Bell, Dr A. NHMRC Draft Information Paper Submission

The literature that the Information Paper relies on to make its conclusions on noise from windfarms can be characterised as scientifically incomplete and misleading. 

This submission is by Dr Andrew Bell, a Visiting Fellow at the John Curtin School of Medical Research who has undertaken decades of research into the mechanisms of hearing, particularly the response of the cochlea to sound pressure.

Dr Bell’s qualifications include an MSc in auditory science and a PhD in cochlear mechanics, both from the Australian National University, with an initial training in physics (BSc from the University of Melbourne). He is member of the Australian Acoustics Society and the Acoustical Society of America, and has published in both their journals. Since 1992 Dr Bell has published peer-reviewed papers on spontaneous otoacoustic emissions from the human cochlea and on the fundamentals of hearing. His research has focused on the crucial role that acoustic pressure plays in stimulating the cochlea, and he has extended that work to include considerations of how pressure is also a vital parameter for understanding the role of the middle ear muscles. Dr Bell is currently the Consulting Editor for the Journal of Hearing Science and is involved with research collaboration with the Polish Institute of Physiology and Pathology of Hearing, one of the largest clinical and research centres in Europe.


The literature that the Information Paper relies on to make its conclusions on noise from windfarms can be characterised as scientifically incomplete and misleading. Section 6.1 on Noise contains incorrect statements based on faulty inferences, and therefore the conclusions reached are incorrect. This submission describes these errors and indicates a way forward to resolve them.

This submission focuses on the annoyance from windfarms and provides an explanation in terms of the constructive interference of tonal infrasound from synchronised turbines operating within a wind farm. The reasoning is set out below.


The Paper does not specifically define the meaning of infrasound. It chooses to classify the sound emitted from a wind turbine as falling in the audible frequency range 20–30 Hz and 200–1000 Hz (paragraph 5 of Section 6.1). The Paper refers to ‘infrasound’ in paragraph 7, and one must infer that frequencies below 20 Hz are being referred to. It is true that, as stated in para 7, “infrasound is considered by some to be an important component of noise from wind farms”, and in this context I draw the attention of the panel to the recent work of Thorne (2013) and Cummings (2013) who point out how constructive interference from synchronised wind turbines can lead to ‘heightened noise zones’. In this submission I extend this concept to infrasound, where, because of the much longer wavelengths involved (a tone of 1 Hz has a wavelength of 330 metres), there is much greater potential for causing problems.

Annoyance from infrasound not adequately addressed by the South Australian study

The Paper downgrades the importance of infrasound in producing annoyance in people living nearby by relying too strongly on an inadequate and flawed study. The Paper states that “Evidence suggests that levels of infrasound are no higher in environments near wind turbines than in a range of other environments,” a statement supported by reference to “a South Australian study” (Evans et al. 2013), on which the NHMRC strongly relies for its original Systematic Review. In the following, I point out major inconsistencies in Evans et al.

Ignoring strong tonal infrasound at blade passing frequency

Evans et al. measured infrasound emitted from two windfarms, and the Bluff Wind Farm showed particularly interesting results. The researchers found that at Bluff Wind Farm the infrasound contained clear peaks at the blade pass frequency of 0.8 Hz and its harmonics at 1.6 Hz and 2.5 Hz, and the authors state, correctly, that these spectral peaks are a characteristic signature of the revolving blades of a wind turbine. The peaks can be clearly seen in Figure 29 of their work. They are also visible in Figures C3–C8, and in one figure for Clements Gap (lower trace of Figure C9).

I want to emphasise an important but generally unappreciated property of these peaks. First, they have a tonal quality because the turbine blades in a wind farm are, for reasons of generation efficiency, regulated by a feedback circuit to maintain a fairly constant rotational speed of 15–17 rpm, irrespective of wind speed (16 rpm = 3.75 sec/rev = 1.25 sec/rev for 3 blades = 0.8 Hz.) Moreover, all the generators in a wind farm are turning at this same speed, constituting a coherent source of infrasound energy. In other words, with the generators in synchrony, the sound emitted by each generator will bear a constant phase relationship to the others. In this circumstance, it is inevitable that the waves will constructively interfere at various points (and destructively interfere at others).

