National Academies Press: OpenBook

Acoustic Beamforming: Mapping Sources of Truck Noise (2009)

Chapter: Chapter 3 - Research Findings

« Previous: Chapter 2 - Research Approach
Page 12
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 12
Page 13
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 13
Page 14
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 14
Page 15
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 15
Page 16
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 16
Page 17
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 17
Page 18
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 18
Page 19
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 19
Page 20
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 20
Page 21
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 21
Page 22
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 22
Page 23
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 23
Page 24
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 24
Page 25
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 25
Page 26
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 26
Page 27
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 27
Page 28
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 28
Page 29
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 29
Page 30
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 30
Page 31
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 31
Page 32
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 32
Page 33
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 33
Page 34
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 34
Page 35
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 35
Page 36
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 36
Page 37
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 37
Page 38
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 38
Page 39
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 39
Page 40
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 40
Page 41
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 41
Page 42
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 42
Page 43
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 43
Page 44
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 44
Page 45
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 45
Page 46
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 46
Page 47
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 47
Page 48
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 48
Page 49
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 49
Page 50
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 50
Page 51
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 51
Page 52
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 52
Page 53
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 53
Page 54
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 54
Page 55
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 55
Page 56
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 56
Page 57
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 57
Page 58
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 58
Page 59
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 59
Page 60
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 60
Page 61
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 61
Page 62
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 62
Page 63
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 63
Page 64
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 64
Page 65
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 65
Page 66
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 66
Page 67
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 67
Page 68
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 68
Page 69
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 69
Page 70
Suggested Citation:"Chapter 3 - Research Findings." National Academies of Sciences, Engineering, and Medicine. 2009. Acoustic Beamforming: Mapping Sources of Truck Noise. Washington, DC: The National Academies Press. doi: 10.17226/14311.
×
Page 70

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

12 3.1 Review of Beamforming for Vehicle Noise Source Identification A recent development in microphone array application for localization of sound sources of moving motor vehicles (21) showed that 124 microphones arranged in a two-dimensional snowflake array made it possible to locate sound sources dur- ing car passbys with a resolution of 1 to 1.3 ft (0.3 to 0.4m) in a frequency range of 500 to 4500 Hz. With an overall array size of approximately 12 by 12 ft (3.7 m by 3.7 m) and a dis- tance of 8 ft (2.4 m) between the array and test object, the two-dimensional distributions of the sound pressure level were measured during passbys of an automobile at speeds from 50 to 110 mph (80 to 180 km/h). In the frequency range between 300 and 500 Hz, however, the array’s spatial resolution was worsened by a factor of 2. Also, even with a simple, stationary point source in front of the center of the array, the false sources of lower magnitude, so-called ghost images, begin to appear in the resulting source maps at frequencies of 1000 Hz and above despite the array side lobes suppressed by the 10 to 15 dB array gain. Barsikow suggests that the frequency range of the array could be extended by increasing the number of microphones to a total of about 160 (21). Commercial beamforming wheel arrays, instrumentation, and software packages were developed by Brüel & Kjær (22, 23). The wheel arrays consist of a number of spokes (from 7 to 15), each holding up to six microphones. Three typical wheel- array designs of 42, 66, and 90 channels with different perfor- mance levels cover different applications from general purpose to automotive components to entire vehicle. The geometry of each array is optimized for minimum side lobe levels over a certain frequency range. A 90-channel Brüel & Kjær wheel array with a PULSE beamforming software package were used by I&R for truck noise testing on IT’s endurance track performed for Caltrans in 2005 (20). Useful results were obtained, but there were lim- its associated with the system being general purpose, rather than optimized for the specific type of measurements needed for moving trucks. In general, however, the circular spiral arrays (24–28) offer adequate side lobe suppression with a minimum number of sensors. As such, this class of arrays can be used to provide affordable performance over a broad frequency range. The cir- cular spiral microphone array was used as the baseline for the beamforming design in the current study. 3.2 Development of Experimental Design 3.2.1 Noise Mapping Technique Development Specific parameters that were considered relative to the beamforming technique developed in this study included the system requirements for heavy trucks (e.g., frequency range of interest, horizontal and vertical spatial resolution required, source to array distance) and design parameters for beamform- ing measurement application (e.g., number of microphones, array size and geometry, sample size and averaging time, pro- cessing software). A factor considered, both as a requirement and a design parameter, was the frequency resolution. One final factor is that, for highway barrier design purposes, verti- cal resolution is much more important than horizontal. The design of candidate array options was built on lessons learned during the array-based demonstration tests of truck noise source localization conducted by Blake and Donavan under Caltrans sponsorship in 2005 (20). The current design study had four major objectives: • Extend the directivity gain to lower frequencies than used in the Caltrans demonstration of 2005. • Meet or exceed the array gain performance of the Brüel & Kjær array used in the 2005 test. C H A P T E R 3 Research Findings

• Optimize array performance for an affordable 70-element array. • Tailor array shape to optimize: – Vertical directivity through vertical array aperture; – Horizontal directivity through combined array aperture and cross-range spreading loss during vehicle passby; – Array height (vertical aperture) for assembly/disassem- bly in the field controlled by available facility; – Side lobe suppression to minimize “false” source local- izations at the high end of frequency range of interest. The low frequency limit on array design was found to be 250 Hz, which should localize sources (potentially exhaust muffler related) to the upper or lower half of the cab area. Degraded performance is unavoidable at lower frequencies, given practicalities of array handling in the field. The beamforming software package was developed to calcu- late both two-dimensional source image distributions pro- jected in a vertical plane at the truck side and one-dimensional vertical sound source distributions determined at a distance of 20 ft (6 m). All processing was done in the frequency–time domain in which the received signal is sectioned into roughly 15 time segments of 0.1 to 0.15 s duration. Within each time segment, the spectrum of the sound at each array micro- phone was digitally computed in equivalent 1 Hz bands. These spectral levels are used to compute either sound pres- sure levels or inter-element cross-spectral densities. The processed sound pressure levels are used to evaluate the ver- tical profiles. The cross-spectral densities are used to localize in elevation (up and down) and cross-range (left or right) to give the two-dimensional source map in the vertical plane. To do this step, the array is digitally steered to its left and to its right. Because the arrival direction of the sound “looks” at the truck both approaching and leaving the array zone, an adjustment in Doppler shift is made at each frequency and for each arrival angle of the sound from the truck. Additionally, a correction is made for the range geometry spreading loss (i.e., 1/r2) at each of the time segments. The array is mathematically modeled as an acoustic lens that focuses on various points in the vertical plane in order to “scan” for the sources. The narrow band images, determined for 1 Hz band width, were summed to obtain images and sound levels in one-third octave frequency bands. All microphone signals, the photocell signals, and a recording time code were simultaneously recorded to provide accurate time resolution of the images at each moment of the passby. A digital photograph of a truck was geometrically scaled to the image plane dimensions using a pre-determined scale factor. 3.2.2 Microphone Array Design The design strategy for the microphone array in this study needed to accommodate many factors, among them: • Develop a circular spiral baseline array that approximates or surpasses the Brüel & Kjær array directivity for the truck application as used in the demonstration runs of 2005. • Reduce the element number to 70 or 72, depending on the numbers of spokes and circles in the array. • To ensure comparable side lobe structure with a reduced number of elements, reduce array circular area proportion- ately. This proportionate reduction provides comparable aperture area per element. • Deform the circular spiral array by expanding the vertical aperture using the chart of spot width vs. frequency (see Figure 4) developed for the study, to select a trial major axis of the ellipse. Maintain element number and array area while deforming. • Selectively introduce tilt of the spokes and spiral angle, to improve side lobe suppression as necessary. Figure 1 illus- trates the essential geometry of an array of the type used in this study. The notional array that is illustrated in Figure 1 represents one of the possible approximations to the array previously used in the Caltrans demonstration tests (20). This array perfor- mance was the baseline for the current design study. The cir- cular array for this application, however, provides unnecessary localization in the horizontal direction and insufficient local- ization in the vertical direction. Given the intrinsic horizontal localization afforded by the passby itself, a deformed array was considered to optimize the effective localization vertically and horizontally for enhanced performance at low frequencies. Deformation of the parent circular array into an ellipse pro- vides a means to tailor these arrays for truck passby. This facet of the array design represents one of the new products of the current project. The spiral angle and the spoke configuration introduce spatial irregularity into the circular array, which allows required side lobe suppression. All of the arrays shown in Figure 2 have the same area (i.e., AB = D2) and all have the same area density of ele- ments. Figure 3 shows that the lobe structures of these arrays at 1000 Hz are virtually unaffected, as apparent from the illustration. The projection of the lobe structure has a shape outline that follows the outline of the array, but is rotated 90 degrees relative to it (i.e., an increase in array length along the x-axis results in a narrower beam in x; con- versely, a reduction in array length along the y-axis results in a wider beam in y). (In this example, x represents the ver- tical axis and y represents the horizontal axis on a plane par- allel to the array plane.) Figure 4 provides a chart that gives the approximate lobe width, here called “spot width,” as it refers to the localization in the truck side plane 20 ft (6 m) from the array plane. As shown, lengths on the order of 12 ft (3.7 m) are required to localize to within ±5 ft (1.5 m). 13

In Figure 5, the spot width is illustrated for the major and minor axes of the alternative arrays shown in Figure 2 [i.e., Ellipse 1, Ellipse 2, and the 7 ft (2.1 m) circular array] as a ref- erence. Given the design requirement of constant area for all arrays, the degradation in horizontal beam width for the ben- efit of vertical discrimination is clearly seen. 3.2.3 Balance Between Array Aperture and Spherical Spreading Loss As the truck sources pass by the microphone array, there is a sound level change at the array microphones that is simply due to the spherical spreading loss resulting from the varying 14 Approx. Spiral Angle (b)(a) Figure 1. Notional 90-element circular spiral array with 15 spokes of 6 elements each and inner and outer radii of 19.7 and 47.6 in. (0.5 m and 1.2 m), respectively: (a) array pattern, (b) directivity pattern of array projected on x,y plane parallel to array plane and situated 20 ft (6 m) in front of it. (c) Ellipse 2(b) Ellipse 1(a) Circular D= 7 ft A= 9.1 ft B= 5.37 ft A= 11.8 ft B= 4.1 ft Figure 2. Patterns for various arrays of 70 elements distributed around 14 spokes of 5 elements each: (a) circular spiral, (b) Ellipse 1, (c) Ellipse 2. A is the major axis, B is the minor axis. The spiral angle is 0.5 degrees.

15 (b) Ellipse 1(a) Circular (c) Ellipse 2 Figure 3. Directivity pattern projections on the parallel x,y plane 20 ft (6 m) in front of the array plane shown in Figure 2 at 1000 Hz. 0.1 1 10 100 000010001001 Frequency, Hz Sp ot W id th , ft La=4 ft La=6 ft La=8 ft La=10 ft La=12 ft Array Length or Diameter Sp ot W id th , ft Figure 4. Width of the focus spot of arrays of various lengths or diameters as function of frequency at range of 20 ft (6 m).

16 0.1 1 10 100 000010001001 Frequency, Hz Sp ot W id th , (- 6d B) , ft Circular Ellipse 1:Major Ellipse 1:Minor Ellipse 2:Major Ellipse 2:Minor Sp ot W id th , (- 6d B) , ft Figure 5. Total width of the −6 dB-down focus spots for major and minor axes of arrays shown in Figure 2. -15 -10 -5 0 5 -80 -60 -40 -20 0 20 40 60 80 Cross-Range, ft Va ria tio n in S PL , d B Ellipse 2 Range - Limited Ellipse 1 Circle Figure 6. Horizontal profile of sound pressure level (SPL) due to passby of simple source at 20 ft (6 m) from array compared with cross-range apertures of array alternatives at 250 Hz. distance. Ignoring possible effects of source directivity, which are probably of little concern in the 250 Hz region of interest, this variation in the sound level provides some localization along the truck. Figure 6 illustrates this effect. The −6 dB-down points for the passby are indicated along with the −6 dB-down points (at 250 Hz) for the three arrays that are shown in Fig- ure 2. Although the combined effect of spreading loss and directivity gain has not been calculated, the approximate sound profile can be estimated. Figures 7 and 8 provide these approx- imate profiles for Ellipses 1 and 2. In all cases, the closest point of approach (CPA) is 20 ft (6 m). These sketches show that spreading loss provides added dis- crimination as the horizontal directivity is reduced to provide for an increased vertical dimension of the array. One cannot arbitrarily increase the horizontal dimension because this would result in degradation of the side lobe structure.

Table 1 summarizes these results for 250 and 465 Hz. The frequency of 465 Hz was used because it was one of the analy- sis bands used in the 2005 Caltrans study, and the study team anticipated that some future design evaluations might need to be compared to data obtained at this frequency. The numbers in parentheses are the lobe widths for the array alone, without benefit of the spreading losses. These values are theoretical, and comparison with the experimental ones is provided in Section 3.5.1 (see Figure 39). The lobe width for a hypotheti- cal point array (a single omni-directional microphone) is also shown in the table for comparison. 3.2.4 Design Conclusions The following conclusions were drawn from the design analysis described previously for the frequency band of highest A-weighted sound levels of emissions (during cruise): • A 70-element elliptical array provides adequate aperture with acceptable side lobe suppression. • The aspect ratio 1.7 of an elliptic aperture array provides beam patterns that are geometrically similar to the array shape at all frequencies of interest. 17 -15 -10 -5 0 5 -80 -60 -40 -20 0 20 40 60 80 Cross-Range, ft Va ria tio n in SP L (d B) Ellipse 2 Range - Limited Ellipse 1 Circle Approximate Combined Pattern: Ellipse 1 Va ria tio n in SP L (d B) Va ria tio n in SP L (d B) -15 -10 -5 0 5 -80 -60 -40 -20 0 20 40 60 80 Cross-Range, ft Va ria tio n in SP L (d B) Ellipse 2 Range - Limited Ellipse 1 Circle Approximate Combined Pattern: Ellipse 2 Figure 7. Approximate profile of sound pressure level for simple source passing by Ellipse 1 array at 250 Hz. Figure 8. Approximate profile of sound pressure level for simple source passing by Ellipse 2 array at 250 Hz.

