CASE CONTROL STUDY IN LOW FREQUENCY SOUND MEASUREMENTS

G.P. van den Berg

INTRODUCTION

Measurements of broad and narrow band sound levels have been made in 36 Dutch dwellings in 1998. In 19 dwellings there were complaints about low frequency (LF) noise, in 17 others no complaints had been reported. The complainants had taken part in an earlier study investigating personal characteristics of complainants about LF noise [1]. Both that study and the study reported here [2] have been supported by the Dutch Ministry of Housing, Spatial Planning and the Environment. The purpose of this study was to assess the level and spectral distribution of (LF) sounds in dwellings. To this end a systematic approach has been developed based on earlier measurements [3].

MATERIALS AND METHODS

Cases / controls. Cases are residents from all over the Netherlands with complaints about low frequency noise. They have partaken in an earlier study investigating personal characteristics of LF noise complainants such as age, gender, occupation, hearing threshold (at frequencies above 200 Hz), sound sensitivity, time spent home, etc. Complainants and their partners (as a control group) also completed enquiries about psychological and health status [1]. Most of them had heard a LF sound for a long time, typically several years, either more or less continuously (especially at night), or for days or weeks with 'silent' intervals in between. In most cases complainants did not know the source of the noise. For the measurements in the study reported here, a control group (without LF noise complaints) was composed based on a comparable distribution over the country, types of dwelling and surroundings.

Measurement instruments. For the measurements in the dwellings a TASCAM DA-P1 digital recorder was used with a high quality Sennheiser MKH 20 P48 microphone. The microphone was hung in rubber bands on a tripod to provide vibration protection. Recordings of typically appr. 10 minutes were made and analysed afterwards with a Larson Davis 2800 analyser. The entire measurement chain could measure down to a few hertz and had a flat frequency response (± 0.5 dB) at frequencies of 20 Hz up to several kilohertz. The instrument noise level was at least 10 dB below the ISO hearing threshold in the LF region. The LF region considered here is that part of the spectrum with frequencies 200 Hz, with an emphasis on 20 - 100 Hz.

Much care has been taken to avoid disturbing sounds (preferably no one in room, no clocks etc.). Unavoidable disturbances (such as passing cars) have not been taken into the analyses afterwards. Disturbances were identified by listening to the recordings and by analysis of time histories of the overall C- and A-weighted levels of the recorded sound.

Measurement time and place. Complainants have recorded the sounds themselves after practical ("hands on") instructions to operate the recorder. The microphone was put at a position where complainants were positive the LF sound was usually present. In almost all locations a position could be chosen in a corner of the bedroom (height 1 to 1.5 m, about 0.4 to 0.5 m from both walls); this was also the standard position in the controls' dwellings. The equipment was left long enough (days, sometimes up to some weeks) to ensure relevant recordings could be made. This procedure guaranteed that the recording indeed contained the LF sounds present at the time the LF sound causing the complaints was heard. The method has been adopted in a recent guideline for measuring LF sounds in case of complaints, issued by the Dutch Noise Annoyance Foundation (NSG) [4,5].

Analysis and assessment method. All recordings have been analyzed, leaving out the part(s) with unwanted (disturbing) sounds, to determine a set of broad band equivalent sound levels: the A- en C-weighted level L_Aeq and L_Ceq, and the LF part of the A-weighted level L _Aeq(LF) (L _Aeq for f_{1/3 octave} 100 Hz). Also, equivalent unweighted 1/3 octave band sound levels were determined as well as line spectra in the LF region with a line width 0,4 Hz. The 1/3 octave band sound levels were compared to the ISO median hearing threshold level for otologically selected young adults [6] (hereafter: ISO threshold) so as to have a measure of audibility. The median hearing threshold for unselected people aged 50 to 60 -the average age of complainants- has been determined recently and is 9.5 dB above the ISO threshold. The threshold for the best hearing 10% of this group is 4.5 dB below the ISO threshold, for the best hearing 5% still appr. 2 dB lower [4,7]. At frequencies above the LF region the hearing threshold levels of complainants are as expected according to their age [1].

Also, the 1/3 octave band sound levels have been compared to criterion values according to Vercammen, the German DIN and the Swedish Socialstyrelsen to assess LF noise [8,9,10]. These criteria as such will not be dealt with here, but results will be given in according to these criteria. DIN has different criteria for tonal and non-tonal sound, of which the non-tonal one is used in this study; in only one case, according to DIN, there was tonal sound above the criterion (but the non-tonal sound level also exceeded the criterion).

The line spectra were used to more precisely determine frequencies of narrow band components (a more detailed analysis of increased narrow band sound levels then is presented here is in preparation).

RESULTS

A total number of 93 sound recordings have been made: 34 at night and 25 at daytime in (19) dwellings of cases (= complainants); 17 at night and 17 at daytime in (17) dwellings of controls. Of the 34 nighttime recordings in cases' dwellings 27 were recorded at a time when the LF noise was heard by the complainants, 7 were recorded at a time of 'LF silence'. Daytime recordings (most of them around noon) were made to compare the sound levels at that time with those at night. Each recording is about 10 minutes. Disturbing sounds, typically for a few minutes per recording, have not been taken into the analyses or data reported here.
All sound levels reported here are equivalent broad band or 1/3 octave band sound pressure levels per recording (without disturbances). The analysed time per recording (without disturbances) is appr. 8 minutes.

