Experimental analysis

on gear noise and Transmission Error

D. Rémond, C. Bard and D. Play

Laboratoire Conception et Analyse des Systèmes Mécaniques

I.N.S.A. de Lyon, Bât. 113
20, Avenue Albert EINSTEIN
69 621 VILLEURBANNE Cédex, FRANCE

Abstract
In geared mechanisms, gear noise becomes more and more outlined
because of reduction of acoustical radiations of coupled mechanisms. The
complexity of acoustical models, the mechanical imbricated degree of
geared mechanisms and the coupling of governing parameters lead us to
have an experimental approach for gear noise analysis. The main excitation
for gearbox is recognised to be the Transmission Error, which seems to be
the right node between gear geometry and noise characteristics. The aim of
the first part of this paper is to define experimental influences of geometrical
and operating condition parameters on gear noise through a Yates
experimental design.
A second part will present an original approach of Transmission Error
measurements with optical encoders and some results on dynamic and
kinematic behaviours of a simple geared system. Then, the results on gear
noise will be correlated with the Transmission Error through classical
spectral analyses. These correlations show that consequential effects on
noise due to geometrical modifications or deformations under load are
taken into account by Transmission Error values. Other parameters like
misalignments are not fully described by the Transmission Error

Experimental analysis on gear noise and Transmission Error 09/05/2025 1

measurement device. All the tests have been performed on an open-loop
test rig, with different spur and helical gears, in real operating conditions
(speed varying from 1000 to 6000 rpm and applied torque from 0 to 120
Nm).

1. Introduction.
Since few years, geared power transmissions are concerned with a
new technologic pressure for the reduction of noise level. This kind of
mechanism is now considered as a major source of excitation and radiation
in modern mechanical architectures. Traditionnal operating conditions of
geared power transmissions lead to numerous vibrations and acoustical
excitations, spread over the audible spectrum. Moreover, the complexity of
gear noise modelization is based on the lack of efficient acoustical models,
the high coupling between the different physical phenomena of acoustical
radiations, the geometrical and mechanical complexity of geared
mechanisms and the compelling need to take into account couplings with
structure (casing, bearings, shafts,...). Such reasons induce to have an
experimental approach of gear noise generation in order to draw up models
which link geometrical and operating conditions to acoustic radiations.
Gear Transmission Error is a specific gear quantity unanimously
considered as a powerful indicator of gear vibrations and noise. It is greatly
used in gear geometry testings and gear behaviour qualifications because
gear geometrical errors are gathered together into one measurable quantity.
But these standardtests are performed under light loads, far from the
working loads considered in the operating conditions. On the other hand,
Transmission Error is now recognised to be the main dynamic excitation
induced by geared systems. It is commonly assumed that gear noise is
deeply linked with Transmission Error, either in magnitude or in frequency

Experimental analysis on gear noise and Transmission Error 09/05/2025 2

localisation. Likewise, Transmission Error represents the proper and the
deeper characteristic which can be proposed by mechanical models of
geared mechanisms. Thus, Transmission Error can be seen as a good
geometry indicator of the mechanical behaviour but also as a good
measurable quantity representative of the gear excitation. From a general
point of view, better is the gear geometry, lower is Transmission Error,
better is gear acoustical behaviour (22), (27).
Technics used for gear Transmission Error measurements during
motion are separated into two main categories when geared systems work
under real conditions of applied torque and speed (7), (14), (21). The
torsionnal accelerometer technic uses two circumferencially placed
accelerometers mounted on a rigid disk clamped to the pinion. Transmission
Error is then obtained by a substraction of tangential accelerations and a
double integration. The main limitations of such a device are the need of
telemetry, the spoiling of Transmission Error in the case of large vibrations,
and the measurement only on one shaft. On the other side, the optical
encoder technic uses two encoders mounted at the end or on each geared
shaft. Transmission Error is then obtained directly by comparing the signal
given by each encoder. The main limitation of this second technic is the
angular speed effect which needs an increase of frequencies of encoder
signals. Further, natural frequencies of encoders and their coupling shaft
effects must be taken into account. However, an original encoder technic (3)
allows to separate the measurement resolution (number of encoder lines)
and the measurement precision (assumed by a high frequency digitalizing
clock). Consequently, many improvements in gearing researches have been
performed with the use of the Transmission Error as an indicator, in order to
optimize tooth profile modifications ((14), (22), (25)), and to decrease
experimental amplitude of gear vibrations ((24) and (3)). However, it must be

