Content uploaded by Stefan Schoenwald
Author content
All content in this area was uploaded by Stefan Schoenwald on Oct 12, 2014
Content may be subject to copyright.
Analysis on
Structure-borne Sound Transmission at
Junctions of Solid Wood Double Walls with
Continuous Floors
Stefan Schoenwald
Empa Swiss Federal Laboratories for Materials Science and Technology, Dübendorf, Switzerland.
Berndt Zeitler, Ivan Sabourin
National Research Council Canada, Construction, Ottawa, Canada.
Summary
Structure-borne sound transmission across a cross-junction of double solid timber walls with a
solid timber floor was analyzed in a recent research project. Both, the double walls as well as
floor slab, were of so-called Cross Laminated Timber (CLT). The floor slab was continuous
across the junction for structural reasons and thus, formed a sound bridge between the elements of
the double wall. To gain a better understanding of the contributions of sound transmission
between the wall and floor elements from the different possible paths, a thorough analysis was
conducted. Hereby, direct sound transmission through, and radiation efficiencies of, the CLT
elements were measured in a direct sound transmission facility; as well as, structure-borne sound
transmission between CLT elements was measured on a junction mock-up. The experimental data
was used as in-put data and for validation of the engineering model of EN 12354/ISO 15712 for
the prediction of flanking sound insulation in buildings. The test procedures, analysis and results
of this research project are presented here. PACS no. 43.55.Rg, 43.40.At
1. Introduction
Cross Laminated Timber (CLT) elements are a building
system that is already widely used in Europe. The
elements consist of an odd number of layers of wooden
battens (lamellas) that are laid-out cross wise and glued
together to form solid wood panels for floors and walls.
CLTs recently emerged in the North American Market as
a new wood construction technology for multi-storey
wooden buildings. Properties and design details of
Canadian CLT structures were assessed in a
multidisciplinary project at the National Research Council
Canada that included besides sound insulation also
structural, fire safety and heat-and-moisture transfer
aspects.
The sound insulation system performance of CLT
buildings was studied first in test series on the direct
airborne and impact sound insulation of CLT walls and
floors with various added layers, for example with
gypsum board linings and floor toppings. In parallel,
structure-borne sound transmission between coupled
elements was investigated at separate junction mock-ups.
The methods of ISO 15712:2005 “Building Acoustics –
Estimation of acoustic performance of buildings from the
performance of elements”, or better known in Europe as
EN 12354, were then utilized with the measured input data
to predict sound transmission along relevant flanking
paths and finally to obtain the apparent sound insulation in
buildings. The greatest challenges when applying the
methods of EN 12354 in this project were that some of the
standards conditions, like moderate damping of the
building elements, were not fulfilled for the tested CLT
elements. Also, measured input data and final results were
rated in units according to ASTM standards instead of
ISO-standards as described in EN 12354. An earlier paper
already describes these problems and their solution using
examples of simple Cross- (X-) and T-junctions of CLT
walls and floors [1].
In this paper the applied methods, as well as the basic
assumptions are briefly reviewed. Then flanking sound
transmission across an X-junction, consisting of a single
leaf continuous CLT-floor with two double leaf CLT walls
attached is investigated. The relative importance of the
different flanking paths is discussed with special attention
on the reduction of the apparent sound insulation due to
the coupling of the two wall leafs by the continuous floor
slab.
2. Flanking sound transmission at cavity
wall-floor-junction
This paper focuses on flanking transmission across a wall-
floor-X-junction where the walls consists of double leaf
cavity walls made out of two 3-ply 78 mm thick CLT
FORUM ACUSTICUM 2014 Schoenwald, Stefan: Junctions of Solid Wood Double Walls
7–12 September, Krakow
elements that were separated by a 25 mm air gap. The gap
between the elements was filled with 25 mm glass fiber
batts and there were no direct structural connections
between the elements when the walls were tested for
direct sound reduction index in the NRC-Wall Sound
Transmission Facility. In buildings in Canada the floor
slabs between the storeys at the top and bottom of these
walls often are continuous across the air gap for structural
strength and for fire blocking, as depicted in Figure 1.
Figure 1. Flanking transmission between two vertically
adjacent rooms across a junction of a cavity wall with
continuous floor slab.
Figure 2. Flanking transmission between two horizontally
adjacent rooms across a junction of a cavity wall with
continuous floor slab.
