1. Introduction

Wake vortices are compact regions of air characterized by strong rotational motion. They create a strong velocity flow field that, in the case of an aircraft having an encounter with them, can lead to a severe upset that manifests itself in a vertical downwash or rapid uncommanded movements of rolling or pitching. Air traffic control has defined a classification that prescribes minimum distances between successive aircraft to avoid hazardous encounters. Although the separation distances ensure safety, they have an adverse effect on airport capacity, which becomes more and more important with continuously growing air traffic.

Many sensor technologies for detecting wake vortices have been studied and evaluated over the last 25 yr, including microwave and millimeter-wave radar, RASS, sodar, lidar, and anemometer-based ground wind lines. Each of the above falls short with respect to one or more of the following acceptance criteria: vortex detection sensitivity, vortex track capability, all-weather operation, automatic operation, real-time measurement, airport safety constraints, and cost. Of the preceding sensors, lidar has demonstrated particularly robust detection of wake vortices in clear weather and in light snow and rain. RASS (Rubin 2000) has been shown to satisfy all of the above criteria except cost. As a result, there is a continuing interest in new wake vortex sensor technologies.

Wake vortices in ground effect are often heard to emit audible sound at airports. Sound recordings made by the author in March 2002 at JFK International Airport indicate that wake vortices emit infrasound that is more than 40 dB stronger than audible wake vortex sound. In this context, infrasound refers to sound frequencies below 20 Hz. The primary objectives of the effort reported on herein was to determine whether infrasound emitted by wake vortices is a reliable indicator of wake vortex presence and whether wake vortex infrasound could be reliably detected in a background of airport and wind noise.

To make these determinations, sound recordings were made of arriving aircraft as they passed over the middle marker of runway 31R and during a follow-on period of 50–70 s. Preceding the JFK tests, recordings of sound produced by atmospheric wind were made in August 2000 on top of Mount Washington (New Hampshire), where the highest winds ever measured on the earth’s surface (104 m s⁻¹) were previously recorded. Additional wind sound recordings were made in February 2002 at Jones Beach (New York), an area virtually free of vertical obstacles to wind flow, a feature particularly important for measuring the sound produced by lower-speed winds.

Previous wake vortex studies have presented acoustic spectra for wake vortices out of ground effect. For example, Michel and Böhning (2002) have published sound spectra using a non-real-time phased microphone array. These spectra were obtained 6 and 12 s following aircraft passage with the array focused at an altitude of 40 m.

The present paper differs from the existing literature in four ways. First, a single microphone at ground level was used to make all recordings. Second, acoustic spectra are presented for wake vortices in ground effect. Third, wake vortex spectra are compared to sound spectra generated by aircraft and ambient wind. Fourth, plots of infrasound emitted by wake vortices following aircraft passage are presented for several types of aircraft.

2. Experimental details

The recording microphone was enclosed in a two-stage microphone blimp for all measurements. The blimp reduced wind noise and had no effect on microphone response up to about 8 kHz. Wind data were collected at three sites: Mount Washington, Jones Beach, and JFK Airport.

On Mount Washington, the microphone-blimp assembly was located on the roof of the Mount Washington Observatory behind a low wall that was closest to the arriving wind, in which direction the treeless terrain fell off smoothly. The wall blocked wind from directly impacting the microphone-blimp assembly. Wind speeds varied from 8 to 30 m s⁻¹. Following amplification, the microphone output was recorded on a Sony deck.

At Jones Beach, a flat wind screen made of 5-cm-thick foam shielded the microphone-blimp assembly from direct wind. Both rested on the pavement at the center of a very large empty parking lot. The battery-powered recording equipment was located about 30 m from the microphone in a leeward direction. Jones Beach is located on a long sandbar in the Atlantic Ocean off the coast of Long Island. The parking lot is surrounded by seawater on three sides, all a considerable distance away. Wind speeds varied between 2.5 and 8.8 m s⁻¹. The wind was consistently from the northwest, in which direction there are no structures of any type that might interact with the wind, either on the sandbar or on the directly opposite mainland for many miles inland. Sea wave noise was also not a contaminating sound source since the sandbar at this location is separated from the mainland by a narrow channel less than 90 m wide. The shortness of the fetch, the distance over which the wind interacts with the seawater surface, results in very small waves on the sandbar beach, which itself was 800 m away from the microphone site in a northwesterly direction.

