Noise indicators and limits for intermittent indoor low frequency noise Torben Holm Pedersen 1 SenseLab, FORCE Technology Venlighedsvej 4, 2970, Hørsholm, Denmark Frank Pedersen 2 The Danish Environmental Protection Agency Tolderlundsvej 5, 5000 Odense, Denmark

ABSTRACT In Denmark indoor low frequency (LF) noise is measured with the indicator L Aeq,10min in the frequency range 10-160 Hz (L ALFeq,10min ). Recent complaints related to short intermittent low frequency noise events (sounds from underground metro, light rail, neighbouring fitness centre…) could indicate that the correlation between L ALFeq,10min and the experienced annoyance is poor. A literature study for LF descriptors and noise limits in other countries has been made. Listening tests on annoyance of various types of intermittent indoor LF noise types is performed and the results are correlated with alternative relevant noise indicators. Based on the results, noise indicators and noise limits are proposed for rail traffic in tunnels and for activities other than traffic.

1. INTRODUCTION

In 1997, the Danish Environmental Protection Agency issued guidelines for measuring and assessing low-frequency noise, infrasound and vibrations in external environments, cf. [14]. For low frequency noise the A-weighted equivalent level in the range 10-160 Hz, L ALFeq,10min , was proposed as an indicator. The limit values are applicable indoor: In living rooms 25 dB during day/evening time and 20 dB for the nighttime. If impulses in the noise are clearly audible, the limits are tightened by 5 dB.

Recent incidents indicate that a 10 minute equivalent level is not representative of the annoyance caused by short-term noise, such as vibration-generated noise from passing metro trains in tunnels or noise from gyms due to frequent throwing or dropping of weights. Based on similar incidents during the last 20 years in general and citizens being highly annoyed by low-frequency noise below the noise limits from metro trains in particular, the Danish Environmental Protection Agency asked FORCE technology to conduct an investigation of measurement and integration times for short-term, low- frequency noise events. This investigation included a literature study and a listening test. 2. LITERATURE STUDY

The literature study is reported in [1], with a list of 25 references, some of these are listed here (references [2] to [12]). Both laboratory studies (listening tests) and field surveys are included.

A general conclusion among several of the studies is that the number and frequency of events have an impact on the annoyance and an impact on sleep if there is noise at night. In addition, it is also

1 thp@forcetechnology.com 2 frape@mst.dk

general that methods that take the time variation into account show the best correlation with the annoyance. In relation to sleep disturbance, the maximum level is a relevant indicator.

It can also be concluded that analyses of short-duration, low-frequency events are complex. The definition of such events is broad, and the same indicator may not be optimal for measurements of both impulses and pass-byes as two examples of short-duration, low-frequency events. It is also not a given that the same indicator is optimal for both annoyance in the day and evening and sleep disturbances at night.

Examples of independent variables, besides the level, in the analysis of short-duration, low- frequency events are:

− The repetition of events, including whether the interval between events is constant or

variable − Frequency content − Onset and decay time − The prominence (relative to the background noise) − Duration.

The literature study shows that the Danish penalty of 5 dB for clearly audible impulses can be justified. However, the studies indicate that further investigations should be made of whether a penalty of 5 dB is enough. In relation to short-duration, low-frequency events that relate to pass-by noise (which normally is not characterized as an impulse), the studies show more mixed results.

The literature study indicated that this area should be investigated further. 3. PRESENT NOISE INDICATORS AND NOISE LIMITS

As part of the literature study [1], noise limits and indicators in different countries were investigated. A review of different countries' rules and limit values, which can be applied to short- duration, low-frequency events, leaves the impression that there are not many common features.

Denmark is one of the only countries to have specific guideline values for low-frequency noise measured as L Aeq,10min in the frequency range 10-160 Hz, with 5 dB penalty for impulsive sounds [14], se Table 1.

Table 1. Danish indoor (living rooms) noise limits, in dB, for activities other than transport according to [13] and [14]. Limits in parenthesis apply for L Aeq,t . If the sound includes prominent tones or impulses - for L ALFeq only if the sound includes prominent impulses

Time L Aeq,t t L Amax,F L ALFeq,t t L ALFmax,F Day and evening 07-18 30 (25) 8 hours - 25 (20) 10 min. - Night 18-07 25 (20) ½ hour 40 20 (15) 10 min. -

Table 2 shows the Danish indoor noise limits for traffic noise in new residences with closed windows but open fresh air valves, [15].