Broad band sound levels. Table 1 presents an overview of the (arithmetic) average equivalent broad band sound levels in cases' and controls' dwellings at day and night. Figure 1 shows cumulative distributions of A- and C-weighted equivalent sound levels of all recordings separated in time (day/night) and group (cases/controls). As expected, daytime sound levels are higher than nighttime levels. In controls' dwellings the daytime sound levels tend to be higher than in those of cases, at night they are practically equal. The difference L_Ceq - L_Aeq, used (e.g. by DIN [9] ) to distinguish between situations with and without substantial LF sound, evidently does not distinguish between situations with and without complaints (cases / controls). Unexpectedly, this difference is smallest in cases' dwellings at night, highest in controls' dwellings in daytime.
 

Spectral levels. In figure 2 1/3 octave band sound levels of all nighttime recordings in dwellings of cases have been plotted. Recordings at a time when no LF sound was heard are not included. To have an indication of audibility, the ISO threshold has been added. In figure 3 spectra depicting the median value and the median value plus and minus one standard deviation (s.d.) of all 1/3 octave band sound levels in figure 2 have been plotted. The area between the upper and lower lines thus contains 74% of all 1/3 octave band sound levels. No 1/3 octave band sound level is above the ISO threshold for f_{1/3 octave} < 40 Hz: therefore, no significant (with respect to audibility) infrasound is present. For f_{1/3 octave} 80 Hz at least 50% of all 1/3 octave band sound levels are below the ISO threshold, and for f_{1/3 octave} 300 Hz all 1/3 octave levels are above this threshold. To figure 3 the same statistics of the daytime recordings have been added. In the lower part of figure 3 the differences between controls' and cases' median day and nighttime 1/3 octave band sound levels are plotted. There is a marked difference for daytime sound levels: in controls' dwellings the median level is appr. 5 dB higher at frequencies 250 Hz. This difference is related to the difference in cases' and controls' daytime L_Ceq distributions (figure 1).

Criteria. The results may be compared to the criteria according to DIN, Socialstyrelsen and Vercammen. The ISO threshold that is used here as a reference curve for audibility, serves as a fourth criterion. Only nighttime recordings where the LF sound indeed is heard are considered.

The average scores given in tabel 2 are the (arithmetic) average differences between the values determined according to a criterion and the criterion value. For the DIN criterion this is the difference between 25 dB(A) and the sum of all 1/3 octave band levels with 10 f_{1/3 octave} 100 Hz; for the other criteria it is the distance of the maximum 1/3 octave band level to the criterion.

All scores in tabel 2 are below the criterion values, except for the ISO threshold. This regards the average scores: in fact, in 10 to 25% of the dwellings sound levels do exceed the criterion values, again except for the ISO threshold that is exceeded in 60 to 70% of the dwellings. There is little difference between cases and controls, although controls tend to have higher average scores.

Correlation coefficients of the scores on the four criteria (Vercammen, Socialstyrelsen, DIN, ISO threshold) with L_Ceq are within 2% of 0.92.

CLASSIFICATION OF NIGHTTIME LF SOUND LEVELS

The measured nighttime sound levels in complainants' dwellings can be classified according to the scores on criteria for LF noise. Also, they can be classified according to the presence of audible LF tonal components. This was implemented by identifying in the 1/3 octave band spectra local maxima (relative to the ISO threshold) with a sound level above or not less than 5 dB below the ISO threshold. Then, in narrow band spectra the frequency and frequency bandwidth of peaks causing the maxima were identified. They could be divided in two classes: narrow peaks of less than appr. 3 Hz bandwidth ('tones') and broader peaks with bandwidths of appr. 3 to 10 Hz ('broad tones').

 

'no' LF sound. Only nighttime measurements in complainants' dwellings have been considered.

• In some cases (15%: 3 out of 19 complainants' dwellings) there is 'considerable' LF sound: all criteria are exceeded, the ISO threshold with 13 to 18 dB. The average sound level in these cases is 57 dB(C), 34 dB(A). Even according to standard (non LF) Dutch limits the sound level would be considered high. It is noteworthy that in two of these three cases the sound source was unknown.
• In most cases (60%) there is 'some' LF sound: there is a LF sound above or just below the ISO threshold (excess 11 to -3 dB). No or just one other criterion is (just) exceeded. The average sound level in these dwellings is 42 dB(C) and 24 dB(A). In 3 of these 11 cases there is no clear relation between the spectral levels and the reported audibility.
• Sometimes (25%) 'no' LF sound can be identified dat reasonably can explain LF noise complaints: the LF sound level is well (> 6 dB) below the ISO threshold, no criterion is exceeded and no obvious spectral component can be identified. The average sound level in these 5 dwellings is 36 dB(C) and 21 dB(A). In some recordings the sound level is closer to the ISO threshold, but the LF sound was perceptible to the complainant also at a time when a lower level was measured. The lower level then was taken as decisive.

