Multisite bio-stimulating implants magnetoelectrically
powered and individually programmed by a single
Zhanghao Yu, Joshua C. Chen, Yan He, Fatima T. Alrashdan,
Benjamin W. Avants, Amanda Singer, Jacob T. Robinson,
Rice University, Houston, TX
Implantable bioelectronics for electrically modulating activities of
specific cells have shown great success and exciting potential in
treating a wide range of diseases. Some of the most representative
therapies are cardiac pacemakers and neuromodulators for motor
function restoration, pain relief and neural disorder treatment [1,2].
While several wireless miniaturized bio-stimulators have been
demonstrated [3-6], most of them lack the capability of coordinated
multisite stimulation, which is shown to be more effective in many
scenarios [1,2]. Equipping an implant with electrode/LED arrays is a
straightforward approach to add extra stimulation channels [7-9], but
the deployment flexibility of stimulating spots is limited due to leads.
 shows a wired retinal stimulator array to scale up the driving
capability and ensure synchronization, but the heavy use of leads
severely limits its applications. A two-site heart pacing system  is
proposed with two independently powered and controlled implants
for flexible leadless deployment. Because the implants are
inductively powered by two transmitters (TXs) with frequency
multiplexing, they face stricter EM exposure constrains for power
transmission, more challenging device synchronization, and limited
scalability to more implants. To circumvent these problems, this
paper presents a hardware platform for coordinated and miniaturized
multisite stimulating implants, wirelessly powered and controlled by
a single TX. Magnetoelectric (ME) wireless power transfer with high
power and efficiency, low body absorption, and less sensitivity to
misalignment [4,5], is co-designed with a robust SoC to enable
reliable operation and individual programmability of the implants. The
presented system features: (1) robust operation with 2V source
amplitude variations, covering up to 40mm distance between TX and
implants; (2) individual addressability and programmability of each
implant, leveraging PUF IDs; (3) >90% chip efficiency for 1.5-to-3.5V
stimulation with fully programmable parameters; (4) no extra TX
output power required for additional implants; (5) miniaturized
implants with 6.2mm3 volume and 30mg mass.
The implant integrates a ME film, a capacitor, on-board electrodes,
and a SoC. The SoC interfaces with the ME film to receive power
and data and drives programmable stimulation. ME induced voltage
is rectified to Vrect and then converted by an adaptive switched-
capacitor power converter (SCPC), which provides proper voltage
and buffers energy on the off-chip capacitor for stimulation, and
provides VDD_H as a high-voltage supply for SoC. A 1V supply VDD_L
is generated by LDO (Fig.1 bottom). Each implant cycles through
charging, data transfer and stimulation phases. To maintain reliable
synchronized operation of multiple implants under different ME
voltages caused by different implantation depth and body movement,
the phase transitions are solely controlled by the TX with a short
notch of magnetic field (Fig.2 top). Comparator outputs in the active
rectifier are reused as watchdog signals to detect the notches.
Meanwhile, a global clock is extracted from the source by sensitive
clock recovery circuit, ensuring synchronization among all implants.
Individually programing every implant by a shared TX is critical for
effective and flexible stimulating therapies. Downlink data transferred
by ASK modulation contains a preamble for real-time demodulation
threshold calibration, an 8-bit ID for addressing, and a 19-bit data
payload for calibration and stimulation settings. The data update
controller checks the ID in packet against the on-chip ID to decide
whether to accept the new data. The on-chip 8-bit ID is realized with
CMOS physical unclonable functions (PUF) leveraging transistor
intrinsic variations to cheaply generate and store device-specific IDs
(Fig.2 bottom). A inverter chain based PUF design with native NMOS
regulation  is employed. Because of the narrow operating
temperature range and the native voltage regulation, 15-cycle
temporal majority voting (TMV) is sufficient to filter out thermal noise
and ensure PUF’s reliability. The ID generation is triggered by power-
on reset signal and clock gated after the ID is loaded to registers.
Variations of input voltage and
power of implants, caused by their
distance and misalignment with the
TX, are unavoidable in practice,
especially for multisite implants.
Thus, robust power recovery to
support stimulation across a wide
source conditions is highly desired.