Whereas Figure 1 shows that for flanking transmission
between vertically adjacent rooms, the number of flanking
paths for a junction is equal as for a junction with a single
leaf wall, Figure 2 shows that for horizontally adjacent
rooms an additional fourth flanking path “Dfd” is created
due to the structural coupling of the double leaf wall to the
continuous floor slab. In Figure 1 and 2 R is sound
reduction index, where the subscript letters “D” (direct)
and “F” (flanking) denote the elements that are involved
in the flanking path. Uppercase letters indicate the source
room and lowercase letters the receive room.
The unknown significance of this “Dfd”-flanking path in
the apparent sound insulation is investigated here utilizing
the methods of the ISO 15712 framework that are
described in the following.
3. Applied Methods
3.1. ISO 15712 prediction method
ISO 15712-1 is a prediction method to estimate the
apparent airborne sound transmission in a building, as is
perceived by the building occupants. The apparent sound
transmission is the energy sum of direct transmission
through the partition and 3 flanking paths at each junction.
ISO 15712-2 gives a corresponding model for apparent
impact sound insulation. Both models are closely related,
use essentially the same calculation procedure as well as
input data. This paper focuses on airborne sound
insulation between two adjacent rooms in a building and
the governing equation for the detailed prediction of the
flanking sound reduction index Rij involving element i and
element j is given in Equation 1 below.
ji
s
situijvsitujsitui
situjsitui
ij
SS
S
DRR
RR
Rlg10
2
,,,,
,,
++∆+∆+
+
=
(1)
As input data the direct resonant sound reduction index Ri
of the bare elements and the incremental change due to
additional layers
∆
R, for example due to floor toppings or
gypsum board wall linings have to be known that describe
coupling between the sound field in the source and receive
room and the building structure. The direction averaged
velocity level difference , describes the structural
power flow between the coupled flanking elements i and j
and the last term the geometry with the area, where
index s denotes the separating element. Index “situ”
indicates that the values are from data measured in the
laboratory already adjusted to simulate the field situation
that is to be predicted.
3.2. Data for element performance
Measured sound reduction index data has to be transferred
from the laboratory to the actual field situation. First, the
non-resonant component of airborne sound transmission,
that may become significant well below coincidence
frequency, should be removed. This component
constitutes a forced displacement of the element by an
incident sound wave that is not proportional to the element
damping. Thus, it does not induce any flexural energy in
the structure that is transmitted across the building
junction. The current ISO 15712 does not give a practical
correction method for non-resonant transmission, as in
most cases of concrete-masonry construction, for which
the method was originally intended, the coincidence
frequency is low and slightly conservative prediction
results are obtained by omitting this correction. However,
in the latest drafts for a revision of EN 12354 [2] a
correction term is suggested (Equation 2) that is based on
measured sound radiation efficiency σair for airborne and
σstruc for structure-born excitation.
struc
air
labnlabresn
RR
σ
σ
lg10
,,,
+=
(2)
As the resonant sound reduction index is proportional to
the damping of the elements, in cases where the total loss
factor of the building element installed in the lab differs
significantly from the field situation due to different edge
losses a further correction needs to be applied to transfer
the lab data to in-situ. For the considered CLT-elements
the internal loss factors are relatively high (0.04 to 0.06)
and govern the total loss factor. Therefore, the adjustment
for the edge losses is omitted and lab data for the element
performance is directly used for the prediction as it is
already discussed in [1].
FORUM ACUSTICUM 2014 Schoenwald, Stefan: Junctions of Solid Wood Double Walls
7–12 September, Krakow
The performance of the CLT elements was tested in NRC
Construction’s Wall and Floor Direct Sound Transmission
Facilities. In both facilities floor and wall test specimens
were installed in movable frames between two isolated
chambers of the transmission suite. The size of the wall
specimens was 3.60 m width by 2.40 m height and of the
floors 4.70 m by 3.78 m. The sizes of the chambers of the
wall facility is 250 m3 and 140 m3 and in the floor facility
approx. 175 m3 for both chambers. Both facilities are
equipped with computer controlled sound and data
acquisition systems that move the microphones to
9 measurement positions in each room.
Besides the standardized measurement of sound reduction
index according to the ASTM-protocols also the radiation
efficiencies were measured in both facilities. The sound
pressure levels and reverberation times in the rooms were
measured with the automated measurement system. The
velocity levels of the elements were measured in the floor
facility with accelerometers and in the wall facility with a
scanning laser Doppler vibrometer. For airborne
excitation the specimens were exposed to white noise
from the installed sound system. For structure-borne
excitation of the walls an electro-dynamic shaker driven
with white noise and for floors an ISO-tapping machine
was used.