• Vertical directivity provides a beam focus spot 50 in. (1.3 m) wide (−6 dB-down) at 465 Hz. • The array provides excellent side lobe suppression of approximately −14 dB over the range of 250 to 2250 Hz, and approximately −11 dB at 8000 Hz. • A minimum spiral angle is needed to suppress side lobes, so element supports may be radial. • Side lobes at high frequencies seem well distributed (i.e., appear amorphous), which minimizes the likelihood of unwanted highlights that may lead to false source images. With regard to low-frequency performance, the following conclusions were drawn: • With a 12 ft (3.7 m) major (vertical) axis of the ellipse, the vertical beam half height (−6 dB) is about 4.5 ft (1.4 m) at 250 Hz. This beam size will allow imaging resolution to the upper half of a large truck cab. • The horizontal effective beam width during passby, includ- ing both beam width and spherical source spreading loss, would be about 9 ft (2.7 m) for the minor axis [horizontal dimension of 4.1 ft (1.2 m)]. The elliptical array described herein represents a new result of this study. Based on the design study for the Ellipse 2 array optimized for the current application, a 70-element array (schematically shown in Figure 9) was selected for experimental engineering and implementation through the proof-of-concept testing, as described in the following sections. 18 Table 1. Width of net focus spots for array alternatives including cross-range spreading. Frequency (Hz) Direction Lobe Width (ft) for Array Design* Circular Ellipse 1 Ellipse 2 Point 250 Vertical 19 14.5 10.8 30 Horizontal 18 (19) 21 (28) 25 (43) 30 465 Vertical 9 7.3 5.6 30 Horizontal 8 (9) 12 (13) 15 (18) 30 * Numbers in parentheses represent values for the array alone -80 -60 -40 -20 0 20 40 60 80 -80 -60 -40 -20 0 20 40 60 80 lx, in ly , in Figure 9. 70-element elliptical array.

3.3 Experimental Microphone Array Engineering 3.3.1 Mechanical Design As the result of the development described in the previous sections, a 70-microphone elliptical array was designed, with an aspect ratio of 1.7, a width of 4 ft (1.2 m), and a height of 12 ft (3.7 m). This design should provide source resolution down to 250 Hz with side lobe suppression of −12 to −14 dB. The microphone array assembly and data acquisition system were further developed by Wyle. Due to the significant size of the array aperture, it was assembled on a metal frame consist- ing of three separate sections vertically mounted together and installed on a four-wheel metal base. The sections, each approximately 4 ft by 3 ft (1.2 m by 0.9 m), could be easily dis- assembled for shipping. Fourteen PVC pipe spokes, each hold- ing 5 microphones, were mounted on the frame sections, providing the 70-microphone elliptical pattern that was designed. The identical lower and upper frame sections hold five spokes each, with five microphones mounted equidistantly on each spoke. The middle frame section holds four spokes with five microphones each. During the field tests of the array, additional microphones were mounted along the central ver- tical axis of the middle frame section, raising the total number of microphones to 73 or 77 for some tests. The assembly with additional microphones is shown in Figure 10. 3.3.2 Data Acquisition System The array was equipped with the 0.24 in. PCB Piezotronics Series 130 array microphones with 0.25 in. ICP® preampli- fiers Model 130P10 or integral ICP® preamplifiers. The microphones/preamplifiers were inserted in holes predrilled in the spokes of the array, each provided with a 3.5 in. windscreen. Prior to array assembly, the microphones were phase-calibrated in pairs using a Brüel & Kjær Type 51AB sound intensity calibrator. The data acquisition system was completed using a National Instruments Model PXI-1044 embedded controller chassis with 12 data acquisition cards providing analog-to-digital con- version for a total of 80 data channels for signal recording. The measurement signals from the array microphones were indi- vidually fed into the PXI channels through 50 ft (15 m) long microphone cables. A controller onboard the PXI chassis enables synchronization of all the data acquisition cards, pro- viding simultaneous recording across all channels. The soft- ware for running the system in real time and transferring data from the PXI to a laptop computer for post-processing was developed by Dr. William Blake and Wyle. During the proof-of-concept testing described in Section 3.4, one of the PXI channels was used for recording the time signal from an ESE Model ES-292 GPS-based time code generator (IRIG). Another channel received a signal from a pair of photo- cells installed on tripods near the microphone array to regis- ter truck passbys. The photocells were Banner Engineering Model SM31EL/SM31RL mini-beam emitter and receiver. Another PXI channel was used for recording signals from IT’s vehicle tracking system. The system includes a Banner Engi- neering Model Q45BB6DLQ infrared linear position sensor attached to the front bumper of the truck. This photosensor detects white strips painted every 5 ft (1.5 m) along the track pavement and telemeters a series of signals to a remote receiver. The signals from the receiver were fed into the PXI. Truck speed was then determined by the distance between the strips and the time between the voltage pulses received in the signals. 3.3.3 Preliminary Testing Prior to full-range proof-of-concept testing scheduled at IT’s facilities in Ft. Wayne, Indiana, a preliminary testing of the experimental beamforming microphone array was performed 19 Figure 10. Experimental microphone array (note seven additional microphones in the array center).

The preliminary testing results indicated that the array per- formed generally as expected for a truncated assembly. Lacking vertical symmetry, its directivity in both vertical and horizon- tal axes was found reasonable for an array of this size. The study team also concluded that the array setting in both vertical and horizontal directions (level and square) is important for image localization and reducing distortion of the images obtained. Overall, the preliminary testing served as a first practical veri- fication of the experimental beamforming array performance. 3.4 Proof-of-Concept Testing The proof-of-concept tests were conducted during two 1-week periods in July 2006 at IT’s Truck Development and Technology Center (TDTC) in Ft. Wayne, Indiana. The TDTC is a controlled environment with test tracks and shop facilities for working on truck setups. IT provided access to its outdoor testing facilities, coordination with its ongoing testing activi- ties, and a representative sample of trucks selected from the vehicle test matrix. IT also provided a professional driver for the tests and technicians to perform all work on the trucks. The first week of testing was conducted at the low-speed passby sound pad. After a 1-week interval of preliminary data processing and analysis, the second week of tests was con- ducted at the high-speed proving grounds/endurance track. Both test phases were documented using digital videotape recording. Weather at the test sites was monitored to avoid testing during periods of rain or high winds. 3.4.1 Low-Speed Tests The microphone array, data acquisition system, and other equipment were transported to and assembled at IT’s low- 20 Figure 11. Preliminary testing of experimental microphone array (lower section). Figure 12. Loudspeaker image at frequency of (a) 315 Hz and (b) 740 Hz. -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 sx/m sy /m OTO Hz Band levels Re 20muPa, SegNo = 7 23 24 25 26 27 28 29 30 31 32 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 sx/m sy /m OTO Hz Band levels Re 20muPa, SegNo = 7 22 23 24 25 26 27 28 29 30 31 (b) 740 Hz(a) 315 Hz locally in Virginia. The overall goal was to ensure that the microphone array, data acquisition system, and software worked properly prior to the actual proof-of-concept test- ing. The test was performed in a residential outdoor area, on a grassy lawn free of obstructions. Only the lower section of the array with 25 microphones was tested using a Mackie Model SRM450 loudspeaker with a pink noise generator, as well as a lawnmower and a gas-powered string trimmer as noise sources in static positions in front of the array. These noise sources were placed alone and in combinations at varying distances from the array, both at on-axis and off- axis locations relative to the array. The data were then processed with the beamforming software, to make sure that the different sources could be correctly identified. One of the test settings is shown in Figure 11. The resultant images of the loudspeaker at two different frequencies are shown in Figure 12.

speed sound pad designed for conducting standard truck passby noise-emission measurements. At this site, the trucks drive on a single lane with sealed, dense-graded asphalt pave- ment and a maximum speed limit of 35 mph (56 km/h). The microphone array was placed on the asphalt diamond mea- surement pad, parallel to the direction of truck travel, at a dis- tance of 20 ft (6 m) from the edge of the driving lane, as shown in Figure 13. Each measurement channel of the system was calibrated using a Brüel & Kjær Type 4231 acoustic calibrator. Initially a series of stationary loudspeaker tests were per- formed, using a Mackie Model SRM450 loudspeaker and a CESVA Model BP012 omni-directional dodecahedron loud- speaker. Sound from a pink noise generator was played through the speakers at a high volume. The speakers were used for initial evaluation of the fully assembled system as known “point” sources placed at several on- and off-axis locations with different distances and heights in front of the array. One example of such a test setup is shown in Figure 14. After the stationary on-the-ground loudspeaker tests were completed, the CESVA loudspeaker was placed on a truck. 21 Figure 13. Microphone array at low-speed track. Figure 14. Omni-directional speaker in front of microphone array at low-speed track. Figure 15. The 4400 truck with omni-directional speaker. The speaker was secured to the rear frame of an Inter- national® 4400 medium utility truck (with no flatbed; 2002 model year). The truck was powered with an International® DT466 250 hp engine and was equipped with an automatic transmission, a single horizontal under-frame muffler and a horizontal tailpipe. The truck was equipped with four Goodyear G124 low-profile tires on the single drive (rear) axle and two Goodyear G159 tires on the steer (front) axle. The truck with mounted omni-directional loudspeaker is shown in Figure 15. To evaluate the array/beamforming system performance, a number of tests were performed with the truck stationary and moving both with and without the speaker (pink noise signal) turned on. These tests were intended to determine the array’s ability to identify the truck noise sources (e.g., engine, tires, and exhaust) in comparison with the loudspeaker, as the truck passed by the array at a constant speed. Several speeds,

22 Speed (mph) rpm Gear Speaker 0 (stationary) idle Neutral Off/On 0 (stationary) 2000 Neutral Off/On 18 1500 2nd Off/On 25 2000 2nd Off/On 31 2500 2nd Off/On 27 Cruise 4th Off Table 2. Tests of the 4400 truck with speaker at the low-speed track. Table 3. Tests of the 9200i truck at the low-speed track. Speed (mph) rpm Gear Notes 0 Idle Neutral 0 650 Neutral 0 1200 Neutral 0 1800 Neutral Stationary 15 1200 6th 25 1200 7th 30 1200 8th Constant rpm, varied gear 13 650 7th 20 1200 7th 31 1850 7th Constant gear, varied rpm 29 Coast down Neutral 29 Cruise 7th Acceleration 700 7th Figure 16. The 9200i truck at the low-speed track. Figure 17. The 5900 truck in front of microphone array at the low-speed track. revolutions per minute, and gear settings used for the truck runs are summarized in Table 2. The next truck tested at the low-speed track was an Inter- national® 9200i Eagle truck (2006 model year). This is a heavy- duty truck used for long hauls. It was powered with a Cummins ISX 450 hp engine and had a single horizontal muffler and a vertical tailpipe, which was mounted to the rear passenger side of the cab. The truck tractor with no trailer was used, equipped with eight Goodyear G372 tires on tandem drive (rear) axles and two Goodyear G395 tires on the steer (front) axle. The truck was also equipped with aero skirts under the cabin. The truck is shown in Figure 16. The truck tests, performed with no loudspeaker, are summarized in Table 3. The testing then continued with the third truck, an Interna- tional® 5900 PAYSTAR™ (2004 model year), which is a “severe service” truck designed to have other equipment mounted on the frame, such as a concrete-mixer drum and associated equipment, and to be used with very heavy payloads. It was powered with a Cummins ISX 565 hp engine. For these tests, the truck had dual vertical mufflers and tailpipes mounted to the back sides of the cab. Only the truck tractor was driven, with approximately 7000 lbs of weight added on the rear sec- tion of the tractor frame. There were eight Goodyear G167 tires on tandem drive (rear) axles and two Goodyear G287 tires on the steer (front) axle. In this case again, a loudspeaker was not used in the testing. The truck is shown in Figure 17. The truck test conditions are listed in Table 4. The measurement data collected during the testing at IT’s low-speed track were preliminarily reviewed in the labora- tory the following week and determined suitable for full post-processing and analysis. The key results of the analysis are presented and discussed in Section 3.5. The microphone array and data acquisition system performed as expected and required no major adjustments for the next stage of the proof- of-concept testing at IT’s high-speed endurance track. 3.4.2 High-Speed Tests For high-speed testing, the microphone array and data acquisition system were transported to IT’s endurance track. It consists of a 1 mi (0.6 km) long multiple-lane loop designed for conducting long-term truck testing with a maximum speed limit of 50 mph (80 km/h). The microphone array was placed in the sealed asphalt area along the middle part of the track, at

