sensors

Article Comparison between and Non-Resonance Type Piezoelectric Acoustic Absorbers

Joo Young Pyun , Young Hun Kim , Soo Won Kwon, Won Young Choi and Kwan Kyu Park * Department of Convergence Mechanical Engineering, Hanyang University, Seoul 04763, Korea; [email protected] (J.Y.P.); [email protected] (Y.H.K.); [email protected] (S.W.K.); [email protected] (W.Y.C.) * Correspondence: [email protected]

 Received: 26 November 2019; Accepted: 18 December 2019; Published: 20 December 2019 

Abstract: In this study, piezoelectric acoustic absorbers employing two receivers and one transmitter with a controller were evaluated. Based on the target and resonance of the system, resonance and non-resonance models were designed and fabricated. With a lateral size less than half the wavelength, the model had stacked structures of lossy acoustic windows, polyvinylidene difluoride, and lead zirconate titanate-5A. The structures of both models were identical, except that the resonance model had steel backing material to adjust the center . Both models were analyzed in the frequency and time domains, and the effectiveness of the absorbers was compared at the target and off-target frequencies. Both models were fabricated and acoustically and electrically characterized. Their reflection reduction ratios were evaluated in the quasi-continuous- and time-transient modes.

Keywords: resonance model; non-resonance model; piezoelectric material

1. Introduction Piezoelectric transducers are used in various fields, such as nondestructive evaluation, image processing, acoustic signal detection, and energy harvesting [1–4]. navigation and ranging (SONAR) is a technology for acoustic signal detection that can be used to detect objects under water. has been developed to prevent detection by such SONAR systems. Therefore, studies have focused on reducing the reflection of sound to prevent detection by SONAR systems. Wedge shaped structures or coatings have been used as passive sound absorbers to minimize underwater acoustic reflection. The coating is durable, but needs improvement. Therefore, minimizing detection in the low frequency range is challenging. In the low frequency range, the wavelength of the signal is several tens of centimeters. The thickness of the wedge shaped structure or coating should be at least several tens of centimeters to minimize acoustic reflection. Attaching a thick, passive sound absorbing material to a submarine hull increases the weight of the submarine and interferes with its propulsion. Sound can also be reduced using the wave scattering method. However, this technique only works under hydrostatic pressure [5,6]. Therefore, active sound absorbing materials have been developed to overcome the disadvantages of passive sound absorbing materials [7–10]. Lafluer et al. used piezorubber to realize an active sound absorbing material [11], whereas Howarth et al. used piezoelectric composites to reduce reflection. Although this material exhibited high performance at one frequency (5.4 kHz), it was not an active sound absorbing material composed of one structure. [12]. Chang et al. reduced sound reflection using two layers of a 1–3 piezoelectric composite. This resulted in an attenuation performance of 20 dB in the range of 6–10 kHz [13]. Accordingly, piezoelectric acoustic absorbers can be classified into resonance and non-resonance types based on

Sensors 2020, 20, 47; doi:10.3390/s20010047 www.mdpi.com/journal/sensors Sensors 2020, 20, 47 2 of 17

Sensors 2019, 19, x FOR PEER REVIEW 2 of 17 previous studies. However, it is difficult to compare them due to the different approaches involved. Therefore,the different we believe approaches that only involved. resonance Therefore, and non-resonance we believe models that on withly similarresonance structures and non-resonance should be compared.models with In this similar study, structures resonance andshould non-resonance be compared. models In this suitable study, for resonance the target and frequency non-resonance were designedmodels via suitable mathematical for the analysis.target frequency Based on were the designs, designed the via resonance mathematical and non-resonance analysis. Based models on the weredesigns, fabricated the resonance and evaluated. and non-reso The absorbernance was models configured were fabricated in the form and of tiles,evaluated. and the The lateral absorber size of was theconfigured structure was in the equal form to of half tiles, of theand wavelength. the lateral size Our of proposed the structure resonance was equal and to non-resonance half of the wavelength. models wereOur fabricated proposed by stackingresonance commercially and non-resonance available leadmodel zirconates were titanatefabricated (PZT). by Variousstacking studies commercially have previouslyavailable investigated lead zirconate the titanate stacking (PZT). or direct Various fabrication studies of have devices previously [14–19]. investigated the stacking or direct fabrication of devices [14–19]. 2. Materials and Methods 2. Materials and Methods 2.1. Concept 2.1. Concept In the case of a piezoelectric material, along the thickness direction is observed when anIn electrodethe case ofis a piezoelectric applied, and material, the sound vibration spreads along around the thickness both the direction front and is observed rear sides when of an theelectrode piezoelectric is applied, material. and Thethe sound resonance spreads frequency around rangeboth the of afront piezoelectric and rear sides material of the depends piezoelectric on itsmaterial. thickness. The The resonance thickness frequency of a piezoelectric range of materiala piezoelectric and its material resonance depends frequency on its are thickness. inversely The proportional,thickness of i.e., a piezoelectric the thinner thematerial piezoelectric and its resona material,nce thefrequency higher are is itsinversely resonance proportional, frequency i.e., and the vicethinner versa. the piezoelectric material, the higher is its resonance frequency and vice versa. InIn this this paper, paper, we we present present a modela model that that resonates resonates in thein the low low frequency frequency region region of theof the target target frequencyfrequency and and a non-resonance a non-resonance model model that deviates that deviat from thees targetfrom the frequency. target Thefrequency. proposed The resonance proposed andresonance non-resonance and non-resonance models incorporated models incorporated a function that a function cancelled that a certaincancelled portion a certain of the portion incident of the soundincident waves. sound waves. TheThe incident incident and and reflected reflected waves waves mustmust bebe separated to to cancel cancel the the incident incident sound sound wave. wave. Two Tworeceiving receiving sensors sensors were were required required to toseparate separate the the incident incident and and reflected reflected waves, waves, as asthe the signals signals were weremeasured measured based based on on the the overlap overlap between between these these two waves.waves. TheThe incidentincident ( P(+P) and) and reflected reflected (P –()P– ) acousticacoustic pressures pressures were were the the input input and and output output of theof the system, system, respectively respectively.(Figure (Figure1). We 1) We calculated calculated the the incidentincident and and reflected reflected acoustic acoustic sensitivities sensitivities based based on on the the diff differenterent receiving receiving sensitivities sensitivities of theof the two two sensorssensors obtained obtained using using the the Krimholtz–Leedom–Matthaei Krimholtz–Leedom–Matthaei (KLM) (KLM) model. model.

Figure 1. FigureIncident 1. Incident and and reflected reflected waves. waves. The proposed resonance and non-resonance models are shown in Figure2. Generally, PZTs with The proposed resonance and non-resonance models are shown in Figure 2. Generally, PZTs with low resonance frequencies are difficult to manufacture. Therefore, the frequency was reduced by low resonance frequencies are difficult to manufacture. Therefore, the frequency was reduced by attaching steel to the rear side of the PZT. The model with steel at the rear side of the PZT was the attaching steel to the rear side of the PZT. The model with steel at the rear side of the PZT was the resonance model, whereas that without steel was the non-resonance model, as shown in Figure2. resonance model, whereas that without steel was the non-resonance model, as shown in Figure 2. Both models consisted of one transmitting sensor, two receiving sensors, and three acoustic windows. Both models consisted of one transmitting sensor, two receiving sensors, and three acoustic windows. Polyvinylidene difluoride (PVDF) could be used as a receiving sensor to measure the sound pressure Polyvinylidene difluoride (PVDF) could be used as a receiving sensor to measure the sound pressure without disturbing the sound propagation. The transmitter used PZT-5A, which is insensitive to without disturbing the sound propagation. The transmitter used PZT-5A, which is insensitive to temperature changes and has excellent transmission capabilities. The resonance and non-resonance temperature changes and has excellent transmission capabilities. The resonance and non-resonance models presented herein were used underwater. Therefore, a physical gap was required between the

Sensors 2020, 20, 47 3 of 17

Sensors 2019, 19, x FOR PEER REVIEW 3 of 17 models presented herein were used underwater. Therefore, a physical gap was required between the two receiving sensors. The The acoustic acoustic window window material material Rho-c Rho-c was was used used because because it it is is a lossy absorber with a physical gap and an impedance similar to that of water.water.