Experimental analysis on gear noise and Transmission Error 09/05/2025 3

pointed out that the correlation between gear noise and Transmission Error
is only based upon global measurements of these two quantities, with
specific gear geometry and under specific operating conditions. So, the
main purpose of this paper is to analyse the correlation between gear
Transmission Error and the radiated gear noise through an experimental
approach. Then, it is also reachable to understand how gear noise is
generated and/or how dynamic Transmission Error governs acoustic
radiations in real operating conditions. In order to perform this analysis,
Influences of gear parameters on noise will be specified and compared to
the influences of the same parameters on the transfer function obtained
between Transmission Error and gear noise data. This paper is organised in
three main steps : a gear noise analysis of gears operating under real
conditions, a setup and validation of a new gear Transmission Error
measurement device with optical encoders, and finally an analysis of
correlations. This paper will present, in a synthetic way, results of these
experiments and conclusions about Transmission Error to noise correlation.

2. Experimental setup and data analysis.

2.1. Test bench configuration and measurements.
A detailed description of the test bench is done elsewhere (16). It is
constituted (Fig. 1) of two geared shafts (1,2). Each shaft has a 60 mm
diameter and is mounted on two rolling bearings. One shaft mounting (1) is
fixed on an intermediate plate which allows to impose small misalignments
between gear axes. Vertical misalignment or deviation (VM) will be referred
to a rotation of the shaft (1) in the vertical plane (side view on fig. 1), ,the
other horizontal misalignment, (HM) or inclination, takes place in the
horizontal plane (front view on fig. 1). Notice that the gear center distance
can be modified.

Experimental analysis on gear noise and Transmission Error 09/05/2025 4

The input rotating speed is set between 100 and 6000 rpm and is feedback
controlled. The applied torque varies independently of the gear velocity,
from 0 to 150 Nm. The active part of the rig is mounted on a rigid frame and
isolated from external noise by a drum (about 40 dB attenuation). A specific
critical speed of the test bench, known to be about 4800 rpm, has been
carefully avoided all along the tests.

Side View

microphone accelerometer 1
isolation drum

accelerometer 2
shaft 1
shaft 2

rigid frame
600 80 80 600 mm

intermediate plate

Front View

Experimental analysis on gear noise and Transmission Error 09/05/2025 5

 gearbox casing shaft 2
gear 2

encoder 2
78 mm

encoder 1
gear 1
shaft 1
Fig.1 : Schematic arrangement of the experimental apparatus.

Experimental analysis on gear noise and Transmission Error 09/05/2025 6

 Parts C1 and C2 (Fig. 1) are 2048 lines optical encoders, they allow to
take at least 40 informations during a gear profile meshing period (the
largest number of teeth of the tested gears is 52). Encoder signals are
processed through a specific acquisition board and timer frequency has
been fixed at 40 MHz, in accordance with the magnitude of angular
displacements. Angular position of each encoder (and consequently position
of the linked gear) is obtained by an accurate location (with timer) of each
encoder rising edge (see Fig. 2). The frequency of encoder signals runs from
3 KHz to 200 Khz when the gear angular speed goes from 100 rpm to 6000
rpm. Specific torsion modes are obviously introduced by the links placed
between gears and encoders and are taken into account during testings.

wheel encoder signal

pinion encoder signal

timer signal
time

3 
wheel position
2  versus time

time

3 

2 
pinion position
 versus time
time

Fig. 2 : Building of angular position laws for pinion and wheel.

Experimental analysis on gear noise and Transmission Error 09/05/2025 7

 Simultaneous analysis of the two encoder signals allows the
computation of the gear Transmission Error; notice that the gear ratio is
taken into account numerically during the calculation. The reconstitution of
Transmission Error from these signals is obtained by re-sampling each
angular position with linear interpolation, at a constant rate, and in a
synchronous way with noise signal.
A classical acoustic pressure microphone (type BK 4165) is situated 750
mm far from the tooth contact along the pressure line. Bearing accelerations
were also measured with piezoelectric accelerometers (model 2224C)
essentially to follow the global behaviour of the geared shafts. Gear noise,
bearing accelerations in two directions were monitored simultaneously on a
data acquisition board. The Transmission Error acquisition board was driven
by the master DAQ board in order to have simultaneous acquisition on each
signal.