a distance of 20 ft (6 m) from the edge of the nearest driving lane of unsealed, dense-graded asphalt. First, the stationary CESVA omni-directional speaker tests were performed at this location for initial check-out and additional array perfor- mance validation. After that, the truck passby tests were carried out in 3 days. The trucks (all by International®) tested at the high-speed endurance track are listed in Table 5 (not in the order of testing). The International® 4400 truck was the same as the one tested before at the low-speed track. The other trucks tested at the high-speed track were not tested at the low-speed track. The 9200i truck with a 450 hp engine, shown in Figure 18, had front and rear Goodyear G395 tires and a short vertical tailpipe, unlike the truck of the same type tested at the low-speed track (compare with Figure 16). This 9200i truck also was not equipped with aero skirts covering a fuel tank and a horizontal part of the exhaust pipe under the cabin. To test a truck with no muffler, the exhaust system of another 9200i Eagle truck (with 370 hp engine) was modi- fied by replacing its vertical muffler with a pipe, as seen in Figure 19. This truck was also equipped with an aerodynamic wind fairing over the cabin. The International® 5900i tractor truck was tested at the high-speed track in three different configurations: (1) bobtail with eight Michelin XZE tires on the drive axles; (2) bobtail with eight Goodyear G164 RTD aggressive tread, high-traction (noisy) tires on the drive axles; and (3) loaded trailer with the same aggressive tread tires on the drive axles. Figures 20 and 21 show the truck without and with the trailer, respectively. Unlike the truck of this type (5900 PAYSTAR) tested at the low-speed track, the 5900i tested at the high-speed track was equipped with a more powerful engine, but had a single muf- fler and a tailpipe. Tables 6 through 11 summarize the tests performed for each truck at the high-speed track. The tables present the test con- ditions of the trucks, including the speed, engine rpm, and gear. Also shown in the tables is the overall A-weighted sound level (LA) measured with a single microphone at the distances of 25 and 50 ft (7.5 and 15 m) from the track for each truck run, as described in Section 3.4.3. For the stationary tests, LA is the measured time-averaged sound level; for the passby tests, LA is the maximum sound level measured during the single truck run. No tests were performed at this track with a loudspeaker mounted on a truck. 3.4.3 Passby and Intensity Measurements In addition to the array measurements of the test trucks for the stationary and passby truck conditions, conventional single-microphone and sound intensity measurements were conducted. These measurements are described in this section. To complement the beamforming results, single-micro- phone sound pressure level measurements were performed at distances of 25 ft (7.5 m) and 50 ft (15 m) from the centerline of vehicle travel. These microphones were set on a line perpen- dicular to the direction of travel offset 5 ft 8 in. (1.7 m) from the center of the array, as shown in Figure 22. The micro- phones were adjusted to a height of 5 ft (1.5 m) above the pave- ment. The microphones were 0.5 in. (12.5 mm) Larson-Davis (LD) Model 2541 fitted onto 0.5 in. (12.5 mm) LD Model PRM900C microphone preamplifiers. The signals from these microphones were fed into an LD 3000 two-channel real-time 23 Table 4. Tests of the 5900 truck at the low-speed track. Speed (mph) rpm Gear Notes 0 650 Neutral 0 1500 Neutral Stationary 0 2000 Neutral 14 1500 4th 25 1500 5th 28 1500 6th Constant rpm, varied gear 10 650 5th 20 1500 5th 26 2000 5th Constant gear, varied rpm 30 Coast down Neutral 30 Cruise 7th–8th From 14 From 1500 4th Acceleration Table 5. Trucks tested at the high-speed track. Truck Type (model year) Engine Type Front/Rear Tires (no.) Exhaust Configuration 4400 (2002) International DT466 250 hp G159 (2) G124 (4) Single horizontal muffler & horizontal tailpipe 9200i (2006) Cummins ISX 450 hp G395 (2) G395 (8) Single horizontal muffler & short vertical tailpipe 9200i Eagle (2004) Cummins ISM 370 hp G395 (2) G372 (8) Single modified vertical tailpipe with no muffler 5900i (2006) Cummins ISX 600 hp G286 (2) Michelin XZE (8) Single vertical muffler & vertical tailpipe 5900i w/o and with loaded trailer (2006) Cummins ISX 600 hp G286 (2) G164 RTD (8) Single vertical muffler & vertical tailpipe

analyzer. The signals were also recorded on a Sony TCD D-100 two-channel DAT recorder for data backup and any further data reduction required later. For the stationary tests of either the trucks or the loud- speaker, linear time averages of the sound pressure levels were obtained for a 15 s time period. These were analyzed in one- third octave frequency bands from 20 to 20,000 Hz. The average overall A-weighted sound levels measured for the stationary conditions on the high-speed track were given pre- viously in Tables 6 through 11 for both the 25 and 50 ft micro- phone positions. For the passby truck tests, one-third octave band spectra were captured every 0.100 s with a 0.125 s expo- nential (fast response) averaging time applied to the signals for a period of 10 s. These data were then used to examine the over- all A-weighted time history of the passby event, as well as the one-third octave band spectrum at the time of maximum over- all A-weighted level during the passby. The maximum overall A-weighted sound levels measured for the truck passby condi- tions on the high-speed track were also given in Tables 6 through 11. Note that the 25 and 50 ft measurements in Tables 6 through 11 are generally consistent with inverse square law, but there are some variations. Near-field effects associated with source size may be a factor in the variations. This phenomenon is par- ticularly likely for the data in Table 8, where the modified exhaust resulted in a strong exhaust source significantly above the engine source. Near-field effects are a concern for passby measurements with a single microphone. They are not, how- ever, an issue for beamforming measurements which are intended to discriminate among a distribution of sources. For each of the test trucks under no-load stationary condi- tions, sound intensity averaged over areas to the side of the 24 Figure 18. The 9200i truck with short vertical tailpipe at the high-speed track. Figure 19. The 9200i Eagle truck with drag fairing and no muffler at the high-speed track. Figure 20. The 5900i truck with no trailer at the high-speed track. Figure 21. The 5900i truck with trailer at the high-speed track.

25 Table 6. Tests of the 4400 truck at the high-speed track. LA (dBA) at Speed (mph) rpm Gear Notes 25 ft 50 ft 7.261.96lartueN)eldi(0470 7.564.27lartueN05110 6.667.37lartueN00410 0 2100 Neutral Stationary 79.3 72.6 3.474.08evirD001253 50 1400 Drive Automatic transmission 83.6 77.6 35 2100 3rd 7.473.08 50 2100 4th Constant rpm, varied gear 84.3 78.0 From 50 * Throttle off Coast down 80.9 73.3 From 10 to 15 1500 to 2450 2nd Acceleration 81.9 76.2 From 22 to 31 1800 to 2450 2nd Acceleration 83.7 78.7 * Not recorded Table 7. Tests of the 9200i truck at the high-speed track. Table 8. Tests of the 9200i Eagle truck (modified exhaust, no muffler) at the high-speed track. Speed (mph) rpm Gear Notes LA (dBA) at 25 ft 50 ft 0 Idle Neutral ** ** 2.866.47lartueN00110 2.277.87lartueN00410 7.176.87lartueN00510 0 1800 Neutral Stationary 81.1 74.1 35 1400 8th 77.8 72.3 46 1800 8th 84.2 77.9 50 1500 9th 83.0 77.5 50 1100 10th 82.6 76.5 35 2100 3rd ** ** 50 2100 4th Constant rpm, varied gear ** ** From 50 * 9th Coast down 81.4 74.7 From 50 * 9th Compression brake 83.6 77.3 46 From 900 to 1800 8 th Per SAE J366 84.5 78.5 * Not recorded **Missing data Speed (mph) rpm Gear Notes LA (dBA) at 25 ft 50 ft 5.366.86lartueNeldI0 8.276.77lartueN00510 9.476.97lartueN00610 0 1900 Neutral Stationary 81.7 76.7 35 1900 7th 85.2 79.9 35 1500 8th 82.6 76.7 45 1900 8th 89.2 84.3 50 1600 9th 89.6 82.1 From 50 * * Coast down 80.0 75.6 From 50 1400 9th Compression brake 96.2 90.9 * * * Low-speed acceleration 89.2 85.9 * Not recorded trucks was measured. The sound intensity probe used for this purpose consisted of two 0.5 in. (12.5 mm) phase-matched condenser microphones spaced 0.625 in. (16 mm) apart in a side-by-side configuration. The microphones were a G.R.A.S. 40AI intensity microphone pair fitted to LD Model PRM900C 0.5 in. (12.5 mm) preamplifiers. Signals from these micro- phones were input to a LD 3000 two-channel analyzer for immediate sound intensity measurement in one-third octave bands. To determine the average sound intensity for an area in the plane parallel to the side of a vehicle, the probe was manu- ally swept over specific subareas, as shown in Figures 23 and 24, several times as linear averaging was performed over a period

26 Table 9. Tests of the 5900i truck (standard tires) at the high-speed track. Speed (mph) rpm Gear Notes LA (dBA) at 25 ft 50 ft 3.065.66lartueNeldI0 7.862.67lartueN00410 7.479.97lartueN00810 0 2000 Neutral Stationary 81.1 74.2 35 1800 6th 81.0 75.3 35 1400 7th 79.6 73.8 50 2000 7th 84.3 78.5 50 1400 8th 84.7 78.9 From 50 * * Coast down 81.8 76.2 From 50 1400 8th Compression brake 83.3 77.6 * Not recorded Table 10. Tests of the 5900i truck (aggressive tread rear tires) at the high-speed track. Table 11. Tests of the 5900i truck (aggressive tread rear tires) with trailer at the high-speed track. Speed (mph) rpm Gear Notes LA (dBA) at 25 ft 50 ft 35 1800 6th 81.4 76.1 35 1400 7th 79.3 73.7 50 2000 7th 85.5 79.9 50 1400 8th 83.5 77.8 From 50 * Neutral Coast down 81.8 77.1 From 50 1400 8th Compression brake 83.4 78.2 13 to 16 From 1300 4th Per SAE J366 80.5 75.2 * Not recorded LA (dBA) at Speed (mph) rpm Gear Notes 25 ft 50 ft 35 1800 6th 82.7 77.2 35 1400 7th 82.2 76.7 50 2000 7th 88.0 82.8 50 1400 8th 87.2 82.4 From 50 * Neutral Coast down 86.9 81.4 From 50 * Neutral Compression brake 87.4 81.3 50 1400 8th - Full throttle 87.9 82.4 35 1400 8th - Full throttle 84.2 78.5 * Not recorded Figure 22. Positioning of passby measurement microphones. Figure 23. Sound intensity scan areas for the 9200i truck. 50 ft Microphone 25 ft Microphone 3 124 5 6

for the areas indicated in Figure 26. From these data, it is appar- ent that the subareas corresponding to the hood and upper cab produce lower levels of sound power, and the sound intensity is negative at the lower frequencies. The latter is an indication that in these subareas the sound energy flows in the opposite direction (toward the structure rather than from it) because of the presence of the other, more powerful sound source(s) in the vicinity. As a result, for subsequent trucks the hood and upper cab subareas were excluded from the analysis. For the 4400 truck, the sound power contribution was also determined for the loudspeaker, and its relationship to the total for the truck only is given in Figure 27. The sound power breakdown for the stock 9200i truck for the subareas of Figure 23 is shown in Figure 28. For this truck the contributions from the muffler and exhaust outlet are quite 27 Figure 24. Sound intensity scan of muffler (subarea 3) for the 9200i truck. Figure 25. Sound power levels for subareas and total of the 4400 truck at 2000 rpm. 55 60 65 70 75 80 85 90 95 100 10 0 12 5 16 0 20 0 25 0 31 5 40 0 50 0 63 0 80 0 10 00 12 50 16 00 20 00 25 00 31 50 40 00 50 00 1/3-Octave Band Center Frequency, Hz A -w ei gh te d So un d Po w er L ev el , d BA Upper Cab - 2 Hood - 3 Muffler - 4 Lower Cab - 5 Wheel Well - 6 Total Truck Figure 26. Sound intensity scan areas for the 4400 truck with loudspeaker. 5 4 1 3 6 2 5 of 40 s. Once the average intensity was determined for each subarea, the sound power radiated through each surface was calculated. These calculations were then summed to deter- mine the total sound power radiated on the side of the truck facing the array. In this manner, the contribution of each sub- area to the total sound power radiated to the side of the truck was evaluated. This methodology was initially applied to the 4400 truck when tested at the low-speed track with the loudspeaker installed. The results for the truck alone are shown in Figure 25

28 Figure 27. Sound power levels of the 4400 truck and loudspeaker. 55 60 65 70 75 80 85 90 95 100 10 0 12 5 16 0 20 0 25 0 31 5 40 0 50 0 63 0 80 0 10 00 12 50 16 00 20 00 25 00 31 50 40 00 50 00 1/3-Octave Band Center Frequency, Hz A -w ei gh te d So un d Po w er L ev el , d BA Speaker Truck Total Figure 28. Sound power levels for subareas and total of the 9200i truck at 1800 rpm. 50 55 60 65 70 75 80 85 90 95 10 0 12 5 16 0 20 0 25 0 31 5 40 0 50 0 63 0 80 0 10 00 12 50 16 00 20 00 25 00 31 50 40 00 50 00 1/3-Octave Band Center Frequency, Hz A -w ei gh te d So un d Po w er L ev el , d BA Wheel Well - 1 Gas Tank - 2 Muffler - 3 Muffler Area - 4 Exhaust Pipe - 5 Exhaust Outlet - 6 Total low compared to the sound power coming from underneath the truck and through the wheel well. Only at very low frequen- cies of 100 to 160 Hz is there a significant contribution of the exhaust outlet. For the 5900i truck, trends similar to the stock 9200i truck are seen in Figures 29 and 30, except that there is somewhat more contribution from the muffler. For the case of the 9200i Eagle truck with the modified “straight-through” exhaust (no muffler), a significantly different contribution of subareas can be seen in Figures 31 and 32. In this case, radiation from the exhaust pipe still remains low; however, sound radiation from the exhaust outlet subarea dominates the total radiation

between 160 and 500 Hz, with the sound levels 6 to 15 dB greater than any of the other subareas. Even at the higher frequencies, the contribution of the exhaust outlet remains high. Upon completion of testing at IT’s high-speed track, the measurement data collected was post-processed and analyzed in the laboratory. The key results of the analysis are presented and discussed in Section 3.5. Again, the microphone array and data acquisition system performed as expected and required no major adjustments. From experimenting with the number of 29 Figure 29. Sound intensity scan areas for the 5900i truck. 1 23 4 5 Figure 30. Sound power levels for subareas and total of the 5900i truck at 1400 rpm. 50 55 60 65 70 75 80 85 90 95 10 0 12 5 16 0 20 0 25 0 31 5 40 0 50 0 63 0 80 0 10 00 12 50 16 00 20 00 25 00 31 50 40 00 50 00 1/3-Octave Band Center Frequency, Hz A -w ei gh te d So un d Po w er L ev el , dB A Front Wheel Well - 1 Between Fender & Gas Tank - 2 Gas Tank Area - 3 Muffler - 4 Exhaust Outlet - 5 Total Figure 31. Sound intensity scans areas for the 9200i Eagle truck with straight-through exhaust (no muffler). 2 1 3 4 array microphones, however, the study team determined that increasing the number of microphones to 73 or 77, by adding units at and near the center of the array, was beneficial for the array’s beamforming performance and resulting noise source mapping. Also, assembling the array in the field, which involved handling multiple cable connections between the microphones and PXI unit, was found to be labor intensive and time consuming. These aspects of the first field experience were addressed prior to subsequent application of the array for the roadside truck noise measurements.

