Wideband Linear Power Amplifier for High-Frequency ... - IEEE Xplore

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 4,

April

2012

825

Correspondence Wideband Linear Power Amplifier for High-Frequency Ultrasonic Coded Excitation Imaging Jinhyoung Park, Changhong Hu, Xiang Li, Qifa Zhou, and K. Kirk Shung Abstract—Linear power amplifiers are critical components in ultrasonic imaging systems that implement chirp-coded excitation. Bench-top commercial power amplifiers are usually used in academic laboratories for high-frequency ultrasound imaging, and the imaging performance depends greatly on these general-purpose instruments. To achieve a wide dynamic range, a power amplifier consisting of two stages is developed for chirp-coded ultrasound imaging applications through the implementation of custom-designed broadband 1:1 transformers and the optimization of feedback circuits. The amplifier has broad bandwidth (5 to 135 MHz), maintaining a linearity up to the 1-dB gain compression point (P1dB) of 41.5 dBm, allowing 16 dBm input power level at 60 MHz. The mean and the maximum values of output third-order intercept points (OIP3) are 51.8 and 53.5 dBm, respectively, between 20 and 110 MHz. With 12 dBm input power, the gain of the amplifier varies between 24 and 29 dB, offering a uniformity which would allow excitation of a 70-MHz single-element transducer with windowed chirp-coded bursts sweeping from 40 to 100 MHz. The performance in high-frequency ultrasound imaging is evaluated with a wire phantom. Echo signal-to-noise ratio (eSNR) of the designed amplifier is 7 dB better than a commercial amplifier, and spatial resolution is maintained.

I. Introduction

C

oded excitation that uses an elongated burst has long been known to be capable of enhancing the penetration depth of ultrasound while maintaining its spatial resolution. O’Donnell et al. [1] demonstrated an improvement of SNR by 15 dB with coded excitation over conventional pulse-echo ultrasound in the clinical ultrasound frequency range. There are two types of coded excitation approaches: phase modulation and frequency modulation. Chiao et al. [2] compared these two methods and discussed the advantages and the disadvantages of each approach. Although systems for phase modulation transmissions are simpler to implement utilizing switching circuits [3], the levels of range side lobes are affected by phase distortion during acoustic propagation in a nonlinear media. In contrast, frequency modulations, i.e., chirp-coded excitation, result in smaller range side lobes (30 MHz) imaging is more crucial than in lower frequency ultrasound ( 50 dBm], broad bandwidth (>130%, high-frequency cutoff is over 100 MHz) and high-voltage output [1-dB gain compression point (P1dB) > 41 dBm] with a flat gain distribution over a bandwidth. To the best of our knowledge, such devices do not yet exist in the market. In this paper, the design and the fabrication of an AWA for high-frequency chirp-coded excitation imaging is reported. The previously stated requirements are achieved by implementing custom-designed broadband 1:1 transformers and optimizing negative-feedback networks. The performance of the designed AWA is assessed in terms of its bandwidth, linearity, and low-noise operation. A wire target is imaged by integrating the proposed AWA into a high-frequency ultrasound imaging system, called an ultrasound biomicroscope (UBM) [9], demonstrating excellent spatial resolution and echo signal-to-noise ratio (eSNR) using a 70MHz single-element transducer. III. Implementation The AWA consists of two stages; Fig. 1 shows a schematic diagram of the implemented circuit and the parts

© 2012 IEEE

826


vol. 59, no. 4,

April

2012

Fig. 1. Schematics of broadband high-frequency linear arbitrary waveform amplifier and the component list used for fabrication.

list. The first stage is a class-A amplifier intended to amplify an input signal with the least distortion. This stage drives a broadband signal to the second stage, adapting a custom-designed 1:1 transformer. The second stage is a class-AB amplifier. A push-pull circuit topology maximizes the amplification rate of the linear power amplifier and the final output is coupled with a center-tapped 1:1 transformer. In each stage, a negative-feedback circuit is placed to flatten gain distribution over 100 MHz and a phase margin of a simplified transfer function is designed to be over 90° for a stable operation. A. Class-A Amplifier An N-type RF MOSFET (Q1, MRF134, Motorola, Schaumburg, IL) is an active component in the first stage. A series connection of a resistor (R1), an inductor (L2), and a capacitor (C1) located between the drain and the gate of the transistor is employed for negative-feedback. An inherent gate-drain capacitance (Cgd) between the gate and the drain in the RF MOSFET couples with the feedback circuit to flatten gain distributions. Using Miller’s theorem [10], the connection between the gate and the drain can be broken and the feedback reactance divided by a constant gain is placed at a position parallel with the input port. The reactance can be calculated by

