You are on page 1of 4

Ultrasound Energy Harvesting System

for Deep Implanted-Medical-Devices (IMDs)
Francesco Mazzilli∗, Prakash E. Thoppay† , Vincent Praplan∗ , and Catherine Dehollain∗
Ecole Polytechnique F´ed´erale de Lausanne, RFIC group, Lausanne, 1015, Switzerland

Marvell, Etoy, 1163, Switzerland

CONTROL
UNIT

Abstract—It is desirable in active medical implants to
derive energy from external sources to charge a rechargeable
battery. In this paper we have developed a novel system
to transfer energy via ultrasound to a deep implanted
medical device. Hence, an external base station is designed
to transmitt energy and a 64-channel high-voltage driver
is proposed for a spherical transducer array. Moreover, a
shunt-C class-E power amplifier (PA) is employed as core
element for the driver, showing a drain efficiency (DE) of
71% and a power added efficiency (PAE) of 57% including
gate-driver switching loss. In addition, a cascaded of two
low-drop-out (LDO) regulators is used within the implanted
device to reduce rectifier ripple and to set the charge voltage
for the micro-energy cell to 4.1 V. The LDOs are implemented in a CMOS 0.18 µm high-voltage (HV) technology
and measurement along with simulated results are reported.

POWER
MANAGEMENT
UNIT

PA

2

PA
FPGA

IMPLANT
acoustic
waves

4

5

6
DC/DC

3

PA
d

1

VBATT

COUT

2

PA

L

PA
PA

64

EXT. ENVIRONMENT

SUBCUTANEOUS

PATIENT’s SKIN

Fig. 1.

Index Terms—CMOS, energy harvesting, implanted medical device, power management, healthcare, sensor networks.

Ultrasound energy harvesting system chain.

The paper discusses the class-E PA design as well as
power management details within the implanted device.
In Section II the ultrasound energy harvesting system is
proposed. In Section III an analysis and tuning methodology to calibrate 64-PA is described. In Section IV the
LDO architecture to improve power-supply-rejection-ratio
(PSRR) is shown. Lastly, the conclusions are given in
section V.

I. I NTRODUCTION
Requirements for active implanted medical devices
(IMDs) deeply introduced into the human body are ruled
by the implant size (1-3 cm3 ) and by the type of coupling
used to transmitt energy [1]. RF and magnetic induced
coupling are heavily attenuated in water, whereas ultrasound is a valid option to overcome this limit despite the
complexity to choose the carrier frequency [2].
The selection of the wavelength (λ) for delivering energy is guided by the transducer design. For instance a
spherical array transducer enables to focus the ultrasound
(US) beam without beamforming electronics. However, to
track the implant when it goes out-of-focus, each array
element needs a power amplifer (PA) to shape the acoustic
field. A good candidate in energy transmission is the classE power amplifier with a theoretical power efficiency up
to 100 % [3].
However, it is more challenging the choice of the CMOS
technology used for energy conversion in the implant. To
avoid oxide breakdown, HV transistors are employed as
the input available voltage can be expected to vary from
few millivolts to tens of volt. In medical ultrasound as the
carrier frequency is setted around 1 MHz, two series LDO
are used to further improve supply noise rejection at low
frequency.

978-1-4673-0219-7/12/$31.00 ©2012 IEEE

1

PA

II. U LTRASOUND W IRELESS P OWER T RANSFER
Figure 1 shows the block diagram of the US wireless
power transfer (WPT) system, a 64-elements spherical
transducer array is used for the control unit (CU) while a 6elements flat transducer array is employed for the implant.
The natural focus of the spherical array, point of maximum
intensity, is located at distance (d) of 11 cm and the focal
area measured at -6 dB is 2 x 5 mm2 . The implanted
transducer has a total active area of 5 x 10 mm2 (SAtot )
and the active area per element of the array is 5 x 1 mm2
(SA). Hence rectifiers should be connected to add the
current yielded by the elements of the array since the focal
area is smaller than SAtot ,. To boost the rectifier output
voltage an off-chip shunt inductor (L) is used to resonate
with the tranducer and the rectifier imaginary part.
For US-WPT a governing equation is nonexisting to
estimate the power link budget as function of frequency
selection and transducer design by considering all the
chain which includes the external PA, the attenuation