30 Figure 32. Sound power levels for subareas and total of the 9200i Eagle truck with straight-through exhaust (no muffler) at 1500 rpm. 50 55 60 65 70 75 80 85 90 95 10 0 12 5 16 0 20 0 25 0 31 5 40 0 50 0 63 0 80 0 10 00 12 50 16 00 20 00 25 00 31 50 40 00 50 00 1/3-Octave Band Center Frequency, Hz A -w ei gh te d So un d Po w er L ev el , d BA Front Wheel Well -1 Gas Tank Area - 2 Exhaust Pipe - 3 Exhaust Outlet - 4 Total

3.5 Proof-of-Concept Test Results This section discusses the proof-of-concept test results in four categories: • Beamformer calibrations with the omni-directional (spher- ical) source, in which in-situ frequency response and steer- ing characteristics of the array are established; • Benchmark localization measurements with a moving truck with and without the spherical source on board, in which the array’s ability to discriminate among multiple sources is verified and the rudiments of developing tempo- ral histories during passby are established; • Comparisons with the intensity measurements made on stationary trucks, in which the array’s ability to localize the principal (i.e., largest magnitude) source on the truck and identify the lesser sources is established; and • Example results from low- and high-speed track passbys, in which localization and characteristics of individual noise sources for the tested moving trucks are evaluated. 3.5.1 Beamformer Calibrations with Spherical Source The beamformer characteristics of the array were calculated for comparison to the test results using a theoretical model of the received signals at each array element with a simple source above ground with ground reflection, for which the ground reflection coefficient was assumed to be between 0.7 and 1.0. Figure 33 shows a side-view diagram of the array, spherical sound source and its image in the ground half space, direct acoustic propagation path, and two example ray traces for the reflected path that were used in the analytical model. Fig- ure 34 is an illustration of the coordinate axis orientations of the source and array planes. The latter figure also labels the characteristics to be found in the images that will be discussed in this section. Without going into details yet, note that the ground plane is at y = 0 and the array is positioned above it. The physical sources are all also above ground, with all reflected image sources below the ground plane. These characteristics are shown in Figure 34. All images of a single acoustic source will therefore show a single “hot spot” above ground with its mir- ror image below ground, as illustrated. Grating lobe effects in the array, a coherent (localized) background source, or a non- specular reflection may contribute an artifact at low level, also as illustrated. Figures 35 through 38 show corresponding calculated and measured images of the spherical source at a series of frequen- cies that serve to define the spatial resolution of the measure- ment array. In these figures and all other source images presented in this section, numbers in the color bar legend indi- cate approximately equivalent one-third octave band sound levels in decibels. At 922 Hz, for example, Figure 35 shows elliptical spots whose major and minor axes are complemen- tary to those of the array. The vertical (−6 dB) width of the spot is about 0.38 m (1.2 ft), while the horizontal width is about 0.67 m (2.2 ft) at a range of 5.8 m (19 ft) which corresponds approximately to the side of the truck or the closer wheel track and an elevation of 1.98 m (6.5 ft). Figures 36 and 37 show examples at lower and higher frequencies, respectively, for the same source at this range and elevation. The two low frequen- cies define the lower limit of the array performance. The two high frequencies define the upper limit of the array perfor- mance. These figures show that the array performs adequately between approximately 250 and 2000 Hz. Figure 38 shows good focusing at a longer range of 7.62 m (25.2 ft), which corresponds to the center of the passby lane. 31 Figure 33. Side view of spherical source and its image in the ground half space. Ys Image Ground Plane (GRC) Array Plane Spherical (Point) Source Direct Acoustic Path Reflected Path Figure 34. Array and typical image orientation for spherical source emission. Array Spherical Source Principal Image of Source Artifact Elevation of Source’s Center X Z Y

32 Figure 35. Images of (a) measured and (b) calculated signals for spherical source emission at 922 Hz [source elevation 1.98 m (6.5 ft), array stand-off at road side 5.8 m (19 ft)]. (b) Calculated(a) Measured sx/m sy /m Measured Equivalent Ls, OTO at F=922.8516Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 82 83 84 85 86 87 88 89 90 91 sx/m sy /m Theoretical Equivalent Ls, OTO at F=922.8516Hz and GRC= 0.7 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 82 83 84 85 86 87 88 89 90 91 Figure 36. Images of (a) measured and (b) calculated signals at two low frequencies. (a) Measured (b) Calculated sx/m sy /m Measured Equivalent Ls, OTO at F=276.8555Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 88 89 90 91 92 93 94 95 96 97 sx/m sy /m Theoretical Equivalent Ls, OTO at F=276.8555Hz and GRC= 0.65 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 88 89 90 91 92 93 94 95 96 97 sx/m sy /m Measured Equivalent Ls, OTO at F=230.7129Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 91 92 93 94 95 96 97 98 99 100 sx/m sy /m Theoretical Equivalent Ls, OTO at F=230.7129Hz and GRC= 1 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 91 92 93 94 95 96 97 98 99 100 231 Hz 277 Hz

1ST REVISE Figure 38. Images of (a) measured and (b) calculated signals for spherical source emission at 922 Hz [source elevation 1.98 m (6.5 ft), array stand-off at road side 7.62 m (25 ft)]. (a) Measured sx/m sy /m Measured Equivalent Ls, OTO at F=2122.5586Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 76 77 78 79 80 81 82 83 84 85 sx/m sy /m Theoretical Equivalent Ls, OTO at F=2122.5586Hz and GRC= 0.7 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 76 77 78 79 80 81 82 83 84 85 sx/m sy /m Measured Equivalent Ls, OTO at F=1707.2754Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 81 82 83 84 85 86 87 88 89 90 sx/m sy /m Theoretical Equivalent Ls, OTO at F=1707.2754Hz and GRC= 0.7 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 81 82 83 84 85 86 87 88 89 90 1707 Hz 2123 Hz (b) Calculated (a) Measured sx/m sy /m Theoretical Equivalent Ls, OTO at F=922.8516Hz and GRC= 0.7 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 81 82 83 84 85 86 87 88 89 90 sx/m sy /m Measured Equivalent Ls, OTO at F=922.8516Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 81 82 83 84 85 86 87 88 89 90 (b) Calculated Figure 37. Images of (a) measured and (b) calculated signals at two high frequencies. 33

Figure 39 allows assessment of the array’s actual measured res- olution performance compared with the predicted design per- formance. The measured values for the −6 dB major and minor axes coordinates of the spots at each frequency are shown as points, while the calculated values are shown as lines. This fig- ure illustrates that the acoustic focusing performance of the array is substantially as predicted. The array’s ability to steer is illustrated in Figures 40 and 41. The steering at elevation of 1.24 m (4 ft) and an offset of 7.62 m (25.2 ft) is shown in Figure 40 looking ahead and at 21.8 degrees to the right [xs = 3.05 m (10 ft)]. Similarly, Figure 41 shows, for an elevation of 1.98 m (6.5 ft), the array’s ability to steer 45 degrees off axis [see lower images for xs = 5.8 m (19 ft) at an offset of 5.8 m (19 ft)]. These illustrations show that, for a spherical source, the array reliably images an omni-directional source at steering angles up to 45 degrees off axis, although at this angle, compared with 21.8 degrees (upper images), there is a slight parallax. As seen in the lower image of Figure 41(a), the source is localized about 1.5 m (5.9 ft) closer than actually located, with the maximum emission measured at a horizon- tal cross range of xs = 4.3 m (14.1 ft) rather than the actual xs = 5.8 m (19 ft). This discrepancy does not occur at 21.8 degrees. Also, note that the measured spot sound levels (84 to 91 dB) are within approximately ±3 dB of one another in all of these images at 922 Hz. Because the purpose of these measurements was to test localization, not to calibrate array levels, no attempt was made to precisely set the source to the same sound level for each of these measurements. This allowed the sound levels for all runs to be set to a nominal setting, which resulted in certain mea- surement variability. Thus, in summary and within measure- ment repeatability, the array steering provided accurate localization and amplitude measurement to within a few deci- bels at steering angles approaching 45 degrees. In other words, when scanning the region of −6 to +6 m (−20 to +20 ft) in cross range and −6 to +6 m (−20 to +20 ft) in elevation at a stand-off of 6 m (20 ft) from the track, the array will localize and accu- rately define the magnitude of a stationary source. However, as the following subsection will discuss, truck noise sources are more complex, and that complexity is apparent in the images. Finally, note that with the spherical source at a range of 20.2 m (66 ft) the array still performs well, as shown in Figure 42. 3.5.2 Benchmark Measurements of Spherical Source on Moving Truck with Competing Truck Noise As a further test of the array steering, with the additional measurement complexity of the moving source and competing truck sources, the omni-directional loudspeaker was strapped to the bed of the 4400 truck on the low-speed track, as was shown in Figure 15. Figure 43 mimics Figure 34 but with the coordinate system and array overlaying an appropriately scaled photograph of a truck (in this case, the 9200i truck) stationed at the closest point of approach (CPA). As with the calibration source, y = 0 corresponds to actual ground. Figure 44 shows the image of the 4400 truck stationary and idling with the spherical source mounted on the truck’s bed and activated. The speaker spot and its ground reflection can be seen clearly, as well as engine noise escaping through the wheel well and also by ground reflection. Note that in this case the engine noise reflection image appears about 0.5 m (1.6 ft) below the ground plane. Figure 45 shows the image of the truck moving to the right at 25 mph (40 km/h), still with the speaker activated. The spheri- cal source and its reflected image are just as clearly represented, 34 Figure 39. Measured and calculated total spot width (–6 dB) at broad side for elliptical array at 6 m range. Frequency, Hz

but now with a little horizontal smearing as the result of the truck motion over the measurement time. The engine noise seems to have shifted in its directionality, now being propagated via ground reflection, for which the reflection image appears to be again at approximately −0.5 m (−1.6 ft). Details of the time resolution governances during passby will be discussed in Sec- tion 3.5.4.1, as these can affect the noted smearing. In Figure 46, the truck is now passing to the left at 25 mph (40 km/h) with the speaker deactivated. Now no source appears on the truck’s bed, yet engine noise is still apparent, although with a new propaga- tion path. The comparison of Figures 45 and 46 is a simple illus- tration of a reoccurring facet of the truck sound: variability with engine/fan operating condition and right-left asymmetry in the source distribution and propagation paths. 3.5.3 Benchmark Parallel Array-Based and Acoustic Intensity Measurements for Stationary Trucks Three trucks with different engine equipment line-ups were examined with both acoustic intensity measurements and array-based measurements in order to develop a fundamental recognition base for defining the truck sources using the array. In the following paragraphs, the relative intensity levels and 35 Figure 40. Images of (a) measured and (b) calculated signals for spherical source emission at 922 Hz for two cross-range source locations xs = 0 and xs = 3.05 m [source elevation 1.24 m (4 ft), offset 7.62 m (25 ft)]. xs=0 xs=3.05 m (a) Measured sx/m sy /m Theoretical Equivalent Ls, OTO at F=922.8516Hz and GRC= 1 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 75 76 77 78 79 80 81 82 83 84 sx/m sy /m Measured Equivalent Ls, OTO at F=922.8516Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 75 76 77 78 79 80 81 82 83 84 sx/m sy /m Theoretical Equivalent Ls, OTO at F=922.8516Hz and GRC= 1 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 78 79 80 81 82 83 84 85 86 87 sx/m sy /m Measured Equivalent Ls, OTO at F=922.8516Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 78 79 80 81 82 83 84 85 86 87 (b) Calculated

Figure 41. Images of (a) measured and (b) calculated signals for spherical source emission at 922 Hz for two cross-range source locations [source elevation 1.98 m (6.5 ft)]. xs=3.05 m, zs=7.62 m (21.8 deg.) sx/m sy /m Measured Equivalent Ls, OTO at F=922.8516Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 77 78 79 80 81 82 83 84 85 86 sx/m sy /m Theoretical Equivalent Ls, OTO at F=922.8516Hz and GRC= 1 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 77 78 79 80 81 82 83 84 85 86 xs=5.8 m, zs=5.8 m (45 deg.) sx/m sy /m Measured Equivalent Ls, OTO at F=922.8516Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 79 80 81 82 83 84 85 86 87 88 sx/m sy /m Theoretical Equivalent Ls, OTO at F=922.8516Hz and GRC= 1 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 79 80 81 82 83 84 85 86 87 88 (a) Measured (b) Calculated (a) Measured (b) Calculated sx/m sy /m Theoretical Equivalent Ls, OTO at F=922.8516Hz and GRC= 1 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 69 70 71 72 73 74 75 76 77 78 sx/m sy /m Measured Equivalent Ls, OTO at F=922.8516Hz -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 69 70 71 72 73 74 75 76 77 78 Figure 42. Images of (a) measured and (b) calculated signals for spherical source emission at 922 Hz for 20.2 m (66 ft) array stand-off at road side [source elevation 1.98 m (6.5 ft)].

37 Figure 43. Illustration of the array and truck coordinate orientations during passby event at the instant of CPA. Array Y X Z relative array-based acoustic levels are compared to establish the credibility of the array in rank-ordering the truck sound sources. Note that in all cases to be discussed, intensity levels are in one-third octave frequency bands, while the source images are generally in substantially narrower bandwidths: about 64 Hz constant bandwidth is used with the levels adjusted to provide approximate one-third octave band equivalent sound levels. The array-based and intensity meas- urements are essentially different and yield identical results only under the ideal conditions of a single-path plane incident wave and a single source. As implemented here, the intensity measurement is made over a plane and provides the average acoustic power which passes across that plane. The planar array captures sound pressures of all arriving ray bundles from the distribution of sources in front of it and, by implementing inter-element phase delays, “localizes” the regions where the rays appear to converge. These regions are called the “sources.” Actual directivity patterns of the sources and the existence of multiple acoustic paths (e.g., ground reflection) could cause discrepancies. Figure 47 shows images at four frequencies for the sta- tionary 4400 truck with the engine set to 2000 rpm, cooling fan running and the spherical source activated. At 922 Hz, as noted previously, the engine compartment sound trans- mitted through the wheel well, lower cab, and by ground reflection is nearly of equal magnitude to that generated by the spherical source. The spherical source dominates all truck sources at 231 Hz and 600 Hz. At 1937 Hz, the contri- butions to the sound are from the spherical source and from beneath the cab, but the side lobe effects start to contami- nate the source map. These observations agree with Figure 27 in showing complete dominance of the sound by the spherical source below 630 Hz and by the engine noise con- tributions within 10 dB of the spherical source levels at frequencies above 1000 Hz. Table 12 summarizes these results. The sound power levels (PWL) relative to 10−12 W Figure 44. Image at 922 Hz of the 4400 truck idling stationary with engine at 2000 rpm and spherical source activated.

