Sie sind auf Seite 1von 3

Analyzing sound levels of waves

Arvid Lyngstad
September 23, 2019

1 Summary
2 Waves were recorded at the pier in Trondheim, and
3 this text analyzes the sound pressure of the recording
4 using common sound level descriptors.

5 1 Introduction
6 The goal of this text is to analyze and discuss a record-
7 ing of waves. To restrict the data set, only 30 min-
8 utes of the recording is presented in the results, using
9 the descriptors Leq , short Leq , Lτ both fast and slow,
10 Lmax along with an analysis of fractile sound levels. Figure 1: The distance from the source.
11 The project succeeded in some of the descriptors, but
12 not all were implemented. Before and after the main recording, two short 42

recordings were made with a 94 dB calibrator. This 43

was used to find the pressure levels, based on a note 44

13 2 Materials and methods on calibration [4]. Using the formula 45

14 The recording took place at the pier in Trondheim on !


|p|
15 the 7th of September, between 8:30 pm and 9:30 pm dBSP L = 20log10 √
16 for a total of one hour of material. The wind was, 2pref
17 according to a weather forecast website [1], 2 m s from where |p| is the amplitude of a harmonic sound wave 46
18 west north-west. The microphone had an orientation and pref = 2 · 10−5 is the reference pressure, the value 47
19 towards north-west pointing at an angle downwards. for the pressure associated with 94 dB can be found 48
20 The same website also gave the temperature to be 10 to be 1.0024P a. This value will be used together with 49

21 C at the beginning of the recording, and 8 ◦ C at the the results to find the correlation between the values 50
22 end. The microphone was set up in a staircase leading stored in the recorded material and the real pressure 51
23 into the water, with rocks or concrete walls in every p(t) of the main recording. It is shown in [4] that the 52
24 direction except for up and out towards the water. correlation is 53
25 The weather was clear, with light clouds occasionally.
26 Fig. 1 shows approximate distance to the immediate pcal
27 surroundings and sources. p(t) = x(t)
xcal
28 The set-up, as seen in figure 2, was simple. A wind-
29 screen was put on a condenser microphone, to shield for a sample at time t. 54

30 from the direct wind blowing into the microphone. In this analysis, the trapezoidal rule will be used to 55

31 The signal was then transmitted through an XLR- evaluate integrals, using the sample values as the dis- 56

32 cable into a Zoom H5 PCM-recorder, and stored there crete points. 57

33 with a sample rate of 48 kHz. The recording was


34 then transferred to a computer and analyzed using 3 Results 58
35 the Wavio module [2] with Python 3.6.3 with Ana-
36 conda. In figure 3, the maximum values for xcal was xcal,1 = 59

37 The analysis was carried through with basis in sec- 2342708 for the first calibration and xcal,2 = 2412240 60

38 tion 4 of a lecture paper on environmental acoustics for the second calibration. This gives the result that 61

39 [3], with formulas 1, 2 and 3 being the crucial ones for


40 this analysis, in addition to the description of fractile
41 sound levels. p(t) ≈ 4.2162 · 10−7 x(t)
2
The Leq -value for each of the three recordings were 62

computed to be 90.45 dB re p0 for the first calibration, 63

and 91.09 dB re p0 for the second calibration. The 64

main recording was chosen to be between the 27th and 65

57th minute of the whole recording, or the samples in 66

the range of 77760000 to 164160000. The value of Leq 67

of the main recording was 62.95 dB re p0 . 68

4 Discussion 69

The difference in calibration from the first to the last 70

is small, but present. This difference could be ex- 71

plained by the microphone having a slightly different 72

position relative to the calibrator, enough to see a dif- 73

ference in sound level. The soundscape was very af- 74

fected by the waves crashing into the walls and stair- 75

Figure 2: The recorder connected to the microphone case, and to the human ear, these sounds are very 76

on a stand. prominent. In the background,during the recording, 77

there are cars driving by on a road 15 − 20 meters be- 78

hind the microphone, at some points people talking 79

loud when passing by was unintentionally recorded. 80

At one time during the chosen 30 minutes a train horn 81

can be heard, and towards the end a small bird flew 82

by, chirping. But the only sounds that influences the 83

recording in a big way are the waves, the way they 84

crash into the staircase and cars in the distance. The 85

windscreen shielded most of the wind, but it can still 86

be heard as noise. When hearing the recording, it 87

sounds like many of the lower frequencies of the site 88

was lost, most likely due to microphones often having 89

a shaped frequency response, with lower frequencies 90

being subdued. The wind is also less prominent in the 91

recording than that of the site. 92

From the computed equivalent sound levels, it is 93

clear that the calibration device was louder than the 94

waves. Figure 4 seems agrees with the value of 62.95 95


Figure 3: Samples 250000 to 250100 from both cali- dB re p0 for the main recording, as the values for 96
brations. short Leq zig-zags around it. 97

The difficulties during recording was to find a place 98

with as little people as possible, and urging the ones 99

who got too near not to stay for long without being 100

too loud. In the final recording, this can still be heard. 101

The post-processing ended up proving more tricky 102

than the recording. When choosing what software 103

to work with, a more familiar environment of Python 104

was chosen instead of Scilab, which meant that ev- 105

erything had to be programmed from scratch. Errors 106

relating to numerical precision when dealing with very 107

small or large numbers showed up due to errors in the 108

programming. The results presented had no problems 109

with this, as it was eventually corrected, but much of 110

the time spent on this project went into correcting 111

these errors, and less on implementing new descrip- 112

Figure 4: The short Leq s with an interval of 1s. tors. Because of this, not all descriptors planned to 113

implement were implemented, which means that in 114

the future there is potential for improvement in allo- 115

cating time for the project. 116


3
117 This project has some sources of uncertainty, the
118 most obvious being the numerical integration. Be-
119 cause the samples are distinct, no data is known be-
120 tween the samples. The uncertainty arising from this
121 is suppressed due to the modern standard of having
122 a high sample rate of 48 kHz, so any missed sounds
123 will not be audible to the human ear. Another un-
124 certainty is the measurement of the distance to the
125 source, but this value is mostly to get a qualitative
126 idea of it, and doesn’t affect the results. If the goal
127 was to find the actual sound pressure at the source,
128 the distance would be crucial to know with a high
129 degree of certainty.

130 5 Conclusion
131 The goal of recording waves was achieved, and so was
132 using some descriptors to analyze the sound levels in
133 the recorded material. The processing offered chal-
134 lenges with programming, but in the end some results
135 were available. The descriptors for fast and slow time-
136 weighted sound level wasn’t successfully implemented,
137 nor were the fractile sound levels. In the future it will
138 be wise to spend more time on a project such as this.

139 References
140 [1] [Internet]
141 https://www.yr.no/sted/Norge
142 /Trøndelag/Trondheim/Ila/
143 Accessed 7th of September 2019
144 [2] Warren Weckesser
145 Wavio module for Python [Internet]
146 available at:
147 Wavio on pypi.org
148 Accessed 20th of September 2019
149 [3] Guillaume Dutilleux
150 Environmental Acoustics [Internet]
151 available at (with restricted access):
152 Environmental acoustics on
153 ntnu.blackboard.com
154 Accessed 23rd of September 2019
155 [4] Guillaume Dutilleux
156 TTT4180 - A note on acoustic calibration
157 [Internet]
158 available at (with restricted access):
159 TTT4180 - A note on acoustic calibration
160 on ntnu.blackboard.com
161 Accessed 23rd of September 2019

Das könnte Ihnen auch gefallen