2865

while the remaining 63 % present a shift only of ±1 V from VGOAL . model CS Figure 4 represents the histogram of the number of amplifiers versus the measured output voltage before and after using the tuning methodology. Drain efficiency (eqn. Figure 5 shows the prototype efficiency versus the DC power supply (VCC ) while VDD is kept constant. 4. Xt = −j290Ω). eq. VOUT VOUT .0 for ΔV1 . VOUT IDS Ct CSHUNT CP LP Rt Fig. 1a) and power added efficiency (eqn. tuning. ZVS condition VDRAIN ZVS VDRAIN Increase CSHUNT I DS VCC VDRAIN = ZVS VOUT = VGOAL Decrease CS VDRAIN Figure 2 shows the shunt-C class-E power amplifier along with the driving amplifer and a transducer equivalent model represented by a parallel branch constituted by the capacitor (Ct ) in series with the resistor (Rt ). Number of amplifiers versus the output voltage w/ and w/o To evaluate the efficiency of the shunt-C class-E power amplifier a prototype is realized and the transducer is replaced by its series lumped equivalent model (Rt = 120Ω. Shunt-C class-E power amplifier with transducer equivalent 30 Number of amplifiers The parallel inductor LP resonates with the transducer at 1 MHz.0 . the following approach is used to design the power amplifier and to determine the input voltage range at the DC-toDC converter. The series inductor LS increases the power amplifier efficiency since the AC current through LP decreases and as consequence the iron loss decreases. model. the parallel capacitors CSHUN T and CP helps to achieve zero voltage switching (ZVS). henceforth in this section a tuning strategy is proposed.0 VGOAL . 64-PA are fabricated on standard substrate FR4 due to the lowfrequency requirement. A 37 % of the number of PAs reach VOUT = VGOAL = 19 Vpeak.VGOAL > ȴV2 Increase CS VGOAL – VOUT < ȴV1 Decrease CP VOUT VOUT < VGOAL START ȴV1 = ȴV1 + ȴV1. the series capacitor CS prevents any DC feedthrough. ΔV2.0 ȴV2 = ȴV2. Figure 3 shows the tuning methodology adopted to calibrate 64-PAs. POUT.RF PVCC POUT. Tuning methodology for the calibration of multiple shunt-C class-E power amplifer. ΔV2 are chosen within the range [0. two main loops can be distinguished to set respectively the output voltage (VOUT ) of the PA to the desired voltage (VGOAL ) and to achieve ZVS for best efficiency. A narrowband model is assumed for the piezoelectric (PZT) tranducers and the implant is always presumed at the spatial peak of maximum intensity. VOUT t 3.VOUT > ȴV1 III. Fig. 2. w/o tuning w/ tuning 25 VGOAL = 19 Vpeak 20 15 10 5 0 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 VOUT [Vp] A. Hence.1 ÷ 1] V and their values also dermine the final maximum distance between VOUT and VGOAL .0 VOUT .VGOAL < ȴV2 VOUT > VGOAL ȴV1 = ȴV1. 3. The shunt-C class-E power amplifier is designed with Advanced Design System (ADS) and an average value for Rt and Ct is considered. Tuning Methodology Design equations have been developed for the class-E power amplifier [4]. The initial conditions ΔV1.66VCC > ZVS RFC VDD VDRAIN 0 LS t < ZVS B.RF − PDY N P AE = PVCC DE = 2866 (1a) (1b) . Measurements trans. To set the value for the capacitors the magnitude |VOUT − VGOAL | is compared with ΔV1 or ΔV2 . S HUNT-C C LASS -E P OWER A MPLIFIER Decrease CP END 0 VDRAIN > ZVS VDRAIN VDRAIN < ZVS Fig.0 VOUT VGOAL Increase CP VOUT ȴV2 = ȴV2 + ȴV2. 1b) are measured for the power amplifier.in the body and the load of the implant.