Some researchers have already pointed to the fact that wind turbines tend to operate in perfect synchrony, even if they initially are out of synchrony [McSwiggan et al. (2008); Cidras (2002); Katayama et al. (2006)]. Some have also pointed to the possibility of the sound from these multiple sources being coherent [Thorne (2003); Cummings (2013)] due to entrainment (the phenomenon observed by Huyghens whereby two small pendulum clocks attached to the same wall come into synchrony). However, this low frequency synchrony and its implications are unappreciated by the Information Paper. If the sources add together coherently, the sound can be much louder, and its intensity fall less with distance, than predicted by assuming the sound is incoherent. For example, two wind turbines would have an intensity 6 dB louder than a single turbine, and four could have an intensity 12 dB louder. When one considers that a single wind farm may have more than 100 wind generators, the possibilities become serious indeed. Moreover, the array becomes a line source rather than a point source, so that the intensity falls off much more slowly with distance (1/r rather than 1/r2). The problem is exacerbated by the wavelength of a pure 0.8 Hz sound wave being about 400 metres, which is about the same distance that wind generators tend to be spaced (about 5 hub heights, which is 5 × 80 m = 400 m). This last factor has only recently come to be recognised [Thorne (2013)]. Thorne uses the concept of “Heightened Noise Zones” (HNZs) to refer to nodes of constructive interference that occur when turbines operate in synchrony. Although Thorne makes reference to the blade passing frequency, his concern is with low frequency sound, not infrasound, and he uses frequencies of 20, 48, and 66 Hz in his simulations of possible interference patterns. If we consider infrasound of 1.6 Hz (200 m wavelength), the effects seen by Thorne would be even more pronounced, yet to this researcher’s knowledge this idea of infrasound coherence does not seem to have been considered in the windfarm literature.

Implications of constructive interference at 1.6 Hz

The implications of constructive interference at 1.6 Hz are profound. It means that infrasound pressure at the blade passing frequency (and its harmonics) could be extremely high and, because the attenuation of infrasound by air is extremely low, and because infrasound will only fall off as 1/r if coming from a line source, the pressure wave could propagate many kilometres. Of course, it is also possible that coherent sources could destructively interfere, but given the ‘quasi-crystalline’ arrangement of wind turbines in a farm, it will give rise to an alternating pattern of constructive and destructive points (nodes and antinodes). In terms of impact on local residents, the pattern of nodes and antinodes will depend on location, and will shift with wind, temperature, and atmospheric stability. There will also be fluctuations in phase on the scale of minutes due to the blades randomly slipping in and out of synchrony, which will mean, at any given location, jumps in phase and shifts from a node condition to an antinode.

Such sudden phase shifts tally with the descriptions of people living near a windfarm who complain of sounds like “sneakers in a drier” or a “jet that never lands” [Cummings (2013)]. Other terms listed by Cummings include thumping, pounding, roaring, and rumbling. These type of descriptions match the perceptions that would be generated by low frequency coherent sources which have fluctuating phases.

Broadband infrasound measurements ignore tonal components

If the coherent infrasound hypothesis is true, then the method of measuring the broadband sound pressure level over 0.15–20 Hz as done by Evans et al. is inappropriate for infrasound audibility. The authors measured the broadband infrasound as about 50–60 dB SPL and then compared it to natural sources over the same band. Their conclusion, endorsed by both the Systematic Review and the Information Paper, is that there were “similar levels of infrasound at rural locations close to wind turbines, rural locations away from wind turbines, and at a number of urban locations” (Section 6.1, Paragraph 7). The inference is that the wind turbines will not be heard due to “masking” (Evans et al., p.37), but this is not necessarily the case. It is similar to claiming that the sound of a symphony orchestra, averaged over 10 minutes and across 20–20,000 Hz, is about 85 dB, whereas the sound of a triangle is only 60 dB, so therefore the triangle will not be audible.