2.2. Signal processing and data analysis of the gear Transmission Error.
For Transmission Error signals (TE), the low frequency component due
to the gear eccentricity effect spoils other frequencies, particularly those
from meshing process (3). Substraction of this low frequency allows to
obtain the typical TE signal presented in the figure 3. A specific indicator

called WT has been choosen in order to validate the TE measurement. WT

is the mean value of the ratios Wij Tij , where Wij is a single peak-valley

magnitude and Tij the associated time delay. This last indicator has been

chosen because it gives a precise indication of level increase in TE signal

both for low and high frequencies (3).

Experimental analysis on gear noise and Transmission Error 09/05/2025 8

 Transmission error (Rd)
T.E

W ij

Tij

Time (s)

Fig. 3 : Typical example of Transmission Error measurement without

excentricity component.

In a first set of experiments, profile modifications, helix angle and

misalignment effects have been considered in this part on TE using WT

indicator. Profile modifications and helix angle are known to be fundamental
parameters in TE reduction. This is clearly established by results presented
in the figure 4 where the dynamic behaviour improvement induce by helical
gear is shown. A mean reduction of 16.0 % (profile modifications) and 78.0
% (helix angle) is obtained outside peak zones. The corresponding peak
reduction is respectively of 60.0 % and 95.0 %.
Misalignment between gear axes defines gear positioning errors
resulting from manufacturing errors and shafts compliance. Lowest critical
speeds (see on the left of vertical broken line in the figures) are attributed to
the dynamic behaviour of the encoder mounting (4). They are not modified
by misalignment, only an increase of peak magnitude is observed (Fig. 6).
The main critical speed (3250 rpm), associated to the overall mechanical
Experimental analysis on gear noise and Transmission Error 09/05/2025 9
transmission behaviour, completely disappears when a vertical
misalignment (VM) of 4' is applied to the spur gear pair (arrow 1 on Fig. 5).
Although this effect cannot be explained, it has to be related to the results
on noise analysis in the next part of the paper.

Fig. 4 : Evolution of Transmission Error level

with speed for three gears pairs (50 rpm steps).

Experimental analysis on gear noise and Transmission Error 09/05/2025 10

 WT (Rd/s)

VM = 0’

VM = 4’

Speed (rpm)

Fig. 5 : Evolution of Transmission Error level versus speed

with vertical misalignment (spur gear =2.0, =0.0).

Effects of horizontal misalignment or inclination (HM) are as influent as

vertical misalignment on the mechanical resonances of the TE signal. The
main critical speed at 3250 rpm is modified by horizontal misalignment and
the critical speed due to encoder mounting is just increased in magnitude.
Two significant increases of TE level (arrows 2 & 3 on figure 7) can be
seen when a vertical misalignment (or deviation) of 4' is applied to the
helical gear pair. Notice the change in vertical scale between the two figures
6 and 7 when helical gears introduce overlap.
These results clearly show a large sensitivity of TE under different
mechanical tested conditions. Then, further correlation analyses must
integrate such variations of the Transmission Error signal.

Experimental analysis on gear noise and Transmission Error 09/05/2025 11

 WT (Rd/s)

HM = 0’

HM = 4’

Speed (rpm)

Fig. 6 : Evolution of Transmission Error level versus speed

with vertical misalignment (spur gear =2.0, =0.0).

WT (Rd/s)

VM = 0’

VM = 4’

3
2

Speed (rpm)

Fig. 7 : Evolution of Transmission Error level versus speed

with vertical misalignment (helical gear =2.0, =2.0).

Experimental analysis on gear noise and Transmission Error 09/05/2025 12

N o i s e l e v e l (d B )

H 1

H 3 H 4
H 2

H 5

H i

F re q u e n c y (H z )

Fig. 8 : Visualization of spectral analysis for gear noise signal.

Data processing for both gear noise and TE signals in correlation

analyses, consists in performing a classical spectral analysis with 16 time
samples of 1024 points. For example in Fig. 8, the magnitudes of the five
5
first harmonics of mesh frequency (called H1, H2, H )
i=1
i and the global

spectrum magnitude (called hi) are taken into account.

2.3. Tested gear characteristics.

All the AGMA 12 tested gears have the same global characteristics : a
78 mm center distance, a 17.5 mm width. But a set of parameters is

considered for each test : contact ratio , overlap ratio , pitch module and

profile modifications with a symetric crowning of 8 m. The following table

(Tab. 1) gathers the tested gears with their main differences.