38 6 4 2 0 sy /m sx/m –2 –4 –6 83 82 81 80 78 79 77 76 75 74 6420–2–4–6 Figure 45. Image at 922 Hz of the 4400 truck traveling to the right at 25 mph with engine at 2000 rpm and spherical source activated. 6 4 2 0 sy /m sx/m –2 –4 –6 72 71 70 69 67 68 66 65 64 63 6420–2–4–6 Figure 46. Image at 982 Hz of the 4400 truck traveling to the left at 25 mph (40 km/h) with engine at 2000 rpm and deactivated spherical source.

39 Figure 47. A series of images for source distribution of the 4400 truck stationary opposite the array with engine at 2000 rpm and spherical source activated. f = 231 Hz f = 600 Hz f = 922 Hz f = 1937 Hz Table 12. Image sound pressure levels and intensity levels (dB) for the stationary 4400 truck with engine at 2000 rpm and spherical source. Frequency (Hz) Spherical Source Wheel Well Lower Cab Image Intensity Image Intensity Image Intensity Image Intensity 231 250 95 96 <86 87 <86 86 600 630 89 91 <80 87 <80 85 922 1000 80 84 81 85 76 83 1430 1251 to 1600 79 85 77 85 76 83 1938 2000 84 85 77 85 82 83

40 Figure 48. A series of images for source distribution of the 9200i Eagle truck with “straight-through” exhaust (no muffler) and engine at 1500 rpm, stationary opposite the array. f = 415 Hz f = 645 Hz f = 923 Hz f = 1523 Hz are presented in the table at the one-third octave band cen- ter frequencies. The array-measured image sound levels are presented as equivalent one-third octave band levels using the relationship where G(f) is the 1 Hz spectrum level, p0 is the reference pres- sure of 20 µPa, 0.233f is the bandwidth of a one-third octave band at frequency f, and the frequencies are selected to be within the one-third octave frequency bands of the compared intensity levels. The relative levels Ls and PWL should be equivalent under the ideal conditions noted above, and for the spherical source they essentially are very close. With only one exception, L G f f ps = ( )( )10 0 233 110 02log . ( )• the PWL and Ls are within 4 dB in their absolute values for the spherical source. The relative rankings of the levels by both measurement techniques, from the spherical source to truck sources, are generally consistent. Note that in these images, the color scale range is 10 dB in order to avoid the appearance of array side lobes which would contaminate the image. The array side lobes (normalized to a main response of unity) are calcu- lated to be approximately a fraction (0.2 to 0.25) of the main lobe (roughly −14 to −12 dB), as shown in Figure 3(c). For the stationary 9200 truck, the images show significant low-frequency energy being emitted from above the cab, as seen in Figure 48. No spherical source was installed on this truck. Its vertical exhaust muffler was replaced with a pipe (“straight-through” exhaust) before the measurement, and this

41 Table 13. Image sound pressure levels and intensity levels (dB) for the stationary 9200i truck with engine at 1500 rpm, without muffler. Frequency (Hz) Exhaust Wheel Well Lower Cab Image Intensity Image Intensity Image Intensity Image Intensity 415 400 76 81 <66 73 <66 73 645 630 76 84 72 (ref) 80 67 77 922 1000 72 85 80 85 76 81 1523 1600 73 84 72 (ref) 84 69 (ref) 77 1938 2000 69 84 71 83 62 77 modification made the exhaust noise prevalent at low frequen- cies. At these frequencies, the location of the source is clearly directly above the cab, not at the exhaust pipe opening. The aerodynamic fairing was open at the back facing the exhaust pipe, which likely excited acoustic volume resonances in the fairing cavity thus amplifying the sound and shifting the source to the fairing. This will be discussed further when the results of the passby for this truck are examined. Noise by the vari- ous paths from the engine compartment appears dominant at 923 Hz. Direct radiation from the exhaust, rather than from the exhaust-excited hollow wind fairing appears equally important to the engine noise through ground reflection at 1523 Hz. Note that in the images, the ground-reflected engine noise sources again appear about 0.5 m (1.6 ft) below ground, similar to the location seen with the 4400 truck. The sound power levels presented for this truck in Figure 32 generally cor- roborate these results by showing dominant exhaust noise at frequencies below 800 Hz. Above this frequency, exhaust and wheel-well noises are generally of comparable order. Table 13 compares the sound levels discussed for this truck, with the notation “(ref)” in the cases where the levels at the array are by ground reflection. Finally, the 5900 truck with the engine set to 1,400 rpm and the cooling fan running provides a relatively demanding com- parison because no single source is dominant. Rather, sound emanates from the engine compartment via multiple paths that are all of relatively similar importance. Figure 49 shows these resolved direct and reflected-path sources. Here again, the locations of ground-reflected sources appear 0.5 to 1 m (1.6 to 3.3 ft) below ground. This location indicates that, in general, engine compartment sources via ground reflection appear at positions −1 m < y < −0.5 m (−3.3 ft < y < −1.6 ft) depending on truck and frequency: the higher the frequency, the closer the reflected sources to the ground surface. This general observa- tion may provide an important distinction between engine and tire noise sources because the latter should occur at the road surface and appear in the zone within approximately −0.5 m < y < +0.5 m (−1.6 ft < y < +1.6 ft). Figure 30, which resulted from the sound intensity measurements described in Section 3.4.3, shows that the wheel well, gas tank area (below the cab door), and the region between the gas tank and the wheel well all contribute as pathways for engine sound. The array was unable to discriminate among these paths except at 1799 Hz, because these pathway “sources” are generally rather distrib- uted and the array horizontal beam is too broad (the beam width is greater than 6 ft (1.8 m) at −6 dB below 1000 Hz, see Figure 39) except at high frequencies. Table 14 quantifies these comparisons with reasonable agreement in the image sound level and nearly equal relative ranking of “sources” by the sound intensity measurements. Further studies in this area might suggest some potential source- or path-targeting treatments. 3.5.4 Example Results from Low- and High-Speed Track Passbys 3.5.4.1 Analysis Technique for Low- and High-Speed Track Passbys For passby measurements, the arrangements such as the example photographed in Figure 22 were used, with photo- cells placed on either side or both sides of the array to mark the instant of the front bumper passing a known point in the passby track. Figure 50 illustrates the geometric details, where L1 and L2 denote optional locations of the photocells relative to the array for determining truck position. Differ- ent locations were experimentally examined and used on the low- and high-speed tracks, although positions were fixed in each case. The cell that indicated approach (i.e., to the left with a rightward approach), used at the high-speed track, was found to give the best timing. The length L1 for the first cell was 5 ft (1.5 m) and the length L2 for the second cell was about 20 ft (6 m), as used at the high-speed track. The first cell was always used to mark bumper position in the approaching truck, which was desired as close to the array center as possible without interfering with the micro- phones. The second cell was used primarily to infer the aver- age speed of truck passby during maneuvers such as braking

truck interrupted the light ray, which provided an immedi- ate projection of the truck length onto the time record as shown in Figures 50 and 51. This feature is useful in inter- preting the images. The collected data runs were longer than needed and, to conserve computer memory, the samples were truncated once the passby photocell signal was examined. Typical truncation limits are shown in Figure 51, in which the first and last sec- onds of data are discarded (shaded areas). In this case, the time to the CPA is 1.25 s after eliminating the first second of the record as indicated by the instant the bumper crosses the photocell signal. The retained data window shown in this exam- ple is 3 s; typically, 2 to 3 s were retained, depending on the speed and length of the truck. With the retained data, the truck can be tracked as it approaches and recedes. The increasing and decreasing sound level is due to range (distance) change, as illustrated in Figure 52 for the same run. In this case, for the 42 Table 14. Image sound pressure levels and intensity levels (dB) for the stationary 5900i truck with engine at 1400 rpm. Frequency (Hz) Wheel Well Gas Tank Region/ Lower Cab Image Intensity Image Intensity Image Intensity 461 500 72 76 * 76 830 800 78 81 * 82 1245 1250 81 83 74 83 1799 2000 74 80 69 81 * Included in wheel well Figure 49. A series of images for source distribution of the 5900i truck stationary opposite the array with engine at 1400 rpm. f = 461 Hz f = 830 Hz f = 1245 Hz f = 1799 Hz and acceleration. It was also used to confirm timing and position. To ensure capturing a complete cab plus trailer length between cells, the second cell was placed further down track relative to the array center than the first. The photocell signal provided a step function as the passing

43 Figure 50. Diagram of a typical passby geometry. L=Vs t Array Center L1 L2 Vehicle Path Front Bumper Rear Bumper 6 m offset to closest wheel track Photo Cell Photo Cell Figure 51. Example of a passby signal record. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Bumper Rear Truck transit time across array centerRetained Passby data Figure 52. Range from array center to frontbumper reference as a function of time during passby. 0 0.5 1 1.5 2 2.5 3 6 8 10 12 14 16 18 20 tr,R Time, s R an ge , m 3 s retained, the range started at about 15 m (49 ft), closed to 6 m (20 ft) at the CPA, and then opened to 19 m (62 ft). As part of the data analysis, the range from the array center to the front bumper is calculated as reference at each sample time. Because the acoustic sources on the truck are distributed along the wheel base and otherwise are not known a priori, the range correction approximated to the bumper as a reference. Thus, for any actual source position the actual range correc- tion could be larger or smaller than calculated. An example of corrected and uncorrected sound pressure autospectra for one microphone in the array is shown in Figure 53. The two- dimensional displays of autospectra for a representative sensor in the array are shown in color bar in 1 Hz bands relative to 1 Pa for the passby of Figures 51 and 52. The record is broken into 14 segments of time (the last 13 are plotted) with the range cal- culated for each. The autospectrum at the top of the figure is uncorrected for range and shows a maximum level at about 1.65 s into the run. At any frequency, the sound level varies by about 15 dB through the run. When range-corrected in the manner described previously, the resulting corrected spectrum, shown at the bottom of the figure, demonstrates less variabil- ity. There is over-correction at the beginning and end of the record, however, owing to the uncertainty in the actual range and to the possibility that the tails of the record near 0 and 3 s are contaminated with some background noise. There is a hint of contamination in approximately the first 1.2 s of the overall run, where the rise with time is less steep than later in the run. Of course, once the sources have been localized, it is straightfor- ward to recalculate the range correction to obtain a more pre- cise pressure level history. This iterative step has not been taken because this demonstration was made in conjunction with stan- dard passby measurement procedure that provides absolute A- weighted sound level and the array provides a relative breakout of the contributions with only approximate overall level.

The procedure for analyzing the passby starts with deter- mining the resolution at various times during the record. To develop a temporal gradient with multiple segments of ade- quate statistical sampling and frequency resolution, a balance must be made among the number of Fast Fourier Transform (FFT) frequencies (nfft), the number of ensembles (noens), and the number of time segments along the record (nofiles). Thus, if the total sample size is N (in the example case, 3 s taken at 6000 samples per second), then In the majority of cases examined during passby, typical values of these parameters are nfft = 128, noens = 10 to 15, nofiles = 15 to 20 for total sample sizes N up to about 20,000. Fewer nofiles allows a greater number of ensembles or greater frequency resolution, but it gives a much longer time segment that averages over too much of a passby. Too high a value of nfft provides a narrower bandwidth and associated spectral variability. As a practical matter, given the trade-off between some spectral smoothing and some time resolution desired, the above-noted parameter values gave the best results. 3.5.4.2 Passby Evaluations of the 5900i Truck: Localization of Engine Compartment and Tire Noise Passby measurements with the 5900i truck at 50 mph (80 km/h) show the ability of the array to discriminate among various propagation paths of engine noise and tire noise. Fig- ure 54 shows truck noise images at 868 Hz for a passby speed of 50 mph with the engine at 1400 and 2000 rpm at an instant N = nfft noens nofiles• • ( )2 opposite to the vertical array axis “A.” A few “spot” areas are indicated in the images. The areas designated “F” are located at the front of the truck, just behind the front wheel and below the ground plane. The source levels associated with these spots differ by roughly 6 dB. The spot labeled “T” in the 1400 rpm run is located near the road surface plane and appears to be due to the rear tires. The streaked spots labeled “D” lie below the ground plane between the front and rear wheels and appear to be within 1 dB for the two runs. These sources could be exhaust or drive-train noise. As noted previously, sound generated in the engine compartment and elsewhere on the truck, when reflected off the ground, appear to the array as sources below ground by 0.5 to 1 m (1.6 to 3.3 ft). Compari- son of the lower image in Figure 54 with the images in Figure 49 for this truck stationary—with no tire or drive-train noise and the engine operating at 1400 rpm—shows that, for simi- lar frequencies, the levels just behind the front wheel are about 75 to 77 dB in each case. During passby at 50 mph (80 km/h), however, the other sources dominate over the engine noise. At 2000 rpm, the engine noise now competes with the other sources, having increased to about 83 or 84 dB. This increase is consistent with the observation of sources associated with this forward wheel-well acoustic propagation path. Images at other frequencies for the engine speed of 1400 rpm are shown in Figure 55, which illustrates the localization of sources at the front and rear tires. As noted previously in the dis- cussion of the array calibrations (e.g., Figure 40), the array will successfully localize sources while steering off axis, but the pro- jection of the sound field onto the microphone plane further elongates and slightly rotates the appearance of the point spread spot function. Keeping these image features in mind, clear con- 44 0.5 1 1.5 2 2.5 2000 4000 6000 8000 10000 tk,log10(Fk),10*log10((GG5( :,:))) -70 -60 -50 -40 -30 0.5 1 1.5 2 2.5 2000 4000 6000 8000 10000 tk,log10(Fk),10*log10((GG5corr( :,:))) -70 -60 -50 -40 -30 Fr eq ue nc y, H z Time, s (a) (b) Figure 53. Two-dimensional displays of autospectra for a representative sensor in the array: (a) uncorrected; (b) range-corrected.

