sensors Article Miniaturized 0.13-µm CMOS Front-End Analog for AlN PMUT Arrays Iván Zamora , Eyglis Ledesma , Arantxa Uranga and Núria Barniol * Departament d’Enginyeria Electrònica, Universitat Autónoma de Barcelona, 08193 Bellaterra, Spain; ivan.zamora@uab.es (I.Z.); eyglis.ledesma@uab.es (E.L.); arantxa.uranga@uab.es (A.U.) * Correspondence: nuria.barniol@uab.cat  Received: 30 January 2020; Accepted: 20 February 2020; Published: 22 February 2020  Abstract: This paper presents an analog front-end transceiver for an ultrasound imaging system based on a high-voltage (HV) transmitter, a low-noise front-end amplifier (RX), and a complementary-metal-oxide-semiconductor, aluminum nitride, piezoelectric micromachined ultrasonic transducer (CMOS-AlN-PMUT). The system was designed using the 0.13-µm Silterra CMOS process and the MEMS-on-CMOS platform, which allowed for the implementation of an AlN PMUT on top of the CMOS-integrated circuit. The HV transmitter drives a column of six 80-µm-square PMUTs excited with 32 V in order to generate enough acoustic pressure at a 2.1-mm axial distance. On the reception side, another six 80-µm-square PMUT columns convert the received echo into an electric charge that is amplified by the receiver front-end amplifier. A comparative analysis between a voltage front-end amplifier (VA) based on capacitive integration and a charge-sensitive front-end amplifier (CSA) is presented. Electrical and acoustic experiments successfully demonstrated the functionality of the designed low-power analog front-end circuitry, which outperformed a state-of-the art front-end application-specific integrated circuit (ASIC) in terms of power consumption, noise performance, and area. Keywords: ultrasound; PMUT; high-voltage (HV) transmitter; low-voltage receiver (RX) amplifier; ultrasound application-specific integrated circuit (ASIC); monolithical integration; CMOS; MEMS 1. Introduction Ultrasound, since it was discovered, has been a widely used tool for multiple applications such as medical echography and nondestructive testing. However, today the ultrasound sensing market is showing an impressive resurgence: new applications along with improved manufacturing capabilities and advanced technological readiness are driving the growth of micromachined ultrasound transducers (MUTs). Volumetric medical imaging [1], in vivo and in vitro neuromodulation ultrasound [2], fingerprint sensing [3], and gesture recognition [4] are some new applications based on MUTs. Nowadays, two different micromachined ultrasound transducers can be found in the literature [5]: the first one is based on a capacitive resonant element (CMUT) that consists of a thin metallized suspended membrane over a cavity with a rigid metallized substrate. A DC voltage is applied between two electrodes, which causes the membrane to deflect toward the substrate due to the electrostatic force. Therefore, an ultrasound can be generated in the surrounding medium from the vibration of the membrane when AC voltage is imposed. CMUTs, although they can be integrated over a preprocessed complementary-metal-oxide-semiconductor (CMOS) substrate, have the following drawbacks: they require a large DC bias, different transmission and reception arrays may be required for imaging applications, and they have higher equivalent capacitance. The second transducer is based on piezoelectric materials (PMUTs), in which (in contrast to CMUTs) the deflection of the membrane is produced by the lateral strain generated from a piezoelectric actuation, whereby the membrane Sensors 2020, 20, 1205; doi:10.3390/s20041205 www.mdpi.com/journal/sensors Sensors 2020, 20, 1205 2 of 18 Sensors 2020, 20, x FOR PEER REVIEW 2 of 19 deflection must of the include membrane at least is produced one piezoelectric by the layer lateral as well as strain generated a passive elastic from layer.aAlthough piezoelectric PMUTsactuation, have whereby the membrane must include at least one piezoelectric layer as well as a passive low electromechanical coupling coefficients and produce little output acoustic pressure with narrow elastic layer. Although PMUTs bandwidths compared have to low CMUTs,electromechanical PMUTs do not coupling require DC coefficients bias, they and have produce little output fewer geometric and acousticconstraints, design pressure withand narrow the same bandwidths transmissioncompared to CMUTs, and reception PMUTs arrays do notcoupling with equal require DC maybias, be they for used have fewer geometric imaging and applications [5].design constraints, Additionally, and the same the capability transmission of using and reception a low thermal process forarrays the with equal coupling piezoelectrical maysuch material, be used for imaging as aluminum applications nitride [5]. Additionally, (AlN), allows the capability for the possibility of using a of monolithically low thermal process integrating PMUTs on CMOS.for the piezoelectrical material, such as aluminum nitride (AlN), allows for the possibility of monolithically integrating PMUTs on CMOS. Figure 1 shows a typical block diagram for ultrasound imaging applications using PMUTs. On the Figure 1 shows a typical block diagram for ultrasound imaging applications using PMUTs. On transmission side, the high-voltage (HV) pulser is able to drive the ultrasound transducer to emit the transmission side, the high-voltage (HV) pulser is able to drive the ultrasound transducer to emit enough acoustic pressure. In order to focus on a specific point, a transmission beamforming controller enough acoustic pressure. In order to focus on a specific point, a transmission beamforming controller is used to generate the respective delays. A high-voltage DC–DC Converter may be required to is used to generate the respective delays. A high-voltage DC–DC Converter may be required to generate the supply voltage of the HV pulser. On the reception path, isolation switches included generate the supply voltage of the HV pulser. On the reception path, isolation switches included in in the front-end amplifier functional block are used to isolate the HV present in the transmission of the front-end amplifier functional block are used to isolate the HV present in the transmission of low- low-voltage circuits used in the reception. Due to the weak electrical signals generated by the PMUTs, voltage circuits used in the reception. Due to the weak electrical signals generated by the PMUTs, a a low-noise amplifier (LNA) is employed to amplify these signals. After amplification, the amplitudes low-noise amplifier (LNA) is employed to amplify these signals. After amplification, the amplitudes of these signals are extracted by an envelope detector and captured by a sample-and-hold circuit, which of these signals are extracted by an envelope detector and captured by a sample-and-hold circuit, will later be digitized and processed. In order to obtain an image, most ultrasound measurements are which will later be digitized and processed. In order to obtain an image, most ultrasound based on the pulse echo method, where a generated short ultrasound pulse is propagated in a specified measurements are based on the pulse echo method, where a generated short ultrasound pulse is direction and is partly reflected wherever there is a change in the acoustic properties of the medium. propagated in a specified direction and is partly reflected wherever there is a change in the acoustic The reflected ultrasound echo is processed in order to obtain information about distance, shape, and properties of the medium. The reflected ultrasound echo is processed in order to obtain information the physical properties of the target. An image is formed by mapping the echo strength versus travel about distance, shape, and the physical properties of the target. An image is formed by mapping the time (proportional to the distance). echo strength versus travel time (proportional to the distance). Figure1.1. Image Figure Image of of an an ultrasonic ultrasonicsystem systemblock blockdiagram. diagram. The integration The integration of of PMUTs PMUTs andand CMOS CMOS to to