Pop Sound Protection and Reduction for Small Microphone Enclosures. Bachelor Thesis

Transcription

1 Pop Sound Protection and Reduction for Small Microphone Enclosures Bachelor Thesis Martin BLASS Michael PIROLT Graz, April 5, 2014 Signal Processing and Speech Communication Laboratory Graz University of Technology Head of the institute: Univ.-Prof. Dipl.-Ing. Dr.techn. Gernot Kubin Supervisor: Dipl.-Ing. Dr.techn. Martin Hagmüller

2 STATUTORY DECLARATION I declare that I have authored this thesis independently, that I have not used other than the declared sources/resources and that I have explicitly marked all material which has been quoted either literally or by content from the used sources Date Name Date Name

3 Contents 1. Speech acquisition and pop sound reduction 5 2. Pop sounds Definition and generation of plosives Frequency properties Impact on microphone diaphragm Acoustical and mechanical pop protection Pop filters and wind shields Baffle systems Hoop systems Foam systems Microphone construction Transducer principle Design of the cage, capsule and diaphragm Pop sensitivity measurement Measurement according to EN Reasons for simplified measurement Equivalent pop level EPL Measurement equipment Pop generator Isolation box Microphones Acoustic filters Measuring process Microphone sensitivity Analog microphones Digital microphones Electret condenser microphones Measured and calculated EPL Verification of sufficiently blunt edges Results Equivalent Pop Levels Pop attenuation

4 Signal analysis of measured pop signals Time domain Frequency domain Discussion EPL and pop attenuation Signal properties Conclusion 58 Appendix A. MATLAB source code 60 A.1. MATLAB code for FFT analysis of pop signals A.2. MATLAB code for evaluating the frequency response of the acoustic filters used Appendix B. Datasheets and manuals 64 4

5 1. Speech acquisition and pop sound reduction Due to the development in memory size and processing speed of personal computers, voice recognition software has become a fundamental and expandable field of research. Spoken words are transferred into analog signals with the help of microphones. The computer converts these signals into the digital domain. Digital signals can now be used to execute different commands, for example opening a file, based on the recognition of the words spoken. It has been shown that the reaction of the microphone towards pop sounds is critically important for the act of recognition. If the analog signal is distorted by plosive sounds or wind noise, the speech recognition software is unable to identify the desired command [1]. Especially for real-time applications, such as programs that convert recorded speech into a transcript, it is necessary to guarantee accuracy in speech recognition. Therefore the system must be equipped with some kind of pop protection, which is commonly performed by employing a windscreen around the microphone. For many years traditional pop screens made out of foam or a stretched fabric have been in use and are sufficient for most applications. However, under certain circumstances a wind screen is not practical or adequate, for example when using a lapel or a visually discreet microphone. Here it is desired to find nano-materials that can be used as a miniature pop shield or furthermore replace the mechanical screen by an electronic pop suppression circuit [1]. Another scenario for the use of pop sound suppression is the close talking situation. The main problem is caused by the speech output of the user itself due to distance of the microphone. When the distance is too small, an undesired sound, known as puff noise or pop sound is generated by the high air velocities of certain speech sound, such as plosives. When the airstream that is caused by sounds such as /p/, /t/ and /k/ hits the protective grid, which is commonly provided to cover the microphone capsule, turbulence is created, leading to distortion. In order to reduce distortion, one could either enlarge the distance between microphone and speaker, or use some kind of pop protection filter. If we now take a look at hand-held devices or boom microphones, pop sound reduction through increasing distance is often not an option. For example boom microphones, which are used in noisy places such as cockpits, are often arranged as differential microphones. They are intended to be used very closely to the speaker s lips in order to maximize the noise cancellation effect of the enclosed capsule. Even for microphones in 5

6 ordinary mobile phones it is desired to deliver speech signals without any distortion due to pop sounds. As we see, there is a vast amount of small devices which are designed to record or perceive human speech and as consequence to the progress in circuit design, those devices are rapidly getting smaller in the near future [2]. Besides conventional pop filtering designs, which are commonly bulky structures used in studio recording sessions or live broadcasting, very little research has been done to reduce the amount of pop noise with small sized filters. Therefore the purpose of this bachelor thesis is to find and compare different materials, which can be used as pop filters in small devices. The following chapters cover a general investigation of pop sounds, different methods of pop filtering and a pop sensitivity measurement, which reveals advantages and disadvantages of various acoustic filters and describe their impact on the microphone. 6

7 2. Pop sounds 2.1. Definition and generation of plosives In order to understand the need of suppressing and reducing pop sounds it is necessary to define its origins. The phonetic term for pop sound is plosive, whereas pop is associated with the sound of the release burst and aspiration of stop consonants /p h /, /t h /, and /k h / [3]. In general the glottis is understood to be the sound source for speech, while the vocal tract acts like a filter modifying the produced sound. Therefore, the process of speech production can be described with the help of a source-filter model. The air flow that is needed for the glottis to act as a sound source is originated from the lungs. The diaphragm pushes the ribcage in order to pump the air through the trachea straight to the larynx, where the glottis is situated. This leads to vibrations of the vocal cords at a frequency of Hz and further to a glottis output of periodic pulses. The speech signal is dynamically shaped and controlled by articulators along the path, such as velum, tongue, teeth and lips [4]. Figure 2.1.: Human articulatory system [4] 7