Table 2. Danish indoor noise limits, L den in dB for road and rail noise in new residences for sound classes A, B and C, according to [15]. The values for evening are estimated as L den – 5 dB and in the night as L den – 10 dB, but the actual values depend on the traffic distribution.

Class C Class B Class A L den 33 28 23 Evening (L Aeq =28) (L Aeq =23) (L Aeq =18) Night (L Aeq =23) (L Aeq =18) (L Aeq =13)

According to [1], there are several Swedish limits. Traffic management targets for structure borne noise state that L Amax, SLOW = 30 dB as a starting point must not be exceeded in living spaces. For new constructions, noise calculations must aim to be 3-5 dB below the limit, i.e. L Amax, SLOW = 25-27 dB. The measurements on stimuli used in the listening test show that the maximum values with time weighting F are 2-3 dB higher than with time weighting S. I.e. if 2 dB is added, A-weighted maximum values of 27-29 dB are obtained with time weighting F. The Swedish Transport Administration sets a limit at night of 32 dB. Furthermore, The National Board of Health and Welfare has recommended criterion curves for noise between 31.5 Hz and 200 Hz.

The corresponding Norwegian limits are L Amax, FAST = 27-32 dB for residences in sound classes B- C. In Norway there are also criteria for the number of incidents (maximum 10) for outdoor activities (roads, railways, industry) and a special criterion curve for the 1/1 octave frequency bands 31.5, 63 and 125 Hz with maximum values for the 1/1 octave bands. These are linked to low-frequency noise in the residential environment in connection with sound class definitions, [16].

It is characteristic that most of the countries studied have rules for maximum values with respectively time weighting F (Fast) or S (Slow). In Norway, L A5% (which denotes the noise level that has been exceeded for 5% of the time) has been used for construction sites. L A5% has the advantage that the value converges so that the result gets closer to the true value with an increasing number if measurements, while the maximum values may change when new events occur.

In New South Wales, Australia, the term “Intrusive Noise Level” is used. The limit value is defined as a level above the background noise level based on of L Aeq,15min (the background noise can also be determined from measurements of L A90% ). In the case of low-frequency noise, the noise level is adjusted by 2 dB during the day and 5 dB at night.

In summary, it can be said that the following indicators and criteria are used in the countries surveyed:

− Maximum value (integration time S or F, or as statistical level L A5% ) − Criteria values in the individual octaves or 1/3-octaves at the lowest frequencies − Adjustment for impulse characteristics − The number of incidents. 4. LISTENING TEST

4.1 Purpose

The purpose of the listening test was to enlighten the relation between the annoyance of short- duration, low-frequency events, and alternative noise indicators in order to provide a qualified proposal for measurement procedures, integration times and proposals for limit values.

Proposals for limit values must be based on the same level of protection of citizens as being defined in the present guidelines [14]. Thus, it must be assumed that the noise annoyance of low-frequency noise will be significant already slightly above the hearing threshold, while the noise annoyance from non-low-frequency noise is assumed to be graduated, so that it is not necessarily annoying when it is slightly audible. 4.2 Details of the listening test

The listening test was performed in accordance with a former developed guideline for listening tests, [17].

The listening test was performed in a listening booth with a background noise level below 10-12 dB(A). Level calibrated headphones (Sennheiser HD 650) which were equalized within +/- 1 dB in the range 10-1000 Hz was used.

21 assessors, 14 men and 8 women, participated in the test. They were naive (not trained) persons, living in apartments and had normal hearing for their age. Ages 23-62 years, average 39,5 years.

The duration of each stimulus was 1 minute including silent intervals between the sounds. The presentation order of the stimuli was randomized individually for each assessor. The total number of stimuli incl. repetitions was 66.

The assessors were asked to imagine that they were sitting and relaxing at home when hearing these sounds. They should assess how annoying the noise would be if it continued the same way, the whole evening. The answering scale is shown in Figure 1.

Figure 1. The answering scale for the assessment of annoyance. The labels on the scale (which were in Danish) were the labels defined in ISO/TS 15 666, [18].

The assessors were reading a book during the presentations of stimuli. 4.2 Stimuli

All stimuli were from natural noise sources recorded in nearby residences with low noise measuring equipment. The level of recorded stimuli was adjusted to have max levels (L Amax,F ) as stated in Table 3.