Sensors 2019, 19, x FOR PEER REVIEW 3 of 17 two receiving sensors. The acoustic window material Rho-c was used because it is a lossy absorber with a physical gap and an impedance similar to that of water. Figure 2 2.. ConceptConcept behind behind the the ( (aa)) resonance resonance and and ( b) non-resonance models. 2.2. Analytical Model 2.2. Analytical Model We adopted a mathematical analysis that required less time compared to the finite element method We adopted a mathematical analysis that required less time compared to the finite element to analyze the resonance and non-resonance models composed of multiple layers. method to analyze the resonance and non-resonance models composed of multiple layers. Some basic mathematical analysis methods have been developed for analyzing piezoelectric Some basic mathematical analysis methods have been developed for analyzing piezoelectric transducers [20–22], of which the KLM model is the simplest. In the KLM model, transducers are transducers [20–22], of which the KLM model is the simplest. In the KLM model, transducers are divided into two parts, namely their front and rear ends. In addition, matching layers can be added to divided into two parts, namely their front and rear ends. In addition, matching layers can be added the front and rear ends of the transducer. Therefore, the KLM equivalent circuit model can be used to to the front and rearFigure ends 2. Conceptof the transducer. behind the Theref(a) resonanceore, the and KLM (b) non-resonance equivalent circuit models. model can be used analyze a multi-layer model. As shown in Figure3, our model consisted of one transmitting sensor and to analyze a multi-layer model. As shown in Figure 3, our model consisted of one transmitting sensor two receiving sensors, with three piezoelectric materials and three matching layers. The equivalent and2.2. Analyticaltwo receiving Model sensors, with three piezoelectric materials and three matching layers. The models included , a transformer, piezoelectric materials, and matching layers. equivalentWe adopted models aincluded mathematical capacitors, analysis a transforme that requr,ired piezoelectric less time material compareds, and to thematching finite layers.element method to analyze the resonance and non-resonanceεA models composed of multiple layers. c = 1 (1) Some basic mathematical analysis methods0 haved been developed for analyzing piezoelectric transducers [20–22], of which the KLM model is the simplest. In the KLM model, transducers are –c0 divided into two parts, namely their front cand0 = rear ends.  In addition, matching layers can be added(2) k2sinc ω to the front and rear ends of the transducer. Thereft ore, ωthe0 KLM equivalent circuit model can be used to analyze a multi-layer model. As shown in Figurekt 3, our model consisted of one transmitting sensor φ = p sinc( f /2 f0) (3) and two receiving sensors, with three piezoelectric2 f0c0Zc materials and three matching layers. The equivalent models included capacitors, a transformer, piezoelectric materials, and matching layers.

Figure 3. The Krimholtz–Leedom–Matthaei (KLM) model.

(1) 𝜀𝐴 𝑐 = 𝑑

–𝑐 (2) 𝑐 = 𝑘 𝑠𝑖𝑛𝑐𝜔 𝜔

(3) 𝜙= 𝑠𝑖𝑛𝑐(𝑓/2𝑓) FigureFigure 3.3. The Krimholtz–Leedom–Matthaei (KLM)(KLM) model.model.

Equations (1) and (2) were used to calculate the , and Equation (3) shows (1) the 𝜀𝐴 transformer operation operation that that converts converts the the mechanic mechanical𝑐 =al properties properties to acoustic to acoustic properties. properties. Here, Here, 𝑐 isc theis 𝑑 0 staticthe static , capacitance, 𝜀 isε theis the , permittivity, 𝐴A is1 isthe the area, area, and and 𝑑d isis the the thickness of the piezoelectricpiezoelectric material. k𝑘 is the electrical coupling constant. 𝑓f is–𝑐 the resonance frequency. 𝑠𝑖𝑛𝑐sinc((x𝑥))==sin(𝜋𝑥x)(2)/𝜋𝑥 x, material. t is the electrical coupling constant.𝑐 = 0 is the resonance frequency. sin π /π , 𝑘 𝑠𝑖𝑛𝑐𝜔 and 𝜔 =2𝜋𝑓 . The electrical ratio is represented by𝜔 𝜙 . The propagation constant γ can be calculated based on the velocity of sound and the quality factor. (3) 𝜙= 2𝜋 𝑠𝑖𝑛𝑐(𝑓𝑗/2𝑓) (4) γ =𝑗1 − 𝑉 2𝑄 Equations (1) and (2) were used to calculate the capacitances, and Equation (3) shows the transformer operation that converts the mechanical properties to acoustic properties. Here, 𝑐 is the static capacitance, 𝜀 is the permittivity, 𝐴 is the area, and 𝑑 is the thickness of the piezoelectric material. 𝑘 is the electrical coupling constant. 𝑓 is the resonance frequency. 𝑠𝑖𝑛𝑐(𝑥) =sin(𝜋𝑥) /𝜋𝑥, and 𝜔 =2𝜋𝑓 . The electrical ratio is represented by 𝜙 . The propagation constant γ can be calculated based on the velocity of sound and the quality factor. 2𝜋 𝑗 (4) γ = 𝑗 1 − 𝑉 2𝑄

Sensors 2020, 20, 47 4 of 17

and ω0 = 2π f0. The electrical ratio is represented by φ. The propagation constant γ can be calculated based on the velocity of sound and the quality factor. ! 2π j γ = j 1 (4) V − 2Q " # cosh γt Zsinhγt M0 = 1 (5) Z sinhγt cosh γt  –j   1 ωc  M1 =  0  (6)  0 1 

 j   1 ω−c  M2 =  0  (7)  0 1  " # ∅ 0 M3 = 1 (8) 0 ∅ " # 1 0 M = (9) 4 1 1 Zba

 γ5t5 γ5t5   cosh 2 Z5sinh 2  M5 =  γ5t5 γ5t5  (10)  1 sinh cosh  Z5 2 2 " # coshγiti Zisinhγiti M = , i = 6, 7, , 10 (11) i 1 sinhγ t coshγ t Zi i i i i ···

Matrices M1–M3 represent the electrical parts of the piezoelectric transducer; M4–M5 represent the mechanical parts; M7 and M9 represent the receiving sensors; M6, M8, and M10 represent the acoustic window matrices; M1 and M2 represent the capacitances. The values of c0 and c0 should be calculated to express the matrices M1 and M2. The permittivity ε, area A1, and thickness d of the piezoelectric material should be considered to obtain the value of c0. The matrix M3 represents the transformer. The transformer is transformed into an acoustic system when the capacitances exhibit electrical performance. M4 and M5 can be used to analyze the piezoelectric material. The piezoelectric matrix contains additional terms such as Z5, γ5, and t5, which are the impedance, propagation constant, and thickness of the piezoelectric material, respectively. M4 and M5 represent the front and rear sides of the piezoelectric material, respectively. Therefore, the total thickness of the piezoelectric material should be substituted by half the actual thickness. M5 and M6 can be derived from the delay line, M0.

Mtotal = M1M2M3M4M5M6M7M8M9M10 (12)

The matrix Mtotal can be constructed as shown in Equation (12). Subsequently, Mtotal is employed to construct the matrices A–D as shown in Equations (13) to (16).

A = Mtotal (1, 1) (13)

B = Mtotal (1, 2) (14)

C = Mtotal (2, 1) (15)

D = Mtotal (2, 2) (16)

V AZF + B Ze = = (17) I CZF + D Sensors 2019, 19, x FOR PEER REVIEW 5 of 17 Sensors 2020, 20, 47 5 of 17 –𝑉 (18) 𝑣= 𝐴𝑍–+𝐵V v = (18) AZF + B The of the transducer 𝑍 can be calculated using Equation (17). The surface velocityThe of electrical the transducer impedance 𝑣 can of be the calculated transducer usingZe can Equation be calculated (18). Therefore, using Equation the transmit (17). Thesensitivity surface ofvelocity the transducer of the transducer can be calculatedv can be calculated using Equations using Equation (17) and (18).(18). Therefore, the transmit sensitivity of the transducerIn the receiving can be model calculated of the using sensor, Equations the matrix (17) andM is (18). constructed in the same way as in the transmissionIn the receiving model. The model of the 𝑉 sensor,can be thecalculated matrix Mbased is constructed on the value in theof 𝑀 same , waywhich as is in the the producttransmission of the model.matrix of The each voltage property,V can the be calculated applied based to the on transducer, the value of andMtotal the, value which of is the the . product of the matrix of each property, the force applied to the transducer, and the value of the resistor.(19) 𝑅 𝑉= 𝐹 𝑍(𝐶𝑅 + 𝐴)R+𝐷𝑅s +𝐵 V = Fs (19) Z (CR + A) + DR + B The transmit characteristics, receiving characteristics,f s ands input impedances of the resonance and non-resonanceThe transmit characteristics, models were receiving calculated characteristics, using the andKLM input method. impedances The frequency of the resonance response and characteristicsnon-resonance were models evaluated. were calculated The pole using and zero the KLMwithin method. the frequency The frequency were specified. response We characteristics developed andwere simulated evaluated. a time The poletransient and zeromodel. within the frequency were specified. We developed and simulated a time transient model. 2.3. Control System 2.3.A Control basic Systemactive control system was built using the feedback controller design with two receiving sensorsA basic(Figure active 4a). control The previous system wasdesign, built shown using thein Figure feedback 4b, controller was built design based with on the two receiving receiving sensitivitysensors (Figure of the4a). sensor The (S), previous design, sensitiv shownity in(R), Figure and 4transmitb, was builtsensitivity based on(T). the The receiving incident – acousticsensitivity pressure of the sensor(P ) and (S), total reflection reflected sensitivity acoustic (R), pressure and transmit (P ) sensitivitywere the input (T). The and incident output acoustic of the – system,pressure respectively. (P+) and total The reflectedincident ( acousticS ) and reflected pressure ( (SP–) )acoustic were the sensitivities input and were output calculated of the system, based onrespectively. the sensitivity The of incident the sensor (S+ )(S and) and reflected the thickness (S–) acoustic of the acoustic sensitivities window were (d calculated) using Equations based on (20) the andsensitivity (21), as ofshown the sensor below. (S ) and the thickness of the acoustic window (d) using Equations (20) and (21), as shown below. S = 𝑗2𝑆𝑠𝑖𝑛(𝑘𝑑)𝑒 (20) S+ = j2Ssin(kd)ejkd (20)

S S=−− =𝑗2𝑆𝑠𝑖𝑛j2Ssin(𝑘𝑑(kd) ) (21) (21) −

FigureFigure 4. 4. BlockBlock diagram diagram of of the the feedback feedback control control (a (a) )control control system system concept concept [12] [12 ]((bb)) previous previous design design ((cc)) new new design. design.

