If you understand, for example, what LCeq10 is you can probably skip this primer on noise. If you don’t you ought to read through this, otherwise you won’t fully appreciate the nature of the issues. I won’t go into that much detail, but some understanding of basic principles is needed to fully appreciate how noise regulations are structured, and where their weaknesses might be. My discussions, being reality and evidence based necessarily become somewhat technical.
Decibels, or dB, are used in a variety of scientific and engineering fields and at their most basic describe any phenomenon where there’s a logarithmic relationship between signals. The two main fields that we laypeople come across them is in electricity (i.e. remember signal-to-noise ratios in stereo stuff?) and sound, but everything I talk about will be sound-related. The “deci” in decibels indicates the base for the logarithm is 10 – in simpler terms, every increase of 10 db means the signal has increased by a factor of 10. A result is an increase of 3 db means the signal has roughly doubled. A complication is that decibels are by convention generally used to describe relative power levels. Unfortunately the response of humans to sound is more accurately described by pressure level differences, not power level differences. Luckily, the power level ratio can be simply squared-rooted to obtain the pressure level ratio. Translating that back to decibels, it now takes 20 db to get a 10-fold increase in the pressure level, and a 6 db change in levels represents a doubling. When people talk about different db noise levels, they are almost always talking about pressure levels.
In practical terms, here are some everyday samples of how loud things are – these are all sound pressure levels (SPL’s).
|Examples of Everyday Sound Levels|
|Weakest sound heard||0dB|
|Whisper quiet library||30dB|
|A Quiet Bedroom||25-30dB|
|Kitchen with refrigerator running||40dB|
|Quiet village street||45-50dB|
|Normal conversation (3-5′)||60-70dB|
|Telephone dial tone||80dB|
|City traffic (inside car)||85dB|
|Train whistle at 500′, Truck Traffic||90dB|
|Subway train at 200′||95dB|
|Power mower at 3′||107dB|
|Power saw at 3′||110dB|
|Sandblasting, Loud Rock Concert||115dB|
Just as important as absolute sound level is the variability of a sound. I mentioned above that 6 dB represents a doubling of a sound’s level, but notice that our human perception is that a sound doubles with a 10 dB increase.
|Perceptions of Increases in Decibel Level|
|Barely Perceptible Change||3dB|
|Clearly Noticeable Change||5dB|
|About Twice as Loud||10dB|
|About Four Times as Loud||20dB|
The primary measure of sound is the decibel but a decibel by itself, while useful, is not enough to reflect the impact of a sound upon a person. The duration and variability of the sound is often more important than the level by itself. Contained within the literature are any number of different time spans along with techniques to reduce the naturally-occurring variability within that span to one number that reasonably represents the sound level. For wind turbines the two most important measures are L90 and Leq, where “L” represents a decibel level over some period of time. In addition to the “90” or the “eq” the “L” is often supplemented with an “A” or a “C” to indicate which weighting is being used. Also added might be an indication of the time period, usually in minutes. Thus LAeq30 would be the A-weighted average over a 30 minute period.
Since many noise regulations specify noise levels above the ambient, it becomes important to be able to specify what the ambient level is. L90 is generally used to do this. A sound level meter takes some number of samples during a period of time, typically several samples per second. The L90 result is the sound level above which 90% of those samples lie. This technique results in throwing away all transient noises, leaving only the softest 10%. Professional sound meters automatically calculate this. For us amateurs, we could put the samples into a spreadsheet, sort them into descending order and then slide down to the value 10% from the bottom.
Many noise regulations specify average noise levels, and so it becomes important to be able to specify one number that represents that average. Given the logarithmic nature of the decibel scale, we just can’t add the samples and divide by the number of samples. To do so would generate a number that would not accurately reflect intermittent loud sounds. Having 100 samples consisting of 90 30-db and 10 120-db is not equivalent to 39 db. Instead the samples are converted back from their logarithmic-based values to the raw sound levels, averaged, and then converted back into decibels. As before, professional sound meters calculate this automatically, and a spreadsheet can also be used to do this computation. For the example above, the result is 110 db. As a further more extreme example, just 1 120-db sample and 99 30-db samples generates an Leq of 100 db.
