NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 1

Rev. 03/25/2011 total pages 11

3. Subjective Perception and Objective NVH Measures

3.0 Introduction

The basic objective of engineering for NVH is to provide the customer with a subjectively
pleasant noise and vibration experience. This includes:

- A sufficiently quiet interior for conversation and listening to music
- A freedom from harsh jerks and jolts
- A freedom from aggravating noises such as squeaks and rattles
- A freedom from noises and vibrations that suggest potential problems or low quality

In order to design for these objectives it is necessary to assess them quantitatively with objective
measures, and this task will be the subject of the following section.

3.1 General Concept of Subjective Objective Correlation

It is usual to relate subjective and objective measures by comparing the average the customer
ratings of objectively measure sound and vibration. We expect a roughly linear relationship
between the ideal metric and the average customer rating as shown in figure 3.1.


Objective Level
S
u
b
j
e
c
t
i
v
e

R
a
t
i
n
g
Raw data
Fit

Figure 3.1: Concept of Subjective Objective Correlation

Here the subjective rating decreases as the objective measure, such as loudness, increases.

This method was used in the past to develop standard subjective measures such as loudness. Very
sophisticated techniques have been developed to make it possible to obtain correlation using a
relatively small numbers of evaluators, but these techniques are beyond the scope of this section.





NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 2

Rev. 03/25/2011 total pages 11
3.2 Physical Measures of Noise and Vibration

By physical measures we mean measures obtained directly using instrumentation, such as
microphones and accelerometers, without making any adjustment for human sensitivity to noise and
vibration.

3.2.1 Physical Noise (Sound) Measures

Sound Pressure (Pascals Pa)

The basic physical unit of sound pressure is the Pascal (Pa), 1 Pascal = 1 N/sq.m . Typical vehicle
interior sound pressures are in the range 0.1 to 1 Pa. This is small compared to the static atmospheric
pressure which is approximately of 5000 Pa.

Sound Pressure Level SPL Decibels-dB

The range of audible sound pressures is very wide, and for this reason it is common to use the Sound
Pressure Level (SPL), which is the logarithm of the ratio of RMS sound pressure to a reference
pressure:


(
(

=
ref
rms
P
P
SPL log 20 dB

where Pref=20 Micro-Pascals . This formula can be written in the more convenient form:

( ) | |dB P SPL
rms
log 20 94 + =

From this we can see that a 1 Pa sound pressure (rms) is equivalent to 94 dB.

Figure 3.2.1 shows a sine wave with peak amplitude 1 Pa, which corresponds to 0.707 Pa rms. So the
SPL is:

SPL = 94 + 20log(.707) = 91 dB














Figure 3.2.1: 91 dBL sine wave
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.01 0.02 0.03 0.04 0.05
S
o
u
n
d

P
r
e
s
s
u
r
e

P
a
Time (seconds)
91 dB sine wave
NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 3

Rev. 03/25/2011 total pages 11

Effect of Measurement on Measured Quantity

There are several ways that the measurement technique can affect the measurement:

(a) Mass of driver and passenger may alter vehicle body response to inputs
(b) Acoustic damping and reflection due driver and passenger may slightly change
above 1000 hz
(c) Microphone may modify sound pressure if its diameter is large enough
(above 5000 Hz for a microphone)
(d) Type of microphone used (free-field or diffuse) can affect results, mainly above 5000 Hz.

For these reasons care should be taken that the measurement technique is consistent from test to
test.

Artificial heads are sometimes used for data collection. These should have little effect below 1000
hz, but at higher frequencies they will change the acoustic field due to reflection and absorption. This
type of measurement is completely different from the type of measurement characterized as free-
field or diffuse. The user should therefore be careful to follow the manufacturers instructions.

Effect of Location

Care should be taken to ensure consistency in measurement location because the sound level will
vary somewhat from location to location. This effect is most marked in narrow band spectra, and
least marked in wide-band overall levels.

3.2.2 Physical Vibration Measures

Acceleration

Acceleration is the most common measure of vibration since measurements are usually made with
accelerometers. Vehicle interior levels are usually less than 0.5 m/s/s, but the expected level
depends on the operating condition.

Velocity

Velocity is usually found by time integration of the acceleration signal, or its equivalent. Above 15 hz
the human sensitivity to velocity is roughly constant, and it is usually the preferred measure. The
target interior level is usually less than 1 mm/s.

Displacement

Displacement is seldom used as a vibration measure except occasionally in the event that visible
vibration is an annoyance.