X(s) =

A(sL2 + R1) , (1) (s C gdL2 + sC gdR1 + 1) 2

where X(s) is reactance and A is the constant gain of the amplifier. Because the input resistance networks (network of input source impedance and R13, R14, and R15 in Fig. 1) are connected with X(s) in series, a final gain function can be estimated by (2). Note that C1 blocks dc current from the drain to the gate and its value (6.5 μF) is large enough to be ignored. H (s) =

X(s) X(s) + R s

(sL2 + R1) , [s 2AR sC gdL2 + s(AR sC gdR1 + L2) + AR s + R1] (2) =

where Rs is the input resistance. In (1), the locations of zero and poles are decided by the L2 and R1 in the feedback. Zero at 120 MHz and both poles at 196 MHz can be found if L2 and R1 are 0.38 μH and 287 Ω, respectively. Therefore, the gain of the amplifier remains flat until 120 MHz. The phase margin for H(s) is kept over 90°, which results in a stable condition of the amplifier with negative-feedback. The inductors L2, L3, and L4 are fabricated by winding 18-gauge enamel magnet wire (2701MG18, Electronix Express, Rahway, NJ), having a low resistance (less than 0.0007 Ω/cm), around a plastic tube of 9 mm diameter; the plastic tube is then removed to form an air-coupled inductor. The value of the inductor is roughly estimated from (3), but the optimal winding number is found by

park et al.: power amplifier for high-frequency ultrasonic coded excitation imaging

changing one or two turns around the calculated value while observing the output signal of the amplifier.

L=

d 2n 2 , (3) 0.0254(18d + 40l )

where L is inductance in microhenries, d is the coil diameter in meters, l is the coil length in meters, and n is the number of turns. The inductances for L2 and L3 (L4) in the implemented circuit are 0.36 and 0.28 μH, respectively. DC voltages at gates of RF MOSFETs (Vg) decide bias points of input signals. Biasing in the first stage is accomplished through a 1-kΩ trimmer (R14/R15) after a 15-VDC (Vdc) regulator (LM340T-15, National Semiconductor, Santa Clara, CA). The Vg is set to 4.8 Vdc to allow an input voltage level as high as 4 Vpp with minimized distortions. After optimization of the feedback circuit and bias, a broadband 1:1 transformer (T1) is coupled with the first stage’s output signals to allow them to be transferred to the second stage. B. Wideband Transformer Fabrication Transformers carry different coupling efficiencies between the primary and the secondary coils depending on the utilized frequency range [11]. Fig. 2 shows transformer equivalent circuits in low-frequency [Fig. 2(a)] and highfrequency ranges [Fig. 2(b)]. In the low-frequency range, inductance in the primary coil takes the major role in transferring a signal to the secondary, and the higher number of turns around a core yields the higher signal transfer efficiency. In the implemented circuit (Fig. 1), the output resistance (23 Ω for MRF134) of Q1 is serially connected to the primary port of T1 [11]. Therefore, the voltage level loaded on the primary coil can also be increased by using a core having high permeability or wrapping more turns. The inductance can be calculated by [12]

L = 0.4πN p µ r

Ae −8 10 , IM = 2π fL, (4) Ie

where L is the inductance of the coil in henries, Np is the number of turns, μr is the relative permeability of a core, Ae is the effective cross-sectional area of the core in square centimeters, Ie is the average magnetic path length in the core in centimeters, IM is the impedance of the core, and f is the frequency. The relative permeability (μr) of the utilized nickel-zinc (NiZn) toroid core (5943001201, Fair-Rite, Wallkill, NY) at 1 MHz is 750, which results in 53 µH from (4) for eight turns of an enamel magnet wire (2700MG24, Electronix Express) [13]. At 1 MHz, the magnitude of inductive impedance is 330 Ω, which is relatively higher than the output resistance of Q1. In contrast, high-frequency response is affected by winding capacitance, formed between the wires and core’s surfaces, and leakage inductance of a transformer. The estimated values of the capacitance and the leakage inductance can be calculated by [11]