At JFK, sound was recorded during aircraft arrivals on 5 and 6 March 2002 at the U.S. Department of Transportation (DOT) Volpe Center Wake Vortex Test Site, which is adjacent to the middle marker of runway 31R. Aircraft are 60 m above the ground at the middle marker. To prevent vortex flow from reaching the microphone, the microphone-blimp assembly was completely enclosed in a 5-cm-thick porous-foam-covered enclosure whose frame was made of polyvinyl chloride (PVC) pipe. The enclosure sat on a flat sandy surface about 120 m from the glide slope centerline and 1.05 km from the runway threshold.

Each JFK recording was initiated just before an aircraft reached the middle marker and terminated 50–70 s later. Reverse engine thrust noise was not heard by the author during recordings because of the high level of background airport noise. Landing aircraft typically initiate reverse engine thrust about 1.3 km past the middle marker, in which case reverse thrust noise arriving at the microphone would be attenuated by at least 20 dB more than aircraft noise generated at the middle marker.

Using the time display on the recorder, the starting and ending times of the interval in which a vortex was physically sensed at the recording site were noted. In this context, physical sensing meant that three cues were simultaneously satisfied: 1) visible agitation of vegetation around the recording site, 2) vortex wind flow sensed on the author’s face and hands, and 3) aurally sensed vortex sound. During recordings made on 5 March, the wind speed was estimated to be 4–7 m s⁻¹ up the glide slope, with a small crosswind component toward the microphone. At about 4 p.m. local time the wind shifted to being totally crosswind toward the microphone. On 6 March, the wind continued transverse to the glide slope at an estimated speed of 4–7 m s⁻¹ throughout the day.

3. Instrumentation

Wind speeds on top of Mount Washington were obtained by noting the starting and ending time for each recording interval and referring afterward to recorded observatory wind speed data (clocks having been previously synchronized). The observatory wind speed sensor was located about 5 m off to the side of the recording site. At Jones Beach, wind speed was measured with a Skymate Wind Meter made by Speedtech Instruments. At JFK, wind direction was obtained from an airport wind sock, and wind speed was estimated by the author as noted earlier.

All recordings were made with a single Shure SM81 condenser microphone whose response is flat from 20 Hz to 20 kHz. The microphone’s response below 20 Hz was provided to the author by Shure (2000, personal communication). The SM81 has a high clipping level and low distortion over a wide range of load impedances. The microphone head was completely enclosed by a Shure two-stage A81WS microphone blimp.

The SM81 microphone output was amplified by a Shure FP23 battery-powered amplifier that is designed for severe field conditions. It has a gain of 66 dB that is adjustable in 11 steps, a dynamic range of at least 122 dB at the +18 dB gain setting, and an audio response that is essentially flat from 10 Hz to 22 kHz. Shure provided response data for the amplifier below 10 Hz.

The FP23 was connected to a Sony MDS-JE630 MiniDisc Deck. The deck’s frequency response is flat within 0.3 dB from 5 Hz to 20 kHz. Its sampling frequency is 44.1 kHz. Output signal-to-noise ratio (SNR) is over 98 dB during playback, and wow and flutter are below measurable limits. The deck panel contains two peak level meters, one for each stereo channel. Only one channel was used to record data. To maximize recorded SNR the FP23 gain was adjusted so that the Sony peak level meter flashed briefly about once every 15–30 s.

All sound spectra were computed using the fast Fourier transform (FFT) algorithm. For this purpose, Sony recordings were played through a PICO ADC-100, 12-bit A/D converter, which is flat from 0 to 50 kHz, into a USB port of a Windows XP Gateway 700X computer. The FFT size was 4096 and the sampling frequency was 10 528 Hz. PICO and Excel software were used to make the spectral plots, which were corrected for Shure microphone and amplifier falloff below 20 Hz.

The magnitude of sound recorded during each JFK aircraft landing was plotted versus time in two ways, again using PICO and Excel software. The first set of plots was obtained by playing Sony recordings through a 200-Hz analog low-pass filter into a PICO ADC-100 converter that sampled the data at a 500-Hz rate. The converter was connected to a USB port on a Gateway 300X computer with a Windows 2000 Professional operating system. To generate the second set of plots, the digitized sound already stored in the computer was processed by a filter that approximately matched the infrasound spectrum generated by wake vortices. A matched filter optimizes signal detection in a background of white Gaussian noise (DiFranco and Rubin 1968). The matched filter passband was determined by a digital filter that cut off sound above 20 Hz at 18 dB per octave and the Shure microphone, whose response falls off rapidly below 7 Hz.