Experimental analysis on gear noise and Transmission Error 09/05/2025 13

 Overlap ratio     

Z1 / Z2 (51/52) (36/38) (36/38) (48/49)

Working pitch module 1.444 mm 2.031 mm 2.031 mm 1.380 mm

 =  NPM 1 3 2 -

 =  WPM 1 1 - -

 =  NPM 1 (*) - 1 1 (*)

 =  WPM 1 (*) - 1 -

 =  NPM 1 - - -

 =  WPM 1 - - -

NPM : No Profile Modification WPM : With Profile Modifications

(*) = gear pairs used for correlations
Table 1 : Description of the set of tested gears.

3. Results on gear noise parametric analysis.

On each following graph (Fig. 9, etc ...), the curves give the general
tendancy of gear noise level fluctuation versus parameter variation. The
upper part is concerned with the global variations versus the studied
parameter and the lower part of each figure gives the relative variations for
the two tested running conditions. As the results are obtained from an
experimental design, the associated vertical ‘error bars’ give the range of
dispersion due to modification of the other parameters. Further, the
reduction of noise level between the two considered conditions (with or

Experimental analysis on gear noise and Transmission Error 09/05/2025 14

without profile modifications in figure 9 for example) is given in the bottom
part of each graph. Then, the four graphs of each figure represent the
influence of one parameter (i.e torque in Fig. 9) on either the mesh
frequency magnitude H1, the second harmonic magnitude H2, or the
summation of the five first harmonic magnitudes Hi, the global spectral
level hi. These four graphs describe the global characteristics of the gear
noise and their evolution versus a specific parameter. Notice that in the
litterature, similar results are presented in parts and concluding remarks can
be in opposition regarding either the harmonic magnitude or the global
spectral level. Consequently, a general tendancy can be extracted from all
the results given by the experimental study. The gear noise global level hi
shows variations versus speed, on the opposite side, applied torque induces
evolution on the harmonic levels. For example in the figure 9, applied torque
induces harmonic level increase, but it leaves global level approximately
constant. In the figure 10, harmonic levels are practically constant when
speed changes but global level increases with speed.

Experimental analysis on gear noise and Transmission Error 09/05/2025 15

 without profile modifications
with profile modifications

Fig. 9 : Effect of profile modifications on gear noise versus torque.

Experimental analysis on gear noise and Transmission Error 09/05/2025 16

 The improving effect of tooth profile modifications is shown versus
applied torque in the figure 9. Notice that the gear profile modifications have
been designed for the full load. Thus, considering the bottom part of each
graph, the gear noise reduction is clearly more effective, for the full load, on
the harmonic magnitudes and practically not on the total sound level. This
can be attributed to the elimination of the tooth impact at the contact entry
during profile meshing. With the original profile, the impact due to
interference gives a sound pressure pic at the beginning of the tooth
meshing while a modified profile smoothes the sound pressure distribution.
Then, considering a Fourier decomposition of signal, a more uniform shape
of the sound pressure distribution decreases the levels of harmonics.
Consequently tooth modifications, designed in order to give a theoritical
involute profile under loads, limit the effect of deformations due to load and
their influence on the emitted harmonic noise.
The figure 10 presents the same kind of results but versus gear angular
speed independently to torque variations. The harmonic levels are nearly
constant when speed increases but the total sound level raises up. The
improvement, due to gear profile modifications, is located at the high speed
and only for the harmonic levels. This point is not completly understood. But
it can be assumed that a speed increase leads to an increase of the dynamic
tooth load. Thus, the profile modifications limits the tooth interferences as in
the case of the torque effect.

Experimental analysis on gear noise and Transmission Error 09/05/2025 17

 w ithout profile modifications
w ith profile modifications

Fig. 10 : Effect of profile modifications on gear noise versus speed.

Experimental analysis on gear noise and Transmission Error 09/05/2025 18

 The figure 11 shows the difference on gear noise induced by a
decrease of the pitch module, all the other geometrical parameters staying
roughly the same (particularly gear width). On these curves, the variations of
the harmonic levels with applied torque appear different for the two
geometries but are nearly the same when no load is applied. The gear noise
decrease can reach more than 15 dB when load is applied.

module = 2.104 mm
module = 1.515 mm

Fig. 11 : Effect of module on gear noise versus torque.

Experimental analysis on gear noise and Transmission Error 09/05/2025 19

 For low pitch module, gear noise and its harmonic levels are less
dependant on applied load, but there is a greater data scattering, due to a
modification of the dynamic behaviour of the whole mechanical device. Here
again, the flexibility of teeth can improve the acoustic behaviour of a geared
transmission by reducing effects of load deformations. The evolutions and
differences between the two gear designs are almost constant with speed,
and the shape of the curves are the same as in the figure 10.