8 A plosive is an oral occlusive that is produced by a build-up of pressure in the oral cavity. The occlusion itself is performed by one single articulator, which is either the lips, the tongue blade or the tongue body (fig. 2.2). While air pressure is preserved and increased at the closure, silence is produced until the constriction of the vocal tract is suddenly released, causing the air to rush through. Simultaneously noise is generated by the fast moving jet of air which leads to turbulence at the constriction [4]. This turbulence noise is an aero-acoustic phenomenon that is generated by the fluctuating pressures in turbulent flow conditions [5] and is also called frication. Figure 2.2.: Production of stop consonants, from left to right: /p/, /t/, /k/ [6] The corresponding phonetic terms of plosives are bilabial /p/, alveolar /t/ and velar /k/. Furthermore one can distinguish between voiced and unvoiced plosives. Voiced plosives, e.g. /b, d, g/ require an increase in sub-glottal pressure in order to meet conditions for voicing. These type of stop consonants are not relevant in this thesis because they do not apply to problem of pop noise. For unvoiced plosives, e.g. /p, t, k/ the vocal cords allow more air to flow through the vocal tract, causing a whole sequence of sound phenomena [5]. An aspirated, unvoiced plosive, such as /p h / in /pa/, follows the sequence: silence, while pressure builds up behind the point of closure release, whose burst induces a transient response frication, as there is a rapid flow through a small opening near the point of closure aspiration, while there is considerable air-flow but no significant constriction in the vocal tract, as the vocal folds are being adducted voicing, which begins once the vocal folds have been sufficiently adducted [5] 8

9 This sequence is performed in about 50 ms and is shown in fig. 2.3: Figure 2.3.: Chronological sequence of an unvoiced plosive [6] Figure 2.4 illustrates the generation of the turbulent air jet caused by the utterance of a bilabial plosive. A high-speed video camera was used to record smoke particles (left), which were compared to a simulated particle front (right) [3]. It is obvious, that the velocity of the air stream reaches its maximum right after the release and is then decreasing with time. This velocity maximum of about 40 m/s always occurs within 5 cm distance from the orifice. 9

10 Figure 2.4.: High speed video (left) and simulated velocity field (right) of the plosive /pa/. From top to bottom the time (in ms) of the image is: 0, 5, 11, 21, 35, 51, 75 and 121 [3] 10

11 2.2. Frequency properties Although the nature of plosives has been investigated over many decades, a plosive is not as predictable as other types of speech sounds. Plosives tend to be non-deterministic, primarily noisy, time variant and highly transient. For the purpose of reducing the energy of a pop sound before it hits the microphone s diaphragm it is useful to observe the signal in the frequency domain. In the following averaged short-term power spectra are used to observe plosives in the frequency domain. The troughs in the burst spectrum of the bilabial plosive /p/ are located at 1.1 khz, 2.2 khz, 3.3 khz and 4.4 khz. This can be explained by approximating the vocal tract to a tube with half-wave resonances (distance lips - glottis 16 cm) [5]. Figure 2.5 shows the power spectra of plosive bursts. Figure 2.5.: Ensemble-averaged spectra of plosive bursts, 512-point Hanning window [5] In general a bilabial burst shows low-pass characteristics, whereas alveolar /t/ and velar /k/ are likely to have center frequencies depending on the vowel that is following. Especially the latter type of plosive indicates a strong dependency: For the syllables /ki/ or /ke/ the center frequencies are near 3 khz, but for /ka/, /ko/ and /ku/ they are situated at 1.5 khz and below. Alveolar bursts however tend to have resonances in higher frequencies, mostly around 3.5 to 4 khz (see fig. 2.6). This property can be exploited for e.g. speech recognition algorithms. Studies that focus on the human perception of vowel-plosive relation have shown that the information for the distinction between plosives is based on their center frequencies. In a presentation 11

12 Figure 2.6.: Center frequencies of /t/ and /k/ depending on following vowel [4] of a particular study, listeners are asked to recognize unvoiced bursts followed by different vowels while the center frequency of the burst is changed in ascending steps of 360 Hz. The result shows that high frequency bursts are recognized as /t/, whereas low frequency bursts as /p/. Bursts are likely to be associated with /k/ when their center frequency is similar to the second resonance frequency of the vowel. This is depicted in figure 2.7. The x-axis shows the vowels, the y-axis the center frequencies [4]: Figure 2.7.: Recognition of synthetic plosive to vowel transitions [4] 12

13 2.3. Impact on microphone diaphragm It is known that plosives contain a turbulent fluid flow that is propagating through air at a much lower speed than the speed of sound and hence the actual signal. Turbulence that is high in pressure has the potential to exceed the dynamic range of a microphone amplifier input or even produce a displacement of the microphone diaphragm. The consequence is a non-linear distortion that is deteriorating the signal. Especially miniature microphones, such as lapel and MEMS microphones, tend to have very high pop sensitivity. There is basically no further protection than a thin metal gauze, which is why the diaphragm is freely accessible for the pop stimulus [7]. 13