Sensors 2019, 19, x FOR PEER REVIEW 6 of 17

Sensors 2020, 20, 47 6 of 17 Unlike in the previous design (Figure 4c), we calculated the new S and S– values from the different receiving sensitivities of the two sensors calculated using the KLM model. S was + calculatedUnlike using in the Equation previous (22), design as there (Figure were4c), we calculateddifferences the at the new twoS sensors.and S– Svalues– was calculated from the + diusingfferent Equation receiving (23). sensitivities of the two sensors calculated using the KLM model. S was calculated using Equation (22), as there were phase differences at the two sensors. S– was calculated using S =𝑆 𝑒 −𝑆 (22) Equation (23). + jkd S = SBe SA (22) S =−S𝑒−– (23) + –jkd S− = S e (23) The gain 𝐺 was maintained constant, with −its optimal value obtained after designing the model for theThe target gain Gfrequencywas maintained [12]. constant, with its optimal value obtained after designing the model for the target frequency [12]. 2.4. Analysis Results 2.4. Analysis Results The transmit and receiving sensitivities of the models were analyzed in the . All The transmit and receiving sensitivities of the models were analyzed in the frequency domain. the dimensions were normalized with respect to the wavelength of the target frequency, 𝜆, and the All the dimensions were normalized with respect to the wavelength of the target frequency, λ, and period of the target frequency, T, to evaluate the performance in the normalized dimension. The the period of the target frequency, T, to evaluate the performance in the normalized dimension. The dimensions of the resonance and non-resonance models were the same as those of the fabricated dimensions of the resonance and non-resonance models were the same as those of the fabricated structure discussed in the previous section. Both models had the same cross-sectional area of structure discussed in the previous section. Both models had the same cross-sectional area of 0.36𝜆× 0.36𝜆. The heights of the resonance and non-resonance models were 2.7𝜆 and 1.2𝜆, respectively. 0.36λ 0.36λ. The heights of the resonance and non-resonance models were 2.7λ and 1.2λ, respectively. The parameters× used in our simulations are listed in Table 1. The parameters used in our simulations are listed in Table1. Table 1. Simulation parameters. Table 1. Simulation parameters. Resonance (Thickness: λ) Non-resonance Model (Thickness: λ) Resonance (Thickness: λ) Non-Resonance Model (Thickness: λ) Acoustic Window 0.27 0.27 AcousticPVDF Window 0.002 0.27 0.27 0.002 AcousticPVDF Window 0.002 0.27 0.002 0.27 Acoustic Window 0.27 0.27 PVDFPVDF 0.002 0.002 0.002 0.002 AcousticAcoustic Window Window 0.27 0.27 0.27 0.27 PZTPZT 0.4 0.4 0.4 SteelSteel 1.52 1.52 - -

EachEach model model had had one one transmit transmit element element (PZT) (PZT) and and two two receiving receiving elements elements (PVDFs). (PVDFs). TheThe input input impedancesimpedances of of the the transmit transmit elements elements of of both both models models were were calculated calculated as as shown shown in in Figure Figure5 5to to evaluate evaluate thethe frequency frequency response response of of the the system. system. The The input input impedance impedance confirmed confirmed that that the the electrical resonance of theof resonancethe resonance model model was atwas the at target the target frequency frequencyf0. In addition,𝑓. In addition, the input the impedances input impedances of both models of both weremodels affected were byaffected the steel by blocks,the steel which blocks, reduced which r theeduced resonance the resonance frequency frequency of the PZT of the from PZT3.6 fromf0 to f3.6𝑓0 at theto cost𝑓 at of the the cost system of the bandwidth. system bandwidth. Notably, both Notably, models both employed models the employed same piezoelectric the same piezoelectric transmitter andtransmitter thus had and identical thus ohadff-resonance identical characteristicsoff-resonance suchcharacteristics as capacitance. such as The capacitance. output pressure The ofoutput the systempressure for of a the small system voltage for a input small can voltage be calculated input can as be shown calculated in Figure as shown6. As in the Figure resonance 6. As the model resonance had amodel high quality had a high factor quality around factor the around resonance, the resonance, it had a higher it had transmit a higher sensitivity transmit sensitivity (150 Pa/V (150 at f0 Pa/V) than at that𝑓 ) ofthan the non-resonancethat of the non-resonance model (102 Pa model/V at 3.6 (102f0). ThePa/V transmit at 3.6𝑓 sensitivity ). The transmit results demonstratedsensitivity results the benefitdemonstrated of the resonance the benefit structure of the atresonance the target structure frequency, at asthe this target structure frequency, had 15 as times this structure higher transmit had 15 sensitivitytimes higher than transmit the non-resonance sensitivity than structure the non-resonance (10 Pa/V at f 0structure). (10 Pa/V at 𝑓).

FigureFigure 5. 5.Simulation Simulation of of the the input input impedances impedances of of the the ( a(a)) resonance resonance and and ( b(b)) non-resonance non-resonance models. models.

Sensors 2019, 19, x FOR PEER REVIEW 7 of 17 SensorsSensors2020 2019, ,20 19,, 47 x FOR PEER REVIEW 7 ofof 1717 Sensors 2019, 19, x FOR PEER REVIEW 7 of 17

Figure 6. Simulation of the transmit sensitivities of the (a) resonance and (b) non-resonance models. Figure 6. Simulation of the transmit sensitivities of the (a) resonance and (b) non-resonance models. Figure 6. Simulation of the transmit sensitivities of the (a) resonance and (b) non-resonance models. TheFigure receiving 6. Simulation sensitivities of the transmit of the twosensitivities receivers of theexhibited (a) resonance an excellent and (b) broadnon-resonance frequency models. response. (FiguresTheThe receiving7receiving and 8) sensitivitiesThesensitivities thickness ofof theofthe the twotwo PVDF receiversreceivers film exhibitedexhibited was 0.002 ananλ ;excellent excellentthus, its broad broadresonance frequencyfrequency frequency response.response. was The receiving sensitivities of the two receivers exhibited an excellent broad frequency response. (Figuressignificantly(Figures7 7and and higher8) 8) The The than thickness thickn the frequencyess of of the the PVDFrange PVDF of film intere film was st.was 0.002Due 0.002 to λtheλ;; thus, lossythus, itsfactorits resonance resonance of the acoustic frequencyfrequency window, waswas (Figures 7 and 8) The thickness of the PVDF film was 0.002λ; thus, its resonance frequency was significantlythesignificantly receiver closehigher to than thanthe thetransmitter the frequency frequency (PVDF1) range range of had intere of a interest. lowerst. Due receiving to Due the tolossy thesensitivity factor lossy of factor than the acoustic that of the of the acousticwindow, outer significantly higher than the frequency range of interest. Due to the lossy factor of the acoustic window, window,receiverthe receiver the(PVDF2) receiverclose to(Figures closethe transmitter to 7 the and transmitter 8). (PVDF1) Overall, (PVDF1) had no a differencelower had a receiving lower was receiving observedsensitivity sensitivity between than that than the of that receivingthe of outer the the receiver close to the transmitter (PVDF1) had a lower receiving sensitivity than that of the outer outersensitivitiesreceiver receiver (PVDF2) of (PVDF2) the (Figuresresonance (Figures 7 andand7 and non-resonance 8).8 ).Overall, Overall, no nomo difference didels,fference except was was that observed observed the resonance between model thethe receivingpresentedreceiving receiver (PVDF2) (Figures 7 and 8). Overall, no difference was observed between the receiving sensitivitiesdramaticsensitivities changes ofof thethe in resonanceresonance the receiving andand non-resonance sensitivitynon-resonance around models,mo thedels, resona exceptexceptnce thatthat frequencies thethe resonanceresonance of the model modelstructure. presentedpresented sensitivities of the resonance and non-resonance models, except that the resonance model presented dramaticdramatic changeschanges inin thethe receivingreceiving sensitivity around the resona resonancence frequencies frequencies of of the the structure. structure. dramatic changes in the receiving sensitivity around the resonance frequencies of the structure.

Figure 7. Receiving sensitivities of the (a) PVDF1 and (b) PVDF2 resonance models. FigureFigure 7.7. ReceivingReceiving sensitivitiessensitivities ofof thethe ((aa)) PVDF1PVDF1 andand ((bb)) PVDF2 PVDF2 resonance resonance models. models. Figure 7. Receiving sensitivities of the (a) PVDF1 and (b) PVDF2 resonance models.