The operating center distance seems to be very important for gear

noise emission, is the operating center distance. Experiments consists have
been performed with a normal (design) and a reduced centre distance, with
modified and not modified spur gear pairs. The figure 12 presents the
difference between the two conditions which can reach among 10 dB versus
torque.
This decrease of gear noise is large and quasi constant along the load
range, for harmonic levels as for global gear noise level. When tested gears
have modified tooth profile, this parameter remains influent, but the
difference at high torque is less remarquable.
The centre distance parameter contributes to modifiy the angular
transmission continuity and moreover general conditions of meshing with
other phenomena like lubricant effects and damping. This influence of
centre distance can be linked with wellknown modulation effects seen with
high eccentricity on gears or shaft.

Experimental analysis on gear noise and Transmission Error 09/05/2025 20

 center distance = 78.0225 mm
center distance = 78.0135 mm

Fig. 12 : Effect of center distance on gear noise versus torque.

Now in order to underline the complexity of the gear noise

characterisation and modelization, the effects of the operating conditions
like misalignments must be considered. The next figure shows the effect of
vertical misalignment or deviation (VM) on gear noise versus contact ratio
for non modified tooth spur gears.

Experimental analysis on gear noise and Transmission Error 09/05/2025 21

 vertical misalignment = 4’
vertical misalignment = 0’

Fig. 13 : Effect of vertical misalignment on gear noise versus contact ratio

for spur gears without modifications.

The flexibility and the continuity of meshing are here again the two
main characteristics which lead to such a reversal of the acoustic behaviour
of spur gears without tooth modification. For low contact ratio, tooth
geometry is normal and the vertical misalignment leads to a bad effect, so
aligned gears are less noisy than the misaligned ones. On the opposite side
for high contact ratio gears, teeth are tall and thin, and thus vertical
misalignement can improve the gear behaviour like the introduction of a
Experimental analysis on gear noise and Transmission Error 09/05/2025 22
small helix angle does. This point can be seen for tooth modified gears, but
the reverse effect seems to be located just before a contact ratio of 3.
Other influences of parameters like overlap ratio, horizontal
misalignment or inclination (HM) have been tested and the conclusions of
this set of experiments are in good agreement with the bibliography (16).

4. Discussion : correlations between TE and gear noise.

In a third step of the experiments, three gear pairs, marked (*) in the
table 1, have been used with simultaneous noise and TE measurements. The
classical correlation tools applied on these measured signals are defined
from the auto-correlation and cross-correlation functions. Transfer and
coherence functions have been calculated upon 16 sets of 1024 samples
each, weighted with a Hamming window and monitored in the frequency
domain. For this new analysis, the TE signal is processed without filtering
low frequency in order to keep all frequency components. One example of
each function is presented in the figure 14, and the shapes of the coherence
function remain practically the same when running conditions or gear
geometries are modified. The mean value of the coherence functions always
stays up to 0.6 and the harmonic values are greater than 0.7. In other words,
70% of gear noise signal are directly related to the TE signal, and the
remaining 30% can be attributed both to the data scattering dipersions and
non linear effects.
From the transfer function, same indicators as for the previous gear
5
noise analysis (called H1, H2, H
i=1
i and hi) have been processed in order to

quantify the modifications induced by the different geometries and the

running conditions. These indicators (with 70% of coherence) are well
Experimental analysis on gear noise and Transmission Error 09/05/2025 23
located in the frequency domain with the tooth number and the angular
speed of each gear. These indicators are then processed in an experimental
design using Yates method which leads to the relative influence of each
change and their coupling effects. Each influence can be separated into a
linear effect (L) and a quadratic effect (Q) related to the studied parameter.
For coupling influences of the parameter, the two effects are gathered in a
global effect (T), that is to say LL+LQ+QL+QQ. Torque, gear pair and
misalignment conditions have been considered.

H1 H2 H3 H4 H5

Experimental analysis on gear noise and Transmission Error 09/05/2025 24

 Fig. 14 : Examples of transfer and coherence functions
for spur gear (51/52, NPM, 1000rpm, 0daNm).

First of all, the transfer function between the Transmission Error and
the gear noise present a non constant shape function when running
conditions change (speed, torque, misalignment). For different gear
geometries, the figure 15 shows two transfer functions with two different
applied load. Modifications occur on the curve shapes and are related to
amplitude differences. In order to keep the results about noise variation with
parameters from the noise analysis, the amplitude of the defined indicators
has been chosen.