form form aasingle single chip chipallows allows forfor aareduction reduction in inparasitic parasitic elements, size, weight, and power consumption of the overall system. Recent elements, size, weight, and power consumption of the overall system. Recent works [3,6–8] have works [3,6–8] have presented aa single-chip presented single-chip ultrasonic fingerprint fingerprint sensor sensor based based on onPMUTs PMUTsthat thatare aredirectly directlybonded bondedtotoa aCMOS CMOSreadout readoutapplication-specific application-specificintegrated integrated circuit circuit (ASIC) (ASIC) using Al-Ge Al-Ge eutectic eutectic bonding. bonding. This This integration strategy (presented by Reference [8]) is a considerable improvement integration strategy (presented by Reference [8]) is a considerable improvement over the wire bonding over the wire bondingand method method andthe reduces reduces theparasitic electrical electricalelements, parasitic but elements, but itexpensive it is a very is a very process, expensiveandprocess, due to and the due to the presence of special eutectic bonding, the fill factor is affected. In order to presence of special eutectic bonding, the fill factor is affected. In order to overcome these drawbacks,overcome these drawbacks, the Silterra foundry has implemented a MEMS-on-CMOS platform, where an AlN-based PMUT can be fabricated over CMOS preprocessed wafers. This monolithic integration of a PMUT Sensors 2020, 20, 1205 3 of 18 the Silterra foundry has implemented a MEMS-on-CMOS platform, where an AlN-based PMUT can be fabricated over CMOS preprocessed wafers. This monolithic integration of a PMUT over the CMOS reduces fabrication complexity, minimizes the interconnections between PMUT neighbors and the CMOS electronic circuitry, and reduces the parasitic elements; consequently, a higher fill factor and improved signal integrity are possible. In this work, we present the design of a 0.13-µm CMOS front-end transmitter and receiver using the MEMS-on-CMOS technological platform from Silterra [9]. This paper is organized as follows: Section 2 describes the design of the system, and in particular, Section 2.1 presents the ultrasonic transducer; Section 2.2 describes the high-voltage transmitter circuitry and its simulation results; Section 2.3 describes two of the front-end amplifiers that are most used to convert the electric charge generated by PMUT into voltage; and Section 2.4 compares both reception (RX) approaches, making use of the simulation results. Section 3 presents the electrical characterization of the circuitry, and Section 4 shows the acoustic experiments, including an ultrasound image. 2. Front-End Analog Circuit Design 2.1. Ultrasonic Transducer As proof of concept for the integrated front-end circuitry, an 80-µm/side AlN PMUT in a 6 × 6 array configuration was used [10]. The electrical capacitance of a single PMUT was estimated at around 200 fF, adding up to a total of 1.2 pF per row. The obtained resonance frequency in fluorinert (FC-70) was in the 3-MHz range, with a bandwidth around 1 MHz [10]. 2.2. High-Voltage Transmitter The transmitter circuit is in charge of the generation of a large voltage signal in order to excite the transducer at its resonance frequency (fr ) and provide enough acoustic pressure for a high signal-to-noise ratio in the echo measurements. We set the needed actuation voltage to a maximum voltage of 32 V to guarantee enough output pressure [10]. Additionally, and taking into account the fact that the square wave increases the effective amplitude 1.27 times compared to a sine wave, we chose a squared-wave transmitter circuit with a 32-V peak voltage and with rising and falling times less than 100 ns to guarantee at least a 50% duty cycle and a 3-MHz driving frequency (according to the expected PMUT resonance frequency in fluorinert). Previous transmitter circuits [11,12] have used a conventional level shifter that consists of an nMOS differential pair and cross-coupled pMOS transistors. This structure, although it is very simple, has two main problems. On the one hand, the input transistors (nMOS) need to be large in order to overpower the pMOS load when the inputs change, resulting in high input capacitance and therefore limiting the maximum operation speed. On the other hand, a large crowbar current flows when both the nMOS and pMOS transistors are conducting, affecting the efficiency and power consumption of the system. In Reference [13], a charge-recycling HV pulser (CRHV) was presented in order to improve the power consumption, although it needs four control signals with different logic levels per channel, which makes the circuitry more complex. In order to solve the issues mentioned above, we chose the circuit proposed by Reference [14]. This circuit has the basic structure of a conventional level shifter (see Figure 2), but introduces additional high-voltage devices (Msw+ and Msw-) in a series with latching transistors (M3 and M4) in order to avoid the crowbar current. These new transistors work as switches controlled by the high-voltage signals Vsw+ and Vsw-, which inhibit the current of M3 and M4 when the nMOS devices (M1 and M2) are driving the current. The control signals (Vsw+ and Vsw-) are obtained from the output of a second dynamic level shifter, where its input signal is delayed with respect to Vin+ [15]. Sensors 2020, 20, x1205 FOR PEER REVIEW 44of of 19 18 Figure 2. Schematic of the high-voltage transmitter. (a) Complete architecture to excite the piezoelectric micromachined ultrasonic transducer (PMUT). (b) An inverter circuit. (c) Cascaded inverter used as a delay element and (d) an output buffer using an inverter string to drive large capacitive loads. the high-voltage 2.2.Schematic signals of the Vsw+ andtransmitter. Vsw-, which inhibit the currenttoofexcite M3 and M4 when the Figure Figure Schematic of high-voltage the high-voltage (a) Complete transmitter. architecture (a) Complete architecturethetopiezoelectric excite the nMOS devices micromachined piezoelectric (M1 and M2) are driving ultrasonic transducer micromachined ultrasonic the current. (PMUT). The control (b) An(PMUT). transducer signals inverter circuit. (b) An (c) (Vsw+ Cascaded inverter and Vsw-) inverter circuit. are obtained used as a (c) Cascaded frominverter the output delay used as a delay element and (d) an output buffer using an inverter string to drive large to element of a andsecond (d) an dynamic output level buffer shifter, using an where inverter its string input to signal drive large is delayed capacitive with loads. respect Vin+ capacitive [15]. loads. In order In order to optimize the the maximum maximum operation operation speed speed and and dynamic dynamic powerpower consumption, consumption, aa properproper dimensioning dimensioning of the nMOS and pMOS transistor the high-voltage signals Vsw+ and Vsw-, which inhibit the sizes was done. Figure Figure 3 shows 3 shows current of M3 the theandsimulated simulated M4 when source source the and and sink sink currents and the rise and fall times in the output node (Vout+ node). nMOS devices (M1 and M2) are driving the current. The control signals (Vsw+ and Vsw-) are obtained It It can can be be seen seen how how the the presence of the Msw+ of the Msw+ from the output and Msw- and Msw- of a second devices dynamicdevices allows levelallowsfor the shifter,for removal the removal where of the its input of crowbar the is signal current. crowbar A delayedcurrent. high-voltage A high- with respect to inverter voltage Vin+ buffer was inverter [15]. designed buffer to drive to was designed large capacitive drive loads. The large capacitive dimensions loads. of the MOSFETs The dimensions selected of the MOSFETs for this selected circuit for this In order toare shown circuit optimize inshown arethe Table 1. in Table maximum 1. operation speed and dynamic power consumption, a proper dimensioning of the nMOS and pMOS transistor sizes was done. Figure 3 shows the simulated source and sink currents and the rise and fall times in the output node (Vout+ node). It can be seen how the presence of