Table 3. The stimuli and levels used in the listening tests

Source Number of events Nominal L Amax,F , dB Repetitions

Bakery 1, 2 ,3 30 40 50 1

Fitness centre 1, 2 ,3 30 40 50 1

Steps 1, 2 ,3 30 40 50 1

Heating plant Cont. 25 35 45 2

Metro (tunnel) 1, 2 ,3, 4 30 40 50 1

Train (tunnel) 1, 2 ,3 30 40 50 1

Road traffic (indoor)* Cont. 25 30 35 40 45 50 2

*Reference sound

Notatall —Sighty Moderately Very——_—atremely How annoyingdo you ° perceive the sound? Annoyancescore. ot 3 3 6 7 8 9 Not visibieto the listeners

Pilot tests showed that the silence in the listening booths was felt unnatural if the assessors should imagine being in their own flat. Therefore, constant traffic noise heard indoor at a level of 20 dB(A) was added during the whole listening session. Two examples of level recordings of the stimuli are shown in Figure 2 and spectra of the stimuli are shown in Figure 3.

Figure 2. Examples on A-weighed level recordings with time weighting F, for two stimuli with 3 sound events and maximum levels of 40 dB(A). The duration of all stimuli was 1 minute. The red lines indicate added background noise.

Figure 3. A-weighed third octave band spectra of the stimuli. The levels of the spectra are visually adjusted to an average of 20 dB(A) in the range 100-400 Hz for easier comparison.

4.3 Noise indicators tested

Measurements were made on all stimuli with the following noise indicators:

− A-weighted sound pressure level with time weighting F:

o Maximum level, L Amax,F o 1%, 5% and 10% fractile levels, L A01,F L A05,F L A10,F o Equivalent sound level L Aeq,1min.

Felaive Aseighted1/3-ocavebandleel eb o BR Ee BRREORESSSERERE

− A-weighted sound pressure level with time weighting S:

o Maximum level, L Amax,S o 1%, 5% and 10% fractiles, L A01,S L A05,S L A10,S

− A-weighted low frequency level with time weighting F:

o Maximum level, L ALFmax,F o 1%, 5% and 10% fractile levels, L ALF01,F L ALF05,F L ALF10,F o Equivalent sound level L ALFeq,1min.

− L Geq , G-weighted equivalent sound pressure level

− Hearing threshold weighted sound pressure level:

o L HTeq , equivalent hearing threshold weighted level o L HTeq with (ISO) graduated tone- or impulse penalty

− L Aeq with graduated impulse penalty, K I − L Aeq with graduated tone penalty, K T − L Aeq with (DK) graduated tone- and impulse penalty, L rTI − L Aeq with (ISO) graduated tone- and impulse penalty, L rTI − L Aeq with (ISO) graduated tone- or impulse penalty, L rT/I − L Aeq with (DK) 5 dB tone- and impulse penalty − L ALFeq with some of the penalties above

The hearing threshold weighted measurements were introduced as a supplement to the A-weighted measurements because the levels tested are very low and because the A-weighting overestimates the loudness level at low levels at low frequencies. At 10 Hz, the A-weighting overestimates the level by almost 30 dB relative to the hearing threshold, see Figure 4.

Figure 4. The characteristics of the hearing threshold and the A-weighting, [19].

In this paper only results for indicators with the highest correlations with annoyance will be shown. 5. CONSIDERATIONS ON LISTENING TESTS IN RELATION TO REAL LIFE

The participating listeners were representative in the sense that they were average persons (not (acoustic) students or experts), living in flats and ages 23-62. The stimuli were recordings from real life of real sources in residences, and the stimuli were presented at relevant levels with representative background noise. The duration of the stimuli was only 1 minute, which is short compared to e.g. a whole evening and the stimuli were heard in a laboratory setting and not experienced in the listeners own homes.

el 8 Bebesssessos

Based on the above it is found reasonable to assume that the results are representative for the suitability of alternative noise indicators to predict the experienced annoyance in general.

Whether the absolute annoyance ratings are directly related to the experience of the same sounds at home is not so evident. To enlighten this issue, we have looked at the results from the reference noise, which is a known sound source for all listeners. The results are shown in Figure 5.

The dose-response curve (model) in this and succeeding graphs is the result of a logistic regression. R 2 is the fraction of explained variance by the logistic regression.

Figure 5. Dose-response curve for the refence noise, continuous traffic noise heard indoor. The points indicate the average annoyance score for each of the repetitions. The pink shaded area is the approximate 95% confidence interval for the model.

From Table 2 it is seen that the average noise limit for transportation noise in the evening in residences in class A and B is approximately L Aeq = 21 dB. When the listeners are exposed to that level in the listening test it is seen from the dose-response curve in Figure 5, that this corresponds to an annoyance score on 1. This score is therefore assumed to be a reasonable annoyance level for proposing a noise limit.