Cw =

(0.225Aε) , 3t

827

LS =

10.6N 2(MT)(4c + a) × 2, (5) 4b × 10 9

where Cw is the winding capacitance in picofarads, A is the area of winding element in square inches, ε is the dielectric constant of insulation under winding (usually 3 to 4 for organic materials), t is the thickness in inches of insulation under winding, N is turns in winding, MT is the mean length of turn, a is the total winding height, b is the winding width, and c is the insulation space. Note that leakage inductance and the winding capacitance are located in series and in parallel to the mutually coupled transformer windings as shown in Fig. 2(b). The calculated values of Cw and LS for the implemented transformer are 3.23 pF and 0.76 µH, respectively, resulting in 101 MHz resonance frequency. The bandwidth of the fabricated transformer is measured by inputting a 5-Vpp continuous wave generated by a function generator (AFG3251, Tektronix Inc., Beaverton, OR) to the transformer. The output voltage level is recorded by an oscilloscope in a single-channel mode (9350AL, Le Croy Corp., Chestnut Ridge, NY) at 10 MHz increments from 1 to 150 MHz. Fig. 3 shows the result of the measurement. The measured −3-dB bandwidth is from 1 to 90 MHz. C. Class-AB Amplifier The second stage is a class-AB amplifier. A transferred signal from the first stage is separated into positive and negative signals by two diode (D7, D8) networks, and a push-pull circuit maximizes the amplification rate of the amplifier with two identical N-type RF MOSFETs (Q2, Q3; MRF148, Motorola). Negative-feedback circuits (C8, L3, R10, C9, L4, R11) are also placed between the drain and the gate of RF MOSFETs. Based on (2), a bandwidth of 114 MHz can be achieved with L3 (L4) and R10 (R11) being 0.28 µH and 220 Ω respectively. Bias voltage level is set to 3.2 Vdc, close to the operational threshold voltage (3 Vdc), with a 1-kΩ trimmer (R17/R18) to amplify positive or negative signals separately in the whole active operational region of the RF MOSFET [14].

Fig. 2. Transformer equivalent circuit at (a) low frequency and (b) high frequency. R1 is the output resistance of a source, R2 is the loading impedance in the secondary coil, XN is the primary coil inductance, XL is the leakage inductance, and XC is the winding capacitance.

828


Fig. 3. Power loss of the fabricated transformers between the primary and the secondary coil. T1 and T2 are 1:1 transformers in the middle and final port of the arbitrary waveform amplifier, respectively.

D. Output Transformer The final output in the second stage is coupled with a load utilizing the same transformer with T1. However, the output transformer (T2) is tapped in the center of the primary coil to convert the phase of negative signals amplified in Q3 by 180°, and the separately amplified signals in the push-pull topology can be merged at the final output load. III. Measurement A. Electrical Performance Measurement To measure the bandwidth of the AWA, a 250 kHz to 40 GHz signal generator (E8257C, Agilent Technologies Inc., Santa Clara, CA) is used as the input signal, the frequency of which is changed from 1 to 150 MHz in 5-MHz steps at different power levels of 0, 6, 12, and 16 dBm. The output peak power is measured by a spectrum analyzer (8596E, Hewlett Packard Co., Palo Alto, CA). A 30-dB attenuator (HAT-30+, Mini-Circuits, Brooklyn, NY), a 20-dB attenuator (HAT-20+, Mini-Circuits), and a 15-dB attenuator (HAT-15+, Mini-Circuits) are serially connected between the input port of the spectrum analyzer and the output of the AWA. To verify the performance of the attenuator, the attenuation variance under the measurement condition is tested. The bandwidth measurements are repeated, connecting the output of the signal generator directly to the attenuators. The measured attenuation is constant at 57 dB, independent of the given frequency and the power levels. The gain (in decibels) at a frequency is acquired by

gain = OE + 57 − IE, (6)

where OE is the read power in the spectrum analyzer and IE is the input power level. The flexibility of the AWA is demonstrated in Fig. 4 showing 6 different types of waveforms: a 60-MHz single-cycle, a four-cycle, an amplitude inverted four-cycle, an eight-cycle sine burst, 13-bit Barker code [15], and Hanning-windowed chirp acquired by