Figure 8. Receiving sensitivities of the (a) PVDF1 and (b) PVDF2 non-resonance models. Figure 8. Receiving sensitivities of the (a) PVDF1 and (b) PVDF2 non-resonance models. Figure 8. Receiving sensitivities of the (a) PVDF1 and (b) PVDF2 non-resonance models. The frequency response characteristics of both models were converted into a time transient model, The frequencyFigure 8. Receiving response sensitivities characteristics of the (ofa) PVDF1both models and (b) PVDF2were converted non-resonance into models. a time transient which was attached to a feedback controller, and time transient analysis was performed. The gain model,The which frequency was attached response to characteristicsa feedback controller, of both and models time transientwere converted analysis into was a performed. time transient The of theThe controller frequency was response pre-calculated characteristics based on theof both receiving models and were transmit converted sensitivities into a of time the modeltransient in gainmodel, of thewhich controller was attached was pre-calculated to a feedback based controller, on the andreceiving time transient and transmit analysis sensitivities was performed. of the model The themodel, frequency which response.was attached A tone-burst to a feedback sine controller, waveform and was time applied transient to both analysis models, was and performed. the reflected The ingain the of frequency the controller response. was pre-calculated A tone-burst sinebased wavefo on therm receiving was applied and transmit to both models, sensitivities and theof the reflected model soundgain of was the controller calculated. was As pre-calculated the system was based considered on the receiving linear, 1 and Pa ,transmit sensitivities 20 cycles of at thef0 sound model soundin the frequency was calculated. response. As theA tone-burst system was sine consi wavefoderedrm linear, was applied 1 Pa amplitude, to both models, 20 cycles and atthe 𝑓 reflected sound wavein the were frequency applied response. to the model. A tone-burst The typical sine wavefo responserm of was the applied reflected to sound both models, wave from and thethe modelreflected is wavesound were was appliedcalculated. to the As model. the system The typicalwas consi responsedered oflinear, the reflected 1 Pa amplitude, sound wave 20 cycles from atthe 𝑓 model sound is shownsound inwas Figure calculated.9. When As the the controller system was was consi switcheddered o linear,ff, 20% 1 of Pa the amplitude, sound was 20 reflected cycles at because 𝑓 sound of shownwave were in Figure applied 9. Whento the model.the controller The typical was switchedresponse off,of the 20% reflected of the sound sound was wave reflected from the because model ofis thewave lossy were acoustic applied windows to the model. used in The the typical model. response The reflected of the sound reflected level sound dramatically wave from reduced the withmodel the is theshown lossy in acousticFigure 9. windows When the used controller in the model.was switched The reflected off, 20% sound of the level sound dramatically was reflected reduced because with of actuationshown in ofFigure the transmitter. 9. When the In controller the continuous was switched wave (CW) off, condition,20% of the the sound reflected was reflected wave was because reduced of the lossyactuation acoustic of the windows transmitter. used inIn thethe model.continuo Theus reflectedwave (CW) sound condition, level dramatically the reflected reduced wave withwas tothe 0.7% lossy of acoustic the input windows pressure. used However, in the model. the reflected The reflected sound was sound approximately level dramatically 10% ( reduced20 dB) in with the reducedthe actuation to 0.7% of ofthe the transmitter. input pressure. In the Howeve continuor, theus reflected wave (CW) sound condition, was approximately the reflected− 10% wave (−20 wasdB) timethe actuation transient regionof the oftransmitter. the burst dueIn the to thecontinuo low fractionalus wave bandwidth(CW) condition, (FBW) the of thereflected system. wave Table was2 inreduced the time to 0.7%transient of the region input pressure.of the burst Howeve due tor, the reflectedlow fractional sound bandwidth was approximately (FBW) of 10% the (system.−20 dB) showsreduced the to reflection 0.7% of the control input simulation pressure. Howeve results forr, the the reflected resonance sound and non-resonancewas approximately model 10% in ( the−20 CW dB) Tablein the 2time shows transient the reflection region controlof the burst simulation due to re thesults low for fractional the resonance bandwidth and non-resonance (FBW) of the model system. in andin the transient time transient regions. region of the burst due to the low fractional bandwidth (FBW) of the system. theTable CW 2 showsand transient the reflection regions. control simulation results for the resonance and non-resonance model in Table 2 shows the reflection control simulation results for the resonance and non-resonance model in the CW and transient regions. the CW and transient regions.

SensorsSensors2020 2019, ,20 19,, 47 x FOR PEER REVIEW 8 ofof 1717 Sensors 2019, 19, x FOR PEER REVIEW 8 of 17

Figure 9. Simulations showing the cancellation results. (a) Resonance and (b) non-resonance model. FigureFigure 9. 9.Simulations Simulations showing showing the the cancellation cancellation results. results. ( a(a)) Resonance Resonance and and ( b(b)) non-resonance non-resonance model. model. TableTable 2.2. ReflectionReflection controlcontrol ofof thethe resonanceresonance andand non-resonancenon-resonance models. models. CW,CW, continuous continuous wave. wave. Table 2. Reflection control of the resonance and non-resonance models. CW, continuous wave. ResonanceResonance Non-Resonance Non-Resonance Model Model Reflection uncontrol (Pa) Resonance 0.15 Non-Resonance 0.2 Model Reflection uncontrol (Pa) 0.15 0.2 Reflection uncontrol (Pa) 0.15 0.2 ReflectionReflection control control (Pa): (Pa): transient transient region region 0.12 0.12 0.1 0.1 ReflectionE ff Effectect control of actuationof actuation (Pa): (dB)transient (dB) region 1.9− 0.121.9 − 0.166 − − ReflectionReflection Effect control control of (Pa):actuation (Pa): CW CW region (dB) region 0.03 0.03−1.9 0.007 0.007−6 ReflectionEffect of control actuation (Pa): (dB) CW region 14 0.03 0.00729 Effect of actuation (dB) − −14 −−29 Effect of actuation (dB) −14 −29 TheThe modelsmodels werewere evaluatedevaluated atat thethe ooff-targetff-target frequency.frequency. TheThe inputinput frequencyfrequency waswas changedchanged fromfrom The models were evaluated at the off-target frequency. The input frequency was changed from 0.80.8𝑓f0 to to1.2 1.2𝑓f0, and, and the the reflection reflection ratio ratio of of both both models models was was calculated. calculated. TheThe non-resonancenon-resonance modelmodel 0.8𝑓 1.2𝑓 exhibitedexhibited to aa better , and performance the reflection by by 4 4 dBratio dB as as comparedof compared both models with with the thewas resonance resonance calculated. model model The at frequenciesnon-resonance at frequencies below belowmodel 𝑓. exhibited a better performance by 4 dB as compared with the resonance model at frequencies below 𝑓 . fInterestingly,0. Interestingly, both both models models showed showed a asi similarmilar cancellation cancellation performance performance above above 𝑓f0.. Overall,Overall, thethe non-resonancenon-resonanceInterestingly, modelbothmodel models presentedpresented showed aa broaderbroader a si cancellationcancellationmilar cancellation responseresponse performance comparedcompared withwith above thethe resonance𝑓 . Overall, model. the non-resonanceAsAs shown shown in model in Figure Figure presented 10, the10, reflectionthe a broaderreflection ratio cancellation inratio the timein the response transient time transient regioncompared was region with as follows. thewas resonance as The follows. resonance model. The As shown in Figure 10, the reflection ratio in the time transient region was as follows. The modelresonance was attenuatedmodel was byattenuated18 dB at byf0 ,− and18 dB the at non-resonance 𝑓, and the non-resonance model was attenuated model was by attenuated20 dB. In the by resonance model was attenuated− by −18 dB at 𝑓 , and the non-resonance model was− attenuated by CW−20 region,dB. In the the CW resonance region, the model resonance was attenuated model was by attenuated30 dB at byf0, − and30 dB the at non-resonance 𝑓, and the non-resonance model was −20 dB. In the CW region, the resonance model was attenuated− by −30 dB at 𝑓 , and the non-resonance attenuatedmodel was by attenuated43 dB at byf0 .−43 dB at 𝑓. model was attenuated− by −43 dB at 𝑓.

FigureFigure 10.10. ReflectionReflection ratiosratios ofof thethe resonanceresonance andand non-resonancenon-resonance models.models. ((aa)) TransientTransient regionregion andand (b(Figureb)) CWCW region.region.10. Reflection ratios of the resonance and non-resonance models. (a) Transient region and (b) CW region. Figure 11 shows the effect of the controller. In the time transient region, the resonance model was Figure 11 shows the effect of the controller. In the time transient region, the resonance model attenuated by 2 dB at f0. The non-resonance model was attenuated by 6 dB at f0. In the CW region, was attenuatedFigure 11− showsby −2 dBthe at effect 𝑓. The of the non-resonance controller. In model the time was transient attenuated− region, by −6 the dB resonanceat 𝑓. In the model CW the resonance model was attenuated by 15 dB at f0, and the non-resonance model was attenuated by region,was attenuated the resonance by −2 dBmo atdel 𝑓 was. The attenuated non-resonance− by − 15model dB atwas 𝑓 ,attenuated and the non-resonance by −6 dB at 𝑓 model. In the wasCW 29 dB at f0. In the case of the designed controller, the results for the two models showed that the −attenuatedregion, the byresonance −29 dB atmo 𝑓del. In was the attenuatedcase of the by designed −15 dB controat 𝑓, ller,and thethe resultsnon-resonance for the two model models was effect was drastically reduced at the off-resonance frequency. However, the deviation from the peak in showedattenuated that by the −29 effectdB at was𝑓. Indrastically the case ofreduced the designed at the controoff-resonanceller, the resultsfrequency. for the However, two models the the non-resonance model was less attenuated than that in the resonance model. deviationshowed that from the the effect peak was in drasticallythe non-resonance reduced modelat the off-resonancewas less attenuated frequency. than However, that in thethe resonancedeviation model.from the peak in the non-resonance model was less attenuated than that in the resonance model.