6. RESULTS

6.1 Correlation between annoyance and noise indicators

Table 4. Correlation coefficients between the mean annoyance score and some of the noise indicators. K I and K T indicate adjustments for prominent impulses and/or tones.

Indicator All sources

Indicator Tunnel Indicator Non transport L Aeq +K I +K T (ISO) 0,89 L ALFeq 0,93 L HTeq 0,95 L Aeq +K I 0,88 L Aeq 0,92 L Aeq +K I 0,94 L ALFeq +K I +K T (ISO) 0,86 L ALFeq +K I 0,92 L Aeq +K I +K T (ISO) 0,91 L ALFeq +KI 0,85 L Aeq +K I 0,89 L Aeq 0,89 L HTeq 0,82 L HTeq 0,89 L Amax,F 0,88 L Aeq 0,83 L ALFeq +K I +K T (ISO) 0,86 L ALFeq +K I +K T (ISO) 0,85

‘01-0 S108 Souekowe sbeiony —

L ALFeq 0,80 L ALFmax,F 0,84 L ALFeq +K I 0,84 L Amax,F 0,8 L Aeq +K I +K T (ISO) 0,83 L ALFmax,F 0,83 L ALFmax,F 0,8 L Amax,F 0,79 L ALFeq 0,82

From Table 4 it is seen that the highest-ranking indicators are different whether all sources are included or whether the sources are divided into noise from trains (including metros) in tunnels as one part and non-transport sources as another part. Therefore, it was decided to divide the sources in “tunnels” and other activities. Furthermore, the correlation increases when the sources are divided into these subgroups. 6.2 Trains in tunnels This group includes metros and other trains in tunnels. Dose-response curves based on logistic regression of the results from the listening test are shown in figures 6 to 8. The vertical lines indicate the 95% confidence intervals of the points.

ant a 01-0 21098 a2ueKouue aBeiany

Figure 6. Dose-response curve for low frequency A-weighted L eq , L ALFeq , for rail traffic in tunnels.

Figure 7. Dose-response curve for L Aeq , for rail traffic in tunnels.

i in 01-0 21098 sauexowe aBesony

Figure 8. Dose-response curve for L Amax,F , for rail traffic in tunnels.

0 (01-0 a1095 saueKowwe oBesony

For indoor noise from metro and other trains in tunnels, L Aeq is preferred to L ALFeq because the noise from the Metro contains energy above the low frequency range 10-160 Hz. The results indicate that a proposal for indoor noise limits in the evening when the noise is worst, could be:

L Aeq,10min = 20 dB and *L Amax,F = 30 dB

* It is recommended that L Amax,F is replaced with the statistical level L A05,F as this converge against a stable value with increasing number of observations, in contradiction to L Amax,F . For the stimuli in this test, the difference was less than 1 dB between these two indicators.

The proposal is in line with other existing noise limits in Denmark, Norway, and Sweden, see clause 3.

6.3 Activities other than traffic

‘01-0 31098 Souekou jue aBeiany x» o Lee *Ki OF +Ky JB » 10

Figure 9. Dose-response curve for hearing threshold weighted L eq , L HTeq with adjustment for prominent impulses or tones (the largest of the two), for activities other than traffic.

The group “Activities other than traffic” includes noise from commercial, industrial and leisure activities. Dose-response curves based on logistic regression of the results from the listening test are shown in the figures 9-12. The vertical lines indicate the 95% confidence intervals of the points.

01-0 108 198 Souerouue aBeieny ry Lig tKi0r# Ky 6B Ey 0

Figure 10. Dose-response curve for L Aeq with adjustment for prominent impulses or tones (the largest of the two), for activities other than traffic.

01-0 91098 souerowve

Figure 11. Dose-response curve for L Amax,F for activities other than traffic.

Figure 12. Dose-response curve for low frequency A-weighted L eq , L ALFeq , for activities other than traffic.

The indicator with the best fit to the data from the listening test is the hearing threshold weighted L eq , L HTeq with adjustments for impulses or tones (the largest of the two) and this is an interesting indicator for low level noise. Future investigations of noise should include this indicator. It is not a standardized indicator; therefore, it is not (yet) suggested for use in the administration of environmental noise.

For indoor noise from activities other than transport a proposal (within the uncertainty of the results) for indoor noise limits in the evening when the noise is worst, could be:

»” Laurea *Ki or Ky AB » 0 01-0 21098 sauexowe abesony

L r,10min = 25 dB and *L Amax,F = 30 dB and L ALFeq,10min = 20 dB

where L r : L Aeq + K I or K T (the largest)

* It is recommended that L Amax,F is replaced with the statistical level L A05,F as this converge against a stable value with increasing number of observations, in contradiction to L Amax,F . For the stimuli in this test, the difference was less than 1 dB between these two indicators.