The tonal signature of a triangle can be distinctly heard above the surrounding instruments, and the reason is the distinctive frequency and narrow bandwidth of the triangle’s sound. The same argument, I suggest, can be put about hearing a tonal 0.8 Hz infrasound among surrounding broadband noise. It may not be heard as a pure tone because of the way the ear hears, perhaps being heard as a modulation of all the other ambient sounds. Nevertheless, it is quite possible that the effect is not “inaudible”, since experiments done in acoustic chambers with pure tones have severe limitations, as discussed in the penultimate section. The suggestion then is that annoyance may not be related to the broadband G-weighted level of infrasound between 0.8 and 20 Hz; rather, the annoyance may be related more to the prominence of tonal infrasound at blade-passing frequencies. With this perspective, lower G-weighted levels may be more annoying than higher ones if the tonal component dominates.

Action of middle ear muscles explains why tonal infrasound is a problem

There is the possibility that other auditory mechanisms, involving the middle-ear muscles, may be at work in detecting low-frequency sound (Bell 2014). This work gives credence to the complaints of residents living next to windfarms, suggesting their reports should be given greater credence when weighed against acoustic measurements that may not be valid or appropriate.

The ear is more than a microphone. The cochlea and middle ear work together to facilitate hearing. The middle ear is the “gateway” to the astoundingly sensitive cochlea, and it must quickly and silently control the gain of the sound entering the cochlea. This researcher has set out in a peer-reviewed paper (Bell 2012) how the middle ear muscles play a vital role in hearing, and the important factor at work is absolute pressure. The middle ear muscles act as a “gain setting” mechanism for the cochlea, positioning the ear drum at the mid-point of its operating characteristic by adjusting the pressure in the fluids filling the cochlea. The cochlea operates over a dynamic range of 120 dB (a million million times), meaning that, in terms of signal detection, some gain control is necessary. This paper provides an understanding of how a sudden low-frequency impulse, with absolute amplitude measured at perhaps 60–80 dB (lin) will disturb the set point of the ear drum. The ear drum reacts linearly to sound pressure, irrespective of frequency, so that a large disturbance will cause the middle ear muscles to react and “hunt” for a new set point. In other words although a 1 Hz wave of 60 dB SPL may be inaudible in terms of the cochlea, it will still have appreciable effects on the ear drum. The amplitude of that 60 dB 1 Hz wave (in terms of pressure) will have the same amplitude as a 60 dB sound of 1 kHz, even though the 1 Hz wave is below audibility. The cochlea may not be giving a response at that low frequency, but the ear drum and the attached middle ear muscles will be working just as hard to maintain a comfortable set-point.

Having the middle ear muscles continually active is likely to lead to annoyance and sleep disturbance. It explains why people near wind farms feel “pressure” in their ears, and why no one finds it pleasant to travel in a car at speed with the windows down (which generates large infrasound signals). Certainly few people would choose to sleep in a car under such conditions, nor would they find it easy to sleep near another generator of infrasound, a windfarm.

Evidence that measurements by Evans et al. were inappropriate

Let us now look at specific evidence in Evans et al. indicating that their measurements may not have been appropriate or their logic strictly true.

A. No difference between ‘on’ and ‘off’

First, on p. 37 there is a curious anomaly: in order to determine whether the Bluff Wind Farm was generating infrasound, the researchers arranged for the wind turbines to be switched off. As a result, they found that [para 4, p.37]: “At Location 8 near the Bluff Wind Farm (Figure 29), the [0.8 Hz, 1.6 Hz, and 2.5 Hz] peaks were detected at a similar level during both operational and shutdown periods. Indeed, at 1.6 Hz in Figure 29 there is less than a 3 dB variation between all measurements. This seems very strange, given that the peaks are distinctive signatures of wind turbines. However, because the levels did not change when the wind turbines were shut down, Evans et al. came to the conclusion that the peaks must not have been due to the Bluff Wind Farm after all. In other words, the convergence of all the 1.6 Hz measurements must just have been due to “an extraneous source” (p. 58). What is this extraneous source? The likelihood is that it was another distant windfarm (North Brown Hill) and that L8 happened to be at an infrasound node.