14 3. Acoustical and mechanical pop protection In order to mitigate unwanted distortion artifacts, especially pop sounds caused by a singer, speaker or other turbulent air streams, different prior art solutions can be found Pop filters and wind shields There are several, often competing approaches in the design and construction of pop filters. The first important requirement is that a filter device should be as effective as possible in diminishing pop artifacts of plosive consonants. Secondly, such an apparatus should cause the fewest anomalies possible in the fidelity of the recording. Especially the latter point demands a careful selection and application of the used materials. Another request for a good pop filtering system would include the persistence of the material and the cost of production. Conventional devices incorporate several methods for shielding a microphone diaphragm from plosive sound distortion or wind noise. Some of them are listed below [8] Baffle systems One conventional design method incorporates a baffle system that is an integral part of the microphone capsule assembly. In those constructions, a series of one or more physical baffles between the receiving end of the capsule and the diaphragm create a twisting path that acts as a series of barriers. The sound waves have to travel around the barriers to reach the microphone diaphragm. In theory, the excess displacement of air resulting in the unwanted distortion plosive is dissipated by the series of baffles. There are many different examples for the use of baffles in microphones. One is as mentioned above when used between the receiving end of the capsule and the diaphragm. Here a stiff and perforated material is being used. This sponge-like structure, which can be made of light material such as plastic, contains multiple cavities, which create a twisting path. In general baffle systems are acting similar to foam designs, which will be explained in detail in another section. [10] However, baffles can also be carried out by placing a rigid perforated structure for enclosing the whole microphone. This version normally uses macro-pore light metal material such as aluminum. On the one hand such structures act as a grid, protecting 14

15 Figure 3.1.: casual microphone baffle acting as a protective grid [9] the microphone itself, but on the other hand the amount of plosive artifact reduction is minimal, which makes baffles a hardly used solution for the reduction of pop noise Hoop systems This design method consists of one or more layers of a permeable fabric that is stretched over a hoop-shaped frame. The most commonly used materials are Lycra or spandex, which both are very flexible synthetic fibers. The fabric layer is held in place by a system of tightly-fitting concentric hoops, with the fabric edges secured by the pressure between the two hoops, which can be comprised by various materials such as plastic or light metal. This hoop is mounted to the microphone stand or the microphone itself by a piece coiled metal, the shape retention of which allows placing the hoop-type structure between the mouth of the vocalist or announcer and the capsule and diaphragm of the microphone. The hoop is fixed to one end of the length of coiled metal and the other end incorporates a clip, which is affixed to the microphone stand or the microphone body. Such coiled metal pieces are also called goose-necks. The fabric covered hoop is positioned with the flat face of the hoop assembly facing the vocalist s or speaker s mouth, so that the sound waves resulting in air pressure changes hit the flat surface of the stretched fabric at a 90 angle. As mentioned in the chapter before, plosive sounds create an unwanted excess air movement. In theory this excess air movement is dissipated and reflected by the fabric. Of course the fabric must be designed in a way that it is permeable to the point, that desired sound can pass through to the capsule and diaphragm without being unevenly attenuated over frequency. In order to guarantee high fidelity the hoop must be positioned at least 5 cm spaced apart from the device. Although the hoop type of pop filter or windscreen is commonly considered to provide good artifact reduction while affording a high fidelity of the sound that does reach the 15

16 Figure 3.2.: hoop filter mounted on a gooseneck [11] diaphragm, it has been found out that conventional designs are moderately efficient, at best. To increase their impact, two of these hoop screens can be used in succession, with a small airspace separating them. Another approach is also changing the flat surface structure or the filter. It is proposed, that a fundamental physical flaw exists in conventional designs due to the substantially orthogonal orientation of such devices, related to both the audio source and the microphone diaphragm. It is further proposed, that by providing multiple non-orthogonally airfoil surfaces, spaced apart from a microphone diaphragm, plosive artifacts may be more effectively deflected away from the diaphragm and the speaker, and thus reduce the impact of such artifacts on the diaphragm. Some of those non-orthogonally structures comprise a conic or pyramidic shape either with or without a rounded apex and are commonly made of a spandex fabric stretched over a support structure which is carried out as a wire frame [8]. However, hoop screens are very common and are especially being used for studio recordings of vocals. But the fact that such structures are quite bulky and also require additional airspace, makes them inappropriate for other areas of application, for instance they cannot be used in combination with any kind of hand held device. Therefore, a better solution to reduce pop noise for small microphone enclosures needs to be found Foam systems Another type of design is made of open cell foam. The main principle is similar to baffle systems, so open cell foam designs can be implemented in almost the same way. It can either be an integral part of the capsule assembly, when used between the microphone grid and the capsule, or an external piece of foam in a variety of shapes with a hollow area into which the microphone is inserted. In theory, the network of foam cells acts as a complex baffle, which prevents the excess displacement of air from a plosive from reaching the microphone diaphragm, yet still allows the desired normal sound waves to pass through to the capsule and diaphragm [8]. 16

