Вы находитесь на странице: 1из 5

UNTREF, Sound Engineering, Acoustics Instruments & Measurements

March 2013, Argentina



Univesidad Nacional de Tres de Febrero, Sound Engineering, Caseros, Argentina. facundo.ramon@gmail.com

1. INTRODUCTION The relative directivity of a sound source describes how the source distributes the sound energy through the space. Directivity ! is defined as the logarithmic difference between the effective pressure !! obtained in a determinated angle of incidence (generally the main acoustic radiation axis) and the effective pressure !(! ) obtained in an angular position ! [1], ! ! = 20 !"#
! (! ) !!


Computer MacBook Pro; Software Audacity with Aurora Plug-ins; Apache OpenOffice; External sound card Focusrite Saffire Pro 14; Two Microphones Earthworks M50; Sound level meter type 1 Svantek 959; Turntable Outline; Powered loudspeaker KRK Rokit 8; Violoncello; Absorbent foam; Tape measure; Large room; 2.2 Step by step 2.2.1 Loudspeaker In order to get the maximum distance to the walls, the sound source was located at the geometric center of the room over the turntable (Fig. 1). This left a distance of 3.5 m from the loudspeaker to the lateral walls.

It depends exclusively on the constructive characteristics of the source. And it is a relative measurement; every SPL (sound pressure level) is obtained under the same conditions and with the same equipment as the previous one. Therefore there is no need to use a standardized measurement system, the only condition is to maintain the same configuration during the entire procedure. Nevertheless, it is necessary to separate the sound source from the sound field and to measure under acoustical free field condition in order to avoid distortions caused by reflections. The objective of this paper is to determinate the directivity of an electro-acoustic source (loudspeaker) and an instrument (violoncello) without using an anechoic chamber and under no laboratory conditions. It is intended to wean the sound source from the reflections of the environment by measuring in a large room and analyzing only the direct sound. 2. PROCEDURE There are no established standards for directivity measurements. The following procedure is considered the optimum way to use a specific list of limited resources in order to obtain the most reliable results. 2.1 Instrumentation The measurement was made with the following instrumentation:

Figure 1: Setup for loudspeaker.

To avoid the generation of comb filter, due to the interference between floor reflections and the direct sound, the microphone and the loudspeaker were placed at floor level [2]. Absorbent foam was used to avoid diffraction on the shape of the turntable.

The ceiling has a double-layered structure made of floating panels of micro perforated material, it was at 2.3 m from the loudspeakers acoustic center, and the concrete ceil was at 2.8 m from the same point. Despite the absorbing surface the ceiling panel represented, the distance to the concrete ceil was considered the shortest one and the time lapse between the direct sound and the first reflection was calculated with it. The microphone was at 1.5 m from the source, this is enough distance to avoid the near field [3] and to obtain a considerable time lapse between direct sound and first reflection. Its angular position was 0 degrees in reference to the acoustic horizontal axis of the loudspeaker. Also, the microphone was covered with absorbent foam to minimize the effect of any undesired reflection or pollutant sound. The sound level meter was placed near the microphone during the noise floor level and signal measurements and then removed. It was set with slow time averaging and no weighing curve. The noise level shown in table 1 is the equivalent continuous level of 1 minute, or Leq1min. The loudspeakers amplifier and digitals amplifiers stages were leveled to get a signal to noise ratio of 18 dB at the microphone position and with the source at 0 degrees of incidence angle. The signal level shown in table 1 is the Leq10sec. After the initial setting none of the amplifying levels were modified during the entire procedure. It was decided to use a pink noise signal, digitally generated with the software Audacity, to excite the source. Figure 2 shows its spectrum analysis, two points are labeled to show the decay of -3 dB per octave. It was obtained with Audacity Spectrum Analyzer.
(,% (#% &,% &#% !"#$%&'% ()*$%)+%

The measurement was made twice. Firs with the loudspeaker in vertical position and then with the loudspeaker flipped sideway (fig. 3). Thereby the horizontal and vertical plane directivity was obtained. Because of the symmetry the loudspeaker has on its vertical plane, the first measurement was made from 0 degrees to 180 degrees. The second one was from 0 to 350 degrees.

Figure 3: Loudspeaker arrangement for the vertical plane directivity.



The main difference with the previous measurement is the inability to reproduce with the source the same signal twice. Even with a professional performer it is not possible to generate exactly the same sound several times with a musical instrument. Therefore, it becomes necessary to have a constant reference microphone near the source and always at the same incident angle for relativizing the results. Two microphones were used; the first one was at a constant position relative to the instrument. The second one was at the same position as in the previous measurement (fig 4).


),% )#% ,,% ,#% !,% !#% (#% (##% !&'$% (###%

Figure 2: Pink Noises spectrum.

Once everything was set, the polar pattern measurements started. Each measurement was made every 10 degrees. The pink noise signal was reproduced with the loudspeaker at a constant level during 1 second (no more time was needed because only the first milliseconds were going to be analyzed), then, with the microphone and the computer, the response was captured, recorded and labeled for later analysis.