4. Results

Atmospheric wind noise and aircraft noise are the principal sources of sound interfering with the detection of wake vortex sound at airports. Figure 1 shows four sound spectra (in dB Hz⁻¹) generated by atmospheric wind, two of which were recorded on top of Mount Washington, one at Jones Beach, and one at JFK. The decibel level is referenced to a sound pressure level (SPL) of 2 × 10⁻⁴ μb. The spectral resolution is 2.57 Hz, corresponding to a sampling frequency of 10 528 Hz and a 4096-point FFT. Each spectrum is an average of about six FFTs.

Figure 1 shows that wind sound is loudest in the infrasound band and that the spectral shape of wind sound is essentially independent of wind speed and geographic location. If the corner frequency of each spectrum is defined by the intersection of a line approximating spectral falloff above 50 Hz with a horizontal line approximating spectral amplitude between 4 and 10 Hz, the corner frequency decreases with increasing wind speed. This implies that the peak intensity moves toward lower infrasound frequencies with increasing wind speed.

In Figs. 2–7, the magnitude of 200-Hz low-pass-filtered sound is plotted versus time for each of six aircraft landings. To obtain Figs. 8–13 , the sound plotted in each of Figs. 2–7 were wake vortex-matched filtered before being magnitude detected, as previously mentioned.

The vertical arrow shown on each sound-versus-time plot indicates when the arriving aircraft passed over the middle marker according to the author’s notes as pinpointed by the aircraft noise peak. A double-headed arrow marks the interval over which the vortex pressure field was physically sensed and audible vortex sound was heard, as described earlier. The date of each sound recording is shown in the lower right corner of each figure.

Figures 14–17 each display acoustic spectra for (peak) aircraft sound, wake vortex sound (when a vortex was physically sensed), and wind sound. These spectra were obtained from the same recordings used to produce Figs. 2–5, respectively.

The following description of wake vortex rollup and descent toward the ground (Hallock and Eberle 1977) will be referred to below. The wake of an aircraft in high-lift configuration (take-off and landing phase) initially contains multiple concentrated vortices that emanate from lifting surface discontinuities (wing tip, flap tips, engine nacelles, horizontal tail plane, fuselage, etc.). In the near field, they interact with each other while rotating around each other and eventually merge into a pair of symmetrical counterrotating vortices that are separated by approximately three-quarters of the generating aircraft’s wingspan.

The vortex pair descends because of mutual velocity induction. The initial descent rate is proportional to aircraft weight and inversely proportional to flight speed and the square of the wingspan. Near the ground, the descent rate decreases and the vortices level off at a height approximately equal to one-half of their initial separation, hereinafter referred to as equilibrium height. A vortex enters ground effect when its height is about a wingspan above the ground, which is approximately equal to 2.5 times equilibrium height. This means that the vortices of heavy aircraft are mostly in ground effect right after they are generated at the middle marker.

Figure 8 shows the effect of a vortex-matched filter on the sound plotted in Fig. 2 that was recorded during the arrival of a B747. The peak aircraft sound around 17 s is reduced by a factor of 10 between the two figures. This reduction is substantially understated because the sound in Fig. 2 has already passed through a 200-Hz low-pass filter before being sampled. Figure 8 also shows that the peak B747 aircraft sound recorded 120 m away is smaller than the peak vortex sound, suggesting that noise from other aircraft around the airport, including reverse engine thrust noise, is also significantly attenuated by wake vortex-matched filtering.

As Fig. 8 shows, the recorded infrasound peaked during the time the vortex pressure field was physically sensed, between 30 and 35 s. This is the only example where infrasound totally disappeared after the vortex pressure field ceased to be sensed. Figure 8 also shows two small pronounced sound peaks at 21 and 26 s. Based on their timing relative to the aircraft sound peak, it is conjectured that the first small sound peak was generated during vortex rollup and the second during early descent of the primary vortex pair.

Crosswind advection and vortex interaction with the ground combine to move a vortex in ground effect away from the flight corridor. In calm wind, vortex interaction with the ground causes the two vortices to move in opposite directions at speeds of approximately 1–2 m s⁻¹. For crosswinds greater than 3.5 m s⁻¹, both vortices move in the direction of the crosswind, with the downwind vortex moving slightly faster and the upwind vortex moving slightly slower than the crosswind.