C=0daNm

C=12daNm

Experimental analysis on gear noise and Transmission Error 09/05/2025 25

 Fig. 15 : Differences in transfer functions for spur gear
induced by torque increase (51/52, NPM, 1000rpm, HM=0, VM=0)..

Consequently, speed has a great influence on the variations of

harmonic magnitude (see table 2). This effect is due to the TE measurements
which induce a variation of sensibility with speed rotation of shaft.
Measurement accuracy moves from 1/1172 at 1000 rpm to 1/195 at 6000
rpm, as the data resolution stays constant at about 40 points per tooth on
each gear. So, the first results on transfer functions have been analysed at
low speed (1000 rpm) in order to avoid speed influence.

H1 H2 H3 H4 H5 RMS

speed (L+Q) 54.5 29.1 30.3 13.8 48.6 27.3

torque (L+Q) 0.5 1.2 1.8 1.7 1.3 1.2
misalignment (L+Q) 0.9 0.2 2.2 5.6 4.4 3
gear (L+Q) 2.2 18.2 28 29.7 7.4 21.6
speed-misalignment (T) 9.6 1.5 2.1 11 5 8.9
gear-misalignment (T) 11.5 13.6 7.1 13.1 8.4 10.9
speed-gear (T) 3.8 18.5 3.8 2.3 0.6 2.8
residual 17 17.7 24.7 22.8 24.3 24.3

Table 2 : Parameter influences on transfer function harmonic (in

percentage).

From all the transfer functions, harmonic indicators have been taken as
inputs for experimental design analysis. The analysis mixed up the torque
(0-60-120 Nm) and the misalignment conditions (HM=0 & VM=0 - HM=4' &
VM=0 - HM=0 & VM=4') for each gear pairs, and the more important
parameter influences are reported in the table 3.
Experimental analysis on gear noise and Transmission Error 09/05/2025 26
Experimental analysis on gear noise and Transmission Error 09/05/2025 27
 Speed 1000 rpm H1 H2 H3 H4 H5 RMS

torque (L+Q) 1.7 4.9 9.1 3.1 10 9.8

gear (L+Q) 6.8 0.8 18.7 20.4 15.3 15.1
misalignment (L+Q) 44.2 16.2 15.6 50 20.6 34.0
gear-misalignment (T) 31.8 54.6 39.0 15.3 34.1 23.9
residual 15.5 23.5 17.6 11.2 20 17.2

Table 3 : Parameter influences on transfer function harmonic

at speed 1000 rpm (in percentage).

If one parameter is influent (>10 %), its change induces modifications

on transfer function and thus, TE is not sufficient to give a right description
of gear noise radiation and vice-versa.
From this set of experiments, the torque has no influence on the first
harmonics of the transfer function. This fact indicates that torque
modifications have a direct influence on noise which is taken into account by
the Transmission Error. Modifications on TE signal induced by torque are
directly linked with gear noise modifications. Then, gear geometry is not the
most influent factor. This result points out that gear geometry modifications,
as deformations due to torque, have a direct influence on noise which is
taken into account by the Transmission Error.
On the other hand, misalignment conditions are, in comparison, more
influential on the magnitude of the first harmonics, and in addition with a
coupling effect in relation with gear geometries. That is to say their effects
on noise are not fully included into the TE parameter. They induced other
phenomena like sliding or friction that produce geometric modifications of
mechanical transmission. The coupling effect between gear geometry and
misalignments leads us to say that design modifications can give a
Experimental analysis on gear noise and Transmission Error 09/05/2025 28
beneficial effect on this dependence of noise to misalignment, but the TE
parameter is not able to fully translate this result. Only a part of these
modifications on noise are correlated with the Transmission Error.

In the table 4, all the experiments are done at a 2250 rpm speed level.
At this specific speed, the torque remains a non governing parameter for the
modification of the transfer function. So, this conclusion which relates the
TE as a good indicator of gear noise remains still true. But, the main
difference which appears at this speed level is the great influence of the gear
geometry. This fact is related to the modifications of the gear dynamic
behaviour seen in the part 2.2 of this paper. At this speed, one can see that
this modifications lead to a small difference on the curves given in figures 4,
5, 6 and 7. This point could be related to the fact that misalignments have no
influence on the three first harmonics of the transfer function, but its
coupling effect remains influential. Here again, this result can be compared
with the differences of the curves represented on the figures 4 to 7.