The authors note (p.37) that there is the possibility that “the peaks in the spectrum during the shutdown resulted from operation of North Brown Hill Wind Farm” which was 8 km from the measurement location and which was “very faintly audible” during the shutdown. But instead of trusting the evidence of their ears, they chose to believe that sound pressure levels must decrease due to the shutdown and discounted that infrasound would propagate 8 km. Here it is worth emphasising that the authors claim, on the basis of their measurements, that the wind farm is inaudible and cannot cause problems, while at the same time noting the evidence of their ears that they could in fact hear a wind generator 8 km away. What they were not expecting is that an array of wind generators 8 km away could generate substantially more infrasound (at 1.6 Hz) than a generator (or array of generators) some 1.5 km away. Naturally, the annoyance of a wind generator increases appreciably at night against a quiet background, as many researchers have noted (Thorne 2013). It is peculiar that scientific evidence, subject to mistaken interpretation, can be taken as more convincing than the evidence of a person’s own ears.

Similarly, Figures C3–C9 show the same signature peaks at 0.8 Hz, 1.6 Hz, and 2.5 Hz at two locations. Again, Evans et al. overlook their significance in simply noting that the peaks were often there whether the windfarms were operating or not, and infer (p.58) that they must be due to “an extraneous source (which was also present during the shutdown periods)”. Furthermore, it is significant that in Figure C3, which shows measurements for low wind speeds (0 to 3 m/s, when wind turbines do not rotate), the extraneous source and its distinctive peaks at 0.8, 1.6, and 2.5 Hz is still present at L8, once more probably due to a more distant windfarm at which the wind was above 3 m/s.

B. Country more noisy than city at 1.6Hz

There is another line of evidence in Evans et al. which points to how a phased array of wind generators, with their infrasound tones, is likely to cause more disturbance than broadband noise. In Figure 31 (p.39), all the 1.6 Hz measurements, both in the country and in the city, are plotted together. The 5 country locations, all near windfarms, can be seen to have an average loudness at 1.6 Hz of 57 dB. In comparison, 11 city locations are seen to have an average loudness at this frequency of 52 dB. In other words, the country locations are 5 dB louder at this frequency than the 11 city ones. (In this comparison, it was considered prudent not to include the L11 location at Myponga because the inside and outside figures differ by 14 dB, which is anomalous when the indoor and outdoor figures for the country sites are the same; nevertheless, even if the L11 figures are included, the country locations are still louder.) On this measure, it is clear that the country locations and their windfarms are appreciably louder than the city ones. In a quiet rural setting, this extra infrasonic tone might be heard as intrusive and disturbing. The equivalent G-weighted broadband figures (Figure 1, p.iv) therefore miss the crucial disturbing signal.

Towards a better approach to the problem

The above considerations give strong indications that wind farm noise is annoying and cannot be ignored. Unlike eyes, which have eyelids to shut out visual input, the auditory system cannot be turned off. In the middle of the night, the regular large amplitude pressure peaks generated at 0.8, 1.6, and 2.5 Hz by an array of wind generators will impinge on the ear drum and activate the middle ear muscles. It is very difficult to sleep with a tap dripping once per second, and it could be equally difficult to sleep with an infrasound pulse periodically activating the middle ear muscles, even if it cannot be “heard”. The explanation – constant middle ear muscle activity – is similar to why people find it difficult to sleep in a moving car with the window down.

Using inappropriate measurements, like using A or G weightings of broadband noise, does not make the signal go away, even if it makes the tonal signal seem small in comparison with the broadband background. For the middle ear muscles, which operate as a gain control mechanism to keep the ear drum and the cochlea at an optimum operating point, the most relevant measure is the absolute sound pressure level (plain decibels without any weighting).