Location

It is difficult to directly measure the vibration as felt by the customer except for vibration at the
steering wheel. As a result surrogate measurements are often used, for example, seat track vibration
is frequently used as a surrogate for driver or passenger seat vibration. Instrumented seat cushions
are sometimes used for a direct measurement but the results depend on the weight of the person or
dummy placed on the cushion. One possibility is to measure the transmissibility of the seat and
combine this with seat track measurements.

NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 4

Rev. 03/25/2011 total pages 11
In the case of the steering wheel, the vibration level may vary considerably between the top and
bottom of the wheel. Given that drivers vary where they place their hands, the main issue is
consistency i.e. always measure at the same location and in the same direction.

Direction

Vertical and lateral motion is usually the of most concern to the customer, but in some cases, such as
low frequency brake shudder or harshness due to impact strips, the fore-aft acceleration may be
important.

Affect of Measurement

As mentioned in the section on sound measurement, some caution needs to be observed relative to
the effect of mass loading. In particularly, steering column vibration may be noticeably affected by
the mass loading of the drivers hands and arms.


NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 5

Rev. 03/25/2011 total pages 11
3.3 Psycho-Acoustics

3.3.1 General Comments

Objective

The objective of psycho-acoustics is to convert physical sound pressure measurements into objective
measures of human response, in particular:

- Loudness
- Annoyance
- Speech Interference
- Sound quality (sqeak, rattle,........)

Variation & Correlation

Human beings differ in their response to noise and vibration and so measures of human perception
must be less precise than physical measures. A number of standard measures have been developed,
for example loudness in sones, based on the correlation of subjective response to objective measure
for a large number of test subjects. The value of these measures for automotive purposes depends on
how well they correlate to customer evaluations of vehicles for specific operating conditions.

3.3.2 Psycho Acoustic Measures

Loudness Level of a Pure Tone Phons

Researchers have developed loudness measures for pure tones (sine waves) relative to pure tones at
1000 Hz. The unit is the phon, defined by:

Loudness Level of 1000 hz tone (phons) = SPL of tone (dB)

The loudness level of tones at other frequencies as then defined as:

Loudness of a tone in phons = the SPL (dB) of an equally loud tone at 1000 hz

Robinson and Dadson developed standard loudness curves:


2
) ( ) ( dB c dB b a L + + =

where the coefficients a, b, and c depend on frequency and are given in a table. For example:

Example of Computing Loudness Leve1:

Pure tone parameters SPL=70 dB Freq= 50 hz

Coefficients (from table): a=-73 b=2.021 c=-.00407

Computed Loudness L 48.5 phon

Equal Loudness Curves

The typical loudness level of tones in an operating automobile with good NVH are in the range 40 to
60 phon, and the equal loudness curves for these levels are shown graphically as in figure 2.
NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 6

Rev. 03/25/2011 total pages 11















Figure 2: Sound pressure required to produce Loudness Levels of 40 and 60 phon

Linear Loudness of a Pure Tone Sones

Phons are a logarithmic measure. Increasing the loudness of a tone by 10 phons makes it sound twice
as loud. Sones are a linear measure which show this effect:


10
40
2

=
P
S P=Loudness in phones

This definition ensures that the loudness in sones doubles for every 10 phon increase in loudness
level. The loudness level of 40 phon (a quiet sound) is arbitrarily set equal to 1 sone. The formula for
phons in terms of sones is:

) log( 2 . 33 40
) 2 log(
) log(
10 40 S
S
P + = + =

So 1 sone is equivalent to 40 phon, 2 sones is equivalent to 50 phon, and so on.

Threshold of Hearing

It is clear that the ears sensitivity is low at low frequencies. For example, at 20 hz it requires more
than 90 dB to produce the loudness level of a 40 dB tone at 1000 hz (40 phon).















Figure 3: Threshold of Hearing Minimum Audible Field (MAF)
30
40
50
60
70
80
90
100
110
2
0
3
1
.
5
5
0
8
0
1
2
5
2
0
0
3
1
5
5
0
0
8
0
0
1
2
5
0
2
0
0
0
3
1
5
0
5
0
0
0
8
0
0
0
1
2
5
0
0
Frequency hz
S
P
L

-

d
B
Loudness = 40 dB @1000 hz
Loudness = 60 dB at 1000 hz
-20
0
20
40
60
80
2
0
3
1
.
5
5
0
8
0
1
2
5
2
0
0
3
1
5
5
0
0
8
0
0
1
2
5
0
2
0
0
0
3
1
5
0
5
0
0
0
8
0
0
0
1
2
5
0
0
Frequency hz
S
o
u
n
d

P
r
e
s
s
u
r
e

L
e
v
e
l


-

d
B
Age 18 - 25 years
Age 60 years
NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 7

Rev. 03/25/2011 total pages 11
The threshold of hearing curves were established by having the test subjects judge the sound
pressure level at which a tone (sine-wave at specific frequency) was just audible. This curve is shown
in figure 3.