Dispersion refers to how a sound spreads out in the atmosphere. We all know that a sound’s intensity decreases as we get further from its source, and this section attempts to explain the principles behind that. Because sound is necessarily studied in a real physical environment it becomes very complicated to describe all its behaviors. I’ll stick to the major (sometimes perfect-world) points, as always focusing on what I think is important to understanding wind turbines’ characteristics.
The intensity of any emission will dissipate as the distance from the source increases. The most common expression of this is that the intensity of sound (or light, for that matter) goes down as the square of the distance: in other words, if you double your distance the sound will decrease to a fourth of what it was. A slightly more accurate expression would be that sound goes down as one less power than the number of dimensions it is spread out over. In the above example the sound is presumed to originate at a point and spread out over 3 dimensions. Thus the decrease is calculated as a power of 2, or squared. If the sound spreads out evenly in all 3 dimensions this would be called “spherical”. Here are some other examples.
- Example #1. A laser goes out in a straight line – only one dimension. Applying our rule, it dissipates not at all. This assumes a perfect vacuum and rod, never practical possibilities.
- Example #2. The sound is not a point source, but rather a line source – like a long violin string (or more commonly, a road). The long string takes away one of the dimensions, so the sound can only disperse over 2 dimensions. This leads to linear dispersion, where if you double your distance the sound decreases by half.
- Example #3. The sound can’t spread in 3 dimensions – like there’s a temperature inversion that prevents the sound from spreading upward, thus providing only 2 dimensions for the spreading. This is “cylindrical” dispersion and leads to linear decreases.
- Example #4. More familar to us is where the source is point on the ground. The sound still spreads in 3 dimensions, but only through half the sphere, so this would be called “hemispherical”. The “dimensions minus one” rule still applies, so the sound would dissipate as the square of the distance. But ground absorbs (or doesn’t) the sound as it spreads. Now everything gets trickier. If the ground absorbed all the sound, you’d be left with the distance-squared behavior of the perfect sphere. If it absorbed none of the sound, the sound would still dissipate as the square of the distance, but would be a constant 6 dB higher.
As example #4 starts to show, the behavior of sound in the real atmosphere with real grass and trees around is not so simple to calculate. In addition the atmosphere itself absorbs some of the sound energy as heat, so the dissipation is somewhat higher than my perfect-world examples have implied. And I haven’t even gotten into reflections and obstructions. Generally speaking, there exist computer modeling programs that can take into account most of these factors.
As example #2 shows, where there is a sound that is not a point source (and in reality there is never a true point source) the sound will not dissipate as quickly as you might otherwise expect. The computational complexity of handling arbitrary non-point sources is so high that for most purposes point (or a somewhat-arbitrary almost-a-point) sources are assumed, and many times the results are good enough for government work. I’ll get into more details in the Regulation Problems section, but for now just imagine what might result when the point source is really a disk almost 100m in diameter, and you are only 350m away from it.
As most of us know, the human ear can at best detect sounds roughly in the 20hz to 20,000hz range. We also know that as we get older the range decreases, especially at the high end. Acoustic engineers have long measured the human response to different frequencies and know that it is not linear, even within the range of what we can detect. Generally speaking, the ear is most responsive to sounds in the our mid-range, roughly 3000-6000hz. The equipment (assuming it’s high quality) used to measure sound can do so in a very linear fashion. To make what it measures more indicative of what we hear frequency weights are applied. The two most important are “A” and “C” weights, which are shown below, along with the B and D weights.
Common Frequency Weights
The most widely used weighting is the “A” scale, shown in green above. When it is applied to either decibels or sound levels, an “A” is appended, i.e. dBa, or LA. It’s purpose is to de-emphasize those frequencies that are not so well detected by the average human ear. As you can see from the curves above, a major part of what it does is decrease low-frequency sounds. In the second chart you can get an idea of what kinds of sounds a turbine actually produces. The major reason the A scale is used seems to be historical. It was originally designed for use at low sound levels, but is now used for more general purposes – including annoyance measurements, where it is not appropriate at all.
What Turbines Produce
As you can see, wind turbines produce a lot of low frequency noise, which is de-emphasized by the A scale. Bill Palmer has written a very readable explanation of these differences. The World Health Organization recognizes this failing, and recommends using the C scale whenever there is a lot of low-frequency sound present. As you can see, the C scale above does not de-emphasize low frequencies nearly as much as the A scale does.