JOURNAL OF LOW FREQUENCY NOISE, VIBRATION AND ACTIVE CONTROL

Pages 15 – 23

Sound insulation of dwellings at low frequencies Dan Hoffmeyer1 and Jørgen Jakobsen2 1DELTA, Venlighedsvej 4, DK-2970 Hørsholm ([email protected]) 2Dansh Environmental Protection Agency, Strandgade 29, DK-1401 København K Email:[email protected] Received 22nd March 2010

ABSTRACT The paper gives a summary of some recent measurements of sound insulation of dwellings at low frequencies (8 – 200 Hz), and compares these data to results from earlier measurements and to selected data from the literature. Data are given for the expected minimum sound insulation of typical Danish dwellings at low frequencies.

1. BACKGROUND As a part of a project "Low Frequency Noise from Large Wind Turbines" undertaken for the Danish Energy Agency by a group under management by DELTA, a series of measurements of low frequency sound insulation (difference of sound level indoors and outdoors) in typical Danish dwellings was performed. These measurements were combined with measurements of noise emission from wind turbines and calculation of noise propagation to allow for an evaluation of the possible impact of the low frequency noise from wind turbines in neighbouring dwellings. These measurements are reported in [1 – 3]. The noise insulation measurements were performed by the use of different measurement methods. One of the methods operated with indoors noise measurements in four positions at the smallest possible distance (0,01 – 0,02 m) from the three-dimensional corners of the room, where the walls meet the ceiling or the floor. This measurement method has been suggested in [4], where it is named 3D-measurements. It has been found that the noise level in the higher end of the low frequency range can be up to 10 dB higher in the corners than in areas more distant from the corners. Measurements were also performed at indoor positions representing he parts of the room normally occupied by the occupants. The Danish Environmental Protection Agency (Danish EPA) asked DELTA, who had performed the measurements, to make supplementary analyses of the data obtained by use of the measurement method recommended by the Danish EPA for general measurements of low frequency noise [5]. These data would be comparable to a set of data obtained earlier for assessment of low frequency noise from high-speed ferries [6]. 2. NEW MEASUREMENTS OF SOUND INSULATION AT LOW FREQUENCIES In the project "Low Frequency Noise from Large Wind Turbines" sound insulation measurements were made in nine rooms (five living rooms and four smaller rooms, typical bedrooms] in five dwellings [1]. The dwellings were selected to represent typical homes at the countryside which could be subject to noise form wind turbines. One of the dwellings was a traditional farmhouse, while others represented modern dwellings, some with large panoramic view windows, and some had a lightweight façade construction like typical summer cottages. A large loudspeaker was placed at some distance (at least 5m) from the dwelling with the angle of sound incidence Vol. 29 No. 1 2010

15

Sound insulation of dwellings at low frequencies about 45o, emitting broadband pink noise, limited to the range below 250 Hz and equalized to compensate for the loudspeaker characteristics. Simultaneous measurements were made of the sound outdoors – with a microphone placed directly upon the façade – and indoors in a number of positions in each room. With the loudspeaker turned off, the background noise was measured in the same positions, and all the measurement results were corrected for the background noise. When the signal from the loudspeaker was less than 6 dB above the background noise, the correction was limited to 1.3 dB. This may result in an overestimate of the indoor sound level and hence an underestimate of the level difference (sound insulation). Generally the correction was only limited in this way at frequency bands 8 and 10 Hz, in a few cases also at 12.5 Hz. When the difference between the levels with the loudspeaker on and off was less than 1.3 dB, the result was discarded. The outdoor sound level, measured very close to the façade, was subtracted by 6 dB to express the level of the incident sound. The sound insulation is expressed as the level difference, which is the difference between the outdoor level of the incident sound and the energy average of the sound level in three or four positions, each corrected for background noise as described. It was not attempted to normalise or to correct for the acoustic conditions of the measurement room. To allow for the different assessment methods, several indoor measurement positions were used in each room. In addition to the four positions at the smallest possible distance from three-dimensional corners of the room as suggested in [4], five positions were chosen to allow for measurements in accordance with the international standard for measurement of sound isolation of facades [7] and the recommended Danish method for measurement of low frequency noise [6]. Two of the positions were near a corner of the room at 0,5 – 1 m distance, and three positions were chosen to represent the spaces typically occupied by the occupants. According to the recommended measurement procedure [6], the average of the sound in three positions including one position near a corner shall represent the sound level in the room. The five measurement positions could be arranged in nine different ways to fulfil this recommendation, three of these using two corner positions and only one position further form the walls. Below in Figures 1 and 2 all the results are shown for two rooms. In the figures, the curves marked "LFM 1" to "LFM 3" use two positions near a corner and "LFM 4" to "LFM 9" use one position near a corner and two positions in representative areas. "3D" is the measurement where four microphones were placed at the smallest possible distance from the corners of the room.