17 Figure 3.3.: alternative pop filter design comprised by conic shape [8] The reduction of plosive artifacts is accomplished in the following way: In the foam material the connection between the individual hollow spaces of the pores provide sound travel distances of different lengths between two oppositely positioned surfaces of such a foam body. When a sound wave impacts in the form of a pressure front onto the surface of a foam body, the sound which enters the foam body at one side has no longer the character of a pressure front when exiting at the opposite side because of the different lengths of the individual paths of the sound. Instead the sound exits from different channels with some time delay. The reduced pressure front character on the exiting side results in reduced pop artifacts [12]. As mentioned above the pop filter can either be implemented as an external foam sock or an internal foam layer inside the capsule assembly, but for both cases usually the same material is being used. This material is polyurethane, which is an open-cell foam material made out of polyester. Open cell foam materials of polyurethane provide a skeleton structure having a multiplicity of columns which define a large number of hexagonal cavities, amounting to about 97 percent of the total volume of the foam. During the manufacture of the foam material the size of the cavities or pores within the foam can be varied within wide limits so that an optimum pore size can be obtained for any purpose [13]. In addition, polyurethane is quite inexpensive and easy to shape, which makes it an ideal material. The external open cell foam filters, also referred to as windshields, are 17

18 ball- or cube-shaped configurations which slip on microphones and are used especially for on-site reporting. Of course there is a vast amount of different external pop filter designs, but most combine the following designs: They comprise a skeleton structure of wire on which the foam material is mounted. The wire frame holds the material in shape and provides stability. In order to attenuate wind noise at even relatively high velocities, a fur-like coating is arranged externally over the foam material. Those techniques are often used in the film and television industry, nevertheless we face the same problem as we have for hoop systems due to the dimension of the apparatus. The combined materials generate a relatively thick layer, which reduces the fidelity of the microphone [12]. Figure 3.4.: external piece of foam used for pop sound reduction [14] This leads us to embedded open cell foam filters in the microphone housing. The main principle here is placing a layer of foam between the protective grid of the microphone and the microphone capsule. The effectiveness of such foam filters depends in the first case upon the thickness of the material. One the one hand, a layer that is too thin will hardly make an impact on reducing the pop noise. On the other hand a layer too thick will degrade the sound quality. In most applications this thickness is about a 1/4 inch (6.35 mm), but of course this depends on the number of pores per inch of the used material. The foam material is selected so that its pore size will not absorb sound waves to any appreciable degree within the audible range of the human ear, 60 to 100 ppi (pores per inch) has proven to be particularly satisfactory in the most devices. In order to mitigate turbulence the layer of foam must be placed apart from the microphone diaphragm. The resulting air space also has a great influence on the effectiveness of the device. Furthermore, the spacer must have a certain acoustic radiation impedance in order to reduce the puff noise transmitted to the microphone capsule. Shape and dimension of the spacer should be chosen in a way that thickness equals about 1/4 of the smallest wavelength one wants to record. So compared to hoop systems, a much smaller spacer can be implemented without creating turbulence. Additionally, the spacer can comprise one or more open areas located laterally of the spacer, which allows puff noise to exit the housing. This should improve the plosive reduction. [2], [15]. Internal open cell foam filters can also be extended with an additional layer of material. It has been found that, when arranging two such foam covers spaced from one another, an improvement of the pop shield effect occurs without changing or dampening 18

19 1... air flow generated by speaker 2... protective grid 3... open cell foam layer 4... microphone capsule 5... spacer 6... open area 7... housing Figure 3.5.: embedded piece of foam inside a microphone housing the output level or the frequency characteristics of the microphone. The two foam bodies can even be comprised of open-cell polyurethane foam with 80 ppi, while the spacer in between the two layers can either be an air space or a layer out of a second open-cell foam with 20 to 40 ppi [12]. Nevertheless the biggest problem caused by foam material is their durability. Foam socks, as well as internal foam layers, tend to physically deteriorate over time. The material starts to crumble and as a result, foam particles often fall into the microphone head or even the microphone capsule, causing damage and reducing performance. Therefore, it is necessary to replace the foam material every now and then. On the other hand the main advantage is that internal foam filters can be combined even with small microphone enclosures, while filtering unwanted pop artifacts [16] Microphone construction When focusing on the microphone itself, the construction as well as size and transducer principle make a notable impact on pop noise. Martin Schneider, who carried out a pop sensitivity measurement in 1993, investigated these properties and came to results presented in the following subsections Transducer principle Pop sensitivity is directly correlated with the transducer principle [7]. A distinction is made between pressure and pressure-gradient transducing microphones. Pressure transducing condenser microphones have an enclosed volume behind its diaphragm, thus they have one entry port for the acoustic signal, which is in front of the diaphragm. They respond uniformly to sound pressure from every direction, which means their polar pattern is omnidirectional. These types of microphones tend to have their resonance frequency in the upper range of their frequency range. This is due to a high diaphragm tension. When a pop sound or any other low frequency turbulence hits the microphone 19