It can be seen that at 1000 hz the lowest sound that can be heard has an SPL of about 4 dB, while at
20 hz the lowest sound which can be heard has an SPL of greater than 70 dB.

It must be noted that their results were for a specific population, the standard group consisting of 18
to 25 year olds with no obvious hearing injuries and ears clear of wax. Since people tend to lose their
ability to hear high frequency sound as they age, Robinson and Dadson also established standard
corrections for age. Apart from age effects, one can expect that people living or working in noisy
environments tend to loose hearing capability due to noise damage.

Weighting Systems

Weighting systems were developed so that sound level meters could give an immediate estimate of
the subjective loudness without the need for calculation. At the time the weighting curves were
developed the implementation could only be by means of analogue filters, and it was necessary to
make a variety of simplifications. The main weighting systems are :

A weighting: Approximates the 40 phon equal loudness contour. Used for many
regulatory purposes e.g. Pass By noise regulations, Airport Noise
regulations. Usual in NVH research papers.

B weighting: Approximates the 70 phon equal loudness contour

C weighting: Approximates the 100 phon contour

D weighting: A weighting system related to equal noisiness contours.
Noisiness is a measure similar to loudness and popular for evaluating
airport noise noisiness is rarely used in automotive applications.

The weighting system acts roughly as an inverse equal loudness contour. If the ears sensitivity has
dropped by X dB at a certain frequency then the weighting lowers the spectrum level by X dB.


















Figure 4: Comparison of the A-weighting System to 40-phon equal loudness weighting

-60
-50
-40
-30
-20
-10
0
10
20
2
0
3
1
.
5
5
0
8
0
1
2
5
2
0
0
3
1
5
5
0
0
8
0
0
1
2
5
0
2
0
0
0
3
1
5
0
5
0
0
0
8
0
0
0
1
2
5
0
0
Frequency hz
W
e
i
g
h
t
i
n
g

F
a
c
t
o
r

d
B
"A-Weight"
"40-phon Weight"
NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 8

Rev. 03/25/2011 total pages 11
Figure 4 shows a comparison of the A-weighting system to a 40-phon equal loudness weighting. It
can be seen that A-weighting agrees with the 40 phon contour at 1000 hz and at 20 hz., but is up to
10 dB in error in some frequency ranges. This problem arises from the relatively simple analogue
filtering required for sound level meters when the standard was set. It is particularly problematic for
the range 80-500 Hz which is often important for NVH problems.

Effect of Similarity of Vehicle Spectra

One thing that helps is that for each operating condition the main difference between vehicles is the
overall level of noise and vibration. To a first approximation the shape of the frequency spectrum
tends to be the same. As a result it is generally believed that the customers response is mainly based
on overall level (about 80%) and only secondarily on sound and vibration quality (spectrum shape).
Consequently the metrics measuring customer perception of noise level may not to be very precise.
For example, while the loudness level measure dBA is frequently criticized as an inaccurate, it often
works quite well if only because the evaluated spectra are similar in shape and mainly differ in
overall level.

Beats

It is quite common in automotive applications that there are two important tones with nearly
identical frequencies. If the frequency separation of the tones is less than 10 hz the resultant sound
will be perceived as a tone at the average frequency modulated at the difference or beat frequency.
The amplitude of the modulation depends on the relative amplitude of the two tones, and is a
maximum when the tones have equal amplitude as shown in figure 5.




















Figure 5. Beats Due to Two Equally Loud Tones 5 hz apart.

In the event the tones have unequal amplitude the level of modulation is reduced as is shown in
figure 6 for the case where one tone has half the amplitude of the other tone:

The modulation effect is known as beating, and the resulting pulsation as beats. Beats affect
customer acceptance of NVH and are strongly to be avoided. The loudness of beats can be estimated
from the instantaneous sone value of the modulated tone.



Beats Due Equal Amplitude Sine Waves
97.5 hz and 102.5 hz
-3
-2
-1
0
1
2
3
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time - Seconds
A
m
p
l
i
t
u
d
e
NVH Basics (Intro-3) Subjective Perception and Objective Measures Page 9

Rev. 03/25/2011 total pages 11

















Figure 6: Beats Caused by Two Tones, One with the Amplitude of the Other


Rumbling

Rumbling is a similar effect to beating but results from 3 or more frequencies close together in
frequency.

