Page 24 - Winter2021

AS THE WIND BLOWS
Figure 5. A range of commonly used microphone windscreen types. Illustrated by Jessica Reeves.
lower frequencies. Despite their use, accurate characteriza- tion of the acoustic environment is challenging, especially in windy conditions. Recordings are often censored when the mean wind speed at the microphone height is greater than a predetermined threshold (Mennitt et al., 2014).
In the motion picture industry, windscreens came into routine use once productions included sound recording outdoors. Early forms of “wind gags” were spherical or streamlined frames covered with one or more layers of fine-meshed muslin or silk (Figure 6) (Clark, 1931). The enclosed volume of air acts as a pressure chamber that maintains a high correlation between turbulent fluctua- tions entering the inlets of a pressure-gradient transducer (Wuttke, 1992). Due to their directional characteristics, these types of transducers are typically used in sound- recording practice.
Although the human eardrum is naturally shielded from wind noise by the pinna and ear canal, wind noise can be a significant problem for users of digital hearing aids. Unmit- igated, wind noise can saturate hearing aid microphones, limit intelligibility of speech, and make outdoor sounds inaudible. Because windscreens are not feasible, wind noise mitigation has focused on digital signal processing. On a single microphone, wind noise may be detected from its statistical properties. In two-microphone algorithms, the relative incoherence of turbulent pressure is used to discrim- inate wind noise from sound. Once detected, wind noise is suppressed through a variety of algorithms, including single-channel noise reduction, microphone switching, or
binaural algorithms in wirelessly paired hearing aids (Luo and Nehorai, 2006; Launer et al., 2016).
Windscreen Physics: Better Than Average?
The means by which windscreens reduce wind noise may seem obvious at first: the microphone is sheltered from the wind. However, wind stagnating on the windscreen still cre- ates pressure sources on the surface that the microphone receives. At high frequencies, wind noise reduction is caused by averaging these sources over the surface. Eddies that are smaller than the windscreen will average with only partial coherence due to the properties of turbulence. Length scale considerations show that mean squared pressure fluctua- tions should decrease with the inverse square of frequency (van den Berg, 2006). Therefore, the bigger the windscreen, the greater the wind noise reduction.
The surface-averaging hypothesis fails at low frequen- cies, however, because significant wind noise reduction is also observed for eddy length scales much larger than the windscreen. To explain this, Phelps (1938) proposed that stagnation pressure fluctuations are distributed over the windscreen in the same way as the pressure field mod- eled by the mean flow. Morgan (1993) developed a model based on this hypothesis by using steady pressure distribu- tions measured in a wind tunnel. Similarly, Zheng and Tan (2003) used computational fluid dynamics to consider a range of wind velocities under Phelps’ (1938) hypothesis.
Phelps’ (1938) hypothesis predicts a large negative correla- tion between the pressure fluctuations on the windward and
Figure 6. Photographs of “wind gags” for wind noise reduction in early motion picture sound recordings (Clark, 1931).
24 Acoustics Today • Winter 2021