Sensors 2020, 20, 47 9 of 17 Sensors 2019, 19, x FOR PEER REVIEW 9 of 17

FigureFigure 11.11.E Effectffect of of the the controller controller of of the the resonance resonance and and non-resonance non-resonance models. models. (a) Transient(a) Transient region region and (andb) CW (b) region. CW region. 2.5. Fabrication 2.5. Fabrication Our aim was to develop resonance and non-resonance models suitable for operation in the low Our aim was to develop resonance and non-resonance models suitable for operation in the frequency range. low frequency range. The resonance and non-resonance models required a thick PZT layer in order to be suitable for The resonance and non-resonance models required a thick PZT layer in order to be suitable for the the low frequency range. PZT is a piezoelectric material; hence, it should be polled to vibrate due to low frequency range. PZT is a piezoelectric material; hence, it should be polled to vibrate due to the the voltage difference. However, it is not feasible to obtain PZT with a polling of tens of centimeters. voltage difference. However, it is not feasible to obtain PZT with a polling of tens of centimeters. Therefore, we used a thick PZT layer wherever possible and attached a backing material to the rear Therefore, we used a thick PZT layer wherever possible and attached a backing material to the rear side side of the PZT to increase the output performance [23]. Thus, we present a model that achieved of the PZT to increase the output performance [23]. Thus, we present a model that achieved the target the target frequency. The resonance and non-resonance models were composed of ten layers of frequency. The resonance and non-resonance models were composed of ten layers of commercially commercially available PZT-5A, excluding the low frequency transducers. As shown in Figure 12a, low available PZT-5A, excluding the low frequency transducers. As shown in Figure 12a, low frequency frequency piezoelectric material based transducers were fabricated. Accordingly, 2 mm thick PZT-5A piezoelectric material based transducers were fabricated. Accordingly, 2 mm thick PZT-5A (T180-A4N0-2929, Piezo.com, Woburn, MA, USA) was stacked to implement the transducer for low (T180-A4N0-2929, Piezo.com, Woburn, MA, USA) was stacked to implement the transducer for low frequency operation. As the resonance and non-resonance models would be used underwater, instead frequency operation. As the resonance and non-resonance models would be used underwater, instead of PZT-5H, which has excellent output performance, PZT-5A was employed due to its insensitivity to of PZT-5H, which has excellent output performance, PZT-5A was employed due to its insensitivity to temperature changes. When stacking PZT-5A, EPOTEK-301 (EPOTEK 301, Epoxy technology, Billerica, temperature changes. When stacking PZT-5A, EPOTEK-301 (EPOTEK 301, Epoxy technology, Billerica, MA, USA), which has a low viscosity and excellent adhesion, was used (Figure 12a). EPOTEK-301 MA, USA), which has a low viscosity and excellent adhesion, was used (Figure 12a). EPOTEK-301 consists of Parts a and b, which were mixed using a weight ratio of 4:1 and stirred for 5 min. We used consists of Parts a and b, which were mixed using a weight ratio of 4:1 and stirred for 5 min. We used Parts a and b after they were sufficiently shaken. Subsequently, the bubbles formed were removed Parts a and b after they were sufficiently shaken. Subsequently, the bubbles formed were removed using a vacuum pump. The interior of the vacuum pump was maintained at 100 kPa for 15 min. The using a vacuum pump. The interior of the vacuum pump was maintained at 100 kPa for 15 min. The degassed EPOTEK-301 was thinly applied on top of PZT-5A. If a large amount of EPOTEK-301 were degassed EPOTEK-301 was thinly applied on top of PZT-5A. If a large amount of EPOTEK-301 were applied, it might be difficult to establish a connection between the piezoelectric elements due to the applied, it might be difficult to establish a connection between the piezoelectric elements due to the thickness of the epoxy. Subsequently, the epoxy spilled on the PZT side was wiped off using a wiper, thickness of the epoxy. Subsequently, the epoxy spilled on the PZT side was wiped off using a wiper, as the vibration mode of the PZT may change if the epoxy flowing down the side of the PZT is not as the vibration mode of the PZT may change if the epoxy flowing down the side of the PZT is not removed. Each corner of PZT-5A was adjusted using a pair of tweezers and a slide 30 min after removed. Each corner of PZT-5A was adjusted using a pair of tweezers and a slide glass 30 min after the application of EPOTEK-301. the application of EPOTEK-301. If the cured PZT was not subjected to any slipping caused by the epoxy, a weight of 100 g was If the cured PZT was not subjected to any slipping caused by the epoxy, a weight of 100 g was added and then cured at 65 C for 2 h. Whenever the PZT was stacked, the same operation described added and then cured at 65◦ °C for 2 h. Whenever the PZT was stacked, the same operation described above was performed, and the electrical impedance was measured. We verified whether the resonance above was performed, and the electrical impedance was measured. We verified whether the frequency of the device decreased when the devices were stacked. Furthermore, when laminating resonance frequency of the device decreased when the devices were stacked. Furthermore, when the PZT-5A, it was ensured that each side faced upward. PZT was stacked in one direction using laminating the PZT-5A, it was ensured that each side faced upward. PZT was stacked in one direction EPOTEK-301. The sides of the device were wrapped with Kapton tape when the PZT-5A was stacked using EPOTEK-301. The sides of the device were wrapped with Kapton tape when the PZT-5A was up to ten layers. stacked up to ten layers. Ten layers were constructed with PZT, and subsequently, a mold was built. A 3 mm thick acrylic Ten layers were constructed with PZT, and subsequently, a mold was built. A 3 mm thick acrylic plate was incorporated on each side of the mold using a cutter (Figure 12b). When each side plate was incorporated on each side of the mold using a laser cutter (Figure 12b). When each side was was completed, it was assembled to complete the mold. The appearance of the finished mold was completed, it was assembled to complete the mold. The appearance of the finished mold was fixed fixed with Kapton tape. The prefabricated frame can be conveniently removed later. When the mold with Kapton tape. The prefabricated frame can be conveniently removed later. When the mold was was developed as one piece, as opposed to assembling several parts, it was prefabricated because developed as one piece, as opposed to assembling several parts, it was prefabricated because the the sensor was damaged when the mold was removed. In this study, we present a horizontal mold, sensor was damaged when the mold was removed. In this study, we present a horizontal mold, because when the layers were stacked vertically, the acoustic window hardened and caused shrinkage, because when the layers were stacked vertically, the acoustic window hardened and caused

Sensors 2019, 19, x FOR PEER REVIEW 11 of 17

SensorsOnce2020 the, 20 ,upper 47 part of the piezoelectric based resonance model was constructed, the ground10 of 17 was connected to the bottom of the stacked piezoelectric element via the conductive epoxy. Subsequently, the section connecting the acoustic window and electric wire was coated with epoxy resulting in a variable height. After the mold was completed, its interior was cleaned with an acetone resin for 5 min. This process was repeated two more times. Finally, we attached steel to the rear end swab to remove the powder from laser cutting (Figure 12c). of the PZT-5A stacked device using EPOTEK 301 to fabricate the resonance model further (Figure 12e).

FigureFigure 12. 12. FabricationFabrication process: process: (a) (Stacka) Stack PZT, PZT, (b) ( bacrylic) acrylic mold, mold, (c) (cmold) mold assembly assembly and and taping, taping, (d()d transmit) transmit and and receiving receiving sensor sensor connection, connection, and and (e ()e curing) curing acoustic acoustic window. window.

Subsequently, PZT and PVDF were placed inside the mold, and the wires were connected One hour after fixing, a weight of 200 g was added. The of the resonance model was (Figure 12d). The transmitting sensor (PZT-5A) was placed at the bottom of the mold and was fixed connected to the steel using a conductive epoxy, as soldering was not effective due to the lubricant using double sided tape. The receiving sensors (PVDF) were designed to fit on both sides of the mold, used in the steel processing. When the conductive epoxy was cured, the epoxy resin was applied to requiring no additional attachments. The electrode parts of the transmitter and receiver were hardened the conductive part for 5 min. This was necessary to fix and coat the ground. The resonance and over one day at room temperature after connecting the wires with the conductive epoxy. Then, the two non-resonance models were fabricated using the aforementioned processes. receiving sensors (PVDF) were grounded. The ground of the stacked sensor PZT-5A, the transmitter, 2.6.was Experimental connected Setup after the mold was completely cured. Conductive epoxy consisting of Parts A and B should be mixed for 1 min in a ratio of 1:1. It is recommended that the conductive epoxy be mixed on theWe box measured tape, as it the penetrates transmission the paper. characteristics Hot plates of canthe beresonance used to cureand non-resonance the conductive models epoxy faster;in a snakehowever, shaped this tank method (Figure is not 13), recommended. which was composed The hardening of 10 mm on thick the hot acrylic plate and showed had a that total the length receiving of 5.8sensor m. When was measuring sensitive to the heat, transmission resulting incharacteristic a deformeds of appearance. a device in Therefore,a tank of length the method 1 m, reflection of curing occurredat room at temperature the end surface was selected. of the tank, In this thus case, maki eachng wire it difficult used a coaxialto determine cable, andthe thetransmission wires were characteristicsattached to the accurately. edge of the Therefore, sensor to the ensure measurements that the sound were waves performed of the sensorin a tank propagated of a length without of 5.8interference. m to minimize The coaxialthe reflection cable was from thicker the end than su arface standard of the wire tank. and As was shown fixed byin bendingFigure 13, the the wire experimentwith a pair was of tweezers.conducted Thisby filling ensured the thattank itwi didth water. not float The from resonance the sensor and andnon-resonance was attached models to the wereconductive located at epoxy. the center We used of the a largetank and amount excited of conductiveat 30 Vpp with epoxy ten and cycle attempted sine waves to obtained fix the wire using only a functionbased on generator the weight and of amplifier. the conductive A low epoxy. frequenc However,y hydrophone this method was placed did not at succeed.a distance Therefore, of 600 mm the fromwire the height resonance was adjusted and non-resonance accurately before models. applying Low frequency the conductive hydrophones epoxy. After were the used conductive to measure epoxy thewas desired cured, afrequency multimeter range was used to toevaluate check whether the ou thetput electrodes performances were connected. of the resonance and non-resonanceThe acoustic models. window (Rho-c) existing between the transmitter and the receiver was prepared. The acoustic window (Aptflex-F21, Precision Acoustics, London, UK) consisted of Parts A and B, which had to be shaken sufficiently before they could be used. Even if the weight ratio was adjusted, these parts did not perform as well as the original material of the acoustic window if not sufficiently shaken, and the acoustic window was gray without being black. Parts A and B were placed in a disposable plastic cup, with the weight ratio set to 3.35:1, and stirred for 5 min using wooden chopsticks. This stirring generated a chemical reaction that produced heat. Even if the parts were mixed for 5 min,