Figure 4: Setup for violoncello.

The SPL generated by the instrument was measured at the microphone position with the sound level meter.

The performer was asked to play each string once in every position and each measurement was made every 15 degrees. The reference level was obtained with the near microphone and the far one obtained the level affected by the angle of incidence. The difference between the two signals is the information needed to calculate the polar pattern, the data processing is explained in the following section. The measurement was made from 0 to 345 degrees on the horizontal plane directivity. It was not practicable, because of matters of time and space, to measure the vertical plane directivity. 2.3 Data Processing All the measurements were recorded digitally and stored in the computer. It was decided to use sample rate of 44.1 kHz and 16 bit of resolution for the storing. The quantization noise obtained with 16 bit of resolution is low with respect to the noise floor level measured. No more resolution was needed because the signal to noise ratio was determined by the noise floor [4]. The audio files were edited using Audacity software and processed with Aurora Acoustical Parameters Plug-in. 2.3.1 Loudspeaker



The signal duration used for the analysis was 15 milliseconds, during that time lapse it is supposed that no reflections have arrived to the microphone yet and signal corresponds to the direct sound (see table 1). The audio file was analyzed with Aurora Acoustical Parameters Plug-in. The analysis consisted of filtering the signal per octave band and obtaining the true RMS (Root mean square) level of each band. The software references the maximum of all the signals analyzed to 120 dB and use that maximum for all the signals, in other words, there is an amplitude relation between all the signals analyzed, but because there is no need of calibration, this value its just referential. The results were stored in a .csv file, then, with Apache OpenOffice Calc the following calculation was made. The values shown are the difference obtained with the following subtraction, which responds to equation (1). ! (! ) = 20 log !! 20 log !! (2)

The sound generated by the violoncello is tonal, so it becomes necessary to study its spectrum to understand the fluctuations in RMS level per octave band. Every string was analyzed with Audacity Spectrum Analyzer and the results are shown from figures 7 to 10. The audio files were grouped by string played and angle of rotation. Due to the characteristics of the instrument, it is not possible to generate a sound with strong attack. In consequence, it is useful to study the first 15 milliseconds of audio signal because the signal to noise ratio obtained is not enough. Therefore, the time window for the analysis was taken 15 milliseconds before the instant of maximum pressure in the reference microphone. At this point both of the microphones are affected by the reflections of the near surfaces, however, the source is reaching its maximum energy emission value so it can be assumed that the source energy exceeds the energy of the reflections, then, the difference between the reference microphone and the fixed one is due to the directivity mostly. This assumption can be made because the rooms dimensions are large enough to avoid the generation of stationary wave at the frequencies of analysis (from the octave band centered in 125 Hz onwards). In this case, there were eight audio files per angle rotated: one audio file for the reference microphone and another for the fixed one, per each one of the four strings. The audio file was analyzed with Aurora Acoustical Parameters and stored in a .csv file. Each string was treated separately and the directivity pattern was obtained as follow: The fixed microphone octave band level was differenced with the reference microphone octave band level () = !"#$% ! !"#"$"%&"() (3)

where !"#$% (! ) is the fixed microphones RMS value at a specific octave band and !"#"$"%&"() is the RMS level of the reference value (always at the same position refered to the instrument) at the same octave band. This was repeated for each octave band at each string and at each angular position. The values of directivity where obtained with the following equation, which also responds to equation 1. ! ! = ! (0) (4)

This way, the level obtained with the first measurement of each octave band, corresponding to the acoustic main axis, were referenced to 0 dB in the polar pattern graphic. The bands analyzed were from 250 Hz to 4 kHz.

This way, the directivity at 0 degrees is referenced to 0 dB and the results are independents from the variations the interpreter makes on the signal. Because of the variations on the spectrum of each string, it was decided to average the directivity of the four strings. 3. RESULTS The next table shows the results of noise floor level, signal level, distance to the first reflective surface and time before the first reflection.
Table 1: Initial conditions

&&%$ &"%$ &,%$ &+%$ &#%$ &*%$ &)%$ &(%$ &'%$ "%%$ "&%$ ""%$ ",%$ "+%$ "#%$


'%$ #$ %$ !#$ !&%$ !&#$ !"%$ !"#$


)%$ *%$ #%$ +%$ ,%$ "%$ &%$ %$ ,#%$ ,+%$ ,,%$ ,"%$ ,&%$ ,%%$ "'%$ "#%$ #%%$ &-$ "-$ +-$

Noise Floor (Leq 1 min) Signal (Leq 10 sec) (Pink Noise) Signal (Leq 5 sec) (Violoncello) Distance to 1st reflective surface Time before 1st reflection

66 [dB Z] SPL 84 [dB Z] SPL 79 [dB Z] SPL 2.8 [m] 16 [ms]





Figure 6: Vertical Polar Pattern from the loudspeaker.

&+# )+# *+# !+# (+# "+# '+# %+# !"# $&'# '%%# $%&# %()# ()$# *$+#

The next polar graphics shows the directivity of the sources.