3.3. Data for junction coupling
The flexural power transmission between two coupled
elements i and j at a building junction is given by the
direction averaged velocity level difference , that
depends on the junction details and is proportional to
geometry (junction length lij and area of elements S) and
their total loss factor. Therefore, geometry and loss factor
may have to be considered, when measured velocity level
difference data is transferred from the laboratory to the
field situation. However, as already discussed in [1], the
change in total loss factor is very small and negligible for
the junctions of CLT elements considered in this study.
Equation 3 according to the draft of EN 12354 [2] is
applied to normalize the measured , to the so-called
normalized direction-averaged junction velocity level
difference ,, and transfer back to a field situation for
prediction.
situjsitui
ij
nijvsituijv
SS
ll
DD
,,
0
,,,,
lg10−=
(3)
CLT elements are connected with nailed metal plates or
long wood screws at a spacing specified by structural
engineers for the specific axial and lateral loads. These
connections are very complex. Therefore, in this research
project ,,-data were measured on wall-wall and floor-
wall junctions that were assembled in a lab environment
as shown in Figure 3 for a floor-wall cross junction.
The 175 mm 5-ply CLT floor slab (total length 8.60 m) is
continuous across the walls that are located at app. 0.50 m
of centre and the outer lamellas are oriented perpendicular
to the walls. The upper and lower walls were both 3.60 m
wide by 2.40 m high.
Figure 3. Set-up for measuring ,, at CLT floor-wall
junctions.
The floor slab was supported by the lower wall in the
middle and at its free edges by shores. Soft rubber pads
were placed between the CLT floor and the shores as
vibration isolators. The lower wall was standing on a
massive rigid concrete floor only supported by its
connections to the CLT floor at its upper edges. The top
wall is connected at the bottom to the CLT floor and a
steel frame spans over its top to apply with hydraulic jacks
an axial load to the upper wall. This is done with a force
that simulates the weight of additional upper stories.
Loading was found to have an effect on flanking sound
transmission in wood frame construction, however, the
effect was limited as the initial loading of the first storey
causes the greatest difference and further loading does not
change the acoustical propagation properties. For CLT
construction the loading had negligible effect.
The size of the elements was according to ISO 10848-4
and the indirect measurement method was utilized. All
coupled elements were instrumented with
4 accelerometers and the velocity response was measured
on all channels simultaneously, while one of the elements
was excited during a 30s time interval with multiple
hammer blows applied in a predefined excitation area of
app. 1 m2. On each element four different excitation areas
and four sets of receiver points were used - in total the
velocity was averaged over 16 measurement points on
each element.
4. Test results
4.1. Sound insulation of CLT elements
The measured sound reduction index of the 78 mm 3-ply
single wall, 3-ply CLT cavity wall, and the 175 mm 5-ply
floor are presented in Figure 4. The results of the single
elements are further corrected below coincidence
frequency by removing the non-resonant component
before using as input data to predict flanking sound
insulation.
FORUM ACUSTICUM 2014 Schoenwald, Stefan: Junctions of Solid Wood Double Walls
7–12 September, Krakow
Figure 4. Direct sound reduction index of CLT wall and floor
elements.
The sound reduction index of the 5-ply CLT floor
(Rw = 41 dB) is higher than for the 3-ply CLT wall
(Rw = 33 dB) due to the greater mass. Thickness
resonances cause the distinct dip around 3150 Hz in the
floor result. Although the cavity wall has approximately
the same mass as the floor it performs much better
(Rw = 47 dB) because of the decoupling of the wall leaves
above its mass-spring-mass resonance at around 80 Hz.
The resonant sound reduction index of the 5-ply floor is
only marginally greater than the uncorrected original data
as the coincidence frequency is low around 200 Hz. For
the 3-ply CLT with its coincidence frequency around 800
Hz the effect of the non-resonant component is much
greater. This is also reflected by the single number
ratings.
4.2. Junction attenuation
At the cavity-wall-floor-junction 13 different , were
measured between the elements as indicated in Figure 5
and 6 and afterwards normalized to ,,. As connection
details at the upper and lower wall were the same – nailed
90 mm angle brackets spaced 300 mm on both wall sides –
the ,, paths denoted by the same color were averaged
to reduce to 5 unique data sets for the analysis.
Figure 5. Dv,Ff,n for vertical (red), horizontal (blue) and
diagonal (green) transmission at cavity-wall-floor- junction;
and denotation of the elements.
Note there are two sets of ,,, or ,, respectively, –
one for vertical and one for horizontal transmission – for
the junction with the double leaf wall as shown in
Figure 6, whereas there is only a single set of data for the
same junction with a single leaf wall. Also, in this double
leaf wall case ,, in Figure 5 is equal for vertical and
diagonal transmission.