Sensors 2020, 20, 47 11 of 17 the mixing did not generate heat, and the mixture was not cured even after the curing time, if the quantity was insufficient. Therefore, it is recommended that the mixture be stirred as fast as it is mixed. Subsequently, the bubbles generated while mixing were discharged with a vacuum pump outward. At this time, the vacuum pump was maintained at 100 kPa for 15 min. While the bubbles escaped, the interior of the mold was cleaned with acetone, using a cotton swab. The interior of the mold was wiped a few times. Acetone acted as a mold release. The acoustic window had to be prepared in less than 20 min, because after 20 min, it would harden inside the plastic container and become unusable. The acoustic window was then slowly poured into the mold. If the acoustic window exceeded the height of the mold, its height could be adjusted using a slide glass. The height was immediately flattened/reduced after pouring the acoustic window into the mold. However, if this process was delayed by a few minutes, the acoustic window would stick to the slide glass and exit the mold. Following the planarization operation, the acoustic window inside the mold exhibited several large and small bubbles even after bubble removal. The bubbles were popped using a toothpick. Once popped, bubbles were still generated, which had to be popped several times. A bubble in the acoustic window was popped using a toothpick, and when the acoustic window was completely attached to the toothpick, it was rolled up into a circular shape using the toothpick. The flattening process was then repeated. The acoustic window was left to cure for 1 to 2 days at room temperature, without covering the mold. Once the acoustic windows had cured, the mold was removed using a knife. Considerable care was exercised during this process to avoid damaging the sensor. Care was also taken when removing the lowest part of the horizontal mold. The mold was carefully removed, particularly where the PVDF was inserted. The mold in the PVDF region had to be removed faster than in the other parts to avoid splitting the PVDF into two (Figure 12e). Once the upper part of the piezoelectric based resonance model was constructed, the ground was connected to the bottom of the stacked piezoelectric element via the conductive epoxy. Subsequently, the section connecting the acoustic window and electric wire was coated with epoxy resin for 5 min. This process was repeated two more times. Finally, we attached steel to the rear end of the PZT-5A stacked device using EPOTEK 301 to fabricate the resonance model further (Figure 12e). One hour after fixing, a weight of 200 g was added. The ground of the resonance model was connected to the steel using a conductive epoxy, as soldering was not effective due to the lubricant used in the steel processing. When the conductive epoxy was cured, the epoxy resin was applied to the conductive part for 5 min. This was necessary to fix and coat the ground. The resonance and non-resonance models were fabricated using the aforementioned processes.

2.6. Experimental Setup We measured the transmission characteristics of the resonance and non-resonance models in a snake shaped tank (Figure 13), which was composed of 10 mm thick acrylic and had a total length of 5.8 m. When measuring the transmission characteristics of a device in a tank of length 1 m, reflection occurred at the end surface of the tank, thus making it difficult to determine the transmission characteristics accurately. Therefore, the measurements were performed in a tank of a length of 5.8 m to minimize the reflection from the end surface of the tank. As shown in Figure 13, the experiment was conducted by filling the tank with water. The resonance and non-resonance models were located at the center of the tank and excited at 30 Vpp with ten cycle sine waves obtained using a function generator and amplifier. A low frequency hydrophone was placed at a distance of 600 mm from the resonance and non-resonance models. Low frequency hydrophones were used to measure the desired frequency range to evaluate the output performances of the resonance and non-resonance models. Sensors 2020, 20, 47 12 of 17 Sensors 2019, 19, x FOR PEER REVIEW 12 of 17

Figure 13. Evaluation of the transmission characteristics of the resonance and non-resonance models.

The experimentalexperimental setupsetup usedused for for the the cancellation cancellation experiment experiment is is shown shown in in Figure Figure 14 .14. As As shown shown in Figurein Figure 14a, 14a, a rubber a rubber tube tube with with inner inner and and outer oute diametersr diameters of 35 of mm 35 and mm 45 and mm, 45respectively, mm, respectively, and a totaland lengtha total oflength 4.8 m of was 4.8 usedm was as aused water as tank.a water The tank. left sideThe ofleft the side rubber of the tube rubber was moldedtube was with molded glass with and theglass right and side the withright rubber.side with The rubber. hydrophone The hydrophone was located was at alocated distance at ofa distance 1 m from of the 1 m right from end the of right the Sensorsrubberend of 2019 the tube., 19rubber, Stackedx FOR tube. PEER PZT REVIEWStacked based devicesPZT based for transmissiondevices for transmission were located were at both located ends at of both the rubber ends13 of tube.of the 17 rubber tube. As shown in Figure 14a,b, the resonance and non-resonance models were evaluated. As shown in Figure 14c (Step 1), a function generator and amplifier were used to activate the device at the right end section. At 70 Vpp, a with 30 cycles was applied to the right end of the device, and the signal reflected from the glass located on the left side of the rubber tube was measured using a hydrophone. After measuring the signal from the right side, the function generator connected to the right sensor was switched off. In Step 2 (Figure 14c), a function generator and amplifier were used to activate the resonance and non-resonance models located at the left end section. At 70 Vpp, a sine wave with 30 cycles was applied to the non-resonance model. The signal obtained at this time was measured using a hydrophone. After verifying the sensor measurement results on the left and right sides, the amplitude and delay time to be applied to the resonance and non-resonance models were determined. In Step 3 (Figure 14c), the amplitude and delay time were modified and measured again. This process was repeatedly performed, as the appropriate amplitude and delay time would increase the cancellation effect. Subsequently, two function generators and amplifiers were connected to each device to resonate with the two devices simultaneously. Finally, we verified whether the desired area

had been cancelled. Figure 14. ExperimentalExperimental setup setup showing showing the the cancellation cancellation in in the the resonance resonance and and non-resonance non-resonance models. models. (a) experiment setup of resonance model ((b)) non-resonancenon-resonance modelmodel ((cc)) experimentexperiment step.step. As shown in Figure 14a,b, the resonance and non-resonance models were evaluated. As shown in 3. Results Figure 14c (Step 1), a function generator and amplifier were used to activate the device at the right 3.1.end Experimental section. At 70Results Vpp, a sine wave with 30 cycles was applied to the right end of the device, and the signal reflected from the glass located on the left side of the rubber tube was measured using a hydrophone.The resonance After measuringand non-resonance the signal models from the were right fabricated side, the with function the generatorsame size connectedand device. to The the resonanceright sensor model was switched shown in off Figure. In Step 15 2 ha (Figured a width, 14c), alength, function and generator height andof 0.36 amplifier𝜆, 0.36 were𝜆, and used 1.8𝜆 to, respectively.activate theresonance The non-resonance and non-resonance model shown models in Figu locatedre 15b at had the a leftwidth, end length, section. and At height 70 Vpp, of a0.36 sine𝜆, 0.36wave𝜆, withand 1.2 30𝜆 cycles, respectively. was applied The wavelength to the non-resonance was set based model. on the The specific signal frequency obtained atand this the time velocity was ofmeasured sound in using water. a hydrophone. After verifying the sensor measurement results on the left and right sides, the amplitude and delay time to be applied to the resonance and non-resonance models were

Figure 15. Fabricated (a) resonance and (b) non-resonance models.

The input impedance was measured after the fabrication of the resonance and non-resonance models. The impedance measurements were performed in air, and the measurement frequency range was 0.3𝑓 to 3.3𝑓 . In Figure 16a,b, the resonance frequency can be observed at 0.97𝑓 and 2.17𝑓, respectively.

Figure 16. Input impedance of the (a) resonance and (b) non-resonance models.

SensorsSensors 20192019,, 1919,, xx FORFOR PEERPEER REVIEWREVIEW 1313 ofof 1717

Sensors 2020, 20, 47 13 of 17 determined. In Step 3 (Figure 14c), the amplitude and delay time were modified and measured again. This process was repeatedly performed, as the appropriate amplitude and delay time would increase the cancellation effect. Subsequently, two function generators and amplifiers were connected to each deviceFigureFigure to resonate 14.14. ExperimentalExperimental with the twosetupsetup devices showingshowing simultaneously. thethe cancellationcancellation in Finally,in thethe resonanceresonance we verified andand non-resonancenon-resonance whether thedesired models.models. area had been((aa)) experimentexperiment cancelled. setupsetup ofof resonanceresonance modelmodel ((bb)) non-resonancenon-resonance modelmodel ((cc)) experimentexperiment step.step.