45 Figure 54. Source distribution at frequency of 868 Hz of the 5900i truck moving to the right at 50 mph with engine at (a) 2000 rpm and (b) 1400 rpm. F F T D (a) Engine at 2000 rpm (b) Engine at 1400 rpm Figure 55. Source distributions for the 5900i truck moving to the right at 50 mph with engine at 1400 rpm at frequencies: (a) 695 Hz, (b) 868 Hz, and (c) 1346 Hz indicated in (d) the truck noise spectrum. Arrows indicate locations of frequencies (a) (b) (d)(c)

tributions from the tires are seen at 695 and 1346 Hz; this oper- ating point [lower engine speed and 50 mph (80 km/h) vehicle speed] and these frequencies were selected to minimize the influence of the engine and maximize that of the tire noise. The directional nature of these sounds is illustrated in Fig- ure 56 for the same truck speed of 50 mph but the engine at higher rpm of 2000. At the top of the figure is a pair of images at 868 Hz from a single passby, but with the truck at two dif- ferent positions relative to the array. The line labeled “A” in this figure denotes the location on the truck of the array main response axis, so that the tires are directly opposite the array on Figure 56(a) and just to the right of the array on Figure 56(b). Thus, Figure 56(b) shows just behind the front tire, while on Figure 56(a) the perspective is to the side of the tire. These two images lead to the conclusion that the tire noise is directive slightly to the rear, say about 20 degrees off the per- pendicular to the truck side-plane. Similarly, in these views, the sound from rear tires appears to be directive forward. The images also suggest rearward directivity of the sound from the engine compartment: see the relative levels of the ground reflection in the upper images of the figure. Figure 56(c) was obtained during a passby for this truck on the following day but with different drive axle tires (G167), which are known to be noisier from other research (23, 29). The proof-of-concept testing cannot confirm, however, those results. 3.5.4.3 Passby Evaluations of the 9200i Truck: Localization of Engine Compartment and Exhaust Noise Figures 16 and 19 show the two body styles of the 9200i series truck that were tested. The style tested at the low-speed 46 (a) (b) (c) Figure 56. Source distributions for the 5900i truck moving to the right at 50 mph with engine at 2000 rpm. Frequency, Hz Figure 57. Sound autospectra for different tested models of the 9200i truck at 31 mph with engine at 1850 rpm (dots denote frequencies of images below). track (Figure 16) had a side skirt below the cab and a raised roof at the cab top. The version tested at the high-speed track (Figure 19) had no skirt and an aerodynamic fairing with an open back was set on top of the cab. This truck was tested with a standard vertical muffler installation and with the muffler replaced with a pipe. The data presented in the fol- lowing paragraphs were collected for the truck speed of 30 to 35 mph with the engine at 1850 to 1900 rpm. Line plots of autospectra at the time for which the maxi- mum overall sound pressure level was recorded are shown in Figure 57 for each of the three passbys for these truck mod- els. Except for some details in the spectra, the skirted and

47 (a) (b) (c) Figure 58. Source distribution at 665 to 668 Hz for various models of the 9200i truck with muffler traveling at 31 to 35 mph. unskirted trucks with a muffler installed provided essentially comparable spectra. With the muffler removed, however, the measured sound pressure levels increased substantially at fre- quencies below 800 Hz. This frequency range is marked by the occurrence of a series of tones (peaks) at multiples of roughly 90 to 100 Hz: this general periodicity is clear in the sample spectrum shown for the truck with no muffler in Figure 57. Images of the noise sources of these truck models are pre- sented in Figures 58 through 62 for three frequencies denoted by the dots in Figure 57: 655 to 668 Hz, 709 Hz, and 975 to 980 Hz. These frequencies were selected on the basis of their proximity to the approximately 700 Hz harmonic of the tones. Of note in Figure 58, at 665 to 668 Hz all trucks with a muffler installed, whether traveling to the right [Figure 58(a) and 58(b)] or to the left [Figure 58(c)], have the same char- acteristic of combined engine and tire noise, with engine noise apparently dominant. This behavior is similar to that in the 975 to 980 Hz range, as shown in Figure 59. Propagation of sound through the wheel well and around the skirt via ground reflection (in both frequency ranges) appears to gen- erally characterize the natures of the received sound levels. Although the source distributions in rightward [Figures 58(b) and 59(b)] and leftward [Figures 58(c) and 59(c)] travel are both localizable to the lower cab, the detailed source distributions lack symmetry, as discussed previously for the 5900i truck. With the muffler removed, the expected localization of sound from the region of the exhaust opening is apparent in Figures 60 to 62 for the three frequencies, where the sound source maps for the truck with a muffler are shown on the left side of each figure and labeled (a), and the source maps for the truck without a muffler are on the right and labeled (b). At all three frequencies, the ground reflection of the exhaust noise is perfectly clear. Tire noise at the rear tire is apparent in Figure 60(a) with the sound level of 79 dB, but is only 75 to 76 dB in Figure 61(b). The engine noise or forward tire noise in the two truck runs are at the sound levels of 82 and 79 dB, respectively.

The main point of these figures, however, is the nature of the exhaust noise. As noted previously, exhaust noise is marked by high tonal content and the source is localized just forward of the stack—above the cab. The sound is conjectured to actually be amplified by the aerodynamic fairing, which acts as a res- onator of the noise emitted from the stack. The source is per- haps a series of volume resonances of the hollow fairing cavity excited by a primary exhaust efflux source. The fundamental half-wave resonance of a 6 ft (1.8 m) cavity would be at about 90 Hz, consistent with the observed approximate lowest peak frequency of the exhaust noise autospectrum (Figure 57). The sound from the stationary truck was also noted to have these features, as discussed previously for Figure 48. Note that the foregoing discussion for this truck is concerned primarily with evaluation of the engine compartment and exhaust noise. In Figures 59 through 62, the color bars indicat- ing the measured sound levels show a range of approximately 10 dB for each image; however, the ranges are of various values for images (a) and (b) in order to illustrate clearly the differ- ences between these noise sources. As a result, the other sources that remain nearly the same in both compared cases, such as the tire–pavement noise in Figures 60 and 61, appear to be of much different strength if judged simply by the image color rather than by the actual values of the measured sound level. 3.5.4.4 Evaluations of the Truck Acoustic Source Level During Passby as a Function of Vertical Elevation The acquired acoustic data together with the recording of the exact truck position with time during the passby can be used to generate a map of the sound level as a function of ver- tical elevation in the truck plane. This map provides a time his- tory of the vertical density of sources at each time increment during the passby. The time history can then be readily inter- preted by relative motion as location along the truck. The cal- 48 Figure 59. Source distribution at 975 to 980 Hz for various models of the 9200i truck with muffler traveling at 31 to 35 mph. (a) (b) (c)

49 (a) (b) (a) (b) Figure 60. Source distribution at 665 Hz for the 9200i Eagle truck: (a) with muffler and (b) without muffler. Figure 61. Source distribution at 709 Hz for the 9200i Eagle truck: (a) with muffler and (b) without muffler.

culation can be done for each frequency giving a data set of sound pressure level as a function of time during passby (or location along the truck), elevation in the truck plane, and fre- quency. This type of processing was fully developed for analy- sis of roadside measurements. During proof-of-concept testing, preliminary examples were generated for the 4400 truck with and without the onboard spherical source activated. To per- form the calculation the array beam was trained to x = 0 (the horizontal position dead ahead) and to a series of vertical posi- tions y = yindex above the road surface, and acoustic spectra was obtained as a function of time increment during the passby for each elevation. This procedure gives a time–frequency record of all sound arriving at the array center from each of the tar- geted locations on the truck. The process is repeated for the sequence of values of y = yindex up to a maximum that, in this case, was set to about 4 m (13 ft) above the surface of the road. The increment for indexing is about 0.5 m (1.6 ft) and a total of nine vertical positions are used. Figure 63 shows a time record of the vertical scans for the stationary, idling truck with its spherical source activated. The source was 4 ft (1.2 m) above the road surface. Because the truck is stationary, the source levels at the different elevations are essentially constant except for possible small propagation variability associated with the outdoor acoustic field. These vertical scans show a concentration of source level in the y = 1 m (3 ft) pixels with markedly reduced levels in the y = 2 m (6.5 ft) (and greater) pixels. The vertical discrimination is smaller for the levels at 1406 Hz than at 937 Hz, given the beam widths plotted in Figure 39 [i.e., ∼3 m (10 ft) and ∼2 m (6.5 ft) for the frequencies of 937 Hz and 1406 Hz, respec- tively]. Also, noting the image plot for this data set in Figure 44, when the array is trained to y = 0, the reflection from the road surface is included in the aperture of the array, espe- cially at 900 to 1000 Hz. Thus in Figure 63(a) there is little variation between levels in the pixels at y = 0 and 0.5 m (0 and 1.6 ft), while there is noticeable difference in Figure 63(b) for which the total beam width is only 2 m (i.e., ±1 m) [6.5 ft (i.e., ±3.25 ft)]. Given that the truck is stationary, the source level distribution is constant over time and appears as a stripe along the y = 1 m (3 ft) pixels. The frequencies discussed in these figures are slightly dif- ferent than those discussed in Figures 44 through 46 because a narrower bandwidth was used here for which the FFT gives narrower analysis bands. The vertical scans were used for the roadside measurements to provide vertical distributions of A-weighted one-third octave band levels using summations of narrower bandwidth spectra, so the data presented here are preliminary to that application. By bracketing the 922 Hz point of Figures 45 and 46, say, differences between 922 Hz and 937 Hz for this truck were determined to be negligible. For the 4400 truck passing to the right at 25 mph (40 km/h) with the onboard source activated, Figure 45 shows an image that localizes the spherical source at a point about 4.6 m (15 ft) behind the bumper. Thus, when the truck is vertically scanned during passby in the manner described previously, the map- ping of levels that is shown in Figure 64 is obtained. In this case, 50 (a) (b) Figure 62. Source distribution at 975 Hz for the 9200i Eagle truck: (a) with muffler and (b) without muffler.

51 (a) f = 937 Hz (b) f = 1406 Hz Figure 63. Vertical scan for the 4400 truck stationary, the engine at 2200 rpm, and the spherical source activated at frequency: (a) 937 Hz and (b) 1406 Hz. The time axis is in seconds relative to an arbitrary reference. Figure 64. Vertical scan at 937 Hz for the 4400 truck passing to the right with the engine at 2200 rpm and the spherical source activated. The time axis is converted to bumper position relative to the array center. Truck photo is reversed on the horizontal scale with its ~6 m (20 ft) wheel base indicated. Vertical axis of the acoustic scan is expanded with the 1 m mark on the truck indicated with the line.

52 Figure 65. Vertical scan at 937 Hz for the 4400 truck passing to the right with the engine at 2200 rpm and the spherical source deactivated. The time axis is converted to bumper position relative to the array center. The truck photo is reversed on the horizontal scale with its ∼6 m (20 ft) wheel base indicated. Vertical axis of the acoustic scan is expanded with the 1 m mark on the truck indicated with the line. the time scale has been interpreted for the reader as position along the truck. (The scale has also been reversed, so the image is presented as if the truck were passing to the left.) This is done by taking advantage of the known translational velocity of the truck by which the bumper position is known at all times dur- ing the event. Thus as the truck passes, the array scans the truck. The horizontal scale can then be plotted as a horizontal position axis for which the reader has been oriented relative to the truck by the small photograph that has been sized to the horizontal scale of the color chart and positioned to reflect the position on the truck to which the array is “looking” for each horizontal pixel. Consistent with Figure 45, Figure 64 shows higher levels between 1 and 5 m (3 and 16 ft) behind the front bumper and 1 m (3 ft) above ground. Figure 44 also shows sources in the vicinity of the wheel well and reflection that does not appear in Figure 45, probably due to effects of directivity at the time instant of this image. However, Figure 64 shows a distribu- tion of sound sources in the forward extremity of the truck, because this depiction of the data gives the continuous com- plete map of the received sound versus time, regardless of directivity. Note that the depiction of the type shown in Fig- ure 45 would show these other sources at another instant. When the onboard spherical source is deactivated, its con- tribution disappears from the vertical scan. This phenomenon is illustrated in Figure 65 for which the corresponding image is given in Figure 46. Figure 64 is a combination of the distribu- tions that are shown in Figures 63 and 65. Also, note that the sound source distribution for this low-speed passby is concen- trated to the 1 m (3 ft) elevation from the surface of the road.