Speed 2250 rpm H1 H2 H3 H4 H5 RMS

torque (L+Q) 0.4 1.9 3.9 4.7 9.2 1.6

gear (L+Q) 38.7 40.5 37.6 28.6 17.8 31.8
misalignment (L+Q) 0.4 7.4 5 19.6 13.7 20.8
gear-misalignment (T) 47.9 30.6 34.2 19.8 39.8 34.3
residual 12.6 19.6 19.3 27.3 19.5 11.5

Table 4 : Parameter influences on transfer function harmonic

at speed 2250 rpm (in percentage).

Experimental analysis on gear noise and Transmission Error 09/05/2025 29

6. Conclusions.

A lot of works have been done on the behaviour of gears and gear
noise. However, all the results and single conclusions cannot give a strong
way to design new quiet gear transmission. Two kinds of difficulty rapidly
appear as the gear noise is caracterised from different indicators and
relations between gear noise and gear parameters causes are often limited
to one or two parameter interactions. Further, the need of a link indicator
between the gear geometrical parameters and the gear noise behaviour lead
us to introduce the gear Transmission Error under real operating conditions.
In order to understand gear noise generation and to validate the
assumption that gear noise is directly related to the Transmission Error,
simultaneous measurements of these two signals have been performed for
different gear geometries and misalignment conditions. Then, classical
spectral correlation techniques have been applied and have lead to the
determination of transfer or coherence functions.
From these functions, quantitative indicators, significant for noise
radiations, have been defined and several experimental design analysis
were performed. The parameters of these analysis have been chosen as
speed, torque, gear geometry and misalignment.
From the parameter study :
- torque influences the level of gear noise harmonics while the
global level (RMS) of gear noise stays practically constant,
- in contrary, speed influences the global RMS gear noise but not
the harmonic levels,

Experimental analysis on gear noise and Transmission Error 09/05/2025 30

 - a strong interaction of parameters is seen on the gear noise
behaviour.
From the correlation study :
- the great influence of speed is due, for a large part, to the
Transmission Error measurement device, its dynamic range being
open ended with shaft speed,
- effects like torque deformations or gear design modifications are
taken into account by the Transmission Error measurement. For
these effects, transfer function can be considered with a constant
shape on harmonic levels, compared with other parameter
influences,
- some other effects like coupling between geometry and
misalignment have repercussions on noise not taken into account
by Transmission Error.

The expected continuations of this work go forward several

directions the improvements of Transmission Error measurement
techniques, the improvement of correlation technique and the definition of
some indicators more reliable and synthetic.
From a more general point of view, the results have clearly shown that the
improvements of dynamic models of geared systems must be made in
relation with the Transmission Error numerical simulation.

Acknowledgements.
The authors would like to thank P. Soleilhac and P. Bayle, technicians, for
their efficient help during the entire experimental program. A part of the

Experimental analysis on gear noise and Transmission Error 09/05/2025 31

tests has been performed under the Brite Euram contract RHINO. Financial
support have been also provided by EUROCOPTER Company.

References.
(1) Attia A.Y., "Noise of involute helical gears", ASME Journal of
Engineering for Industry, Feb, (1969), pp.165-171.
(2) Attia A.Y., "Noise of gears a comparative study", Proc. International
Power Transmission and Gearing Conference, Chicago, (1989), pp. 143-
148.
(3) Bard C., Rémond D. and Play D., "New Transmission Error measurement
for heavy load gears", Proc. of International Gearing Conference,
Newcastle, 7-9 September, (1994), pp. 393-400.
(4) Bard C., "Modélisation du comportement dynamique des transmissions
par engrenages", Thèse de doctorat, Institut National des Sciences
Appliquées de Lyon, (1995), No 95-ISAL-0031.
(5) Goupy J., "La Méthode des Plans d'Expériences", Dunod Ed., Paris,
(1988).
(6) Houser D.R., "Gear noise : state of the art", Proc. Inter Noise 88,
Avignon, France, (1988), p. 516.
(7) Houser D.R. and Blankenship G.W., "Methods for measuring
Transmission Error under load and at operating speeds. SAE Off-
Highway Conference, Milwaukee, September, (1989).
(8) Kong X., "Gear train monitoring by Transmission Error method". Ph. D.
Thesis : Mechanical Engineering, University of Wisconsin-Madison,
(1987).