3.3.3. ResultsResultsResults

3.1.3.1.3.1. ExperimentalExperimentalExperimental ResultsResultsResults TheTheThe resonanceresonance and and non-resonancenon-resonance non-resonance modelsmodels models werewere were fabricatedfabricated fabricated withwith with thethe the samesame same sizesize size andand and device.device. device. TheThe Theresonanceresonance resonance modelmodel model shownshown shown inin Figure inFigure Figure 1515 15haha dhadd aa width, awidth, width, length,length, length, andand and heightheight height ofof of 0.360.36 0.36𝜆𝜆λ,, , 0.360.36 0.36𝜆𝜆λ,,, and andand 1.81.8𝜆1.8𝜆λ,,, respectively.respectively.respectively. The TheThe non-resonance non-resonancenon-resonance model modelmodel shown shownshown in inin Figure FiguFigurere 15 15b15bb had hadhad a aa width, width,width, length, length,length, and andand height heightheight of ofof 0.36 0.360.36λ𝜆𝜆,,, 0.360.360.36λ𝜆𝜆,,, and andand 1.2 1.21.2λ𝜆𝜆,,, respectively.respectively. The TheThe wavelength wavelengthwavelength was waswas set setset based basedbased on onon the thethe specificspecificspecific frequency frequencyfrequency and andand the thethe velocity velocityvelocity ofofof sound soundsound in inin water. water.water.

FigureFigureFigure 15. 15.15. Fabricated FabricatedFabricated ( (a(aa))) resonance resonanceresonance and andand ( (b(bb))) non-resonance non-resonancenon-resonance models. models.models.

TheTheThe inputinputinput impedanceimpedanceimpedance waswaswas measuredmeasuredmeasured afterafterafter thethethe fabrication fabrfabricationication ofofof the thethe resonance resonanceresonance and andand non-resonance non-resonancenon-resonance models.models.models. TheThe The impedanceimpedance impedance measurementsmeasurements measurements werewere were performeperforme performeddd inin air,air, in and air,and andthethe measurementmeasurement the measurement frequencyfrequency frequency rangerange rangewaswas 0.3𝑓0.3𝑓 was 0.3 toto f 03.3𝑓3.3𝑓to 3.3.. fInIn0. Figure InFigure Figure 16a,b,16a,b, 16a,b, thethe the resonanceresonance resonance frequencyfrequency frequency cancan can bebe be observedobserved observed atat at 0.97𝑓0.970.97𝑓f0 and andand 2.172.17𝑓2.17𝑓f0,,, respectively. respectively.respectively.

FigureFigureFigure 16. 16.16. Input InputInput impedance impedanceimpedance of ofof the thethe ( (a(aa))) resonance resonanceresonance and andand ( (b(bb))) non-resonance non-resonancenon-resonance models. models.models.

The transmit characteristics of the resonance and non-resonance models were measured inside the snake shaped tank containing water. The measurement frequency range was 0.3 f0 to 3.3 f0. As shown in Figure 17a, the resonance model exhibited a performance of 699 Pa/V at 1.13 f0. As shown in Figure 17b, the non-resonance model exhibited a performance of 452 Pa/V at 2.1 f0. The resonance model demonstrated transmit characteristics that were 1.55 higher than those of the non-resonance model. Sensors 2019, 19, x FOR PEER REVIEW 14 of 17

The transmit characteristics of the resonance and non-resonance models were measured inside the snake shaped tank containing water. The measurement frequency range was 0.3𝑓 to 3.3𝑓. As shown in Figure 17a, the resonance model exhibited a performance of 699 Pa/V at 1.13𝑓 . As shown in Figure 17b, the non-resonance model exhibited a performance of 452 Pa/V at 2.1𝑓. The resonance model demonstrated transmit characteristics that were 1.55 higher than those of the non-resonance model. Sensors 2019, 19, x FOR PEER REVIEW 14 of 17 Sensors 2019, 19, x FOR PEER REVIEW 14 of 17 The transmit characteristics of the resonance and non-resonance models were measured inside the The transmit characteristics of the resonance and non-resonance models were measured inside the snake shaped tank containing water. The measurement frequency range was 0.3𝑓 to 3.3𝑓 . As shown snake shaped tank containing water. The measurement frequency range was 0.3𝑓 to 3.3𝑓. As shown in Figure 17a, the resonance model exhibited a performance of 699 Pa/V at 1.13𝑓 . As shown in in Figure 17a, the resonance model exhibited a performance of 699 Pa/V at 1.13𝑓. As shown in Figure 17b, the non-resonance model exhibited a performance of 452 Pa/V at 2.1𝑓 . The resonance Figure 17b, the non-resonance model exhibited a performance of 452 Pa/V at 2.1𝑓. The resonance model demonstrated transmit characteristics that were 1.55 higher than those of the model demonstrated transmit characteristics that were 1.55 higher than those of the Sensorsnon-resonance2020, 20, 47 model. 14 of 17 non-resonance model.

Figure 17. Transmit sensitivities of the (a) resonance and (b) non-resonance models.

Figure 18 shows the signal measured by the receiving sensors of the hydrophone and the resonance and non-resonance models when the external sound waves were transmitted to the transducer. When the external transducer was excited at 30 Vpp, the signal was received by the hydrophone, PVDF1,Figure and 17. TransmitPVDF2 sensitivities inserted of inside the (a) resonance the resonance and (b) non-resonance and non-resonance models. models. The FigureFigure 17.17. TransmitTransmit sensitivitiessensitivities ofof thethe ((aa)) resonanceresonance and and ( b(b)) non-resonance non-resonance models. models. signal received at the hydrophone, as shown in Figure 18a, was used to verify our resonance and Figure 18 shows the signal measured by the receiving sensors of the hydrophone and the non-resonanceFigureFigure models. 18 18 shows shows Figure the the signal signal 18b,c measured measured shows by the aby receivingscan the receivingof sensors the receivedsensors of the hydrophone of signalsthe hydrophone and of thethe resonance andresonance the and resonance and non-resonance models when the external sound waves were transmitted to the andresonance non-resonance and non-resonance models when models the external when sound the external waves were sound transmitted waves were to the transmitted transducer. Whento the non-resonancetransducer. models. When Figurethe external 18c showstransducer that was the exci signalsted at 30of Vpp,the receiving the signal sewasnsors received differed by the by 1/4 of thetransducer. external transducerWhen the external was excited transducer at 30 Vpp, was the exci signalted at was 30 receivedVpp, the by signal the hydrophone, was received PVDF1, by the the wavelength.hydrophone, However, PVDF1, andin thePVDF2 case inserted of the inside reso nancethe resonance model, and it wasnon-resonance difficult models.to observe The the 1/4 andhydrophone, PVDF2 inserted PVDF1, inside and PVDF2 the resonance inserted and inside non-resonance the resonance models. and non-resonance The signal received models. at theThe signal received at the hydrophone, as shown in Figure 18a, was used to verify our resonance and wavelengthhydrophone,signal difference received as shownat in the the hydrophone, in receiving Figure 18a, assensor was shown used signals in to Figure verify 18a, our resonancewas used to and verify non-resonance our resonance models. and non-resonance models. Figure 18b,c shows a scan of the received signals of the resonance and Figurenon-resonance 18b,c shows models. a scan ofFigure the received 18b,c shows signals a of scan the resonance of the received and non-resonance signals of the models. resonance Figure 18andc non-resonance models. Figure 18c shows that the signals of the receiving sensors differed by 1/4 of showsnon-resonance that the signalsmodels. of Figure the receiving 18c shows sensors that the diff sieredgnals by of 1 /the4 of receiving the wavelength. sensors differed However, by in1/4 the of the wavelength. However, in the case of the resonance model, it was difficult to observe the 1/4 casethe wavelength. of the resonance However, model, in it wasthe dicasefficult of the toobserve resonance the model, 1/4 wavelength it was difficult difference to inobserve the receiving the 1/4 wavelength difference in the receiving sensor signals sensorwavelength signals difference in the receiving sensor signals

Figure 18. Received signal from the (a) hydrophone, (b) PVDF1 and PVDF2 resonance models, and

(c) PVDF1 and PVDF2 non-resonance models. FigureFigure 18.18. ReceivedReceived signal signal from from the the (a ()a hydrophone,) hydrophone, (b) ( bPVDF1) PVDF1 and and PVDF2 PVDF2 resonance resonance models, models, and Figure 18. Received signal from the (a) hydrophone, (b) PVDF1 and PVDF2 resonance models, and and(c) PVDF1 (c) PVDF1 and and PVDF2 PVDF2 non-resonance non-resonance models. models. We conducted(c) PVDF1 cancellation and PVDF2 non-resonance experiments models. (Figure 19) on the resonance and non-resonance models. Table 3 showsWeWe conductedthe conducted reflection cancellation cancellation control experiments expe resultsriments for (Figure (Figure the 19 resonance19)) on on thethe resonanceresonance and non-resonance andand non-resonancenon-resonance model models.models. in the CW We conducted cancellation experiments (Figure 19) on the resonance and non-resonance models. TableTable3 shows3 shows the the reflection reflection control control results results for thefor resonancethe resonance and and non-resonance non-resonance model model in the in CW the and CW and transientTable 3regions. shows the reflection control results for the resonance and non-resonance model in the CW transientand transient regions. regions. and transient regions.

Figure 19. Cancellation. Figure 19. Cancellation. FigureFigure 19. Cancellation.Cancellation.