20 there will only be a limited excursion of the diaphragm. This signifies a relatively low pop sensitivity. In comparison, a pressure gradient microphone receives the sound pressure difference of a sound wave, which has entered from the front and the back of the capsule. The diaphragm is under less mechanical tension, which results in a resonance frequency that lies somewhere in the middle of the working frequency range. Therefore it is more susceptible to low frequency bursts, which yields a higher pop sensitivity [7] Design of the cage, capsule and diaphragm The pop sensitivity of pressure gradient microphones can be slightly attenuated by a relatively large wire-mesh cage surrounding the microphone capsule [7]. The reason why large cages have this dampening effect is because of acting similar to a pressure chamber. This leads to smaller dislocations of the diaphragm. Further, combinations of different mesh layers inside a cage have shown good results in pop reduction. As an example of a particular measurement, a large cage with 3 mesh layers reduced the pop sound pressure level by 7 db. Another advantage of larger microphone cages is that the diaphragm is not as exposed to internal reflections as in smaller constructions. In this measurement, it was also shown that a central support of the diaphragm results in less pop sensitivity. In general these supporting constructions can be found in broadcast and vocalist microphones. Combined with a well-constructed mesh network pop sounds can be systematically attenuated while having little influence on the actual acoustic signal [7]. Finally a ranking of microphones can be made, from highest to lowest pop sensitivity: 1. miniature, cylindrical microphones, 2. large diaphragm studio microphones, free diaphragm 3. large diaphragm studio microphones, centrally supported diaphragm 4. dedicated vocalist microphones [7] In the following measurement of this thesis particularly miniature, cylindrical microphones will be examined. 20

21 4. Pop sensitivity measurement Experiments in the past proved that the human speech does not deliver acceptable results for a standardized pop signal. Thus, for an objective and reproducible measurement a pop generator is needed. Standard test equipment and measuring methods have been developed and are defined in DIN EN : Measurement according to EN In this standard a simplified measurement is proposed in Annex B. In a research work of Dr. Wollherr from 1991 it was shown, that this measurement has given the best results in simulating the air stream of a plosive [17]. Besides being tolerant regarding the position of the microphone under test, it yields a good conformity with subjective hearing tests [18]. The goal is to measure the reaction of a microphone to a defined pop stimulus, which comes as close as possible to human plosives. According to figure 4.1, a sine signal with a frequency of 5 Hz is fed to a loudspeaker. The loudspeaker, by preference a sub-woofer with a very low resonance frequency, is enclosed by a metal plate forming a pressure chamber. Inside this enclosed chamber a sound pressure level of 140 db has to be produced. The resulting air motion is forced to escape through 9 holes with a diameter of 4.4 mm, arranged in a quadratic array. The distance between two adjacent holes is 10 mm, measured from the center of the holes. It is recommended to blunt the edges of the holes in order to decrease the occurrence of turbulence [19]. Another hole, that is 30 mm apart from the three-by-three hole arrangement, is needed for the calibration microphone, which is used to determine the sound pressure level inside of the pressure chamber. The diameter of a typical measurement microphone is 12.7 mm, as a consequence the diameter of the hole was chosen to be 13 mm. The microphone under test is placed 10 cm away from the middle of the hole arrangement. The measurement setup of EN60268 is depicted in figure 4.1. It is further proposed, in case of measuring the coefficient of transmission, to use some kind of isolation box, inside which 5 cm of sound absorbing material is attached to the walls. Though determining the microphone s transmission factor is not necessary for the measurement, we considered using this box to be appropriate for better free field conditions and also reducing impact from diffuse sound (fig. 4.2). 21

22 Figure 4.1.: Measurement setup [19] Figure 4.2.: Measurement setup for coefficient of transmission [19] In the general case the results of this measurement are A-weighted. In this thesis only small microphones which are highly sensitive to pop stimuli, are being tested. Since the equivalent pop levels are expected to be far beyond the 40 db equal-loudness contours, it is reasonable not to use any weighting. The results of an A-weighting would only conform to low level signals, accordingly it seems hard to justify the use of A-weighted data only to qualify pop sensitivity [7]. 22

23 Reasons for simplified measurement Setup I, according to EN60268, chapter 18.5, proposes a single pop impulse that is generated by a low-pass filtered step function (fig. 4.3). The loudspeaker is covered by a solid plate as in the simplified setup II from Annex B, but with one single hole with a diameter of 10 mm (fig. 4.4). Experiments from the past performed by Dr. Wollherr have shown that setup I yields more inaccurate results than setup II. Figure 4.3.: Measurement setup I [19] Figure 4.4.: Vent of measurement setup I [19] If a microphone is positioned inside the pressure chamber the course of the pressure over time is continuously smooth, as shown in figure 4.5 at the bottom. Exactly at the vent, inharmonic components are added to the pressure signal. This can be derived from the slightly uneven waveform (fig. 4.5, second from bottom). In a distance of more than 2 cm from the hole the signal features a superposition of noise which is due to turbulence of the pop stimulus (fig. 4.5, upper two graphs). 23

24 Figure 4.5.: Sound pressure over time inside and outside the loudspeaker chamber [18] However, the amount of turbulence is not the same in every position in front of the vent. The closer a microphone capsule is placed to the vent, the greater the mean velocity of the air stream is. This is what has been pointed out in section 2.1. Besides, there is a certain amount of fluctuation of air stream velocity that is highly dependent from the position off-axis. In figure 4.6 these two quantities are pointed out, for a microphone in a distance of 5 mm (drawn through line), respectively 50 mm (dashed line) [18]. Bearing this in mind, microphones could not easily be classified in terms of pop sensitivity, if they were differently positioned or different in size. Setup II on the other hand, uses a three-by-three hole array as it has been described in 4.1. The pop stimulus is generated by a quasi-stationary 5 Hz signal that guarantees simpler data evaluation with a standard level meter than the step function impulse response of setup I [7]. 24