the Msw+ and Msw- devices allows for the removal of the crowbar current. A high- voltage inverter buffer was designed to drive large capacitive loads. The dimensions of the MOSFETs selected for this circuit are shown in Table 1. Figure 3. Voltage and source and sink currents in the output node (Vout+) [15]. Figure 3. Voltage and source and sink currents in the output node (Vout+) [15]. Figure 3. Voltage and source and sink currents in the output node (Vout+) [15]. Sensors 2020, 20, x FOR PEER REVIEW 5 of 19 Sensors 2020, 20, 1205 5 of 18 Table 1. Dimensions of the MOSFET devices used in the transmitter. MOSFETs Aspect Ratio (W/L) (μm/μm) Table 1. Dimensions of the MOSFET devices used in the transmitter. M1, M2 20/5.75 M3, M4, Msw+, Msw- MOSFETs 10/4 Aspect Ratio (W/L) (µm/µm) M5, M6 10/5.75 M1, M2 20/5.75 M7, M8, M9, M10 M3, M4, Msw+, Msw- 4.7/4 10/4 M11 M5, M6 10/1 10/5.75 M7, M8, M9, M10 M12 4.7/4 25/1 M11 M13 10/1 1/4 M12 25/1 M14 2.5/4 M13 1/4 M15 M14 1/3 2.5/4 M16 M15 2.5/3 1/3 M16 M17 2.5/3 10/4.8 M17 M18 10/4.8 25/4 M18 25/4 M19 M19 20/4.8 20/4.8 M20 M20 50/4 50/4 In most imaging ultrasound applications, the transducer is an array of PMUTs. To obtain enough In most imaging ultrasound applications, the transducer is an array of PMUTs. To obtain enough acoustic pressure level, several PMUTs of the array emit simultaneously, forming a transmit channel. acoustic pressure level, several PMUTs of the array emit simultaneously, forming a transmit channel. In order to reduce the size of the die, one HV transmitter is used per channel, and therefore it has to In order to reduce the size of the die, one HV transmitter is used per channel, and therefore it has to be be able to drive a large load capacitance at the resonance frequency. Figure 4 shows the simulation able to drive a large load capacitance at the resonance frequency. Figure 4 shows the simulation results results of the maximum operation frequency of our HV transmitter, taking into account the designed of the maximum operation frequency of our HV transmitter, taking into account the designed output output buffer, as a function of the load capacitance. Taking into account that the capacitance buffer, as a function of the load capacitance. Taking into account that the capacitance associated with associated with an individual PMUT was in the range of 200 fF, we could make Figure 4, which an individual PMUT was in the range of 200 fF, we could make Figure 4, which indicates the maximum indicates the maximum number of PMUTs simultaneously actuated at each frequency (e.g., in the number of PMUTs simultaneously actuated at each frequency (e.g., in the case of 3 MHz, the maximum case of 3 MHz, the maximum capacitance corresponded to 37 pF, which is equivalent to 185 capacitance corresponded to 37 pF, which is equivalent to 185 individual PMUTs, a reasonable number individual PMUTs, a reasonable number for a linear array configuration). for a linear array configuration). With this circuit, 2.9 ns of a 10–90% rise/fall time was obtained for 1.2 pF (equivalent to one row With this circuit, 2.9 ns of a 10–90% rise/fall time was obtained for 1.2 pF (equivalent to one row of of a 6 × 6 PMUT array) of load capacitance, achieving the design requirements. a 6 × 6 PMUT array) of load capacitance, achieving the design requirements. Figure 4. Maximum transmitter operating frequency as a function of load capacitance (simulation Figure 4. Maximum transmitter operating frequency as a function of load capacitance (simulation results). results). 2.3. Front-End Amplifier 2.3. Front-End Amplifier The operation of PMUTs as a sensor is governed by the direct piezoelectric effect, where an The charge electrical operation of PMUTswhen is generated as a the sensor is governed acoustic by theatdirect wave arrives piezoelectric the PMUT. In ordereffect, wherethis to measure an electrical electric charge charge, is generated it was necessarywhen the acoustic to design wave a low-noise arrives (LNA) amplifier at the PMUT. In order at the first to the stage of measure this front-end electric charge, it was necessary to design a low-noise amplifier (LNA) at the first stage PMUT receiver circuit that was a tradeoff between power supply, noise performance, gain, bandwidth, of the front- Sensors 2020, 20, 1205 6 of 18 dynamic range, and die size. Transimpedance amplifier topologies (TIAs) are widely used to amplify the weak signal generated by relatively high-impedance CMUTs [12,16]. Nevertheless, piezoelectric transducers (PZTs and PMUTs) with similar sizes present much lower equivalent impedance around their resonance frequency, whereas this topology suffers from a low noise/power tradeoff when it is chosen as a front-end amplifier for PMUT devices. A capacitive feedback voltage amplifier (CFVA) with a split-capacitor feedback network has been a very common topology in readout integrated circuits (ICs) for PZT transducers [17,18] (in order to sense the transducer’s voltage rather than its electric charge). Nevertheless, this topology suffers from a large area due to the use of several capacitors. This work presents a comparative analysis of two of the most used front-end amplifier configurations in sensors based on PMUTs: a voltage amplifier (VA) based on capacitive integration [11,13,15] and a charge-sensitive amplifier (CSA) [3,19]. 2.3.1. Voltage Amplifier (VA) Description Figure 5a shows a scheme of a front-end amplifier based on capacitive integration. The accumulated electric charge on the input equivalent capacitance (the parallel between CPMUT , the electric capacitance of the PMUT; Cin , the input capacitance of the voltage amplifier; and Cp , all of the parasitic capacitances) is converted into voltage and is later amplified by the designed voltage amplifier. This relationship can be expressed by Equation (1), where A0 is the open-loop amplifier voltage gain and QE is the electric charge generated by the PMUT: QE Vout = A0 (1) CPMUT + CP + Cin To find a tradeoff between power consumption, gain, bandwidth, and area, a single-ended input self-biased push–pull configuration was selected as the VA topology (see Figure 5c). With this configuration, a smaller possible area could be used for the amplifier, which allows for a pitch-matched circuit with a single PMUT. The input-referred noise current of this topology was derived from a small-signal equivalent model considering all parasitic capacitance (see Equation (2)): s(C +C )   sCin (Cf + Cin ) in,n + in,p [1 + g f +gout ] ds,n ds,p in,in (s) =   (2) sCf s(C C +C C +C C )    A0 [Cf gm,n + gm,p + Cin gds,n + gds,p ] 1 − g +g [1 +  in f out f  out in ] m,n m,p Cf gm,n +gm,p +Cin gds,n +gds,p where in,n and in,p are the square root of the mean square current noise for nMOS and pMOS transistors; gm,n and gm,p are the transconductance for the M1 and M2 transistors, respectively; gds,n and gds,p are the channel conductance of M1 and M2, respectively; Cf is the feedback capacitance (in this topology, the sum of gate-to-drain capacitances); and Cout is the output capacitance of the circuit (approximately the load capacitance). Due to the small current generated by the PMUT (nanoamperes), the minimization of amplifier noise is a must. In order to minimize the input current noise and maximize the transimpedance gain, the pMOS M1 and nMOS M2 transistors were correctly sized. To maximize the dynamic range, the operation point of the amplifier was fixed to VDD /2 by a pMOS M3 transistor connected between the input and output (operating in the subthreshold region, which functions as very high-impedance resistance) [20]. A source follower buffer was added to match the output impedance to 50 Ω for testing. Sensors 2020, 20, 1205 7 of 18 Sensors 2020, 20, x FOR PEER REVIEW 7 of 19 (a) (b) (c) Figure5.5. PMUT Figure PMUT front-end front-end amplifier amplifier configurations: configurations: (a) (a) aa voltage voltage amplifier amplifier (VA) (VA)based basedononcapacitive capacitive integration, (b) a CSA amplifier, and (c) a proposed schematic for the open-loop voltage amplifier, integration, (b) a CSA amplifier, and (c) a proposed schematic for the open-loop voltage amplifier, A A0,, 0 forthe for theVA VAandandCSA CSAreceivers. receivers. 