Sensors 2020, 20, 47 15 of 17

Table 3. Reflection control of the resonance and non-resonance models.

Resonance Non-Resonance Model Reflection uncontrol (mV) 152 116 Reflection control (mV): transient region 144 76 Effect of actuation (dB) 0.5 4 − − Reflection control (mV): CW region 40 16 Effect of actuation (dB) 12 17 − −

4. Discussion This paper aimed to present the advantages and disadvantages of acoustic absorbers having the same parameters when operating in resonance or non-resonance modes. As described above, an absorber in resonance mode had a high transmit sensitivity relative to its resonance frequency. Thus, when a high output sound was incident, high cancellation may be demonstrated. When low pressure was applied, such as SONAR sound waves transmitted from a long distance, the non-resonance type showed excellent overall performance. In the case of the resonance type, the degrading performance was negligible except over the range of a small proportion of the resonance frequency due to the low FBW. The non-resonance type had a broader frequency response, except that it had a limited transmit sensitivity. Even in the non-resonance type, the FBW at the resonance frequency showed a narrow characteristic. The operating frequency was off-the resonance, which generally indicates a wideband output characteristic. Two sensor approaches to separating input and output are miniaturized methods widely used for acoustic characterization. Although they are mathematically concise and the design of the controller is easy, they have several disadvantages. First, in the two-sensor design, a 1/4 wavelength distance between sensors is required; thus, low frequency structure applications have thick layers. Furthermore, despite the wide FBW shown in Figure 17, this design only worked well around the designed target frequency. This paper presented a relatively simple control method for comparing the two types of absorbers. In addition, a modern cancellation method, such as the FxLMS (filtered – x LMS) algorithm, which has high accuracy, is expected to demonstrate excellent cancellation performance. Thus, it can be assumed that the non-resonance type will show excellent performance. We manufactured the resonance and non-resonance models inside our laboratory, without outsourcing to other companies. Therefore, even when the piezoelectric elements were stacked, the alignment was not even. Thus, it appeared that the piezoelectric element was affected during vibration.

5. Conclusions In this study, we analyzed and fabricated two different types of piezoelectric absorbers. The absorbers consisted of two receiving sensors composed of PVDF films and one transmitter made of PZT-5A. Both models were analytically analyzed in the frequency and time domains. With the conventional PZT and PVDF structures, the absorbers achieved a reflection ratio of greater than 17 dB in the CW condition and 4 dB in the time transient region. The resonance absorbers had − − high sensitivities around the resonance, albeit at the expense of the bandwidth. The non-resonance absorber exhibited a better frequency response; however, it required a high transmit voltage. The reflected sound wave was mainly observed in the time transient region when using a conventional feedback controller. Consequently, the resonance structure had a higher reflection ratio due to the narrow frequency response. This work showed that the fractional bandwidth of the system had an important role in feedback based piezoelectric absorbers.

Author Contributions: Conceptualization and supervision, K.K.P.; simulation, Y.H.K.; fabrication, J.Y.P., S.W.K., and W.Y.C.; experiment work, J.Y.P., Y.H.K., and W.Y.C.; data analysis, J.Y.P., Y.H.K., and S.W.K.; validation, J.Y.P., Y.H.K., and W.Y.C.; writing and original draft preparation, J.Y.P.; writing and review and editing, all authors. All authors have read and agreed to the published version of the manuscript. Sensors 2020, 20, 47 16 of 17

Funding: This work was supported by the Defense Acquisition Program Administration and Agency for Defense Development in Korea under the contract UD170023DD. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Yildiz, F.; Zhu, J.; Pecen, R.; Guo, L. Energy scavenging for sensor nodes with a focus on rotation to electricity conversion. ASEE Annu. Conf. Expo. Conf. Proc. 2007, 2153, 12.613.1–12.613.13. 2. Qiu, Z.; Fang, H.; O’Leary, R.; Gachagan, A.; Moldovan, A. Broadband Piezocrystal Transducer Array for Non-Destructive Evaluation Imaging Applications. In Proceedings of the 2018 IEEE International Ultrasonics Symposium (IUS), Kobe, Japan, 22–25 October 2018; pp. 1–4. 3. Gomes, R.; Feitosa, F.R.; Souto, C.; Lima, B.A.; Junior, J.A.; Cunha, M.; Sobrinho, J.M.B.; Dubois, J.-M. Crack Detection in High-Velocity Oxygen-Fuel-Sprayed Al59.2Cu25.5Fe12.3B3 Quasicrystalline Coatings Using Piezoelectric Active Sensors. J. Mater. Eng. Perform. 2019, 28, 5649–5660. [CrossRef] 4. Ke, Q.; Liew, W.H.; Tao, H.; Wu, J.; Yao, K. KNNS-BNZH Lead-Free 1–3 Piezoelectric Composite for Ultrasonic and Photoacoustic Imaging. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 2019, 66, 1395–1401. [CrossRef] [PubMed] 5. Meng, H.; Wen, J.; Zhao, H.; Wen, X. Optimization of locally resonant acoustic on underwater sound absorption characteristics. J. Sound Vib. 2012, 331, 4406–4416. [CrossRef] 6. Varadan, V.K.; Varadan, V.V.; Su, J.H.; Pillai, T.A.K. Comparison of sound scattering by rigid and elastic obstacles in water. J. Acoust. Soc. Am. 1982, 71, 1377–1383. [CrossRef] 7. Gentilman, R.L.; Fiore, D.; Pham-Nguyen, H.; Serwatka, W.J.; Pazol, B.G.; Near, C.D.; McGuire, P.T.; Bowen, L.J. 1–3 Piezocomposite Smart Panels for Active Surface Control. Smart Struct. Mater. 1996 Ind. Commer. Appl. Smart Struct. Technol. 1996, 2721, 234–239. 8. Corsaro, R.D.; Houston, B.; Bucaro, J.A. Sensor—actuator tile for underwater surface impedance control studies. J. Acoust. Soc. Am. 1997, 102, 1573–1581. [CrossRef] 9. Zhang, J.; Zhang, W.; Varadan, V.K.; Varadan, V.V. Active underwater acoustic cancellation with pressure sensor and piezocomposite transducer. J. Acoust. Soc. Am. 2000, 108, 2638. [CrossRef] 10. Wu, Z.; Bao, X.Q.; Varadan, V.K.; Varadan, V.V. Broadband active acoustic absorbing coating with an adaptive digital controller. Smart Mater. Struct. 1993, 2, 40–46. [CrossRef] 11. Shields, F.D.; Hendrix, J.E. Acoustically active surfaces using piezorubber. J. Acoust. Soc. Am. 1991, 90, 1230–1237. 12. Howarth, T.R.; Varadan, V.K.; Bao, X.; Varadan, V.V. Piezocomposite coating for active underwater sound reduction. J. Acoust. Soc. Am. 2005, 91, 823–831. [CrossRef] 13. Chang, W.; Kwon, D.; Park, J.; Jeon, B. An active underwater acoustic reflection control system. In Proceedings of the 2008 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems, Seoul, Korea, 20–22 August 2008; pp. 583–587. 14. Nieto, E.; Fernandez, J.F.; Moure, C.; Duran, P. Multilayer piezoelectric devices based on PZT. J. Mater. Sci. Mater. . 1996, 7, 55–60. [CrossRef] 15. Goldfarb, M.; Celanovic, N. Modeling Piezoelectric Stack Actuators for Control of Micromanipulation. IEEE Control. Syst. 1997, 17, 69–79. 16. Yoon, C.-B.; Koh, Y.-H.; Park, G.-T.; Kim, H.-E. Multilayer Actuator Composed of PZN-PZT and PZN-PZT/Ag Fabricated by Co-Extrusion Process. J. Am. Ceram. Soc. 2005, 88, 1625–1627. [CrossRef] 17. Sahoo, B.; Jaleel, V.A.; Panda, P.K. Development of PZT powders by wet chemical method and fabrication of multilayered stacks/actuators. Mater. Sci. Eng. B Solid-State Mater. Adv. Technol. 2006, 126, 80–85. [CrossRef] 18. Ahmad, K.A.; Osman, M.K.; Hussain, Z.; Abdullah, M.F.; Manaf, A.A.; Abdullah, N. Design and characterization piezoelectric acoustic transducer for sonar application. In Proceedings of the 2018-8th IEEE International Conference on Control System Computing Engineering ICCSCE, Penang, Malaysia, 23–25 November 2019; pp. 233–237. 19. Yao, K.; Zhu, W. Improved preparation procedure and properties for a multilayer piezoelectric thick-film actuator. Sens. Actuators A Phys. 1998, 71, 139–143. [CrossRef] 20. Krimholtz, R.; Leedom, D.A.; Matthaei, G.L. Erratum: New equivalent circuits for elementary piezoelectric transducers. Electron. Lett. 1970, 6, 560. [CrossRef] Sensors 2020, 20, 47 17 of 17

21. Castillo, M.; Acevedo, P.; Moreno, E. KLM model for lossy piezoelectric transducers. Ultrasonics 2003, 41, 671–679. [CrossRef] 22. Redwood, M. Experiments with the Electrical Analog of a Piezoelectric Transducer. J. Acoust. Soc. Am. 1964, 36, 1872–1880. [CrossRef] 23. Kossoff, G. The Effects of Backing and Matching on the Performance of Piezoelectric Transducers. IEEE Trans. Sonics Ultrason. 1966, 13, 20–30. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).