Simply generating a high enough
voltage for stimulation driver
(VDD_stim) may ensure robustness
but will suffer from high power loss
and thus heat dissipation . Alternatively, unregulated voltage
stimulation by directly driving electrodes with charged capacitors has
high efficiency but sacrifices precise charge deposition control . To
achieve the desired robustness and efficiency without a complicated
feedback and reconfiguration loop as in , the proposed SCPC
directly generate a VDD_stim that is 10% higher than the desired
stimulus amplitude, and relies on the off-chip capacitor and a
regulator-style stimulation driver to support regulated mono- and bi-
phasic stimulus. Regulation of VDD_stim is realized by disconnecting
the capacitor from SCPC, the core of which is a charge pump and a
charging controller, once it reaches the desired level. High-speed
amplifiers inside the stimulation driver regulates the stimulus. To
save power, the amplifier will only be turned on in the stimulation
phase. The SCPC also includes an always-on high voltage selector
to generate VDD_H, which connects VDD_H to the higher one between
Vrect and VDD_stim and guarantees cold startup using Vrect (Fig. 3).
Fig.4 captures the operation waveforms of the implant. VDD_stim is
charged up and regulated to 2.75V, then drops to 2.15V after the
2.5V, 1.2ms bi-phasic stimulation. It is verified that the implant
maintains its operation with maximum stimulation amplitude (3.5V)
under large ME source variations (1.5–2.7V). 90% stimulating
efficiency is achieved as long as the amplitude is larger than 1.5V.
Power transfer at various distances are measured, which shows a
maximum TX-RX distance of 40mm and a highest power transfer
efficiency (PTE) of 1.03%. Individual programming of two implants
by a single TX is illustrated in Fig.5 (top left).
An in-vitro test with the a 2cm thick porcine tissue as a medium is
conducted (Fig.5, top right), which demonstrates flexible implant
deployment covering a space with 35mm radius, and synchronized
stimulations by two implants with programmed 0.01-to-0.8ms delays.
Based on simulation of the specific absorption rate (SAR) and the
electric field induction in a coil-generated 330kHz magnetic field, a
magnetic strength of 0.1mT, which is enough to sustain implant’s
functionality, can be delivered to a depth of 60mm without violating
the IEEE C95.1-2019 standards (unrestricted environment).
The proposed system is further validated in-vivo using a transgenic
line of Hydra vulgaris as a model for muscle stimulation and a rat
model for neural stimulation. Hydra naturally express a calcium
sensitive fluorescent protein, GCaMP7b, as well as voltage-gated ion
channels. To model synchronous stimulation of muscle tissue, two
hydra are used. In order to synchronize the muscle contractions, we
provide 3.5V, 20Hz, 1.2ms pulse width, biphasic stimulation pulse
trains. This results in >200% GCaMP7b fluorescence increases
which demonstrates activation of ion channels resulting in stimulus
aligned muscle contractions in both organisms (Fig.6 top left). We
also stimulate the sciatic nerve of the rat with varying amplitudes. A
graded response in the intensity of the rat leg kick is measured with
EMG of the plantar muscles (Fig.6 top right). The comparison table
with other bio-stimulating systems is given in Fig. 6 (bottom).
 S. Harkema et al., Lancet, 2011, DOI: 10.1016/S0140-6736(11)60547-3.
 H. Lyu et al., Sci. Rep., 2020, DOI: 10.1038/s41598-020-59017-z.
 D. K. Piech et al., Nat. BME., 2020, DOI: 10.1038/s41551-020-0518-9.
 A. Singer et al., Neuron, 2020, DOI: 10.1016/j.neuron.2020.05.019.
 Z. Yu et al., TBioCAS, 2020, DOI: 10.1109/TBCAS.2020.3037862.
 H. M. Lee et al., JSSC, 2015, DOI: 10.1109/JSSC.2014.2355814.
 Y. K. Lo et al., ISSCC, 2016, DOI: 10.1109/ISSCC.2016.7418067.
 Y. Jia et al., ISSCC, 2018, DOI: 10.1109/ISSCC.2018.8310387.
 Y. Jia et al., ISSCC, 2020, DOI: 10.1109/ISSCC19947.2020.9063065.
 T. Tokuda et al., TBioCAS, 2010, DOI: 10.1109/TBCAS.2010.2078508.
 D. Li et al., ISSCC, 2019, DOI: 10.1109/ISSCC.2019.8662537.
 H. M. Lee et al., JSSC, 2013, DOI: 10.1109/JSSC.2013.2266862.
Capacitor for Rectifier
© 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
B: Wired Array
of Stimulators 
C: Multiple TXs +
Stim. Implants Pairs 
+ High Channel Density
Fixed Channel Quantity
+ Flexible Deployment
Higher Body Absorption
+ Flexible Deployment
+ Higher Efficiency
+ Good Synchronization
+ Good Scalability
+ High Channel Density
High Risks of Infection
Fixed Channel Quantity
A: Single Implant with
Electrode/LED Array [7-9]
Proposed Implants Power Managem ent
Amp. & Timing
CLKG & CLKstim
This Work: Single TX +
Multiple Stim. Implants
Coil + -
Magnetoelectric (ME) Effects
Fig. 1. Concepts of multisite bio-stimulation and various multisite
stimulating system structures; illustration of the implant, architecture
of its SoC, and principles of ME power transfer for multiple RXs.