25 Figure 4.6.: Air stream velocity and its amount of fluctuation [18] In another measurement a contrasting juxtaposition between these two setup types has been made (fig. 4.7). A microphone was placed in front of the vent (solid line), resp. hole-array on the axis as well as 10 mm off-axis (dotted line). Setup II yields much better alignment of the frequency responses than setup I, where there is an almost constant difference of the EPL by approximately 10 db. For a microphone with a metal wire gauze there is still no satisfying resemblance of the frequency responses within Hz in setup I. Again, setup II proves to be more consistent. Additionally in subjective hearing tests setup II presented good results in terms of similarity to real plosives [18], because the air stream at the hole-array comes close to the conditions of a human mouth. Figure 4.7.: Comparison of both measurement setups: setup I (left), setup II (right) [18] Due to these facts it is more reasonable to use the simplified measuring setup from EN60268, Annex B, as it also proves to be less critical when it comes to reproducibility of the measurement. 25

26 Equivalent pop level EPL The equivalent pop level, further abbreviated EPL, is a sound pressure level which determines a microphone s sensitivity towards a pop stimulus. The EPL is also used to categorize the attenuation that is given for a certain pop screen, wind shield or any other pop damping material. For comparability, the sound pressure of a pop sound will be normalized respectively to the free-field sensitivity at 1 khz. The EPL is calculated as follows: ( ) Vpop EP L = 20 log + 94 [db SPL] (4.1) V sens EP L... equivalent pop level, V pop... voltage measured at the microphone output, V sens... mic. output voltage at 1 khz and 1 Pa (free-field sensitivity: 1 Pa = 94 db SPL) log() is the logarithm to the base of 10 [7] The free-field sensitivity of a microphone is measured at 1 Pa. Related to the threshold of hearing of 20 µpa, this yields a sound pressure level of 94 db: ( ) 1 Pa 20 log = 93, 98 db (4.2) 20 µpa The sensitivity S of a microphone is typically given as S = V sens 1 Pa in [ ] mv P a (4.3) V sens = S 1 Pa (4.4) Entering the quantities of 4.2 and 4.4 into equation 4.1 gives where [ ( ) ( )] Vpop 1 Pa EP L = 20 log + log (4.5) S 1 Pa 20 µpa [ ( ) ( )] Vpop 1 EP L = 20 log + log (4.6) S 20 µpa ( ) Ppop EP L = 20 log in [db SPL] (4.7) 20 µpa (4.8) P pop = V pop S... pop sound pressure in [Pa] (4.9) 26

27 4.2. Measurement equipment The following chapter lists and describes the components of the measurement setup according to EN : 2010 used specifically for this bachelor thesis. The data sheets of all devices are listed in the appendix (B). Audio interface: Focusrite Saffire PRO 40 Amplifier: Apart PA4060 Loudspeaker: Visaton W Ω Audio analyzer: NTi Audio XL2 The different microphones used in this experiment are the following: NTi Audio M2230 ICC Electret Condenser Microphones (omnidirectional characteristic) ICC Electret Condenser Microphones (unidirectional characteristic) ST Microelectronics MP34DT01 MEMS audio sensor omnidirectional digital microphone with STM32F407 microcontroller Philips LFH3500 SpeechMike Pop generator Since the pop generator must be run with a 5 Hz sinusoidal signal at 140 db, a speaker with a high resilience and a low resonant frequency (36 Hz) was chosen. Furthermore, the low cost of this apparatus made it suitable for this purpose, so in case of damage it could be replaced more easily. Figure 4.8 shows the dimensions of the speaker. Figure 4.8.: dimensions of the Visaton W Ω 27

28 As explained in 4.1, the speaker must be covered with a plate to generate the desired air stream. Different to the suggested metal, the plate is made of acrylic glass, in consequence of the simplicity of handling, which is also leading to lower costs and definitely to a lower weight of the pop generator. The acrylic plate is furnished with vents and an opening for the measuring microphone according to 4.1. It is displayed in fig Figure 4.9.: holes at the center of the covering plate Finally, the two components were combined and a holder was applied. Fig and fig show front and rear view of the pop generator construction. To ensure accurate signal transmission a 6.35 mm mono jack was mounted to the socket for the loudspeaker. Figure 4.10.: front view of the pop generator Figure 4.11.: rear view of the pop generator The 5 Hz sinusoidal signal is created on a PC using a waveform generation software TestGenerator in Steinberg Nuendo 4. By means of the audio interface Focusrite Saffire PRO 40 the input signal is fed into the pop generator via a 6.35 mm TS cable. Since this configuration results in a maximum sound pressure level of 120 db only, an additional amplifier needs to be added. The Apart PA4060 power amplifier produces an SPL inside the enclosed chamber of the pop generator of up to 140 db. This is detected by the NTi Audio M2230 measurement microphone and analyzed by the NTi Audio XL2. To measure this sound pressure level, the measurement microphone has to be placed into the designated opening, as can be seen in fig

29 Figure 4.12.: detecting the sound pressure level inside the pop generator Isolation box Reaching an SPL of 140 db inside the pressure chamber and thereby fulfilling the standard, the different microphones and acoustic filters can now be measured. Therefore, they are placed inside an isolation box as stated in section 4.1 and fig The isolation box consists of a press board cube with mm inner dimensions. Furthermore, 5 cm of acoustically absorbing material is mounted to the interior walls. For this purpose convoluted foam is used. To hold the microphone and the pop filter in place, alligator clips are fixed to an improvised stand, so they can be moved toward or away from the surface of the pop generator. The inside of the isolation box and the corresponding mounting can be observed in fig and Figure 4.13.: isolation box Figure 4.14.: alligator clips used as microphone and filter mounting 29

