer
characteristics on the one hand and about the relevant
sources on the other hand can be made available from
the measured data.
TRANSFER PATH INFORMATION

Structure borne Paths

As shown in Figure 8 a strong knocking pattern at 1.61.7 kHz is transferred via the right engine mount in ydirection. The effective mount transfer functions related
to the two attachment points show no specific weakness
at the frequency of interest, neither do the vibroacoustic
transfer functions from the body attachment points to the
ears (Figure 11 and Figure 12). Thus the knocking
pattern must be mainly due to strong excitation from the
engine.

right up
right dn
left
front
top
rear up
rear dn

right up
right dn
left
front
top
rear up
rear dn

Figure 13: Binaural airborne transfer functions of the

engine compartment, top: left channel, bottom: right
channel
Figure 11: Effective mount transfer function (F/a) of right
engine mount (y), front (red) and rear (green)
attachment point

REFINEMENTS
In order to develop refinements in principle, noise
reduction can be applied at the source and at the
transfer path:

Reduction at the source

o Use of absorbers
o Change
of
excitation
(combustion
calibration, piston clearance etc.)
Reduction at the transfer path
o Change of mount
o Improved body stiffness
o Modified sound insulation package
o mending of leaks

BTPS results can directly lead to solutions for

improvements and the method can be used to predict
potential for optimization.
Experimental approach
Figure 12: Vibro-acoustic transfer function (p/F) of right
engine mount (y), front (red) and rear (green)
attachment point, top: left channel, bottom: right channel
Airborne Paths
Listening to the airborne interior noise contributions
reveals that the 3-4 kHz combustion noise transferred
from the upper rear side of the engine is most disturbing.
This problem can be directly related to insufficient
attenuation of the respective transfer path, as shown in
Figure 13. Further the 1.8-2 kHz pattern detected in the
interior noise can be related to a peak in the lower right
side transfer function.

As an example, an absorber was used on the engine

side of the right mount in order to reduce excitation
levels. Figure 14 shows that the measured acceleration
(y-dir) at the engine side of the mount carries mainly the
1.5-1.8 kHz pattern. With the absorber attached, the
acceleration is significantly reduced at the frequency of
interest. Interior noise measurements reveal the effect
on the overall sound quality, Figure 15. This result
validates the significance of the transfer path predicted
by the BTPS.

Predictive Approach
The transfer path model has also been used to predict
the overall effect of specific path modifications as
presented in Figure 17.

Figure 14: Vibration at right engine mount bracket (y),

left: original state, right: absorber attached

Figure 17: Interior noise synthesis, left original, right:

synthetically improved airborne attenuation of upper rear
engine side, top: left channel, bottom: right channel

FFT: 512, Mod-FFT: 512, Overlay: 0.50

FFT: 512, Mod-FFT: 512, Overlay: 0.50

mDKI: 0.503

0.8

0.6

0.4

1.0

0.5 EO
1
40
2
30
20
10

49.9 dB(A)

mDKI: 0.483

0.8

0.5 EO
1
40
2
Level [dB(A)]

50.0 dB(A)

Modulation Index

1.0

Level [dB(A)]

Figure 16 shows the absorber effect on the critical

modulations for the artificial heads right hand side ear.
The NBMA plots display a significant reduction in the 1.5
to 1.8 kHz region. Further the 3 to 3.5 kHz region is
improved with regard to 2nd E.O. modulation; thereby the
JD Power score can be enhanced by 0.2 scale points.

Modulation Index

Figure 15: Interior noise measurement, left: original,

right: with absorber at right engine mount bracket, top:
left channel, bottom: right channel

The idle interior noise synthesis of the original state is

compared to a synthesis based on a virtual improvement
of the airborne noise DKI contributors (compare Figure
10) upper rear side and bottom. The most disturbing
combustion noise pattern at 3-4 kHz could be reduced
significantly, leading to a JD Power score improvement
of 0.35 scale points.