Figure 8 shows that the pressure field arrived at the microphone site about 16 s after aircraft passage. The wingspan of a B747 is 60 m, and the center of each primary vortex during initial descent is separated from the glide slope centerline by about three-eighths of a wingspan, or 22 m. Since the microphone was located 120 m from the glide slope centerline, the downwind vortex traveled 98 m to reach the microphone. The wind during the B747 passage was predominantly transverse to the glide slope and toward the microphone. An average wind speed of 6 m s⁻¹ is consistent with the vortex reaching the microphone in 16 s.

To obtain the spectra in Fig. 14, the spectrum of B747 aircraft sound was computed around 17 s, B747 vortex sound around 33 s, and wind sound around 40 s, with reference to the plot in Fig. 8. The shapes of the B747 vortex spectrum and the wind spectrum are similar except for a pronounced peak in the vortex spectrum between 10 and 20 Hz. Figure 14 shows that B747 sound in the infrasound band is smaller than that for vortex sound, which is consistent with the magnitude relationship shown in Fig. 8. Figure 14 also shows a slower falloff rate of aircraft sound in the audio band compared to vortex sound, which confirms the benefit of eliminating sound above 20 Hz.

Figure 3 shows sound recorded during an MD80 landing at JFK before being matched filtered, and Fig. 9 after being matched filtered. The aircraft sound peak at 6 s in Fig. 3 is again reduced by a factor of 10 in Fig. 9.

During this recording, the speed of the crosswind is estimated to have been less than 1 m s⁻¹, so that the MD80 vortex arrived at the microphone site well after the recording was terminated, which explains why no vortex pressure field was physically sensed during the recording. The infrasound peak at 53 s in Fig. 9 is significantly larger than the MD80 aircraft infrasound peak and is presumed to have been generated by an MD80 vortex prior to its arrival at the microphone. This infrasound continues past the end of the recording.

In Fig. 15, the MD80 aircraft sound spectrum was computed around 6 s, the wind spectrum around 25 s, and the MD80 vortex spectrum around 53 s, with reference to Fig. 9. The vortex spectrum shape, as before, is similar to the wind spectrum shape except for a small peak between 10 and 20 Hz. The slower falloff rate of the aircraft spectrum in the audio band relative to the vortex spectrum again shows the benefit of vortex-matched filtering.

Figures 4 and 10 show sound recorded during a B767 landing. Vortex-matched filtering once again reduced the aircraft sound peak around 17 s by a factor of 10. In this example, as Fig. 10 shows, the peak vortex infrasound was smaller than the peak aircraft infrasound. This may be related to the fact that a weak vortex pressure field was sensed in this case. Also, vortex infrasound continued after the vortex pressure field was no longer sensed.

Figure 10 shows that the vortex pressure field was physically sensed about 25 s after aircraft passage. This is consistent with an average crosswind speed of 4 m s⁻¹, which is within the range of estimated crosswind speeds for that day.

The spectral densities shown in Fig. 16 were calculated for aircraft sound around 17 s, wind sound around 28 s, and vortex sound around 38 s, with respect to the plot in Fig. 10. The B767 wake vortex spectrum exhibits a very broad peak extending from 40 Hz down to 10 Hz. Figure 10(correctly) shows that the magnitude of vortex infrasound at 38 s is weaker than aircraft infrasound at 17 s, while Fig. 16 indicates otherwise. This apparent contradiction is due to the short duration of the aircraft sound peak. Each FFT was obtained by averaging the recorded sound over a reasonably short time interval. In this case, the actual FFT averaging interval was probably somewhat wider than the aircraft peak sound duration, resulting in a reduced and perhaps slightly inaccurate aircraft sound spectrum.

Figures 5, 17 and 11 contain plots for an A320 aircraft. During this recording, the wind direction, as in the previous example, was toward the microphone. Many of the earlier comments are also applicable to this example. Note that the peak aircraft sound in Fig. 5 is reduced in Fig. 11 by a factor of 6. The smaller reduction may be characteristic of A320 aircraft noise. Figure 11(correctly) shows that the aircraft infrasound peak is larger than the vortex infrasound peak, while Fig. 17 indicates otherwise. This again is an artifact of the narrow aircraft sound peak at 12 s, as previously explained.

The explanation of the small sound bumps at 16 and 23 s is the same as before. Finally, an average crosswind speed of 4 m s⁻¹ toward the microphone, the same as in the previous example, would account for physical sensing of the vortex pressure field at the microphone site after an elapsed time of about 25 s. Also as before, vortex infrasound continued after the vortex pressure field was no longer physically sensed.