2.3.2. 2.3.2. Charge-Sensitive Charge-Sensitive Amplifier Amplifier (CSA) (CSA)Description Description The The second second approach approach is is the the charge-sensitive charge-sensitive amplifier amplifier shown shown in in Figure Figure 5b, 5b, where where the the output output voltage, voltage, given given by byEquation Equation(3), (3),only onlydepends dependson onthe thefeedback feedbackcapacitance capacitance(C (Cf ), f), considering considering that that the the input impedance of the amplifier is much smaller than the equivalent input impedance of the amplifier is much smaller than the equivalent impedance given by PMUT impedance given by PMUT and the andparasitic elements: the parasitic elements: Q QE (3)( V Vout = = .C . C f 3) Sincethe Since theinput inputimpedance impedanceisisproportional proportionaltoto1/(C 1/(C inin++ CCff*A00), it is necessary to choose an an amplifier amplifier topologythat topology thatguarantees guaranteesaahigh highopen openlooploopgain gainAA00 (with (with minimum minimum dimensions) dimensions) to to decrease decrease Cin Cin and and the area. the area. This reduction in the the input input capacitance capacitance impedance impedance allows allows forfor the the use use of of aa low-feedback low-feedback capacitance,CCf ,f,thus capacitance, thusachieving achieving a maximum a maximum charge charge transfer transfer and andSNR.SNR. Therefore, Therefore, the samethe single-ended same single- endedself-biased input input self-biasedpush–pull push–pull configurationconfiguration used in used the VA in was the VA the was rightthe right selection. selection. Inorder In ordertotoselectselectthetheCC , wetook f , fwe tookinto into account account thetheinputinput referred referred noise, noise, the the charge charge transfer transfer fromfrom the the PMUT PMUT to theto the feedback feedback capacitance, capacitance, and the and the gain. The gain. The input-referred input-referred noise currentnoise of thiscurrent of this configuration isconfiguration given by the is given byexpression: following the following expression: s(C + C ) i , + i , i , (s) = (  s(Cf + Cin ) in,nsC + in,p in,in (s)g =, + g , 1 − g , + gsC,  4) (4) f This equation shows that the input noise g + g m,ncurrent 1 − m,p is directlygm,n +gproportional to Cf, and therefore the m,p feedback capacitance must be selected to be as low as possible. In order to perform good signal This equation processing, at leastshows 85% ofthat thethe inputtransfer charge noise current must be is guaranteed. directly proportional Considering to Caf , C and therefore PMUT = 1.2 pF the feedbackto (equivalent capacitance one row of must six be selected 80-μm to be [10]), PMUTs as lowaas possible. In parametric AC order to perform analysis good signal was performed to processing, at least 85% of the charge transfer must be guaranteed. compute the charge transfer. Figure 6 shows that for a Cf higher than 400 fF, the amplifier behaves Considering a C PMUT = 1.2 pF as (equivalent a CSA andto one row ofatsix guarantees least80-µman 85%PMUTs [10]),transfer. charge a parametric AC analysis Therefore, in orderwastoperformed maximizetothe compute output the charge transfer. Figure 6 shows that for a C f higher than 400 fF, voltage given by Equation (3) and to find a tradeoff between noise performance and charge transfer, the amplifier behaves as a CSA and guarantees 400 fF was selected as the feedback capacitance. In this case, a pMOS M3 transistor was used to by at least an 85% charge transfer. Therefore, in order to maximize the output voltage given fix Equation (3) and to find a tradeoff between noise performance and the operation point and was dimensioned (without affecting the bias) to avoid any leakage current charge transfer, 400 fF was selected as thecould that feedback charge capacitance. Cf, leading to In saturation this case, aofpMOS M3 transistor the amplifier. The same wassource-follower used to fix the operation buffer waspointused and was dimensioned (without to match the output impedance to 50 Ω. affecting the bias) to avoid any leakage current that could charge Cf , leading to saturation of the amplifier. The same source-follower buffer was used to match the output impedance to 50 Ω. Sensors 2020, 20, x FOR PEER REVIEW 8 of 19 Sensors Sensors 20,20, 2020, 2020, 1205PEER REVIEW x FOR 8 of8 19 of 18 Figure Figure 6. 6. Charge Charge transfer transfer obtained obtained for for different feedback capacitances different feedback capacitances (C (Cff).). Figure 6. Charge transfer obtained for different feedback capacitances (Cf). In Figure 7, the final final layout layoutof ofthe theLNA LNAbased basedon onaaVA VAisisshown showntogether togetherwith anan with optical image optical imageof In thethe of Figure PMUT PMUT 7, the array, final layout demonstrating array, of demonstrating the LNA based its capabilities on for for its capabilities a VA a minuteis shown areaarea a minute together andand with an a pitch-matched optical system a pitch-matched image (for(for system the of the the CMOS CMOS PMUT array, demonstrating its capabilities for a minute area and a pitch-matched system (for receiver). receiver). the CMOS receiver). (a) (b) (a) (b) Figure Figure 7. 7. (a) Layout of (a) Layout of the theLNA LNA(VA(VAtopology) topology)with withanan output output buffer. buffer. TheThe LNALNA dimensions dimensions are are 23 µm23 Figure μm 7. (a) of widthLayout of the LNA (VA topology) with an output buffer. The LNA dimensions are 23 of width andand 25 μm 25 µm of length. of length. (b) Optical (b) Optical image image of a of 6 ×a 66 (square) × 6 (square) 80-μm/side 80-µm/side PMUTPMUT array. array. μm of width and 25 μm of length. (b) Optical image of a 6 × 6 (square) 80-μm/side PMUT array. 2.4. Front-End 2.4. Front-End Amplifier Amplifier Simulation Simulation Results Results 2.4. Front-End Amplifier Simulation Results All simulations All simulations were were donedone withwith 14 14 pF pF and and 5050 ΩΩ asas the the load load impedance impedance (input (input impedance impedance of of the the All simulations oscilloscope). The oscilloscope). were The PMUT PMUTwas done with was modeled 14 modeledas pF and 50 asaa currentΩ as the current source load source in impedance in parallel parallel with (input with its impedance its electrical of electrical capacitance. the capacitance. oscilloscope). The assumed The PMUT operation was modeled frequency was as a the current resonance source in frequency The assumed operation frequency was the resonance frequency of the PMUT array in liquid (3 parallel of thewith PMUT its electrical array in capacitance. liquid (3MHz), MHz), The assumed operation frequency which was given by Reference [10]. was the resonance frequency of the PMUT array in liquid (3 MHz), which was given by Reference [10]. 