Figure 1.

16

Ten different results of level difference from a room “Vaerløse Stue”.

JOURNAL OF LOW FREQUENCY NOISE, VIBRATION AND ACTIVE CONTROL

Dan Hoffmeyer and Jørgen Jakobsen

Figure 2.

Eleven different results of level difference from a room "Tulstrup værelse". The level difference marked LFM 10 is obtained using only two positions near the corners.

It is seen that all the measurement results generally agree at the lowest frequencies, up to about 50 Hz. At higher frequencies the measurements using the three dimensional corners (3D) gives consistently lower level differences than the other arrangements. This is because the average sound level measured in the "3D-positions" are generally 5 – 7 dB, occasionally up to 10 dB, higher than the other averages. On the other hand, the nine different arrangements of three measurement positions show a good mutual agreement within the entire frequency range. The three different averages using two microphone positions near corners and one further from the walls (LFM 1 – LFM 3) do not deviate in any significant way from the other results. According to the original report [1] it has not been possible to establish the connection between the measured level differences at low frequencies and the building acoustic conditions of the building, in particular the type of the façade and the type and size of the windows. Results from dwellings with lightweight facades and / or large window areas did not deviate significantly from the results from heavier buildings with small windows. Hence it was chosen to regard the buildings as equally typical Danish dwellings, and to analyse all the data statistically. In the present situation three different averages of the results from each dwelling have been calculated: one for the "3D-measurements", one for one of the measurement procedures using two positions near a corner and one supplementary position (arrangement LFM 1), and finally an average for one of the measurement procedures using one position near a corner and two positions in typical occupied areas (arrangement LFM 5). The three averages are illustrated in Figure 3 below.

Figure 3.

Comparison of the averaged level differences from all nine rooms, using respectively the measurement procedures “3D’, ‘LFM 1’, and “LFM 5”. Vol. 29 No. 1 2010

17

Sound insulation of dwellings at low frequencies Here it is evident that the level difference measured by use of the "3D-positions" is 5 – 10 dB lower than the other two level differences at the frequency band 63 Hz and above. The two results using the LFM 1 and LFM 5 procedure respectively agree closely (generally within 1 dB) except for deviations up to 3 dB in the frequency range 31.5 – 50 Hz. Possibly the sound level in this frequency range is higher near the corners than elsewhere in the rooms, so the levels using the LFM 1 procedure using two positions near the corners are slightly higher than the procedures where only one position near a corner is used. 3. EARLIER MEASUREMENTS OF SOUND INSULATION AT LOW FREQUENCIES In 1996 the Danish Environmental Protection Agency asked DELTA to measure the sound insulation at low frequencies in a number of typical dwellings and to compare these measurements to similar results from the literature. The measurements and the results were to be used as background for a Statutory Order setting limits for the low frequency noise from high-speed ferries [6]. The measurements and the analysis were published in Danish in [8]. Nine dwellings were selected to represent typical dwellings with a sea view, having large windows or glass facades, and measurements were made in 17 rooms. The measurements were made in a way similar to the new measurements described above. However, only three measurement positions were chosen indoors, in accordance with the recommended procedure in [5]. Corrections of up to 6 dB for background noise were allowed, corresponding to a "signal-to-background noise" ratio of only 1,3 dB, since it was found that both the signal and the background noise were stationary. Naturally such large corrections results in a large uncertainty. Generally there were no problems with background noise in the frequency range above 20 – 25 Hz.

Figure 4.

Average level difference from measurements in 17 rooms in 1996 compared to the data in Figure 3.