By replacing equation 4a into equation 4b.RFtot [2].Where POUT. VOUT 1 and VBAT T can be expressed as follows: 35 2 m18 Fig. The US-WPT link efficiency can be expressed as the ratio between the received electrical power Pload and the delivered electrical power POUT. PSRR Analysis A methodology to analyze PSRR is proposed in [6] by mean of control-theory. T WO S TAGE LDO Figure 6 depicts the linear DC-to-DC converter that sets the charging voltage of the battery (see Figure 1).4 W.RFtot equals to 5. DE PAE m17 Rf2 V3 vSS vb2 R1 iOUT 60 55 Rf1 m9 V3 OTA2 vref vout1 M V1 m11 m7 vout1 vb3 70 Efficiency [%] m5 vref OTA1 75 Fig.RF is the delivered power to Rt . P SRRi are defined by P SRR1 = − AAp1 A3 A4 P SRR3 = − Ap3 and P SRR4 = Ap4 . The rectifier output voltage (Vrec ) is regulated to VOUT 1 = 4. To evaluate PSRR with respect to Vrec . The supply gains are written as Ap1 = V1 /Vrec . A2 (1 − Ap1 ) + Ap2 A1 A2 Vref + Vrec (3a) 1 + β 1 A1 A2 1 + β 1 A1 A2 A4 (1 − Ap3 ) + Ap4 A3 A4 = Vref + VOU T 1 (3b) 1 + β 2 A3 A4 1 + β 2 A3 A4 VOU T 1 = The 64-PAs are able to deliver an acoustic intensity (Iac ) in water of 23 W/cm2 at 11 cm with PVCC . Assuming a conjugate load matching at each element of the implanted tranducer ends.1 V.3 dB re.tot equals to 8 W and POUT. vrec vSS m10 m8 m1 m2 m3 m4 m16 65 m12 m13 CSHUNT = 550 pF Cp = 690 pF CS = 1 nF f0 = 1 MHz (DC = 50 %) Lp = LS = 68 µH 50 45 40 m14 3 4 5 6 VCC [V] m15 R2 BATTERY MODEL vBATT vb4 ESR Rself CBATT Linear DC-to-DC converter to improve PSRR. a cell equivalent series resistance of 180 Ω (ESR) and a self-discharge resistor (Rself ) of 10 MΩ representing the battery leakage. So. vb1 m6 7 8 9 10 Efficiency of the tuned shunt-C class-E power amplifier.RFtot VBAT T R1 Where the attenuation factors are defined as β1 = R1 +R f1 R2 and β2 = R2 +Rf 2 . Figure 7 represents an equivalent network of the two-series LDO to analyse PSRR at low frequency. The parameters 1 2 . [5]: Pload Iac × SA × ηae = (2) ηlink = POUT. VBAT T can be expressed Where IV. as follows: VBAT T = 1 β2   + 1 1+ β1 P SRRLDO2  Vref 1 P SRRLDO1 P SRRLDO2  Vrec β1  (5) At low frequency any variation in Vrec is further suppressed at VBAT T by using two LDOs in series as shown by the term (P SRRLDO1 × P SRRLDO2 )−1 .RFtot POUT. The micro-energy cell is represented by its Thevenin equivalent model with an equivalent capacitance of 277 mF (CBAT T ). At low frequency PSRR 2867 . Ap3 = V3 /VOUT 1 and Ap4 = VBAT T /VOUT 1 . 1V /μP a and its associated amplifier. Ap2 = VOUT 1 /Vrec . the maximum theoretical US-WPT link efficiency is 10.6%.2 V and VBAT T = 4. B.2 V.18 μm HV technology to enhance PSRR. P SRR2 = AAp2 . equations 3 can be rearragend as: Where ηae = 50% is the acousto-electric efficiency for one element of the implanted tranducer array. Vref 1 Vrec + β1 β1 P SRRLDO1 Vref VOU T 1 1 ≈ + β2 β2 P SRRLDO2 VOU T 1 ≈ (4a) VBAT T (4b) 1 1 1 1 P SRRLDO1 = A1 + A1 P SRR2 + P SRR1 and 1 1 1 1 P SRRLDO2 = A3 + A3 P SRR4 + P SRR3 .5 V A. The acoustic intensity is measured via the hydrophone HGL0200 (ONDA) which average sensitivity is -264. Experimental Results PSRR of the proposed LDO is measured using the methodology proposed in [7]. VDD = 4. assuming β1 A1 A2  1 and β2 A3 A4  1. the reference voltage (Vref ) is provided externally and it is set to 1. the ripple on DE is less that 10% on all the scale of VCC whereas the ripple on PAE is less that 10% for VCC larger than 3 V. A current mirror OTA (OTA1) and a basic OTA (OTA2) are designed in a CMOS 0. PVCC is the DC power and PDY N is the dynamic power consumed by the driver during the charge/discharge of the gate-tosource and gate-to-drain capacitances of the power MOS. 5. 6. Hence. According to Figure 5.