(

vol. 59, no. 4,

April

2012

)

αt   chirp(t) = cos  2π f i + t  ×W Hanning, (7)  2 

where fi is the lowest frequency of the chirp, α is the rate of the frequency increase, t is time, and WHanning is the Hanning window. Note that the frequency is linearly modulated from 35 to 105 MHz. The signals generated using Matlab (Matlab 7.0, The MathWorks Inc., Natick, MA) are loaded in a function generator (AFG3251, Tektronix Inc.) at a 2 GHz sampling rate to minimize the aliasing artifact. The function generator gives an input signal to the AWA. Output signals from the AWA are recorded by an oscilloscope (9350AL, Le Croy Corp.) through a 30-dB attenuator (HAT-30+, Mini-Circuits). Phase distortion for the chirped output is evaluated by subtracting the input phase from the output phase and is quantified, in percent, by

distortion =

abs(IP − OP) , (8) π(NS)

where IP is the phase of the input signal, OP is the phase of the output signal, and NS is the number of samples. For evaluations of linearity of the AWA, two parameters, P1dB and OIP3 are measured. The P1dB is measured at 60 MHz, which is the center frequency of the AWA. Long-duration sine bursts of 400 cycles are used to drive the AWA while changing the input voltage from 0.6 to 5.0 Vpp in 0.2-Vpp increments. The output Vpp after the attenuator is recorded to measure the P1dB, which can be found by drawing the gain curve compared with the input voltage level. OIP3 can be measured by two-tone test [16]. Using a two-channel function generator (AFG3252, Tektronix Inc.), two different continuous signals with 10 MHz difference in the center frequency are summed with a T-connector (413592–6, Tyco Electronics Corp., Berwyn, PA). The summed signal becomes an input to AWA and the output signal is recorded by an oscilloscope (9350AL, Le Croy Corp.) with 30-dB attenuator (HAT-30+, Mini-Circuits). The input voltage at each frequency for the two-tone test is 3.6 Vpp. Power levels of third-order intermodulation distortions (IP3) and fundamental signals are recorded by reading the power spectrum at corresponding frequency. Note that the intermodulation frequency is 2f1 − f2 and 2f2 − f1 if the fundamental frequency values used for the two-tones are f1 and f2. Then, OIP3 can be estimated by

OIP3 = FP +

(FP − IP3) , (9) 2

where FP is the power value of the fundamental signal. OIP3 is measured at every 10 MHz increase from 20 to 110 MHz. SNR is predicted at P1dB with a 400-cycle sine pulse at 60 MHz center frequency. To reduce the estimation error, the measurement has been repeated 10 times under the


829

Fig. 4. Output of designed arbitrary waveform amplifier using input waveform of (a) 60-MHz one-cycle sine burst (b) 60-MHz 4-cycle sine burst, (c) 60-MHz phase-inverted 4-cycle sine burst, (d) 60-MHz 8-cycle sine burst, (e) 60-MHz 13-bit Barker code, (f) Hanning-windowed chirp with frequency from 35 to 105 MHz, and (g) its spectrum. Dashed lines are input and solid lines are output signals, referring to the vertical scale on the right and on the left, respectively.

same experimental conditions and the results are averaged. SNR is quantified by

 ∑10 max(abs(e(i)))   , (10) SNR = 20log 10  i =1 10 ∑i =1 σ(N (i))  

where σ is a standard deviation, e(i) is the signal at the ith trial, N(i) is the noise at the ith trial, and abs is the absolute value function. The noise level of the AWA is measured separately without the attenuator. The oscilloscope’s vertical resolution is 8 bit. Therefore, the noise level with the attenuator is below the minimum value of the resolution of the oscilloscope. At the same setting without the attenuator and with the output of the function generator off, the output noise level of the AWA is measured. The SNR is measured for both the designed AWA and the commercial AWA. When the SNR of a commercial AWA is estimated with an input voltage of 220 mV, which is the

rated input to the AWA, the gain is 50 dB, producing a voltage of 75 Vpp. B. Ultrasound Performance Measurement The implemented AWA is integrated into a UBM performing as a pulse generator with a function generator (AFG3251, Tektronix Inc.) for evaluation. Fig. 5 shows the block diagram of the UBM. In this system, the protection circuit [17] is a fast diode (PMBD7000, NXP Semiconductors, Eindhoven, The Netherlands) which has a recovery time of less than 4 ns to minimize signal distortion. For chirp-coded excitation imaging, a 1-µs-long Hanning-windowed chirp is created using Matlab (Matlab 7.0) and the waveform is loaded into the function generator. The frequency range of the windowed chirp is from 40 to 100 MHz. The created waveform is amplified by a power amplifier and excites a 70-MHz LiNbO3 single element transducer having a focal distance of 5.36 mm, f-number

