The most complete measurement of wind turbine sound I've found so far is from the Mars Hill Wind Farm. When built it was the largest wind farm in New England, with 28 1.5 MW turbines. It has also been one of the most contentious, with many complaints from the mis-led neighbors. In response to the complaints, an independent consultant was hired to measure actual noise levels during normal operations. The "2nd Quarter" report was published on November 2, 2007, and most of my numbers are from that report. The "3rd Quarter" report was published April 11, 2008 and didn't reveal any important new facts.

The reports are long and technical, and they are on both this site and the original marshillwind.com site. To me the following two tables, one from November 2007 and the updated table from April 2008, are useful in compressing the entire reports into one page. These charts list the different testing locations and a summary of the noise levels found at each. The second column provides the peak noise levels that were attributed to wind turbine operations. Maine, in its infinite wisdom, allowed the Mars Hill farm to produce 50dbA at a neighbor's home, a full 10dbA above Ontario's limit. One site, MP-1, is above even that very generous limit. The owners at MP-1 must be miserable. But most noise regulations and recommendations have a 40dbA max, and almost all of the locations experience noise above that. Also notice that the pre-construction model estimates are quite far under the actual results.

To get some sense of how far away from the turbines you have to get to down to 40dbA, I've sorted the rows by distance, obtained a few additional numbers, and eliminated a couple of locations where the 2ndQ numbers looked invalid (one meter looked like it was inconsistently high, and another location suffered from a lawn mower), leaving six valid-appearing locations. It is unclear if the effects of the topology of Mars Hill would be better or worse than Amherst Island, Amherst being much flatter than Mars Hill. Also note that the turbines for AI are almost 50% larger than those at Mars Hill, and there are potentially more of them. One of the purposes of this study was to compare the pre-operational computed levels with the actuals, and the last column provides the original estimate, which was always under the measured result. This is true even though the models are supposed to represent the "worst case" and the measured results were the average of a 24-hour period.

Location	Distance(m) to nearest turbine	LAeq average	LA1 average	Comments	Non-farm noise level	Modelled Estimate
MP-1	260	54.3	58.9	Upwind	40	51
MP-8	390	49.4	53.9	Sideways	39	47.5
MP-6A	390	46.2	50.5	Downwind	33	42
MP-7A	800	44.8	46.9	Downwind	32	41
MP-4A	1050	43.6	48.8	Upwind	34	37
MP-2	1900	39.7	47.1	Upwind	30	35

I included the LA1 averages because the WHO Guidelines recommend its use as more representative of human disturbance for the type of noise that wind turbines produce.

Also note the deltas between the non-farm noise level and the LAeq levels. That is how much noise, on a fairly constant basis, is added to the background level. Given that every 6db is a doubling of sound levels, the closest site had it's background quadrupled, while even the furthest had its background level more than doubled. The experience at Mars Hill provides a fair amount of support for a minimum setback of 1500 meters.

The Mars Hill study came off of marshillwind.com, the site maintained by the energy company, and can be obtained from there. I also have both the 2nd Quarter Report, 1.7mb and the 3rd Quarter Report, 9.0mb available from this site.

Here's an interesting comment letter concerning how the modeling was done.

You'd think with all this noise someone would be complaining, and maybe there'd be health impacts too.