2 bit8 bits
Check Gap Data
6 bits 4 bits 5 bits 3 bits
Data Transmission Protocol and Individually Programm ing Scheme
Operating Scheme an d Synchronized Timing Reference s Generation
PUF with Tempor al Majority Voting (TMV)
PUF[i]   
Errors due to Noise
PUF Output (V)
Data Transfer Stimulation
Fig. 2. Diagram of SoC operating phase transition and clock
recovery; diagram of data recovery circuitry and schematics of PUF.
Adaptive Switched-Ca pacitor Power Converter (SC PC)
Stimulation Dr iver
Pin & Vrect: A > B > C
Low Efficiency, Sensitive to Vin Variations Power Overhead of Precise Detection
Bad Regulation, Sensitive to Vin Variations
Adaptive SCPC Stim.
Good efficiency and regulation
Robust to input variations
Regulation of VDD_stim
Fig. 3. Principles of the proposed highly efficient voltage stimulation;
schematics of adaptive power converter and stimulation driver.
Stimulation at 20Hz with Rload = 1kΩ
Measured Stimulating EfficiencyMeasured Stimulation with Source Variation s
0 5 10 15
4 Stim. Amplitude
Stimulation Amplitude Setting
Stim. Efficiency (%)
VME =2.7V VME =2.2V VME =1.5V
phase transi tion
Drops due to
Measured Power Delivery vs. Distances
0 5 10 15 20 25 30 35 40 45
TX-RX Distance (mm)
Peak PTE (%)
Operating Distance > 40mm
(Vrect > 1.4V)
Fig. 4. Measurements of operating waveforms of the proposed
implant; 3.5V stimulation with varying ME voltage; stimulating
efficiency with various amplitudes; power delivery versus distances.
Progmaming of Multiple Implants with a Si ngle TX
Synchronized Stimulations with Programma ble
Amplitude, Pulse Width and Delay (In-Vitro)
Stimulation:A Amp. Change:
Local Safety Analysis of the 330-kHz
Magnetic Field in COMSOL
In-Vitro Test with Porcine Tissue
Thickness: 2cm 30
10 20 30
Lateral & Longitudinal
Power Supply AFG
Skin 67.1 0.00913
Fat 56.2 0.15
Muscle 36.6 0.51
Bone 10.6 0.00819
Max. Electric Field (V/m)
Max. SAR (W/kg)
Fig. 5. Measurement of individual programmability of implants with
a single TX; safety analysis; in-vitro test with porcine tissue and
shmoo plot of operating space; synchronized stimulations and
stimulations with programmable delays.
Hydra Contractions with & without Stimulation Rat EMG Response vs. Stimulation Amplitude
Permanent Magnet (Bias)
(1.2mS pulse width,
This Work ISSCC'16  ISSCC'18  Sci. Rep.'20  ISSCC'20  Nat. BME.'20 
Neural / Cardiac Neural Neural Cardiac Neural Neural
180 180 350 180 350 65
Magnetoelectric Inductive Ind uctive Inductive Inductive
Single TX +
Two TXs +
Voltage Current Optical Voltage Optical, Current Current
Amplitude 3.5V, 4b 0.5mA, 7b N/A 3V, Fixed 0.77mA, 5b 0.4mA, 3b
Pulse Width 1.2ms, 4b 8ms 2ms, 2b 0.3ms, 3 levels 0.64ms, 4b Continuous
Frequency Continuous Continuous 10Hz, 2b Continuous 400Hz, 4b Continuous
Delay 0.8ms, 5b No No No No No
PUF ID Multiplexer Multiplexer Differe nt fcarrier Multiplexer N/A
9864 300 320700 4
1 x 0.8 5.7 x 4.4 1 x 1 0.85 x 0.45 5 x 3 1 x 1
6.2 500 12.2 10.1 N/A 1.7
40 N/A 7 60 N/A 55
Implant Volume (mm3)
Max. RX-TX Distance (mm)
SoC Power (μW)
Chip Size (mm2)
Fig. 6. Synchronous muscle activations of Hydra in response to
electrical stimulations; EMG response of rat with various stimulation
amplitudes; comparisons with state-of-the-art stimulating systems.