'(%$ ''%$ '"%$ '&%$ '%%$ ",%$ "+%$ "*%$ ")%$ "#%$ "(%$ "'%$ ""%$ "&%$ "%%$ &,%$ &+%$ &*%$ &)%$ &#%$ &(%$ !&#$ !"%$ !"#$ !#$ !&%$ '#%$ %$ %$ &%$ "%$ '%$ (%$ #%$ )%$ *%$ +%$ ,%$ &%%$ &&%$ &"%$ &'%$ "#%$ #%%$ &-$ "-$ (-$

$+# +# %+# %++# %+++#

Figure 7: 4th string frequency spectrum.

('# )'# "'# +'# !"# $!%# &!'#

%()# *(*#

((&# )(+#

!"#$% %

*'# %'# &'# $'# '# &'# &''# &'''#

Figure 5: Horizontal Polar Pattern from the loudspeaker.

&#$ +#$ (#$ *#$ !"#$% "#$ '#$ )#$ %#$ !#$ #$ %#$

Figure 8: 3rd string frequency spectrum.

%&#$ ''!$ !"#$ ()%$ !#%%$

%##$ !&'$%


Figure 9: 2nd string frequency spectrum.

+'$ )'$ *'$ ('$ !"#$

%&'$ (#($ )*!$ "')*$ "+#+$


#'$ %'$ &'$ !'$ "'$ '$ !'$ !''$ !&'$% !'''$

Figure 10: 1st string frequency spectrum.

#$ "(&$ ""#$ "'&$ '#$ &$ #$ !&$ "##$ !'#$ !'&$ %,&$ !%#$ !%&$ %*#$ !"#$ +#$ *&$ '%&$ %&#$ &##$ %&&$ '#&$ )#$ '&$ "#$ (&$



%%&$ %'#$ '+&$ ',#$ ')&$ '&#$


Figure 11: Horizontal Polar Pattern from the Violoncello at 125[Hz], 250[Hz] and 500[Hz].
#$ "(&$ ""#$ "'&$ '#$ &$ #$ !&$ "##$ !'#$ !'&$ %,&$ !%#$ !%&$ %*#$ !"#$ +#$ *&$ '-$ %-$ (-$ %&&$ '#&$ )#$ '&$ "#$ (&$

(250 Hz), and the vertical plane (fig. 6) shows the areas where there is a destructive interference between the two speakers (2 kHz) and it also shows more backward radiation at low frequencies than a high frequencies. But this method is imitated at low frequencies, the lectures at the 63 Hz or 125 Hz centered octave band were no possible because the short integration time window. The polar pattern of the instrument is not as clear as that of the loudspeaker. However it shows the expected results. A characteristic of the violoncello is the back plate radiation at 300 Hz caused by the vibration of its rear [5], which is shown in the figure 11 for low frequencies at 250 Hz. It also shows the directivity at high frequency (fig. 12). But there is no correlation between some of the points measured in one position and the next one (fig. 12, the 1 kHz octave band between 105 and 135 degrees), and this is associated with the low signal to noise ratio achieved mainly at high frequencies. The noise floor is an important fact to consider, the room where the measurements were made is located near a train station and in a noisy district. That was not a major problem for the electroacoustic source because it can be amplified and get over the noise, but it was for the natural source. Notwithstanding, the use of one microphone as a reference and another one on a fixed position for measuring a natural source did not represent any inconvenient and eliminated the problem of no having enough microphones to cover the entire circumference. Accordingly, despite there is no standard guideline for polar pattern measurements, it can be concluded that is important being isolated as much as possible from the external noise and having, as close as possible, free-field condition. 5. REFERENCES [1] Pueo B. O., Rom M. R. Electroacstica Altavoces y Micrfonos. First edition, pp. 51-73. Pearson. Madrid. Espaa. 2003. [2] Serway R. A., Jewett J. W., Fsica para ciencias e ingeniera. Volume I, pp 500-520. Cengage Learning. Mexico D. F. Mexico. 2008. [3] Beranek L. L. Acstica. Spanish version, pp. 104-120. H.A.S.A. Buenos Aires. Argentina. 1961. [4] Oppenheim A. V., Willsky A. S., Hamid Nawab S. Seales y Sistemas Segunda edicin. Second edition. Pearson. Edo. de Mxico. Mexico. 1998. [5] Sound On Sound, Online magazine, Recording Strings. http://www.soundonsound.com/sos/apr99/articles/rec strings.htm, April, 1999.



%%&$ %'#$ '+&$ ',#$ ')&$ '&#$


Figure 12: Horizontal Polar Pattern from the Violoncello at 1[kHz], 2[kHz] and 4[kHz].

4. DISCUSSION The results obtained for the electro-acoustic source are the expected ones, they show the regular behavior of 2-way loudspeaker described by literature [1][3]. The horizontal plane (fig. 5) shows it is cardioid at high frequencies (1 kHz, 2 kHz and 4 kHz) and it turns omnidirectional at low frequencies