830


vol. 59, no. 4,

April

2012

Fig. 6. Gain distribution of the implemented arbitrary waveform amplifier. The gain values are measured for the different input power levels of 0 dBm (diamond), 6 dBm (square), 12 dBm (triangle), and 16 dBm (×). Fig. 5. Block diagram of the implemented linear ultrasound biomicroscope (UBM). The fabricated arbitrary waveform amplifier is integrated into the UBM as a part of the pulse generator.

of 2, and 66% bandwidth. For a comparison, an image using a commercial AWA (75A250A, Amplifier Research Corp., Souderton, PA) is acquired and the performance is compared with the implemented AWA by scanning wire targets. In addition to chirp-coded excitation imaging, images using a 70-MHz single-cycle sine burst are acquired and the result obtained with the custom-designed AWA is also compared with the commercial AWA. The short burst or the windowed chirp code burst is transmitted at a 1  kHz pulse repetition rate while the transducer is moving over a wire target at a speed of 10 cm/s. The diameter of the tungsten wire target is 20 μm and three wires are diagonally aligned in depth. The second wire among the three is located at the physical focal distance of the single element transducer, which is 5.36 mm. The echoes from the targets are digitized by an analogto-digital converter (ADC; CS12400, GaGe Applied Technologies Inc., Lachine, QC, Canada) at 400 MHz sampling rate. The software developed using LabView (LabView 5.0, National Instruments Corp., Austin, TX) postprocesses and displays the echo in decibel scale. When the transmission pulse is a windowed chirp, compression with a matched filter which is the time-inverted transmitted waveform [18] is performed during postprocessing. After the image is acquired, echo signal-to-noise ratio (eSNR) is calculated by [19]

 max(abs(1 frame of signal))  eSNR = 20 × log 10   , (11) σb  

where σ b is a standard deviation of background noise. To measure the eSNR, a noise frame is acquired using a 0-V transmission.

IV. Results Fig. 6 shows the gain distribution over the frequency range from 1 to 150 MHz at different input power levels of 0, 6, 12, and 16 dBm. At 16 dBm input, the −3-dB band-

width is 185% from 5 to 135 MHz, excluding a region from 35 to 45 MHz where the gain overshoots, and the difference of the gain between the maximum and the minimum is 5 dB. Because of the high cutoff frequency adjusted by the resistance value in the feedback networks (R1, R10, R11) and the designed transformers, the gain values in the mid-band region are slightly reduced as a consequence of the broadened bandwidth. Fig. 7 shows linearity parameters of P1dB and OIP3. In Fig. 7(a), the solid line is the voltage gain at 60 MHz and the dotted line is the output power level. The gain drops from 26 to 25 dB at the power level of 41.5 dBm, which is P1dB, maintaining a linear amplification rate. The output voltage level is 75 Vpp, and the output power of the AWA can be estimated to be 13 W. At P1dB, SNR of the output signal is 83.8 dB and the standard deviation of the white noise (σnoise) is only 2.2 mV. In contrast, these values for the commercial amplifier are 61.5 dB and 29.2 mV, respectively. Note that another commercial amplifier (ENI325LA, Electronics and Innovation Ltd., Rochester, NY), which can perform 41.5 dBm of P1dB, also has a similar noise level (>20 mV) as 75A250A. In Fig. 7(b), the value of OIP3 at 60 MHz is 53.5 dBm and the average value 51.8 dBm, which means the harmonic power level becomes equivalent to the fundamental when the output voltage level is 220 Vpp. Therefore, the developed AWA is hardly distorted by harmonic signals for an output power level under P1dB. Power consumption of the AWA is 32 W caused by static dc currents of 800 mA with 40 Vdc. For this reason, an aluminum plate is attached to the transistors as a heat sink. Fig. 4 demonstrates outputs from the designed AWA with 6 types of inputs generated in the function generator. Dashed lines are input signals and solid lines are output signals. Fig. 4(a) is the output signal for a 60-MHz singlecycle sine burst. Figs. 4(b) and 4(c) show the 4-cycle sine burst and its 180° phase-inverted waveform for tissue harmonic ultrasound imaging. In the frequency domain, the bandwidths of the two outputs change by 0.05% compared with the input signals. Therefore, the second-harmonic signal can be isolated by blocking the fundamental frequency utilizing these output signals. Fig. 4(d) is for an 8-cycle sine pulse for a pulsed Doppler application. The amplitude of each cycle is almost the same, allowing the