From narrow band analysis it can be concluded that in 5 dwellings with 'considerable' or 'some' LF sound there is tonal sound (appr. 48 Hz and once 100 Hz; plus harmonics), in 5 dwellings there is sound in a relatively narrow frequency band ('broad tones' at 35-40 and 75-80 Hz, in one case at appr. 155 Hz). In three cases the LF sound is non-tonal. However, although in most cases (75%) a LF sound above or just below the ISO threshold could be demonstrated, there is no proof that that sound indeed was the cause of complaints.

In dwellings with 'no' LF sound the silence, the relative absence of (LF) sound, seems characteristic and possibly a factor related to the complaints. Maybe some of these complainants hear very well or there is a lack of indoor sound masking body sounds such as blood flow or LF tinnitus.
There is no obvious relation between the classification given and the number of complainants in a household. In cases were 'some' or 'no' LF sound can be demonstrated either both or just one of the adults present may claim to perceive a LF sound.

DISCUSSION AND CONCLUSIONS

In most cases of complaints about LF noise a LF sound can be identified that may explain complaints. It is as yet not proven that these LF sounds are indeed the cause of complaints. In most cases this sound is not obvious to others (e.g. a house mate or an investigator) and the source is unknown -even to the complainant. The measured LF sounds have frequency components related to the electric power frequency (50 Hz; 47 - 49 Hz from asynchronic motors) and the number of revolutions of car engines (appr. 2000 - 3000 min^-1 = 33 - 50 Hz) and their harmonics. Thus at least some of the 'sounds of silence' appear to be the sounds of engines.

In some cases no LF sound can be demonstrated that can reasonably explain LF noise complaints. The cause of complaints then cannot be explained by indoor LF sound, unless (in some cases) the complainant hears extremely well.

On average the dwellings of complainants are more quiet than dwellings without complaints (controls), especially at daytime and at low frequencies. As a consequence complainants score less (1 - 4 dB) on proposed LF noise criteria. In most cases these criteria are not exceeded. The criteria therefore have no simple relation to complaints: complaints are not always accompanied by high scores (most cases) and high scores need not be accompanied by complaints (some controls).

In view of the differences in sound levels between complainants' and controls' dwellings, complaints probably are influenced by other factors. For some a persistent LF sound attracts an attention that eventually leads to serious complaints, in contrast to others (house mates, controls) who apparently do not hear or notice it. This may correspond to the fact that complainants consider themselves sensitive to sound.

It is possible that a slight sound is more striking in complainants' dwellings because there is less masking sound. This however would apply mostly during the day when sound levels in cases' dwellings are lower than those in controls' dwellings, whereas complaints usually arise out of a lack of sleep at night.
 
 

REFERENCES

1. C. Gielkens-Sijstermans, T.H. Collijn, A.W. Jongmans-Liedekerken: Sensitivity for low frequency sounds: a study into possible factors (Gevoeligheid voor laagfrequent geluid; een studie naar mogelijke factoren), GGD Oostelijk Zuid-Limburg, 1998 (in Dutch).
   An abstract is published as: A.W. Jongmans-Liedekerken, C. Gielkens-Sijstermans: Susceptibility to low frequency noise, proceedings ISEE/ISEA'99, Athens
2. G.P. van den Berg, P.W.G. Altena, R.R. Nederhoed: 'Silent sound': low frequency sound in dwellings ('Stil geluid': laagfrequent geluid in woningen), Natuurkundewinkel RuG (Rapport NWU-83), ISBN 90 367 10 464, 1999 (in Dutch)
3. G.P. van den Berg: Sound exposure measurements in cases of low frequency noise complaints, proc. Internoise 1998, Christchurch
4. G.P. van den Berg, W. Passchier-Vermeer: Assessment of low frequency noise complaints, proc. Internoise 1999, Fort Lauderdale
5. NSG: Nederlandse Stichting Geluidhinder: Guideline for Low Frequency Noise (Richtlijn Laag-frequent Geluid), Delft 1999 (in Dutch)
6. ISO: ISO 226:Acoustics-Equal Loudness contours for otologically normal listeners, part 1
7. W. Passchier-Vermeer: Assessment of low frequency noise in dwellings (Beoordeling laagfrequent geluid in woningen), TNO Preventie en Gezondheid (rapport 98.028), 1998 (in Dutch)
8. M.L.S. Vercammen: Low Frequency Noise Limits, J Low Freq Noise Vib 11 no.1 pp. 7-13, 1992
9. DIN: Messung und Bewertung tieffrequenter Geräuschimmissionen in der Nachbarschaft, Deutsches Institut für Normierung (DIN 45680), Berlijn, 1997
10. Socialstyrelsen: Indoor Noise and High Sound-Levels: General Guidelines issued by the Swedish National Board of Health and Welfare (Socialstyrelsen), 1996