If the human ear can hear wind turbine noise, while the measurements say the noise is inaudible, then better measurement techniques are clearly needed. The Information Paper’s position – that because instruments do not measure anything above ambient background levels then the sound cannot be heard – is based on inappropriate measures and incorrect interpretation. It ignores the primary evidence of our senses (including in some cases the ears of the experimenters). In this way,

…the statement fails to adequately address the reports of people who live close to windfarms and who are troubled by what they hear. Authorities and the community of scientists should listen to them and find the source of the problem. Dismissing the reports as contrived is not the basis of good science or good policy.

It is likely that the apparent paradox between our ears and our instruments can be resolved by recognising the tonal nature of infrasound at the blade passing frequency (0.8 Hz) and its harmonics (1.6 Hz and 2.5 Hz). There need to be more measurements made at these frequencies, and at far greater distances from the wind turbines, up to tens of kilometres. Measurements need to be made at multiple points so that a map of infrasound nodes and antinodes under various conditions can be made. Some theoretical studies of the interference patterns caused by the quasi-crystalline arrangement of wind turbines, and the effect of various meteorological factors, would also provide more insight. A relatively simple solution the problem of windfarm noise exists: ensure that the blades never operate in synchrony.

It is clear that infrasound hearing thresholds have been inadequately determined.

Measurements made in compression chambers using pure tones do not reflect the windfarm situation in which there are multiple infrasound sources superimposed on higher frequency sounds. Windfarms generate tonal infrasound (most strongly at 1.6 Hz, with a wavelength of 200 m) from many discrete sources, which interfere with varying phase relationships. The phases depend on meteorological factors and the degree of synchronisation of the turbines, and these effects can be expected to vary from moment to moment. Rapidly fluctuating phase is a likely explanation for the “boots in the dryer” perception, and is a situation that has not been tested in pressure chambers.

Concluding remarks

I thank the review panel for the opportunity to comment on this important issue, and trust that these comments will prompt the panel to reconsider some of their scientifically unsupportable statements, notably para 7 of Section 6.1 discussed above. In addition, in the Summary in Section 7.2.3, the following also require modification.

Bullet point 4 in Section 7.2.3 needs correction: wind farm noise can be heard at more than 500–1500 m, particularly at night. This point should be modified in the light of the review’s recommendation on p.20 calling for more research into “wind turbine signature” and for field measurements “ranging from 500 m to 3 km and beyond”.

Bullet point 5 in Section 7.2.3 needs correction: infrasound differs from other natural sources in having strong tonal components at 0.8, 1.6, and 2.5 Hz. Its effects, including annoyance and disturbed sleep, should be studied by simulated wind turbine noise generated by a speaker, as recommended by the panel on p. 20. This researcher suggests that theoretical modelling of the phase coherence of wind farms at blade passing frequencies would also be appropriate. Further study of the cochlea’s input gain control mechanism – the middle ear muscle system – would also extend our understanding of how infrasound is heard.


Bell, A. (2012). How do middle ear muscles protect the cochlea? Reconsideration of the intralabyrinthine pressure theory. Journal of Hearing Science 1(2), 9–23.

Bell, A (2014). Annoyance from wind turbines: role of the middle ear muscles. Letter to the Editor. Acoustics Australia 42, 60.

Cidras, J (2002). Synchronization of asynchronous wind turbines. IEEE Tansactions on Power Systems 17, 1162-1169.

Cummings, J (2013). The variability factor in wind turbine noise. Fifth Intl Conference on Wind Turbine Noise. [Available online at ]

Evans, T, Cooper, J & Lenchine, V (2013). Infrasound level near windfarms and in other environments. Report by EPA South Australia and Resonate Acoustics.

Katayama, N, Takata, G, Miyake, M & Nanahara, T (2006) Theoretical study on synchronization phenomena of wind turbines in a wind farm. Electrical Engineering in Japan 155: 9–18.

McSwiggan, D, Litttler, T, Morrow, DJ & Kennedy, J (2008). A study of tower shadow effect on fixed-speed wind turbines. Universities Power Engineering Conference, Padova.

Thorne, R (2013) Wind farm noise and human perception: a review. Noise Measurement Services, Enoggera. [Available online at ]

Download the submission