8. ACKNOWLEDGEMENTS

This investigation was financed by The Danish Environmental Protection Agency. 9. REFERENCES

[1] M. Herlufsen, M. B. Hansen, and P. Finne,”Measurement and integration times for short

duration low frequency noise events. Part 1: Litterature study,” (In Danish) 2021.

[2] A. Moorhouse, D. Waddington, and M. Adams, “Proposed criteria for the assessment of low

frequency noise disturbance,” Manchester, 2005.

[3] T. Poulsen and F. R. Mortensen, “Laboratory Evaluation of Annoyance of Low Frequency

Noise,” 2002.

[4] C. H. Kasess, T. Maly, P. Majdak, and H. Waubke, “The relation between psychoacoustical

factors and annoyance under different noise reduction conditions for railway noise,” J. Acoust. Soc. Am. , vol. 141, no. 5, pp. 3151–3163, 2017, doi: 10.1121/1.4982878.

[5] P.-A. Vallin, C. Marquis-Favre, J. Bleuse, and L.-A. Gille, “Railway noise annoyance

modeling: Accounting for noise sensitivity and different acoustical features,” J. Acoust. Soc. Am. , vol. 144, no. 6, pp. 3381–3390, 2018, doi: 10.1121/1.5082296.

[6] B. Park, J. Y. Jeon, S. Choi, and J. Park, “Short-term noise annoyance assessment in passenger

compartments of high-speed trains under sudden variation,” Appl. Acoust. , vol. 97, no. October 2018, pp. 46–53, 2015, doi: 10.1016/j.apacoust.2015.04.007.

[7] G. M. Aasvang, T. Moum, and B. Engdahl, “Self-reported sleep disturbances due to railway

noise: Exposure-response relationships for nighttime equivalent and maximum noise levels,” J. Acoust. Soc. Am. , vol. 124, no. 1, pp. 257–268, 2008, doi: 10.1121/1.2932074.

[8] G. M. Aasvang, B. Øverland, R. Ursin, and T. Moum, “A field study of effects of road traffic

and railway noise on polysomnographic sleep parameters,” J. Acoust. Soc. Am. , vol. 129, no. 6, pp. 3716–3726, 2011, doi: 10.1121/1.3583547.

[9] M. Vernet, “Comparison between train noise and road noise annoyance during sleep,” J. Sound

Vib. , 1983, doi: 10.1016/0022-460X(83)90571-0.

[10] E. Öhström and R. Rylander, “Sleep disturbance by road traffic noise-A laboratory study on

number of noise events,” J. Sound Vib. , 1990, doi: 10.1016/0022-460X(90)90570-P.

[11] S. Pennig et al. , “Annoyance and self-reported sleep disturbance due to night-time railway noise

examined in the field,” J. Acoust. Soc. Am. , vol. 132, no. 5, pp. 3109–3117, 2012, doi: 10.1016/j.jtbi.2008.09.025.

[12] M. Brink et al. , “A survey on exposure-response relationships for road, rail, and aircraft noise

annoyance: Differences between continuous and intermittent noise,” Environ. Int. , vol. 125, no. January, pp. 277–290, 2019, doi: 10.1016/j.envint.2019.01.043.

[13] The Danish Environmental Protection Agency (Miljøstyrelsen), "Guideline on external noise

from activities” (in Danish). No. 5/1984.

[14] The Danish Environmental Protection Agency (Miljøstyrelsen), "Low frequency noise, infra-

sound and vibrations in the external environment" (in Danish). Orientering nr. 9/1997. 1997.

[15] Building regulations BR18 (in Danish). Bygningsreglementet.dk,

[16] Norwegian Standard (in Norwegian), “Lydforhold i bygninger - Lydklasser for ulike

bygningstyper”.

[17] T. H. Pedersen and R. S. H. Skov, “Guideline: Listening tests for measurement of the relative

annoyance and the annoyance potential of noise,” 2020. https://forcetechnology.com/- /media/force-technology-media/pdf-files/unnumbered/senselab/118-20470---landsbyggefonden- --guideline-listening-tests---tc-101522.pdf?la=en

[18] ISO/TS 15666:2021. Acoustics — Assessment of noise annoyance by means of social and

socio-acoustic surveys. 2021.

[19] T. H. Pedersen, “Low Frequency Noise From Large Wind Turbines - A procedure for

evaluation of the audibility for low frequency sound and a literature study,” 2008.