0.6

0.4

10
0

10

0.2

0.6

0.4

0.2
0

0.0

0.0
2
3
Carrier Frequency [kHz]

mDKI: 0.517

0.5 EO
1
30
2
Level [dB(A)]

Modulation Index

0.6

0.4

1.0

20

10

41.4 dB(A)

mDKI: 0.498

0.8
Modulation Index

40.3 dB(A)

0.8

0.6

0.4

0.5 EO
1
30
2

20

10

0
0.0

0.0
1

2
3
Carrier Frequency [kHz]

mDKI: 0.498

0.8

0.6

0.4

2
3
Carrier Frequency [kHz]

Figure 16: NBMAs for interior noise measurements, left:

original, right: with absorber at right engine mount
bracket, top: left channel, bottom: right channel

2
3
Carrier Frequency [kHz]

FFT: 512, Mod-FFT: 512, Overlay: 0.50

1.0

0.5 EO
1
40
2
30
20
10

49.0 dB(A)

mDKI: 0.479

0.8

0.6

0.4

0.5 EO
1
40
2
30
20
10

0.2

0
0.0

0.2

0.2

49.1 dB(A)

FFT: 512, Mod-FFT: 512, Overlay: 0.50

FFT: 512, Mod-FFT: 512, Overlay: 0.50

1.0

2
3
Carrier Frequency [kHz]

0.2
1

Level [dB(A)]

FFT: 512, Mod-FFT: 512, Overlay: 0.50

1.0

20

10

2
3
Carrier Frequency [kHz]

Level [dB(A)]

0.4

20

0.8

0.5 EO
1
30
2

Modulation Index

0.6

mDKI: 0.490

Modulation Index

Level [dB(A)]

Modulation Index

0.8

39.9 dB(A)

0.0

0.0

FFT: 512, Mod-FFT: 512, Overlay: 0.50

1.0

0.5 EO
1
30
2

Level [dB(A)]

mDKI: 0.494

Modulation Index

40.2 dB(A)

Level [dB(A)]

FFT: 512, Mod-FFT: 512, Overlay: 0.50

20

0.2

0.2

1.0

30

0
0.0

2
3
Carrier Frequency [kHz]

2
3
Carrier Frequency [kHz]

Figure 18: NBMAs for interior noise synthesis left:

original,
right:
synthetically
improved
airborne
attenuation of top and upper rear engine side, top: left
channel, bottom: right channel

INVESTIGATIONS ON ENGINE LEVEL

A most effective noise refinement is achieved if vehicle
transfer path optimizations are complemented by
acoustic improvement of the sources. Several
investigations can be accomplished at engine level in
order to deduce modifications. If a certain noise problem
may be addressed both at the engine and at the transfer
path, interior noise simulations enable the engineer to
select the most effective and/or most economical
solution.

Furthermore, high-resolution time and frequency-based

analyses of cylinder pressures reveals characteristic
properties of the excitation, namely pressure gradient
and oscillations due to cylinder cavity resonances. Both
are important factors influencing Diesel sound quality [6].
Figure 20 illustrates that a Wavelet analysis can be used
to generate a spectrogram of the excitation frequencies
due to a single firing event. The noticeable spot at 3-4
kHz is caused by combustion bowl resonances and
corresponds to the knocking pattern described above.

With the results of vehicle level investigations already

present, engine measurements can be focused on
certain source locations or noise patterns. If applicable,
engine test bench measurement points (i.e. sensor
positions) should include but not be limited to the
points used in the vehicle investigations.
ENGINE TEST BENCH MEASUREMENTS
Data from standard measurement procedures can be
combined and used with advanced analysis methods.
Usually available data include:

far field and/or near field microphone signals

cylinder pressures
acceleration at engine mounts
crank angle

As an example the correlation of cylinder pressures and

microphone signals is demonstrated in Figure 19. Due to
the high resolution of the analysis the modulated noise
pattern can be assigned to a certain cylinder if the
distance to the microphone is taken into account.

Figure 20: Wavelet analysis of cylinder pressure during

firing event showing cavity resonance
Additional measurements provide helpful information
regarding the engine operation and transfer of
impulsiveness:

acceleration at cylinder block and head surfaces

acceleration at main bearings
injectors current

COMBUSTION NOISE SEPARATION

Figure 19: Cylinder pressure traces and high-resolution

spectrogram of test bench microphone signal, noise
pattern can be associated to specific cylinder

Diesel engine sound quality greatly depends on

combustion noise characteristics. Thus separating the
noise contributions that are originating by combustion
excitation from the total engine noise is crucial for NVH
optimization. Several methods exist for this task and
related applications [6, 7, 8, 9, 10]. An easy procedure
that proved sufficient for most applications is based on
measuring engine noise during motored (dragged)
operation and during fired operation at various load
points. The total combustion-related noise may further
be split up into direct combustion noise (due to cylinder
pressure alone) and indirect combustion noise (due to
piston slap and other mechanical effects that are exited
by cylinder pressure).