Considerable effort was made to find a connection between the sound insulation and building acoustics as well as the room acoustics properties of the buildings, but in vain. In the analysis the data were split into two groups, one with a lower sound insulation and one with a higher sound insulation. The standard deviation of the results within these groups was slightly smaller than the standard deviation of the data in the total material. This was the main argument for the procedure chosen. In the frequency range 20 – 50 Hz, the average of the "low" insulation was 6 – 10 dB lower than the average of the "high" insulation, both at lower and higher frequencies no significant difference was found. Next a statistical method was applied to determine the level difference that would be exceeded in at least 90% of the dwellings. This was simplified into 1/1 –octave data and included in an annex to the Statutory Order as the recommended method for the determination of noise level due to high speed ferries. 18

JOURNAL OF LOW FREQUENCY NOISE, VIBRATION AND ACTIVE CONTROL

Dan Hoffmeyer and Jørgen Jakobsen In Figure 4, the average of all the measured level differences from the 17 rooms is compared to the recent measurements from Figure 3. It is seen that the two data sets compare favourable, and in particular there is good agreement between the old data and the new data obtained by procedure LFM 5. This is not surprising since this is one of the procedures in close agreement with the recommendations in [5], which was also followed in the old measurements. Only in the range around 100 Hz, the new measurements showed a significantly higher level difference than the old. The statistical method mentioned above is not entirely transparent, and in the present situation a simpler and more straightforward procedure was adapted. Here the standard deviation was calculated for the same group of data as was averaged, and by subtraction of one standard deviation an estimate of the level difference exceeded by 80 – 90% of the data was obtained. For this measure, the symbol ∆Lσ is used here. Figure 5 shows ∆Lσ for the same data as in Figure 4.

Figure 5.

Level difference exceeded by 80 – 90% of the data (∆Lσ ) from the old and the new measurements applying three different measurement procedures.

4. INFORMATION FROM THE LITERATURE ON SOUND INSULATION AT LOW FREQUENCIES In the 1996 project for the Danish Environmental Protection Agency also data from the literature was investigate. Usually sound insulation measurements are made in the frequency range above 100 Hz, so only few data were found about sound insulation of dwellings at low frequencies. Dutch data In an investigation in the Netherlands in 1990, the level difference at low frequencies in two rooms (one living room and one sleeping room) in each of seven typical dwellings was measured. The measurements were made by using a large servo loudspeaker placed outdoors and measuring the sound in each room by slowly scanning a diagonal in the room. In addition attempts were made to determine the indoor noise generated by a vibrator placed outdoors, generating structure borne noise. In a paper [9] describing the project, the average of the measured level differences in all seven living rooms and all seven sleeping rooms were shown in a figure. In the same figure, the average minus two standard deviations was illustrated. The paper concluded that the level differences of the sleeping rooms were generally smaller due to their smaller volume. The figure showed however that the average level difference of the living rooms was not significantly different from the average level difference of the sleeping rooms, but the standard deviations of the data from living rooms were larger. The data were read off the figure and are shown in Table I below.

Vol. 29 No. 1 2010

19

Sound insulation of dwellings at low frequencies Table I. Average level difference (in dB) of facades to living rooms and sleeping rooms, and standard deviation of the data in the two groups. From [9] Frequency, Hz Site

16

20

25

31.5

40

50

63

80

100

125

160

Living room, avg

6

14

22

22

21

23

22

25

28

30

26

Living room, std.

3,5

2

4

5,5

2

4

2

4

4

3

5

Sleep, avg.

16

20

22

17

17

22

21

23

26

29

26

Sleep, std.

9

6

4,5

4,5

4,5

4,5` 6

7

4

6,5

5

British data In two reports about low frequency road traffic noise and vibrations from TRRL [10, 11], the road traffic noise was measured simultaneously outdoors and indoors of two dwellings where the inhabitants had complained about noise and vibration. The measurements were made in only one indoor position, so the uncertainty can be taken as considerable. The main focus of the investigations was put on the ground borne vibrations from road traffic, so possibly the indoor noise level was influenced by structure-borne noise as well as by airborne noise through the façade. The level differences in the two dwellings were: Table II Level difference in dB measured with road traffic noise in two British dwellings. From [10, 11]. Frequency, Hz Site