30 Microphones Besides the NTi Audio M2230, which is only used for calibration purposes, four different microphones were tested in this measurement assembly. The ICC Electret Condenser Microphones come in two different versions: omnidirectional and uni-directional. Both types are being used in this measurement, but since they are optically identical the following pictures represent the omni- as well as the unidirectional microphone. Further, the omnidirectional type is being referred as mic 1 and the uni-directional as mic 2. Figures 4.15 and 4.16 show two different views of the ICC Electret Condenser Microphone. Figure 4.15.: mic 1 resp. mic 2 (side view) Figure 4.16.: mic 1 resp. mic 2 (profile) For a better comprehension of proportions there is a ruler placed near the microphone. The ST Microelectronics MP34DT01 MEMS microphone is displayed in fig and 4.18, furthermore referred as mic 3: Figure 4.17.: mic 3 (side view) Figure 4.18.: mic 3 (profile) Aside from that, a dictation device, the Philips LFH3500 SpeechMike, is tested regarding its reaction towards pop sounds, as it is normally used very close to the lips of the speaker. Since this apparatus is quite bulky, it cannot be used in combination with the mounting system in the isolation box. Instead a typical microphone stand was turned into a mounting. The isolation box was not used for this measurement. Additionally, customized filters had to be created for this microphone, as will be pointed 30

31 out in subsection In fig and 4.20 the Philips LFH3500 SpeechMike, in the following labeled mic 4, is displayed in front view and profile. Figure 4.19.: mic 4 (front view) Figure 4.20.: mic 4 (profile) Acoustic filters As seen in the figures above, the electret and the MEMS microphones both have a diameter of less than 10 mm. Combining these two microphones with an conventional pop screen of fabric or wire gauze would not be adequate. No filters of such small diameters are being sold, so they had to be customized for the purpose of the measurement. For this reason 5 different pop filters were made out of 3 different materials. For filt 1 and filt 2 spandex materials were stretched over a wire framed hoop, resulting in two hoop pop filters. Figure 4.21 shows the front view of filt 1, a hoop filter with 30 mm diameter and 2 mm in depth. Figure 4.22 shows the filt 2, also a hoop filter with 10 mm diameter and 4 mm in depth. Figure 4.21.: filt 1 (front view) Figure 4.22.: filt 2 (front view) For filt 3 and filt 4 a layer of polyurethane material was used. filt 3 has a size of about mm and can be seen in figures 4.23 and 4.24 below. 31

32 Figure 4.23.: filt 3 (front view) Figure 4.24.: filt 3 (side view) The second foam filter, filt 4 is thinner having a size of approx mm. filt 4 is displayed in 4.25 and 4.26 below. Figure 4.26.: filt 4 (side view) Figure 4.25.: filt 4 (front view) The last miniature filter is comprised of a special high porous material called Aerogel. Aerogel is a silicate based ultralight material derived from a gel, in which the liquid component of the gel has been replaced with a gas. About 95 to 98 % of this material is represented by pores, all under 100 nm, which, in theory, makes it a perfect pop filtering material. But this also brings along the drawback of this material, since its high amount of pores result in a very fragile structure. It is almost impossible to handle, even small pressure applied could lead to a demolition of the structure [20]. The Aerogel filter, filt 5, has a diameter of 20 mm and a depth of 8 mm. A thinner filter was not tested due to the risk of destroying the fragile structure. Figure 4.27 shows the Aerogel pop filter filt 5. 32

33 Figure 4.27.: filt 5 - Aerogel (front view) Due to the already mentioned bulky nature of the Philips LFH3500 SpeechMike, additional filters with a larger diameter were exclusively used for this device. In the following figures 4.28 and 4.29 filt 6 is depicted. It is comprised of polyurethane material with a size of mm. Figure 4.28.: filt 6 (front view) Figure 4.29.: filt 6 (side view) Additionally a metal gauze, referring to baffle systems in 3, was being used in combination with the dictating machine. The trapezoidal baffle filter filt 7 can be seen in fig Finally the dictation microphone was combined with a customary hoop filter with a diameter of approx. 150 mm. Figure 4.31 shows the filt 8, a commonly available fabric filter. 33

34 Figure 4.30.: filt 7, a baffle filter Figure 4.31.: filt 8, a fabric hoop filter 4.3. Measuring process In this section the methods of obtaining measured values and calculating measurement results are described Microphone sensitivity In order to calculate the EPL according to 4.1.2, the microphone sensitivity for a sinusoidal signal at 1 khz and 1 Pa must be given. In general this quantity is available in the unit [mv/pa] and can be taken from the data sheet or manual. If the sensitivity is specified in dbv, dbfs (in the case of digital microphones) or any other unit, a conversion is necessary Analog microphones The sensitivity of an analog microphone tells how many volts the output signal will be for a standardized SPL at 1 Pa. The linear unit of [mv/pa] can be expressed logarithmically in decibel, e.g. dbv with respect to 1 V [21]. Equation 4.10 shows this connectedness: ( ) Sensitivity mv/pa Sensitivity dbv = 20 log (4.10) Output AREF where Output AREF is the reference output ratio of 1000 mv/pa [21]. Rearranging this equation it gives the sensitivity in [mv/pa]: Sensitivity mv/pa = 10 ( Sensitivity dbv/20) Output AREF (4.11) The sensitivity of mic 4 is given as S = 37 dbv. Expressed in [mv/pa] this yields: Sensitivity mv/pa = / mv/pa (4.12) = mv/pa (4.13) 34