STRUCTURE ATTENUATION
Structure attenuation (SA) is a useful measure for
quantifying acoustic properties of the engine block and
head regarding airborne radiation of impulsiveness. SA
is defined by the relation of radiated combustion noise
sound pressure level to cylinder pressure level:

SA[dB ] =

Cylinder pressure level

Combustion noise level

Several methods exist for the determination of SA or,

alternatively, the combustion noise transfer function [9,
10].
With combustion excitation being a broad band source,
weaknesses in structure attenuation almost always lead
to specific patterns in radiated noise. If such frequencies
coincide with weaknesses in the vehicle sound package
(airborne transfer functions) this will likely result in
noticeable combustion noise patterns in the interior.
Thus with both transfer functions and structure
attenuation data available, countermeasures can be
designed either in the car or in the engine. Although SA
is usually applied to radiated noise, a similar
investigation can be carried out for the structure borne
combustion noise.
INTERIOR NOISE SYNTHESIS OF ENGINE
MODIFICATIONS
Synthesis of interior noise (BTPS) is possible based also
on test bench signals if respective transfer path input
points are used in the vehicle and on the test bench.
Differences in fixture impedance and room acoustics
compared to the engine compartment have to be
compensated for given operating conditions. Thus it is
much easier to assess the customer-relevant effects of
engine modifications, namely in the interior noise, while
comparing test bench signals alone may lead to wrong
conclusions.

CONCLUSION
Impulsive noise caused by Diesel combustion is strongly
influencing the perceived vehicle interior noise sound
quality. Impulsiveness requires time-based analysis
methods some of which have been demonstrated in this
paper. A sound quality metric for Diesel knocking has
been developed that correlates well with subjective
evaluation including JD Power APEAL score.
Time-based Binaural Transfer Path Analysis and
Synthesis allow tracking of impulsiveness along the
vehicle noise paths and identifying acoustic weaknesses
in the power train installation. If the results are combined
with acoustic engine test bench data, sources for
different transfer paths can be assessed and noise
patterns can be related to properties of engine
components or the combustion process itself. Effects of
source modifications are evaluated considering the

customer i.e. in the interior noise. If all these factors and

presented methods are considered in combination,
competitive solutions for sound quality improvements
can be found in a shorter timeframe and require less
hardware changes.

REFERENCES
1. R. Heinrichs, U. Groemping: Customer Driven diesel
Vehicle Sound Quality. Inter-Noise 2004.
2. T. Leonhard, R. Heinrichs, K.-H. Brger: Diesel
Powertrain Sound Quality. Aachener Kolloquium
2004.
3. K. Genuit: A New Approach to Objective
Determination of Noise Quality Based on Relative
Parameters. Inter-Noise 1996.
4. K. Genuit, J. Poggenburg: The Design of Vehicle
Interior Noise Using Binaural Transfer Path Analysis.
SAE NVC19, 1999
5. R. Sottek, D. Riemann, P. Sellerbeck: Virtual
Binaural Auralization of Vehicle Interior Sounds.
DAGA 2004
6. G. Shu, H. Wei, X. Liang: The Relationship between
the Combustion Noise and the High-Frequency
Oscillation Pressure of Internal Combustion Engine.
Inter-Noise 2004
7. J, Antoni, R. Boustany, F. Gautier, S. Wang: Source
Separation in Diesel Engines with the Cyclic Wiener
Filter. EuroNoise 2006
8. C. Renard, L. Polac: Combustion Noise and Piston
Slap Noise: Identification of Two Sources
Responsible for Diesel Engines Sound Signature.
Daga 2004
9. K. Nakashima, Y. Yajima, M. Yamamoto:
Measurement of Structural Attenuation of a Diesel
Engine and its Applications for Reduction of Noise
and Vibration. SAE 912710, 1991.
10. S. Ge-qun, W. Hai-qiao, H. Rui: The Transfer
Function of Combustion Noise in DI-Diesel Engine.
SAE 2005-01-2486, 2005.