Figure 6. Dv,Fd,n or Dv,Df,n: left: for vertical transmission right:
for horizontal transmission; and denotation of the elements
Figure 7 presents ,, for the 5 situations from 50 Hz to
3150 Hz as above this frequency range the signal to
background noise ratio was not sufficient on some of the
receiving elements. It is obvious that junction attenuation
is smallest (in average around 0 dB) for the “Ff”-path in
the continuous floor slab and greatest (in average around
20 dB) for the vertical and diagonal “Ff”-paths between
the upper and lower walls. , , for the vertical and
diagonal case agrees well, however, it tends to be about 2-
3 dB greater for the diagonal. ,, for horizontal and
vertical transmission lay in the middle with in average
about 10 dB. There is a good agreement of both data sets
below 100 Hz. Then, the difference between the two data
sets increases to about 8 dB in the middle frequency range
with higher values for the horizontal case, before both
data sets approach each other above 1000 Hz. This
indicates that at low frequencies both wall leaves act as a
single plate and move with the same velocity. In the mid-
frequency range both leaves do not move equally and
structure-borne sound transmission to the farther element
(e.g. from element 1 to 2 B in Figure 6) in case of
horizontal transmission is more attenuated. This is due to
the transmission losses to the closer element (e.g. from
element 1 to 2 A) and the blocking of energy by this
element. In the high frequency range the wall leaves are
less coupled to the floor because they form a series of
point connections, as can be seen also at the decrease of
,, along the floor slab and its increase for the walls.
Figure 7. Measured Dv,ij,n at cavity-wall-floor-junction.
To better understand junction dynamics, ,, of the
cavity-wall-floor-junction are compared to the results of
FORUM ACUSTICUM 2014 Schoenwald, Stefan: Junctions of Solid Wood Double Walls
7–12 September, Krakow
the same junction with a single leaf 3-ply wall attached to
the continuous floor slab. In Figure 8 ,, are presented
for horizontal and vertical transmission. For the
horizontal case the single wall (black) and cavity wall
(blue) agree very well. This indicates that the coupling of
the walls to the floor slab is very weak, and transmission
in the continuous floor slab is only marginally effected by
the attached walls. For the vertical case, transmission
between the single walls (black x) is more attenuated
below 250 Hz and less above 630 Hz. In the low
frequency range the impedance difference to the floor is
greater for the single leaf wall as it has only half of the
mass and therefore structure-borne sound transmission is
more attenuated which gives greater , ,. In the high
frequency range coupling of the single walls to the floor is
greater, as the double number of connectors is used –
brackets are attached on both sides of the leaf – whereas
for the double wall the connectors are only located on one
side.
Figure 8. Comparison Dv,Ff,n for single-leaf-wall and cavity-
wall.
Figure 9. Comparison Dv,Fd,n and Dv,Df,n for single-leaf-wall
and cavity-wall.
Not so clear are the results for , , in Figure 9. For the
single leaf wall junction ,, for horizontal and vertical
transmission are identical and are expected to agree best
with the data for the vertical for the double leaf wall
junction. At low frequencies the attenuation is greater for
the single wall junction, which also supports the findings
above for ,,, however, in the mid frequency range the
values lay in-between the vertical and horizontal case and
approach the vertical case only at high frequencies. For
the matter of illustration the measurement uncertainty is
indicated by the dashed lines, however, only for the two
data sets of the double wall. For the single wall data the
uncertainty has about the same magnitude. The
uncertainty for all cases being much less than the
difference of the results between 160 Hz and 1250 Hz,
indicates that the measured differences are significant.
This suggests that for a double wall junction, attenuation
is less for the leaf that is closer to the excited floor and
greater for the leaf that is farther away.
4.3. Coupling of the leaves by the floor
As next step the importance of the flanking sound
reduction path, “Dfd”, controlled by the coupling of the
two wall leaves, is investigated. Since it is not possible to
measure this path directly on the specimen as set up in the
transmission test facility, it was predicted utilizing the
EN 12354-methods with some assumptions. First, it was
assumed that structure-borne sound transmitted through
the floor for example horizontally from one leaf to the
other leaf of the upper wall (Dfd) should be about the
same as from one leaf of the upper wall (2A) to one leaf of
the lower (4A or 4B) when the connection details are the
same like in the presented case. These ,, are known
from measurement and together with resonant sound
reduction index for the single leaf 3-ply CLT wall, the
flanking sound reduction index of path “Dfd” was
estimated with Equation 1.