Experimental analysis on gear noise and Transmission Error 09/05/2025 32

(9) Kubo A., Kiyono S., "Vibrational Excitation of Cylindrical Involute Gears
due to Tooth Form Error", Bulletin of JSME, Vol.23, No.183, (1980), pp.
1536-1543.
(10) Kubo A., Kiyono S. and Fujino M., " On Analysis and Prediction of
Machine Vibration Caused by Gear Meshing (1st report)", Bulletin of
JSME, Vol. 29, No. 258, (1986), pp. 4424-4429.
(11) Kubo A., Kuboki T. and Nonaka T., "Estimation of Transmission Error of
Cylindricla Involute Gears by Tooth Contact Pattern", JSME Int. Journal,
Series III, Vol. 34, No. 2, (1991), pp. 252-259.
(12) Mark W. D., "Analysis of the vibration of gears systems : basic theory",.
Journal of the acoustical society of America, No 63, (1978), pp 1409-
1430.
(13) Max J., "Méthodes et Techniques du Traitement du Signal", Masson Ed.,
Paris, (1989).
(14) Munro R.G., "A review of the theory and measurement of gear
Transmission Error", Proc. Institution of Mechanical Engineers, Int. Conf.
Gearbox Noise and Vibration, Cambridge, 9-11 April, (1990), pp.3-10
(15) Régie Nationale des Usines RENAULT, "Procédé et appareil de mesure
de l'erreur cinématique d'un couple d'engrenages". (1991) Brevet
européen N° 91401630.8.
(16) Rémond D., "Contribution à l'étude et l'analyse expérimentale du bruit
d'engrènement - Développement et application de la Transformée en
Ondelettes Rapide". Thèse de doctorat, Institut National des Sciences
Appliquées de Lyon, (1991), No 91-ISAL-0067.
(17) Rémond D., Bard C. and Play D., "A systematic approach of the effect of
design parameters on spectral and wavelet gear noise analysis",
MPT'91, JSME Int. Conf. Motions and Power Transmissions, Hiroshima,
Japan, (1991), pp.128-134.

Experimental analysis on gear noise and Transmission Error 09/05/2025 33

(18) Rémond D., Vélex P. and Sabot J., "Comportement dynamique et
acoustique des transmissions par engrenages", CETIM, Senlis, (1993),
p. 189.
(19) Rémond D., Bard C. and Play D., "Gear noise and Transmission Error
correlations in real operating conditions", Proc. Euro Noise 95, Lyon,
France, (1995) pp. 1049-1054.
(20) Smith J.D., "A modular system for Transmission Error measurement",
Proc. Institution of Mech. Engineers, Vol 202, N° C6, (1988), p. 439.
(21) Smith J.D. and Echeverria-Villagomez J.S., "Comparing encoder and
accelerometer measurement of Transmission Error or torsionnal
vibration", Proc. Institution of Mechanical Engineers, Int. Conf. Gearbox
Noise and Vibration, Cambridge, 9-11 April, (1990), pp. 43-50.
(22) Tavakoli M.S. and HOUSER D.R., "Optimum Profile Modifications for the
Minimization of Statis Transmission Errors of Spur Gears", Journal of
Mechanism, Transmissions and Automation in Design, December,
(1984), Vol. 108, pp.86-94.
(23) Thompson A.M., "Origin of Gear Noise", Proc. First World Congress on
Gearing, Paris, (1977), Vol. 2, pp. 1809-1817.
(24) Umezawa K., Suzuki T., HOUJOH H. and Bagiasna K., "Influence of
misalignment on vibration of helical gear", Proc. Second World
Congress on Gearing, Paris, (1986), pp. 615-626.
(25) Umezawa K., "The Performance diagrams for the Vibration of helical
gears", Proc. 1989 International Power transmission and Gearing
Conference, Chicago, (1989), pp.399-408.
(26) Vinayak H., Houser D. R., "A comparison of analytical predictions with
experimental measurements of Transmission Error of misaligned
gears", A.S.M.E Int. Power transmission conf., (1992), Vol. 1, pp 11-18

Experimental analysis on gear noise and Transmission Error 09/05/2025 34

(27) Welbourn D. B., "Gear errors and their resultant noise spectra". 1969-70,
Proc. of the Institution of Mech. Engineers, Vol. 184, (1970),
pp 131-139.
(28) Winter H., Gerber H., "Investigation on the excitation of the vibration
and noise at spur and helical gears". 5th A.S.M.E Int. Power
Transmission and Gearing Conf., Chicago, (1989), pp 467-470.

Experimental analysis on gear noise and Transmission Error 09/05/2025 35