DBN - The De Bilt Infrasound Array

Infrasound is being measured by the Seismology Division of KNMI since February 1991. Experiments started with an array of woofers in KNMI's back-yard. The "De Bilt Infrasound Array" (DBN) was upgraded during the summer of 2002. The six instruments of DBN then consisted of electret microphones. These were replaced by KNMI microbarometers (mb). The KNMI mb's have a much lower low frequency cut-off of 500 seconds and are more suitable for durable outside operation than electret microphones. Figure 1 shows the pvc housing being installed, mostly below the earth's surface, in KNMI's back-yard.

Figure 1: The installation of pvc tubes for housing the KNMI mb's. DBN is installed in KNMI's back-yard between several meteorological measurement devices.

Configuring an infrasound array involves calculating the precise interstation distance and total aperture of the array. The lowest frequency of air pressure fluctuations which can be recovered by DBN is limited by the size of the back-yard. With approximately 100 meters available for the aperture, DBN is configured as shown in Figure 2.

Figure 2: The layout of DBN. Six KNMI mb's are placed two by two on the sides of an equilateral triangle.

The instruments are laid out such that an optimal array response is achieved. The array response is modeled by letting a monochromatic plane wave vertically incident on the array. The array response reflects the ability of the array to resolve a certain source. Figure 3 shows the array response of DBN to a 1 and 5 Hz plane wave.

Figure 3: The array response of DBN to a vertically incident plane wave of 1 Hz (left) and 5 Hz (right).

As follows from Figure 3, DBN's response has the characteristics of an optimal array for frequencies higher than 5 Hz. The response is circular, delta-like peaked and side lobes are of low amplitude and far from the main lob. Frequencies lower than 5 Hz will certainly be resolved but localization of the energy will become less accurate due to the broadness of the main lob.

Measuring infrasound means dealing with very low signal-to-noise ratios, often lower than 1. Wind is the main source of noise. The wind noise per instrument is reduced by attaching porous hoses. Doing so, the atmosphere above an instrument is sampled over an area rather than one point. Wind noise, which has small correlation lengths, will be cancelled out while the coherent infrasound signal will remain unaffected. The final installation of a microbarometer can be seen in Figure 4.

Figure 4: Top view of the noise reducer installation consisting of porous hoses (black). In order to sample not too close to the instrument, impermeable hose (green) is used for the first meter. The KNMI microbarometer is installed beneath the pvc cover, where also the noise reducer is attached.

The data acquisition system consists of a PC with Real Time Linux (RTLinux) as operating system. Data are converted from analog to digital signals through a National Instruments PCI AD card. This card is controlled under RTLinux by Comedi, the Linux Control and Measurement Device Interface. Acquired data are accessible through various tcp/ip protocols as telnet, ftp and http.

Figure 5 shows one of the first strong sonic events detected by DBN. The zero time of the time axis is 2002, August 23 23h58m30.094s local time (UT+2h). After 79.5 seconds, a beautiful N wave arrives at DBN. MB 03 and 04 are the first to measure the air pressure fluctuations, therefore the event has a northeastern origin. The N wave is caused by a plane flying through the sound barrier, a sonic boom. The pressure increase is generated at the front of the plane, while a pressure decrease is caused by the plane's back going through the sound barrier. After the N wave, up to 81 seconds, scattered energy travels over DBN. An U wave is visible around 83 seconds. The U wave shape results from reflection of the N wave, in the lower atmosphere, causing a 90 degrees phase shift.

Figure 5: Infrasound recordings from 2002, August 23, the time axis zero time is 23h58m30.094s local time. The data are bandpass filtered with a second order Butterworth filter with corner frequencies of 0.5 and 5 HZ. Coherent energy is clearly visible in all traces. The wave train starts with and N wave caused by compression of air at the front of the plane and extension at the back. Around 83 seconds an U wave arrives. The amplitude of the N phase is roughly 0.6 pascals (Pa). Also note the small travel time differences over the array which will be used for localizing the source.

Travel time differences of the infrasonic energy over the array are used to characterize an event. Events are characterized in the time and frequency domain in terms of back azimuth (bearing with respect to the North) and apparent sound speed. The apparent sound speed is the propagation speed of the infrasonic wave as measured by the mb's, this is the horizontal fraction of the true sound speed. Therefore, the apparent sound speed is a measure of the angle of incidence of the wave. The continuous recordings of the six mb's are automatically processed to detect coherent signals. The derived, so-called, Fisher ratio gives the statistical likehood of a coherent signal having traveled over the array.
Figure 6 shows the results from the time domain Fisher analysis for the sonic boom data. A significant increase in Fisher ratio is calculated around 79 seconds, corresponding to the sonic boom. Both the N and U wave give high values. The middle frames show stable solutions for the apparent sound speed and back azimuth. For the N wave 315 m/s and 70 deg are resolved, while the U wave has a apparent sound speed of 360 m/s. This higher velocity is caused by its steeper angle of incidence due to reflection. The top frame shows the best beam. The best beam is constructed by time aligning the signals and summing them. The best beam leads to an increase (square root) in signal-to-noise ratio, although this is not necessary in this very high signal-to-noise ratio event.

Figure 6: Time domain Fisher analysis of the infrasound data. The time axis is the same for all frames. The lower frame gives the Fisher ratio as function of time. A major increase of the F ratio starting at 79 seconds indicates a coherent wave traveling over the array. The middle two frames give the resolved apparent sound velocity and back azimuth, being respectively 315 m/s and 70 degrees for the first coherent arrival. The top frame shows the best beam, which is the sum of the 6 time aligned traces. This analysis is only part of the complete automatic analysis of infrasound on 2002, August 23.

The event is also analyzed in the frequency domain. The results are shown in Figure 7. Travel time differences are translated to phase differences in the frequency domain through the Fourier transform. The event is detected at maximum spectral power by performing beamforming in the slowness (p) domain. The slowness vector (white) resolves 325 m/s, being its length, and a back azimuth of 70 degrees, being its angle with respect to the North. Frequencies between 1.5 and 5 Hz were used in this broad band frequency analysis.

Figure 7: Broad band frequency domain analysis of the data, between 1.5 and 5 Hz. The white slowness vector points to the maximum spectral power. The length of the slowness vector gives a velocity of 315 m/s while its angle with respect to the North resolves a back azimuth of 70 degrees.

The Deelen Infrasound Array also detected the event. Localization can thus be done through cross bearing. The map in Figure 8 shows the location of the air plane at the Dutch-German border.

Figure 8: Localization of the sonic event through cross bearing.

Here, only one event on 2002, August 23 has been analyzed, the complete event list shows a huge amount of infrasonic activity.

August 2002
Läslo Evers