Annema. june 2010. IEEE Transactions on.8) 4. a linear dropout regulator was fabricated in a CMOS 0. Table I is a summary of the measured performance of the proposed LDO. to recharge a microbattery in the implant. Acar. rec measured at VBAT T is 10 dB above PSRR measured at VOUT 1 (Figure 8(a)).18 μm HV technology. Ultrasonic vs. et al.0e+06 (b) Fig. 10 dB 30 25 0. 65 PSRR [dB] 0÷9 (Vrec ) 10 (iOU T ) 4. 2010 International Conference on.05 (VBAT T ) 300 [8] 200 (= Vrec -VBAT T ) 210 0. In EMBC 2010. In Body Sensor Networks (BSN).1±0. The authors thank IMASONIC that has provided the transducers and INSERM for his support during acoustic field measurements.08 (simulated 0. dec.vrec vref A1 shift in the pmos threshold voltage from −1. 8.C. inductive power delivery for miniature biomedical implants. C ONCLUSION A novel system to transfer energy via ultrasound to deep implanted medical devices has been proposed. Power sources and electrical recharging strategies for implantable medical devices. In-vitro platform to study ultrasound as source for wireless energy transfer and communication for implanted medical devices. 54(12):2706 –2717. 2010.03) 55 45 VOUT1 VBATT VBATT (Cext = 10 μF) 35 25 15 5 1.S. Circuits and Systems. while at high frequency both PSRRs dropped down to few dBs (Figure 8(b)). [7] Texas Instrument. 70 65 VOUT1 VBATT PSRR [dB] 60 55 50 45 ACKNOWLEDGMENT 40 35 The research leading to these results has received funding from the European Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement n.0e+04 2. Equivalent network of the linear DC-to-DC converter to improve PSRR. sep 1990. Circuits and Systems I: Regular Papers. To compensate for the very low frequency pole introduced by the micro-cell battery and to improve PSRR at high frequency. Analytical design equations for class-e power amplifiers. RF Power Amplifiers for Mobile Communications.5e+05 1. Power supply rejection ratio in operational transconductance amplifiers.com. Moreover. a parallel bypass capacitor with low ESR can be added (Figure 8(b)).6) 2. Denisov and E. A. The driving electronic for the external transducers showed a DE up to 71% for VCC = 3 V and it was capable of delivering Iac = 23 W/cm2 at 11 cm by leading to a theoretical ηlink = 10. 2008.infinitepowersolutions. Steyaert. New York:Springer. Reynaert and M. and B.4 V (typical) to −1. the variation is due to a 2868 . 10. Steyaert and W. 2009.1007/s11708-0080016-3. [4] M. V. pages 84 –89. Yeatman. Measured PSRR: (a) at low frequency [50 Hz : 1 kHz] and (b) at high frequency [10 kHz : 1 MHz]. Frontiers of Energy and Power Engineering in China. 37(9):1077 –1084. Measured results are very close to simulated results (typical case).J. pages 3751 –3754.M. 7. 2007.6) 0.0e+05 f [Hz] 7.7 (simulated 1. [6] M. [2] A.1 100 200 300 400 500 600 700 800 900 1000 f [Hz] (a) 85 R EFERENCES 75 [1] Xiaojuan Wei and Jing Liu. sept. 224009.6%.2 (simulated 0.2 V as reported by the process control monitor (PCM) characterization report of the foundry.5e+05 5. 2006.66 (simulated 0. http://www. [3] P.J. 2:1–13. [5] Mazzilli F. IEEE Transactions on. The LDO was successfully tested and measured along with simulated results were reported. [8] Infinity Power Solutions. Ldo psrr measurement simplified: Slaa414. m10 OTA1 v1 vout1 A2 Rf1 TABLE I S UMMARY OF LDO SPECIFICATIONS AND PERFORMANCE R1 vSS vout1 m17 OTA2 vref A3 Parameter Input voltage (V) Load current (mA) Output voltage (V) Battery capacity (μAh) Drop voltage (mV) Ground current (μA) Load regulation @ VOU T 1 (mV /mA) OU T 1 Line regulation ΔV (mV /V ) ΔVrec Load regulation @ VBAT T (mV /mA) BAT T Line regulation ΔV (mV /V ) ΔV v3 vBATT A4 Rf2 ESR R2 Rself vSS CBATT Fig. Sansen. Nauta.