2.4.1. Transimpedance 2.4.1. Transimpedance Gain Gain andand Input Input Impedance Impedance 2.4.1. Transimpedance Gain and Input Impedance Figure 88shows Figure showsthethe transimpedance transimpedance gaingain of both of both approaches approaches (the voltage (the voltage amplifier amplifier and charge- and Figure sensitive 8 charge-sensitive shows amplifier) the transimpedance amplifier) and the and the voltage voltage gain of both open-loop open-loop approaches gain gain A0. ItAcan (the 0 . It be canseenvoltage be seen thatthatamplifier the the VAVA and charge- amplifier amplifier had had a sensitive a higher amplifier) transimpedanceand the voltage gain (≈23 dB Ω), open-loop gain since theA higher transimpedance gain (≈23 dB Ω), since the output voltage of the VA depended not only on athe 0. It canvoltage output be seen of that the the VA VA amplifier depended not had only on higher transimpedance the amplifier amplifier open-loop open-loop gaingain gain A(≈23 AdB 0, but Ω), also 0 , also but since on theonthe output the voltage integration integration of theofof thecurrent the current VAindepended theininput not the input only on the capacitance. capacitance. The amplifier The CSA open-loop transimpedancegain A , but 0gain also dependedon the integration only on the of value theof current CSA transimpedance gain depended only on the value of the feedback capacitance (it needs to bethe in feedback the input capacitance. capacitance (it needs The to be CSA transimpedance lower than the total gain VA depended input only capacitance). on the However, value the ofVAthe feedback amplifier lower than the total VA input capacitance). However, the VA amplifier frequency response was capacitance frequency (it response needs was to be limited lower by the limitedthan bythe open the total loopopen VA loopinput bandwidth capacitance). (21.7 bandwidth MHz). However, (21.7InMHz). the case In of the the the case VA CSA, of amplifier its thefrequency CSA, its frequency behavior frequency response was was limited behavior by was limited limited by by its input the open impedance its loop input and bandwidth the equivalent impedance (21.7 the and MHz). impedance In the equivalent ofcase PMUTs, of the impedance plusCSA, of its the frequency parasitic PMUTs, behavior capacitance. plus was At the the parasitic limited by its input impedance and the equivalent impedance of PMUTs, plus the parasitic Sensors 2020, 20, x FOR PEER REVIEW 9 of 19 Sensors 2020, 20, x FOR PEER REVIEW 9 of 19 Sensors 2020, 20, At capacitance. 1205the frequency where both impedances were similar (around 140 MHz), the electric 9 of 18 capacitance. At the frequency where both impedances were similar (around 140 MHz), the electric charge generated by the PMUT was divided, and therefore the CSA stopped behaving like a charge- charge generated sensitive by the PMUT was divided, and therefore the CSA stopped behaving like a charge- amplifier. frequency sensitive where both impedances were similar (around 140 MHz), the electric charge generated by amplifier. the PMUT was divided, and therefore the CSA stopped behaving like a charge-sensitive amplifier. Figure 8. Figure 8. Frequency Frequency response response ofof both both amplifiers amplifiers (VA (VA and and CSA) CSA) (left (left axis) axis) along along with with the the push–pull push–pull Figure 8. Frequency open-loop response response of A of both amplifiers (VA and CSA) (left axis) along with the push–pull 0 (right axis). open-loop response of A0 (right axis). open-loop response of A0 (right axis). Figure9 9shows Figure shows thethe dependency dependency of theoftransimpedance the transimpedancegain in gain in both both cases (VAcases (VAamplifiers), and CSA and CSA Figure 9 shows amplifiers), thethe with and dependency PMUTs and of thecapacitances. parasitic transimpedance gain ingiven Asgiven expected both cases (VA(1)and CSA with the PMUTs parasitic capacitances. As expected EquationsEquations andVA (1) and (3), the (3),had the amplifiers), with the PMUTs and parasitic capacitances. As expected given Equations (1) and (3), the a steeper slope (∆TIGain = 7.5 dBΩ), since it amplified the generated input voltage through the VA had a steeper slope (ΔTIGain = 7.5 dBΩ), since it amplified the generated input voltage through VAthehadintegration a steeper slope the(ΔTIGain = 7.5 dBΩ), all since theitcapacitances amplified thepresent generated input voltage of thethrough integration of theofPMUT PMUT current current in the in all of of capacitances in the present in the input ofinput amplifier. the amplifier. Instead, theInstead, integration of the PMUT current in all of the capacitances present in the input of the amplifier. the transimpedance the transimpedance gain of thegain CSAofhadtheaCSA had linear a linear dependency dependency withslope with the lower the lower = (ΔTIGain slope (∆TIGain 0.8 dBΩ), Instead, the transimpedance = 0.8still dBΩ), but had still more the VA gain hadofmore the CSA gain.had a linear dependency with the lower slope (ΔTIGain but the VA gain. = 0.8 dBΩ), but still the VA had more gain. Figure 9. Dependence of the transimpedance gain at 3 MHz, with equivalent capacitance given by the Figure 9. Dependence of the transimpedance gain at 3 MHz, with equivalent capacitance given by the PMUT Figure plus parasitic 9. Dependence ofcapacitances. the transimpedance gain at 3 MHz, with equivalent capacitance given by the PMUT plus parasitic capacitances. PMUT plus parasitic capacitances. 2.4.2. Noise Performance and Dynamic Range 2.4.2. Noise Performance and Dynamic Range To compare 2.4.2. Noise the noise Performance and performance Dynamic Range of both amplifiers, the output noise voltage (Vn,out ) and the inputTo compare noise the current noise (In,in performance ), defined of both(2) by Equations amplifiers, the output and (4), were noise taken into voltageThe account. (Vn,out) and the presence of To compare input noise the noise current (I ),performance defined by of both amplifiers, Equations (2) and the (4), output were noise taken intovoltage account.(VThe n,out) and the presence of the feedback capacitor in the CSA amplifier caused its output impedance to be smaller than the VA n,in input the noise current feedback (In,in), defined capacitor in the by Equations CSA amplifier (2) and its caused (4),output were taken into account. impedance to be The presence smaller than ofVA the output impedance, causing the output voltage noise to be smaller (see Figure 10). However, due to the theoutput feedback capacitor causing impedance, in the CSA the amplifier caused output voltage its output noise impedance (see to be smaller than the VA to big difference between their transimpedance gains, thetoinput be smaller Figure noise current 10). However, was lower for the VAduethan output impedance, causing √ the output voltage √ noise to be smaller (see Figure 10). However, due to for the CSA (0.08 pA/ Hz and 0.12 pA/ Hz, respectively, at a 3-MHz frequency). Sensors Sensors2020, 2020,20, 20,xxFOR FORPEER PEERREVIEW REVIEW 10 10 of of 19 19 the thebig big Sensors difference 20, 1205 between difference 2020, betweentheir theirtransimpedance transimpedancegains, gains,the theinput inputnoise noisecurrent currentwas waslower lowerfor forthe the VA VA 10 of 18 than thanfor forthe theCSA CSA(0.08 (0.08pA/√Hz pA/√Hzand and0.12 0.12pA/√Hz, pA/√Hz,respectively, respectively,at ataa3-MHz 3-MHzfrequency). frequency). (a) (a) (b) (b) Figure 10.Computations Figure10. Computationsusing usingEquations Equations(2) (2)and and(4) (4)and andthe thesimulated simulatedRX RXamplifiers’ amplifiers’noise noisespectral spectral density. density.(a) density. (a)Output Outputvoltage noise voltagenoise and noiseand (b) and(b) input (b)input current inputcurrent noise. currentnoise. noise. Figure Figure 11 11shows showsthe shows thesimulated the simulateddynamic simulated dynamicrange dynamic rangefor range foraaa3-MHz for 3-MHzinput 3-MHz inputsignal. input signal. AA maximum signal. maximum inputinput current currentof of100 of 100nA 100 nAand nA and11µA and 1 μA μA for the for for VA the the VA VAand CSA and and CSAwere CSA obtained, were were respectively. obtained, obtained, ThisThis respectively. respectively. greatgreat This difference great was difference difference expected was due to was expected expected theto due due bigthe to difference the big between big difference difference their gains. between between theirIn their theseIn gains. gains. results, In thesethe these CSA the results, results, provided the CSA a dynamicaa CSA provided provided range dynamic10 times dynamic rangehigher range 10 10timesthan times the VA higher higher thanamplifier, than the VAallowing theVA amplifier,itallowing amplifier, to amplify allowing ititPMUT to signals toamplify amplify PMUT PMUT10 times larger signals signals 10 than 10times times those largeramplified larger than thanthose thosebyamplified the VA. by amplified bythe theVA. VA. Figure 11.Simulated Figure11. Simulateddynamic dynamicrange rangeat at33MHz. MHz. 3. Electrical Characterization 3.1. High-Voltage Transmitter The HV transmitter circuit was tested using a signal generator and an oscilloscope. The circuitry, operated from two power supplies (3.3 V and 32 V) and with a 3.3-V-squared input