10

12.5

16

20

25

31.5

40

50

63

80

100

125

160

Slough Rd,

7

9

11

13

13

9

4

12

15

15

18

15

16

Ludlow

15

15

16

17

16

21

14

22

20

21

21

23

-

American data In a paper about low frequency noise from large wind turbines [12], a curve shows the expected minimum and maximum sound level difference of typical American dwellings. The data are based on the authors’ experience, including earlier measurements of level difference at low frequencies from aircraft noise. Table III. Highest and lowest expected level difference in dB of American dwellings. From [12] Frequency, Hz

8

10

12.5

16

20

25

31.5

40

50

63

80

100

125

160

Max

21

20

19

19

20

20

22

23

25

26

26

26

26

26

Min

2

6

6

5

2

5

8

9

10

11

6

13

13

14

Comparison and comments In Figure 6 the data from the literature is compared to the average data from both the new and the old Danish project. To avoid confusion, only one of the recent measurements, LFM-5 is shown. No clear picture is seen from the graphs. Due to the large interval between the "US min" and the "US max" curves these fence in nearly all the other graphs, but little information is gained. The two graphs from the British investigation of road traffic noise show the same general trend as well as a pronounced minimum at 40 Hz, but elsewhere they differ considerably. Due to the insufficient information available about the measurement procedure and the use of only one indoor microphone position, these data must be considered as rather uncertain.

20

JOURNAL OF LOW FREQUENCY NOISE, VIBRATION AND ACTIVE CONTROL

Dan Hoffmeyer and Jørgen Jakobsen

Figure 6.

Comparison between data on level differences of dwellings at low frequencies from the literature (taken from [9 – 12]) and the measured data (taken from Figure 4)

The two curves from the Dutch investigation agree very closely in the frequency range from 50 Hz and above, and apart from the data at 16 Hz they show a general agreement. These curves represent the average of measurements in seven buildings, and the indoors noise level in each case measured by scanning a microphone through a diagonal in the room, so the data are to be regarded as having small uncertainty. There is a significant difference between the Dutch data from the two types of room and the Danish data, both from the recent and old measurements. In general the level differences from the Dutch project are 3 – 8 dB higher than in the Danish investigations. In the original report [8] it was observed that the (old) Danish measurements were in good agreement with the Dutch data from the literature. However, this observation was in error since the Danish data used for the comparison had not been corrected for the "flush-mounted" outdoors microphone. With a microphone placed very close to the façade, a correction of -6 dB is applied to obtain an estimate of the level of the incident sound. Information has been given [13] that the Dutch measurements were also made with the outdoors microphone placed close to the façade and close to the ground as well, and that no corrections were applied to the measured levels. This can explain the divergence between the Dutch and the Danish data sets quite well, since the outdoor levels in the Dutch measurements are probably overestimated by at least 6 dB compared to the "incident field" level assumed for the Danish data. 5. CONCLUSION – REPRESENTATIVE DATA FOR SOUND INSULATION AT LOW FREQUENCIES A good agreement was found between the recent and the earlier measurements of sound level differences of facades at low frequencies, when comparable measurement procedures were applied. This gives the possibility of merging the average level differences from the two investigations for future purposes. However, the data did not agree well with the most trustworthy data from the literature [9]. The reason for this disagreement was found and discussed above, and for the sake of comparison a correction of minus 6 dB has been applied to the data in the comparison below. The level differences found using the "3D-method" are 510 dB lower at frequencies above 63 Hz than data from measurements using positions as recommended by the Danish EPA. The measured data from both of the Danish projects have been treated statistically, and the level difference expected to be exceeded by 80 – 90% typical Danish dwellings (∆Lσ) has been determined. This is shown in Table IV below. In Figure 7 the data are shown together with the corresponding (∆Lσ ) data from the Dutch investigation, where a suggested correction of – 6 dB has been applied. Also

Vol. 29 No. 1 2010

21

Sound insulation of dwellings at low frequencies the 1/1 –octave data given in the Danish Statutory Orders about low frequency noise from high speed ferries are illustrated. It can be seen that both the original data from the Statutory Orders and the Dutch data, corrected with – 6 dB, compare well to the newly established ∆Lσ. Table IV. Level difference in dB expected to be exceeded in 80 – 90% of typical Danish dwellings. Frequency, Hz

8

10

12.5

16

20

25

31.5

40

50

63

80

100

125

160

200

∆Lσ

2,4

1,2

3,2

2,1

3,6

4,6

6,7

7,6

10,3

14,2

17,5

18,4

17,5

18,6

22,4

Figure 7.