3.6 Roadside Testing The beamforming noise mapping technique developed in this project and validated through the proof-of-concept test- ing as described in Sections 3.1 through 3.5 of this chapter was further applied to quantify noise sources for a wide range of trucks under actual road conditions. This second experi- mental task of the project was accomplished through the roadside truck noise measurements using the developed measurement system on an in-service highway. Four related issues are discussed in the following subsections: (1) micro- phone array modifications for the roadside testing, (2) mod- ifications to the test data post-processing algorithm, (3) site selection for the roadside testing, and (4) roadside measure- ment setup. The roadside test results are presented and dis- cussed in Section 3.7. 3.6.1 Microphone Array Modifications The results of the proof-of-concept testing presented previ- ously confirmed that the microphone array, data acquisition system, and beamforming software developed in the course of the study performed generally as expected and required no major adjustments. Based on the field experience obtained during the testing, certain improvements to the system were implemented prior to the roadside testing of the array to sim- plify its setup and tear down. A few minor modifications for the microphone array were necessary to speed up assembly of the array and improve handling of multiple microphone cables between the array and data acquisition system in the field. For the previous proof-of-concept testing, individual 50 ft (15 m) long microphone cables were connected directly between the microphones and data acquisition system, as could be seen in Figure 10. While functional, that setup was laborious for the large number of microphones employed. To make setup more efficient, microphones at each of the three frame sections of the array were pre-wired with a cable junction panel, as shown in Figure 66 for the lower section. When mechanical assembly of the frame sections of the array was complete in the field, three cable bundles similar to professional audio cable snakes, also seen in Figure 66, were connected between the three junction panels and the data acquisition system. For measurements, the total frame and cable connections could now be set up in approximately 30 min compared to several hours previously. The array dimensions and microphone locations remained unchanged from the design described in Section 3.3.1. Each of the 14 radial spokes had 5 microphones, and 7 additional microphones were mounted vertically in the center, such that the total of 77 microphones was used for the roadside testing. The microphone coordinates in the array are provided in Appendix A. 3.6.2 Data Post-Processing Algorithm Modifications The data analysis relies heavily on a time-gated FFT tech- nique that provides a detailed frequency–time decomposition of the truck’s sound levels and frequency content during its passby. The sound pressure spectra are measured in 1 Hz fre- quency bands evenly distributed from 0 Hz to the selected upper limit (generally taken to be 3000 Hz). Numerical effi- ciency and limitations on the processing time provide a con- straint on the number of frequencies analyzed and the number of times for which spectra are calculated. This was discussed in detail in Section 3.5.4. For the earlier proof-of-concept tests, the settings were used that provided a frequency range from 0 to 3000 Hz in 64 steps for each of 15 to 20 time segments. The actual measurement bandwidth used at that time was about 47 Hz (3000/64) at all frequencies, with occasionally broader bandwidths. Once these spectra were obtained, using a simple arithmetic conversion formula the source levels were pre- sented as equivalent A-weighted one-third octave band levels of sound pressure relative to 20 µPa at each indicated fre- quency. This was done to provide levels that could be com- pared approximately to classically processed one-third octave band levels, which were also collected during the passbys. This use of approximate one-third octave band equivalents, though convenient, was not ideal because of its fundamental inability to accurately present one-third octave band levels that contain tones, such as tire tones. Thus, a wrapper code for use in batch-processing and saving the spectral files with frequency intervals much smaller than 47 Hz was developed for the analy- sis of the roadside recordings. Using a frequency range of 200 to 2500 Hz for the desired one-third octave band spectra, and the need for at least 3 frequency samples in each band, the number of digital frequency samples within the 0 to 3000 Hz 53 Figure 66. Microphone cable junction box.

range was increased to 256. This increase brought the interval of frequency analysis down to 11.7 Hz, ensuring the existence of an adequate number of frequency samples in each one- third octave band used for array beamforming. A-weighting factors were then applied to each band sound level. To com- pensate for the increased sample time of the FFT processing, the number of ensembles for analysis was also reduced from that used earlier. The additional code development in this second phase also provided the vertical distribution of the sound received at the array, to infer the vertical distribution of apparent sources above and below the road surface. The processing that was developed consisted of applying the array’s beamformer algo- rithm to each truck passby, but with steering limited only to points directly in front of the array and steered to points along a vertical line extending both above and below the surface of the road. These vertical distributions are consistent with the two-dimensional beamformed images, as confirmed by eval- uations of distributions with the loudspeaker source. A network of codes provided a sequence of steps used in resolving the acoustic images of the passby: display the time records of the passby; truncate the record to roughly 2 s; develop the image functions at 11.7 Hz intervals; and calcu- late A-weighted one-third octave band levels of image values, reference microphone levels, and vertical distributions of source levels. 3.6.3 Test Site Selection Five candidate sites, all on Maryland highways, were consid- ered for the roadside truck noise measurements: • I-70 rest stop north of Frederick • US 15 near the Pennsylvania border, at a southbound rest stop (Mason-Dixon Discovery Center) near Emmitsburg • US 301 near the Delaware border, at a side spot just south of Wilson Street/Strawberry Lane • US 301 at MD 405 Price Station Road • I-895 Harbor Tunnel Throughway in the vicinity of Halethorpe All of the sites were visited prior to the testing, and their suitability for the measurements was evaluated. The follow- ing conditions were considered in this evaluation: • Sufficient number of random traveling trucks in the range of 15 to 20 vehicles per hour, on average, to provide ade- quate time separation and discrimination among passbys. • Sufficient space separation between the opposite directions of travel to minimize overlapping passbys and noise inter- ference from the opposite traffic lanes. • No road curvatures and no grades in the direction of travel that would cause vehicles to decelerate or accelerate. • Adequate, accessible, and safe side spot or shoulder loca- tion for placing the microphone array at a distance of 20 ft (6 m) from the edge of the nearest driving lane, as well as for parking a minivan with the data acquisition system. • The setup should be located on a flat surface at the road ele- vation, with no buildings or other reflective surfaces behind. Based upon the site review and evaluation, the site on US 301 at MD 405 Price Station Road was selected for the roadside testing as meeting all of the criteria. At this location, the opposite directions of travel on the four-lane highway split forming a wide median with trees, while the northbound lanes remain straight, generally flat, and provide a wide, paved turn- ing lane with a shoulder convenient for the measurement array installation. The road pavement in the vicinity of the site was dense-graded asphaltic concrete (DGAC). The posted speed limit at the site was 55 mph. 3.6.4 Roadside Measurement Setup The microphone array sections with pre-assembled micro- phones, the data acquisition system, and other equipment were transported in a minivan to the US 301 site selected for the test- ing on April 9, 2008. The array was assembled in the field and placed on a shoulder at a distance of 20 ft (6 m) from the edge of the nearest northbound driving lane. The data acquisition system was located in the minivan. The overall measurement setup is shown in Figure 67. Each measurement channel of the system was calibrated using a Brüel and Kjær Type 4231 acoustic calibrator. To track a vehicle and determine its speed at the test site, the same two photocells that were used for the proof-of-concept testing were again installed on tripods near the microphone array, with their signals fed into the data acquisition system. Still photographs of vehicles passing the array were taken in order to later relate the vehicle geometry to the sound distribution images. The measurement session was recorded with a video camera for later references as necessary. The results of these measurements are presented and discussed in the next section. 54 Figure 67. Roadside measurement setup.

3.7 Results of Roadside Measurements For initial check-out of the array performance and calibra- tion of the test site geometry, a short test was performed first using a stationary loudspeaker. After this initial test, the actual truck passbys were measured using the microphone array and data acquisition system. In the following subsections, roadside test results are presented and discussed in three categories: (1) calibration of the test site geometry using imaging tests performed with the loudspeaker source; (2) acoustic image and noise source distribution results of the vehicle passbys; and (3) example of truck source modeling for simulating noise propagation. 3.7.1 Calibration of the Test Site Geometry A loudspeaker source, Mackie Model SRM450, was used to calibrate the positioning alignment of the array relative to the road surface in the path of the trucks in the near lane of the highway. The source was mounted on a tripod and the assem- bly was located on the roadway with its center approximately at the expected wheel path of the closest side of the trucks at a height of 4.5 ft (1.4 m) above the road surface. This test was performed during a short break in the traffic, just prior to the commencement of truck measurements. Noise from a pink noise generator was played through the speaker at a high vol- ume for the test. Figure 68 shows acoustic images of the speaker obtained at frequencies of approximately 270, 1000, and 2000 Hz. In these plots, a small offset to the vertical coor- dinate [about 0.27 m (0.9 ft)] was provided to compensate for misalignment of the array acoustic plane with the road sur- face and road crowning. This offset is not (and obviously should not be) a function of time or of frequency. Note the presence of the ground-reflected image source with the sound level slightly less than the direct path image, especially as fre- quency increases. The reflected image of the source is below ground and is 1 to 2 dB lower in level than the direct image. This suggests a reflection coefficient of 0.8, which is compa- rable to that apparent in the proof-of-concept measurements earlier in the study. Overall, the measurement system per- formed as expected and was adequate for conducting the roadside truck measurements. As noted previously, one of the extensions made to the processing was to provide vertical source distributions. Fig- ures 69 to 71 provide vertical profiles of the sound source dis- tributions at frequencies 250, 1000, and 2000 Hz, respectively. In these figures on the left-hand side is a level distribution of the stationary source over a 1.3 s time interval, which demon- strates the (expected) invariance of the sound profile with time for each frequency. On the right-hand side is a line plot of the A-weighted sound level vertical profile at one of the time increments. Comparison of Figures 68 through 71 illus- trates the duality of the source two-dimensional imaging and the vertical distributions in localizing the noise sources. 3.7.2 Image Results of the Vehicle Passbys Of 100 truck passbys recorded in a single day of data acqui- sition, passbys for 59 heavy trucks and 4 medium trucks were analyzed and are discussed here for the definition of source lev- els. Because the objective was to interpret source level images and vertical distributions of the sources, the following require- ments were necessary for a passby to be analyzed: the vehicle in the curb lane, significant time–space separation between passby vehicles to allow for distinct one-to-one vehicle–passby identification, a photograph of the vehicle, and nearly constant vehicle speed. The remaining 37 passbys were also acoustically viable as recorded passby data, but they could not be examined in detail because at least one of the required characteristics for 55 (a) (b) (c) Figure 68. Images of the loudspeaker positioned on the roadway at the test site for frequencies of (a) 270 Hz, (b) 1000 Hz, and (c) 2000 Hz.

56 (a) (b) (a) (b) complete analysis was missing or they were tractors with atyp- ical trailers (boat or car haulers, wide-load modular houses, etc.). Acoustic source maps obtained during truck passbys were then used to provide time histories and spatial distributions of sources and source paths from the engine, exhaust, tires, and certain body components. Individual vehicle speed during passbys varied from 55 to 70 mph (88 to 112 km/h). The total sets of 59 heavy truck passbys and 4 medium truck passbys (to a much lesser extent) provide ensembles of sound levels by vehicle, one-third octave frequency band, time, and vertical position. These profiles can be used to define maxi- mum levels, mean levels, and sound exposure levels—all across the ensembles of trucks. The roadside setup geometry used for these measurements followed that used for the proof-of- concept tests, with improvements in the timing of the truck position relative to the center of the array in order to reduce positioning error. For this, two photocells were positioned at distances 5 ft (1.5 m) before and 20 ft (6 m) after the array Figure 69. Calibration source, vertical scan at 250 Hz: (a) time–level distribution; (b) line plot at 0.62 s. Figure 70. Calibration source, vertical scan at 1000 Hz: (a) time–level distribution; (b) line plot at 0.62 s.

center. The positions of the photocells and array relative to the test track were the same as illustrated in Figures 50 and 51. In the dimensions used at the highway site, the length of the truck as it intermittently cuts the photocell beam is shown by the total length of its signal. Figure 72 shows an example from roadside case 60, which is a heavy truck. The front bumper is again taken for all runs as the reference point on the truck, so that the max- imum sound occurs slightly later (typically about 0.06 s for this highway speed limit) than the bumper passage by the array center. The data processing, including the distance correction for equalizing the passby levels and the Doppler shift for fre- quency correction, is described in Sections 3.5.4.1 and 3.6 and requires these time-accurate truck coordinates. Components of the signal alignment and truncation as relevant to time- synchronizing the time–frequency spectrogram and the pho- tograph of the truck 60 are shown in the figure. The upper traces include the first photocell signal (red), the second photocell signal (green), and the signal from the reference microphone in the array (blue). The truck passing at 58 mph (93 km/h) is positioned so that its bumper is at the first photo- cell. The size of the truck photograph is determined by the speed and the known spacing between the red cones [7.3 m (24 ft)]. The spectrogram is taken approximately every 0.05 s with the 87 Hz frequency intervals. Clearly seen is the corre- spondence between instances of high sound levels at multiples of approximately 450 Hz that are due to the tires, although high background levels are also seen below 200 Hz. Figure 73 shows the distribution of the one-third octave band A-weighted spectra for all 59 heavy trucks. Each spectrum is measured with the single reference microphone stationed at the array distance [20 ft (6 m)] and at height of 1.9 m (6 ft) above the road grade at the time of the maximum overall A- weighted sound level (OASPL) for each truck. The dotted black lines indicate the statistical maximum and minimum levels, and the bold solid black line indicates the mean level across the selected truck population. Truck 60 is also shown for reference, as it is a typical truck with sound levels near the mean of the population. All of the individual trucks are not identified here, but their levels are all shown to indicate the distribution of lev- els within the population. More discussion of the distribution will be presented below. Note that a commonly occurring spec- tral feature is the pair of 500 and 1000 Hz peaks, likely due to tire noise (see also Figure 72). Figure 74 shows the spectra for the four medium trucks in the data set. Again, the prominent 500 and 1000 Hz peaks are present. These spectra are all measured with the single refer- ence microphone stationed at the array distance [20 ft (6 m)] and at height of 1.9 m (6 ft) above the road grade at the time of maximum level for each truck. Although the population is too small for generalization, the levels are slightly lower than for the heavy truck mean levels. That the sound from the typical truck is due to tire noise is emphasized in Figure 75, which shows a full set of images for the one-third octave frequency bands from 315 to 2000 Hz obtained for truck 60. Each point spread function is shown in the upper left corner with the same scale as in the main image. Though the number legends for these are unreadable, they act as tic marks that correlate with the main figure. The color bar provides a relative 10 dB scale for which the reference value is approximately the red dot level for truck 60 in Figure 73. Note that the tractor drive-axle tires and the trailer tires are both contributing, but prevail in different frequency bands. 57 (a) (b) Figure 71. Calibration source, vertical scan at 2000 Hz: (a) time–level distribution; (b) line plot at 0.62 s.

58 Figure 72. Signal alignment and acoustic spectrogram with the photograph of truck 60. Figure 73. One-third octave band sound spectra (in dBA) for 59 heavy trucks.