signal, generated output pulses up to 32 V. Figure 12 shows the input and output waveforms of this circuit, which achieved a rise/fall time of 47.5 ns. Using 14 pF as the load capacitance from the oscilloscope plus 3 pF due to the bonding pad, printed circuit board (PCB) pad, and connectors, this value differed by 3.1. High-Voltage Transmitter The HV transmitter circuit was tested using a signal generator and an oscilloscope. The circuitry, operated from two power supplies (3.3 V and 32 V) and with a 3.3-V-squared input signal, generated output pulses Sensors 2020, up to 32 V. Figure 12 shows the input and output waveforms of this circuit, which 20, 1205 11 of 18 achieved a rise/fall time of 47.5 ns. Using 14 pF as the load capacitance from the oscilloscope plus 3 pF due to the bonding pad, printed circuit board (PCB) pad, and connectors, this value differed by 6.36 ns 6.36 ns compared comparedto tothe thesimulation simulationresult. result.This Thisdifference differencecould have could been have duedue been to the estimation to the of the estimation of load capacitance (parasitic capacitance) as well as due to process variations. For this the load capacitance (parasitic capacitance) as well as due to process variations. For this estimated estimated load capacitance, load the HV capacitance, thetransmitter operated HV transmitter at a maximum operated frequency at a maximum of 5 MHz, frequency of 5achieving a 10–90% MHz, achieving a rise 10– time and a 50–50% latency of 47.5 ns and 31 ns, respectively. 90% rise time and a 50–50% latency of 47.5 ns and 31 ns, respectively. The static The static power power consumption consumption was was negligible, negligible, and and the the dynamic dynamic power power consumption, consumption, defined defined as as CLLV 2 V2f,f,was was3.683.68mW mWfor foraa1.2-pF 1.2-pFload loadcapacitance. capacitance. C Waveform of the HV transmitter input and output. Figure 12. Waveform 3.2. Front-End 3.2. Front-End Amplifiers Amplifiers The RX The RXamplifiers werewere amplifiers characterized in terms characterized inofterms voltage–voltage gain for the VA, of voltage–voltage gaintransimpedance for the VA, gain for the CSA,gain transimpedance and for input thecurrent CSA, andnoiseinput and output currentdynamic noise andrange for both output of them. dynamic range for both of them.Figure 13 shows the measured and postlayout simulation voltage–voltage gain of the voltage amplifier Figure(VA). A measured 13 shows bandwidth the measured andofpostlayout about 22 MHz was achieved. simulation Compared voltage–voltage to the gain postlayout of the voltage simulation amplifier and (VA). Aconsidering measured the technology bandwidth of aboutvariations, 22 MHz the was simulated achieved. values Compared were to close the to the postlayout Sensors 2020, 20, x FOR PEER REVIEW 12 of 19 experimental ones. simulation and considering the technology variations, the simulated values were close to the experimental ones. Figure 13. Transfer function of the VA amplifier. Figure 13. Transfer function of the VA amplifier. In order to compute the gain of the CSA, we employed the charge injection method. A 100-fF capacitor connected in series at the CSA input was used as a test injection capacitor (CTEST). Small voltage steps (ΔV) applied to the CTEST produced charge pulses at the CSA input with a value of CTEST Sensors 2020, 20, 1205 12 of 18 In order to compute the gain of the CSA, we employed the charge injection method. A 100-fF capacitor connected in series at the CSA input was used as a test injection capacitor (CTEST ). Small voltage steps (∆V) applied to the CTEST produced charge pulses at the CSA input with a value of CTEST *∆V. Figure 14 shows the output voltage measured at the buffer output for an input charge sweep from 2.5 fC to 18 fC (corresponding to voltage steps from 25 mV to 180 mV). According to Equation (3), the slope of this graph must be inversely proportional to the feedback capacitance, giving 472 fF with Sensorsa2020, 100% 20,charge transfer. x FOR PEER REVIEWTaking into account this and using the designed feedback capacitance 13 of 19 (Cf = 410 fF in the layout), the obtained charge transfer was 87%, close to the simulated results. Sensors 2020, 20, x FOR PEER REVIEW 13 of 19 Figure 14. Figure Tested results 14. Tested results of of the the output output voltage voltage of of the the CSA. CSA. The noise performance of both the VA and CSA amplifiers was measured by opening the input and measuring the output noise. The input noise was computed by dividing the measured output noise by the corresponding transfer function. The resulting spectral density noise is shown in Figure 15, which was in good agreement with the simulations. The measured output-referred voltage noise density √ √ was 60.68 nV/ Hz and 19.01 nV/ Hz at 3 MHz for the VA and CSA amplifiers, respectively. At the √ √ same frequency, the measured input-referred current noise was 68 fA/ Hz and 147 fA/ Hz for the VA and CSA, respectively. Taking into account the PMUT bandwidth (1 MHz in liquid), an input-referred integrated noise of 59.18 pArms and 142.7 pArms was obtained for the VA and CSA, respectively. Figure 14. Tested results of the output voltage of the CSA. (a) (b) Figure 15. Measured spectral density noise: (a) output-referred voltage noise and (b) input-referred current noise. Figure 16 shows the measurements of the VA and CSA output voltage for different input current levels at 3 MHz. The measured transimpedance gain was 109.22 dBΩ and 99.57 dBΩ, respectively. An input-measured dynamic range of 69 dB and 71 dB was achieved, where the previous measured integrated input noise was taken as the minimum detectable current. (a) (b) Figure 15. Figure 15. Measured Measured spectral spectral density density noise: noise: (a) output-referred voltage (a) output-referred voltage noise noise and and (b) (b) input-referred input-referred current noise. current noise. Figure 16 shows the measurements of the VA and CSA output voltage for different input current levels at 3 MHz. The measured transimpedance gain was 109.22 dBΩ and 99.57 dBΩ, respectively. An input-measured dynamic range of 69 dB and 71 dB was achieved, where the previous measured integrated input noise was taken as the minimum detectable current. Sensors 2020, 20, 1205 13 of 18 Considering that the reception sensitivity of the PMUT array [10] was 5.9 V/MPa, the measured √ √ noise equivalent pressure was 0.84 mPa/ Hz and 1.32 mPa/ Hz at 3 MHz. These values are very low for imaging applications, where the received echoes are in the order of kPa, so a good signal-to-noise ratio can be guaranteed. Figure 16 shows the measurements of the VA and CSA output voltage for different input current levels at 3 MHz. The measured transimpedance gain was 109.22 dBΩ and 99.57 dBΩ, respectively. An input-measured dynamic range of 69 dB and 71 dB was achieved, where the previous measured integrated input noise was taken as the minimum detectable current. Sensors 2020, 20, x FOR PEER REVIEW 14 of 19 Figure 16. Dynamic Dynamic range measurements: the the VA VA amplifier (black) and the CSA amplifier (red). (red). 