Level difference exceeded by 80 - 90% of the data (∆Lσ) from both of the Danish projects. For comparison the similar data (∆Lσ) calculated from the results of the Dutch investigation [9] are shown, corrected by – 6dB. Also shown are the 1/1-octave data from the recommended Danish calculation procedure for low frequency noise from high speed ferries.

6. FINAL WARNING REMARKS – GROUND BORNE NOISE Data for sound insulation of dwellings at low frequencies are to be applied in the assessment of low frequency noise from external noise sources, provided the noise level outside the dwelling can be measured or calculated. Assessment of low frequency noise is based on the indoor noise level. This method of action will work well, provided the noise is exclusively airborne sound. This is expected with noise from e.g. ships, aircraft, and wind turbines. The data represent typical Danish dwellings and express sound insulation in the low end, since 80 – 90% of the dwellings are expected to have higher sound insulation. The uncertainty in calculation of indoor low frequency noise depends on the agreement between the conditions at the actual site (mainly the type and character of buildings) and the dwellings investigated in the two Danish projects. Some industrial noise sources also generate vibrations, and in dwellings close to these sources the vibrations may propagate into the building and generate audible structure-borne sound. Train traffic and road traffic on uneven roads are also known to generate vibration capable of causing structure-borne sound. In some cases where an important contribution to the indoor noise level may be structure-borne sound, the method where the outdoor noise level is used as basis for calculation of the indoor level will be in error. The indoor noise level can be severely underestimated in such cases.

22

JOURNAL OF LOW FREQUENCY NOISE, VIBRATION AND ACTIVE CONTROL

Dan Hoffmeyer and Jørgen Jakobsen REFERENCES [1] EFP-06 Project "Low Frequency Noise from Large Wind Turbine". Measurements of Sound Insulation of Facades. DELTA AV 1097/08, 30. April 2008. [2]

EFP-06 Project "Low Frequency Noise from Large Wind Turbines". Results from Sound Power Measurements. DELTA AV 136/08-rev. 1, 19. December 2008.

[3]

EFP-06 Project "Low Frequency Noise from Large Wind Turbines". Summary and Conclusions on Measurements and Methods. DELTA AV 140/08-rev. 1, 19. December 2008.

[4]

Pedersen, S., Møller, H. And Persson, K.: "Indoor measurements of noise at low frequencies – problems and solutions". Journal of Low Frequency Noise, Vibration and Active Control, Vol. 26, (2) 2007, 249-270.

[5]

Jakobsen, J.: "Danish guidelines on environmental low frequency noise, infrasound and vibration". Journal of Low Frequency Noise, Vibration and Active Control, Vol. 20 (3), 2001, 141-148.

[6]

Søndergaard, B.: "Noise from Catamaran Ferries". Journal of Low Frequency Noise, Vibration and Active Control, Vol. 18 (3) 1999, 123-127. The regulation described in the paper is: Statutory Orders No. 821 of 23rd October 1997 "Environmental permit to ferry service by high-speed ferries" (In Danish).

[7]

EN ISO 140-5:1998, Acoustics – Measurement of sound insulation in buildings and of building elements – Part 6: Field measurements of airborne sound insulation of facade elements and facades, 1998.

[8]

Danish Environmental Protection Agency – Working report 10/1997: Evaluation of low frequency noise from ferries – part 2 (In Danish).

[9]

Vercammen, M.L.S.: "Low-Frequency Noise Limits". Journal of Low Frequency Noise and Vibration, Vol.11 (1) 1992, 7-13.

[10] Martin, D. J. Et al.: Measurement and Analysis of Traffic-induced Vibration in Buildings. TRRL Supplementary Report 402, TRRL Berkshire 1978. [11] Martin, D. J. Et al.: Low Frequency Noise and Building Vibration. TRRL Supplementary Report 429, TRRL Berkshire 1978. [12] Shepherd, K. P. and Hubbard, H. H.: "Physical Characteristics and Perception of Low Frequency Noise from Wind Turbines". Journal of Noise Control Engineering, Vol. 36 (1) 1991, 5-15. [13] Vercammen, M.L.S. personal communication, August 2009.

Vol. 29 No. 1 2010

23