831

Fig. 7. Measurement of linearity specification. (a) Gain change (solid line) over the increased input power, and output power increase (dotted line) at 60 MHz. (b) Output third-order intercept point distribution from 20 to 110 MHz.

detection of the Doppler signal uniformly within a Doppler gate. In the single- or multi-cycle study, positive peak voltage levels are 10 V smaller than the levels in the negative side. This may be caused by a clipped input resulting from the limited bias voltage although the measured output is at P1dB. For coded excitation waveform types, the input voltage level is reduced to 12.9 dBm to prevent clipping in the first stage. Fig. 4(e) is the 13-bit Barker code. Each bit of the code is represented by a 2-cycle 60-MHz sine pulse. Because of the reduced input voltage level, the positive and the negative amplitudes are the same. When the output signal is compressed with a matched filter, the range side lobe level is −21.34 dB, which is 0.9 dB higher than the theoretical value. Fig. 4(f) is for a 5-μs long wideband Hanning-windowed chirp code, sweeping the frequency from 35 to 105 MHz, for driving a 70-MHz high-frequency ultrasound transducer, and Fig. 4(g) is its spectrum. The input and the output signals are overlapped in the time and the frequency domain with minimal phase distortions (4%). The −6-dB bandwidth of the output is 50% and the value is same with the ideal bandwidth of the Hanning

Fig. 8. Images acquired by a windowed chirp (a) with the developed arbitrary waveform amplifier (AWA), (b) with the commercial AWA, (c) single-cycle sine burst with the developed AWA, and (d) with the commercial AWA. The displaying dynamic range is 70 dB.

windowed chirp. In the spectrum over 100 MHz, an unexpected signal which may be caused by an aliasing effect or a harmonic distortion of the designed AWA is observed. The sampling rate of the digital equipment is over 1 GHz and the 3-dB noise bandwidth of the function generator is 250 MHz. Therefore, it is likely that the source of this high-frequency signal is from the harmonic distortion of the AWA. A wire target measurement gives an estimation of spatial resolution and eSNR of images acquired by the UBM with the custom-designed AWA or the commercial AWA. Fig. 8 shows wire target images acquired by chirp-coded excitation [Figs. 8(a) and 8(b)] and 70-MHz single sine burst [Figs. 8(c) and 8(d)]. The images acquired by the

TABLE I. Spatial Resolution and Echo Signal-to-Noise Ratio (eSNR) in the Wire Target Images. Designed AWA Chirp eSNR (dB) Axial resolution (μm) Lateral resolution (μm)

65.7 28.8 (30) 40 (43)

70 MHz singlecycle sine burst 49.3 30.8 (21) 40.0 (43)

Commercial AWA Chirp 58.4 32.7 (30) 50 (43)

AWA = arbitrary waveform amplifier. Theoretical values are provided in parentheses.