This phenomenon suggests that different tire patterns were used on the tractor and trailer of this truck. All nine of these images were made for the same computer settings of the array’s spatial scanning, so any biases in sound level equaliza- tion for distance correction are the same for all frequencies. Thus, the frequency-to-frequency relative comparisons in this figure are valid. The vertical distribution of sound sources at the nearest truck’s side plane can now be assessed in various ways. One way is to examine the vertical profiles of source sound levels at 6 m (20 ft) reference distance for the mean values of all heavy trucks. Figure 76 shows the source height distributions of the mean A-weighted one-third octave band levels for each band from 100 to 2500 Hz and for the mean OASPL. As dis- cussed in Section 3.6.2, these distributions are obtained by mathematically focusing the array at a sequence of vertical positions above and below the road surface directly in front of the array. The sound levels are determined at the instant of the maximum OASPL passby level at the reference micro- phone for each truck, and are evaluated by the microphone array at that instant for each truck. The mean levels for each frequency band and steering elevation over all trucks are then evaluated and plotted in the figure. It is clear from Figure 76 that the vertical distribution for the average truck sound level is dominated by sources between −1 and +1.5 m (−3 and +5 ft). This distribution holds for the frequencies above 315 Hz. Below 200 Hz, although the verti- cal beam width is larger for the array, the source distribution seems to drift upwards. Comparison with the images of Fig- ure 75 confirms that the tire sound sources are positioned just above the road surface. Figure 77 shows the measured vertical source distributions for three individual trucks. These distributions compare rea- sonably well with those obtained in the Caltrans study (30, 31). To illustrate this comparison, Figure 78 reproduces Fig- ure 8 from Donavan et al. (30). The level distributions are similar in shape, although those in Figure 78 show the maxi- mum levels at or slightly below the road surface, while in Fig- ure 77 the maximum levels occur at or slightly above the road surface. This discrepancy is caused by an uncertainty in cali- brating the position of the array’s acoustic axis relative to the ground plane. It is expected that the certainty with which the ground plane can be accurately established from either set of data is within about 0.2 m (0.6 ft). Some contributors to this uncertainty include the distance to the actual sources versus the beamforming analysis plane, the size and directivity of the calibration loudspeaker, the frequency-dependent reflection coefficient for the path of reflected sound between the sources and the array, and the precision with which the array can be aligned because the relative tilt angle of the array and the incli- nation of the road crown could not be measured. This 0.2 m uncertainty is equivalent to effective acoustic inclination error of less than 2 degrees. The resolution of this discrepancy may become the subject of a separate follow-up study. The statistical distributions of the sound levels measured across the population of heavy trucks are provided in the form of histograms in Figure 79 for the OASPL at the reference microphone and at the 0.4 m (1.3 ft) height for the one-third octave frequency bands centered at 500, 1000, and 2000 Hz. All of these levels were recorded at the time of maximum OASPL for each truck. The data for these histograms and for Figure 73 discussed previously are the same. The breadth of 59 Figure 74. One-third octave band sound spectra (in dBA) for four medium trucks.

60 Figure 75. Source image maps of truck 60 for one-third octave frequency bands from 315 through 2000 Hz.

61 Figure 76. Vertical distributions for mean one-third octave band and overall A-weighted sound levels (in dBA) for 59 heavy trucks. Figure 77. Vertical distributions for one-third octave band and overall A-weighted sound levels (in dBA) for individual heavy trucks (truck images are to the same scale): (a) truck 18, (b) truck 31, and (c) truck 54.

62 Figure 78. Vertical distributions for one-third octave band and overall A-weighted sound levels for an example truck at the Lakeview site, CA [from Donavan et al. (30, Fig. 8)]. Figure 79. Statistical distributions of the (a) overall sound levels and one-third octave band sound pressure levels (in dBA) at (b) 500 Hz, (c) 1000 Hz, and (d) 2000 Hz for 59 heavy trucks. (a) (b) (c) (d)

the sound level distribution shown in Figure 79 is identical to that shown in Figure 73 at the corresponding frequencies. The larger spread at 500 Hz is due to the variation of levels in this band apparent in both figures. Other examinations of the vertical distributions of sound sources are also possible. Figure 80 shows the vertical distri- bution of the highest sound levels among the 59 heavy truck passbys in each one-third octave frequency band and for each elevation. These distributions identify the levels for the nois- iest truck at each vertical position. The levels are determined at the instant of the maximum OASPL at the reference micro- phone for each truck and evaluated by the microphone array at that instant for each truck; then the maximum levels at each frequency and steering elevation for all trucks are eval- uated. Note that these distributions are for statistical maxima and do not represent levels from any individual truck sample. A modest increase of approximately 8 dBA can be seen in the near-road sources at 500 Hz due to tire variations. A signifi- cant increase in sound levels at elevations of 3 to 4 m (10 to 13 ft) is also noticeable. It is shown later in this section that the latter increase correlates well with the truck source images localized at the vertical exhaust position. In a similar manner, Figure 81 shows the vertical distribu- tion of the 1 s sound exposure levels (SELs). The levels are determined at the instant of the maximum OASPL at the ref- erence microphone for each truck and evaluated by the micro- phone array at that instant for each truck. Then the maximum A-weighted SELs are determined for each one-third octave frequency band and steering elevation for all of the trucks. These distributions also show the expected dominance by sound sources located near the road surface. Figures 82 through 84 show the same types of vertical dis- tributions for the four medium trucks measured during the roadside testing. Although this small sample does not form a statistically significant population, these results are included here for completeness and to indicate the similar general dis- tributions as obtained for the heavy trucks. For each figure, the sound levels are determined at the instant of the maxi- mum OASPL at the reference microphone for each medium truck and are evaluated by the microphone array at that instant for each truck; then the mean levels, the highest lev- els, and the maximum 1 s SELs, respectively, are determined for each one-third octave frequency band and steering eleva- tion for all four medium trucks. The absence of sound sources at the 2 to 4 m (6.5 to 13 ft) elevations may be due to a lack of samples in the population, but it also is likely due to the absence of vertical exhaust stacks on these trucks. As previously noted, the vertical distribution of the sound sources has its highest levels near the surface of the road, which dominate the levels for the average truck. The maximum lev- els during heavy truck passbys also appear to have significant components at elevations between 2 and 4 m (6.5 to 13 ft). Figure 85 shows the vertical distributions of OASPL for the 59 heavy trucks. In the following paragraphs, a few representa- tive truck passbys are discussed further in more detail. Truck 60 is a typical truck with the sound spectrum near the mean of 63 Figure 80. Vertical distributions of highest A-weighted sound levels in one-third octave bands and of overall sound level (in dBA) for 59 heavy trucks.

for truck 38 in Figure 73, which dominates the level maxima for the population. Truck 38 is a three-axle dump truck that produced both exhaust tones and tire tones over the fre- quency range. The bold red line in Figure 85 shows two max- ima in the vertical OASPL distribution at the elevations of about 0.4 and 3.6 m for this truck. Figure 87 shows the acoustic images of this truck in various frequency bands. As can be seen in the figure, near the road surface the tire noise 64 Figure 81. Vertical distributions of maximum 1 s sound exposure levels (in dBA) in one-third octave bands for 59 heavy trucks. Figure 82. Vertical distributions for mean one-third octave band and overall A-weighted sound levels (in dBA) for four medium trucks. the heavy truck population. Earlier in this section (see discus- sion of Figure 75), the tires on truck 60 were identified as the primary sound originators. Truck 15 is representative of the maximum sound levels, and the image in Figure 86 for this truck also clearly identifies the tires on the tandem trailers as the source originators. Trucks 38 and 50 are uniquely relevant to the low-frequency sound levels. Note the identification of the sound spectrum

appears to be prominent at frequencies below 500 Hz, and the tire noise near the road surface appears to be prominent at higher frequencies. In all these cases, the exhaust noise deter- mines the source levels at the elevation between 3.5 and 4 m above the road surface. The mean sound levels for the total heavy truck population appear to be minimally influenced by the exhaust noise sources; however, the maximum sound 65 Figure 83. Vertical distributions of highest A-weighted sound levels in one-third octave bands and of overall sound level (in dBA) for four medium trucks. Figure 84. Vertical distributions of maximum 1 s sound exposure levels (in dBA) in one-third octave bands for four medium trucks. dominated at higher frequencies (630 and 1250 Hz in this fig- ure), but at lower frequencies (250 Hz) the noise source at the vertical exhaust was predominant for truck 38. Similarly, truck 50 also produced exhaust noise at an elevation of nearly 4 m, as illustrated by the bold magenta line in Figure 85 and the acoustic images of the truck in Figure 88. Similar data analysis for trucks 13 and 57 indicated that the exhaust noise

66 Figure 85. Vertical distributions of overall A-weighted sound levels (in dBA) for heavy trucks. The legend on the left identifies the truck ID numbers. (a) (b) Figure 86. (a) Typical source image and (b) vertical profile of OASPL for truck 15.

67 Figure 87. (a, c, e) Source images and (b, d, f) vertical profiles of OASPL for truck 38 in one-third octave bands centered at 250, 630, and 1250 Hz. (a) (b) (c) (d) (e) (f)

levels of the population are significantly affected by the exhaust noise sources. In an interesting case of truck 36, with the source distribu- tion identified in Figure 85, the highest OASPL was localized below the road surface. Figure 89 shows three acoustic images of tanker truck 36 at various frequencies, which also reveal the source images located below the ground plane. The below- grade source image could be due to a combination of the direct-path tire noise and reflection of tire noise off the bot- tom of the tank. Close inspection of the photograph of this truck showed a distribution of piping and other structure below the tank between the axles, which may explain the pres- ence of the additional weak source in the image at 800 Hz. Note also an increased level of sound in the area of the exhaust stack in the 500 Hz band. A slight shift of the source image from the exhaust stack forward, which can be observed in the figure, is likely a result of a gradual deceleration recorded for this truck. 3.7.3 Example Model of Truck Sources for Simulating Noise Propagation Results of the Vehicle Passbys Figures 76, 77, and 80 through 88 show the vertical distri- butions of noise sources as determined by roadside mea- surements. The key figures are 76 and 80, which show the distributions for heavy trucks. Figure 76 shows the mean A-weighted spectral and overall levels, while Figure 80 shows the distribution of maximum levels. The data plotted in these two figures are presented in Appendix B. The levels in Figures 76 and 80 and Appendix B are not nor- malized to absolute values. They are, rather, relative weight- ings. Application of these results to the Traffic Noise Model (32) requires that the levels be normalized to a 50 ft (15 m) passby distance. The scaling process would be as follows: • At each height convert the levels in Table B-1 or B-2 to energy • Sum the energies at all heights, then normalize to 1 68 Figure 88. (a, c) Source images and (b, d) vertical profiles of OASPL for truck 50 in one-third octave bands centered at 630 and 1000 Hz. (a) (b) (c) (d)

• Multiply the normalized energies by the energy of a total passby measurement [i.e., the levels in Appendix A of the Traffic Noise Model technical manual (32)] This process can be carried out for any number of source heights, as a generalization of the upper and lower emission levels defined by Equations 7 and 8 in Appendix A of the Traf- fic Noise Model technical manual (32). The apparent vertical distribution of truck noise sources can be effectively simulated by a simple system of two uncorrelated sources, similar to that presented in the Traffic Noise Model technical manual (32). This system is illustrated in Figure 90. The ground is assumed to reflect sound from both sources with a reflection coefficient of 0.8, as suggested by the measure- ments made in this project with the calibration source. Note also that this model produced simulated source images that agree well with the measured ones, as was shown previously in Section 3.5.1 (see Figures 35 through 37). The source levels of S1 and S2 for the tires and exhaust, respectively, as well as the two heights ym1 and ym2 are frequency dependent. These quan- tities are also different for determining the mean sound lev- els and the maximum sound levels of the truck population. Table 15 gives the values of the parameters that were used to produce the vertical source distributions shown in Figures 91 and 92. The source levels in Table 15 are all relative to 20 µPa at 6 m and represent equivalent free-field source levels from the individual sources. The values shown are example values for two frequencies only. To extend these values to a broader range of frequencies, one would calculate by an inverse propagation method the levels and vertical locations of the elemental source distributions that are required to replicate the one-third octave band profiles. In general this calculation may require a contin- uum of sources. The vertical profiles of the equivalent sources obtained for Figures 91 and 92 were produced using the same codes that are described in Sections 3.2.1 and 3.6.2 and used to process the 69 Figure 89. Source image for truck 36 in one-third octave bands centered at (a) 500 Hz, (b) 800 Hz, and (c) 1000 Hz. Figure 90. Geometry of simulated truck noise sources. Simple sources S1 and S2 represent the tire and exhaust sources, respectively. Image Sources SourcesObserver Location Ground, Reflection Coefficient = 0.8. ym2 ym1 S2 S1 6 m Table 15. Model parameters for simulation of vertical truck noise source distributions. Sound Level Metric Frequency Band (Hz) Source Level S1 (dBA) Source Elevation ym1 (m) Source Level S2 (dBA) Source Elevation ym2 (m) Mean Levels 500 81 0.8 63 3.5 5.3564.0470001 Maximum Levels 500 89 0.7 81 3.5 1000 84 0.4 77 3.7

measurement data. Note that the measured source distribu- tions all match the simulated ones, although as frequency increases it is clear that additional sources would be necessary to “fill out” the profiles at intermediate elevations. These examples show, nevertheless, that a two-source model does provide a starting point that generally characterizes the appar- ent source profile quite well. To build a complete model of the truck noise sources with the requisite fidelity for a highway noise model, however, one should use the above model exam- ple as a building block. Among other requirements for the adequate precision, such a model should use ground reflection coefficients as a function of frequency, improve fidelity in the vertical distribution at higher frequencies, and account for some beamforming factors inherent in the current data. These beamforming factors have to do with non-unity correlation coefficients between microphone pairs that determine the pre- cise relationship between the steered array output and the sound levels at the reference microphone. 70 Figure 91. Vertical distributions of source levels for 500 Hz as measured at 6 m (20 ft) distance for heavy trucks: (a) mean levels, (b) maximum levels. Figure 92. Vertical distributions of source levels for 1000 Hz as measured at 6 m (20 ft) distance for heavy trucks: (a) mean levels, (b) maximum levels. (b) Maximum levels(a) Mean levels ___ο__ Measured _____ Simulated ___o__ Measured _____ Simulated (a) Mean levels ___o__ Measured _____ Simulated ___o__ Measured _____ Simulated (b) Maximum levels

Next: Chapter 4 - Conclusions and Recommendations »
Acoustic Beamforming: Mapping Sources of Truck Noise Get This Book
×
 Acoustic Beamforming: Mapping Sources of Truck Noise
Buy Paperback | $49.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) Report 635: Acoustic Beamforming: Mapping Sources of Truck Noise explores the acoustic beamforming technique in an attempt to pinpoint and measure noise levels from heavy truck traffic. The beamforming technique uses an elliptical array of more than 70 microphones and data acquisition software to measure noise levels from a variety of noise sources on large trucks—including the engine, tires, mufflers, and exhaust pipes.

READ FREE ONLINE

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!