4. Acoustic Characterization 4. Acoustic Characterization To To validate validate thethe analog analog front-end front-end CMOS CMOS in in its its intended intended application, application, anan acoustic acoustic characterization characterization was was done done using usingfluorinert fluorinert(FC-70, (FC-70,where wherethethemass massdensity density % =ρ 1940 is is kgkg = 1940 −3 m m−3and andthe sound the soundvelocity velocity is cis=c 690 m/s) as a propagation medium. The set-up is shown in Figure 17, where = 690 m/s) as a propagation medium. The set-up is shown in Figure 17, where a grating phantom a grating phantom was was introduced introduced in in aa small small pool, pool, confining confining the the FC-70 FC-70 to to approximately approximately 2.1 2.1 mm mm over the 66 × over the × 66 PMUT PMUT array surface. One row of this array was excited by the designed HV transmitter (four array surface. One row of this array was excited by the designed HV transmitter (four cycles at 3 MHz cycles at 3 MHz with with a 32-V peak-to-peak), and another row was dedicated to reception using the proposed VA as an a 32-V peak-to-peak), and another row was dedicated to reception using the proposed VA as an LNA LNA front-end front-end amplifier. amplifier. In In order order to to obtain obtain aa 2D 2D image, image, thethe grating grating phantom phantom was was shifted shifted in the ×× in the direction and yy direction, direction and direction,withwithsteps stepsofof50 μmand 50µm and100100µm,μm,respectively. respectively.Figure Figure18b 18bshows shows anan image imageof aofsection of this phantom (see Figure 18a). The blue bars correspond with the slots, a section of this phantom (see Figure 18a). The blue bars correspond with the slots, with a width with a width close to 1.1 to close mm 1.1and mm1 mm,and as was expected 1 mm, (physical(physical as was expected width), demonstrating the capability width), demonstrating of the system the capability to of the make an ultrasonic image. system to make an ultrasonic image. (a) (b) Figure 17. Set-up for acoustic characterization using an FC-70: (a) the schematic set-up; (b) an optical image of the experimental set-up using a grating phantom. array surface. One row of this array was excited by the designed HV transmitter (four cycles at 3 MHz with a 32-V peak-to-peak), and another row was dedicated to reception using the proposed VA as an LNA front-end amplifier. In order to obtain a 2D image, the grating phantom was shifted in the × direction and y direction, with steps of 50 μm and 100 μm, respectively. Figure 18b shows an image of a section of this phantom (see Figure 18a). The blue bars correspond with the slots, with a width Sensors 2020, 20, 1205 14 of 18 close to 1.1 mm and 1 mm, as was expected (physical width), demonstrating the capability of the system to make an ultrasonic image. (a) (b) Figure Figure 17. 17. Set-up Set-upfor foracoustic acousticcharacterization characterizationusing using an an FC-70: FC-70: (a) (a)the theschematic schematicset-up; set-up;(b) (b)an anoptical optical image Sensors 2020,of image 20,the of experimental x FOR the set-up set-up using PEER REVIEW experimental using aa grating grating phantom. phantom. 15 of 19 (a) (b) Figure 18. Figure 18. (a) (a) Grating Grating phantom phantom and and (b) (b) an an ultrasonic ultrasonic image image of of aaphantom phantomsection. section. 5. Discussion 5. Discussion In this work, we present the design and characterization of two functional blocks of an imaging In this work, we present the design and characterization of two functional blocks of an imaging ultrasound system: an HV transmitter and a front-end receiver amplifier. In addition, we compared ultrasound system: an HV transmitter and a front-end receiver amplifier. In addition, we compared two of the most used front-end amplifiers in sensor-based PMUTs. two of the most used front-end amplifiers in sensor-based PMUTs. In this section, we will compare these proposals to prior work, taking into account all of the In this section, we will compare these proposals to prior work, taking into account all of the obtained results. obtained results. 5.1. HV Transmitter 5.1. HV Transmitter Table 2 summarizes and compares our transmitter to prior reported works. The performance of Table 2 summarizes and compares our transmitter to prior reported works. The performance of the HV transmitter was measured in terms of its speed, area, power consumption, and load capacitance. the HV transmitter was measured in terms of its speed, area, power consumption, and load In order to compare different systems, we defined a figure-of-merit (FOM) that takes into account all of capacitance. In order to compare different systems, we defined a figure-of-merit (FOM) that takes the key parameters that intervene in the transmitter’s performance (see Equation (5)): into account all of the key parameters that intervene in the transmitter’s performance (see Equation (5)): mA VMAX [V]∗CL [pF] FOMTX [ 2 ]= , (5) mm [ ]∗ [ ] trise [ns]∗area[mm2 ] FOM [ ]= [ ]∗ , (5) [ ] where CL corresponds to the nominal load capacitance, VMAX is the maximum pulsed output where CLand voltage, corresponds to the trise is the risenominal time. load Ourcapacitance, designed HV VMAXtransmitter is the maximum pulsed achieved theoutput voltage, best FOM TX and trise is the rise time. Our designed HV transmitter achieved the best FOMTX (1018 mA/mm ), which 2 translated to the best electrical performance with the highest integration density. This performance was attributable in part to the small area that the transmitter has. Table 2. HV transmitter performance summary and comparisons. Sensors 2020, 20, 1205 15 of 18 (1018 mA/mm2 ), which translated to the best electrical performance with the highest integration density. This performance was attributable in part to the small area that the transmitter has. Table 2. HV transmitter performance summary and comparisons. Parameter This Work [13] (2019) [21] (2017) [3] (2016) [16] (2016) 0.13 µm 0.18 µm 0.18 µm HV 0.18 µm HV TSMC 0.18 µm Process technology HV CMOS CMOS BCD8-SOI CMOS HV CMOS Transducer PMUT PMUT CMUT PMUT CMUT Pulsed output voltage (V) 32 5/13.2 100 24 30 Nominal operation frequency (MHz) 3 5 10 14 5 Nominal load capacitance (pF) 1.2 N/A 9.2 2 2 Area (mm2 ) 0.013 N/A 0.09 1 0.017 1 0.016 Rise time (ns) 2.9 N/A 14 10.56 2 6.6 FOMTX (mA/mm2 ) 1018 - 730 267 568 1 These areas were estimated using a chip micrograph. 2 Obtained as trise = 2.2RCL , where R = Vmax/Imax. 5.2. Front-End Receiver Amplifier To compare both presented LNA topologies to the state-of-the-art topologies, we also defined a figure-of-merit that includes the main parameters of the LNA design. Taking into account the gain, input referred noise, power consumption, operation frequency, and area, we defined the following FOMs: Gain[ V p MHz V] ∗ frequency[MHz] FOMRX_1 [ ]= , (6) V2 ∗A ∗ µm2 input voltage noise [ √nV ]∗Power Consumption [mW]∗area [mm2 ] HZ V p Hz Transimpedance gain[ A ]∗ frequency[MHz] FOMRX_2 [ ]= , (7) mA3 ∗ µm2 input current noise [ √ pA ]∗Power Consumption [mW]∗area [mm2 ] HZ where Equation (6) was used to compare the VA and CFVA topologies and Equation (7) was used to compare the CSA and TIA topologies. Table 3 summarizes the comparisons between the performance of the designed LNAs and that of the other amplifiers reported as being state-of-the-art, demonstrating competitive results. Taking the best computed FOMRX_1,2 , the designed VA and CSA topologies achieved ~2.5× and ~45× higher improvement, respectively, guaranteeing the best electrical performance with a minimum area. These competitive FOM values were possible since our two LNAs (based on a self-biased push–pull amplifier) have a smaller area (almost by 10 times), which makes these designs good candidates to implement a pitch-matched system, where the LNA can be implemented just below its ultrasound transducer. Finally, comparing the two presented LNA implementations in this work, the VA amplifier provided a remarkable improvement in terms of transimpedance gain and input-referred noise, achieving a higher signal-to-noise ratio. Nevertheless, the great dependency of the VA gain on the parasitic capacitance means that the CSA is the more useful option when the amplifier is not fully integrated with the PMUTs, since all of the parasitic capacitances are difficult to efficiently control, providing different gains for the same devices. The VA is more adequate when the amplifier can be directly connected to the sensor, as happens when there is a monolithic integration of the CMOS and PMUTs. Sensors 2020, 20, 1205 16 of 18 Table 3. Measured LNA performance summary and comparisons. Parameter This Work [13] (2019) [18] (2018) [16] (2016) [11] (2015) [17] (2015) Topology VA CSA VA CFVA TIA VA CFVA 0.13 µm HV 0.13 µm HV 0.18 µm 0.18 µm TSMC 0.18 µm 0.18 µm HV 0.18 µm Process technology CMOS CMOS CMOS HV-BCD HV CMOS CMOS CMOS Transducer PMUT PMUT PMUT PZT CMUT PMUT PZT Power supply (V) 1.5 1.5 1.5 1.8 1.8 1.8 1.8 Power consumption 0.3 0.3 0.08 0.79 1.4 N/A 0.135 (mW) Area (10−4 mm2 ) 6 9 N/A 30 1 280 310 60 Voltage–voltage 2 21.8 N/A 29/30/42/53 18 N/A N/A −12/6/24 gain (dB) Transimpedance 116/113.5 109.22 at 3 MHz 99.57 at 3 MHz N/A N/A N/A N/A gain (dBΩ) 110/104 10.2/10.8 Bandwidth (MHz) 22 N/A 10 20 N/A 9.8 10.6/10.5 Input current √ noise 0.08 at 3 MHz 0.15 at 3 MHz N/A N/A 0.41 @ 5 MHz N/A N/A (pA/ Hz) Input voltage √ noise 7.1 at 3 MHz N/A N/A 7.9 at 5 MHz N/A 11 at 0.22 MHz 5.9 at 4 MHz (nV/ Hz) Input dynamic 69 71 90 75 N/A N/A 81 range (dB) FOMRX_1 16674 N/A N/A 949 N/A N/A 6633 3 (MHz/V2 Aµm2 ) FOMRX_2 N/A 4.1*109 N/A N/A 0.09*109 3 N/A N/A (Hz/mA3 µm2 ) 1This area was estimated from a chip micrograph. 