35 Digital microphones The sensitivity of digital microphones, such as the MEMS microphone mic 3 used in this thesis, only depends on a single parameter, maximum acoustic input. The full-scale digital word that these microphones produce at the output (0 dbfs) is mapped to the microphone s maximum acoustic input. This means that the sensitivity is the difference between this maximum input and the 94 db SPL reference. For example, if a digital microphone s maximum input SPL is 120 db, which refers to an output of 0 dbfs, the sensitivity is then calculated as the difference between the output at 94 db SPL and 120 db expressed in dbfs [21]: Sensitivity dbfs = Reference db Input max, db (4.14) Sensitivity dbfs = 94 db 120 db (4.15) Sensitivity dbfs = 26 dbfs (4.16) In general the sensitivity of digital microphones is measured as a percentage of the full-scale output that is generated by a 94 db SPL input. The conversion equation is ( ) Sensitivity %FS Sensitivity dbfs = 20 log (4.17) Output DREF The sensitivity and output level of digital microphones are given as peak levels because they are referred to the full-scale word, which is peak value. For example a 94 db SPL sinusoidal input signal will give a 26 dbfs peak output level, or a 29 dbfs RMS level [21]. For mic 3 the sensitivity is given by 26 dbfs. Since this is a peak value, the corresponding RMS value is 29 dbfs. The reason for calculating the sensitivity for the RMS level is simply to guarantee better comparability to the sensitivity of analog microphones, which is always referring to a sinusoidal reference at 94 db SPL RMS and 1 khz. Assuming 0 dbfs conforms to 18 dbu, as it is stated in the EBU Technical Recommendation R , the sensitivity is calculated as follows: 18 dbu = 0 dbfs (4.18) 11 dbu = 29 dbfs (4.19) ( ) Usens 11 dbu = 20 log (4.20) V / V = U sens (4.21) Thus the sensitivity of mic 3 is mv/pa. U sens = mv (4.22) 35

36 Electret condenser microphones The sensitivity of electret condenser microphones (ECM) is either given in [mv/pa] or in [V/µbar]. In the latter case, a conversion has to be performed to gain the right quantity for calculating the EPL. The relation between Pascal and bar is 1 bar = 10 5 Pa (4.23) 1 µbar = 10 1 Pa (4.24) For the omnidirectional ECM mic 1 the sensitivity is 60 db, where 0 db refers to 1 V/µbar. In linear terms this yields: ( ) S mic 1 60 db = 20 log 1 V/µbar (4.25) S mic 1 = /20 1 V/µbar (4.26) = 10 3 V/µbar (4.27) This term has to be translated into the unit of [mv/pa] using the relation from 4.24: The sensitivity of mic 1 amounts to 10 mv/pa. S mic 1 = 10 3 V/µbar (4.28) = 10 3 /10 1 V/Pa (4.29) = 10 mv/pa (4.30) The same calculation has to be carried out for mic 2 which is unidirectional and has a sensitivity of 67 db: S mic 2 = /20 1 V/µbar (4.31) = V/µbar (4.32) = 4.47 mv/pa (4.33) Table 4.1 shows the the sensitivities of all microphones used in the measurement: Table 4.1.: Sensitivities of the used microphones Microphone mic 1 mic 2 mic 3 mic 4 Sensitivity [mv/pa]

37 Measured and calculated EPL In order to estimate the EPL of the microphone and pop filter combinations displayed in sections and 4.2.4, two different approaches can being used. Since both of the electret condenser microphones have a three-pin XLR connector, they can be directly plugged into the NTi Audio XL2 (4.32), which has already been used for calibration purpose. Figure 4.32.: NTi Audio XL2 - Handheld Audio and Acoustic Analyzer [22] This apparatus simplifies the computation of the EPL value a lot, since it considers the microphones sensitivity. In other words, if the free field sensitivity value at 1 khz, which can be taken from data sheets, is fed into the audio analyzer, it delivers the proper EPL value. It has to be noted, that beside the advice of the norm paper no frequency weighting is being used here (see 4.1). The XL2 has two time weighting modes, after a short testing phase the slow time weighting (1000 ms) delivered more stable and convincing results. This can be referred to the low frequency input signal of 5 Hz: Since one period of the input signal takes 200 ms, the fast time weighted periodic time of 125 ms is not adequate for measuring. Unfortunately, the XL2 audio analyzer is not compatible with the other two microphones, since they only feature a USB connecting and therefore must be plugged directly into a personal computer. After recording the signal, the EPL values must be computed according to equation in due consideration of the microphone s sensitivity. In order to determine the voltage V pop of the signal the software Steinberg Nuendo 4 was used. The signal properties were investigated by the tool Statistics. The average RMS value, e.g db as shown in fig. 4.33, could be expected to be a good approximation compared to the voltage values, which the NTi XL2 is working with at slow time weighting. 37