Figures 6 and 12 show sound recorded during an A340 passage before and after being matched filtered, respectively. The A340 arrived on 5 March shortly after the wind direction shifted toward the microphone at 4 p.m. local time. Figures 6 and 12 show a reduction factor of 7 in aircraft sound between the two plots, almost the same as that for the A320. Figure 12 shows that vortex infrasound is substantially stronger than aircraft infrasound.

A crosswind slightly greater than 4 m s⁻¹ toward the microphone is consistent with physically sensing the vortex pressure field about 22 s after aircraft passage. This example differs from previous examples in that the recorded infrasound dipped sharply while the vortex pressure field was sensed. As in previous examples, vortex infrasound continued well after the pressure field was no longer sensed.

Figures 7 and 13 contain sound plots for a second B767 aircraft. This aircraft arrived about 35 min after the A340 mentioned above, at which time the wind direction was still toward the microphone, but gusting was stronger. Figures 7 and 13 show a reduction factor of almost 8 in aircraft sound between the two plots. Figure 13 shows that vortex sound was again stronger than aircraft sound. The plots in Figs. 7 and 13 are quite similar except for the period between 38 and 46 s. While the author did not hear reverse engine thrust noise during any of the recordings, this interval coincides with the time when B767 reverse engine thrust noise would be expected and is the most likely explanation for the difference between the two figures. A comparison of Figs. 7 and 13 reveals that the sound between 38 and 46 s was reduced by a factor between 15 and 20. If reverse engine thrust noise was in fact the source of the sound, it indicates that reverse engine thrust noise is strongly attenuated by the vortex-matched filter.

A second anomaly in this example is the fact that the elapsed time between aircraft passage and the time a vortex pressure field was physically sensed is about 14 s. This would require a crosswind speed of about 7 m s⁻¹. The author’s notes do not show that the wind increased during this time but does state that wind gusting increased. This is another example where the recorded vortex infrasound continued well beyond the time the vortex pressure field was physically sensed.

No explanation has been given for the fluctuations in vortex-generated sound in the above figures. Such an explanation requires knowledge of the wake vortex sound generating mechanism, which has not yet appeared in the literature, as well as an understanding of vortex behavior. A recent report (Hardin and Wang 2003) analyzes the interaction of a primary wake vortex pair with the ground. The vortices are modeled as two tip vortices of finite length with opposite rotation. The report calculates the radiated sound to be a narrowband tone with a peak frequency near 0.01 Hz. It is obvious that this mechanism did not generate the wake vortex sound spectra shown above.

Powell (1964) showed that sound is generated aerodynamically by the movement of eddies, or vorticity, in an unsteady fluid flow. A cardinal result of Powell’s paper is that vorticity in a slightly compressible fluid induces the whole flow field, both the hydrodynamic part and the acoustic part.

As pointed out earlier, the spectra of wind and vortex sound display a strong similarity. If one assumes that both sounds are generated by the same mechanism, a plausible hypothesis based on Powell’s result is that both sounds are produced by airflow close to the ground interacting with the ground. More particularly, vortex sound is generated within the airflow outside the core that is close to the ground. If this hypothesis is correct, a corollary is that fluctuations in vortex sound are the result of unsteadiness of the vortex flow field and, longer term, vortex structural changes due to aging.

5. Conclusions

Matched filter detection of wake vortex infrasound appears to be a useful method for detecting the presence of wake vortices in ground effect, provided the detection algorithm takes into consideration temporal fluctuations in vortex infrasound magnitude. The flatness of vortex sound spectra at the low end of the infrasound band suggests that detection performance may be improved by using microphones sensitive down to 1 Hz, and perhaps lower.

Infrasound detection of wake vortices has several theoretical advantages. First, vortex-generated sound peaks in the infrasound band. Second, infrasound is unaffected by atmospheric attenuation. Third, aircraft noise, airport noise, and rain noise are mostly concentrated above the infrasound band. Fourth, a small number of microphones may be able to detect wake vortex presence over a large area. Although a vertically pointing acoustic array would theoretically reduce interference from distant airport noise sources, an acoustic array with directivity at infrasound frequencies would be extremely large, and, more importantly, wake vortices in ground effect would lie mostly in the array’s near field.

Operational advantages of infrasound detection of wake vortex presence include airport and aircraft safety, simplicity, low cost, ruggedness, and minimal impact on performance by airport noise and weather. A method that utilizes a small number of microphones to detect and track wake vortices in ground effect around the middle marker has been previously described (U.S. patent application 2003, W. L. Rubin).