2 Including a TGC amplifier as a second stage. 3 Computed considering its higher gain. 6. Conclusions A CMOS circuit was designed and implemented using the novel CMOS-MEMS monolithic integration platform from Silterra in order to drive a 6 × 6 PMUT array. The system presented with competitive performance in comparison to state-of-the-art highlighting area and noise performance. A comprehensive comparison between two RX front-end amplifiers was done, showing that the VA amplifier based on capacitive integration provided remarkable improvements in general terms but was more sensitive to parasitic elements, which will be controlled with monolithic integration. This system was successfully applied in imaging experiments. Author Contributions: conceptualization, I.Z., A.U., and N.B.; methodology, I.Z.; electrical experiments, I.Z. and A.U.; acoustic experiments, E.L.; writing—original draft preparation, I.Z.; writing—review and editing, A.U. and N.B.; supervision, A.U. and N.B.; project administration, N.B. All authors have read and agreed to the published version of the manuscript. Acknowledgments: This work has been partially supported by project TEC2015-66337-R (MINECO/ FEDER). Special acknowledgements are made to E.M.F., C.B.F., and all the other members of the SilTerra’s MEMS & SENSORS technology development team for support in fabricating the wafers. Conflicts of Interest: The authors declare no conflicts of interest. References 1. Janjic, J.; Tan, M.; Daeichin, V.; Noothout, E.; Chen, C.; Chen, Z.; Chang, Z.Y.; Beurskens, R.H.S.H.; van Soest, G.; van der Steen, A.F.W.; et al. A 2D Ultrasound Transducer with Front-End ASIC and Low Cable Count for 3D Forward-Looking Intravascular Imaging: Performance and Characterization. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2018, 65, 1832–1844. [CrossRef] [PubMed] 2. Kim, H.; Kim, S.; Lee, S.; Kang, B.B.; Jo, Y.; Kook, G.; Oh, C.; Kim, M.K.; Kim, H.; Lee, H.J. MEMS Transducers for Non-Invasive Ultrasound Brain Stimulation. In Proceedings of the 2019 20th International Conference on Solid-State Sensors, Actuators and Microsystems & Eurosensors XXXIII (TRANSDUCERS & EUROSENSORS XXXIII), Berlin, Germany, 23–27 June 2019; pp. 242–245. Sensors 2020, 20, 1205 17 of 18 3. Tang, H.-Y.; Lu, Y.; Jiang, X.; Ng, E.J.; Tsai, J.M.; Horsley, D.A.; Boser, B.E. 3-D Ultrasonic Fingerprint Sensor-on-a-Chip. IEEE J. Solid-State Circuits 2016, 51, 2522–2533. [CrossRef] 4. Przybyla, R.J.; Tang, H.-Y.; Shelton, S.E.; Horsley, D.A.; Boser, B.E. 12.1 3D Ultrasonic Gesture Recognition. In Proceedings of the 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC), San Francisco, CA, USA, 9–13 February 2014; pp. 210–211. 5. Qiu, Y.; Gigliotti, J.V.; Wallace, M.; Griggio, F.; Demore, C.E.M.; Cochran, S.; Trolier-McKinstry, S. Piezoelectric micromachined ultrasound transducer (PMUT) arrays for integrated sensing, actuation and imaging. Sensors 2015, 15, 8020–8041. [CrossRef] [PubMed] 6. Jiang, X.; Tang, H.-Y.; Lu, Y.; Ng, E.J.; Tsai, J.M.; Boser, B.E.; Horsley, D.A. Ultrasonic Fingerprint Sensor with Transmit Beamforming Based on a PMUT Array Bonded to CMOS Circuitry. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2017, 64, 1401–1408. [CrossRef] [PubMed] 7. Lu, Y.; Tang, H.; Fung, S.; Wang, Q.; Tsai, J.M.; Daneman, M.; Boser, B.E.; Horsley, D.A. Ultrasonic fingerprint sensor using a piezoelectric micromachined ultrasonic transducer array integrated with complementary metal oxide semiconductor electronics. Appl. Phys. Lett. 2015, 106, 263503. [CrossRef] 8. Jiang, X.; Lu, Y.; Tang, H.Y.; Tsai, J.M.; Ng, E.J.; Daneman, M.J.; Boser, B.E.; Horsley, D.A. Monolithic ultrasound fingerprint sensor. Microsyst. Nanoeng. 2017, 3, 1–8. [CrossRef] [PubMed] 9. Pandian, M.S.; Ferrer, E.M.; Tay, W.S.; Madhaven, V.; Kantimahanti, A.K.; Sobreviela, G.; Uranga, A.; Barniol, N. Thin Film Piezoelectric Devices Integrated on CMOS. In Proceedings of the 2016 Symposium on Piezoelectricity, Acoustic Waves and Device Applications, SPAWDA 2016, Xi’an, China, 21–24 October 2016. 10. Ledesma, E.; Zamora, I.; Torres, F.; Uranga, A.; Tzanov, V.; Bhd, S. ALN Piezoelectric Micromachined Ultrasonic Transducer Array Monolothically Fabricated on Top of Pre-Processed CMOS Substrates. In Proceedings of the 2019 20th International Conference on Solid-State Sensors, Actuators and Microsystems & Eurosensors XXXIII (TRANSDUCERS & EUROSENSORS XXXIII), Berlin, Germany, 23–27 June 2019; pp. 655–658. 11. Przybyla, R.J.; Tang, H.-Y.; Guedes, A.; Shelton, S.E.; Horsley, D.A.; Boser, B.E. 3D Ultrasonic Rangefinder on a Chip. IEEE J. Solid-State Circuits 2015, 50, 320–334. [CrossRef] 12. Chen, K.; Lee, H.-S.; Chandrakasan, A.P.; Sodini, C.G. Ultrasonic Imaging Transceiver Design for CMUT: A Three-Level 30-Vpp Pulse-Shaping Pulser With Improved Efficiency and a Noise-Optimized Receiver. IEEE J. Solid-State Circuits 2013, 48, 2734–2745. [CrossRef] 13. Lee, J.; Lee, K.; Eovino, B.E.; Park, J.H.; Lin, L.; Yoo, H.; Yoo, J. 11.1 A 5.37mW/Channel Pitch-Matched Ultrasound ASIC with Dynamic-Bit-Shared SAR ADC and 13.2V Charge-Recycling TX in Standard CMOS for Intracardiac Echocardiography. In Proceedings of the IEEE International Solid-State Circuits Conference, San Francisco, CA, USA, 17–21 February 2019; pp. 2018–2020. 14. Tang, H.-Y.; Seo, D.; Singhal, U.; Li, X.; Maharbiz, M.M.; Alon, E.; Boser, B.E. Miniaturizing Ultrasonic System for Portable Health Care and Fitness. IEEE Trans. Biomed. Circuits Syst. 2015, 9, 767–776. [CrossRef] [PubMed] 15. Zamora, I.; Ledesma, E.; Uranga, A.; Barniol, N. Fully Integrated CMOS-PMUT Transceiver. In Proceedings of the 2018 25th IEEE International Conference on Electronics Circuits and Systems, Bordeaux, France, 9–12 December 2018; pp. 149–152. 16. Chen, K.; Lee, H.S.; Sodini, C.G. A Column-Row-Parallel ASIC Architecture for 3-D Portable Medical Ultrasonic Imaging. IEEE J. Solid-State Circuits 2016, 51, 738–751. 17. Chen, C.; Chen, Z.; Chang, Z.Y.; Pertijs, M.A.P. A Compact 0.135-mW/Channel LNA Array for Piezoelectric Ultrasound Transducers. In Proceedings of the ESSCIRC Conference 2015—41st European Solid-State Circuits Conference (ESSCIRC), Graz, Austria, 14–18 September 2015. 18. Kang, E.; Ding, Q.; Shabanimotlagh, M.; Kruizinga, P.; Chang, Z.Y.; Noothout, E.; Vos, H.J.; Bosch, J.G.; Verweij, M.D.; De Jong, N.; et al. A Reconfigurable Ultrasound Transceiver ASIC with 24 × 40 Elements for 3-D Carotid Artery Imaging. IEEE J. Solid-State Circuits 2018, 53, 2065–2075. [CrossRef] 19. Tang, H.; Lu, Y.; Fung, S.; Tsai, J.M.; Daneman, M.; Horsley, D.A.; Boser, B.E. Pulse-Echo Ultrasonic Fingerprint Sensor on a Chip. In Proceedings of the 2015 Transducers—2015 18th International Conference on Solid-State Sensors, Actuators and Microsystems (TRANSDUCERS), Anchorage, AK, USA, 21–25 June 2015; pp. 674–677. 20. Riverola, M.; Sobreviela, G.; Torres, F.; Uranga, A.; Barniol, N. Single-resonator dual-frequency BEOL-embedded CMOS-MEMS oscillator with low-power and ultra-compact TIA core. IEEE Electron Device Lett. 2017, 38, 273–276. [CrossRef] Sensors 2020, 20, 1205 18 of 18 21. Sautto, M.; Savoia, A.S.; Quaglia, F.; Caliano, G.; Mazzanti, A. A comparative analysis of CMUT receiving architectures for the design optimization of integrated transceiver front ends. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2017, 64, 826–838. [CrossRef] [PubMed] © 2020 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/).