Figure 10. Effect of floor-coupling on direct sound insulation.
The dashed red line in Figure 10 is the estimated flanking
sound reduction index (Dfd) for the coupling of the double
leaf wall at the top or bottom by a continuous floor. The
black line is the lab measured sound reduction index of
the double wall (Dd) without any connections. The red
solid line is the resultant of both (Dd+Ddf) that shows
coupling the double wall to the continuous floor reduces
the sound insulation only in the frequency range above
1000 Hz.
Further, in Figure 10 the situation with a continuous floor
slab on the top and bottom of the wall, like in a real
building, was predicted and also in this case the effect is
limited to the high frequency range and single number
ratings reduce only by one point.
FORUM ACUSTICUM 2014 Schoenwald, Stefan: Junctions of Solid Wood Double Walls
7–12 September, Krakow
The difference of the sound reduction index with and
without coupling is plotted as red line in Figure 11 in
octave bands. Additionally, the difference of the velocity
level difference between leaf 2A and 2B of the double was
measured at the junction mock-up and a second time when
the two wall leaves when they were disconnected from the
floor and placed on resilient pads to reproduce the
situation in the Wall Sound Transmission Facility when
Rw was measured. This difference of the velocity level
differences is shown as blue line. The good agreement of
both supports the correctness of the assumptions made in
the prediction above.
Figure 11. Comparison of effect of floor-coupling on sound
reduction index and veloctiy level difference.
4.4. System performance in the building
Finally, the apparent sound insulation between two
horizontally adjacent rooms that are separated by a double
3-ply CLT wall is predicted using Equation 1. The
geometry of the rooms is the same as for the junction
mock-up. The 5-ply CLT floor below and above the
rooms is continuous across the junction. Further, it is
assumed that side walls are discontinuous across the
junction and hence flanking involving them is negligible.
Figure 12. Predicted apparent sound insulation between
horizontally adjacent rooms.
Figure 12 shows sound reduction index results at one
junction for the direct path, for the significant flanking
paths, the total flanking, as well as for the apparent. Note
that Df is not shown as it is very similar to Fd. The curves
clearly show that the coupling between the two leaves is
the least significant path, whereas the floor-floor path is
the most dominant with the continuous slab. The flanking
sound insulation is equal to direct below 315 Hz and much
less above. The first measures to improve the sound
insulation would be to apply floating floor toppings and
hung ceilings, thereby reducing transmission through the
Ff path. Such measures will automatically lead to
improvements of the Fd and Df-paths. Furthermore
additional measures at the separating wall will beneficial
the Fd, Df and Dfd paths as well as for the direct sound
insulation that controls the apparent sound reduction at the
low frequency range.
5. Summary and Conclusions
In this paper a brief review on the application of the
ISO 15712 methods for prediction of flanking sound
transmission in Cross Laminated Timber (CLT) was
given. The measurement of the necessary input data for
the element performance as well as assumptions made in
the prediction was outlined. For a junction with double
walls the so-called normalized velocity level differences
,, that describe the attenuation of structure-borne
sound at the junction are different for horizontally and
vertically adjacent rooms and one additional flanking
path, coupling of the wall leaves by the floor has to be
taken into account. The measured ,, demonstrated that
the continuous floor at a double wall junction is most
important flanking path and the vertical wall-wall paths
are least important. For the used connection methods of
the walls (angle brackets) structure-borne sound
transmission is equal for junctions with a single 3-ply and
double 3-ply CLT wall is attached. Differences are
marginally greater for the other ,,, however, data for a
single wall junction could be used as first estimate for a
double wall junction if handled with care. Further a
method to estimate the effect of coupling the wall leaves
with a continuous floor was shown and validated. It was
demonstrated that for the tested junction this coupling was
negligible as the most dominant transmission path was due
to the continuous floor. As first step to improve the
apparent sound transmission either additional floor and
ceiling treatments or a structural break in the floor at the
junction are necessary.
Acknowledgement
This project was conducted at NRC Construction as part
of a client funded research project on mid-rise wooden
construction. The authors acknowledge the support of
Canadian Wood Council, FPInnovations and the Provinces
of Quebec and Ontario.
References
[1] S. Schoenwald, B. Zeitler, I. Sabourin, F. King: Sound
Insulation Performance of Cross Laminated Timber
Building Systems. Proc. of Internoise 2013, Innsbruck,
Austria.
[2] CEN/TC 126/WG 02: Fourth draft of EN 12354-1. Doc.
Number: N 0330, CEN-AFNor