Minimum Detectable Change in Sound Level

Changes in sound level of less than 1 phon are not detectable by the average listener. This implies
that a customer will not be able to distinguish between two vehicles whose loudness differs by less
than one phon. For mid-frequency sounds (200 to 2000 hz) with intermediate loudness (40 60
phon) one can roughly say that sound pressure changes of less than 1dB will be undetectable.

A three phon change in level will be clearly audible, and a 10 phon change will be very pronounced.

Loudness of Broadband Noise

The loudness of broadband noise is usually estimated in one of two different ways. The first is to use
a simple weighting scheme such as A-weighting. The second is to use one of several methods based
on 1/3
rd
octave band analysis.

Weighting Method in the Time Domain

This method simply applies the weighting to the time domain signal prior to any further analysis.

Weighting Method in the Frequency Domain

Weighting is applied to the frequency spectrum by correcting the dB level at each frequency.


Zwickers Sones (ISO 532 1975 (E) Method B)

Zwickers method splits the spectrum into barks that are roughly equivalent to 1/3
rd
octave bands. A
loudness index is computed for each band and the total is computed with a more sophisticated
masking algorithm than that used by Stevens. Nowadays Zwicker sones are usually computed via a
Modulation Due to 97.5 hz Tone and 102.5 hz tone at 1/2 Amplitude
-2
-1
0
1
2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time - Seconds
A
m
p
l
i
t
u
d
e
NVH Basics (Intro-3) Subjective Perception and Objective Measures Page
10

Rev. 03/25/2011 total pages 11
computer algorithm (8) that differs somewhat from the original ISO standard, in particular the
masking algorithm is simplified.

Speech Interference Articulation Index

For automotive applications various versions of the articulation index (AI) have become popular.
This measure was originally developed to estimate the intelligibility of telephone calls in a noisy
environment.

The basic concept is that each 1/3
rd
octave (from 200 hz to 5000 hz center frequency) is assigned a
maximum contribution to AI and a threshold noise level. If the actual noise in the 1/3
rd
octave band is
less than the threshold then the band contribution is the maximum. For each dB of noise above
threshold the contribution to AI is reduced by a 1/30
th
of the maximum. If the noise is 30 or more dB
above threshold the contribution is set to zero. The 1/3
rd
octave contributions are summed to give
the final result that may be expressed as a decimal fraction (<=1) or a percentage (<=100%).

Since this method is not a recognized standard the implementation may depend on the version of
software used for the calculation it is advisable to consult the user manual. Also note that one variant
is to make the contribution negative if the noise is more than 30 dB above threshold.

Time Varying Sounds

The measures described above were developed for stationary sounds, that is sounds which are
either constant in time or have time averages which are constant over reasonable time periods (a few
seconds). The application to rapidly varying sound levels, for example the sound as a vehicle
accelerates, raises a number of technical questions. For example, the sound pressure level will
change during the time required to determine the 1/3
rd
octave band level. Various schemes have
been developed to resolve these issues but they are too advanced for discussion in this section.

Powertrain Sound Quality Measures

While the primary effort in NVH is usually to reduce the level of noise and vibration, the quality of the
sound heard by the customer must also be considered. This is a large area and can only be briefly
summarized here. Since sound quality issues vary with operating condition we will summarise them
under appropriate headings. Powertrain issues are usually most noticeable at idle or when cruising
or accelerating on a smooth road.

Sound Quality and Engine Orders

Engine sounds can generally be analyzed in terms of sine waves at frequencies which are multiples of
the crankshaft rotational speed i.e. RPM/60. Thus a sine wave at 4 times the rotational frequency of
the crankshaft is called 4
th
order. The relative level of the various orders and the way in which their
level varies with rpm is important for sound quality. Here we give two examples.

Percusive Noises (Knocking, tapping, slapping)

These are most noticeable at idle or when cruising at low speed under low load conditions.
The customer can hear each percussive event separately and this gives the impression of low
quality. The frequency spectrum is wide and time domain analysis is useful.

Order Related Rumbling

Noise due to interaction of orders (sine waves) in the 100 600 hz range

NVH Basics (Intro-3) Subjective Perception and Objective Measures Page
11

Rev. 03/25/2011 total pages 11

It is generally thought that about 80% of customer noise concerns are directly related to loudness,
the other 20% arising from the quality of the sound. In some cases the quality of the sound may be
the predominant factor, for example the spectrum of door slam sound may indicate quality, or lack of
quality, to a customer. A large number of sound quality measures have been developed with varying
degrees of correlation to customer evaluation.