70 MHz singlecycle sine burst 44.6 23.1 (21) 40 (43)

832


implemented AWA [Figs. 8(a) and 8(c)] compare favorably with the images acquired by the commercial AWA [Figs. 8(b) and 8(d)]. Note that the dynamic display range of each image is 70 dB and the output voltage level of each transmission is maintained as 70 Vpp for a fair comparison. The gray level in the background of Fig. 8(a) is the highest compared with the others. The eSNR of the image using chirp-coded excitation with the developed AWA is 65.7 dB, close to the maximum (66.2 dB) performance of a 12-bit ADC, and the value is 7 and 21 dB higher than the chirp excitation and single-cycle burst, respectively, using the commercial product. The spatial resolution values between the data by the custom-designed and the commercial AWA do not have significant difference, and are close to the theoretical values given in Table I, where the theoretical values are provided in parentheses. V. Conclusion A broadband, low-noise linear arbitrary waveform amplifier for high-frequency ultrasound imaging via chirpcoded excitation has been designed and implemented. A custom-designed broadband transformer and a negativefeedback circuit expanded the operational bandwidth up to 185% while maintaining linear amplification in lownoise operation. Although the gain fluctuation of 5 dB within the operational bandwidth exists, the designed AWA yielded comparable spatial resolution to a commercial product in the wire target imaging study while a better eSNR is achieved. Therefore, this amplifier can be used not only as an independent system but also as a pulse generator module which allows chirp-coded excitation of imaging transducers at a center frequency up to 100 MHz. References [1] M. O’Donnell, “Coded excitation system for improving the penetration of real-time phased array imaging systems,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 39, no. 3, pp. 341–351, 1992.

vol. 59, no. 4,

April

2012

[2] R. Y. Chiao and H. Xiaohui, “Coded excitation for diagnostic ultrasound: A system developer’s perspective,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 52, no. 2, pp. 160–170, 2005. [3] X. Xu, J. T. Yen, and K. K. Shung, “A low-cost bipolar pulse generator for high-frequency ultrasound applications,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 54, no. 2, pp. 443–447, 2007. [4] K. K. Shung, Diagnostic Ultrasound: Imaging and Blood Flow Measurements. Boca Raton, FL: Taylor & Francis, 2006. [5] C. H. Hu, R. Liu, Q. Zhou, J. Yen, and K. K. Shung, “Coded excitation using biphase coded pulse with mismatched filters for high frequency ultrasound imaging,” Ultrasonics, vol. 44, no. 3, pp. 330–336, 2006. [6] J. Mamou and J. A. Ketterling, “Chirp coded excitation imaging with a high-frequency ultrasound annular array,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 55, no. 2, pp. 508–513, 2008. [7] C. Passmann and H. Ermert, “A 100-MHz ultrasound imaging system for dermatologic and ophthalmologic diagnostics,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 43, no. 4, pp. 545–552, 1996. [8] M. R. Bosisio, J. M. Hasquenoph, L. Sandrin, P. Laugier, S. L. Bridal, and S. Yon, “Real-time chirp coded imaging with a programmable ultrasound biomicroscope,” IEEE Trans. Biomed. Eng., vol. 57, no. 3, pp. 654–664, 2010. [9] F. S. Foster, C. J. Palvlin, K. A. Harasiewicz, D. A. Christopher, and D. H. Turnbull, “Advances in ultrasound biomicroscopy,” Ultrasound Med. Biol., vol. 26, no. 1, pp. 1–27, 2000. [10] A. A. Sedra and K. C. Smith, Microelectronic Circuits, 4th ed., Oxford, UK: Oxford University Press, 1998, pp. 583–619. [11] R. Lee, W. Leo, and C. E. Carter, Electronic Transformers and Circuits, 3rd ed., New York, NY: Wiley, 1988. [12] J. Sevick, Transmission Line Transformers, 2nd ed., Newington, CT: The American Radio Relay League, 1990. [13] C. Trask, “Designing wide-band transformer for HF and VHF power amplifiers,” QEX, pp. 3–15, Mar./Apr. 2005. [14] C. A. Schuler, Electronics, 4th ed., New York, NY: McGraw-Hill, 1994, pp. 171–174. [15] R. H. Barker, “Group synchronization of binary digital systems,” in Communication Theory. W. Jackson, Ed. New York, NY: Academic, 1953, pp. 273–287. [16] D. Seremeta, “Accurate Measurement of LT5514 Third Order Intermodulation Products,” Linear Technology Corp., Milpitas, CA, Application Note 97, 2006. [17] J. K. Poulsen, “Low loss wideband protection circuit for high frequency ultrasound,” in IEEE Ultrasonics Symp., 1999, pp. 823–826. [18] T. Misaridis and J. A. Jensen, “Use of modulation excitation signals in medical ultrasound. Part I: Basic concepts and expected benefits,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 52, no. 2, pp. 177–191, 2005. [19] K. F. Üstüner and G. L. Holley, “Ultrasound imaging system performance assessment,” presented at the 2003 American Association of Physicists in Medicine Annu. Meeting, San Diego, CA, Aug. 2003.